1
0
Fork 0
alistair23-linux/drivers/gpu/drm/drm_gem.c

858 lines
23 KiB
C
Raw Normal View History

/*
* Copyright © 2008 Intel Corporation
*
* 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, sublicense,
* 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 NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS 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:
* Eric Anholt <eric@anholt.net>
*
*/
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/uaccess.h>
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/module.h>
#include <linux/mman.h>
#include <linux/pagemap.h>
#include <linux/shmem_fs.h>
#include <linux/dma-buf.h>
#include <drm/drmP.h>
#include <drm/drm_vma_manager.h>
/** @file drm_gem.c
*
* This file provides some of the base ioctls and library routines for
* the graphics memory manager implemented by each device driver.
*
* Because various devices have different requirements in terms of
* synchronization and migration strategies, implementing that is left up to
* the driver, and all that the general API provides should be generic --
* allocating objects, reading/writing data with the cpu, freeing objects.
* Even there, platform-dependent optimizations for reading/writing data with
* the CPU mean we'll likely hook those out to driver-specific calls. However,
* the DRI2 implementation wants to have at least allocate/mmap be generic.
*
* The goal was to have swap-backed object allocation managed through
* struct file. However, file descriptors as handles to a struct file have
* two major failings:
* - Process limits prevent more than 1024 or so being used at a time by
* default.
* - Inability to allocate high fds will aggravate the X Server's select()
* handling, and likely that of many GL client applications as well.
*
* This led to a plan of using our own integer IDs (called handles, following
* DRM terminology) to mimic fds, and implement the fd syscalls we need as
* ioctls. The objects themselves will still include the struct file so
* that we can transition to fds if the required kernel infrastructure shows
* up at a later date, and as our interface with shmfs for memory allocation.
*/
/*
* We make up offsets for buffer objects so we can recognize them at
* mmap time.
*/
/* pgoff in mmap is an unsigned long, so we need to make sure that
* the faked up offset will fit
*/
#if BITS_PER_LONG == 64
#define DRM_FILE_PAGE_OFFSET_START ((0xFFFFFFFFUL >> PAGE_SHIFT) + 1)
#define DRM_FILE_PAGE_OFFSET_SIZE ((0xFFFFFFFFUL >> PAGE_SHIFT) * 16)
#else
#define DRM_FILE_PAGE_OFFSET_START ((0xFFFFFFFUL >> PAGE_SHIFT) + 1)
#define DRM_FILE_PAGE_OFFSET_SIZE ((0xFFFFFFFUL >> PAGE_SHIFT) * 16)
#endif
/**
* Initialize the GEM device fields
*/
int
drm_gem_init(struct drm_device *dev)
{
struct drm_gem_mm *mm;
mutex_init(&dev->object_name_lock);
idr_init(&dev->object_name_idr);
mm = kzalloc(sizeof(struct drm_gem_mm), GFP_KERNEL);
if (!mm) {
DRM_ERROR("out of memory\n");
return -ENOMEM;
}
dev->mm_private = mm;
drm_vma_offset_manager_init(&mm->vma_manager,
DRM_FILE_PAGE_OFFSET_START,
DRM_FILE_PAGE_OFFSET_SIZE);
return 0;
}
void
drm_gem_destroy(struct drm_device *dev)
{
struct drm_gem_mm *mm = dev->mm_private;
drm_vma_offset_manager_destroy(&mm->vma_manager);
kfree(mm);
dev->mm_private = NULL;
}
/**
* Initialize an already allocated GEM object of the specified size with
* shmfs backing store.
*/
int drm_gem_object_init(struct drm_device *dev,
struct drm_gem_object *obj, size_t size)
{
struct file *filp;
filp = shmem_file_setup("drm mm object", size, VM_NORESERVE);
if (IS_ERR(filp))
return PTR_ERR(filp);
drm_gem_private_object_init(dev, obj, size);
obj->filp = filp;
return 0;
}
EXPORT_SYMBOL(drm_gem_object_init);
/**
* Initialize an already allocated GEM object of the specified size with
* no GEM provided backing store. Instead the caller is responsible for
* backing the object and handling it.
