Skip to contentSkip to conversation

videobuf2: Add missing lock held on vb2_fop_relase

PATCH Superseded Ricardo Ribalda Delgado 3 participants 9 messages Nov 2, 2013 – Nov 6, 2013

Subsystems (3)

v4 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

[ 308.297741] BUG: unable to handle kernel NULL pointer dereference at
0000000000000260
[ 308.297759] IP: [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.297794] PGD 159719067 PUD 158119067 PMD 0
[ 308.297812] Oops: 0000 #1 SMP
[ 308.297826] Modules linked in: qt5023_video videobuf2_dma_sg
qtec_xform videobuf2_vmalloc videobuf2_memops videobuf2_core
qtec_white qtec_mem gpio_xilinx qtec_cmosis qtec_pcie fglrx(PO)
spi_xilinx spi_bitbang qt5023
[ 308.297888] CPU: 1 PID: 2189 Comm: java Tainted: P O 3.11.0-qtec-standard #1
[ 308.297919] Hardware name: QTechnology QT5022/QT5022, BIOS
PM_2.1.0.309 X64 05/23/2013
[ 308.297952] task: ffff8801564e1690 ti: ffff88014dc02000 task.ti:
ffff88014dc02000
[ 308.297962] RIP: 0010:[<ffffffffa07a9fd2>] [<ffffffffa07a9fd2>]
vb2_perform_fileio+0x372/0x610 [videobuf2_core]
[ 308.297985] RSP: 0018:ffff88014dc03df8 EFLAGS: 00010202
[ 308.297995] RAX: 0000000000000000 RBX: ffff880158a23000 RCX: dead000000100100
[ 308.298003] RDX: 0000000000000000 RSI: dead000000200200 RDI: 0000000000000000
[ 308.298012] RBP: ffff88014dc03e58 R08: 0000000000000000 R09: 0000000000000001
[ 308.298020] R10: ffffea00051e8380 R11: ffff88014dc03fd8 R12: ffff880158a23070
[ 308.298029] R13: ffff8801549040b8 R14: 0000000000198000 R15: 0000000001887e60
[ 308.298040] FS: 00007f65130d5700(0000) GS:ffff88015ed00000(0000)
knlGS:0000000000000000
[ 308.298049] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 308.298057] CR2: 0000000000000260 CR3: 0000000159630000 CR4: 00000000000007e0
[ 308.298064] Stack:
[ 308.298071] ffff880156416c00 0000000000198000 0000000000000000
ffff880100000001
[ 308.298087] ffff88014dc03f50 00000000810a79ca 0002000000000001
ffff880154904718
[ 308.298101] ffff880156416c00 0000000000198000 ffff880154904338
ffff88014dc03f50
[ 308.298116] Call Trace:
[ 308.298143] [<ffffffffa07aa3c4>] vb2_read+0x14/0x20 [videobuf2_core]
[ 308.298198] [<ffffffffa07aa494>] vb2_fop_read+0xc4/0x120 [videobuf2_core]
[ 308.298252] [<ffffffff8154ee9e>] v4l2_read+0x7e/0xc0
[ 308.298296] [<ffffffff8116e639>] vfs_read+0xa9/0x160
[ 308.298312] [<ffffffff8116e882>] SyS_read+0x52/0xb0
[ 308.298328] [<ffffffff81784179>] tracesys+0xd0/0xd5
[ 308.298335] Code: e5 d6 ff ff 83 3d be 24 00 00 04 89 c2 4c 8b 45 b0
44 8b 4d b8 0f 8f 20 02 00 00 85 d2 75 32 83 83 78 03 00 00 01 4b 8b
44 c5 48 <8b> 88 60 02 00 00 85 c9 0f 84 b0 00 00 00 8b 40 58 89 c2 41
89
[ 308.298487] RIP [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.298507] RSP <ffff88014dc03df8>
[ 308.298514] CR2: 0000000000000260
[ 308.298526] ---[ end trace e8f01717c96d1e41 ]---

---
v2: Comments by Sylvester Nawrocki

fimc-capture and fimc-lite where calling vb2_fop_release with the lock held. Therefore a new __vb2_fop_release function has been created to be used by drivers that overload the release function.

v3: Comments by Sylvester Nawrocki and Mauro Carvalho Chehab

Use vb2_fop_release_locked instead of __vb2_fop_release

v4: Comments by Sylvester Nawrocki

Rename vb2_fop_release_locked to __vb2_fop_release and fix patch format

 drivers/media/platform/exynos4-is/fimc-capture.c |  2 +-
 drivers/media/platform/exynos4-is/fimc-lite.c    |  2 +-
 drivers/media/v4l2-core/videobuf2-core.c         | 23 ++++++++++++++++++++++-
 include/media/videobuf2-core.h                   |  1 +
 4 files changed, 25 insertions(+), 3 deletions(-)
drivers/media/platform/exynos4-is/fimc-capture.c +1 -1
index fb27ff7..8192fe0 100644
@@ -549,7 +549,7 @@ static int fimc_capture_release(struct file *file)
 		vc->streaming = false;
 	}
 
-	ret = vb2_fop_release(file);
+	ret = __vb2_fop_release(file);
 
 	if (close) {
 		clear_bit(ST_CAPT_BUSY, &fimc->state);
drivers/media/platform/exynos4-is/fimc-lite.c +1 -1
index e5798f7..cbe51cd 100644
@@ -546,7 +546,7 @@ static int fimc_lite_release(struct file *file)
 		mutex_unlock(&entity->parent->graph_mutex);
 	}
 
-	vb2_fop_release(file);
+	__vb2_fop_release(file);
 	pm_runtime_put(&fimc->pdev->dev);
 	clear_bit(ST_FLITE_SUSPENDED, &fimc->state);
 
drivers/media/v4l2-core/videobuf2-core.c +22 -1
index 594c75e..f48d72a 100644
@@ -2619,18 +2619,39 @@ int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma)
 }
 EXPORT_SYMBOL_GPL(vb2_fop_mmap);
 
