Instead of a single if statement with multiple conditions, split it into
several if statements testing only one condition at a time and return true
or false immediately after. This makes it more immediate to reason.
The module's text size also slightly decreases, at least with gcc 14.2.0
on x86_64.
Before:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1897138 161583 16136 2074857 1fa8e9 fs/btrfs/btrfs.ko
After:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1896976 161583 16136 2074695 1fa847 fs/btrfs/btrfs.ko
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is an exported function and therefore it should have a 'btrfs_'
prefix, to make it clear it's btrfs specific, avoid future name collisions
with code outside btrfs, and make its naming consistent with most other
btrfs exported functions.
So add a 'btrfs_' prefix to it and make it return bool instead of int,
since all we need is to return true or false.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The inode_lock field of struct btrfs_root was removed in commit
e2844cce75c9e61("btrfs: remove inode_lock from struct btrfs_root and use
xarray locks") but the related comment haven't been updated.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Sun YangKai <sunk67188@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Another batch of pointer parameter constifications. This is for clarity
and minor addition to type safety. There are no observable effects in the
assembly code and .ko measured on release config.
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 925baeddc5 ("Btrfs: Start btree concurrency work.") added the
comment for the field keep_locks. This got moved later but without the
comment, so move it to the right place and fix the comment style.
Signed-off-by: Sun YangKai <sunk67188@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It's pointless to have a comment above the prototype declarations of
btrfs_ctree_init() and btrfs_ctree_exit() mentioning that they are
declared in ctree.c. This is from the old days when ctree.h was used
to place anything that didn't fit in any other file. So remove it.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have 3 functions that have their prototypes declared in ctree.h but
they are defined at extent-tree.c and they are unrelated to the btree
data structure. Move the prototypes out of ctree.h and into extent-tree.h.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently btrfs_alloc_write_mask() is defined in ctree.h but it's not
related at all to the btree data structure, so move it into fs.h.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently BTRFS_BYTES_TO_BLKS() is defined in ctree.h but it's not related
at all to the btree data structure, so move it into fs.h.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The folio ordered helper macros are defined at ctree.h but this is not
the best place since ctree.{h,c} is all about the btree data structure
implementation and not a generic module. So move these macros into the
fs.h header.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The ctree module is about the implementation of the btree data structure
and not a place holder for generic filesystem things like the csum
algorithm details. Move the functions related to the csum algorithm
details away from ctree.c and into fs.c, which is a far better place for
them. Also fix missing punctuation in comments and change one multiline
comment to a single line comment since everything fits in under 80
characters.
For some reason this also slightly reduces the module's size.
Before this change:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1782126 161045 16920 1960091 1de89b fs/btrfs/btrfs.ko
After this change:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1782094 161045 16920 1960059 1de87b fs/btrfs/btrfs.ko
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have been using the following check
if (generation <= root->root_key.offset)
to make decisions about whether or not to visit a node during snapshot
delete. This is because for normal subvolumes this is set to 0, and for
snapshots it's set to the creation generation. The idea being that if
the generation of the node is less than or equal to our creation
generation then we don't need to visit that node, because it doesn't
belong to us, we can simply drop our reference and move on.
However reloc roots don't have their generation stored in
root->root_key.offset, instead that is the objectid of their
corresponding fs root. This means we can incorrectly not walk into
nodes that need to be dropped when deleting a reloc root.
There are a variety of consequences to making the wrong choice in two
distinct areas.
visit_node_for_delete()
1. False positive. We think we are newer than the block when we really
aren't. We don't visit the node and drop our reference to the node
and carry on. This would result in leaked space.
2. False negative. We do decide to walk down into a block that we
should have just dropped our reference to. However this means that
the child node will have refs > 1, so we will switch to
UPDATE_BACKREF, and then the subsequent walk_down_proc() will notice
that btrfs_header_owner(node) != root->root_key.objectid and it'll
break out of the loop, and then walk_up_proc() will drop our reference,
so this appears to be ok.
do_walk_down()
1. False positive. We are in UPDATE_BACKREF and incorrectly decide that
we are done and don't need to update the backref for our lower nodes.
This is another case that simply won't happen with relocation, as we
only have to do UPDATE_BACKREF if the node below us was shared and
didn't have FULL_BACKREF set, and since we don't own that node
because we're a reloc root we actually won't end up in this case.
2. False negative. Again this is tricky because as described above, we
simply wouldn't be here from relocation, because we don't own any of
the nodes because we never set btrfs_header_owner() to the reloc root
objectid, and we always use FULL_BACKREF, we never actually need to
set FULL_BACKREF on any children.
Having spent a lot of time stressing relocation/snapshot delete recently
I've not seen this pop in practice. But this is objectively incorrect,
so fix this to get the correct starting generation based on the root
we're dropping to keep me from thinking there's a problem here.
CC: stable@vger.kernel.org
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We are close to removing the private_2 flag, so switch btrfs to using
owner_2 for its ordered flag. This is mostly used by buffer head
filesystems, so btrfs can use it because it doesn't use buffer heads.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Link: https://lore.kernel.org/r/20241002040111.1023018-5-willy@infradead.org
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
When doing concurrent lseek(2) system calls against the same file
descriptor, using multiple threads belonging to the same process, we have
a short time window where a race happens and can result in a memory leak.
The race happens like this:
1) A program opens a file descriptor for a file and then spawns two
threads (with the pthreads library for example), lets call them
task A and task B;
2) Task A calls lseek with SEEK_DATA or SEEK_HOLE and ends up at
file.c:find_desired_extent() while holding a read lock on the inode;
3) At the start of find_desired_extent(), it extracts the file's
private_data pointer into a local variable named 'private', which has
a value of NULL;
4) Task B also calls lseek with SEEK_DATA or SEEK_HOLE, locks the inode
in shared mode and enters file.c:find_desired_extent(), where it also
extracts file->private_data into its local variable 'private', which
has a NULL value;
5) Because it saw a NULL file private, task A allocates a private
structure and assigns to the file structure;
6) Task B also saw a NULL file private so it also allocates its own file
private and then assigns it to the same file structure, since both
tasks are using the same file descriptor.
At this point we leak the private structure allocated by task A.
Besides the memory leak, there's also the detail that both tasks end up
using the same cached state record in the private structure (struct
btrfs_file_private::llseek_cached_state), which can result in a
use-after-free problem since one task can free it while the other is
still using it (only one task took a reference count on it). Also, sharing
the cached state is not a good idea since it could result in incorrect
results in the future - right now it should not be a problem because it
end ups being used only in extent-io-tree.c:count_range_bits() where we do
range validation before using the cached state.
Fix this by protecting the private assignment and check of a file while
holding the inode's spinlock and keep track of the task that allocated
the private, so that it's used only by that task in order to prevent
user-after-free issues with the cached state record as well as potentially
using it incorrectly in the future.