*/
void drm_gem_private_object_init(struct drm_device *dev,
struct drm_gem_object *obj, size_t size)
{
BUG_ON((size & (PAGE_SIZE - 1)) != 0);
obj->dev = dev;
obj->filp = NULL;
kref_init(&obj->refcount);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
obj->handle_count = 0;
obj->size = size;
}
EXPORT_SYMBOL(drm_gem_private_object_init);
/**
* Allocate a GEM object of the specified size with shmfs backing store
*/
struct drm_gem_object *
drm_gem_object_alloc(struct drm_device *dev, size_t size)
{
struct drm_gem_object *obj;
obj = kzalloc(sizeof(*obj), GFP_KERNEL);
if (!obj)
goto free;
if (drm_gem_object_init(dev, obj, size) != 0)
goto free;
if (dev->driver->gem_init_object != NULL &&
dev->driver->gem_init_object(obj) != 0) {
goto fput;
}
return obj;
fput:
/* Object_init mangles the global counters - readjust them. */
fput(obj->filp);
free:
kfree(obj);
return NULL;
}
EXPORT_SYMBOL(drm_gem_object_alloc);
static void
drm_gem_remove_prime_handles(struct drm_gem_object *obj, struct drm_file *filp)
{
if (obj->import_attach) {
drm/prime: keep a reference from the handle to exported dma-buf (v6) Currently we have a problem with this: 1. i915: create gem object 2. i915: export gem object to prime 3. radeon: import gem object 4. close prime fd 5. radeon: unref object 6. i915: unref object i915 has an imported object reference in its file priv, that isn't cleaned up properly until fd close. The reference gets added at step 2, but at step 6 we don't have enough info to clean it up. The solution is to take a reference on the dma-buf when we export it, and drop the reference when the gem handle goes away. So when we export a dma_buf from a gem object, we keep track of it with the handle, we take a reference to the dma_buf. When we close the handle (i.e. userspace is finished with the buffer), we drop the reference to the dma_buf, and it gets collected. This patch isn't meant to fix any other problem or bikesheds, and it doesn't fix any races with other scenarios. v1.1: move export symbol line back up. v2: okay I had to do a bit more, as the first patch showed a leak on one of my tests, that I found using the dma-buf debugfs support, the problem case is exporting a buffer twice with the same handle, we'd add another export handle for it unnecessarily, however we now fail if we try to export the same object with a different gem handle, however I'm not sure if that is a case I want to support, and I've gotten the code to WARN_ON if we hit something like that. v2.1: rebase this patch, write better commit msg. v3: cleanup error handling, track import vs export in linked list, these two patches were separate previously, but seem to work better like this. v4: danvet is correct, this code is no longer useful, since the buffer better exist, so remove it. v5: always take a reference to the dma buf object, import or export. (Imre Deak contributed this originally) v6: square the circle, remove import vs export tracking now that there is no difference Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-04-21 17:54:36 -06:00
drm_prime_remove_buf_handle(&filp->prime,
obj->import_attach->dmabuf);
}
if (obj->export_dma_buf) {
drm/prime: keep a reference from the handle to exported dma-buf (v6) Currently we have a problem with this: 1. i915: create gem object 2. i915: export gem object to prime 3. radeon: import gem object 4. close prime fd 5. radeon: unref object 6. i915: unref object i915 has an imported object reference in its file priv, that isn't cleaned up properly until fd close. The reference gets added at step 2, but at step 6 we don't have enough info to clean it up. The solution is to take a reference on the dma-buf when we export it, and drop the reference when the gem handle goes away. So when we export a dma_buf from a gem object, we keep track of it with the handle, we take a reference to the dma_buf. When we close the handle (i.e. userspace is finished with the buffer), we drop the reference to the dma_buf, and it gets collected. This patch isn't meant to fix any other problem or bikesheds, and it doesn't fix any races with other scenarios. v1.1: move export symbol line back up. v2: okay I had to do a bit more, as the first patch showed a leak on one of my tests, that I found using the dma-buf debugfs support, the problem case is exporting a buffer twice with the same handle, we'd add another export handle for it unnecessarily, however we now fail if we try to export the same object with a different gem handle, however I'm not sure if that is a case I want to support, and I've gotten the code to WARN_ON if we hit something like that. v2.1: rebase this patch, write better commit msg. v3: cleanup error handling, track import vs export in linked list, these two patches were separate previously, but seem to work better like this. v4: danvet is correct, this code is no longer useful, since the buffer better exist, so remove it. v5: always take a reference to the dma buf object, import or export. (Imre Deak contributed this originally) v6: square the circle, remove import vs export tracking now that there is no difference Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-04-21 17:54:36 -06:00
drm_prime_remove_buf_handle(&filp->prime,
obj->export_dma_buf);
}
}
static void drm_gem_object_ref_bug(struct kref *list_kref)
{
BUG();
}
/**
* Called after the last handle to the object has been closed
*
* Removes any name for the object. Note that this must be
* called before drm_gem_object_free or we'll be touching
* freed memory
*/
static void drm_gem_object_handle_free(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
/* Remove any name for this object */
if (obj->name) {
idr_remove(&dev->object_name_idr, obj->name);
obj->name = 0;
/*
* The object name held a reference to this object, drop
* that now.
*
* This cannot be the last reference, since the handle holds one too.
*/
kref_put(&obj->refcount, drm_gem_object_ref_bug);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
}
}
static void
drm_gem_object_handle_unreference_unlocked(struct drm_gem_object *obj)
{
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
if (WARN_ON(obj->handle_count == 0))
return;
/*
* Must bump handle count first as this may be the last
* ref, in which case the object would disappear before we
* checked for a name
*/
mutex_lock(&obj->dev->object_name_lock);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
if (--obj->handle_count == 0)
drm_gem_object_handle_free(obj);
mutex_unlock(&obj->dev->object_name_lock);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
drm_gem_object_unreference_unlocked(obj);
}
/**
* Removes the mapping from handle to filp for this object.
*/
int
drm_gem_handle_delete(struct drm_file *filp, u32 handle)
{
struct drm_device *dev;
struct drm_gem_object *obj;
/* This is gross. The idr system doesn't let us try a delete and
* return an error code. It just spews if you fail at deleting.
* So, we have to grab a lock around finding the object and then
* doing the delete on it and dropping the refcount, or the user
* could race us to double-decrement the refcount and cause a
* use-after-free later. Given the frequency of our handle lookups,
* we may want to use ida for number allocation and a hash table
* for the pointers, anyway.
