i915_gem_retire_requests oops on 2.6.30-rc7

i915_gem_retire_requests oops on 2.6.30-rc7

[Alex Bennee]

Hi,

I've just had X crash on me with an attendant oops in the kernel as it did so:

[99666.683322] ------------[ cut here ]------------
[99666.683332] WARNING: at drivers/gpu/drm/i915/i915_gem.c:3832
i915_gem_idle+0x1ae/0x2da()
[99666.683334] Hardware name: System Product Name
[99666.683336] Modules linked in: nls_iso8859_1 nls_cp437 vfat fat
nls_base bridge stp llc bnep rfcomm l2cap bluetooth ipv6 af_packet
snd_pcm_oss snd_mixer_oss snd_seq_oss snd_seq_midi_event snd_seq
snd_seq_device kvm_intel kvm acpi_cpufreq snd_hda_codec_analog
usb_storage usb_libusual snd_hda_intel snd_hda_codec uhci_hcd ehci_hcd
snd_hwdep snd_pcm snd_timer snd soundcore snd_page_alloc firewire_ohci
firewire_core ide_cd_mod cdrom pcspkr evdev usbcore crc_itu_t unix
[99666.683374] Pid: 5330, comm: X Not tainted 2.6.30-rc7-ajb #65
[99666.683376] Call Trace:
[99666.683382]  [<ffffffff803d3db8>] ? i915_gem_retire_requests+0x164/0x184
[99666.683386]  [<ffffffff803d553d>] ? i915_gem_idle+0x1ae/0x2da
[99666.683390]  [<ffffffff802378e9>] warn_slowpath_common+0x7c/0xa9
[99666.683393]  [<ffffffff8023792a>] warn_slowpath_null+0x14/0x16
[99666.683396]  [<ffffffff803d553d>] i915_gem_idle+0x1ae/0x2da
[99666.683399]  [<ffffffff803d5687>] i915_gem_lastclose+0x1e/0x3b
[99666.683402]  [<ffffffff803ce725>] i915_driver_lastclose+0x37/0x5a
[99666.683406]  [<ffffffff803be306>] drm_lastclose+0x47/0x2b1
[99666.683409]  [<ffffffff803bea4e>] drm_release+0x4d1/0x4ee
[99666.683413]  [<ffffffff802ad8ad>] __fput+0xe4/0x19d
[99666.683416]  [<ffffffff802ad983>] fput+0x1d/0x1f
[99666.683419]  [<ffffffff802aac02>] filp_close+0x5e/0x68
[99666.683422]  [<ffffffff80239307>] put_files_struct+0x70/0xc7
[99666.683425]  [<ffffffff802393aa>] exit_files+0x4c/0x55
[99666.683427]  [<ffffffff8023ac40>] do_exit+0x1c7/0x644
[99666.683431]  [<ffffffff80242c98>] ? __dequeue_signal+0x108/0x134
[99666.683434]  [<ffffffff8023b134>] do_group_exit+0x77/0xa4
[99666.683437]  [<ffffffff80244984>] get_signal_to_deliver+0x318/0x338
[99666.683442]  [<ffffffff8020a525>] do_notify_resume+0x90/0x6f9
[99666.683444]  [<ffffffff80242962>] ? lock_task_sighand+0x3a/0x63
[99666.683447]  [<ffffffff802ad959>] ? __fput+0x190/0x19d
[99666.683450]  [<ffffffff80241a32>] ? sigprocmask+0xb1/0xd2
[99666.683453]  [<ffffffff80241ef6>] ? sys_rt_sigprocmask+0x59/0xd3
[99666.683456]  [<ffffffff80241a32>] ? sigprocmask+0xb1/0xd2
[99666.683459]  [<ffffffff8020b235>] sysret_signal+0x6d/0xb7
[99666.683462] ---[ end trace 59cc4070da9f68b2 ]---
[99666.694008] [drm:i915_get_vblank_counter] *ERROR* trying to get
vblank count for disabled pipe 0

I've been seeing these for most of the 2.6.30 series but I thought
they had been banished with the latest updates. Please advise if you
would like to see any other information?

-- 
Alex, homepage: http://www.bennee.com/~alex/
CV: http://www.bennee.com/~alex/cv.php

Re: i915_gem_retire_requests oops on 2.6.30-rc7

[Eric Anholt]

[-- Attachment #1: Type: text/plain, Size: 607 bytes --]

On Tue, 2009-05-26 at 13:20 +0100, Alex Bennee wrote:
> Hi,
> 
> I've just had X crash on me with an attendant oops in the kernel as it did so:

That's not an oops, that's a WARN_ON.  The hardware wasn't caught up by
the time we gave up when shutting down the harwdare for VT switching.
That's bad.  But we don't know how long we should wait, since the GPU
may be hung, and waiting forever would be bad.  So we warn that
Something Bad may have happened.

It's gone with KMS since we never shut down the GPU.

-- 
Eric Anholt
eric@anholt.net                         eric.anholt@intel.com



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: i915_gem_retire_requests oops on 2.6.30-rc7

[Alex Bennee]

2009/5/26 Eric Anholt <eric@anholt.net>:
> On Tue, 2009-05-26 at 13:20 +0100, Alex Bennee wrote:
>> Hi,
>>
>> I've just had X crash on me with an attendant oops in the kernel as it did so:
>
> That's not an oops, that's a WARN_ON.  The hardware wasn't caught up by
> the time we gave up when shutting down the harwdare for VT switching.
> That's bad.  But we don't know how long we should wait, since the GPU
> may be hung, and waiting forever would be bad.  So we warn that
> Something Bad may have happened.

I can restart X afterwards so it can't be that hung. Would you like a
reg dump next the problem occurs?