-int vb2_fop_release(struct file *file)
+static int _vb2_fop_release(struct file *file, bool lock_is_held)
 {
 	struct video_device *vdev = video_devdata(file);
+	struct mutex *lock;
 
 	if (file->private_data == vdev->queue->owner) {
+		if (lock_is_held)
+			lock = NULL;
+		else
+			lock = vdev->queue->lock ?
+				vdev->queue->lock : vdev->lock;
+		if (lock)
+			mutex_lock(lock);
 		vb2_queue_release(vdev->queue);
 		vdev->queue->owner = NULL;
+		if (lock)
+			mutex_unlock(lock);
 	}
 	return v4l2_fh_release(file);
 }
+
+int vb2_fop_release(struct file *file)
+{
+	return _vb2_fop_release(file, false);
+}
 EXPORT_SYMBOL_GPL(vb2_fop_release);
 
+int __vb2_fop_release(struct file *file)
+{
+	return _vb2_fop_release(file, true);
+}
+EXPORT_SYMBOL_GPL(__vb2_fop_release);
+
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
 		size_t count, loff_t *ppos)
 {
include/media/videobuf2-core.h +1 -0
index 6781258..76400fa 100644
@@ -491,6 +491,7 @@ int vb2_ioctl_expbuf(struct file *file, void *priv,
 
 int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma);
 int vb2_fop_release(struct file *file);
+int __vb2_fop_release(struct file *file);
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
 		size_t count, loff_t *ppos);
 ssize_t vb2_fop_read(struct file *file, char __user *buf,
Signature · 1 line
1.8.4.rc3
#1 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Quoted 8 lines · Ricardo Ribalda Delgado

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

Quoted 131 lines
[ 308.297741] BUG: unable to handle kernel NULL pointer dereference at
0000000000000260
[ 308.297759] IP: [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.297794] PGD 159719067 PUD 158119067 PMD 0
[ 308.297812] Oops: 0000 #1 SMP
[ 308.297826] Modules linked in: qt5023_video videobuf2_dma_sg
qtec_xform videobuf2_vmalloc videobuf2_memops videobuf2_core
qtec_white qtec_mem gpio_xilinx qtec_cmosis qtec_pcie fglrx(PO)
spi_xilinx spi_bitbang qt5023
[ 308.297888] CPU: 1 PID: 2189 Comm: java Tainted: P O 3.11.0-qtec-standard #1
[ 308.297919] Hardware name: QTechnology QT5022/QT5022, BIOS
PM_2.1.0.309 X64 05/23/2013
[ 308.297952] task: ffff8801564e1690 ti: ffff88014dc02000 task.ti:
ffff88014dc02000
[ 308.297962] RIP: 0010:[<ffffffffa07a9fd2>] [<ffffffffa07a9fd2>]
vb2_perform_fileio+0x372/0x610 [videobuf2_core]
[ 308.297985] RSP: 0018:ffff88014dc03df8 EFLAGS: 00010202
[ 308.297995] RAX: 0000000000000000 RBX: ffff880158a23000 RCX: dead000000100100
[ 308.298003] RDX: 0000000000000000 RSI: dead000000200200 RDI: 0000000000000000
[ 308.298012] RBP: ffff88014dc03e58 R08: 0000000000000000 R09: 0000000000000001
[ 308.298020] R10: ffffea00051e8380 R11: ffff88014dc03fd8 R12: ffff880158a23070
[ 308.298029] R13: ffff8801549040b8 R14: 0000000000198000 R15: 0000000001887e60
[ 308.298040] FS: 00007f65130d5700(0000) GS:ffff88015ed00000(0000)
knlGS:0000000000000000
[ 308.298049] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 308.298057] CR2: 0000000000000260 CR3: 0000000159630000 CR4: 00000000000007e0
[ 308.298064] Stack:
[ 308.298071] ffff880156416c00 0000000000198000 0000000000000000
ffff880100000001
[ 308.298087] ffff88014dc03f50 00000000810a79ca 0002000000000001
ffff880154904718
[ 308.298101] ffff880156416c00 0000000000198000 ffff880154904338
ffff88014dc03f50
[ 308.298116] Call Trace:
[ 308.298143] [<ffffffffa07aa3c4>] vb2_read+0x14/0x20 [videobuf2_core]
[ 308.298198] [<ffffffffa07aa494>] vb2_fop_read+0xc4/0x120 [videobuf2_core]
[ 308.298252] [<ffffffff8154ee9e>] v4l2_read+0x7e/0xc0
[ 308.298296] [<ffffffff8116e639>] vfs_read+0xa9/0x160
[ 308.298312] [<ffffffff8116e882>] SyS_read+0x52/0xb0
[ 308.298328] [<ffffffff81784179>] tracesys+0xd0/0xd5
[ 308.298335] Code: e5 d6 ff ff 83 3d be 24 00 00 04 89 c2 4c 8b 45 b0
44 8b 4d b8 0f 8f 20 02 00 00 85 d2 75 32 83 83 78 03 00 00 01 4b 8b
44 c5 48 <8b> 88 60 02 00 00 85 c9 0f 84 b0 00 00 00 8b 40 58 89 c2 41
89
[ 308.298487] RIP [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.298507] RSP <ffff88014dc03df8>
[ 308.298514] CR2: 0000000000000260
[ 308.298526] ---[ end trace e8f01717c96d1e41 ]---

---
v2: Comments by Sylvester Nawrocki

fimc-capture and fimc-lite where calling vb2_fop_release with the lock held. Therefore a new __vb2_fop_release function has been created to be used by drivers that overload the release function.