*/
spin_lock(&filp->table_lock);
/* Check if we currently have a reference on the object */
obj = idr_find(&filp->object_idr, handle);
if (obj == NULL) {
spin_unlock(&filp->table_lock);
return -EINVAL;
}
dev = obj->dev;
/* Release reference and decrement refcount. */
idr_remove(&filp->object_idr, handle);
spin_unlock(&filp->table_lock);
drm_gem_remove_prime_handles(obj, filp);
if (dev->driver->gem_close_object)
dev->driver->gem_close_object(obj, filp);
drm_gem_object_handle_unreference_unlocked(obj);
return 0;
}
EXPORT_SYMBOL(drm_gem_handle_delete);
/**
* drm_gem_dumb_destroy - dumb fb callback helper for gem based drivers
*
* This implements the ->dumb_destroy kms driver callback for drivers which use
* gem to manage their backing storage.
*/
int drm_gem_dumb_destroy(struct drm_file *file,
struct drm_device *dev,
uint32_t handle)
{
return drm_gem_handle_delete(file, handle);
}
EXPORT_SYMBOL(drm_gem_dumb_destroy);
/**
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
* drm_gem_handle_create_tail - internal functions to create a handle
*
* This expects the dev->object_name_lock to be held already and will drop it
* before returning. Used to avoid races in establishing new handles when
* importing an object from either an flink name or a dma-buf.
*/
int
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
drm_gem_handle_create_tail(struct drm_file *file_priv,
struct drm_gem_object *obj,
u32 *handlep)
{
struct drm_device *dev = obj->dev;
int ret;
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
WARN_ON(!mutex_is_locked(&dev->object_name_lock));
/*
* Get the user-visible handle using idr. Preload and perform
* allocation under our spinlock.
*/
idr_preload(GFP_KERNEL);
spin_lock(&file_priv->table_lock);
ret = idr_alloc(&file_priv->object_idr, obj, 1, 0, GFP_NOWAIT);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
drm_gem_object_reference(obj);
obj->handle_count++;
spin_unlock(&file_priv->table_lock);
idr_preload_end();
mutex_unlock(&dev->object_name_lock);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
if (ret < 0) {
drm_gem_object_handle_unreference_unlocked(obj);
return ret;
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
}
*handlep = ret;
if (dev->driver->gem_open_object) {
ret = dev->driver->gem_open_object(obj, file_priv);
if (ret) {
drm_gem_handle_delete(file_priv, *handlep);
return ret;
}
}
return 0;
}
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
/**
* Create a handle for this object. This adds a handle reference
* to the object, which includes a regular reference count. Callers
* will likely want to dereference the object afterwards.
*/
int
drm_gem_handle_create(struct drm_file *file_priv,
struct drm_gem_object *obj,
u32 *handlep)
{
mutex_lock(&obj->dev->object_name_lock);
return drm_gem_handle_create_tail(file_priv, obj, handlep);
}
EXPORT_SYMBOL(drm_gem_handle_create);
/**
* drm_gem_free_mmap_offset - release a fake mmap offset for an object
* @obj: obj in question
*
* This routine frees fake offsets allocated by drm_gem_create_mmap_offset().
*/
void
drm_gem_free_mmap_offset(struct drm_gem_object *obj)
{
struct drm_device *dev = obj->dev;
struct drm_gem_mm *mm = dev->mm_private;
drm_vma_offset_remove(&mm->vma_manager, &obj->vma_node);
}
EXPORT_SYMBOL(drm_gem_free_mmap_offset);
/**
* drm_gem_create_mmap_offset_size - create a fake mmap offset for an object
* @obj: obj in question
* @size: the virtual size
*
* GEM memory mapping works by handing back to userspace a fake mmap offset
* it can use in a subsequent mmap(2) call. The DRM core code then looks
* up the object based on the offset and sets up the various memory mapping
* structures.
*
* This routine allocates and attaches a fake offset for @obj, in cases where
* the virtual size differs from the physical size (ie. obj->size). Otherwise
* just use drm_gem_create_mmap_offset().
*/
int
drm_gem_create_mmap_offset_size(struct drm_gem_object *obj, size_t size)
{
struct drm_device *dev = obj->dev;
struct drm_gem_mm *mm = dev->mm_private;
return drm_vma_offset_add(&mm->vma_manager, &obj->vma_node,
size / PAGE_SIZE);
}
EXPORT_SYMBOL(drm_gem_create_mmap_offset_size);
/**
* drm_gem_create_mmap_offset - create a fake mmap offset for an object
* @obj: obj in question
*
* GEM memory mapping works by handing back to userspace a fake mmap offset
* it can use in a subsequent mmap(2) call. The DRM core code then looks
* up the object based on the offset and sets up the various memory mapping
* structures.
*
* This routine allocates and attaches a fake offset for @obj.
*/
int drm_gem_create_mmap_offset(struct drm_gem_object *obj)
{
return drm_gem_create_mmap_offset_size(obj, obj->size);
}
EXPORT_SYMBOL(drm_gem_create_mmap_offset);
/**
* drm_gem_get_pages - helper to allocate backing pages for a GEM object
* from shmem
* @obj: obj in question
* @gfpmask: gfp mask of requested pages
*/
struct page **drm_gem_get_pages(struct drm_gem_object *obj, gfp_t gfpmask)
{
struct inode *inode;
struct address_space *mapping;
struct page *p, **pages;
int i, npages;
/* This is the shared memory object that backs the GEM resource */
inode = file_inode(obj->filp);
mapping = inode->i_mapping;
/* We already BUG_ON() for non-page-aligned sizes in
* drm_gem_object_init(), so we should never hit this unless
* driver author is doing something really wrong:
*/
WARN_ON((obj->size & (PAGE_SIZE - 1)) != 0);
npages = obj->size >> PAGE_SHIFT;
pages = drm_malloc_ab(npages, sizeof(struct page *));
if (pages == NULL)
return ERR_PTR(-ENOMEM);
gfpmask |= mapping_gfp_mask(mapping);
for (i = 0; i < npages; i++) {
p = shmem_read_mapping_page_gfp(mapping, i, gfpmask);
if (IS_ERR(p))
goto fail;
pages[i] = p;
/* There is a hypothetical issue w/ drivers that require
* buffer memory in the low 4GB.. if the pages are un-
* pinned, and swapped out, they can end up swapped back
* in above 4GB. If pages are already in memory, then
* shmem_read_mapping_page_gfp will ignore the gfpmask,
* even if the already in-memory page disobeys the mask.