Fixes: 3c32c7212f ("btrfs: use cached state when looking for delalloc ranges with lseek")
CC: stable@vger.kernel.org # 6.6+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add a DEFINE_FREE for struct btrfs_path. This defines a function that
can be called using the __free attribute. Define a macro
BTRFS_PATH_AUTO_FREE to make the declaration of an auto freeing path
very clear.
The intended use is to define the auto free of path in cases where the
path is allocated somewhere at the beginning and freed either on all
error paths or at the end of the function.
int func() {
BTRFS_PATH_AUTO_FREE(path);
if (...)
return -ERROR;
path = alloc_path();
...
if (...)
return -ERROR;
...
return 0;
}
Signed-off-by: Leo Martins <loemra.dev@gmail.com>
[ update changelog ]
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Continue adding const to parameters. This is for clarity and minor
addition to safety. There are some minor effects, in the assembly code
and .ko measured on release config.
Signed-off-by: David Sterba <dsterba@suse.com>
If we have 2 threads that are using the same file descriptor and one of
them is doing direct IO writes while the other is doing fsync, we have a
race where we can end up either:
1) Attempt a fsync without holding the inode's lock, triggering an
assertion failures when assertions are enabled;
2) Do an invalid memory access from the fsync task because the file private
points to memory allocated on stack by the direct IO task and it may be
used by the fsync task after the stack was destroyed.
The race happens like this:
1) A user space program opens a file descriptor with O_DIRECT;
2) The program spawns 2 threads using libpthread for example;
3) One of the threads uses the file descriptor to do direct IO writes,
while the other calls fsync using the same file descriptor.
4) Call task A the thread doing direct IO writes and task B the thread
doing fsyncs;
5) Task A does a direct IO write, and at btrfs_direct_write() sets the
file's private to an on stack allocated private with the member
'fsync_skip_inode_lock' set to true;
6) Task B enters btrfs_sync_file() and sees that there's a private
structure associated to the file which has 'fsync_skip_inode_lock' set
to true, so it skips locking the inode's VFS lock;
7) Task A completes the direct IO write, and resets the file's private to
NULL since it had no prior private and our private was stack allocated.
Then it unlocks the inode's VFS lock;
8) Task B enters btrfs_get_ordered_extents_for_logging(), then the
assertion that checks the inode's VFS lock is held fails, since task B
never locked it and task A has already unlocked it.
The stack trace produced is the following:
assertion failed: inode_is_locked(&inode->vfs_inode), in fs/btrfs/ordered-data.c:983
------------[ cut here ]------------
kernel BUG at fs/btrfs/ordered-data.c:983!
Oops: invalid opcode: 0000 [#1] PREEMPT SMP PTI
CPU: 9 PID: 5072 Comm: worker Tainted: G U OE 6.10.5-1-default #1 openSUSE Tumbleweed 69f48d427608e1c09e60ea24c6c55e2ca1b049e8
Hardware name: Acer Predator PH315-52/Covini_CFS, BIOS V1.12 07/28/2020
RIP: 0010:btrfs_get_ordered_extents_for_logging.cold+0x1f/0x42 [btrfs]
Code: 50 d6 86 c0 e8 (...)
RSP: 0018:ffff9e4a03dcfc78 EFLAGS: 00010246
RAX: 0000000000000054 RBX: ffff9078a9868e98 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff907dce4a7800 RDI: ffff907dce4a7800
RBP: ffff907805518800 R08: 0000000000000000 R09: ffff9e4a03dcfb38
R10: ffff9e4a03dcfb30 R11: 0000000000000003 R12: ffff907684ae7800
R13: 0000000000000001 R14: ffff90774646b600 R15: 0000000000000000
FS: 00007f04b96006c0(0000) GS:ffff907dce480000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f32acbfc000 CR3: 00000001fd4fa005 CR4: 00000000003726f0
Call Trace:
<TASK>
? __die_body.cold+0x14/0x24
? die+0x2e/0x50
? do_trap+0xca/0x110
? do_error_trap+0x6a/0x90
? btrfs_get_ordered_extents_for_logging.cold+0x1f/0x42 [btrfs bb26272d49b4cdc847cf3f7faadd459b62caee9a]
? exc_invalid_op+0x50/0x70
? btrfs_get_ordered_extents_for_logging.cold+0x1f/0x42 [btrfs bb26272d49b4cdc847cf3f7faadd459b62caee9a]
? asm_exc_invalid_op+0x1a/0x20
? btrfs_get_ordered_extents_for_logging.cold+0x1f/0x42 [btrfs bb26272d49b4cdc847cf3f7faadd459b62caee9a]
? btrfs_get_ordered_extents_for_logging.cold+0x1f/0x42 [btrfs bb26272d49b4cdc847cf3f7faadd459b62caee9a]
btrfs_sync_file+0x21a/0x4d0 [btrfs bb26272d49b4cdc847cf3f7faadd459b62caee9a]
? __seccomp_filter+0x31d/0x4f0
__x64_sys_fdatasync+0x4f/0x90
do_syscall_64+0x82/0x160
? do_futex+0xcb/0x190
? __x64_sys_futex+0x10e/0x1d0
? switch_fpu_return+0x4f/0xd0
? syscall_exit_to_user_mode+0x72/0x220
? do_syscall_64+0x8e/0x160
? syscall_exit_to_user_mode+0x72/0x220
? do_syscall_64+0x8e/0x160
? syscall_exit_to_user_mode+0x72/0x220
? do_syscall_64+0x8e/0x160
? syscall_exit_to_user_mode+0x72/0x220
? do_syscall_64+0x8e/0x160
entry_SYSCALL_64_after_hwframe+0x76/0x7e
Another problem here is if task B grabs the private pointer and then uses
it after task A has finished, since the private was allocated in the stack
of task A, it results in some invalid memory access with a hard to predict
result.
This issue, triggering the assertion, was observed with QEMU workloads by
two users in the Link tags below.
Fix this by not relying on a file's private to pass information to fsync
that it should skip locking the inode and instead pass this information
through a special value stored in current->journal_info. This is safe
because in the relevant section of the direct IO write path we are not
holding a transaction handle, so current->journal_info is NULL.
