linux/drivers/gpu/drm/vmwgfx/ttm_object.c

668 lines
17 KiB
C
Raw Permalink Normal View History

/* SPDX-License-Identifier: GPL-2.0 OR MIT */
/**************************************************************************
*
* Copyright (c) 2009-2023 VMware, Inc., Palo Alto, CA., USA
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sub license, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial portions
* of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*
**************************************************************************/
/*
* Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
*
* While no substantial code is shared, the prime code is inspired by
* drm_prime.c, with
* Authors:
* Dave Airlie <airlied@redhat.com>
* Rob Clark <rob.clark@linaro.org>
*/
/** @file ttm_ref_object.c
*
* Base- and reference object implementation for the various
* ttm objects. Implements reference counting, minimal security checks
* and release on file close.
*/
#define pr_fmt(fmt) "[TTM] " fmt
#include "ttm_object.h"
#include "vmwgfx_drv.h"
#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/slab.h>
#include <linux/atomic.h>
#include <linux/module.h>
#include <linux/hashtable.h>
MODULE_IMPORT_NS("DMA_BUF");
#define VMW_TTM_OBJECT_REF_HT_ORDER 10
/**
* struct ttm_object_file
*
* @tdev: Pointer to the ttm_object_device.
*
* @lock: Lock that protects the ref_list list and the
* ref_hash hash tables.
*
* @ref_list: List of ttm_ref_objects to be destroyed at
* file release.
*
* @ref_hash: Hash tables of ref objects, one per ttm_ref_type,
* for fast lookup of ref objects given a base object.
*
* @refcount: reference/usage count
*/
struct ttm_object_file {
struct ttm_object_device *tdev;
spinlock_t lock;
struct list_head ref_list;
DECLARE_HASHTABLE(ref_hash, VMW_TTM_OBJECT_REF_HT_ORDER);
struct kref refcount;
};
drm/vmwgfx/ttm_object: Demote half-assed headers and fix-up another Fixes the following W=1 kernel build warning(s): drivers/gpu/drm/vmwgfx/ttm_object.c:60: error: Cannot parse struct or union! drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'mem_glob' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'ops' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dmabuf_release' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dma_buf_size' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'idr' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'rcu_head' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'tfile' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Function parameter or member 'dmabuf' not described in 'get_dma_buf_unless_doomed' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Excess function parameter 'dma_buf' description in 'get_dma_buf_unless_doomed' Cc: VMware Graphics <linux-graphics-maintainer@vmware.com> Cc: Roland Scheidegger <sroland@vmware.com> Cc: Zack Rusin <zackr@vmware.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: "Christian König" <christian.koenig@amd.com> Cc: Dave Airlie <airlied@redhat.com> Cc: Rob Clark <rob.clark@linaro.org> Cc: dri-devel@lists.freedesktop.org Cc: linux-media@vger.kernel.org Cc: linaro-mm-sig@lists.linaro.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Zack Rusin <zackr@vmware.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210115181601.3432599-10-lee.jones@linaro.org
2021-01-15 18:15:41 +00:00
/*
* struct ttm_object_device
*
* @object_lock: lock that protects idr.
*
* This is the per-device data structure needed for ttm object management.