*
* It is only a theoretical issue today, because none of
* the devices with this limitation can be populated with
* enough memory to trigger the issue. But this BUG_ON()
* is here as a reminder in case the problem with
* shmem_read_mapping_page_gfp() isn't solved by the time
* it does become a real issue.
*
* See this thread: http://lkml.org/lkml/2011/7/11/238
*/
BUG_ON((gfpmask & __GFP_DMA32) &&
(page_to_pfn(p) >= 0x00100000UL));
}
return pages;
fail:
while (i--)
page_cache_release(pages[i]);
drm_free_large(pages);
return ERR_CAST(p);
}
EXPORT_SYMBOL(drm_gem_get_pages);
/**
* drm_gem_put_pages - helper to free backing pages for a GEM object
* @obj: obj in question
* @pages: pages to free
* @dirty: if true, pages will be marked as dirty
* @accessed: if true, the pages will be marked as accessed
*/
void drm_gem_put_pages(struct drm_gem_object *obj, struct page **pages,
bool dirty, bool accessed)
{
int i, npages;
/* We already BUG_ON() for non-page-aligned sizes in
* drm_gem_object_init(), so we should never hit this unless
* driver author is doing something really wrong:
*/
WARN_ON((obj->size & (PAGE_SIZE - 1)) != 0);
npages = obj->size >> PAGE_SHIFT;
for (i = 0; i < npages; i++) {
if (dirty)
set_page_dirty(pages[i]);
if (accessed)
mark_page_accessed(pages[i]);
/* Undo the reference we took when populating the table */
page_cache_release(pages[i]);
}
drm_free_large(pages);
}
EXPORT_SYMBOL(drm_gem_put_pages);
/** Returns a reference to the object named by the handle. */
struct drm_gem_object *
drm_gem_object_lookup(struct drm_device *dev, struct drm_file *filp,
u32 handle)
{
struct drm_gem_object *obj;
spin_lock(&filp->table_lock);
/* Check if we currently have a reference on the object */
obj = idr_find(&filp->object_idr, handle);
if (obj == NULL) {
spin_unlock(&filp->table_lock);
return NULL;
}
drm_gem_object_reference(obj);
spin_unlock(&filp->table_lock);
return obj;
}
EXPORT_SYMBOL(drm_gem_object_lookup);
/**
* Releases the handle to an mm object.
*/
int
drm_gem_close_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_gem_close *args = data;
int ret;
if (!(dev->driver->driver_features & DRIVER_GEM))
return -ENODEV;
ret = drm_gem_handle_delete(file_priv, args->handle);
return ret;
}
/**
* Create a global name for an object, returning the name.
*
* Note that the name does not hold a reference; when the object
* is freed, the name goes away.
*/
int
drm_gem_flink_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_gem_flink *args = data;
struct drm_gem_object *obj;
int ret;
if (!(dev->driver->driver_features & DRIVER_GEM))
return -ENODEV;
obj = drm_gem_object_lookup(dev, file_priv, args->handle);
if (obj == NULL)
return -ENOENT;
mutex_lock(&dev->object_name_lock);
idr_preload(GFP_KERNEL);
drm/gem: fix up flink name create race This is the 2nd attempt, I've always been a bit dissatisified with the tricky nature of the first one: http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html The issue is that the flink ioctl can race with calling gem_close on the last gem handle. In that case we'll end up with a zero handle count, but an flink name (and it's corresponding reference). Which results in a neat space leak. In my first attempt I've solved this by rechecking the handle count. But fundamentally the issue is that ->handle_count isn't your usual refcount - it can be resurrected from 0 among other things. For those special beasts atomic_t often suggest way more ordering that it actually guarantees. To prevent being tricked by those hairy semantics take the easy way out and simply protect the handle with the existing dev->object_name_lock. With that change implemented it's dead easy to fix the flink vs. gem close reace: When we try to create the name we simply have to check whether there's still officially a gem handle around and if not refuse to create the flink name. Since the handle count decrement and flink name destruction is now also protected by that lock the reace is gone and we can't ever leak the flink reference again. Outside of the drm core only the exynos driver looks at the handle count, and tbh I have no idea why (it's just for debug dmesg output luckily). I've considered inlining the drm_gem_object_handle_free, but I plan to add more name-like things (like the exported dma_buf) to this scheme, so it's clearer to leave the handle freeing in its own function. This is exercised by the new gem_flink_race i-g-t testcase, which on my snb leaks gem objects at a rate of roughly 1k objects/s. v2: Fix up the error path handling in handle_create and make it more robust by simply calling object_handle_unreference. v3: Fix up the handle_unreference logic bug - atomic_dec_and_test retursn 1 for 0. Oops. v4: Squash in inlining of drm_gem_object_handle_reference as suggested by Dave Airlie and add a note that we now have a testcase. Cc: Dave Airlie <airlied@gmail.com> Cc: Inki Dae <inki.dae@samsung.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:37 -06:00
/* prevent races with concurrent gem_close. */
if (obj->handle_count == 0) {
ret = -ENOENT;
goto err;
}
if (!obj->name) {
ret = idr_alloc(&dev->object_name_idr, obj, 1, 0, GFP_NOWAIT);
if (ret < 0)
goto err;
obj->name = ret;
/* Allocate a reference for the name table. */
drm_gem_object_reference(obj);
}
args->name = (uint64_t) obj->name;
ret = 0;
err:
idr_preload_end();
mutex_unlock(&dev->object_name_lock);
drm_gem_object_unreference_unlocked(obj);
return ret;
}
/**
* Open an object using the global name, returning a handle and the size.
