public-mirrors/linux
1
0
Fork 0
mirror of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git synced 2025-08-05 16:54:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
linux/fs/smb/client/cached_dir.c

833 lines
21 KiB
C
Raw Permalink Normal View History

cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
// SPDX-License-Identifier: GPL-2.0
/*
* Functions to handle the cached directory entries
*
* Copyright (c) 2022, Ronnie Sahlberg <lsahlber@redhat.com>
*/
cifs: find and use the dentry for cached non-root directories also This allows us to use cached attributes for the entries in a cached directory for as long as a lease is held on the directory itself. Previously we have always allowed "used cached attributes for 1 second" but this extends this to the lifetime of the lease as well as making the caching safer. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-12 06:13:03 -05:00
#include <linux/namei.h>
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
#include "cifsglob.h"
#include "cifsproto.h"
#include "cifs_debug.h"
#include "smb2proto.h"
#include "cached_dir.h"
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
static struct cached_fid *init_cached_dir(const char *path);
static void free_cached_dir(struct cached_fid *cfid);
cifs: return a single-use cfid if we did not get a lease If we did not get a lease we can still return a single use cfid to the caller. The cfid will not have has_lease set and will thus not be shared with any other concurrent users and will be freed immediately when the caller drops the handle. This avoids extra roundtrips for servers that do not support directory leases where they would first fail to get a cfid with a lease and then fallback to try a normal SMB2_open() Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Cc: stable@vger.kernel.org Reviewed-by: Bharath SM <bharathsm@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-02-17 13:35:01 +10:00
static void smb2_close_cached_fid(struct kref *ref);
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
static void cfids_laundromat_worker(struct work_struct *work);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
struct cached_dir_dentry {
struct list_head entry;
struct dentry *dentry;
};
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
static struct cached_fid *find_or_create_cached_dir(struct cached_fids *cfids,
const char *path,
smb3: allow controlling maximum number of cached directories Allow adjusting the maximum number of cached directories per share (defaults to 16) via mount parm "max_cached_dirs" Signed-off-by: Steve French <stfrench@microsoft.com>
2023-09-01 02:15:14 -05:00
bool lookup_only,
__u32 max_cached_dirs)
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
{
struct cached_fid *cfid;
list_for_each_entry(cfid, &cfids->entries, entry) {
if (!strcmp(cfid->path, path)) {
/*
* If it doesn't have a lease it is either not yet
* fully cached or it may be in the process of
* being deleted due to a lease break.
*/
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (!cfid->time || !cfid->has_lease) {
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
return NULL;
}
kref_get(&cfid->refcount);
return cfid;
}
}
if (lookup_only) {
return NULL;
}
smb3: allow controlling maximum number of cached directories Allow adjusting the maximum number of cached directories per share (defaults to 16) via mount parm "max_cached_dirs" Signed-off-by: Steve French <stfrench@microsoft.com>
2023-09-01 02:15:14 -05:00
if (cfids->num_entries >= max_cached_dirs) {
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
return NULL;
}
cfid = init_cached_dir(path);
if (cfid == NULL) {
return NULL;
}
cfid->cfids = cfids;
cfids->num_entries++;
list_add(&cfid->entry, &cfids->entries);
cfid->on_list = true;
kref_get(&cfid->refcount);
smb: Don't leak cfid when reconnect races with open_cached_dir open_cached_dir() may either race with the tcon reconnection even before compound_send_recv() or directly trigger a reconnection via SMB2_open_init() or SMB_query_info_init(). The reconnection process invokes invalidate_all_cached_dirs() via cifs_mark_open_files_invalid(), which removes all cfids from the cfids->entries list but doesn't drop a ref if has_lease isn't true. This results in the currently-being-constructed cfid not being on the list, but still having a refcount of 2. It leaks if returned from open_cached_dir(). Fix this by setting cfid->has_lease when the ref is actually taken; the cfid will not be used by other threads until it has a valid time. Addresses these kmemleaks: unreferenced object 0xffff8881090c4000 (size 1024): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 32 bytes): 00 01 00 00 00 00 ad de 22 01 00 00 00 00 ad de ........"....... 00 ca 45 22 81 88 ff ff f8 dc 4f 04 81 88 ff ff ..E"......O..... backtrace (crc 6f58c20f): [<ffffffff8b895a1e>] __kmalloc_cache_noprof+0x2be/0x350 [<ffffffff8bda06e3>] open_cached_dir+0x993/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e unreferenced object 0xffff8881044fdcf8 (size 8): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 8 bytes): 00 cc cc cc cc cc cc cc ........ backtrace (crc 10c106a9): [<ffffffff8b89a3d3>] __kmalloc_node_track_caller_noprof+0x363/0x480 [<ffffffff8b7d7256>] kstrdup+0x36/0x60 [<ffffffff8bda0700>] open_cached_dir+0x9b0/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e And addresses these BUG splats when unmounting the SMB filesystem: BUG: Dentry ffff888140590ba0{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] WARNING: CPU: 3 PID: 3433 at fs/dcache.c:1536 umount_check+0xd0/0x100 Modules linked in: CPU: 3 UID: 0 PID: 3433 Comm: bash Not tainted 6.12.0-rc4-g850925a8133c-dirty #49 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:umount_check+0xd0/0x100 Code: 8d 7c 24 40 e8 31 5a f4 ff 49 8b 54 24 40 41 56 49 89 e9 45 89 e8 48 89 d9 41 57 48 89 de 48 c7 c7 80 e7 db ac e8 f0 72 9a ff <0f> 0b 58 31 c0 5a 5b 5d 41 5c 41 5d 41 5e 41 5f e9 2b e5 5d 01 41 RSP: 0018:ffff88811cc27978 EFLAGS: 00010286 RAX: 0000000000000000 RBX: ffff888140590ba0 RCX: ffffffffaaf20bae RDX: dffffc0000000000 RSI: 0000000000000008 RDI: ffff8881f6fb6f40 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed1023984ee3 R10: ffff88811cc2771f R11: 00000000016cfcc0 R12: ffff888134383e08 R13: 0000000000000002 R14: ffff8881462ec668 R15: ffffffffaceab4c0 FS: 00007f23bfa98740(0000) GS:ffff8881f6f80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000556de4a6f808 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> d_walk+0x6a/0x530 shrink_dcache_for_umount+0x6a/0x200 generic_shutdown_super+0x52/0x2a0 kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 Code: ff ff ff ff c3 66 0f 1f 44 00 00 48 8b 0d 11 93 0d 00 f7 d8 64 89 01 b8 ff ff ff ff eb bf 0f 1f 44 00 00 b8 50 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d e9 92 0d 00 f7 d8 64 89 01 48 RSP: 002b:00007ffee9138598 EFLAGS: 00000246 ORIG_RAX: 0000000000000050 RAX: 0000000000000000 RBX: 0000558f1803e9a0 RCX: 00007f23bfb93ae7 RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000558f1803e9a0 RBP: 0000558f1803e600 R08: 0000000000000007 R09: 0000558f17fab610 R10: d91d5ec34ab757b0 R11: 0000000000000246 R12: 0000000000000001 R13: 0000000000000000 R14: 0000000000000015 R15: 0000000000000000 </TASK> irq event stamp: 1163486 hardirqs last enabled at (1163485): [<ffffffffac98d344>] _raw_spin_unlock_irqrestore+0x34/0x60 hardirqs last disabled at (1163486): [<ffffffffac97dcfc>] __schedule+0xc7c/0x19a0 softirqs last enabled at (1163482): [<ffffffffab79a3ee>] __smb_send_rqst+0x3de/0x990 softirqs last disabled at (1163480): [<ffffffffac2314f1>] release_sock+0x21/0xf0 ---[ end trace 0000000000000000 ]--- VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI CPU: 1 UID: 0 PID: 3433 Comm: bash Tainted: G W 6.12.0-rc4-g850925a8133c-dirty #49 Tainted: [W]=WARN Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 </TASK> Modules linked in: ---[ end trace 0000000000000000 ]--- RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 This reproduces eventually with an SMB mount and two shells running these loops concurrently - while true; do cd ~; sleep 1; for i in {1..3}; do cd /mnt/test/subdir; echo $PWD; sleep 1; cd ..; echo $PWD; sleep 1; done; echo ...; done - while true; do iptables -F OUTPUT; mount -t cifs -a; for _ in {0..2}; do ls /mnt/test/subdir/ | wc -l; done; iptables -I OUTPUT -p tcp --dport 445 -j DROP; sleep 10 echo "unmounting"; umount -l -t cifs -a; echo "done unmounting"; sleep 20 echo "recovering"; iptables -F OUTPUT; sleep 10; done Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Fixes: 5c86919455c1 ("smb: client: fix use-after-free in smb2_query_info_compound()") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:26 -08:00
/*
* Set @cfid->has_lease to true during construction so that the lease
* reference can be put in cached_dir_lease_break() due to a potential
* lease break right after the request is sent or while @cfid is still
* being cached, or if a reconnection is triggered during construction.
* Concurrent processes won't be to use it yet due to @cfid->time being
* zero.
*/
cfid->has_lease = true;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
return cfid;
}
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
cifs: find and use the dentry for cached non-root directories also This allows us to use cached attributes for the entries in a cached directory for as long as a lease is held on the directory itself. Previously we have always allowed "used cached attributes for 1 second" but this extends this to the lifetime of the lease as well as making the caching safer. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-12 06:13:03 -05:00
static struct dentry *
path_to_dentry(struct cifs_sb_info *cifs_sb, const char *path)
{
struct dentry *dentry;
const char *s, *p;
char sep;
sep = CIFS_DIR_SEP(cifs_sb);
dentry = dget(cifs_sb->root);
s = path;
do {
struct inode *dir = d_inode(dentry);
struct dentry *child;
if (!S_ISDIR(dir->i_mode)) {
dput(dentry);
dentry = ERR_PTR(-ENOTDIR);
break;
}
/* skip separators */
while (*s == sep)
s++;
if (!*s)
break;
p = s++;
/* next separator */
while (*s && *s != sep)
s++;
VFS: rename lookup_one_len family to lookup_noperm and remove permission check The lookup_one_len family of functions is (now) only used internally by a filesystem on itself either - in a context where permission checking is irrelevant such as by a virtual filesystem populating itself, or xfs accessing its ORPHANAGE or dquota accessing the quota file; or - in a context where a permission check (MAY_EXEC on the parent) has just been performed such as a network filesystem finding in "silly-rename" file in the same directory. This is also the context after the _parentat() functions where currently lookup_one_qstr_excl() is used. So the permission check is pointless. The name "one_len" is unhelpful in understanding the purpose of these functions and should be changed. Most of the callers pass the len as "strlen()" so using a qstr and QSTR() can simplify the code. This patch renames these functions (include lookup_positive_unlocked() which is part of the family despite the name) to have a name based on "lookup_noperm". They are changed to receive a 'struct qstr' instead of separate name and len. In a few cases the use of QSTR() results in a new call to strlen(). try_lookup_noperm() takes a pointer to a qstr instead of the whole qstr. This is consistent with d_hash_and_lookup() (which is nearly identical) and useful for lookup_noperm_unlocked(). The new lookup_noperm_common() doesn't take a qstr yet. That will be tidied up in a subsequent patch. Signed-off-by: NeilBrown <neil@brown.name> Link: https://lore.kernel.org/r/20250319031545.2999807-5-neil@brown.name Signed-off-by: Christian Brauner <brauner@kernel.org>
2025-03-19 14:01:35 +11:00
child = lookup_noperm_positive_unlocked(&QSTR_LEN(p, s - p),
dentry);
cifs: find and use the dentry for cached non-root directories also This allows us to use cached attributes for the entries in a cached directory for as long as a lease is held on the directory itself. Previously we have always allowed "used cached attributes for 1 second" but this extends this to the lifetime of the lease as well as making the caching safer. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-12 06:13:03 -05:00
dput(dentry);
dentry = child;
} while (!IS_ERR(dentry));
return dentry;
}
cifs: fix dentry lookups in directory handle cache Get rid of any prefix paths in @path before lookup_positive_unlocked() as it will call ->lookup() which already adds those prefix paths through build_path_from_dentry(). This has caused a performance regression when mounting shares with a prefix path where readdir(2) would end up retrying several times to open bad directory names that contained duplicate prefix paths. Fix this by skipping any prefix paths in @path before calling lookup_positive_unlocked(). Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-24 13:56:33 -03:00
static const char *path_no_prefix(struct cifs_sb_info *cifs_sb,
const char *path)
{
size_t len = 0;
if (!*path)
return path;
if ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH) &&
cifs_sb->prepath) {
len = strlen(cifs_sb->prepath) + 1;
if (unlikely(len > strlen(path)))
return ERR_PTR(-EINVAL);
}
return path + len;
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
/*
* Open the and cache a directory handle.