v3: Comments by Sylvester Nawrocki and Mauro Carvalho Chehab

Use vb2_fop_release_locked instead of __vb2_fop_release

v4: Comments by Sylvester Nawrocki

Rename vb2_fop_release_locked to __vb2_fop_release and fix patch format

 drivers/media/platform/exynos4-is/fimc-capture.c |  2 +-
 drivers/media/platform/exynos4-is/fimc-lite.c    |  2 +-
 drivers/media/v4l2-core/videobuf2-core.c         | 23 ++++++++++++++++++++++-
 include/media/videobuf2-core.h                   |  1 +
 4 files changed, 25 insertions(+), 3 deletions(-)
drivers/media/platform/exynos4-is/fimc-capture.c +1 -1
index fb27ff7..8192fe0 100644
@@ -549,7 +549,7 @@ static int fimc_capture_release(struct file *file)
 		vc->streaming = false;
 	}
 
-	ret = vb2_fop_release(file);
+	ret = __vb2_fop_release(file);
 
 	if (close) {
 		clear_bit(ST_CAPT_BUSY, &fimc->state);
drivers/media/platform/exynos4-is/fimc-lite.c +1 -1
index e5798f7..cbe51cd 100644
@@ -546,7 +546,7 @@ static int fimc_lite_release(struct file *file)
 		mutex_unlock(&entity->parent->graph_mutex);
 	}
 
-	vb2_fop_release(file);
+	__vb2_fop_release(file);
 	pm_runtime_put(&fimc->pdev->dev);
 	clear_bit(ST_FLITE_SUSPENDED, &fimc->state);
 
drivers/media/v4l2-core/videobuf2-core.c +13 -1
index 594c75e..f48d72a 100644
@@ -2619,18 +2619,39 @@ int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma)
 }
 EXPORT_SYMBOL_GPL(vb2_fop_mmap);
 
-int vb2_fop_release(struct file *file)
+static int _vb2_fop_release(struct file *file, bool lock_is_held)
 {
 	struct video_device *vdev = video_devdata(file);
+	struct mutex *lock;
 
 	if (file->private_data == vdev->queue->owner) {
+		if (lock_is_held)
+			lock = NULL;
+		else
+			lock = vdev->queue->lock ?
+				vdev->queue->lock : vdev->lock;
+		if (lock)
+			mutex_lock(lock);
 		vb2_queue_release(vdev->queue);
 		vdev->queue->owner = NULL;
+		if (lock)
+			mutex_unlock(lock);
 	}
 	return v4l2_fh_release(file);
 }
+
+int vb2_fop_release(struct file *file)

drivers/media/v4l2-core/videobuf2-core.c int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma) · line 2653

Quoted 10 lines
drivers/media/v4l2-core/videobuf2-core.c +8 -0
+{
+	return _vb2_fop_release(file, false);
+}
 EXPORT_SYMBOL_GPL(vb2_fop_release);
 
+int __vb2_fop_release(struct file *file)
+{
+	return _vb2_fop_release(file, true);
+}
+EXPORT_SYMBOL_GPL(__vb2_fop_release);

Sorry for introducing yet another opinion, but I think this is very confusing.

I would do this:

static int __vb2_fop_release(struct file *file, struct mutex *lock)

{       
        struct video_device *vdev = video_devdata(file);
        
        if (file->private_data == vdev->queue->owner) {
		if (lock)
			mutex_lock(lock);
                vb2_queue_release(vdev->queue);
                vdev->queue->owner = NULL;
		if (lock)
			mutex_unlock(lock);
        }
        return v4l2_fh_release(file); 
}       

int vb2_fop_release(struct file *file)

{
	struct video_device *vdev = video_devdata(file);
	struct mutex *lock = vdev->queue->lock ?
				vdev->queue->lock : vdev->lock;

	return __vb2_fop_release(file, lock);
}

EXPORT_SYMBOL_GPL(vb2_fop_release);

int vb2_fop_release_unlock(struct file *file)

{
	return __vb2_fop_release(file, NULL);
}

EXPORT_SYMBOL_GPL(vb2_fop_release_unlock);

Optionally, __vb2_fop_release can be exported and then vb2_fop_release_unlock isn't necessary.

Regards,

Hans

Quoted 17 lines
drivers/media/v4l2-core/videobuf2-core.c +1 -0
+
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
 		size_t count, loff_t *ppos)
 {
include/media/videobuf2-core.h +1 -0
index 6781258..76400fa 100644
@@ -491,6 +491,7 @@ int vb2_ioctl_expbuf(struct file *file, void *priv,
 
 int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma);
 int vb2_fop_release(struct file *file);
+int __vb2_fop_release(struct file *file);
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
 		size_t count, loff_t *ppos);
 ssize_t vb2_fop_read(struct file *file, char __user *buf,
#2 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 12 lines · Hans Verkuil
Quoted 8 lines · Ricardo Ribalda Delgado