*
* This handle (of course) holds a reference to the object, so the object
* will not go away until the handle is deleted.
*/
int
drm_gem_open_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_gem_open *args = data;
struct drm_gem_object *obj;
int ret;
u32 handle;
if (!(dev->driver->driver_features & DRIVER_GEM))
return -ENODEV;
mutex_lock(&dev->object_name_lock);
obj = idr_find(&dev->object_name_idr, (int) args->name);
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
if (obj) {
drm_gem_object_reference(obj);
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
} else {
mutex_unlock(&dev->object_name_lock);
return -ENOENT;
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
}
drm/gem: completely close gem_open vs. gem_close races The gem flink name holds a reference onto the object itself, and this self-reference would prevent an flink'ed object from every being freed. To break that loop we remove the flink name when the last userspace handle disappears, i.e. when obj->handle_count reaches 0. Now in gem_open we drop the dev->object_name_lock between the flink name lookup and actually adding the handle. This means a concurrent gem_close of the last handle could result in the flink name getting reaped right inbetween, i.e. Thread 1 Thread 2 gem_open gem_close flink -> obj lookup handle_count drops to 0 remove flink name create_handle handle_count++ If someone now flinks this object again, we'll get a new flink name. We can close this race by removing the lock dropping and making the entire lookup+handle_create sequence atomic. Unfortunately to still be able to share the handle_create logic this requires a handle_create_tail function which drops the lock - we can't hold the object_name_lock while calling into a driver's ->gem_open callback. Note that for flink fixing this race isn't really important, since racing gem_open against gem_close is clearly a userspace bug. And no matter how the race ends, we won't leak any references. But with dma-buf where the userspace dma-buf fd itself is refcounted this is a valid sequence and hence we should fix it. Therefore this patch here is just a warm-up exercise (and for consistency between flink buffer sharing and dma-buf buffer sharing with self-imports). Also note that this extension of the critical section in gem_open protected by dev->object_name_lock only works because it's now a mutex: A spinlock would conflict with the potential memory allocation in idr_preload(). This is exercises by igt/gem_flink_race/flink_name. Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-08-14 16:02:45 -06:00
/* drm_gem_handle_create_tail unlocks dev->object_name_lock. */
ret = drm_gem_handle_create_tail(file_priv, obj, &handle);
drm_gem_object_unreference_unlocked(obj);
if (ret)
return ret;
args->handle = handle;
args->size = obj->size;
return 0;
}
/**
* Called at device open time, sets up the structure for handling refcounting
* of mm objects.
*/
void
drm_gem_open(struct drm_device *dev, struct drm_file *file_private)
{
idr_init(&file_private->object_idr);
spin_lock_init(&file_private->table_lock);
}
/**
* Called at device close to release the file's
* handle references on objects.
*/
static int
drm_gem_object_release_handle(int id, void *ptr, void *data)
{
struct drm_file *file_priv = data;
struct drm_gem_object *obj = ptr;
struct drm_device *dev = obj->dev;
drm_gem_remove_prime_handles(obj, file_priv);
if (dev->driver->gem_close_object)
dev->driver->gem_close_object(obj, file_priv);
drm_gem_object_handle_unreference_unlocked(obj);
return 0;
}
/**
* Called at close time when the filp is going away.
*
* Releases any remaining references on objects by this filp.
*/
void
drm_gem_release(struct drm_device *dev, struct drm_file *file_private)
{
idr_for_each(&file_private->object_idr,
&drm_gem_object_release_handle, file_private);
idr_destroy(&file_private->object_idr);
}
void
drm_gem_object_release(struct drm_gem_object *obj)
{
if (obj->filp)
fput(obj->filp);
}
EXPORT_SYMBOL(drm_gem_object_release);
/**
* Called after the last reference to the object has been lost.