* If error then *cfid is not initialized.
*/
int open_cached_dir(unsigned int xid, struct cifs_tcon *tcon,
cifs: Do not access tcon->cfids->cfid directly from is_path_accessible cfids will soon keep a list of cached fids so we should not access this directly from outside of cached_dir.c Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:51:18 -05:00
const char *path,
struct cifs_sb_info *cifs_sb,
bool lookup_only, struct cached_fid **ret_cfid)
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
{
struct cifs_ses *ses;
struct TCP_Server_Info *server;
struct cifs_open_parms oparms;
struct smb2_create_rsp *o_rsp = NULL;
struct smb2_query_info_rsp *qi_rsp = NULL;
int resp_buftype[2];
struct smb_rqst rqst[2];
struct kvec rsp_iov[2];
struct kvec open_iov[SMB2_CREATE_IOV_SIZE];
struct kvec qi_iov[1];
int rc, flags = 0;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
__le16 *utf16_path = NULL;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
u8 oplock = SMB2_OPLOCK_LEVEL_II;
struct cifs_fid *pfid;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct dentry *dentry = NULL;
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
struct cached_fid *cfid;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fids *cfids;
cifs: fix dentry lookups in directory handle cache Get rid of any prefix paths in @path before lookup_positive_unlocked() as it will call ->lookup() which already adds those prefix paths through build_path_from_dentry(). This has caused a performance regression when mounting shares with a prefix path where readdir(2) would end up retrying several times to open bad directory names that contained duplicate prefix paths. Fix this by skipping any prefix paths in @path before calling lookup_positive_unlocked(). Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-24 13:56:33 -03:00
const char *npath;
cifs: commands that are retried should have replay flag set MS-SMB2 states that the header flag SMB2_FLAGS_REPLAY_OPERATION needs to be set when a command needs to be retried, so that the server is aware that this is a replay for an operation that appeared before. This can be very important, for example, for state changing operations and opens which get retried following a reconnect; since the client maybe unaware of the status of the previous open. This is particularly important for multichannel scenario, since disconnection of one connection does not mean that the session is lost. The requests can be replayed on another channel. This change also makes use of exponential back-off before replays and also limits the number of retries to "retrans" mount option value. Also, this change does not modify the read/write codepath. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-01-21 03:32:47 +00:00
int retries = 0, cur_sleep = 1;
smb: client: use ParentLeaseKey in open_cached_dir Implement ParentLeaseKey logic in open_cached_dir() by looking up the parent cfid, copying its lease key into the fid struct, and setting the appropriate lease flag. Fixes: f047390a097e ("CIFS: Add create lease v2 context for SMB3") Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-08 13:37:50 -03:00
__le32 lease_flags = 0;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
smb: client: remove unnecessary checks in open_cached_dir() Checks inside open_cached_dir() can be removed because if dir caching is disabled then tcon->cfids is necessarily NULL. Therefore, all other checks are redundant. Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.com> Reviewed-by: Enzo Matsumiya <ematsumiya@suse.de> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-22 22:14:36 -03:00
if (cifs_sb->root == NULL)
return -ENOENT;
if (tcon == NULL)
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
return -EOPNOTSUPP;
ses = tcon->ses;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfids = tcon->cfids;
smb: client: remove unnecessary checks in open_cached_dir() Checks inside open_cached_dir() can be removed because if dir caching is disabled then tcon->cfids is necessarily NULL. Therefore, all other checks are redundant. Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.com> Reviewed-by: Enzo Matsumiya <ematsumiya@suse.de> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-22 22:14:36 -03:00
if (cfids == NULL)
return -EOPNOTSUPP;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
cifs: commands that are retried should have replay flag set MS-SMB2 states that the header flag SMB2_FLAGS_REPLAY_OPERATION needs to be set when a command needs to be retried, so that the server is aware that this is a replay for an operation that appeared before. This can be very important, for example, for state changing operations and opens which get retried following a reconnect; since the client maybe unaware of the status of the previous open. This is particularly important for multichannel scenario, since disconnection of one connection does not mean that the session is lost. The requests can be replayed on another channel. This change also makes use of exponential back-off before replays and also limits the number of retries to "retrans" mount option value. Also, this change does not modify the read/write codepath. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-01-21 03:32:47 +00:00
replay_again:
/* reinitialize for possible replay */
flags = 0;
oplock = SMB2_OPLOCK_LEVEL_II;
server = cifs_pick_channel(ses);
if (!server->ops->new_lease_key)
return -EIO;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
utf16_path = cifs_convert_path_to_utf16(path, cifs_sb);
if (!utf16_path)
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
return -ENOMEM;
smb: client: Avoid race in open_cached_dir with lease breaks A pre-existing valid cfid returned from find_or_create_cached_dir might race with a lease break, meaning open_cached_dir doesn't consider it valid, and thinks it's newly-constructed. This leaks a dentry reference if the allocation occurs before the queued lease break work runs. Avoid the race by extending holding the cfid_list_lock across find_or_create_cached_dir and when the result is checked. Cc: stable@vger.kernel.org Reviewed-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-06 22:28:09 -07:00
spin_lock(&cfids->cfid_list_lock);
smb3: allow controlling maximum number of cached directories Allow adjusting the maximum number of cached directories per share (defaults to 16) via mount parm "max_cached_dirs" Signed-off-by: Steve French <stfrench@microsoft.com>
2023-09-01 02:15:14 -05:00
cfid = find_or_create_cached_dir(cfids, path, lookup_only, tcon->max_cached_dirs);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
if (cfid == NULL) {
smb: client: Avoid race in open_cached_dir with lease breaks A pre-existing valid cfid returned from find_or_create_cached_dir might race with a lease break, meaning open_cached_dir doesn't consider it valid, and thinks it's newly-constructed. This leaks a dentry reference if the allocation occurs before the queued lease break work runs. Avoid the race by extending holding the cfid_list_lock across find_or_create_cached_dir and when the result is checked. Cc: stable@vger.kernel.org Reviewed-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-06 22:28:09 -07:00
spin_unlock(&cfids->cfid_list_lock);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
kfree(utf16_path);
return -ENOENT;
}
/*
smb: Don't leak cfid when reconnect races with open_cached_dir open_cached_dir() may either race with the tcon reconnection even before compound_send_recv() or directly trigger a reconnection via SMB2_open_init() or SMB_query_info_init(). The reconnection process invokes invalidate_all_cached_dirs() via cifs_mark_open_files_invalid(), which removes all cfids from the cfids->entries list but doesn't drop a ref if has_lease isn't true. This results in the currently-being-constructed cfid not being on the list, but still having a refcount of 2. It leaks if returned from open_cached_dir(). Fix this by setting cfid->has_lease when the ref is actually taken; the cfid will not be used by other threads until it has a valid time. Addresses these kmemleaks: unreferenced object 0xffff8881090c4000 (size 1024): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 32 bytes): 00 01 00 00 00 00 ad de 22 01 00 00 00 00 ad de ........"....... 00 ca 45 22 81 88 ff ff f8 dc 4f 04 81 88 ff ff ..E"......O..... backtrace (crc 6f58c20f): [<ffffffff8b895a1e>] __kmalloc_cache_noprof+0x2be/0x350 [<ffffffff8bda06e3>] open_cached_dir+0x993/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e unreferenced object 0xffff8881044fdcf8 (size 8): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 8 bytes): 00 cc cc cc cc cc cc cc ........ backtrace (crc 10c106a9): [<ffffffff8b89a3d3>] __kmalloc_node_track_caller_noprof+0x363/0x480 [<ffffffff8b7d7256>] kstrdup+0x36/0x60 [<ffffffff8bda0700>] open_cached_dir+0x9b0/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e And addresses these BUG splats when unmounting the SMB filesystem: BUG: Dentry ffff888140590ba0{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] WARNING: CPU: 3 PID: 3433 at fs/dcache.c:1536 umount_check+0xd0/0x100 Modules linked in: CPU: 3 UID: 0 PID: 3433 Comm: bash Not tainted 6.12.0-rc4-g850925a8133c-dirty #49 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:umount_check+0xd0/0x100 Code: 8d 7c 24 40 e8 31 5a f4 ff 49 8b 54 24 40 41 56 49 89 e9 45 89 e8 48 89 d9 41 57 48 89 de 48 c7 c7 80 e7 db ac e8 f0 72 9a ff <0f> 0b 58 31 c0 5a 5b 5d 41 5c 41 5d 41 5e 41 5f e9 2b e5 5d 01 41 RSP: 0018:ffff88811cc27978 EFLAGS: 00010286 RAX: 0000000000000000 RBX: ffff888140590ba0 RCX: ffffffffaaf20bae RDX: dffffc0000000000 RSI: 0000000000000008 RDI: ffff8881f6fb6f40 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed1023984ee3 R10: ffff88811cc2771f R11: 00000000016cfcc0 R12: ffff888134383e08 R13: 0000000000000002 R14: ffff8881462ec668 R15: ffffffffaceab4c0 FS: 00007f23bfa98740(0000) GS:ffff8881f6f80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000556de4a6f808 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> d_walk+0x6a/0x530 shrink_dcache_for_umount+0x6a/0x200 generic_shutdown_super+0x52/0x2a0 kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 Code: ff ff ff ff c3 66 0f 1f 44 00 00 48 8b 0d 11 93 0d 00 f7 d8 64 89 01 b8 ff ff ff ff eb bf 0f 1f 44 00 00 b8 50 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d e9 92 0d 00 f7 d8 64 89 01 48 RSP: 002b:00007ffee9138598 EFLAGS: 00000246 ORIG_RAX: 0000000000000050 RAX: 0000000000000000 RBX: 0000558f1803e9a0 RCX: 00007f23bfb93ae7 RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000558f1803e9a0 RBP: 0000558f1803e600 R08: 0000000000000007 R09: 0000558f17fab610 R10: d91d5ec34ab757b0 R11: 0000000000000246 R12: 0000000000000001 R13: 0000000000000000 R14: 0000000000000015 R15: 0000000000000000 </TASK> irq event stamp: 1163486 hardirqs last enabled at (1163485): [<ffffffffac98d344>] _raw_spin_unlock_irqrestore+0x34/0x60 hardirqs last disabled at (1163486): [<ffffffffac97dcfc>] __schedule+0xc7c/0x19a0 softirqs last enabled at (1163482): [<ffffffffab79a3ee>] __smb_send_rqst+0x3de/0x990 softirqs last disabled at (1163480): [<ffffffffac2314f1>] release_sock+0x21/0xf0 ---[ end trace 0000000000000000 ]--- VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI CPU: 1 UID: 0 PID: 3433 Comm: bash Tainted: G W 6.12.0-rc4-g850925a8133c-dirty #49 Tainted: [W]=WARN Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 </TASK> Modules linked in: ---[ end trace 0000000000000000 ]--- RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 This reproduces eventually with an SMB mount and two shells running these loops concurrently - while true; do cd ~; sleep 1; for i in {1..3}; do cd /mnt/test/subdir; echo $PWD; sleep 1; cd ..; echo $PWD; sleep 1; done; echo ...; done - while true; do iptables -F OUTPUT; mount -t cifs -a; for _ in {0..2}; do ls /mnt/test/subdir/ | wc -l; done; iptables -I OUTPUT -p tcp --dport 445 -j DROP; sleep 10 echo "unmounting"; umount -l -t cifs -a; echo "done unmounting"; sleep 20 echo "recovering"; iptables -F OUTPUT; sleep 10; done Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Fixes: 5c86919455c1 ("smb: client: fix use-after-free in smb2_query_info_compound()") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:26 -08:00
* Return cached fid if it is valid (has a lease and has a time).