*/
struct ttm_object_device {
spinlock_t object_lock;
struct dma_buf_ops ops;
void (*dmabuf_release)(struct dma_buf *dma_buf);
struct idr idr;
};
drm/vmwgfx/ttm_object: Demote half-assed headers and fix-up another Fixes the following W=1 kernel build warning(s): drivers/gpu/drm/vmwgfx/ttm_object.c:60: error: Cannot parse struct or union! drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'mem_glob' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'ops' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dmabuf_release' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dma_buf_size' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'idr' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'rcu_head' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'tfile' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Function parameter or member 'dmabuf' not described in 'get_dma_buf_unless_doomed' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Excess function parameter 'dma_buf' description in 'get_dma_buf_unless_doomed' Cc: VMware Graphics <linux-graphics-maintainer@vmware.com> Cc: Roland Scheidegger <sroland@vmware.com> Cc: Zack Rusin <zackr@vmware.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: "Christian König" <christian.koenig@amd.com> Cc: Dave Airlie <airlied@redhat.com> Cc: Rob Clark <rob.clark@linaro.org> Cc: dri-devel@lists.freedesktop.org Cc: linux-media@vger.kernel.org Cc: linaro-mm-sig@lists.linaro.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Zack Rusin <zackr@vmware.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210115181601.3432599-10-lee.jones@linaro.org
2021-01-15 18:15:41 +00:00
/*
* struct ttm_ref_object
*
* @hash: Hash entry for the per-file object reference hash.
*
* @head: List entry for the per-file list of ref-objects.
*
* @kref: Ref count.
*
* @obj: Base object this ref object is referencing.
*
* @ref_type: Type of ref object.
*
* This is similar to an idr object, but it also has a hash table entry
* that allows lookup with a pointer to the referenced object as a key. In
* that way, one can easily detect whether a base object is referenced by
* a particular ttm_object_file. It also carries a ref count to avoid creating
* multiple ref objects if a ttm_object_file references the same base
* object more than once.
*/
struct ttm_ref_object {
struct rcu_head rcu_head;
struct vmwgfx_hash_item hash;
struct list_head head;
struct kref kref;
struct ttm_base_object *obj;
struct ttm_object_file *tfile;
};
static void ttm_prime_dmabuf_release(struct dma_buf *dma_buf);
static inline struct ttm_object_file *
ttm_object_file_ref(struct ttm_object_file *tfile)
{
kref_get(&tfile->refcount);
return tfile;
}
static int ttm_tfile_find_ref_rcu(struct ttm_object_file *tfile,
uint64_t key,
struct vmwgfx_hash_item **p_hash)
{
struct vmwgfx_hash_item *hash;
hash_for_each_possible_rcu(tfile->ref_hash, hash, head, key) {
if (hash->key == key) {
*p_hash = hash;
return 0;
}
}
return -EINVAL;
}
static int ttm_tfile_find_ref(struct ttm_object_file *tfile,
uint64_t key,
struct vmwgfx_hash_item **p_hash)
{
struct vmwgfx_hash_item *hash;
hash_for_each_possible(tfile->ref_hash, hash, head, key) {
if (hash->key == key) {
*p_hash = hash;
return 0;
}
}
return -EINVAL;
}
static void ttm_object_file_destroy(struct kref *kref)
{
struct ttm_object_file *tfile =
container_of(kref, struct ttm_object_file, refcount);
kfree(tfile);
}
static inline void ttm_object_file_unref(struct ttm_object_file **p_tfile)
{
struct ttm_object_file *tfile = *p_tfile;
*p_tfile = NULL;
kref_put(&tfile->refcount, ttm_object_file_destroy);
}
int ttm_base_object_init(struct ttm_object_file *tfile,
struct ttm_base_object *base,
bool shareable,
enum ttm_object_type object_type,
void (*refcount_release) (struct ttm_base_object **))
{
struct ttm_object_device *tdev = tfile->tdev;
int ret;
base->shareable = shareable;
base->tfile = ttm_object_file_ref(tfile);
base->refcount_release = refcount_release;
base->object_type = object_type;
kref_init(&base->refcount);
idr_preload(GFP_KERNEL);
spin_lock(&tdev->object_lock);
ret = idr_alloc(&tdev->idr, base, 1, 0, GFP_NOWAIT);
spin_unlock(&tdev->object_lock);
idr_preload_end();
if (ret < 0)
return ret;
base->handle = ret;
ret = ttm_ref_object_add(tfile, base, NULL, false);
if (unlikely(ret != 0))
goto out_err1;
ttm_base_object_unref(&base);
return 0;
out_err1:
spin_lock(&tdev->object_lock);
idr_remove(&tdev->idr, base->handle);
spin_unlock(&tdev->object_lock);
return ret;
}
static void ttm_release_base(struct kref *kref)
{
struct ttm_base_object *base =
container_of(kref, struct ttm_base_object, refcount);
struct ttm_object_device *tdev = base->tfile->tdev;
spin_lock(&tdev->object_lock);
idr_remove(&tdev->idr, base->handle);
spin_unlock(&tdev->object_lock);
/*
* Note: We don't use synchronize_rcu() here because it's far
* too slow. It's up to the user to free the object using
* call_rcu() or ttm_base_object_kfree().