The following C program triggers the issue:
$ cat repro.c
/* Get the O_DIRECT definition. */
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <stdint.h>
#include <fcntl.h>
#include <errno.h>
#include <string.h>
#include <pthread.h>
static int fd;
static ssize_t do_write(int fd, const void *buf, size_t count, off_t offset)
{
while (count > 0) {
ssize_t ret;
ret = pwrite(fd, buf, count, offset);
if (ret < 0) {
if (errno == EINTR)
continue;
return ret;
}
count -= ret;
buf += ret;
}
return 0;
}
static void *fsync_loop(void *arg)
{
while (1) {
int ret;
ret = fsync(fd);
if (ret != 0) {
perror("Fsync failed");
exit(6);
}
}
}
int main(int argc, char *argv[])
{
long pagesize;
void *write_buf;
pthread_t fsyncer;
int ret;
if (argc != 2) {
fprintf(stderr, "Use: %s <file path>\n", argv[0]);
return 1;
}
fd = open(argv[1], O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT, 0666);
if (fd == -1) {
perror("Failed to open/create file");
return 1;
}
pagesize = sysconf(_SC_PAGE_SIZE);
if (pagesize == -1) {
perror("Failed to get page size");
return 2;
}
ret = posix_memalign(&write_buf, pagesize, pagesize);
if (ret) {
perror("Failed to allocate buffer");
return 3;
}
ret = pthread_create(&fsyncer, NULL, fsync_loop, NULL);
if (ret != 0) {
fprintf(stderr, "Failed to create writer thread: %d\n", ret);
return 4;
}
while (1) {
ret = do_write(fd, write_buf, pagesize, 0);
if (ret != 0) {
perror("Write failed");
exit(5);
}
}
return 0;
}
$ mkfs.btrfs -f /dev/sdi
$ mount /dev/sdi /mnt/sdi
$ timeout 10 ./repro /mnt/sdi/foo
Usually the race is triggered within less than 1 second. A test case for
fstests will follow soon.
Reported-by: Paulo Dias <paulo.miguel.dias@gmail.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=219187
Reported-by: Andreas Jahn <jahn-andi@web.de>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=219199
Reported-by: syzbot+4704b3cc972bd76024f1@syzkaller.appspotmail.com
Link: https://lore.kernel.org/linux-btrfs/00000000000044ff540620d7dee2@google.com/
Fixes: 939b656bc8 ("btrfs: fix corruption after buffer fault in during direct IO append write")
CC: stable@vger.kernel.org # 5.15+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
During an append (O_APPEND write flag) direct IO write if the input buffer
was not previously faulted in, we can corrupt the file in a way that the
final size is unexpected and it includes an unexpected hole.
The problem happens like this:
1) We have an empty file, with size 0, for example;
2) We do an O_APPEND direct IO with a length of 4096 bytes and the input
buffer is not currently faulted in;
3) We enter btrfs_direct_write(), lock the inode and call
generic_write_checks(), which calls generic_write_checks_count(), and
that function sets the iocb position to 0 with the following code:
if (iocb->ki_flags & IOCB_APPEND)
iocb->ki_pos = i_size_read(inode);
4) We call btrfs_dio_write() and enter into iomap, which will end up
calling btrfs_dio_iomap_begin() and that calls
btrfs_get_blocks_direct_write(), where we update the i_size of the
inode to 4096 bytes;
5) After btrfs_dio_iomap_begin() returns, iomap will attempt to access
the page of the write input buffer (at iomap_dio_bio_iter(), with a
call to bio_iov_iter_get_pages()) and fail with -EFAULT, which gets
returned to btrfs at btrfs_direct_write() via btrfs_dio_write();
6) At btrfs_direct_write() we get the -EFAULT error, unlock the inode,
fault in the write buffer and then goto to the label 'relock';
7) We lock again the inode, do all the necessary checks again and call
again generic_write_checks(), which calls generic_write_checks_count()
again, and there we set the iocb's position to 4K, which is the current
i_size of the inode, with the following code pointed above:
if (iocb->ki_flags & IOCB_APPEND)
iocb->ki_pos = i_size_read(inode);
8) Then we go again to btrfs_dio_write() and enter iomap and the write
succeeds, but it wrote to the file range [4K, 8K), leaving a hole in
the [0, 4K) range and an i_size of 8K, which goes against the
expectations of having the data written to the range [0, 4K) and get an
i_size of 4K.
Fix this by not unlocking the inode before faulting in the input buffer,
in case we get -EFAULT or an incomplete write, and not jumping to the
'relock' label after faulting in the buffer - instead jump to a location
immediately before calling iomap, skipping all the write checks and
relocking. This solves this problem and it's fine even in case the input
buffer is memory mapped to the same file range, since only holding the
range locked in the inode's io tree can cause a deadlock, it's safe to
keep the inode lock (VFS lock), as was fixed and described in commit
51bd9563b6 ("btrfs: fix deadlock due to page faults during direct IO
reads and writes").
A sample reproducer provided by a reporter is the following:
$ cat test.c
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include <fcntl.h>
#include <stdio.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <unistd.h>
int main(int argc, char *argv[])
{
if (argc < 2) {
fprintf(stderr, "Usage: %s <test file>\n", argv[0]);
return 1;
}
int fd = open(argv[1], O_WRONLY | O_CREAT | O_TRUNC | O_DIRECT |
O_APPEND, 0644);
if (fd < 0) {
perror("creating test file");
return 1;
}
char *buf = mmap(NULL, 4096, PROT_READ,
MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
ssize_t ret = write(fd, buf, 4096);
if (ret < 0) {
perror("pwritev2");
return 1;
}
struct stat stbuf;
ret = fstat(fd, &stbuf);
if (ret < 0) {
perror("stat");
return 1;
}
printf("size: %llu\n", (unsigned long long)stbuf.st_size);
return stbuf.st_size == 4096 ? 0 : 1;
}
A test case for fstests will be sent soon.