* Otherwise, it is either a new entry or laundromat worker removed it
* from @cfids->entries. Caller will put last reference if the latter.
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
*/
smb: Don't leak cfid when reconnect races with open_cached_dir open_cached_dir() may either race with the tcon reconnection even before compound_send_recv() or directly trigger a reconnection via SMB2_open_init() or SMB_query_info_init(). The reconnection process invokes invalidate_all_cached_dirs() via cifs_mark_open_files_invalid(), which removes all cfids from the cfids->entries list but doesn't drop a ref if has_lease isn't true. This results in the currently-being-constructed cfid not being on the list, but still having a refcount of 2. It leaks if returned from open_cached_dir(). Fix this by setting cfid->has_lease when the ref is actually taken; the cfid will not be used by other threads until it has a valid time. Addresses these kmemleaks: unreferenced object 0xffff8881090c4000 (size 1024): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 32 bytes): 00 01 00 00 00 00 ad de 22 01 00 00 00 00 ad de ........"....... 00 ca 45 22 81 88 ff ff f8 dc 4f 04 81 88 ff ff ..E"......O..... backtrace (crc 6f58c20f): [<ffffffff8b895a1e>] __kmalloc_cache_noprof+0x2be/0x350 [<ffffffff8bda06e3>] open_cached_dir+0x993/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e unreferenced object 0xffff8881044fdcf8 (size 8): comm "bash", pid 1860, jiffies 4295126592 hex dump (first 8 bytes): 00 cc cc cc cc cc cc cc ........ backtrace (crc 10c106a9): [<ffffffff8b89a3d3>] __kmalloc_node_track_caller_noprof+0x363/0x480 [<ffffffff8b7d7256>] kstrdup+0x36/0x60 [<ffffffff8bda0700>] open_cached_dir+0x9b0/0x1fb0 [<ffffffff8bdaa750>] cifs_readdir+0x15a0/0x1d50 [<ffffffff8b9a853f>] iterate_dir+0x28f/0x4b0 [<ffffffff8b9a9aed>] __x64_sys_getdents64+0xfd/0x200 [<ffffffff8cf6da05>] do_syscall_64+0x95/0x1a0 [<ffffffff8d00012f>] entry_SYSCALL_64_after_hwframe+0x76/0x7e And addresses these BUG splats when unmounting the SMB filesystem: BUG: Dentry ffff888140590ba0{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] WARNING: CPU: 3 PID: 3433 at fs/dcache.c:1536 umount_check+0xd0/0x100 Modules linked in: CPU: 3 UID: 0 PID: 3433 Comm: bash Not tainted 6.12.0-rc4-g850925a8133c-dirty #49 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:umount_check+0xd0/0x100 Code: 8d 7c 24 40 e8 31 5a f4 ff 49 8b 54 24 40 41 56 49 89 e9 45 89 e8 48 89 d9 41 57 48 89 de 48 c7 c7 80 e7 db ac e8 f0 72 9a ff <0f> 0b 58 31 c0 5a 5b 5d 41 5c 41 5d 41 5e 41 5f e9 2b e5 5d 01 41 RSP: 0018:ffff88811cc27978 EFLAGS: 00010286 RAX: 0000000000000000 RBX: ffff888140590ba0 RCX: ffffffffaaf20bae RDX: dffffc0000000000 RSI: 0000000000000008 RDI: ffff8881f6fb6f40 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed1023984ee3 R10: ffff88811cc2771f R11: 00000000016cfcc0 R12: ffff888134383e08 R13: 0000000000000002 R14: ffff8881462ec668 R15: ffffffffaceab4c0 FS: 00007f23bfa98740(0000) GS:ffff8881f6f80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000556de4a6f808 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> d_walk+0x6a/0x530 shrink_dcache_for_umount+0x6a/0x200 generic_shutdown_super+0x52/0x2a0 kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 Code: ff ff ff ff c3 66 0f 1f 44 00 00 48 8b 0d 11 93 0d 00 f7 d8 64 89 01 b8 ff ff ff ff eb bf 0f 1f 44 00 00 b8 50 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d e9 92 0d 00 f7 d8 64 89 01 48 RSP: 002b:00007ffee9138598 EFLAGS: 00000246 ORIG_RAX: 0000000000000050 RAX: 0000000000000000 RBX: 0000558f1803e9a0 RCX: 00007f23bfb93ae7 RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000558f1803e9a0 RBP: 0000558f1803e600 R08: 0000000000000007 R09: 0000558f17fab610 R10: d91d5ec34ab757b0 R11: 0000000000000246 R12: 0000000000000001 R13: 0000000000000000 R14: 0000000000000015 R15: 0000000000000000 </TASK> irq event stamp: 1163486 hardirqs last enabled at (1163485): [<ffffffffac98d344>] _raw_spin_unlock_irqrestore+0x34/0x60 hardirqs last disabled at (1163486): [<ffffffffac97dcfc>] __schedule+0xc7c/0x19a0 softirqs last enabled at (1163482): [<ffffffffab79a3ee>] __smb_send_rqst+0x3de/0x990 softirqs last disabled at (1163480): [<ffffffffac2314f1>] release_sock+0x21/0xf0 ---[ end trace 0000000000000000 ]--- VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI CPU: 1 UID: 0 PID: 3433 Comm: bash Tainted: G W 6.12.0-rc4-g850925a8133c-dirty #49 Tainted: [W]=WARN Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 Call Trace: <TASK> kill_anon_super+0x22/0x40 cifs_kill_sb+0x159/0x1e0 deactivate_locked_super+0x66/0xe0 cleanup_mnt+0x140/0x210 task_work_run+0xfb/0x170 syscall_exit_to_user_mode+0x29f/0x2b0 do_syscall_64+0xa1/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e RIP: 0033:0x7f23bfb93ae7 </TASK> Modules linked in: ---[ end trace 0000000000000000 ]--- RIP: 0010:generic_shutdown_super+0x290/0x2a0 Code: e8 15 7c f7 ff 48 8b 5d 28 48 89 df e8 09 7c f7 ff 48 8b 0b 48 89 ee 48 8d 95 68 06 00 00 48 c7 c7 80 7f db ac e8 00 69 af ff <0f> 0b 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 90 90 90 90 90 90 RSP: 0018:ffff88811cc27a50 EFLAGS: 00010246 RAX: 000000000000003e RBX: ffffffffae994420 RCX: 0000000000000027 RDX: 0000000000000000 RSI: ffffffffab06180e RDI: ffff8881f6eb18c8 RBP: ffff8881462ec000 R08: 0000000000000001 R09: ffffed103edd6319 R10: ffff8881f6eb18cb R11: 00000000016d3158 R12: ffff8881462ec9c0 R13: ffff8881462ec050 R14: 0000000000000001 R15: 0000000000000000 FS: 00007f23bfa98740(0000) GS:ffff8881f6e80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8364005d68 CR3: 0000000123c80000 CR4: 0000000000350ef0 This reproduces eventually with an SMB mount and two shells running these loops concurrently - while true; do cd ~; sleep 1; for i in {1..3}; do cd /mnt/test/subdir; echo $PWD; sleep 1; cd ..; echo $PWD; sleep 1; done; echo ...; done - while true; do iptables -F OUTPUT; mount -t cifs -a; for _ in {0..2}; do ls /mnt/test/subdir/ | wc -l; done; iptables -I OUTPUT -p tcp --dport 445 -j DROP; sleep 10 echo "unmounting"; umount -l -t cifs -a; echo "done unmounting"; sleep 20 echo "recovering"; iptables -F OUTPUT; sleep 10; done Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Fixes: 5c86919455c1 ("smb: client: fix use-after-free in smb2_query_info_compound()") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:26 -08:00
if (cfid->has_lease && cfid->time) {
cifs: add new field to track the last access time of cfid The handlecache code today tracks the time at which dir lease was acquired and the laundromat thread uses that to check for old entries to cleanup. However, if a directory is actively accessed, it should not be chosen to expire first. This change adds a new last_access_time field to cfid and uses that to decide expiry of the cfid. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-07-24 22:23:53 -05:00
cfid->last_access_time = jiffies;
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
spin_unlock(&cfids->cfid_list_lock);
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
*ret_cfid = cfid;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
kfree(utf16_path);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
return 0;
}
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
spin_unlock(&cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
smb: client: use ParentLeaseKey in open_cached_dir Implement ParentLeaseKey logic in open_cached_dir() by looking up the parent cfid, copying its lease key into the fid struct, and setting the appropriate lease flag. Fixes: f047390a097e ("CIFS: Add create lease v2 context for SMB3") Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-08 13:37:50 -03:00
pfid = &cfid->fid;
cifs: fix dentry lookups in directory handle cache Get rid of any prefix paths in @path before lookup_positive_unlocked() as it will call ->lookup() which already adds those prefix paths through build_path_from_dentry(). This has caused a performance regression when mounting shares with a prefix path where readdir(2) would end up retrying several times to open bad directory names that contained duplicate prefix paths. Fix this by skipping any prefix paths in @path before calling lookup_positive_unlocked(). Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-24 13:56:33 -03:00
/*
VFS: rename lookup_one_len family to lookup_noperm and remove permission check The lookup_one_len family of functions is (now) only used internally by a filesystem on itself either - in a context where permission checking is irrelevant such as by a virtual filesystem populating itself, or xfs accessing its ORPHANAGE or dquota accessing the quota file; or - in a context where a permission check (MAY_EXEC on the parent) has just been performed such as a network filesystem finding in "silly-rename" file in the same directory. This is also the context after the _parentat() functions where currently lookup_one_qstr_excl() is used. So the permission check is pointless. The name "one_len" is unhelpful in understanding the purpose of these functions and should be changed. Most of the callers pass the len as "strlen()" so using a qstr and QSTR() can simplify the code. This patch renames these functions (include lookup_positive_unlocked() which is part of the family despite the name) to have a name based on "lookup_noperm". They are changed to receive a 'struct qstr' instead of separate name and len. In a few cases the use of QSTR() results in a new call to strlen(). try_lookup_noperm() takes a pointer to a qstr instead of the whole qstr. This is consistent with d_hash_and_lookup() (which is nearly identical) and useful for lookup_noperm_unlocked(). The new lookup_noperm_common() doesn't take a qstr yet. That will be tidied up in a subsequent patch. Signed-off-by: NeilBrown <neil@brown.name> Link: https://lore.kernel.org/r/20250319031545.2999807-5-neil@brown.name Signed-off-by: Christian Brauner <brauner@kernel.org>
2025-03-19 14:01:35 +11:00
* Skip any prefix paths in @path as lookup_noperm_positive_unlocked() ends up
cifs: fix dentry lookups in directory handle cache Get rid of any prefix paths in @path before lookup_positive_unlocked() as it will call ->lookup() which already adds those prefix paths through build_path_from_dentry(). This has caused a performance regression when mounting shares with a prefix path where readdir(2) would end up retrying several times to open bad directory names that contained duplicate prefix paths. Fix this by skipping any prefix paths in @path before calling lookup_positive_unlocked(). Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-24 13:56:33 -03:00
* calling ->lookup() which already adds those through
* build_path_from_dentry(). Also, do it earlier as we might reconnect
* below when trying to send compounded request and then potentially
* having a different prefix path (e.g. after DFS failover).