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

Quoted 136 lines
Quoted 136 lines
[ 308.297741] BUG: unable to handle kernel NULL pointer dereference at
0000000000000260
[ 308.297759] IP: [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.297794] PGD 159719067 PUD 158119067 PMD 0
[ 308.297812] Oops: 0000 #1 SMP
[ 308.297826] Modules linked in: qt5023_video videobuf2_dma_sg
qtec_xform videobuf2_vmalloc videobuf2_memops videobuf2_core
qtec_white qtec_mem gpio_xilinx qtec_cmosis qtec_pcie fglrx(PO)
spi_xilinx spi_bitbang qt5023
[ 308.297888] CPU: 1 PID: 2189 Comm: java Tainted: P O 3.11.0-qtec-standard #1
[ 308.297919] Hardware name: QTechnology QT5022/QT5022, BIOS
PM_2.1.0.309 X64 05/23/2013
[ 308.297952] task: ffff8801564e1690 ti: ffff88014dc02000 task.ti:
ffff88014dc02000
[ 308.297962] RIP: 0010:[<ffffffffa07a9fd2>] [<ffffffffa07a9fd2>]
vb2_perform_fileio+0x372/0x610 [videobuf2_core]
[ 308.297985] RSP: 0018:ffff88014dc03df8 EFLAGS: 00010202
[ 308.297995] RAX: 0000000000000000 RBX: ffff880158a23000 RCX: dead000000100100
[ 308.298003] RDX: 0000000000000000 RSI: dead000000200200 RDI: 0000000000000000
[ 308.298012] RBP: ffff88014dc03e58 R08: 0000000000000000 R09: 0000000000000001
[ 308.298020] R10: ffffea00051e8380 R11: ffff88014dc03fd8 R12: ffff880158a23070
[ 308.298029] R13: ffff8801549040b8 R14: 0000000000198000 R15: 0000000001887e60
[ 308.298040] FS: 00007f65130d5700(0000) GS:ffff88015ed00000(0000)
knlGS:0000000000000000
[ 308.298049] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 308.298057] CR2: 0000000000000260 CR3: 0000000159630000 CR4: 00000000000007e0
[ 308.298064] Stack:
[ 308.298071] ffff880156416c00 0000000000198000 0000000000000000
ffff880100000001
[ 308.298087] ffff88014dc03f50 00000000810a79ca 0002000000000001
ffff880154904718
[ 308.298101] ffff880156416c00 0000000000198000 ffff880154904338
ffff88014dc03f50
[ 308.298116] Call Trace:
[ 308.298143] [<ffffffffa07aa3c4>] vb2_read+0x14/0x20 [videobuf2_core]
[ 308.298198] [<ffffffffa07aa494>] vb2_fop_read+0xc4/0x120 [videobuf2_core]
[ 308.298252] [<ffffffff8154ee9e>] v4l2_read+0x7e/0xc0
[ 308.298296] [<ffffffff8116e639>] vfs_read+0xa9/0x160
[ 308.298312] [<ffffffff8116e882>] SyS_read+0x52/0xb0
[ 308.298328] [<ffffffff81784179>] tracesys+0xd0/0xd5
[ 308.298335] Code: e5 d6 ff ff 83 3d be 24 00 00 04 89 c2 4c 8b 45 b0
44 8b 4d b8 0f 8f 20 02 00 00 85 d2 75 32 83 83 78 03 00 00 01 4b 8b
44 c5 48 <8b> 88 60 02 00 00 85 c9 0f 84 b0 00 00 00 8b 40 58 89 c2 41
89
[ 308.298487] RIP [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.298507] RSP <ffff88014dc03df8>
[ 308.298514] CR2: 0000000000000260
[ 308.298526] ---[ end trace e8f01717c96d1e41 ]---

---
v2: Comments by Sylvester Nawrocki

fimc-capture and fimc-lite where calling vb2_fop_release with the lock held. Therefore a new __vb2_fop_release function has been created to be used by drivers that overload the release function.

v3: Comments by Sylvester Nawrocki and Mauro Carvalho Chehab

Use vb2_fop_release_locked instead of __vb2_fop_release

v4: Comments by Sylvester Nawrocki

Rename vb2_fop_release_locked to __vb2_fop_release and fix patch format

 drivers/media/platform/exynos4-is/fimc-capture.c |  2 +-
 drivers/media/platform/exynos4-is/fimc-lite.c    |  2 +-
 drivers/media/v4l2-core/videobuf2-core.c         | 23 ++++++++++++++++++++++-
 include/media/videobuf2-core.h                   |  1 +
 4 files changed, 25 insertions(+), 3 deletions(-)
drivers/media/platform/exynos4-is/fimc-capture.c +1 -1
index fb27ff7..8192fe0 100644
@@ -549,7 +549,7 @@ static int fimc_capture_release(struct file *file)
              vc->streaming = false;
      }
 
-     ret = vb2_fop_release(file);
+     ret = __vb2_fop_release(file);
 
      if (close) {
              clear_bit(ST_CAPT_BUSY, &fimc->state);
drivers/media/platform/exynos4-is/fimc-lite.c +1 -1
index e5798f7..cbe51cd 100644
@@ -546,7 +546,7 @@ static int fimc_lite_release(struct file *file)
              mutex_unlock(&entity->parent->graph_mutex);
      }
 
-     vb2_fop_release(file);
+     __vb2_fop_release(file);
      pm_runtime_put(&fimc->pdev->dev);
      clear_bit(ST_FLITE_SUSPENDED, &fimc->state);
 
drivers/media/v4l2-core/videobuf2-core.c +16 -1
index 594c75e..f48d72a 100644
@@ -2619,18 +2619,39 @@ int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma)
 }
 EXPORT_SYMBOL_GPL(vb2_fop_mmap);
 
-int vb2_fop_release(struct file *file)
+static int _vb2_fop_release(struct file *file, bool lock_is_held)
 {
      struct video_device *vdev = video_devdata(file);
+     struct mutex *lock;
 
      if (file->private_data == vdev->queue->owner) {
+             if (lock_is_held)
+                     lock = NULL;
+             else
+                     lock = vdev->queue->lock ?
+                             vdev->queue->lock : vdev->lock;
+             if (lock)
+                     mutex_lock(lock);
              vb2_queue_release(vdev->queue);
              vdev->queue->owner = NULL;
+             if (lock)
+                     mutex_unlock(lock);
      }
      return v4l2_fh_release(file);
 }
+
+int vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, false);
+}
 EXPORT_SYMBOL_GPL(vb2_fop_release);
 
Quoted 7 lines

drivers/media/v4l2-core/videobuf2-core.c int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma) · line 2653

Quoted 5 lines
drivers/media/v4l2-core/videobuf2-core.c +5 -0
+int __vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, true);
+}
+EXPORT_SYMBOL_GPL(__vb2_fop_release);

Sorry for introducing yet another opinion, but I think this is very confusing.

It is confusing the lock_held parameter or the __ naming for unlocked versions?