* Must be called holding struct_ mutex
*
* Frees the object
*/
void
drm_gem_object_free(struct kref *kref)
{
struct drm_gem_object *obj = (struct drm_gem_object *) kref;
struct drm_device *dev = obj->dev;
BUG_ON(!mutex_is_locked(&dev->struct_mutex));
if (dev->driver->gem_free_object != NULL)
dev->driver->gem_free_object(obj);
}
EXPORT_SYMBOL(drm_gem_object_free);
void drm_gem_vm_open(struct vm_area_struct *vma)
{
struct drm_gem_object *obj = vma->vm_private_data;
drm_gem_object_reference(obj);
mutex_lock(&obj->dev->struct_mutex);
drm_vm_open_locked(obj->dev, vma);
mutex_unlock(&obj->dev->struct_mutex);
}
EXPORT_SYMBOL(drm_gem_vm_open);
void drm_gem_vm_close(struct vm_area_struct *vma)
{
struct drm_gem_object *obj = vma->vm_private_data;
drm: Fix use-after-free in drm_gem_vm_close() As we may release the last reference, we need to store the device in a local variable in order to unlock afterwards. [ 60.140768] BUG: unable to handle kernel paging request at 6b6b6b9f [ 60.140973] IP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] *pdpt = 0000000024a54001 *pde = 0000000000000000 [ 60.141014] Oops: 0002 [#1] PREEMPT SMP [ 60.141014] last sysfs file: /sys/devices/LNXSYSTM:00/device:00/PNP0A08:00/PNP0C0A:00/power_supply/BAT0/voltage_now [ 60.141014] Modules linked in: uvcvideo ath9k pegasus ath9k_common ath9k_hw hid_egalax ath3k joydev asus_laptop sparse_keymap battery input_polldev [ 60.141014] [ 60.141014] Pid: 771, comm: meego-ux-daemon Not tainted 2.6.37.2-7.1 #1 EXOPC EXOPG06411/EXOPG06411 [ 60.141014] EIP: 0060:[<c1536d11>] EFLAGS: 00010046 CPU: 0 [ 60.141014] EIP is at __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] EAX: 00000100 EBX: 6b6b6b9b ECX: e9b4a1b0 EDX: e4a4e580 [ 60.141014] ESI: db162558 EDI: 00000246 EBP: e480be50 ESP: e480be44 [ 60.141014] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 [ 60.141014] Process meego-ux-daemon (pid: 771, ti=e480a000 task=e9b4a1b0 task.ti=e480a000) [ 60.141014] Stack: [ 60.141014] e4a4e580 db162558 f5a2f838 e480be58 c1536dd0 e480be68 c125ab1b db162558 [ 60.141014] db1624e0 e480be78 c10ba071 db162558 f760241c e480be94 c10bb0bc 000155fe [ 60.141014] f760241c f5a2f838 f5a2f8c8 00000000 e480bea4 c1037c24 00000000 f5a2f838 [ 60.141014] Call Trace: [ 60.141014] [<c1536dd0>] ? mutex_unlock+0x8/0xa [ 60.141014] [<c125ab1b>] ? drm_gem_vm_close+0x39/0x3d [ 60.141014] [<c10ba071>] ? remove_vma+0x2d/0x58 [ 60.141014] [<c10bb0bc>] ? exit_mmap+0x126/0x13f [ 60.141014] [<c1037c24>] ? mmput+0x37/0x9a [ 60.141014] [<c10d450d>] ? exec_mmap+0x178/0x19c [ 60.141014] [<c1537f85>] ? _raw_spin_unlock+0x1d/0x36 [ 60.141014] [<c10d4eb0>] ? flush_old_exec+0x42/0x75 [ 60.141014] [<c1104442>] ? load_elf_binary+0x32a/0x922 [ 60.141014] [<c10d3f76>] ? search_binary_handler+0x200/0x2ea [ 60.141014] [<c10d3ecf>] ? search_binary_handler+0x159/0x2ea [ 60.141014] [<c1104118>] ? load_elf_binary+0x0/0x922 [ 60.141014] [<c10d56b2>] ? do_execve+0x1ff/0x2e6 [ 60.141014] [<c100970e>] ? sys_execve+0x2d/0x55 [ 60.141014] [<c1002a5a>] ? ptregs_execve+0x12/0x18 [ 60.141014] [<c10029dc>] ? sysenter_do_call+0x12/0x3c [ 60.141014] [<c1530000>] ? init_centaur+0x9c/0x1ba [ 60.141014] Code: c1 00 75 0f ba 38 01 00 00 b8 8c 3a 6c c1 e8 cc 2e b0 ff 9c 58 8d 74 26 00 89 c7 fa 90 8d 74 26 00 e8 d2 b4 b2 ff b8 00 01 00 00 <f0> 66 0f c1 43 04 38 e0 74 07 f3 90 8a 43 04 eb f5 83 3d 64 ef [ 60.141014] EIP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 SS:ESP 0068:e480be44 [ 60.141014] CR2: 000000006b6b6b9f Reported-by: Rusty Lynch <rusty.lynch@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-17 16:33:33 -06:00
struct drm_device *dev = obj->dev;