*/
npath = path_no_prefix(cifs_sb, path);
if (IS_ERR(npath)) {
rc = PTR_ERR(npath);
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
goto out;
cifs: fix dentry lookups in directory handle cache Get rid of any prefix paths in @path before lookup_positive_unlocked() as it will call ->lookup() which already adds those prefix paths through build_path_from_dentry(). This has caused a performance regression when mounting shares with a prefix path where readdir(2) would end up retrying several times to open bad directory names that contained duplicate prefix paths. Fix this by skipping any prefix paths in @path before calling lookup_positive_unlocked(). Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-24 13:56:33 -03:00
}
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (!npath[0]) {
dentry = dget(cifs_sb->root);
} else {
dentry = path_to_dentry(cifs_sb, npath);
if (IS_ERR(dentry)) {
rc = -ENOENT;
goto out;
}
smb: client: use ParentLeaseKey in open_cached_dir Implement ParentLeaseKey logic in open_cached_dir() by looking up the parent cfid, copying its lease key into the fid struct, and setting the appropriate lease flag. Fixes: f047390a097e ("CIFS: Add create lease v2 context for SMB3") Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-08 13:37:50 -03:00
if (dentry->d_parent && server->dialect >= SMB30_PROT_ID) {
struct cached_fid *parent_cfid;
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry(parent_cfid, &cfids->entries, entry) {
if (parent_cfid->dentry == dentry->d_parent) {
cifs_dbg(FYI, "found a parent cached file handle\n");
if (parent_cfid->has_lease && parent_cfid->time) {
lease_flags
|= SMB2_LEASE_FLAG_PARENT_LEASE_KEY_SET_LE;
memcpy(pfid->parent_lease_key,
parent_cfid->fid.lease_key,
SMB2_LEASE_KEY_SIZE);
}
break;
}
}
spin_unlock(&cfids->cfid_list_lock);
}
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
}
cfid->dentry = dentry;
smb: Initialize cfid->tcon before performing network ops Avoid leaking a tcon ref when a lease break races with opening the cached directory. Processing the leak break might take a reference to the tcon in cached_dir_lease_break() and then fail to release the ref in cached_dir_offload_close, since cfid->tcon is still NULL. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-26 18:50:31 -06:00
cfid->tcon = tcon;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
/*
* We do not hold the lock for the open because in case
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
* SMB2_open needs to reconnect.
* This is safe because no other thread will be able to get a ref
* to the cfid until we have finished opening the file and (possibly)
* acquired a lease.
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
*/
if (smb3_encryption_required(tcon))
flags |= CIFS_TRANSFORM_REQ;
server->ops->new_lease_key(pfid);
memset(rqst, 0, sizeof(rqst));
resp_buftype[0] = resp_buftype[1] = CIFS_NO_BUFFER;
memset(rsp_iov, 0, sizeof(rsp_iov));
/* Open */
memset(&open_iov, 0, sizeof(open_iov));
rqst[0].rq_iov = open_iov;
rqst[0].rq_nvec = SMB2_CREATE_IOV_SIZE;
cifs: Fix uninitialized memory reads for oparms.mode Use a struct assignment with implicit member initialization Signed-off-by: Volker Lendecke <vl@samba.org> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2023-01-11 12:37:58 +01:00
oparms = (struct cifs_open_parms) {
.tcon = tcon,
cifs: append path to open_enter trace event We do not dump the file path for smb3_open_enter ftrace calls, which is a severe handicap while debugging using ftrace evens. This change adds that info. Unfortunately, we're not updating the path in open params in many places; which I had to do as a part of this change. SMB2_open gets path in utf16 format, but it's easier of path is supplied as char pointer in oparms. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2023-03-17 12:51:17 +00:00
.path = path,
cifs: Fix uninitialized memory reads for oparms.mode Use a struct assignment with implicit member initialization Signed-off-by: Volker Lendecke <vl@samba.org> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2023-01-11 12:37:58 +01:00
.create_options = cifs_create_options(cifs_sb, CREATE_NOT_FILE),
cifs: open_cached_dir(): add FILE_READ_EA to desired access Since smb2_query_eas() reads EA and uses cached directory, open_cached_dir() should request FILE_READ_EA access. Otherwise listxattr() and getxattr() will fail with EACCES (0xc0000022 STATUS_ACCESS_DENIED SMB status). Link: https://bugzilla.kernel.org/show_bug.cgi?id=218543 Cc: stable@vger.kernel.org Signed-off-by: Eugene Korenevsky <ekorenevsky@astralinux.ru> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-03-01 17:53:44 +03:00
.desired_access = FILE_READ_DATA | FILE_READ_ATTRIBUTES |
FILE_READ_EA,
cifs: Fix uninitialized memory reads for oparms.mode Use a struct assignment with implicit member initialization Signed-off-by: Volker Lendecke <vl@samba.org> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2023-01-11 12:37:58 +01:00
.disposition = FILE_OPEN,
.fid = pfid,
smb: client: use ParentLeaseKey in open_cached_dir Implement ParentLeaseKey logic in open_cached_dir() by looking up the parent cfid, copying its lease key into the fid struct, and setting the appropriate lease flag. Fixes: f047390a097e ("CIFS: Add create lease v2 context for SMB3") Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-05-08 13:37:50 -03:00
.lease_flags = lease_flags,
cifs: update the same create_guid on replay File open requests made to the server contain a CreateGuid, which is used by the server to identify the open request. If the same request needs to be replayed, it needs to be sent with the same CreateGuid in the durable handle v2 context. Without doing so, we could end up leaking handles on the server when: 1. multichannel is used AND 2. connection goes down, but not for all channels This is because the replayed open request would have a new CreateGuid and the server will treat this as a new request and open a new handle. This change fixes this by reusing the existing create_guid stored in the cached fid struct. REF: MS-SMB2 4.9 Replay Create Request on an Alternate Channel Fixes: 4f1fffa23769 ("cifs: commands that are retried should have replay flag set") Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-02-09 11:25:42 +00:00
.replay = !!(retries),
cifs: Fix uninitialized memory reads for oparms.mode Use a struct assignment with implicit member initialization Signed-off-by: Volker Lendecke <vl@samba.org> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2023-01-11 12:37:58 +01:00
};
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
rc = SMB2_open_init(tcon, server,
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
&rqst[0], &oplock, &oparms, utf16_path);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
if (rc)
goto oshr_free;
smb2_set_next_command(tcon, &rqst[0]);
memset(&qi_iov, 0, sizeof(qi_iov));
rqst[1].rq_iov = qi_iov;
rqst[1].rq_nvec = 1;
rc = SMB2_query_info_init(tcon, server,
&rqst[1], COMPOUND_FID,
COMPOUND_FID, FILE_ALL_INFORMATION,
SMB2_O_INFO_FILE, 0,
sizeof(struct smb2_file_all_info) +
PATH_MAX * 2, 0, NULL);
if (rc)
goto oshr_free;
smb2_set_related(&rqst[1]);
cifs: commands that are retried should have replay flag set MS-SMB2 states that the header flag SMB2_FLAGS_REPLAY_OPERATION needs to be set when a command needs to be retried, so that the server is aware that this is a replay for an operation that appeared before. This can be very important, for example, for state changing operations and opens which get retried following a reconnect; since the client maybe unaware of the status of the previous open. This is particularly important for multichannel scenario, since disconnection of one connection does not mean that the session is lost. The requests can be replayed on another channel. This change also makes use of exponential back-off before replays and also limits the number of retries to "retrans" mount option value. Also, this change does not modify the read/write codepath. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-01-21 03:32:47 +00:00
if (retries) {
smb2_set_replay(server, &rqst[0]);
smb2_set_replay(server, &rqst[1]);
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
rc = compound_send_recv(xid, ses, server,
flags, 2, rqst,
resp_buftype, rsp_iov);
if (rc) {
if (rc == -EREMCHG) {
tcon->need_reconnect = true;
pr_warn_once("server share %s deleted\n",
smb3: add dynamic trace points for tree disconnect Needed this for debugging a failing xfstest. Also change camel case for "treeName" to "tree_name" in tcon struct. Example trace output (from "trace-cmd record -e smb3_tdis*"): umount-9718 [006] ..... 5909.780244: smb3_tdis_enter: xid=206 sid=0xcf38894e tid=0x3d0b8cf8 path=\\localhost\test umount-9718 [007] ..... 5909.780878: smb3_tdis_done: xid=206 sid=0xcf38894e tid=0x3d0b8cf8 Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-09-21 14:05:53 -05:00
tcon->tree_name);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
goto oshr_free;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: Check the lease context if we actually got a lease Some servers may return that we got a lease in rsp->OplockLevel but then in the lease context contradict this and say we got no lease at all. Thus we need to check the context if we have a lease. Additionally, If we do not get a lease we need to make sure we close the handle before we return an error to the caller. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Cc: stable@vger.kernel.org Reviewed-by: Bharath SM <bharathsm@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-02-17 13:35:00 +10:00
cfid->is_open = true;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
spin_lock(&cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
o_rsp = (struct smb2_create_rsp *)rsp_iov[0].iov_base;
oparms.fid->persistent_fid = o_rsp->PersistentFileId;
oparms.fid->volatile_fid = o_rsp->VolatileFileId;
#ifdef CONFIG_CIFS_DEBUG2
oparms.fid->mid = le64_to_cpu(o_rsp->hdr.MessageId);
#endif /* CIFS_DEBUG2 */
