lkml.org 
[lkml]   [2007]   [Sep]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Subject[BUG]: circular locking depedency in videobuf code
Date
Hi,

I found a bug (circular lock dependency) in generic videobuf code:


Suppose app has two threads, and one calls munmap() on a video buffer , and second calls VIDIOC_QBUF ioctl.
(The actual app that does that is kdetv on exit)


thread1: takes current->mm->mmap_sem, then takes q->lock
thread2: takes q->lock then takes current->mm->mmap_sem

Explanation:

thread 1: munmap takes current->mm->mmap_sem, and calls videobuf_vm_close which tries to take q->lock
thread2: videobuf_qbuf takes q->lock and calls q->ops->buf_prepare(q,buf,field),
that is driver specific, but most(if not all) implementations call videobuf_iolock, that calls videobuf_dma_init_user, and
that function takes current->mm->mmap_sem.

Since buffer queue is the same, as well as current->mm, a deadlock can occur. It happens almost always on my system
Since mm->mmap_sem for kdetv is down forever, commands like ps/top hang, and reboot doesn't work for same reason.
I have FlyVideo2000 card (saa7130HL based), so saa7134 driver is used.

It is a bit difficult for me to fix that bug, so can you help me?
Maybe I should take current->mm->mmap_sem in beginning of videobuf_qbuf and release it immediately, is this a right way?

Looking for your comments,
Best regards,
Maxim Levitsky

PS:

this is a call trace of both threads after the hang (with frame pointers enabled):

<4>[18681.072790] kdetv D c1ef3900 0 8137 4821
<4>[18681.072794] df0b1ee4 00200046 00000818 c1ef3900 00000002 df0b1ecc 6f504f76 00000cd3
<4>[18681.072801] df00e000 c1f3ed40 00000000 00200046 00d269ab 00000001 000187f7 00000000
<4>[18681.072809] 000000ff 00000000 00000000 00000818 f722f5dc 00200246 df00e000 df0b1f24
<4>[18681.072816] Call Trace:
<4>[18681.072818] [<c03aba38>] __mutex_lock_slowpath+0xf8/0x350
<4>[18681.072822] [<c03abcac>] mutex_lock+0x1c/0x20
<4>[18681.072825] [<f940f78d>] videobuf_vm_close+0x9d/0x140 [video_buf]
<4>[18681.072835] [<c0179a3b>] remove_vma+0x2b/0x50
<4>[18681.072839] [<c017a55a>] do_munmap+0x18a/0x1f0
<4>[18681.072843] [<c017a5f0>] sys_munmap+0x30/0x50
<4>[18681.072847] [<c01084ca>] sysenter_past_esp+0x6b/0xb5
<4>[18681.072851] =======================
<4>[18681.072853] kdetv D c1f45900 0 8139 4821
<4>[18681.072857] dc12fd08 00200046 00000a18 c1f45900 00000002 dc12fcf0 6f4ec99b 00000cd3
<4>[18681.072864] c3e26d00 c1f90d40 00000001 00000002 00d269b4 00000001 000144f6 00000000
<4>[18681.072872] 000000ff 00000000 00000000 00000a18 00000002 e48f0b34 c3e26d00 dc12fd30
<4>[18681.072879] Call Trace:
<4>[18681.072881] [<c0224b25>] rwsem_down_failed_common+0x75/0x190
<4>[18681.072885] [<c03acdcd>] rwsem_down_read_failed+0x1d/0x28
<4>[18681.072889] [<c03ace53>] call_rwsem_down_read_failed+0x7/0xc
<4>[18681.072894] [<f940df34>] videobuf_dma_init_user+0xb4/0x150 [video_buf]
<4>[18681.072901] [<f940e2e7>] videobuf_iolock+0xd7/0xe0 [video_buf]
<4>[18681.072907] [<f943de82>] buffer_prepare+0x1b2/0x200 [saa7134]
<4>[18681.072919] [<f940e87d>] videobuf_qbuf+0x20d/0x440 [video_buf]
<4>[18681.072926] [<f943f0a2>] video_do_ioctl+0x562/0x1110 [saa7134]
<4>[18681.072937] [<f93e028a>] video_usercopy+0xda/0x250 [videodev]
<4>[18681.072943] [<f943ccda>] video_ioctl+0x1a/0x20 [saa7134]
<4>[18681.072954] [<c0199f2b>] do_ioctl+0x6b/0x80
<4>[18681.072957] [<c0199f97>] vfs_ioctl+0x57/0x290
<4>[18681.072961] [<c019a209>] sys_ioctl+0x39/0x60
<4>[18681.072964] [<c01084ca>] sysenter_past_esp+0x6b/0xb5
<4>[18681.072969] =======================

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2007-09-09 23:13    [W:0.038 / U:0.828 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site