Hi Nelson,

Nelson A. de Oliveira wrote:

> It seems to be somehow reproducible when using LibreOffice and Chromium.

That's no good. Thanks for reporting it.

[...]
| radeon 0000:01:00.0: GPU lockup CP stall for more than 10000msec
| ------------[ cut here ]------------
| WARNING: at /build/buildd-linux-2.6_2.6.39-2-amd64-kuqdRa/linux-2.6-2.6.39/debian/build/source_amd64_none/drivers/gpu/drm/radeon/radeon_fence.c:246 radeon_fence_wait+0x2a9/0x34b [radeon]()
| Hardware name: Infoway
| GPU lockup (waiting for 0x0001CAAA last fence id 0x0001CAA9)

The upstream report https://bugs.freedesktop.org/show_bug.cgi?id=33643
seems similar.

First, the boring questions:

- can you reproduce it with the kernel from experimental?
- can you reproduce it with the kernel from squeeze?
- if it's possible to test without the virtualbox driver, does
that help?
- could you also attach the output of

sh /usr/share/bug/xserver-xorg-video-radeon/script 3>&1

for reference?

Aside from that, this should probably be investigated upstream, by
filing a new bug at http://bugs.freedesktop.org/, product DRI,
component DRM/Radeon, with as specific and reliable a reproduction
recipe as possible.

Hope that helps, and sorry I have no better ideas.

Regards,
Jonathan

| Modules linked in: md4 hmac nls_utf8 cifs fscache tun powernow_k8 mperf ipt_REJECT cpufreq_conservative cpufreq_userspace cpufreq_stats cpufreq_powersave xt_tcpudp iptable_filter ip_tables x_tables bridge stp ppdev lp vboxnetadp(O) vboxnetflt(O) vboxdrv(O) binfmt_misc fuse loop kvm_amd kvm usbhid hid snd_hda_codec_realtek sg radeon ata_generic snd_hda_intel snd_hda_codec ttm snd_hwdep snd_pcm drm_kms_helper sr_mod cdrom r8169 option usb_wwan usbserial snd_seq snd_timer snd_seq_device drm i2c_algo_bit snd ohci_hcd psmouse parport_pc i2c_piix4 edac_core mii power_supply pata_atiixp soundcore ehci_hcd snd_page_alloc edac_mce_amd usbcore i2c_core k8temp floppy parport evdev serio_raw button processor ext4 mbcache jbd2 crc16 dm_mod sd_mod crc_t10dif ahci libahci libata scsi_mod fan thermal thermal_sys
| Pid: 1226, comm: Xorg Tainted: G O 2.6.39-2-amd64 #1
| Call Trace:
| [<ffffffff810458b4>] ? warn_slowpath_common+0x78/0x8c
| [<ffffffff81045966>] ? warn_slowpath_fmt+0x45/0x4a
| [<ffffffffa0304a60>] ? radeon_fence_wait+0x2a9/0x34b [radeon]
| [<ffffffff8105f351>] ? wake_up_bit+0x20/0x20
| [<ffffffffa03189b9>] ? radeon_ib_get+0x109/0x1ad [radeon]
| [<ffffffffa0319b28>] ? radeon_cs_ioctl+0x81/0x191 [radeon]
| [<ffffffffa020a798>] ? drm_ioctl+0x24e/0x321 [drm]
| [<ffffffffa0319aa7>] ? radeon_cs_parser_init+0x26e/0x26e [radeon]
| [<ffffffff81335ae2>] ? do_page_fault+0x314/0x34c
| [<ffffffff8110899d>] ? do_vfs_ioctl+0x445/0x492
| [<ffffffff81108a35>] ? sys_ioctl+0x4b/0x72
| [<ffffffff81338dd2>] ? system_call_fastpath+0x16/0x1b
| ---[ end trace 5ecacd3c3cd46744 ]---
| radeon 0000:01:00.0: GPU reset succeed
| radeon 0000:01:00.0: ffff880129d00400 unpin not necessary
| ------------[ cut here ]------------
| kernel BUG at /build/buildd-linux-2.6_2.6.39-2-amd64-kuqdRa/linux-2.6-2.6.39/debian/build/source_amd64_none/drivers/gpu/drm/radeon/radeon_object.c:363!



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110705204407.GA7032@elie">http://lists.debian.org/20110705204407.GA7032@elie

Nelson A. de Oliveira wrote:

> I was able to reproduce the crash only when using both Chromium and
> LibreOffice at the same time (it wasn't reproducible not even when
> playing heavier 3D games) thus I thought that indeed it was some kind
> of problem with the X stack or the Kernel. But then it started to
> crash more frequently (even with Kernel versions from stable and
> experimental).
> It happens that the problem was a failing fan of the graphics card.

Thanks for your work tracking this one down.

It's a shame this was so hard to figure out --- I wonder if there's
some temperature sensor the kernel could use to help diagnose this
kind of thing. If some day you come up with some possible improvement
to the kernel in that vein, please feel free to contact
linux-kernel@vger.kernel.org about it (cc-ing me if you want).

Regards,
Jonathan



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110712151802.GB12343@elie">http://lists.debian.org/20110712151802.GB12343@elie