smb: client: fix potential OOBs in smb2_parse_contexts() Validate offsets and lengths before dereferencing create contexts in smb2_parse_contexts(). This fixes following oops when accessing invalid create contexts from server: BUG: unable to handle page fault for address: ffff8881178d8cc3 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 4a01067 P4D 4a01067 PUD 0 Oops: 0000 [#1] PREEMPT SMP NOPTI CPU: 3 PID: 1736 Comm: mount.cifs Not tainted 6.7.0-rc4 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 RIP: 0010:smb2_parse_contexts+0xa0/0x3a0 [cifs] Code: f8 10 75 13 48 b8 93 ad 25 50 9c b4 11 e7 49 39 06 0f 84 d2 00 00 00 8b 45 00 85 c0 74 61 41 29 c5 48 01 c5 41 83 fd 0f 76 55 <0f> b7 7d 04 0f b7 45 06 4c 8d 74 3d 00 66 83 f8 04 75 bc ba 04 00 RSP: 0018:ffffc900007939e0 EFLAGS: 00010216 RAX: ffffc90000793c78 RBX: ffff8880180cc000 RCX: ffffc90000793c90 RDX: ffffc90000793cc0 RSI: ffff8880178d8cc0 RDI: ffff8880180cc000 RBP: ffff8881178d8cbf R08: ffffc90000793c22 R09: 0000000000000000 R10: ffff8880180cc000 R11: 0000000000000024 R12: 0000000000000000 R13: 0000000000000020 R14: 0000000000000000 R15: ffffc90000793c22 FS: 00007f873753cbc0(0000) GS:ffff88806bc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffff8881178d8cc3 CR3: 00000000181ca000 CR4: 0000000000750ef0 PKRU: 55555554 Call Trace: <TASK> ? __die+0x23/0x70 ? page_fault_oops+0x181/0x480 ? search_module_extables+0x19/0x60 ? srso_alias_return_thunk+0x5/0xfbef5 ? exc_page_fault+0x1b6/0x1c0 ? asm_exc_page_fault+0x26/0x30 ? smb2_parse_contexts+0xa0/0x3a0 [cifs] SMB2_open+0x38d/0x5f0 [cifs] ? smb2_is_path_accessible+0x138/0x260 [cifs] smb2_is_path_accessible+0x138/0x260 [cifs] cifs_is_path_remote+0x8d/0x230 [cifs] cifs_mount+0x7e/0x350 [cifs] cifs_smb3_do_mount+0x128/0x780 [cifs] smb3_get_tree+0xd9/0x290 [cifs] vfs_get_tree+0x2c/0x100 ? capable+0x37/0x70 path_mount+0x2d7/0xb80 ? srso_alias_return_thunk+0x5/0xfbef5 ? _raw_spin_unlock_irqrestore+0x44/0x60 __x64_sys_mount+0x11a/0x150 do_syscall_64+0x47/0xf0 entry_SYSCALL_64_after_hwframe+0x6f/0x77 RIP: 0033:0x7f8737657b1e Reported-by: Robert Morris <rtm@csail.mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-12-11 10:26:41 -03:00
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (o_rsp->OplockLevel != SMB2_OPLOCK_LEVEL_LEASE) {
smb: client: fix potential OOBs in smb2_parse_contexts() Validate offsets and lengths before dereferencing create contexts in smb2_parse_contexts(). This fixes following oops when accessing invalid create contexts from server: BUG: unable to handle page fault for address: ffff8881178d8cc3 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 4a01067 P4D 4a01067 PUD 0 Oops: 0000 [#1] PREEMPT SMP NOPTI CPU: 3 PID: 1736 Comm: mount.cifs Not tainted 6.7.0-rc4 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 RIP: 0010:smb2_parse_contexts+0xa0/0x3a0 [cifs] Code: f8 10 75 13 48 b8 93 ad 25 50 9c b4 11 e7 49 39 06 0f 84 d2 00 00 00 8b 45 00 85 c0 74 61 41 29 c5 48 01 c5 41 83 fd 0f 76 55 <0f> b7 7d 04 0f b7 45 06 4c 8d 74 3d 00 66 83 f8 04 75 bc ba 04 00 RSP: 0018:ffffc900007939e0 EFLAGS: 00010216 RAX: ffffc90000793c78 RBX: ffff8880180cc000 RCX: ffffc90000793c90 RDX: ffffc90000793cc0 RSI: ffff8880178d8cc0 RDI: ffff8880180cc000 RBP: ffff8881178d8cbf R08: ffffc90000793c22 R09: 0000000000000000 R10: ffff8880180cc000 R11: 0000000000000024 R12: 0000000000000000 R13: 0000000000000020 R14: 0000000000000000 R15: ffffc90000793c22 FS: 00007f873753cbc0(0000) GS:ffff88806bc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffff8881178d8cc3 CR3: 00000000181ca000 CR4: 0000000000750ef0 PKRU: 55555554 Call Trace: <TASK> ? __die+0x23/0x70 ? page_fault_oops+0x181/0x480 ? search_module_extables+0x19/0x60 ? srso_alias_return_thunk+0x5/0xfbef5 ? exc_page_fault+0x1b6/0x1c0 ? asm_exc_page_fault+0x26/0x30 ? smb2_parse_contexts+0xa0/0x3a0 [cifs] SMB2_open+0x38d/0x5f0 [cifs] ? smb2_is_path_accessible+0x138/0x260 [cifs] smb2_is_path_accessible+0x138/0x260 [cifs] cifs_is_path_remote+0x8d/0x230 [cifs] cifs_mount+0x7e/0x350 [cifs] cifs_smb3_do_mount+0x128/0x780 [cifs] smb3_get_tree+0xd9/0x290 [cifs] vfs_get_tree+0x2c/0x100 ? capable+0x37/0x70 path_mount+0x2d7/0xb80 ? srso_alias_return_thunk+0x5/0xfbef5 ? _raw_spin_unlock_irqrestore+0x44/0x60 __x64_sys_mount+0x11a/0x150 do_syscall_64+0x47/0xf0 entry_SYSCALL_64_after_hwframe+0x6f/0x77 RIP: 0033:0x7f8737657b1e Reported-by: Robert Morris <rtm@csail.mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-12-11 10:26:41 -03:00
spin_unlock(&cfids->cfid_list_lock);
rc = -EINVAL;
goto oshr_free;
}
rc = smb2_parse_contexts(server, rsp_iov,
&oparms.fid->epoch,
oparms.fid->lease_key,
&oplock, NULL, NULL);
if (rc) {
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
spin_unlock(&cfids->cfid_list_lock);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
goto oshr_free;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
smb: client: fix potential OOBs in smb2_parse_contexts() Validate offsets and lengths before dereferencing create contexts in smb2_parse_contexts(). This fixes following oops when accessing invalid create contexts from server: BUG: unable to handle page fault for address: ffff8881178d8cc3 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 4a01067 P4D 4a01067 PUD 0 Oops: 0000 [#1] PREEMPT SMP NOPTI CPU: 3 PID: 1736 Comm: mount.cifs Not tainted 6.7.0-rc4 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 RIP: 0010:smb2_parse_contexts+0xa0/0x3a0 [cifs] Code: f8 10 75 13 48 b8 93 ad 25 50 9c b4 11 e7 49 39 06 0f 84 d2 00 00 00 8b 45 00 85 c0 74 61 41 29 c5 48 01 c5 41 83 fd 0f 76 55 <0f> b7 7d 04 0f b7 45 06 4c 8d 74 3d 00 66 83 f8 04 75 bc ba 04 00 RSP: 0018:ffffc900007939e0 EFLAGS: 00010216 RAX: ffffc90000793c78 RBX: ffff8880180cc000 RCX: ffffc90000793c90 RDX: ffffc90000793cc0 RSI: ffff8880178d8cc0 RDI: ffff8880180cc000 RBP: ffff8881178d8cbf R08: ffffc90000793c22 R09: 0000000000000000 R10: ffff8880180cc000 R11: 0000000000000024 R12: 0000000000000000 R13: 0000000000000020 R14: 0000000000000000 R15: ffffc90000793c22 FS: 00007f873753cbc0(0000) GS:ffff88806bc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: ffff8881178d8cc3 CR3: 00000000181ca000 CR4: 0000000000750ef0 PKRU: 55555554 Call Trace: <TASK> ? __die+0x23/0x70 ? page_fault_oops+0x181/0x480 ? search_module_extables+0x19/0x60 ? srso_alias_return_thunk+0x5/0xfbef5 ? exc_page_fault+0x1b6/0x1c0 ? asm_exc_page_fault+0x26/0x30 ? smb2_parse_contexts+0xa0/0x3a0 [cifs] SMB2_open+0x38d/0x5f0 [cifs] ? smb2_is_path_accessible+0x138/0x260 [cifs] smb2_is_path_accessible+0x138/0x260 [cifs] cifs_is_path_remote+0x8d/0x230 [cifs] cifs_mount+0x7e/0x350 [cifs] cifs_smb3_do_mount+0x128/0x780 [cifs] smb3_get_tree+0xd9/0x290 [cifs] vfs_get_tree+0x2c/0x100 ? capable+0x37/0x70 path_mount+0x2d7/0xb80 ? srso_alias_return_thunk+0x5/0xfbef5 ? _raw_spin_unlock_irqrestore+0x44/0x60 __x64_sys_mount+0x11a/0x150 do_syscall_64+0x47/0xf0 entry_SYSCALL_64_after_hwframe+0x6f/0x77 RIP: 0033:0x7f8737657b1e Reported-by: Robert Morris <rtm@csail.mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-12-11 10:26:41 -03:00
rc = -EINVAL;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (!(oplock & SMB2_LEASE_READ_CACHING_HE)) {
spin_unlock(&cfids->cfid_list_lock);
cifs: Check the lease context if we actually got a lease Some servers may return that we got a lease in rsp->OplockLevel but then in the lease context contradict this and say we got no lease at all. Thus we need to check the context if we have a lease. Additionally, If we do not get a lease we need to make sure we close the handle before we return an error to the caller. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Cc: stable@vger.kernel.org Reviewed-by: Bharath SM <bharathsm@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-02-17 13:35:00 +10:00
goto oshr_free;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
qi_rsp = (struct smb2_query_info_rsp *)rsp_iov[1].iov_base;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (le32_to_cpu(qi_rsp->OutputBufferLength) < sizeof(struct smb2_file_all_info)) {
spin_unlock(&cfids->cfid_list_lock);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
goto oshr_free;
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
if (!smb2_validate_and_copy_iov(
le16_to_cpu(qi_rsp->OutputBufferOffset),
sizeof(struct smb2_file_all_info),
&rsp_iov[1], sizeof(struct smb2_file_all_info),
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
(char *)&cfid->file_all_info))
cfid->file_all_info_is_valid = true;
cifs: find and use the dentry for cached non-root directories also This allows us to use cached attributes for the entries in a cached directory for as long as a lease is held on the directory itself. Previously we have always allowed "used cached attributes for 1 second" but this extends this to the lifetime of the lease as well as making the caching safer. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-12 06:13:03 -05:00
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
cfid->time = jiffies;
cifs: add new field to track the last access time of cfid The handlecache code today tracks the time at which dir lease was acquired and the laundromat thread uses that to check for old entries to cleanup. However, if a directory is actively accessed, it should not be chosen to expire first. This change adds a new last_access_time field to cfid and uses that to decide expiry of the cfid. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-07-24 22:23:53 -05:00
cfid->last_access_time = jiffies;
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
spin_unlock(&cfids->cfid_list_lock);
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
/* At this point the directory handle is fully cached */
rc = 0;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
oshr_free:
SMB2_open_free(&rqst[0]);
SMB2_query_info_free(&rqst[1]);
free_rsp_buf(resp_buftype[0], rsp_iov[0].iov_base);
free_rsp_buf(resp_buftype[1], rsp_iov[1].iov_base);
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
out:
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
if (rc) {
spin_lock(&cfids->cfid_list_lock);
if (cfid->on_list) {
list_del(&cfid->entry);
cfid->on_list = false;
cfids->num_entries--;
}
if (cfid->has_lease) {
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
/*
* We are guaranteed to have two references at this
* point. One for the caller and one for a potential
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
* lease. Release one here, and the second below.