*/
ttm_object_file_unref(&base->tfile);
if (base->refcount_release)
base->refcount_release(&base);
}
void ttm_base_object_unref(struct ttm_base_object **p_base)
{
struct ttm_base_object *base = *p_base;
*p_base = NULL;
kref_put(&base->refcount, ttm_release_base);
}
struct ttm_base_object *ttm_base_object_lookup(struct ttm_object_file *tfile,
uint64_t key)
{
struct ttm_base_object *base = NULL;
struct vmwgfx_hash_item *hash;
int ret;
spin_lock(&tfile->lock);
ret = ttm_tfile_find_ref(tfile, key, &hash);
if (likely(ret == 0)) {
base = hlist_entry(hash, struct ttm_ref_object, hash)->obj;
if (!kref_get_unless_zero(&base->refcount))
base = NULL;
}
spin_unlock(&tfile->lock);
return base;
}
struct ttm_base_object *
ttm_base_object_lookup_for_ref(struct ttm_object_device *tdev, uint64_t key)
{
struct ttm_base_object *base;
rcu_read_lock();
base = idr_find(&tdev->idr, key);
if (base && !kref_get_unless_zero(&base->refcount))
base = NULL;
rcu_read_unlock();
return base;
}
int ttm_ref_object_add(struct ttm_object_file *tfile,
struct ttm_base_object *base,
bool *existed,
bool require_existed)
{
struct ttm_ref_object *ref;
struct vmwgfx_hash_item *hash;
int ret = -EINVAL;
if (base->tfile != tfile && !base->shareable)
return -EPERM;
if (existed != NULL)
*existed = true;
while (ret == -EINVAL) {
rcu_read_lock();
ret = ttm_tfile_find_ref_rcu(tfile, base->handle, &hash);
if (ret == 0) {
ref = hlist_entry(hash, struct ttm_ref_object, hash);
if (kref_get_unless_zero(&ref->kref)) {
rcu_read_unlock();
break;
}
}
rcu_read_unlock();
if (require_existed)
return -EPERM;
ref = kmalloc(sizeof(*ref), GFP_KERNEL);
if (unlikely(ref == NULL)) {
return -ENOMEM;
}
ref->hash.key = base->handle;
ref->obj = base;
ref->tfile = tfile;
kref_init(&ref->kref);
spin_lock(&tfile->lock);
hash_add_rcu(tfile->ref_hash, &ref->hash.head, ref->hash.key);
ret = 0;
list_add_tail(&ref->head, &tfile->ref_list);
kref_get(&base->refcount);
spin_unlock(&tfile->lock);
if (existed != NULL)
*existed = false;
}
return ret;
}
static void __releases(tfile->lock) __acquires(tfile->lock)
ttm_ref_object_release(struct kref *kref)
{
struct ttm_ref_object *ref =
container_of(kref, struct ttm_ref_object, kref);
struct ttm_object_file *tfile = ref->tfile;
hash_del_rcu(&ref->hash.head);
list_del(&ref->head);
spin_unlock(&tfile->lock);
ttm_base_object_unref(&ref->obj);
kfree_rcu(ref, rcu_head);
spin_lock(&tfile->lock);
}
int ttm_ref_object_base_unref(struct ttm_object_file *tfile,
unsigned long key)
{
struct ttm_ref_object *ref;
struct vmwgfx_hash_item *hash;
int ret;
spin_lock(&tfile->lock);
ret = ttm_tfile_find_ref(tfile, key, &hash);
if (unlikely(ret != 0)) {
spin_unlock(&tfile->lock);
return -EINVAL;
}
ref = hlist_entry(hash, struct ttm_ref_object, hash);
kref_put(&ref->kref, ttm_ref_object_release);
spin_unlock(&tfile->lock);
return 0;
}
void ttm_object_file_release(struct ttm_object_file **p_tfile)
{
struct ttm_ref_object *ref;
struct list_head *list;
struct ttm_object_file *tfile = *p_tfile;
*p_tfile = NULL;
spin_lock(&tfile->lock);
/*
* Since we release the lock within the loop, we have to
* restart it from the beginning each time.