Reported-by: Hanna Czenczek <hreitz@redhat.com>
Link: https://lore.kernel.org/linux-btrfs/0b841d46-12fe-4e64-9abb-871d8d0de271@redhat.com/
Fixes: 8184620ae2 ("btrfs: fix lost file sync on direct IO write with nowait and dsync iocb")
CC: stable@vger.kernel.org # 6.1+
Tested-by: Hanna Czenczek <hreitz@redhat.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
KCSAN complains about a data race when accessing the last_trans field of a
root:
[ 199.553628] BUG: KCSAN: data-race in btrfs_record_root_in_trans [btrfs] / record_root_in_trans [btrfs]
[ 199.555186] read to 0x000000008801e308 of 8 bytes by task 2812 on cpu 1:
[ 199.555210] btrfs_record_root_in_trans+0x9a/0x128 [btrfs]
[ 199.555999] start_transaction+0x154/0xcd8 [btrfs]
[ 199.556780] btrfs_join_transaction+0x44/0x60 [btrfs]
[ 199.557559] btrfs_dirty_inode+0x9c/0x140 [btrfs]
[ 199.558339] btrfs_update_time+0x8c/0xb0 [btrfs]
[ 199.559123] touch_atime+0x16c/0x1e0
[ 199.559151] pipe_read+0x6a8/0x7d0
[ 199.559179] vfs_read+0x466/0x498
[ 199.559204] ksys_read+0x108/0x150
[ 199.559230] __s390x_sys_read+0x68/0x88
[ 199.559257] do_syscall+0x1c6/0x210
[ 199.559286] __do_syscall+0xc8/0xf0
[ 199.559318] system_call+0x70/0x98
[ 199.559431] write to 0x000000008801e308 of 8 bytes by task 2808 on cpu 0:
[ 199.559464] record_root_in_trans+0x196/0x228 [btrfs]
[ 199.560236] btrfs_record_root_in_trans+0xfe/0x128 [btrfs]
[ 199.561097] start_transaction+0x154/0xcd8 [btrfs]
[ 199.561927] btrfs_join_transaction+0x44/0x60 [btrfs]
[ 199.562700] btrfs_dirty_inode+0x9c/0x140 [btrfs]
[ 199.563493] btrfs_update_time+0x8c/0xb0 [btrfs]
[ 199.564277] file_update_time+0xb8/0xf0
[ 199.564301] pipe_write+0x8ac/0xab8
[ 199.564326] vfs_write+0x33c/0x588
[ 199.564349] ksys_write+0x108/0x150
[ 199.564372] __s390x_sys_write+0x68/0x88
[ 199.564397] do_syscall+0x1c6/0x210
[ 199.564424] __do_syscall+0xc8/0xf0
[ 199.564452] system_call+0x70/0x98
This is because we update and read last_trans concurrently without any
type of synchronization. This should be generally harmless and in the
worst case it can make us do extra locking (btrfs_record_root_in_trans())
trigger some warnings at ctree.c or do extra work during relocation - this
would probably only happen in case of load or store tearing.
So fix this by always reading and updating the field using READ_ONCE()
and WRITE_ONCE(), this silences KCSAN and prevents load and store tearing.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently we use the spinlock inode_lock from struct btrfs_root to
serialize access to two different data structures:
1) The delayed inodes xarray (struct btrfs_root::delayed_nodes);
2) The inodes xarray (struct btrfs_root::inodes).
Instead of using our own lock, we can use the spinlock that is part of the
xarray implementation, by using the xa_lock() and xa_unlock() APIs and
using the xarray APIs with the double underscore prefix that don't take
the xarray locks and assume the caller is using xa_lock() and xa_unlock().
So remove the spinlock inode_lock from struct btrfs_root and use the
corresponding xarray locks. This brings 2 benefits:
1) We reduce the size of struct btrfs_root, from 1336 bytes down to
1328 bytes on a 64 bits release kernel config;
2) We reduce lock contention by not using anymore the same lock for
changing two different and unrelated xarrays.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently we use a red black tree (rb-tree) to track the currently open
inodes of a root (in struct btrfs_root::inode_tree). This however is not
very efficient when the number of inodes is large since rb-trees are
binary trees. For example for 100K open inodes, the tree has a depth of
17. Besides that, inserting into the tree requires navigating through it
and pulling useless cache lines in the process since the red black tree
nodes are embedded within the btrfs inode - on the other hand, by being
embedded, it requires no extra memory allocations.
We can improve this by using an xarray instead, which is efficient when
indices are densely clustered (such as inode numbers), is more cache
friendly and behaves like a resizable array, with a much better search
and insertion complexity than a red black tree. This only has one small
disadvantage which is that insertion will sometimes require allocating
memory for the xarray - which may fail (not that often since it uses a
kmem_cache) - but on the other hand we can reduce the btrfs inode
structure size by 24 bytes (from 1080 down to 1056 bytes) after removing
the embedded red black tree node, which after the next patches will allow
to reduce the size of the structure to 1024 bytes, meaning we will be able
to store 4 inodes per 4K page instead of 3 inodes.
This change does a straightforward change to use an xarray, and results
in a transaction abort if we can't allocate memory for the xarray when
creating an inode - but the next patch changes things so that we don't
need to abort.
Running the following fs_mark test showed some improvements:
$ cat test.sh
#!/bin/bash
DEV=/dev/nullb0
MNT=/mnt/nullb0
MOUNT_OPTIONS="-o ssd"
FILES=100000
THREADS=$(nproc --all)
echo "performance" | \
tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor
mkfs.btrfs -f $DEV
mount $MOUNT_OPTIONS $DEV $MNT
OPTS="-S 0 -L 5 -n $FILES -s 0 -t $THREADS -k"
for ((i = 1; i <= $THREADS; i++)); do
OPTS="$OPTS -d $MNT/d$i"
done
fs_mark $OPTS
umount $MNT
Before this patch:
FSUse% Count Size Files/sec App Overhead
10 1200000 0 92081.6 12505547
16 2400000 0 138222.6 13067072
23 3600000 0 148833.1 13290336
43 4800000 0 97864.7 13931248
53 6000000 0 85597.3 14384313
After this patch:
FSUse% Count Size Files/sec App Overhead
10 1200000 0 93225.1 12571078
16 2400000 0 146720.3 12805007
23 3600000 0 160626.4 13073835
46 4800000 0 116286.2 13802927
53 6000000 0 90087.9 14754892
The test was run with a release kernel config (Debian's default config).
Also capturing the insertion times into the rb tree and into the xarray,
that is measuring the duration of the old function inode_tree_add() and
the duration of the new btrfs_add_inode_to_root() function, gave the
following results (in nanoseconds):
Before this patch, inode_tree_add() execution times:
Count: 5000000
Range: 0.000 - 5536887.000; Mean: 775.674; Median: 729.000; Stddev: 4820.961
Percentiles: 90th: 1015.000; 95th: 1139.000; 99th: 1397.000
0.000 - 7.816: 40 |
7.816 - 37.858: 209 |
37.858 - 170.278: 6059 |
170.278 - 753.961: 2754890 #####################################################
753.961 - 3326.728: 2232312 ###########################################
3326.728 - 14667.018: 4366 |
14667.018 - 64652.943: 852 |
64652.943 - 284981.761: 550 |
284981.761 - 1256150.914: 221 |
1256150.914 - 5536887.000: 7 |
After this patch, btrfs_add_inode_to_root() execution times:
Count: 5000000
Range: 0.000 - 2900652.000; Mean: 272.148; Median: 241.000; Stddev: 2873.369
Percentiles: 90th: 342.000; 95th: 432.000; 99th: 572.000
0.000 - 7.264: 104 |
7.264 - 33.145: 352 |
33.145 - 140.081: 109606 #
140.081 - 581.930: 4840090 #####################################################
581.930 - 2407.590: 43532 |
2407.590 - 9950.979: 2245 |
9950.979 - 41119.278: 514 |
41119.278 - 169902.616: 155 |
169902.616 - 702018.539: 47 |
702018.539 - 2900652.000: 9 |
Average, percentiles, standard deviation, etc, are all much better.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Do a cleanup in the rest of the headers:
- add forward declarations for types referenced by pointers
- add includes when types need them
This fixes potential compilation problems if the headers are reordered
or the missing includes are not provided indirectly.