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
*/
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
cfid->has_lease = false;
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
kref_put(&cfid->refcount, smb2_close_cached_fid);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
}
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
spin_unlock(&cfids->cfid_list_lock);
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
kref_put(&cfid->refcount, smb2_close_cached_fid);
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
} else {
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
*ret_cfid = cfid;
cifs: Check the lease context if we actually got a lease Some servers may return that we got a lease in rsp->OplockLevel but then in the lease context contradict this and say we got no lease at all. Thus we need to check the context if we have a lease. Additionally, If we do not get a lease we need to make sure we close the handle before we return an error to the caller. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Cc: stable@vger.kernel.org Reviewed-by: Bharath SM <bharathsm@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-02-17 13:35:00 +10:00
atomic_inc(&tcon->num_remote_opens);
}
smb: client: fix use-after-free in smb2_query_info_compound() The following UAF was triggered when running fstests generic/072 with KASAN enabled against Windows Server 2022 and mount options 'multichannel,max_channels=2,vers=3.1.1,mfsymlinks,noperm' BUG: KASAN: slab-use-after-free in smb2_query_info_compound+0x423/0x6d0 [cifs] Read of size 8 at addr ffff888014941048 by task xfs_io/27534 CPU: 0 PID: 27534 Comm: xfs_io Not tainted 6.6.0-rc7 #1 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-3-gd478f380-rebuilt.opensuse.org 04/01/2014 Call Trace: dump_stack_lvl+0x4a/0x80 print_report+0xcf/0x650 ? srso_alias_return_thunk+0x5/0x7f ? srso_alias_return_thunk+0x5/0x7f ? __phys_addr+0x46/0x90 kasan_report+0xda/0x110 ? smb2_query_info_compound+0x423/0x6d0 [cifs] ? smb2_query_info_compound+0x423/0x6d0 [cifs] smb2_query_info_compound+0x423/0x6d0 [cifs] ? __pfx_smb2_query_info_compound+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __stack_depot_save+0x39/0x480 ? kasan_save_stack+0x33/0x60 ? kasan_set_track+0x25/0x30 ? ____kasan_slab_free+0x126/0x170 smb2_queryfs+0xc2/0x2c0 [cifs] ? __pfx_smb2_queryfs+0x10/0x10 [cifs] ? __pfx___lock_acquire+0x10/0x10 smb311_queryfs+0x210/0x220 [cifs] ? __pfx_smb311_queryfs+0x10/0x10 [cifs] ? srso_alias_return_thunk+0x5/0x7f ? __lock_acquire+0x480/0x26c0 ? lock_release+0x1ed/0x640 ? srso_alias_return_thunk+0x5/0x7f ? do_raw_spin_unlock+0x9b/0x100 cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 ? __pfx___do_sys_fstatfs+0x10/0x10 ? srso_alias_return_thunk+0x5/0x7f ? lockdep_hardirqs_on_prepare+0x136/0x200 ? srso_alias_return_thunk+0x5/0x7f do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Allocated by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x8f/0xa0 open_cached_dir+0x71b/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] smb311_queryfs+0x210/0x220 [cifs] cifs_statfs+0x18c/0x4b0 [cifs] statfs_by_dentry+0x9b/0xf0 fd_statfs+0x4e/0xb0 __do_sys_fstatfs+0x7f/0xe0 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 27534: kasan_save_stack+0x33/0x60 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 ____kasan_slab_free+0x126/0x170 slab_free_freelist_hook+0xd0/0x1e0 __kmem_cache_free+0x9d/0x1b0 open_cached_dir+0xff5/0x1240 [cifs] smb2_query_info_compound+0x5c3/0x6d0 [cifs] smb2_queryfs+0xc2/0x2c0 [cifs] This is a race between open_cached_dir() and cached_dir_lease_break() where the cache entry for the open directory handle receives a lease break while creating it. And before returning from open_cached_dir(), we put the last reference of the new @cfid because of !@cfid->has_lease. Besides the UAF, while running xfstests a lot of missed lease breaks have been noticed in tests that run several concurrent statfs(2) calls on those cached fids CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 00000000715bfe83 len 108 CIFS: VFS: Dump pending requests: CIFS: VFS: \\w22-root1.gandalf.test No task to wake, unknown frame... CIFS: VFS: \\w22-root1.gandalf.test Cmd: 18 Err: 0x0 Flags: 0x1... CIFS: VFS: \\w22-root1.gandalf.test smb buf 000000005aa7316e len 108 ... To fix both, in open_cached_dir() ensure that @cfid->has_lease is set right before sending out compounded request so that any potential lease break will be get processed by demultiplex thread while we're still caching @cfid. And, if open failed for some reason, re-check @cfid->has_lease to decide whether or not put lease reference. Cc: stable@vger.kernel.org Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-30 17:19:56 -03:00
kfree(utf16_path);
cifs: helper function to check replayable error codes The code to check for replay is not just -EAGAIN. In some cases, the send request or receive response may result in network errors, which we're now mapping to -ECONNABORTED. This change introduces a helper function which checks if the error returned in one of the above two errors. And all checks for replays will now use this helper. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-01-21 03:32:46 +00:00
cifs: commands that are retried should have replay flag set MS-SMB2 states that the header flag SMB2_FLAGS_REPLAY_OPERATION needs to be set when a command needs to be retried, so that the server is aware that this is a replay for an operation that appeared before. This can be very important, for example, for state changing operations and opens which get retried following a reconnect; since the client maybe unaware of the status of the previous open. This is particularly important for multichannel scenario, since disconnection of one connection does not mean that the session is lost. The requests can be replayed on another channel. This change also makes use of exponential back-off before replays and also limits the number of retries to "retrans" mount option value. Also, this change does not modify the read/write codepath. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-01-21 03:32:47 +00:00
if (is_replayable_error(rc) &&
smb2_should_replay(tcon, &retries, &cur_sleep))
goto replay_again;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
return rc;
}
int open_cached_dir_by_dentry(struct cifs_tcon *tcon,
struct dentry *dentry,
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
struct cached_fid **ret_cfid)
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
{
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
struct cached_fid *cfid;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fids *cfids = tcon->cfids;
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
if (cfids == NULL)
smb: client: change return value in open_cached_dir_by_dentry() if !cfids Change return value from -ENOENT to -EOPNOTSUPP to maintain consistency with the return value of open_cached_dir() for the same case. This change is safe as the only calling function does not differentiate between these return values. Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.com> Reviewed-by: Enzo Matsumiya <ematsumiya@suse.de> Signed-off-by: Henrique Carvalho <henrique.carvalho@suse.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-22 22:14:37 -03:00
return -EOPNOTSUPP;
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry(cfid, &cfids->entries, entry) {
if (dentry && cfid->dentry == dentry) {
smb: cached directories can be more than root file handle Update this log message since cached fids may represent things other than the root of a mount. Fixes: e4029e072673 ("cifs: find and use the dentry for cached non-root directories also") Signed-off-by: Paul Aurich <paul@darkrain42.org> Reviewed-by: Bharath SM <bharathsm@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-08 14:29:02 -08:00
cifs_dbg(FYI, "found a cached file handle by dentry\n");
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
kref_get(&cfid->refcount);
*ret_cfid = cfid;
spin_unlock(&cfids->cfid_list_lock);
return 0;
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
return -ENOENT;
}
static void
smb2_close_cached_fid(struct kref *ref)
{
struct cached_fid *cfid = container_of(ref, struct cached_fid,
refcount);
smb3: retrying on failed server close In the current implementation, CIFS close sends a close to the server and does not check for the success of the server close. This patch adds functionality to check for server close return status and retries in case of an EBUSY or EAGAIN error. This can help avoid handle leaks Cc: stable@vger.kernel.org Signed-off-by: Ritvik Budhiraja <rbudhiraja@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-04-02 14:01:28 -05:00
int rc;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock(&cfid->cfids->cfid_list_lock);
if (cfid->on_list) {
list_del(&cfid->entry);
cfid->on_list = false;
cfid->cfids->num_entries--;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfid->cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
dput(cfid->dentry);
cfid->dentry = NULL;
if (cfid->is_open) {
smb3: retrying on failed server close In the current implementation, CIFS close sends a close to the server and does not check for the success of the server close. This patch adds functionality to check for server close return status and retries in case of an EBUSY or EAGAIN error. This can help avoid handle leaks Cc: stable@vger.kernel.org Signed-off-by: Ritvik Budhiraja <rbudhiraja@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-04-02 14:01:28 -05:00
rc = SMB2_close(0, cfid->tcon, cfid->fid.persistent_fid,
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfid->fid.volatile_fid);
smb3: fix Open files on server counter going negative We were decrementing the count of open files on server twice for the case where we were closing cached directories. Fixes: 8e843bf38f7b ("cifs: return a single-use cfid if we did not get a lease") Cc: stable@vger.kernel.org Acked-by: Bharath SM <bharathsm@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-04-06 23:16:08 -05:00
if (rc) /* should we retry on -EBUSY or -EAGAIN? */
cifs_dbg(VFS, "close cached dir rc %d\n", rc);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
free_cached_dir(cfid);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: drop the lease for cached directories on rmdir or rename When we delete or rename a directory we must also drop any cached lease we have on the directory. Fixes: a350d6e73f5e ("cifs: enable caching of directories for which a lease is held") Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-18 17:39:10 +10:00
void drop_cached_dir_by_name(const unsigned int xid, struct cifs_tcon *tcon,
const char *name, struct cifs_sb_info *cifs_sb)
{
struct cached_fid *cfid = NULL;
int rc;
rc = open_cached_dir(xid, tcon, name, cifs_sb, true, &cfid);
if (rc) {
cifs_dbg(FYI, "no cached dir found for rmdir(%s)\n", name);
return;
}
spin_lock(&cfid->cfids->cfid_list_lock);
if (cfid->has_lease) {
cfid->has_lease = false;
kref_put(&cfid->refcount, smb2_close_cached_fid);
}
spin_unlock(&cfid->cfids->cfid_list_lock);
close_cached_dir(cfid);
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
void close_cached_dir(struct cached_fid *cfid)
{
kref_put(&cfid->refcount, smb2_close_cached_fid);
}
/*
* Called from cifs_kill_sb when we unmount a share
*/
void close_all_cached_dirs(struct cifs_sb_info *cifs_sb)
{
struct rb_root *root = &cifs_sb->tlink_tree;
struct rb_node *node;
struct cached_fid *cfid;
struct cifs_tcon *tcon;
struct tcon_link *tlink;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fids *cfids;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
struct cached_dir_dentry *tmp_list, *q;
LIST_HEAD(entry);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_lock(&cifs_sb->tlink_tree_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
for (node = rb_first(root); node; node = rb_next(node)) {
tlink = rb_entry(node, struct tcon_link, tl_rbnode);
tcon = tlink_tcon(tlink);
if (IS_ERR(tcon))
continue;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfids = tcon->cfids;
if (cfids == NULL)
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
continue;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_lock(&cfids->cfid_list_lock);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
list_for_each_entry(cfid, &cfids->entries, entry) {
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
tmp_list = kmalloc(sizeof(*tmp_list), GFP_ATOMIC);
smb: Log an error when close_all_cached_dirs fails Under low-memory conditions, close_all_cached_dirs() can't move the dentries to a separate list to dput() them once the locks are dropped. This will result in a "Dentry still in use" error, so add an error message that makes it clear this is what happened: [ 495.281119] CIFS: VFS: \\otters.example.com\share Out of memory while dropping dentries [ 495.281595] ------------[ cut here ]------------ [ 495.281887] BUG: Dentry ffff888115531138{i=78,n=/} still in use (2) [unmount of cifs cifs] [ 495.282391] WARNING: CPU: 1 PID: 2329 at fs/dcache.c:1536 umount_check+0xc8/0xf0 Also, bail out of looping through all tcons as soon as a single allocation fails, since we're already in trouble, and kmalloc() attempts for subseqeuent tcons are likely to fail just like the first one did. Signed-off-by: Paul Aurich <paul@darkrain42.org> Acked-by: Bharath SM <bharathsm@microsoft.com> Suggested-by: Ruben Devos <rdevos@oxya.com> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-20 08:01:54 -08:00
if (tmp_list == NULL) {
/*
* If the malloc() fails, we won't drop all
* dentries, and unmounting is likely to trigger
* a 'Dentry still in use' error.