*/
while (!list_empty(&tfile->ref_list)) {
list = tfile->ref_list.next;
ref = list_entry(list, struct ttm_ref_object, head);
ttm_ref_object_release(&ref->kref);
}
spin_unlock(&tfile->lock);
ttm_object_file_unref(&tfile);
}
struct ttm_object_file *ttm_object_file_init(struct ttm_object_device *tdev)
{
struct ttm_object_file *tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
if (unlikely(tfile == NULL))
return NULL;
spin_lock_init(&tfile->lock);
tfile->tdev = tdev;
kref_init(&tfile->refcount);
INIT_LIST_HEAD(&tfile->ref_list);
hash_init(tfile->ref_hash);
return tfile;
}
struct ttm_object_device *
ttm_object_device_init(const struct dma_buf_ops *ops)
{
struct ttm_object_device *tdev = kmalloc(sizeof(*tdev), GFP_KERNEL);
if (unlikely(tdev == NULL))
return NULL;
spin_lock_init(&tdev->object_lock);
/*
* Our base is at VMWGFX_NUM_MOB + 1 because we want to create
* a seperate namespace for GEM handles (which are
* 1..VMWGFX_NUM_MOB) and the surface handles. Some ioctl's
* can take either handle as an argument so we want to
* easily be able to tell whether the handle refers to a
* GEM buffer or a surface.
*/
idr_init_base(&tdev->idr, VMWGFX_NUM_MOB + 1);
tdev->ops = *ops;
tdev->dmabuf_release = tdev->ops.release;
tdev->ops.release = ttm_prime_dmabuf_release;
return tdev;
}
void ttm_object_device_release(struct ttm_object_device **p_tdev)
{
struct ttm_object_device *tdev = *p_tdev;
*p_tdev = NULL;
WARN_ON_ONCE(!idr_is_empty(&tdev->idr));
idr_destroy(&tdev->idr);
kfree(tdev);
}
/**
* get_dma_buf_unless_doomed - get a dma_buf reference if possible.
*
drm/vmwgfx/ttm_object: Demote half-assed headers and fix-up another Fixes the following W=1 kernel build warning(s): drivers/gpu/drm/vmwgfx/ttm_object.c:60: error: Cannot parse struct or union! drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'mem_glob' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'ops' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dmabuf_release' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'dma_buf_size' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:97: warning: Function parameter or member 'idr' not described in 'ttm_object_device' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'rcu_head' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:128: warning: Function parameter or member 'tfile' not described in 'ttm_ref_object' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Function parameter or member 'dmabuf' not described in 'get_dma_buf_unless_doomed' drivers/gpu/drm/vmwgfx/ttm_object.c:582: warning: Excess function parameter 'dma_buf' description in 'get_dma_buf_unless_doomed' Cc: VMware Graphics <linux-graphics-maintainer@vmware.com> Cc: Roland Scheidegger <sroland@vmware.com> Cc: Zack Rusin <zackr@vmware.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: "Christian König" <christian.koenig@amd.com> Cc: Dave Airlie <airlied@redhat.com> Cc: Rob Clark <rob.clark@linaro.org> Cc: dri-devel@lists.freedesktop.org Cc: linux-media@vger.kernel.org Cc: linaro-mm-sig@lists.linaro.org Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Zack Rusin <zackr@vmware.com> Link: https://patchwork.freedesktop.org/patch/msgid/20210115181601.3432599-10-lee.jones@linaro.org
2021-01-15 18:15:41 +00:00
* @dmabuf: Non-refcounted pointer to a struct dma-buf.