drm: Fix use-after-free in drm_gem_vm_close() As we may release the last reference, we need to store the device in a local variable in order to unlock afterwards. [ 60.140768] BUG: unable to handle kernel paging request at 6b6b6b9f [ 60.140973] IP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] *pdpt = 0000000024a54001 *pde = 0000000000000000 [ 60.141014] Oops: 0002 [#1] PREEMPT SMP [ 60.141014] last sysfs file: /sys/devices/LNXSYSTM:00/device:00/PNP0A08:00/PNP0C0A:00/power_supply/BAT0/voltage_now [ 60.141014] Modules linked in: uvcvideo ath9k pegasus ath9k_common ath9k_hw hid_egalax ath3k joydev asus_laptop sparse_keymap battery input_polldev [ 60.141014] [ 60.141014] Pid: 771, comm: meego-ux-daemon Not tainted 2.6.37.2-7.1 #1 EXOPC EXOPG06411/EXOPG06411 [ 60.141014] EIP: 0060:[<c1536d11>] EFLAGS: 00010046 CPU: 0 [ 60.141014] EIP is at __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] EAX: 00000100 EBX: 6b6b6b9b ECX: e9b4a1b0 EDX: e4a4e580 [ 60.141014] ESI: db162558 EDI: 00000246 EBP: e480be50 ESP: e480be44 [ 60.141014] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 [ 60.141014] Process meego-ux-daemon (pid: 771, ti=e480a000 task=e9b4a1b0 task.ti=e480a000) [ 60.141014] Stack: [ 60.141014] e4a4e580 db162558 f5a2f838 e480be58 c1536dd0 e480be68 c125ab1b db162558 [ 60.141014] db1624e0 e480be78 c10ba071 db162558 f760241c e480be94 c10bb0bc 000155fe [ 60.141014] f760241c f5a2f838 f5a2f8c8 00000000 e480bea4 c1037c24 00000000 f5a2f838 [ 60.141014] Call Trace: [ 60.141014] [<c1536dd0>] ? mutex_unlock+0x8/0xa [ 60.141014] [<c125ab1b>] ? drm_gem_vm_close+0x39/0x3d [ 60.141014] [<c10ba071>] ? remove_vma+0x2d/0x58 [ 60.141014] [<c10bb0bc>] ? exit_mmap+0x126/0x13f [ 60.141014] [<c1037c24>] ? mmput+0x37/0x9a [ 60.141014] [<c10d450d>] ? exec_mmap+0x178/0x19c [ 60.141014] [<c1537f85>] ? _raw_spin_unlock+0x1d/0x36 [ 60.141014] [<c10d4eb0>] ? flush_old_exec+0x42/0x75 [ 60.141014] [<c1104442>] ? load_elf_binary+0x32a/0x922 [ 60.141014] [<c10d3f76>] ? search_binary_handler+0x200/0x2ea [ 60.141014] [<c10d3ecf>] ? search_binary_handler+0x159/0x2ea [ 60.141014] [<c1104118>] ? load_elf_binary+0x0/0x922 [ 60.141014] [<c10d56b2>] ? do_execve+0x1ff/0x2e6 [ 60.141014] [<c100970e>] ? sys_execve+0x2d/0x55 [ 60.141014] [<c1002a5a>] ? ptregs_execve+0x12/0x18 [ 60.141014] [<c10029dc>] ? sysenter_do_call+0x12/0x3c [ 60.141014] [<c1530000>] ? init_centaur+0x9c/0x1ba [ 60.141014] Code: c1 00 75 0f ba 38 01 00 00 b8 8c 3a 6c c1 e8 cc 2e b0 ff 9c 58 8d 74 26 00 89 c7 fa 90 8d 74 26 00 e8 d2 b4 b2 ff b8 00 01 00 00 <f0> 66 0f c1 43 04 38 e0 74 07 f3 90 8a 43 04 eb f5 83 3d 64 ef [ 60.141014] EIP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 SS:ESP 0068:e480be44 [ 60.141014] CR2: 000000006b6b6b9f Reported-by: Rusty Lynch <rusty.lynch@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-17 16:33:33 -06:00
mutex_lock(&dev->struct_mutex);
drm_vm_close_locked(obj->dev, vma);
drm_gem_object_unreference(obj);
drm: Fix use-after-free in drm_gem_vm_close() As we may release the last reference, we need to store the device in a local variable in order to unlock afterwards. [ 60.140768] BUG: unable to handle kernel paging request at 6b6b6b9f [ 60.140973] IP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] *pdpt = 0000000024a54001 *pde = 0000000000000000 [ 60.141014] Oops: 0002 [#1] PREEMPT SMP [ 60.141014] last sysfs file: /sys/devices/LNXSYSTM:00/device:00/PNP0A08:00/PNP0C0A:00/power_supply/BAT0/voltage_now [ 60.141014] Modules linked in: uvcvideo ath9k pegasus ath9k_common ath9k_hw hid_egalax ath3k joydev asus_laptop sparse_keymap battery input_polldev [ 60.141014] [ 60.141014] Pid: 771, comm: meego-ux-daemon Not tainted 2.6.37.2-7.1 #1 EXOPC EXOPG06411/EXOPG06411 [ 60.141014] EIP: 0060:[<c1536d11>] EFLAGS: 00010046 CPU: 0 [ 60.141014] EIP is at __mutex_unlock_slowpath+0x5a/0x111 [ 60.141014] EAX: 00000100 EBX: 6b6b6b9b ECX: e9b4a1b0 EDX: e4a4e580 [ 60.141014] ESI: db162558 EDI: 00000246 EBP: e480be50 ESP: e480be44 [ 60.141014] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 [ 60.141014] Process meego-ux-daemon (pid: 771, ti=e480a000 task=e9b4a1b0 task.ti=e480a000) [ 60.141014] Stack: [ 60.141014] e4a4e580 db162558 f5a2f838 e480be58 c1536dd0 e480be68 c125ab1b db162558 [ 60.141014] db1624e0 e480be78 c10ba071 db162558 f760241c e480be94 c10bb0bc 000155fe [ 60.141014] f760241c f5a2f838 f5a2f8c8 00000000 e480bea4 c1037c24 00000000 f5a2f838 [ 60.141014] Call Trace: [ 60.141014] [<c1536dd0>] ? mutex_unlock+0x8/0xa [ 60.141014] [<c125ab1b>] ? drm_gem_vm_close+0x39/0x3d [ 60.141014] [<c10ba071>] ? remove_vma+0x2d/0x58 [ 60.141014] [<c10bb0bc>] ? exit_mmap+0x126/0x13f [ 60.141014] [<c1037c24>] ? mmput+0x37/0x9a [ 60.141014] [<c10d450d>] ? exec_mmap+0x178/0x19c [ 60.141014] [<c1537f85>] ? _raw_spin_unlock+0x1d/0x36 [ 60.141014] [<c10d4eb0>] ? flush_old_exec+0x42/0x75 [ 60.141014] [<c1104442>] ? load_elf_binary+0x32a/0x922 [ 60.141014] [<c10d3f76>] ? search_binary_handler+0x200/0x2ea [ 60.141014] [<c10d3ecf>] ? search_binary_handler+0x159/0x2ea [ 60.141014] [<c1104118>] ? load_elf_binary+0x0/0x922 [ 60.141014] [<c10d56b2>] ? do_execve+0x1ff/0x2e6 [ 60.141014] [<c100970e>] ? sys_execve+0x2d/0x55 [ 60.141014] [<c1002a5a>] ? ptregs_execve+0x12/0x18 [ 60.141014] [<c10029dc>] ? sysenter_do_call+0x12/0x3c [ 60.141014] [<c1530000>] ? init_centaur+0x9c/0x1ba [ 60.141014] Code: c1 00 75 0f ba 38 01 00 00 b8 8c 3a 6c c1 e8 cc 2e b0 ff 9c 58 8d 74 26 00 89 c7 fa 90 8d 74 26 00 e8 d2 b4 b2 ff b8 00 01 00 00 <f0> 66 0f c1 43 04 38 e0 74 07 f3 90 8a 43 04 eb f5 83 3d 64 ef [ 60.141014] EIP: [<c1536d11>] __mutex_unlock_slowpath+0x5a/0x111 SS:ESP 0068:e480be44 [ 60.141014] CR2: 000000006b6b6b9f Reported-by: Rusty Lynch <rusty.lynch@intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-17 16:33:33 -06:00
mutex_unlock(&dev->struct_mutex);
}
EXPORT_SYMBOL(drm_gem_vm_close);
/**
* drm_gem_mmap_obj - memory map a GEM object
* @obj: the GEM object to map
* @obj_size: the object size to be mapped, in bytes
* @vma: VMA for the area to be mapped
*
* Set up the VMA to prepare mapping of the GEM object using the gem_vm_ops
* provided by the driver. Depending on their requirements, drivers can either
* provide a fault handler in their gem_vm_ops (in which case any accesses to
* the object will be trapped, to perform migration, GTT binding, surface
* register allocation, or performance monitoring), or mmap the buffer memory
* synchronously after calling drm_gem_mmap_obj.
*
* This function is mainly intended to implement the DMABUF mmap operation, when
* the GEM object is not looked up based on its fake offset. To implement the
* DRM mmap operation, drivers should use the drm_gem_mmap() function.
*
* NOTE: This function has to be protected with dev->struct_mutex
*
* Return 0 or success or -EINVAL if the object size is smaller than the VMA
* size, or if no gem_vm_ops are provided.
*/
int drm_gem_mmap_obj(struct drm_gem_object *obj, unsigned long obj_size,
struct vm_area_struct *vma)
{
struct drm_device *dev = obj->dev;
lockdep_assert_held(&dev->struct_mutex);
/* Check for valid size. */
if (obj_size < vma->vm_end - vma->vm_start)
return -EINVAL;
if (!dev->driver->gem_vm_ops)
return -EINVAL;
vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
vma->vm_ops = dev->driver->gem_vm_ops;
vma->vm_private_data = obj;
vma->vm_page_prot = pgprot_writecombine(vm_get_page_prot(vma->vm_flags));
/* Take a ref for this mapping of the object, so that the fault
* handler can dereference the mmap offset's pointer to the object.
* This reference is cleaned up by the corresponding vm_close
* (which should happen whether the vma was created by this call, or
* by a vm_open due to mremap or partial unmap or whatever).
*/
drm_gem_object_reference(obj);
drm_vm_open_locked(dev, vma);
return 0;
}
EXPORT_SYMBOL(drm_gem_mmap_obj);
/**
* drm_gem_mmap - memory map routine for GEM objects
* @filp: DRM file pointer
* @vma: VMA for the area to be mapped
*
* If a driver supports GEM object mapping, mmap calls on the DRM file
* descriptor will end up here.
*
* Look up the GEM object based on the offset passed in (vma->vm_pgoff will
* contain the fake offset we created when the GTT map ioctl was called on
* the object) and map it with a call to drm_gem_mmap_obj().
*/
int drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
{
struct drm_file *priv = filp->private_data;
struct drm_device *dev = priv->minor->dev;
struct drm_gem_mm *mm = dev->mm_private;
struct drm_gem_object *obj;
struct drm_vma_offset_node *node;
int ret = 0;
if (drm_device_is_unplugged(dev))
return -ENODEV;
mutex_lock(&dev->struct_mutex);
node = drm_vma_offset_exact_lookup(&mm->vma_manager, vma->vm_pgoff,
vma_pages(vma));
if (!node) {
mutex_unlock(&dev->struct_mutex);
return drm_mmap(filp, vma);
}
obj = container_of(node, struct drm_gem_object, vma_node);
ret = drm_gem_mmap_obj(obj, drm_vma_node_size(node) << PAGE_SHIFT, vma);
mutex_unlock(&dev->struct_mutex);
return ret;
}
EXPORT_SYMBOL(drm_gem_mmap);