*/
cifs_tcon_dbg(VFS, "Out of memory while dropping dentries\n");
spin_unlock(&cfids->cfid_list_lock);
spin_unlock(&cifs_sb->tlink_tree_lock);
goto done;
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_lock(&cfid->fid_lock);
tmp_list->dentry = cfid->dentry;
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
cfid->dentry = NULL;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_unlock(&cfid->fid_lock);
list_add_tail(&tmp_list->entry, &entry);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_unlock(&cfids->cfid_list_lock);
}
spin_unlock(&cifs_sb->tlink_tree_lock);
smb: Log an error when close_all_cached_dirs fails Under low-memory conditions, close_all_cached_dirs() can't move the dentries to a separate list to dput() them once the locks are dropped. This will result in a "Dentry still in use" error, so add an error message that makes it clear this is what happened: [ 495.281119] CIFS: VFS: \\otters.example.com\share Out of memory while dropping dentries [ 495.281595] ------------[ cut here ]------------ [ 495.281887] BUG: Dentry ffff888115531138{i=78,n=/} still in use (2) [unmount of cifs cifs] [ 495.282391] WARNING: CPU: 1 PID: 2329 at fs/dcache.c:1536 umount_check+0xc8/0xf0 Also, bail out of looping through all tcons as soon as a single allocation fails, since we're already in trouble, and kmalloc() attempts for subseqeuent tcons are likely to fail just like the first one did. Signed-off-by: Paul Aurich <paul@darkrain42.org> Acked-by: Bharath SM <bharathsm@microsoft.com> Suggested-by: Ruben Devos <rdevos@oxya.com> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-20 08:01:54 -08:00
done:
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
list_for_each_entry_safe(tmp_list, q, &entry, entry) {
list_del(&tmp_list->entry);
dput(tmp_list->dentry);
kfree(tmp_list);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
/* Flush any pending work that will drop dentries */
flush_workqueue(cfid_put_wq);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
/*
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
* Invalidate all cached dirs when a TCON has been reset
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
* due to a session loss.
*/
void invalidate_all_cached_dirs(struct cifs_tcon *tcon)
{
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fids *cfids = tcon->cfids;
struct cached_fid *cfid, *q;
smb3: do not start laundromat thread when dir leases disabled When no directory lease support, or for IPC shares where directories can not be opened, do not start an unneeded laundromat thread for that mount (it wastes resources). Fixes: d14de8067e3f ("cifs: Add a laundromat thread for cached directories") Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Acked-by: Tom Talpey <tom@talpey.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-09-19 11:35:53 -05:00
if (cfids == NULL)
return;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
/*
* Mark all the cfids as closed, and move them to the cfids->dying list.
* They'll be cleaned up later by cfids_invalidation_worker. Take
* a reference to each cfid during this process.
*/
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry_safe(cfid, q, &cfids->entries, entry) {
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
list_move(&cfid->entry, &cfids->dying);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfids->num_entries--;
cfid->is_open = false;
cifs: set rc to -ENOENT if we can not get a dentry for the cached dir We already set rc to this return code further down in the function but we can set it earlier in order to suppress a smash warning. Also fix a false positive for Coverity. The reason this is a false positive is that this happens during umount after all files and directories have been closed but mosetting on ->on_list to suppress the warning. Reported-by: Dan carpenter <dan.carpenter@oracle.com> Reported-by: coverity-bot <keescook+coverity-bot@chromium.org> Addresses-Coverity-ID: 1525256 ("Concurrent data access violations") Fixes: a350d6e73f5e ("cifs: enable caching of directories for which a lease is held") Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-17 18:48:26 -05:00
cfid->on_list = false;
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
if (cfid->has_lease) {
/*
* The lease was never cancelled from the server,
* so steal that reference.
*/
cfid->has_lease = false;
} else
kref_get(&cfid->refcount);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
/*
* Queue dropping of the dentries once locks have been dropped
*/
if (!list_empty(&cfids->dying))
queue_work(cfid_put_wq, &cfids->invalidation_work);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
static void
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
cached_dir_offload_close(struct work_struct *work)
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
{
struct cached_fid *cfid = container_of(work,
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
struct cached_fid, close_work);
struct cifs_tcon *tcon = cfid->tcon;
WARN_ON(cfid->on_list);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
kref_put(&cfid->refcount, smb2_close_cached_fid);
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
cifs_put_tcon(tcon, netfs_trace_tcon_ref_put_cached_close);
}
/*
* Release the cached directory's dentry, and then queue work to drop cached
* directory itself (closing on server if needed).
*
* Must be called with a reference to the cached_fid and a reference to the
* tcon.
*/
static void cached_dir_put_work(struct work_struct *work)
{
struct cached_fid *cfid = container_of(work, struct cached_fid,
put_work);
struct dentry *dentry;
spin_lock(&cfid->fid_lock);
dentry = cfid->dentry;
cfid->dentry = NULL;
spin_unlock(&cfid->fid_lock);
dput(dentry);
queue_work(serverclose_wq, &cfid->close_work);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
smb: change return type of cached_dir_lease_break() to bool cached_dir_lease_break() has return type as int but only returning true or false. change return type of this function to bool for clarity. Signed-off-by: Bharath SM <bharathsm@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-07-01 00:19:32 +05:30
bool cached_dir_lease_break(struct cifs_tcon *tcon, __u8 lease_key[16])
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
{
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fids *cfids = tcon->cfids;
struct cached_fid *cfid;
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
if (cfids == NULL)
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
return false;
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry(cfid, &cfids->entries, entry) {
if (cfid->has_lease &&
!memcmp(lease_key,
cfid->fid.lease_key,
SMB2_LEASE_KEY_SIZE)) {
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
cfid->has_lease = false;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfid->time = 0;
/*
* We found a lease remove it from the list
* so no threads can access it.
*/
list_del(&cfid->entry);
cfid->on_list = false;
cfids->num_entries--;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
++tcon->tc_count;
trace_smb3_tcon_ref(tcon->debug_id, tcon->tc_count,
netfs_trace_tcon_ref_get_cached_lease_break);
queue_work(cfid_put_wq, &cfid->put_work);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfids->cfid_list_lock);
return true;
}
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfids->cfid_list_lock);
cifs: Move cached-dir functions into a separate file Also rename crfid to cfid to have consistent naming for this variable. This commit does not change any logic. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-10 22:00:08 -05:00
return false;
}
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
static struct cached_fid *init_cached_dir(const char *path)
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
{
struct cached_fid *cfid;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfid = kzalloc(sizeof(*cfid), GFP_ATOMIC);
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
if (!cfid)
return NULL;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
cfid->path = kstrdup(path, GFP_ATOMIC);
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
if (!cfid->path) {
kfree(cfid);
return NULL;
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
INIT_WORK(&cfid->close_work, cached_dir_offload_close);
INIT_WORK(&cfid->put_work, cached_dir_put_work);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
INIT_LIST_HEAD(&cfid->entry);
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
INIT_LIST_HEAD(&cfid->dirents.entries);
mutex_init(&cfid->dirents.de_mutex);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock_init(&cfid->fid_lock);
kref_init(&cfid->refcount);
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
return cfid;
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
static void free_cached_dir(struct cached_fid *cfid)
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
{
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_dirent *dirent, *q;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
WARN_ON(work_pending(&cfid->close_work));
WARN_ON(work_pending(&cfid->put_work));
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
dput(cfid->dentry);
cfid->dentry = NULL;
/*
* Delete all cached dirent names
*/
list_for_each_entry_safe(dirent, q, &cfid->dirents.entries, entry) {
list_del(&dirent->entry);
kfree(dirent->name);
kfree(dirent);
}
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
kfree(cfid->path);
cfid->path = NULL;
kfree(cfid);
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
static void cfids_invalidation_worker(struct work_struct *work)
{
struct cached_fids *cfids = container_of(work, struct cached_fids,
invalidation_work);
struct cached_fid *cfid, *q;
LIST_HEAD(entry);
spin_lock(&cfids->cfid_list_lock);
/* move cfids->dying to the local list */
list_cut_before(&entry, &cfids->dying, &cfids->dying);
spin_unlock(&cfids->cfid_list_lock);
list_for_each_entry_safe(cfid, q, &entry, entry) {
list_del(&cfid->entry);
/* Drop the ref-count acquired in invalidate_all_cached_dirs */
kref_put(&cfid->refcount, smb2_close_cached_fid);
}
}
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
static void cfids_laundromat_worker(struct work_struct *work)
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
{
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
struct cached_fids *cfids;
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
struct cached_fid *cfid, *q;
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
struct dentry *dentry;
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
LIST_HEAD(entry);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
cfids = container_of(work, struct cached_fids, laundromat_work.work);
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry_safe(cfid, q, &cfids->entries, entry) {
cifs: add new field to track the last access time of cfid The handlecache code today tracks the time at which dir lease was acquired and the laundromat thread uses that to check for old entries to cleanup. However, if a directory is actively accessed, it should not be chosen to expire first. This change adds a new last_access_time field to cfid and uses that to decide expiry of the cfid. Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2025-07-24 22:23:53 -05:00
if (cfid->last_access_time &&
time_after(jiffies, cfid->last_access_time + HZ * dir_cache_timeout)) {
smb: client: prevent new fids from being removed by laundromat Check if @cfid->time is set in laundromat so we guarantee that only fully cached fids will be selected for removal. While we're at it, add missing locks to protect access of @cfid fields in order to avoid races with open_cached_dir() and cfids_laundromat_worker(), respectively. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-09 17:37:40 -03:00
cfid->on_list = false;
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
list_move(&cfid->entry, &entry);
cfids->num_entries--;
smb: prevent use-after-free due to open_cached_dir error paths If open_cached_dir() encounters an error parsing the lease from the server, the error handling may race with receiving a lease break, resulting in open_cached_dir() freeing the cfid while the queued work is pending. Update open_cached_dir() to drop refs rather than directly freeing the cfid. Have cached_dir_lease_break(), cfids_laundromat_worker(), and invalidate_all_cached_dirs() clear has_lease immediately while still holding cfids->cfid_list_lock, and then use this to also simplify the reference counting in cfids_laundromat_worker() and invalidate_all_cached_dirs(). Fixes this KASAN splat (which manually injects an error and lease break in open_cached_dir()): ================================================================== BUG: KASAN: slab-use-after-free in smb2_cached_lease_break+0x27/0xb0 Read of size 8 at addr ffff88811cc24c10 by task kworker/3:1/65 CPU: 3 UID: 0 PID: 65 Comm: kworker/3:1 Not tainted 6.12.0-rc6-g255cf264e6e5-dirty #87 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 11/12/2020 Workqueue: cifsiod smb2_cached_lease_break Call Trace: <TASK> dump_stack_lvl+0x77/0xb0 print_report+0xce/0x660 kasan_report+0xd3/0x110 smb2_cached_lease_break+0x27/0xb0 process_one_work+0x50a/0xc50 worker_thread+0x2ba/0x530 kthread+0x17c/0x1c0 ret_from_fork+0x34/0x60 ret_from_fork_asm+0x1a/0x30 </TASK> Allocated by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 __kasan_kmalloc+0xaa/0xb0 open_cached_dir+0xa7d/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 2464: kasan_save_stack+0x33/0x60 kasan_save_track+0x14/0x30 kasan_save_free_info+0x3b/0x60 __kasan_slab_free+0x51/0x70 kfree+0x174/0x520 open_cached_dir+0x97f/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e Last potentially related work creation: kasan_save_stack+0x33/0x60 __kasan_record_aux_stack+0xad/0xc0 insert_work+0x32/0x100 __queue_work+0x5c9/0x870 queue_work_on+0x82/0x90 open_cached_dir+0x1369/0x1fb0 smb2_query_path_info+0x43c/0x6e0 cifs_get_fattr+0x346/0xf10 cifs_get_inode_info+0x157/0x210 cifs_revalidate_dentry_attr+0x2d1/0x460 cifs_getattr+0x173/0x470 vfs_statx_path+0x10f/0x160 vfs_statx+0xe9/0x150 vfs_fstatat+0x5e/0xc0 __do_sys_newfstatat+0x91/0xf0 do_syscall_64+0x95/0x1a0 entry_SYSCALL_64_after_hwframe+0x76/0x7e The buggy address belongs to the object at ffff88811cc24c00 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 16 bytes inside of freed 1024-byte region [ffff88811cc24c00, ffff88811cc25000) Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:27 -08:00
if (cfid->has_lease) {
/*
* Our lease has not yet been cancelled from the
* server. Steal that reference.