*
* Obtain a file reference from a lookup structure that doesn't refcount
* the file, but synchronizes with its release method to make sure it has
* not been freed yet. See for example kref_get_unless_zero documentation.
* Returns true if refcounting succeeds, false otherwise.
*
* Nobody really wants this as a public API yet, so let it mature here
* for some time...
*/
static bool __must_check get_dma_buf_unless_doomed(struct dma_buf *dmabuf)
{
fs: port files to file_ref Port files to rely on file_ref reference to improve scaling and gain overflow protection. - We continue to WARN during get_file() in case a file that is already marked dead is revived as get_file() is only valid if the caller already holds a reference to the file. This hasn't changed just the check changes. - The semantics for epoll and ttm's dmabuf usage have changed. Both epoll and ttm synchronize with __fput() to prevent the underlying file from beeing freed. (1) epoll Explaining epoll is straightforward using a simple diagram. Essentially, the mutex of the epoll instance needs to be taken in both __fput() and around epi_fget() preventing the file from being freed while it is polled or preventing the file from being resurrected. CPU1 CPU2 fput(file) -> __fput(file) -> eventpoll_release(file) -> eventpoll_release_file(file) mutex_lock(&ep->mtx) epi_item_poll() -> epi_fget() -> file_ref_get(file) mutex_unlock(&ep->mtx) mutex_lock(&ep->mtx); __ep_remove() mutex_unlock(&ep->mtx); -> kmem_cache_free(file) (2) ttm dmabuf This explanation is a bit more involved. A regular dmabuf file stashed the dmabuf in file->private_data and the file in dmabuf->file: file->private_data = dmabuf; dmabuf->file = file; The generic release method of a dmabuf file handles file specific things: f_op->release::dma_buf_file_release() while the generic dentry release method of a dmabuf handles dmabuf freeing including driver specific things: dentry->d_release::dma_buf_release() During ttm dmabuf initialization in ttm_object_device_init() the ttm driver copies the provided struct dma_buf_ops into a private location: struct ttm_object_device { spinlock_t object_lock; struct dma_buf_ops ops; void (*dmabuf_release)(struct dma_buf *dma_buf); struct idr idr; }; ttm_object_device_init(const struct dma_buf_ops *ops) { // copy original dma_buf_ops in private location tdev->ops = *ops; // stash the release method of the original struct dma_buf_ops tdev->dmabuf_release = tdev->ops.release; // override the release method in the copy of the struct dma_buf_ops // with ttm's own dmabuf release method tdev->ops.release = ttm_prime_dmabuf_release; } When a new dmabuf is created the struct dma_buf_ops with the overriden release method set to ttm_prime_dmabuf_release is passed in exp_info.ops: DEFINE_DMA_BUF_EXPORT_INFO(exp_info); exp_info.ops = &tdev->ops; exp_info.size = prime->size; exp_info.flags = flags; exp_info.priv = prime; The call to dma_buf_export() then sets mutex_lock_interruptible(&prime->mutex); dma_buf = dma_buf_export(&exp_info) { dmabuf->ops = exp_info->ops; } mutex_unlock(&prime->mutex); which creates a new dmabuf file and then install a file descriptor to it in the callers file descriptor table: ret = dma_buf_fd(dma_buf, flags); When that dmabuf file is closed we now get: fput(file) -> __fput(file) -> f_op->release::dma_buf_file_release() -> dput() -> d_op->d_release::dma_buf_release() -> dmabuf->ops->release::ttm_prime_dmabuf_release() mutex_lock(&prime->mutex); if (prime->dma_buf == dma_buf) prime->dma_buf = NULL; mutex_unlock(&prime->mutex); Where we can see that prime->dma_buf is set to NULL. So when we have the following diagram: CPU1 CPU2 fput(file) -> __fput(file) -> f_op->release::dma_buf_file_release() -> dput() -> d_op->d_release::dma_buf_release() -> dmabuf->ops->release::ttm_prime_dmabuf_release() ttm_prime_handle_to_fd() mutex_lock_interruptible(&prime->mutex) dma_buf = prime->dma_buf dma_buf && get_dma_buf_unless_doomed(dma_buf) -> file_ref_get(dma_buf->file) mutex_unlock(&prime->mutex); mutex_lock(&prime->mutex); if (prime->dma_buf == dma_buf) prime->dma_buf = NULL; mutex_unlock(&prime->mutex); -> kmem_cache_free(file) The logic of the mechanism is the same as for epoll: sync with __fput() preventing the file from being freed. Here the synchronization happens through the ttm instance's prime->mutex. Basically, the lifetime of the dma_buf and the file are tighly coupled. Both (1) and (2) used to call atomic_inc_not_zero() to check whether the file has already been marked dead and then refuse to revive it. This is only safe because both (1) and (2) sync with __fput() and thus prevent kmem_cache_free() on the file being called and thus prevent the file from being immediately recycled due to SLAB_TYPESAFE_BY_RCU. Both (1) and (2) have been ported from atomic_inc_not_zero() to file_ref_get(). That means a file that is already in the process of being marked as FILE_REF_DEAD: file_ref_put() cnt = atomic_long_dec_return() -> __file_ref_put(cnt) if (cnt == FIlE_REF_NOREF) atomic_long_try_cmpxchg_release(cnt, FILE_REF_DEAD) can be revived again: CPU1 CPU2 file_ref_put() cnt = atomic_long_dec_return() -> __file_ref_put(cnt) if (cnt == FIlE_REF_NOREF) file_ref_get() // Brings reference back to FILE_REF_ONEREF atomic_long_add_negative() atomic_long_try_cmpxchg_release(cnt, FILE_REF_DEAD) This is fine and inherent to the file_ref_get()/file_ref_put() semantics. For both (1) and (2) this is safe because __fput() is prevented from making progress if file_ref_get() fails due to the aforementioned synchronization mechanisms. Two cases need to be considered that affect both (1) epoll and (2) ttm dmabuf: (i) fput()'s file_ref_put() and marks the file as FILE_REF_NOREF but before that fput() can mark the file as FILE_REF_DEAD someone manages to sneak in a file_ref_get() and brings the refcount back from FILE_REF_NOREF to FILE_REF_ONEREF. In that case the original fput() doesn't call __fput(). For epoll the poll will finish and for ttm dmabuf the file can be used again. For ttm dambuf this is actually an advantage because it avoids immediately allocating a new dmabuf object. CPU1 CPU2 file_ref_put() cnt = atomic_long_dec_return() -> __file_ref_put(cnt) if (cnt == FIlE_REF_NOREF) file_ref_get() // Brings reference back to FILE_REF_ONEREF atomic_long_add_negative() atomic_long_try_cmpxchg_release(cnt, FILE_REF_DEAD) (ii) fput()'s file_ref_put() marks the file FILE_REF_NOREF and also suceeds in actually marking it FILE_REF_DEAD and then calls into __fput() to free the file. When either (1) or (2) call file_ref_get() they fail as atomic_long_add_negative() will return true. At the same time, both (1) and (2) all file_ref_get() under mutexes that __fput() must also acquire preventing kmem_cache_free() from freeing the file. So while this might be treated as a change in semantics for (1) and (2) it really isn't. It if should end up causing issues this can be fixed by adding a helper that does