Signed-off-by: David Sterba <dsterba@suse.com>
This helper is used in transaction abort or cleanup context and the
callers cannot handle all errors, only do best effort.
btrfs_cleanup_one_transaction
btrfs_destroy_delayed_refs
btrfs_error_unpin_extent_range
btrfs_destroy_pinned_extent
btrfs_error_unpin_extent_range
Signed-off-by: David Sterba <dsterba@suse.com>
The radix-tree has been superseded by the xarray
(https://lwn.net/Articles/745073), this patch converts the
btrfs_root::delayed_nodes, the APIs are used in a simple way.
First idea is to do xa_insert() but this would require GFP_ATOMIC
allocation which we want to avoid if possible. The preload mechanism of
radix-tree can be emulated within the xarray API.
- xa_reserve() with GFP_NOFS outside of the lock, the reserved entry
is inserted atomically at most once
- xa_store() under a lock, in case something races in we can detect that
and xa_load() returns a valid pointer
All uses of xa_load() must check for a valid pointer in case they manage
to get between the xa_reserve() and xa_store(), this is handled in
btrfs_get_delayed_node().
Otherwise the functionality is equivalent, xarray implements the
radix-tree and there should be no performance difference.
The patch continues the efforts started in 253bf57555 ("btrfs: turn
delayed_nodes_tree into an XArray") and fixes the problems with locking
and GFP flags 088aea3b97 ("Revert "btrfs: turn delayed_nodes_tree
into an XArray"").
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Looks like the struct member was added in 2007 in 2.6.29 in commit
87ee04eb0f ("Btrfs: Add simple stripe size parameter") but hasn't been
used at all since. So let's remove it. This was found by tool
https://github.com/jirislaby/clang-struct, then build tested after
removing the struct member.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently btrfs_block_can_be_shared() returns an int that is used as a
boolean. Since it all it needs is to return true or false, and it can't
return errors for example, change the return type from int to bool to
make it a bit more readable and obvious.
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The logged_list[2] and log_extents_lock[2] members of struct btrfs_root
are no longer used, their last use was removed in commit 5636cf7d6d
("btrfs: remove the logged extents infrastructure"). So remove these
fields. This reduces the size of struct btrfs_root, on a release kernel,
from 1392 bytes down to 1352 bytes.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
-----BEGIN PGP SIGNATURE-----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=9TVx
-----END PGP SIGNATURE-----
Merge tag 'for-6.7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs updates from David Sterba:
"New features:
- raid-stripe-tree
New tree for logical file extent mapping where the physical mapping
may not match on multiple devices. This is now used in zoned mode
to implement RAID0/RAID1* profiles, but can be used in non-zoned
mode as well. The support for RAID56 is in development and will
eventually fix the problems with the current implementation. This
is a backward incompatible feature and has to be enabled at mkfs
time.
- simple quota accounting (squota)
A simplified mode of qgroup that accounts all space on the initial
extent owners (a subvolume), the snapshots are then cheap to create
and delete. The deletion of snapshots in fully accounting qgroups
is a known CPU/IO performance bottleneck.
The squota is not suitable for the general use case but works well
for containers where the original subvolume exists for the whole
time. This is a backward incompatible feature as it needs extending
some structures, but can be enabled on an existing filesystem.
- temporary filesystem fsid (temp_fsid)
The fsid identifies a filesystem and is hard coded in the
structures, which disallows mounting the same fsid found on
different devices.
For a single device filesystem this is not strictly necessary, a
new temporary fsid can be generated on mount e.g. after a device is
cloned. This will be used by Steam Deck for root partition A/B
testing, or can be used for VM root images.
Other user visible changes:
- filesystems with partially finished metadata_uuid conversion cannot
be mounted anymore and the uuid fixup has to be done by btrfs-progs
(btrfstune).
Performance improvements:
- reduce reservations for checksum deletions (with enabled free space
tree by factor of 4), on a sample workload on file with many
extents the deletion time decreased by 12%
- make extent state merges more efficient during insertions, reduce
rb-tree iterations (run time of critical functions reduced by 5%)
Core changes:
- the integrity check functionality has been removed, this was a
debugging feature and removal does not affect other integrity
checks like checksums or tree-checker
- space reservation changes:
- more efficient delayed ref reservations, this avoids building up
too much work or overusing or exhausting the global block
reserve in some situations
- move delayed refs reservation to the transaction start time,
this prevents some ENOSPC corner cases related to exhaustion of
global reserve
- improvements in reducing excessive reservations for block group
items
- adjust overcommit logic in near full situations, account for one
more chunk to eventually allocate metadata chunk, this is mostly
relevant for small filesystems (<10GiB)
- single device filesystems are scanned but not registered (except
seed devices), this allows temp_fsid to work
- qgroup iterations do not need GFP_ATOMIC allocations anymore
- cleanups, refactoring, reduced data structure size, function
parameter simplifications, error handling fixes"
* tag 'for-6.7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux: (156 commits)
btrfs: open code timespec64 in struct btrfs_inode
btrfs: remove redundant log root tree index assignment during log sync
btrfs: remove redundant initialization of variable dirty in btrfs_update_time()
btrfs: sysfs: show temp_fsid feature
btrfs: disable the device add feature for temp-fsid
btrfs: disable the seed feature for temp-fsid
btrfs: update comment for temp-fsid, fsid, and metadata_uuid
btrfs: remove pointless empty log context list check when syncing log
btrfs: update comment for struct btrfs_inode::lock
btrfs: remove pointless barrier from btrfs_sync_file()
btrfs: add and use helpers for reading and writing last_trans_committed
btrfs: add and use helpers for reading and writing fs_info->generation
btrfs: add and use helpers for reading and writing log_transid
btrfs: add and use helpers for reading and writing last_log_commit
btrfs: support cloned-device mount capability
btrfs: add helper function find_fsid_by_disk
btrfs: stop reserving excessive space for block group item insertions
btrfs: stop reserving excessive space for block group item updates
btrfs: reorder btrfs_inode to fill gaps
btrfs: open code btrfs_ordered_inode_tree in btrfs_inode
...