Quoted 28 lines

I would do this:

static int __vb2_fop_release(struct file *file, struct mutex *lock)

{
        struct video_device *vdev = video_devdata(file);

        if (file->private_data == vdev->queue->owner) {
                if (lock)
                        mutex_lock(lock);
                vb2_queue_release(vdev->queue);
                vdev->queue->owner = NULL;
                if (lock)
                        mutex_unlock(lock);
        }
        return v4l2_fh_release(file);
}

int vb2_fop_release(struct file *file)

{
        struct video_device *vdev = video_devdata(file);
        struct mutex *lock = vdev->queue->lock ?
                                vdev->queue->lock : vdev->lock;

        return __vb2_fop_release(file, lock);
}

EXPORT_SYMBOL_GPL(vb2_fop_release);

Quoted 9 lines

int vb2_fop_release_unlock(struct file *file)

{
        return __vb2_fop_release(file, NULL);
}

EXPORT_SYMBOL_GPL(vb2_fop_release_unlock);

Optionally, __vb2_fop_release can be exported and then vb2_fop_release_unlock isn't necessary.

i dont have any strong opinion in any direction. All I really care is that the oops is fixed :).

If your concern about the patch is the is_lock_held function, I can make a patch with the params on your proposal and the __naming as Sylvester suggested, so everyone is happy.

Sylvester, Hanshat do you think?

Thanks for your comments!

Quoted 22 lines

Regards,

Hans

Quoted 17 lines
drivers/media/v4l2-core/videobuf2-core.c +1 -0
+
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
              size_t count, loff_t *ppos)
 {
include/media/videobuf2-core.h +1 -0
index 6781258..76400fa 100644
@@ -491,6 +491,7 @@ int vb2_ioctl_expbuf(struct file *file, void *priv,
 
 int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma);
 int vb2_fop_release(struct file *file);
+int __vb2_fop_release(struct file *file);
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
              size_t count, loff_t *ppos);
 ssize_t vb2_fop_read(struct file *file, char __user *buf,
Signature · 1 line
Ricardo Ribalda
#3 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Hi,

Quoted 6 lines · Ricardo Ribalda Delgado

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

We're discussing v4 actually ;)

Quoted 18 lines · Hans Verkuil
Quoted 12 lines · Ricardo Ribalda Delgado
Quoted 8 lines

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

[...]

Quoted 76 lines
Quoted 28 lines
Quoted 28 lines
drivers/media/v4l2-core/videobuf2-core.c +16 -1
-int vb2_fop_release(struct file *file)
+static int _vb2_fop_release(struct file *file, bool lock_is_held)
 {
      struct video_device *vdev = video_devdata(file);
+     struct mutex *lock;
 
      if (file->private_data == vdev->queue->owner) {
+             if (lock_is_held)
+                     lock = NULL;
+             else
+                     lock = vdev->queue->lock ?
+                             vdev->queue->lock : vdev->lock;
+             if (lock)
+                     mutex_lock(lock);
              vb2_queue_release(vdev->queue);
              vdev->queue->owner = NULL;
+             if (lock)
+                     mutex_unlock(lock);
      }
      return v4l2_fh_release(file);
 }
+
+int vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, false);
+}
 EXPORT_SYMBOL_GPL(vb2_fop_release);
Quoted 7 lines

drivers/media/v4l2-core/videobuf2-core.c int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma) · line 2653

Quoted 5 lines
drivers/media/v4l2-core/videobuf2-core.c +5 -0
+int __vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, true);
+}
+EXPORT_SYMBOL_GPL(__vb2_fop_release);

Sorry for introducing yet another opinion, but I think this is very confusing.

It is confusing the lock_held parameter or the __ naming for unlocked versions?

Quoted 28 lines

I would do this:

static int __vb2_fop_release(struct file *file, struct mutex *lock)

{
        struct video_device *vdev = video_devdata(file);

        if (file->private_data == vdev->queue->owner) {
                if (lock)
                        mutex_lock(lock);
                vb2_queue_release(vdev->queue);
                vdev->queue->owner = NULL;
                if (lock)
                        mutex_unlock(lock);
        }
        return v4l2_fh_release(file);
}

int vb2_fop_release(struct file *file)

{
        struct video_device *vdev = video_devdata(file);
        struct mutex *lock = vdev->queue->lock ?
                                vdev->queue->lock : vdev->lock;

        return __vb2_fop_release(file, lock);
}

EXPORT_SYMBOL_GPL(vb2_fop_release);

Quoted 9 lines

int vb2_fop_release_unlock(struct file *file)

{
        return __vb2_fop_release(file, NULL);
}

EXPORT_SYMBOL_GPL(vb2_fop_release_unlock);

Optionally, __vb2_fop_release can be exported and then vb2_fop_release_unlock isn't necessary.

Quoted 8 lines

i dont have any strong opinion in any direction. All I really care is that the oops is fixed :).

If your concern about the patch is the is_lock_held function, I can make a patch with the params on your proposal and the __naming as Sylvester suggested, so everyone is happy.

Sylvester, Hanshat do you think?

I like Hans' proposal, probably it's better to export __vb2_fop_release(), so it can be also used in cases like em28xx, to make things a bit more clear. But I'm fine with vb2_fop_release_unlock() as well, don't really have strong preference. It's up to you.

--
Thanks,
Sylwester

#4 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

In reply to Ricardo Ribalda Delgado
Quoted 18 lines · Ricardo Ribalda Delgado

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 10 lines · Hans Verkuil
Quoted 8 lines · Ricardo Ribalda Delgado

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

Quoted 7 lines
Quoted 2 lines

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

static int em28xx_v4l2_close(struct file *filp)

{
        struct em28xx_fh *fh  = filp->private_data;
        struct em28xx    *dev = fh->dev;
        int              errCode;

        em28xx_videodbg("users=%d\n", dev->users);

        mutex_lock(&dev->lock);
        vb2_fop_release(filp);
	...

vb2_fop_release(filp) will, with your patch, also try to get dev->lock.

