Bug#570466: xserver-xorg-video-radeon: randomly hangs when using kms/dri2

[Brice Goglin]

hobbes at poukram.net wrote:
> Package: xserver-xorg-video-radeon
> Version: 1:6.12.99+git20100201.a887818f-1
> Severity: normal
>
> Hello,
>
> when using kms/dri2 with the free radeon driver, my display hangs after
> a random (but always quite short) delay. I have not been able to find a
> precise trigger but will report back if a pattern shows up.
>
> The system is still accessible via network, and X consumes 100% CPU. I
> have this in my Xorg.0.log after the hang :
>
> -------------------
> [mi] EQ overflowing. The server is probably stuck in an infinite loop.
>
> Backtrace:
> 0: /usr/bin/X (xorg_backtrace+0x3b) [0x80e696b]
> 1: /usr/bin/X (mieqEnqueue+0x1ab) [0x80e628b]
> 2: /usr/bin/X (xf86PostMotionEventP+0xcc) [0x80c104c]
> 3: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6fe7000+0x4801) [0xb6feb801]
> 4: /usr/lib/xorg/modules/input/evdev_drv.so (0xb6fe7000+0x4af6) [0xb6febaf6]
> 5: /usr/bin/X (0x8048000+0x6bb2f) [0x80b3b2f]
> 6: /usr/bin/X (0x8048000+0x11d194) [0x8165194]
> 7: (vdso) (__kernel_sigreturn+0x0) [0xb77db400]
> 8: /usr/lib/libdrm.so.2 (drmCommandWriteRead+0x3b) [0xb738018b]
> 9: /usr/lib/libdrm_radeon.so.1 (0xb77c0000+0x123a) [0xb77c123a]
> 10: /usr/lib/libdrm_radeon.so.1 (0xb77c0000+0x128d) [0xb77c128d]
> 11: /usr/lib/libdrm_radeon.so.1 (radeon_bo_map+0x1b) [0xb77c29fb]
> 12: /usr/lib/xorg/modules/drivers/radeon_drv.so (0xb7288000+0xb63a4) [0xb733e3a4]
> 13: /usr/lib/xorg/modules/libexa.so (0xb738d000+0x782a) [0xb739482a]
> 14: /usr/bin/X (0x8048000+0x15baef) [0x81a3aef]
> 15: /usr/bin/X (0x8048000+0x2b7c4) [0x80737c4]
> 16: /usr/bin/X (0x8048000+0x2b017) [0x8073017]
> 17: /usr/bin/X (0x8048000+0x1e95a) [0x806695a]
> 18: /lib/i686/cmov/libc.so.6 (__libc_start_main+0xe5) [0xb74ecb55]
> 19: /usr/bin/X (0x8048000+0x1e541) [0x8066541]
> -------------------
>
> I straced the X process, it indeed seems stuck in an infinite loop :
>
> ioctl(9, 0xc0086464, 0xbfdf4628)        = -1 EBUSY (Device or resource busy)
> ioctl(9, 0xc0086464, 0xbfdf4628)        = -1 EBUSY (Device or resource busy)
> --- SIGALRM (Alarm clock) @ 0 (0) ---
> sigreturn()                             = ? (mask now [])
> ioctl(9, 0xc0086464, 0xbfdf4628)        = -1 EBUSY (Device or resource busy)
> ioctl(9, 0xc0086464, 0xbfdf4628)        = -1 EBUSY (Device or resource busy)
>
>
> The ioctl line comes thousands of times, and sometimes the SIGALRM line
> (I almost missed it).
>   

Can you try with a 2.6.33-rc kernel ?

By the way, your bug reports seem to have always the same wrongly
formatted From: line.

Brice