>
> It's gone with KMS since we never shut down the GPU.
>

Although I'm not switching to KMS yet due to the other issues I'm
seeing (see xorg mails). But with KMS enabled we still see some things
like:

[  660.644103] [drm] TMDS-8: set mode  2f
[  960.272242] INFO: task events/0:9 blocked for more than 120 seconds.
[  960.272245] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[  960.272248] events/0      D ffff8800b70a58c0     0     9      2
[  960.272254]  ffff8800b70b5de0 0000000000000046 ffff8800b103dc41
ffff880001028980
[  960.272260]  0000000000004000 0000000000010dc0 000000000000ca38
ffff8800b0d31678
[  960.272265]  ffff8800b70b5d60 ffff8800b70a58c0 ffffffff80648350
ffff8800b70a5c48
[  960.272270] Call Trace:
[  960.272280]  [<ffffffff804bfce6>] ? thread_return+0x3e/0xaa
[  960.272285]  [<ffffffff804c0589>] __mutex_lock_slowpath+0xdf/0x129
[  960.272290]  [<ffffffff804c092d>] mutex_lock+0x23/0x3b
[  960.272295]  [<ffffffff803cecd3>] i915_gem_retire_work_handler+0x2d/0x6b
[  960.272301]  [<ffffffff80247666>] worker_thread+0x176/0x20f
[  960.272306]  [<ffffffff803ceca6>] ? i915_gem_retire_work_handler+0x0/0x6b
[  960.272311]  [<ffffffff8024b463>] ? autoremove_wake_function+0x0/0x3d
[  960.272315]  [<ffffffff802474f0>] ? worker_thread+0x0/0x20f
[  960.272319]  [<ffffffff802474f0>] ? worker_thread+0x0/0x20f
[  960.272323]  [<ffffffff8024b061>] kthread+0x5b/0x88
[  960.272329]  [<ffffffff8020c0ba>] child_rip+0xa/0x20
[  960.272333]  [<ffffffff8024b006>] ? kthread+0x0/0x88
[  960.272337]  [<ffffffff8020c0b0>] ? child_rip+0x0/0x20

So are you sure this is a !KMS related thing?

-- 
Alex, homepage: http://www.bennee.com/~alex/
CV: http://www.bennee.com/~alex/cv.php

Re: i915_gem_retire_requests oops on 2.6.30-rc7

[Eric Anholt]

[-- Attachment #1: Type: text/plain, Size: 2696 bytes --]

On Wed, 2009-05-27 at 08:50 +0100, Alex Bennee wrote:
> 2009/5/26 Eric Anholt <eric@anholt.net>:
> > On Tue, 2009-05-26 at 13:20 +0100, Alex Bennee wrote:
> >> Hi,
> >>
> >> I've just had X crash on me with an attendant oops in the kernel as it did so:
> >
> > That's not an oops, that's a WARN_ON.  The hardware wasn't caught up by
> > the time we gave up when shutting down the harwdare for VT switching.
> > That's bad.  But we don't know how long we should wait, since the GPU
> > may be hung, and waiting forever would be bad.  So we warn that
> > Something Bad may have happened.
> 
> I can restart X afterwards so it can't be that hung. Would you like a
> reg dump next the problem occurs?
> 
> >
> > It's gone with KMS since we never shut down the GPU.
> >
> 
> Although I'm not switching to KMS yet due to the other issues I'm
> seeing (see xorg mails). But with KMS enabled we still see some things
> like:
> 
> [  660.644103] [drm] TMDS-8: set mode  2f
> [  960.272242] INFO: task events/0:9 blocked for more than 120 seconds.
> [  960.272245] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
> disables this message.
> [  960.272248] events/0      D ffff8800b70a58c0     0     9      2
> [  960.272254]  ffff8800b70b5de0 0000000000000046 ffff8800b103dc41
> ffff880001028980
> [  960.272260]  0000000000004000 0000000000010dc0 000000000000ca38
> ffff8800b0d31678
> [  960.272265]  ffff8800b70b5d60 ffff8800b70a58c0 ffffffff80648350
> ffff8800b70a5c48
> [  960.272270] Call Trace:
> [  960.272280]  [<ffffffff804bfce6>] ? thread_return+0x3e/0xaa
> [  960.272285]  [<ffffffff804c0589>] __mutex_lock_slowpath+0xdf/0x129
> [  960.272290]  [<ffffffff804c092d>] mutex_lock+0x23/0x3b
> [  960.272295]  [<ffffffff803cecd3>] i915_gem_retire_work_handler+0x2d/0x6b
> [  960.272301]  [<ffffffff80247666>] worker_thread+0x176/0x20f
> [  960.272306]  [<ffffffff803ceca6>] ? i915_gem_retire_work_handler+0x0/0x6b
> [  960.272311]  [<ffffffff8024b463>] ? autoremove_wake_function+0x0/0x3d
> [  960.272315]  [<ffffffff802474f0>] ? worker_thread+0x0/0x20f
> [  960.272319]  [<ffffffff802474f0>] ? worker_thread+0x0/0x20f
> [  960.272323]  [<ffffffff8024b061>] kthread+0x5b/0x88
> [  960.272329]  [<ffffffff8020c0ba>] child_rip+0xa/0x20
> [  960.272333]  [<ffffffff8024b006>] ? kthread+0x0/0x88
> [  960.272337]  [<ffffffff8020c0b0>] ? child_rip+0x0/0x20
> 
> So are you sure this is a !KMS related thing?

Yes, that looks like something else entirely.

If you have GPU hangs with KMS, please report them at
bugs.freedesktop.org, with intel_gpu_dump output.

-- 
Eric Anholt
eric@anholt.net                         eric.anholt@intel.com



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]