Sylwester's comment re em28xx is incorrect.

Quoted 145 lines
Quoted 137 lines
Quoted 137 lines

[ 308.297741] BUG: unable to handle kernel NULL pointer dereference at
0000000000000260
[ 308.297759] IP: [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.297794] PGD 159719067 PUD 158119067 PMD 0
[ 308.297812] Oops: 0000 #1 SMP
[ 308.297826] Modules linked in: qt5023_video videobuf2_dma_sg
qtec_xform videobuf2_vmalloc videobuf2_memops videobuf2_core
qtec_white qtec_mem gpio_xilinx qtec_cmosis qtec_pcie fglrx(PO)
spi_xilinx spi_bitbang qt5023
[ 308.297888] CPU: 1 PID: 2189 Comm: java Tainted: P O 3.11.0-qtec-standard #1
[ 308.297919] Hardware name: QTechnology QT5022/QT5022, BIOS
PM_2.1.0.309 X64 05/23/2013
[ 308.297952] task: ffff8801564e1690 ti: ffff88014dc02000 task.ti:
ffff88014dc02000
[ 308.297962] RIP: 0010:[<ffffffffa07a9fd2>] [<ffffffffa07a9fd2>]
vb2_perform_fileio+0x372/0x610 [videobuf2_core]
[ 308.297985] RSP: 0018:ffff88014dc03df8 EFLAGS: 00010202
[ 308.297995] RAX: 0000000000000000 RBX: ffff880158a23000 RCX: dead000000100100
[ 308.298003] RDX: 0000000000000000 RSI: dead000000200200 RDI: 0000000000000000
[ 308.298012] RBP: ffff88014dc03e58 R08: 0000000000000000 R09: 0000000000000001
[ 308.298020] R10: ffffea00051e8380 R11: ffff88014dc03fd8 R12: ffff880158a23070
[ 308.298029] R13: ffff8801549040b8 R14: 0000000000198000 R15: 0000000001887e60
[ 308.298040] FS: 00007f65130d5700(0000) GS:ffff88015ed00000(0000)
knlGS:0000000000000000
[ 308.298049] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 308.298057] CR2: 0000000000000260 CR3: 0000000159630000 CR4: 00000000000007e0
[ 308.298064] Stack:
[ 308.298071] ffff880156416c00 0000000000198000 0000000000000000
ffff880100000001
[ 308.298087] ffff88014dc03f50 00000000810a79ca 0002000000000001
ffff880154904718
[ 308.298101] ffff880156416c00 0000000000198000 ffff880154904338
ffff88014dc03f50
[ 308.298116] Call Trace:
[ 308.298143] [<ffffffffa07aa3c4>] vb2_read+0x14/0x20 [videobuf2_core]
[ 308.298198] [<ffffffffa07aa494>] vb2_fop_read+0xc4/0x120 [videobuf2_core]
[ 308.298252] [<ffffffff8154ee9e>] v4l2_read+0x7e/0xc0
[ 308.298296] [<ffffffff8116e639>] vfs_read+0xa9/0x160
[ 308.298312] [<ffffffff8116e882>] SyS_read+0x52/0xb0
[ 308.298328] [<ffffffff81784179>] tracesys+0xd0/0xd5
[ 308.298335] Code: e5 d6 ff ff 83 3d be 24 00 00 04 89 c2 4c 8b 45 b0
44 8b 4d b8 0f 8f 20 02 00 00 85 d2 75 32 83 83 78 03 00 00 01 4b 8b
44 c5 48 <8b> 88 60 02 00 00 85 c9 0f 84 b0 00 00 00 8b 40 58 89 c2 41
89
[ 308.298487] RIP [<ffffffffa07a9fd2>] vb2_perform_fileio+0x372/0x610
[videobuf2_core]
[ 308.298507] RSP <ffff88014dc03df8>
[ 308.298514] CR2: 0000000000000260
[ 308.298526] ---[ end trace e8f01717c96d1e41 ]---

---
v2: Comments by Sylvester Nawrocki

fimc-capture and fimc-lite where calling vb2_fop_release with the lock held. Therefore a new __vb2_fop_release function has been created to be used by drivers that overload the release function.

v3: Comments by Sylvester Nawrocki and Mauro Carvalho Chehab

Use vb2_fop_release_locked instead of __vb2_fop_release

v4: Comments by Sylvester Nawrocki

Rename vb2_fop_release_locked to __vb2_fop_release and fix patch format

 drivers/media/platform/exynos4-is/fimc-capture.c |  2 +-
 drivers/media/platform/exynos4-is/fimc-lite.c    |  2 +-
 drivers/media/v4l2-core/videobuf2-core.c         | 23 ++++++++++++++++++++++-
 include/media/videobuf2-core.h                   |  1 +
 4 files changed, 25 insertions(+), 3 deletions(-)
drivers/media/platform/exynos4-is/fimc-capture.c +1 -1
index fb27ff7..8192fe0 100644
@@ -549,7 +549,7 @@ static int fimc_capture_release(struct file *file)
              vc->streaming = false;
      }
 
-     ret = vb2_fop_release(file);
+     ret = __vb2_fop_release(file);
 
      if (close) {
              clear_bit(ST_CAPT_BUSY, &fimc->state);
drivers/media/platform/exynos4-is/fimc-lite.c +1 -1
index e5798f7..cbe51cd 100644
@@ -546,7 +546,7 @@ static int fimc_lite_release(struct file *file)
              mutex_unlock(&entity->parent->graph_mutex);
      }
 
-     vb2_fop_release(file);
+     __vb2_fop_release(file);
      pm_runtime_put(&fimc->pdev->dev);
      clear_bit(ST_FLITE_SUSPENDED, &fimc->state);
 
drivers/media/v4l2-core/videobuf2-core.c +16 -1
index 594c75e..f48d72a 100644
@@ -2619,18 +2619,39 @@ int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma)
 }
 EXPORT_SYMBOL_GPL(vb2_fop_mmap);
 