When creating a snapshot of a subvolume that was created in the current
transaction, we can end up not persisting a dirty extent buffer that is
referenced by the snapshot, resulting in IO errors due to checksum failures
when trying to read the extent buffer later from disk. A sequence of steps
that leads to this is the following:
1) At ioctl.c:create_subvol() we allocate an extent buffer, with logical
address 36007936, for the leaf/root of a new subvolume that has an ID
of 291. We mark the extent buffer as dirty, and at this point the
subvolume tree has a single node/leaf which is also its root (level 0);
2) We no longer commit the transaction used to create the subvolume at
create_subvol(). We used to, but that was recently removed in
commit 1b53e51a4a ("btrfs: don't commit transaction for every subvol
create");
3) The transaction used to create the subvolume has an ID of 33, so the
extent buffer 36007936 has a generation of 33;
4) Several updates happen to subvolume 291 during transaction 33, several
files created and its tree height changes from 0 to 1, so we end up with
a new root at level 1 and the extent buffer 36007936 is now a leaf of
that new root node, which is extent buffer 36048896.
The commit root remains as 36007936, since we are still at transaction
33;
5) Creation of a snapshot of subvolume 291, with an ID of 292, starts at
ioctl.c:create_snapshot(). This triggers a commit of transaction 33 and
we end up at transaction.c:create_pending_snapshot(), in the critical
section of a transaction commit.
There we COW the root of subvolume 291, which is extent buffer 36048896.
The COW operation returns extent buffer 36048896, since there's no need
to COW because the extent buffer was created in this transaction and it
was not written yet.
The we call btrfs_copy_root() against the root node 36048896. During
this operation we allocate a new extent buffer to turn into the root
node of the snapshot, copy the contents of the root node 36048896 into
this snapshot root extent buffer, set the owner to 292 (the ID of the
snapshot), etc, and then we call btrfs_inc_ref(). This will create a
delayed reference for each leaf pointed by the root node with a
reference root of 292 - this includes a reference for the leaf
36007936.
After that we set the bit BTRFS_ROOT_FORCE_COW in the root's state.
Then we call btrfs_insert_dir_item(), to create the directory entry in
in the tree of subvolume 291 that points to the snapshot. This ends up
needing to modify leaf 36007936 to insert the respective directory
items. Because the bit BTRFS_ROOT_FORCE_COW is set for the root's state,
we need to COW the leaf. We end up at btrfs_force_cow_block() and then
at update_ref_for_cow().
At update_ref_for_cow() we call btrfs_block_can_be_shared() which
returns false, despite the fact the leaf 36007936 is shared - the
subvolume's root and the snapshot's root point to that leaf. The
reason that it incorrectly returns false is because the commit root
of the subvolume is extent buffer 36007936 - it was the initial root
of the subvolume when we created it. So btrfs_block_can_be_shared()
which has the following logic:
int btrfs_block_can_be_shared(struct btrfs_root *root,
struct extent_buffer *buf)
{
if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
buf != root->node && buf != root->commit_root &&
(btrfs_header_generation(buf) <=
btrfs_root_last_snapshot(&root->root_item) ||
btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
return 1;
return 0;
}
Returns false (0) since 'buf' (extent buffer 36007936) matches the
root's commit root.
As a result, at update_ref_for_cow(), we don't check for the number
of references for extent buffer 36007936, we just assume it's not
shared and therefore that it has only 1 reference, so we set the local
variable 'refs' to 1.
Later on, in the final if-else statement at update_ref_for_cow():
static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf,
struct extent_buffer *cow,
int *last_ref)
{
(...)
if (refs > 1) {
(...)
} else {
(...)
btrfs_clear_buffer_dirty(trans, buf);
*last_ref = 1;
}
}
So we mark the extent buffer 36007936 as not dirty, and as a result
we don't write it to disk later in the transaction commit, despite the
fact that the snapshot's root points to it.
Attempting to access the leaf or dumping the tree for example shows
that the extent buffer was not written:
$ btrfs inspect-internal dump-tree -t 292 /dev/sdb
btrfs-progs v6.2.2
file tree key (292 ROOT_ITEM 33)
node 36110336 level 1 items 2 free space 119 generation 33 owner 292
node 36110336 flags 0x1(WRITTEN) backref revision 1
checksum stored a8103e3e
checksum calced a8103e3e
fs uuid 90c9a46f-ae9f-4626-9aff-0cbf3e2e3a79
chunk uuid e8c9c885-78f4-4d31-85fe-89e5f5fd4a07
key (256 INODE_ITEM 0) block 36007936 gen 33
key (257 EXTENT_DATA 0) block 36052992 gen 33
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
total bytes 107374182400
bytes used 38572032
uuid 90c9a46f-ae9f-4626-9aff-0cbf3e2e3a79
The respective on disk region is full of zeroes as the device was
trimmed at mkfs time.
Obviously 'btrfs check' also detects and complains about this:
$ btrfs check /dev/sdb
Opening filesystem to check...
Checking filesystem on /dev/sdb
UUID: 90c9a46f-ae9f-4626-9aff-0cbf3e2e3a79
generation: 33 (33)
[1/7] checking root items
[2/7] checking extents
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
bad tree block 36007936, bytenr mismatch, want=36007936, have=0
owner ref check failed [36007936 4096]
ERROR: errors found in extent allocation tree or chunk allocation
[3/7] checking free space tree
[4/7] checking fs roots
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
checksum verify failed on 36007936 wanted 0x00000000 found 0x86005f29
bad tree block 36007936, bytenr mismatch, want=36007936, have=0
The following tree block(s) is corrupted in tree 292:
tree block bytenr: 36110336, level: 1, node key: (256, 1, 0)
root 292 root dir 256 not found
ERROR: errors found in fs roots
found 38572032 bytes used, error(s) found
total csum bytes: 16048
total tree bytes: 1265664
total fs tree bytes: 1118208
total extent tree bytes: 65536
btree space waste bytes: 562598
file data blocks allocated: 65978368
referenced 36569088
Fix this by updating btrfs_block_can_be_shared() to consider that an
extent buffer may be shared if it matches the commit root and if its
generation matches the current transaction's generation.
This can be reproduced with the following script:
$ cat test.sh
#!/bin/bash
MNT=/mnt/sdi
DEV=/dev/sdi
# Use a filesystem with a 64K node size so that we have the same node
# size on every machine regardless of its page size (on x86_64 default
# node size is 16K due to the 4K page size, while on PPC it's 64K by
# default). This way we can make sure we are able to create a btree for
# the subvolume with a height of 2.
mkfs.btrfs -f -n 64K $DEV
mount $DEV $MNT
btrfs subvolume create $MNT/subvol
# Create a few empty files on the subvolume, this bumps its btree
# height to 2 (root node at level 1 and 2 leaves).
for ((i = 1; i <= 300; i++)); do
echo -n > $MNT/subvol/file_$i
done
btrfs subvolume snapshot -r $MNT/subvol $MNT/subvol/snap
umount $DEV
btrfs check $DEV
Running it on a 6.5 kernel (or any 6.6-rc kernel at the moment):
$ ./test.sh
Create subvolume '/mnt/sdi/subvol'
Create a readonly snapshot of '/mnt/sdi/subvol' in '/mnt/sdi/subvol/snap'
Opening filesystem to check...