something like: long cnt = atomic_long_read(&ref->refcnt); do { if (cnt < 0) return false; } while (!atomic_long_try_cmpxchg(&ref->refcnt, &cnt, cnt + 1)); return true; which would block FILE_REF_NOREF to FILE_REF_ONEREF transitions. - Jann correctly pointed out that kmem_cache_zalloc() cannot be used anymore once files have been ported to file_ref_t. The kmem_cache_zalloc() call will memset() the whole struct file to zero when it is reallocated. This will also set file->f_ref to zero which mens that a concurrent file_ref_get() can return true: CPU1 CPU2 __get_file_rcu() rcu_dereference_raw() close() [frees file] alloc_empty_file() kmem_cache_zalloc() [reallocates same file] memset(..., 0, ...) file_ref_get() [increments 0->1, returns true] init_file() file_ref_init(..., 1) [sets to 0] rcu_dereference_raw() fput() file_ref_put() [decrements 0->FILE_REF_NOREF, frees file] [UAF] causing a concurrent __get_file_rcu() call to acquire a reference to the file that is about to be reallocated and immediately freeing it on realizing that it has been recycled. This causes a UAF for the task that reallocated/recycled the file. This is prevented by switching from kmem_cache_zalloc() to kmem_cache_alloc() and initializing the fields manually. With file->f_ref initialized last. Note that a memset() also isn't guaranteed to atomically update an unsigned long so it's theoretically possible to see torn and therefore bogus counter values. Link: https://lore.kernel.org/r/20241007-brauner-file-rcuref-v2-3-387e24dc9163@kernel.org Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-10-07 16:23:59 +02:00
return file_ref_get(&dmabuf->file->f_ref);
}
/**
* ttm_prime_refcount_release - refcount release method for a prime object.
*
* @p_base: Pointer to ttm_base_object pointer.
*
* This is a wrapper that calls the refcount_release founction of the
* underlying object. At the same time it cleans up the prime object.
* This function is called when all references to the base object we
* derive from are gone.
*/
static void ttm_prime_refcount_release(struct ttm_base_object **p_base)
{
struct ttm_base_object *base = *p_base;
struct ttm_prime_object *prime;
*p_base = NULL;
prime = container_of(base, struct ttm_prime_object, base);
BUG_ON(prime->dma_buf != NULL);
mutex_destroy(&prime->mutex);
if (prime->refcount_release)
prime->refcount_release(&base);
}
/**
* ttm_prime_dmabuf_release - Release method for the dma-bufs we export
*
* @dma_buf:
*
* This function first calls the dma_buf release method the driver
* provides. Then it cleans up our dma_buf pointer used for lookup,
* and finally releases the reference the dma_buf has on our base
* object.
*/
static void ttm_prime_dmabuf_release(struct dma_buf *dma_buf)
{
struct ttm_prime_object *prime =
(struct ttm_prime_object *) dma_buf->priv;
struct ttm_base_object *base = &prime->base;
struct ttm_object_device *tdev = base->tfile->tdev;
if (tdev->dmabuf_release)
tdev->dmabuf_release(dma_buf);
mutex_lock(&prime->mutex);
if (prime->dma_buf == dma_buf)
prime->dma_buf = NULL;
mutex_unlock(&prime->mutex);
ttm_base_object_unref(&base);
}
/**
* ttm_prime_fd_to_handle - Get a base object handle from a prime fd
*
* @tfile: A struct ttm_object_file identifying the caller.
* @fd: The prime / dmabuf fd.
* @handle: The returned handle.
*
* This function returns a handle to an object that previously exported
* a dma-buf. Note that we don't handle imports yet, because we simply
* have no consumers of that implementation.