-int vb2_fop_release(struct file *file)
+static int _vb2_fop_release(struct file *file, bool lock_is_held)
 {
      struct video_device *vdev = video_devdata(file);
+     struct mutex *lock;
 
      if (file->private_data == vdev->queue->owner) {
+             if (lock_is_held)
+                     lock = NULL;
+             else
+                     lock = vdev->queue->lock ?
+                             vdev->queue->lock : vdev->lock;
+             if (lock)
+                     mutex_lock(lock);
              vb2_queue_release(vdev->queue);
              vdev->queue->owner = NULL;
+             if (lock)
+                     mutex_unlock(lock);
      }
      return v4l2_fh_release(file);
 }
+
+int vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, false);
+}
 EXPORT_SYMBOL_GPL(vb2_fop_release);
 
Quoted 7 lines

drivers/media/v4l2-core/videobuf2-core.c int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma) · line 2653

Quoted 5 lines
drivers/media/v4l2-core/videobuf2-core.c +5 -0
+int __vb2_fop_release(struct file *file)
+{
+     return _vb2_fop_release(file, true);
+}
+EXPORT_SYMBOL_GPL(__vb2_fop_release);

Sorry for introducing yet another opinion, but I think this is very confusing.

Quoted 1 line

It is confusing the lock_held parameter or the __ naming for unlocked versions?

The naming for the most part. Having a normal function, a '_' prefix and a '__' prefix is a bit over the top IMHO. I prefer to have a clearly named _unlock version.

Regards,

Hans

Quoted 75 lines
Quoted 29 lines

I would do this:

static int __vb2_fop_release(struct file *file, struct mutex *lock)

{
        struct video_device *vdev = video_devdata(file);

        if (file->private_data == vdev->queue->owner) {
                if (lock)
                        mutex_lock(lock);
                vb2_queue_release(vdev->queue);
                vdev->queue->owner = NULL;
                if (lock)
                        mutex_unlock(lock);
        }
        return v4l2_fh_release(file);
}

int vb2_fop_release(struct file *file)

{
        struct video_device *vdev = video_devdata(file);
        struct mutex *lock = vdev->queue->lock ?
                                vdev->queue->lock : vdev->lock;

        return __vb2_fop_release(file, lock);
}

EXPORT_SYMBOL_GPL(vb2_fop_release);

Quoted 9 lines

int vb2_fop_release_unlock(struct file *file)

{
        return __vb2_fop_release(file, NULL);
}

EXPORT_SYMBOL_GPL(vb2_fop_release_unlock);

Optionally, __vb2_fop_release can be exported and then vb2_fop_release_unlock isn't necessary.

i dont have any strong opinion in any direction. All I really care is that the oops is fixed :).

If your concern about the patch is the is_lock_held function, I can make a patch with the params on your proposal and the __naming as Sylvester suggested, so everyone is happy.

Sylvester, Hanshat do you think?

Thanks for your comments!

Quoted 22 lines

Regards,

Hans

Quoted 17 lines
drivers/media/v4l2-core/videobuf2-core.c +1 -0
+
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
              size_t count, loff_t *ppos)
 {
include/media/videobuf2-core.h +1 -0
index 6781258..76400fa 100644
@@ -491,6 +491,7 @@ int vb2_ioctl_expbuf(struct file *file, void *priv,
 
 int vb2_fop_mmap(struct file *file, struct vm_area_struct *vma);
 int vb2_fop_release(struct file *file);
+int __vb2_fop_release(struct file *file);
 ssize_t vb2_fop_write(struct file *file, char __user *buf,
              size_t count, loff_t *ppos);
 ssize_t vb2_fop_read(struct file *file, char __user *buf,
#5 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Quoted 39 lines · Hans Verkuil
Quoted 18 lines · Ricardo Ribalda Delgado
Quoted 18 lines

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 10 lines · Hans Verkuil
Quoted 10 lines · Ricardo Ribalda Delgado
Quoted 8 lines
Quoted 8 lines

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

static int em28xx_v4l2_close(struct file *filp)

{
        struct em28xx_fh *fh  = filp->private_data;
        struct em28xx    *dev = fh->dev;
        int              errCode;

        em28xx_videodbg("users=%d\n", dev->users);

        mutex_lock(&dev->lock);
        vb2_fop_release(filp);
	...
Quoted 3 lines

vb2_fop_release(filp) will, with your patch, also try to get dev->lock.

Sylwester's comment re em28xx is incorrect.

dev->lock is not used as the video queue lock:

$ git grep "lock =" drivers/media/usb/em28xx/
...
drivers/media/usb/em28xx/em28xx-video.c:        dev->vdev->queue->lock = &dev->vb_queue_lock;
drivers/media/usb/em28xx/em28xx-video.c:                dev->vbi_dev->queue->lock = &dev->vb_vbi_queue_lock;

There is a separate mutex for the video queue which needs to be acquired independently.

--
Regards,
Sylwester

#6 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Quoted 46 lines · Sylwester Nawrocki
Quoted 39 lines · Hans Verkuil
Quoted 18 lines · Ricardo Ribalda Delgado
Quoted 18 lines

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 10 lines · Hans Verkuil
Quoted 10 lines · Ricardo Ribalda Delgado
Quoted 8 lines
Quoted 8 lines

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

static int em28xx_v4l2_close(struct file *filp)

{
        struct em28xx_fh *fh  = filp->private_data;
        struct em28xx    *dev = fh->dev;
        int              errCode;

        em28xx_videodbg("users=%d\n", dev->users);

        mutex_lock(&dev->lock);
        vb2_fop_release(filp);
	...
Quoted 3 lines

vb2_fop_release(filp) will, with your patch, also try to get dev->lock.