Checking filesystem on /dev/sdi
UUID: bbdde2ff-7d02-45ca-8a73-3c36f23755a1
[1/7] checking root items
[2/7] checking extents
parent transid verify failed on 30539776 wanted 7 found 5
parent transid verify failed on 30539776 wanted 7 found 5
parent transid verify failed on 30539776 wanted 7 found 5
Ignoring transid failure
owner ref check failed [30539776 65536]
ERROR: errors found in extent allocation tree or chunk allocation
[3/7] checking free space tree
[4/7] checking fs roots
parent transid verify failed on 30539776 wanted 7 found 5
Ignoring transid failure
Wrong key of child node/leaf, wanted: (256, 1, 0), have: (2, 132, 0)
Wrong generation of child node/leaf, wanted: 5, have: 7
root 257 root dir 256 not found
ERROR: errors found in fs roots
found 917504 bytes used, error(s) found
total csum bytes: 0
total tree bytes: 851968
total fs tree bytes: 393216
total extent tree bytes: 65536
btree space waste bytes: 736550
file data blocks allocated: 0
referenced 0
A test case for fstests will follow soon.
Fixes: 1b53e51a4a ("btrfs: don't commit transaction for every subvol create")
CC: stable@vger.kernel.org # 6.5+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently the log_transid field of a root is always modified while holding
the root's log_mutex locked. Most readers of a root's log_transid are also
holding the root's log_mutex locked, however there is one exception which
is btrfs_set_inode_last_trans() where we don't take the lock to avoid
blocking several operations if log syncing is happening in parallel.
Any races here should be harmless, and in the worst case they may cause a
fsync to log an inode when it's not really needed, so nothing bad from a
functional perspective.
To avoid data race warnings from tools like KCSAN and other issues such
as load and store tearing (amongst others, see [1]), create helpers to
access the log_transid field of a root using READ_ONCE() and WRITE_ONCE(),
and use these helpers where needed.
[1] https://lwn.net/Articles/793253/
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently, the last_log_commit of a root can be accessed concurrently
without any lock protection. Readers can be calling btrfs_inode_in_log()
early in a fsync call, which reads a root's last_log_commit, while a
writer can change the last_log_commit while a log tree if being synced,
at btrfs_sync_log(). Any races here should be harmless, and in the worst
case they may cause a fsync to log an inode when it's not really needed,
so nothing bad from a functional perspective.
To avoid data race warnings from tools like KCSAN and other issues such
as load and store tearing (amongst others, see [1]), create helpers to
access the last_log_commit field of a root using READ_ONCE() and
WRITE_ONCE(), and use these helpers everywhere.
[1] https://lwn.net/Articles/793253/
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_realloc_node() is only used by the defrag code. Nowadays we have a
defrag.c file, so move it, and its helper close_blocks(), into defrag.c.
During the move also do a few minor cosmetic changes:
1) Change the return value of close_blocks() from int to bool;
2) Use SZ_32K instead of 32768 at close_blocks();
3) Make some variables const in btrfs_realloc_node(), 'blocksize' and
'end_slot';
4) Get rid of 'parent_nritems' variable, in both places where it was
used it could be replaced by calling btrfs_header_nritems(parent);
5) Change the type of a couple variables from int to bool;
6) Rename variable 'err' to 'ret', as that's the most common name we
use to track the return value of a function;
7) Move some variables from the top scope to the scope of the for loop
where they are used.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Export comp_keys() out of ctree.c, as btrfs_comp_keys(), so that in a
later patch we can move out defrag specific code from ctree.c into
defrag.c.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Rename and export __btrfs_cow_block() as btrfs_force_cow_block(). This is
to allow to move defrag specific code out of ctree.c and into defrag.c in
one of the next patches.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Relocation data allocations are quite tricky for simple quotas. The
basic data relocation sequence is (ignoring details that aren't relevant
to this fix):
- create a fake relocation data fs root
- create a fake relocation inode in that root
- for each data extent:
- preallocate a data extent on behalf of the fake inode
- copy over the data
- for each extent
- swap the refs so that the original file extent now refers to the new
extent item
- drop the fake root, dropping its refs on the old extents, which lets
us delete them.
Done naively, this results in storing an extent item in the extent tree
whose owner_ref points at the relocation data root and a no-op squota
recording, since the reloc root is not a legit fstree. So far, that's
OK. The problem comes when you do the swap, and leave an extent item
owned by this bogus root as the real permanent extents of the file. If
the file then drops that ref, we free it and no-op account that against
the fake relocation root. Essentially, this means that relocation is
simple quota "extent laundering", since we re-own the extents into a
fake root.
Simple quotas very intentionally doesn't have a mechanism for
transferring ownership of extents, as that is exactly the complicated
thing we are trying to avoid with the new design. Further, it cannot be
correctly done in this case, since at the time you create the new
"real" refs, there is no way to know which was the original owner before
relocation unless we track it.
Therefore, it makes more sense to trick the preallocation to handle
relocation as a special case and note the proper owner ref from the
beginning. That way, we never write out an extent item without the
correct owner ref that it will eventually have.
This could be done by wiring a special root parameter all the way
through the allocation code path, but to avoid that special case
touching all the code, take advantage of the serial nature of relocation
to store the src root on the relocation root object. Then when we finish
the prealloc, if it happens to be this case, prepare the delayed ref
appropriately.
We must also add logic to handle relocating adjacent extents with
different owning roots. Those cannot be preallocated together in a
cluster as it would lose the separate ownership information.
This is obviously a smelly bit of code, but I think it is the best
solution to the problem, given the relocation implementation.
Signed-off-by: Boris Burkov <boris@bur.io>
Signed-off-by: David Sterba <dsterba@suse.com>
When marking an extent buffer as dirty, at btrfs_mark_buffer_dirty(),
we check if its generation matches the running transaction and if not we
just print a warning. Such mismatch is an indicator that something really
went wrong and only printing a warning message (and stack trace) is not
enough to prevent a corruption. Allowing a transaction to commit with such
an extent buffer will trigger an error if we ever try to read it from disk
due to a generation mismatch with its parent generation.
So abort the current transaction with -EUCLEAN if we notice a generation
mismatch. For this we need to pass a transaction handle to
btrfs_mark_buffer_dirty() which is always available except in test code,
in which case we can pass NULL since it operates on dummy extent buffers
and all test roots have a single node/leaf (root node at level 0).