*/
cfid->has_lease = false;
} else
kref_get(&cfid->refcount);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
}
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
}
spin_unlock(&cfids->cfid_list_lock);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
list_for_each_entry_safe(cfid, q, &entry, entry) {
list_del(&cfid->entry);
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
spin_lock(&cfid->fid_lock);
dentry = cfid->dentry;
cfid->dentry = NULL;
spin_unlock(&cfid->fid_lock);
dput(dentry);
if (cfid->is_open) {
spin_lock(&cifs_tcp_ses_lock);
++cfid->tcon->tc_count;
trace_smb3_tcon_ref(cfid->tcon->debug_id, cfid->tcon->tc_count,
netfs_trace_tcon_ref_get_cached_laundromat);
spin_unlock(&cifs_tcp_ses_lock);
queue_work(serverclose_wq, &cfid->close_work);
} else
/*
* Drop the ref-count from above, either the lease-ref (if there
* was one) or the extra one acquired.
*/
kref_put(&cfid->refcount, smb2_close_cached_fid);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
queue_delayed_work(cfid_put_wq, &cfids->laundromat_work,
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
dir_cache_timeout * HZ);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
}
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
struct cached_fids *init_cached_dirs(void)
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
{
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
struct cached_fids *cfids;
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
cfids = kzalloc(sizeof(*cfids), GFP_KERNEL);
if (!cfids)
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
return NULL;
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock_init(&cfids->cfid_list_lock);
INIT_LIST_HEAD(&cfids->entries);
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
INIT_LIST_HEAD(&cfids->dying);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
INIT_WORK(&cfids->invalidation_work, cfids_invalidation_worker);
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
INIT_DELAYED_WORK(&cfids->laundromat_work, cfids_laundromat_worker);
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
queue_delayed_work(cfid_put_wq, &cfids->laundromat_work,
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
dir_cache_timeout * HZ);
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
return cfids;
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
/*
* Called from tconInfoFree when we are tearing down the tcon.
* There are no active users or open files/directories at this point.
*/
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
void free_cached_dirs(struct cached_fids *cfids)
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
{
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
struct cached_fid *cfid, *q;
cifs: use LIST_HEAD() and list_move() to simplify code list_head can be initialized automatically with LIST_HEAD() instead of calling INIT_LIST_HEAD(). Using list_move() instead of list_del() and list_add(). Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-17 16:55:08 +08:00
LIST_HEAD(entry);
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
smb3: do not start laundromat thread when dir leases disabled When no directory lease support, or for IPC shares where directories can not be opened, do not start an unneeded laundromat thread for that mount (it wastes resources). Fixes: d14de8067e3f ("cifs: Add a laundromat thread for cached directories") Reviewed-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Acked-by: Tom Talpey <tom@talpey.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-09-19 11:35:53 -05:00
if (cfids == NULL)
return;
smb: client: make laundromat a delayed worker By having laundromat kthread processing cached directories on every second turned out to be overkill, especially when having multiple SMB mounts. Relax it by using a delayed worker instead that gets scheduled on every @dir_cache_timeout (default=30) seconds per tcon. This also fixes the 1s delay when tearing down tcon. Signed-off-by: Paulo Alcantara (SUSE) <pc@manguebit.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-10-05 16:04:25 -03:00
cancel_delayed_work_sync(&cfids->laundromat_work);
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
cancel_work_sync(&cfids->invalidation_work);
cifs: Add a laundromat thread for cached directories and drop cached directories after 30 seconds Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2023-07-06 12:32:24 +10:00
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_lock(&cfids->cfid_list_lock);
list_for_each_entry_safe(cfid, q, &cfids->entries, entry) {
cfid->on_list = false;
cfid->is_open = false;
cifs: use LIST_HEAD() and list_move() to simplify code list_head can be initialized automatically with LIST_HEAD() instead of calling INIT_LIST_HEAD(). Using list_move() instead of list_del() and list_add(). Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-17 16:55:08 +08:00
list_move(&cfid->entry, &entry);
cifs: store a pointer to a fid in the cfid structure instead of the struct also create a constructor that takes a path name and stores it in the fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:44 +10:00
}
smb: During unmount, ensure all cached dir instances drop their dentry The unmount process (cifs_kill_sb() calling close_all_cached_dirs()) can race with various cached directory operations, which ultimately results in dentries not being dropped and these kernel BUGs: BUG: Dentry ffff88814f37e358{i=1000000000080,n=/} still in use (2) [unmount of cifs cifs] VFS: Busy inodes after unmount of cifs (cifs) ------------[ cut here ]------------ kernel BUG at fs/super.c:661! This happens when a cfid is in the process of being cleaned up when, and has been removed from the cfids->entries list, including: - Receiving a lease break from the server - Server reconnection triggers invalidate_all_cached_dirs(), which removes all the cfids from the list - The laundromat thread decides to expire an old cfid. To solve these problems, dropping the dentry is done in queued work done in a newly-added cfid_put_wq workqueue, and close_all_cached_dirs() flushes that workqueue after it drops all the dentries of which it's aware. This is a global workqueue (rather than scoped to a mount), but the queued work is minimal. The final cleanup work for cleaning up a cfid is performed via work queued in the serverclose_wq workqueue; this is done separate from dropping the dentries so that close_all_cached_dirs() doesn't block on any server operations. Both of these queued works expect to invoked with a cfid reference and a tcon reference to avoid those objects from being freed while the work is ongoing. While we're here, add proper locking to close_all_cached_dirs(), and locking around the freeing of cfid->dentry. Fixes: ebe98f1447bb ("cifs: enable caching of directories for which a lease is held") Cc: stable@vger.kernel.org Signed-off-by: Paul Aurich <paul@darkrain42.org> Signed-off-by: Steve French <stfrench@microsoft.com>
2024-11-18 13:50:28 -08:00
list_for_each_entry_safe(cfid, q, &cfids->dying, entry) {
cfid->on_list = false;
cfid->is_open = false;
list_move(&cfid->entry, &entry);
}
cifs: enable caching of directories for which a lease is held This expands the directory caching to now cache an open handle for all directories (up to a maximum) and not just the root directory. In this patch, locking and refcounting is intended to work as so: The main function to get a reference to a cached handle is find_or_create_cached_dir() called from open_cached_dir() These functions are protected under the cfid_list_lock spin-lock to make sure we do not race creating new references for cached dirs with deletion of expired ones. An successful open_cached_dir() will take out 2 references to the cfid if this was the very first and successful call to open the directory and it acquired a lease from the server. One reference is for the lease and the other is for the cfid that we return. The is lease reference is tracked by cfid->has_lease. If the directory already has a handle with an active lease, then we just take out one new reference for the cfid and return it. It can happen that we have a thread that tries to open a cached directory where we have a cfid already but we do not, yet, have a working lease. In this case we will just return NULL, and this the caller will fall back to the case when no handle was available. In this model the total number of references we have on a cfid is 1 for while the handle is open and we have a lease, and one additional reference for each open instance of a cfid. Once we get a lease break (cached_dir_lease_break()) we remove the cfid from the list under the spinlock. This prevents any new threads to use it, and we also call smb2_cached_lease_break() via the work_queue in order to drop the reference we got for the lease (we drop it outside of the spin-lock.) Anytime a thread calls close_cached_dir() we also drop a reference to the cfid. When the last reference to the cfid is released smb2_close_cached_fid() will be invoked which will drop the reference ot the dentry we held for this cfid and it will also, if we the handle is open/has a lease also call SMB2_close() to close the handle on the server. Two events require special handling: invalidate_all_cached_dirs() this function is called from SMB2_tdis() and cifs_mark_open_files_invalid(). In both cases the tcon is either gone already or will be shortly so we do not need to actually close the handles. They will be dropped server side as part of the tcon dropping. But we have to be careful about a potential race with a concurrent lease break so we need to take out additional refences to avoid the cfid from being freed while we are still referencing it. free_cached_dirs() which is called from tconInfoFree(). This is called quite late in the umount process so there should no longer be any open handles or files and we can just free all the remaining data. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-10-06 00:14:31 -05:00
spin_unlock(&cfids->cfid_list_lock);
list_for_each_entry_safe(cfid, q, &entry, entry) {
list_del(&cfid->entry);
free_cached_dir(cfid);
}
cifs: Make tcon contain a wrapper structure cached_fids instead of cached_fid This wrapper structure will later be expanded to contain a list of fids that are cached and not just the root fid. Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-31 12:49:42 +10:00
kfree(cfids);
cifs: Add constructor/destructors for tcon->cfid and move the structure definitions into cached_dir.h Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com> Signed-off-by: Steve French <stfrench@microsoft.com>
2022-08-11 19:04:29 -05:00
}
Reference in a new issue Copy permalink