*/
int ttm_prime_fd_to_handle(struct ttm_object_file *tfile,
int fd, u32 *handle)
{
struct ttm_object_device *tdev = tfile->tdev;
struct dma_buf *dma_buf;
struct ttm_prime_object *prime;
struct ttm_base_object *base;
int ret;
dma_buf = dma_buf_get(fd);
if (IS_ERR(dma_buf))
return PTR_ERR(dma_buf);
if (dma_buf->ops != &tdev->ops)
return -ENOSYS;
prime = (struct ttm_prime_object *) dma_buf->priv;
base = &prime->base;
*handle = base->handle;
ret = ttm_ref_object_add(tfile, base, NULL, false);
dma_buf_put(dma_buf);
return ret;
}
/**
* ttm_prime_handle_to_fd - Return a dma_buf fd from a ttm prime object
*
* @tfile: Struct ttm_object_file identifying the caller.
* @handle: Handle to the object we're exporting from.
* @flags: flags for dma-buf creation. We just pass them on.
* @prime_fd: The returned file descriptor.
*
*/
int ttm_prime_handle_to_fd(struct ttm_object_file *tfile,
uint32_t handle, uint32_t flags,
int *prime_fd)
{
struct ttm_object_device *tdev = tfile->tdev;
struct ttm_base_object *base;
struct dma_buf *dma_buf;
struct ttm_prime_object *prime;
int ret;
base = ttm_base_object_lookup(tfile, handle);
if (unlikely(base == NULL ||
base->object_type != ttm_prime_type)) {
ret = -ENOENT;
goto out_unref;
}
prime = container_of(base, struct ttm_prime_object, base);
if (unlikely(!base->shareable)) {
ret = -EPERM;
goto out_unref;
}
ret = mutex_lock_interruptible(&prime->mutex);
if (unlikely(ret != 0)) {
ret = -ERESTARTSYS;
goto out_unref;
}
dma_buf = prime->dma_buf;
if (!dma_buf || !get_dma_buf_unless_doomed(dma_buf)) {
DEFINE_DMA_BUF_EXPORT_INFO(exp_info);
exp_info.ops = &tdev->ops;
exp_info.size = prime->size;
exp_info.flags = flags;
exp_info.priv = prime;
/*
* Need to create a new dma_buf
*/
dma_buf = dma_buf_export(&exp_info);
if (IS_ERR(dma_buf)) {
ret = PTR_ERR(dma_buf);
mutex_unlock(&prime->mutex);
goto out_unref;
}
/*
* dma_buf has taken the base object reference
*/
base = NULL;
prime->dma_buf = dma_buf;
}
mutex_unlock(&prime->mutex);
ret = dma_buf_fd(dma_buf, flags);
if (ret >= 0) {
*prime_fd = ret;
ret = 0;
} else
dma_buf_put(dma_buf);
out_unref:
if (base)
ttm_base_object_unref(&base);
return ret;
}
/**
* ttm_prime_object_init - Initialize a ttm_prime_object
*
* @tfile: struct ttm_object_file identifying the caller
* @size: The size of the dma_bufs we export.
* @prime: The object to be initialized.
* @type: See ttm_base_object_init
* @refcount_release: See ttm_base_object_init
*
* Initializes an object which is compatible with the drm_prime model
* for data sharing between processes and devices.
*/
int ttm_prime_object_init(struct ttm_object_file *tfile, size_t size,
struct ttm_prime_object *prime,
enum ttm_object_type type,
void (*refcount_release) (struct ttm_base_object **))
{
bool shareable = !!(type == VMW_RES_SURFACE);
mutex_init(&prime->mutex);
prime->size = PAGE_ALIGN(size);
prime->real_type = type;
prime->dma_buf = NULL;
prime->refcount_release = refcount_release;
return ttm_base_object_init(tfile, &prime->base, shareable,
ttm_prime_type,
ttm_prime_refcount_release);
}