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We don't need any of these includes in the ctree.h header file for the
header file itself, remove them to clean up ctree.h a little bit.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is related to the name hashing for dir items, move it into
dir-item.h.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Ideally this would be un-inlined, but that is a cleanup for later. For
now move this into inode-item.h, which is where the extref code lives.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This simply sends the same arguments into crc32c(), and is just used in
a few places. Remove this wrapper and directly call crc32c() in these
instances.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is the only place this helper is used, take it out of ctree.h and
move it into free-space-cache.c.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The readdir implementation currently processes always up to the last index
it finds. This however can result in an infinite loop if the directory has
a large number of entries such that they won't all fit in the given buffer
passed to the readdir callback, that is, dir_emit() returns a non-zero
value. Because in that case readdir() will be called again and if in the
meanwhile new directory entries were added and we still can't put all the
remaining entries in the buffer, we keep repeating this over and over.
The following C program and test script reproduce the problem:
$ cat /mnt/readdir_prog.c
#include <sys/types.h>
#include <dirent.h>
#include <stdio.h>
int main(int argc, char *argv[])
{
DIR *dir = opendir(".");
struct dirent *dd;
while ((dd = readdir(dir))) {
printf("%s\n", dd->d_name);
rename(dd->d_name, "TEMPFILE");
rename("TEMPFILE", dd->d_name);
}
closedir(dir);
}
$ gcc -o /mnt/readdir_prog /mnt/readdir_prog.c
$ cat test.sh
#!/bin/bash
DEV=/dev/sdi
MNT=/mnt/sdi
mkfs.btrfs -f $DEV &> /dev/null
#mkfs.xfs -f $DEV &> /dev/null
#mkfs.ext4 -F $DEV &> /dev/null
mount $DEV $MNT
mkdir $MNT/testdir
for ((i = 1; i <= 2000; i++)); do
echo -n > $MNT/testdir/file_$i
done
cd $MNT/testdir
/mnt/readdir_prog
cd /mnt
umount $MNT
This behaviour is surprising to applications and it's unlike ext4, xfs,
tmpfs, vfat and other filesystems, which always finish. In this case where
new entries were added due to renames, some file names may be reported
more than once, but this varies according to each filesystem - for example
ext4 never reported the same file more than once while xfs reports the
first 13 file names twice.
So change our readdir implementation to track the last index number when
opendir() is called and then make readdir() never process beyond that
index number. This gives the same behaviour as ext4.
Reported-by: Rob Landley <rob@landley.net>
Link: https://lore.kernel.org/linux-btrfs/2c8c55ec-04c6-e0dc-9c5c-8c7924778c35@landley.net/
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217681
CC: stable@vger.kernel.org # 6.4+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
At btrfs_del_ptr(), instead of doing a BUG_ON() in case we fail to record
tree mod log operations, do a transaction abort and return the error to
the callers. There's really no need for the BUG_ON() as we can release all
resources in the context of all callers, and we have to abort because other
future tree searches that use the tree mod log (btrfs_search_old_slot())
may get inconsistent results if other operations modify the tree after
that failure and before the tree mod log based search.
This implies btrfs_del_ptr() return an int instead of void, and making all
callers check for returned errors.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This exists internal to ctree.c, however btrfs check needs to use it for
some of its operations. I'd rather not duplicate that code inside of
btrfs check as this is low level and I want to keep this code in one
place, so rename the function to btrfs_del_ptr and export it so that it
can be used inside of btrfs-progs safely. Add a comment to make sure
this doesn't get removed by a future cleanup.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is needed in btrfs-progs for the tools that convert the checksum
types for file systems and a few other things. We don't have it in the
kernel as we just want to get the size for the super blocks type.
However I don't want to have to manually add this every time we sync
ctree.c into btrfs-progs, so add the helper in the kernel with a note so
it doesn't get removed by a later cleanup.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since print-tree infrastructure only prints the content of a tree block,
we can make them to accept const extent buffer pointer.
This removes a forced type convert in extent-tree, where we convert a
const extent buffer pointer to regular one, just to avoid compiler
warning.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_prev_leaf() is not used outside ctree.c, so there's no need to
export it at ctree.h - just make it static at ctree.c and move its
definition above btrfs_search_slot_for_read(), since that function
calls it.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_bin_search() is a simple wrapper that searches for the whole slots
by calling btrfs_generic_bin_search() with the starting slot/first_slot
preset to 0.
This simple wrapper can be open coded as btrfs_bin_search().
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In the search loop of the binary search function, we are doing a division
by 2 of the sum of the high and low slots. Because the slots are integers,
the generated assembly code for it is the following on x86_64:
0x00000000000141f1 <+145>: mov %eax,%ebx
0x00000000000141f3 <+147>: shr $0x1f,%ebx
0x00000000000141f6 <+150>: add %eax,%ebx
0x00000000000141f8 <+152>: sar %ebx
It's a few more instructions than a simple right shift, because signed
integer division needs to round towards zero. However we know that slots
can never be negative (btrfs_header_nritems() returns an u32), so we
can instead use unsigned types for the low and high slots and therefore
use unsigned integer division, which results in a single instruction on
x86_64:
0x00000000000141f0 <+144>: shr %ebx
So use unsigned types for the slots and therefore unsigned division.
This is part of a small patchset comprised of the following two patches:
btrfs: eliminate extra call when doing binary search on extent buffer
btrfs: do unsigned integer division in the extent buffer binary search loop
The following fs_mark test was run on a non-debug kernel (Debian's default
kernel config) before and after applying the patchset:
$ cat test.sh
#!/bin/bash
DEV=/dev/sdi
MNT=/mnt/sdi
MOUNT_OPTIONS="-o ssd"
MKFS_OPTIONS="-O no-holes -R free-space-tree"
FILES=100000
THREADS=$(nproc --all)
FILE_SIZE=0
umount $DEV &> /dev/null
mkfs.btrfs -f $MKFS_OPTIONS $DEV
mount $MOUNT_OPTIONS $DEV $MNT
OPTS="-S 0 -L 6 -n $FILES -s $FILE_SIZE -t $THREADS -k"
for ((i = 1; i <= $THREADS; i++)); do
OPTS="$OPTS -d $MNT/d$i"
done
fs_mark $OPTS
umount $MNT
Results before applying patchset:
FSUse% Count Size Files/sec App Overhead
2 1200000 0 174472.0 11549868
4 2400000 0 253503.0 11694618
4 3600000 0 257833.1 11611508
6 4800000 0 247089.5 11665983
6 6000000 0 211296.1 12121244
10 7200000 0 187330.6 12548565
Results after applying patchset:
FSUse% Count Size Files/sec App Overhead
2 1200000 0 207556.0 11393252
4 2400000 0 266751.1 11347909
4 3600000 0 274397.5 11270058
6 4800000 0 259608.4 11442250
6 6000000 0 238895.8 11635921
8 7200000 0 211942.2 11873825
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