Sylwester's comment re em28xx is incorrect.

dev->lock is not used as the video queue lock:

Quoted 7 lines
$ git grep "lock =" drivers/media/usb/em28xx/
...
drivers/media/usb/em28xx/em28xx-video.c:        dev->vdev->queue->lock = &dev->vb_queue_lock;
drivers/media/usb/em28xx/em28xx-video.c:                dev->vbi_dev->queue->lock = &dev->vb_vbi_queue_lock;

There is a separate mutex for the video queue which needs to be acquired independently.

Darn, I missed that one. I was looking for it in em28xx_vdev_init(), which is where I expected the queue->lock to be set, if there was any.

That said, wouldn't it be a good idea to swap the order:

        vb2_fop_release(filp);
        mutex_lock(&dev->lock);

I don't believe there is a good reason for nesting mutexes here.

Regards,

Hans

Quoted 9 lines

--
Regards,
Sylwester
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

#7 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Hello Hans and Sywester

I have just posted a new patch. I think it fits the suggestions from both of you, please take a look to it and please post any comment.

I will also send a patch about the em28xx, to swap the lock order.

Thanks for your comments

Quoted 79 lines · Hans Verkuil
Quoted 46 lines · Sylwester Nawrocki
Quoted 39 lines · Hans Verkuil
Quoted 18 lines · Ricardo Ribalda Delgado
Quoted 18 lines

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 10 lines · Hans Verkuil
Quoted 10 lines · Ricardo Ribalda Delgado
Quoted 8 lines
Quoted 8 lines

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

static int em28xx_v4l2_close(struct file *filp)

{
        struct em28xx_fh *fh  = filp->private_data;
        struct em28xx    *dev = fh->dev;
        int              errCode;

        em28xx_videodbg("users=%d\n", dev->users);

        mutex_lock(&dev->lock);
        vb2_fop_release(filp);
     ...
Quoted 3 lines

vb2_fop_release(filp) will, with your patch, also try to get dev->lock.

Sylwester's comment re em28xx is incorrect.

dev->lock is not used as the video queue lock:

Quoted 7 lines
$ git grep "lock =" drivers/media/usb/em28xx/
...
drivers/media/usb/em28xx/em28xx-video.c:        dev->vdev->queue->lock = &dev->vb_queue_lock;
drivers/media/usb/em28xx/em28xx-video.c:                dev->vbi_dev->queue->lock = &dev->vb_vbi_queue_lock;

There is a separate mutex for the video queue which needs to be acquired independently.

Darn, I missed that one. I was looking for it in em28xx_vdev_init(), which is where I expected the queue->lock to be set, if there was any.

That said, wouldn't it be a good idea to swap the order:

        vb2_fop_release(filp);
        mutex_lock(&dev->lock);

I don't believe there is a good reason for nesting mutexes here.

Regards,

Hans

Quoted 9 lines

--
Regards,
Sylwester
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 1 line
Ricardo Ribalda
#8 lore raw

PATCH v4 1/1 videobuf2: Add missing lock held on vb2_fop_relase

Here are the patches

https://patchwork.linuxtv.org/patch/20668/
https://patchwork.linuxtv.org/patch/20669/

Thanks!

Quoted 94 lines · Ricardo Ribalda Delgado

Hello Hans and Sywester

I have just posted a new patch. I think it fits the suggestions from both of you, please take a look to it and please post any comment.

I will also send a patch about the em28xx, to swap the lock order.

Thanks for your comments

Quoted 60 lines · Hans Verkuil
Quoted 46 lines · Sylwester Nawrocki
Quoted 39 lines · Hans Verkuil
Quoted 18 lines · Ricardo Ribalda Delgado
Quoted 18 lines

Hello Hans

Thanks for your comments.

Please take a look to v4 of this patch
https://patchwork.linuxtv.org/patch/20529/

Quoted 10 lines · Hans Verkuil
Quoted 10 lines · Ricardo Ribalda Delgado
Quoted 8 lines
Quoted 8 lines

From: Ricardo Ribalda <[email protected]>

vb2_fop_relase does not held the lock although it is modifying the queue->owner field.

This could lead to race conditions on the vb2_perform_io function when multiple applications are accessing the video device via read/write API:

Quoted 7 lines
Quoted 7 lines
Quoted 2 lines
Quoted 2 lines

It's also called directly by drivers/media/usb/em28xx/em28xx-video.c!

em28xx-video does not hold the lock, therefore it can call the normal function. On v2 we made a internal function that should be called if the funciton is called directly by the driver. Please take a look to the old comments. https://patchwork.linuxtv.org/patch/20460/

static int em28xx_v4l2_close(struct file *filp)

{
        struct em28xx_fh *fh  = filp->private_data;
        struct em28xx    *dev = fh->dev;
        int              errCode;

        em28xx_videodbg("users=%d\n", dev->users);

        mutex_lock(&dev->lock);
        vb2_fop_release(filp);
     ...
Quoted 3 lines

vb2_fop_release(filp) will, with your patch, also try to get dev->lock.

Sylwester's comment re em28xx is incorrect.

dev->lock is not used as the video queue lock:

Quoted 7 lines
$ git grep "lock =" drivers/media/usb/em28xx/
...
drivers/media/usb/em28xx/em28xx-video.c:        dev->vdev->queue->lock = &dev->vb_queue_lock;
drivers/media/usb/em28xx/em28xx-video.c:                dev->vbi_dev->queue->lock = &dev->vb_vbi_queue_lock;

There is a separate mutex for the video queue which needs to be acquired independently.

Darn, I missed that one. I was looking for it in em28xx_vdev_init(), which is where I expected the queue->lock to be set, if there was any.

That said, wouldn't it be a good idea to swap the order:

Quoted 19 lines
        vb2_fop_release(filp);
        mutex_lock(&dev->lock);

I don't believe there is a good reason for nesting mutexes here.

Regards,

Hans

Quoted 9 lines

--
Regards,
Sylwester
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Ricardo Ribalda

Signature · 1 line
Ricardo Ribalda