[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

Andreas Boll  changed:

   What|Removed |Added

 Resolution|--- |FIXED
 Status|NEW |RESOLVED

--- Comment #37 from Andreas Boll  ---
Thanks to Heiko for fixing this issue!

Fixed by

commit e933246013eef376804662f3fcf4646c143c6c88
Author: Heiko Przybyl 
Date:   Sun Nov 20 14:42:28 2016 +0100

r600/sb: Fix loop optimization related hangs on eg

-- 
You are receiving this mail because:
You are the assignee for the bug.___
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #36 from Heiko  ---
Btw, I've submitted the patch to the mesa list some time ago:
https://patchwork.freedesktop.org/patch/122534

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #35 from Adam Lyall  ---
I've tested Heiko's patch from comment 34 and it fixed Grid Autosport crashing
for on the latest Mesa from git.

Note that, at least for my Radeon HD5850, without the patch running with
sbsafemath still leads to the game crashing. Only nosb was working.

Grid Autosports benchmark FPS increases a few frames with the patch vs vanilla
with "nosb"

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

Heiko  changed:

   What|Removed |Added

 Attachment #126832|0   |1
is obsolete||

--- Comment #34 from Heiko  ---
Created attachment 127616
  --> https://bugs.freedesktop.org/attachment.cgi?id=127616=edit
Cleaned up version of the porposed fix

Ok, so I've stripped the debugging stuff from the patch. I tested it against
the octodad trace, saints row 4, grid autosport, rocket league, none of them
hanging the GPU. Also I've tested it on system's mesa and didn't have any
unwanted fallout.

Probably can be posted to the mailing list or should I split it up?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #33 from Marek Olšák  ---
The EVENT_WRITE errors are from a different issue that should be fixed in
master now.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #32 from Heiko  ---
(In reply to Frederic Romagne from comment #31)
> I had a bunch of 
> 
> oct. 02 10:06:22 kiss-desktop kernel: [drm:evergreen_packet3_check.isra.14
> [radeon]] *ERROR* bad EVENT_WRITE
> oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_ioctl [radeon]] *ERROR*
> Invalid command stream !
> oct. 02 10:06:22 kiss-desktop steam.desktop[1150]: radeon: The kernel
> rejected CS, see dmesg for more information (-22).
> oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_packet_next_reloc
> [radeon]] *ERROR* No packet3 for relocation for packet at 6094.
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6094]=0xC0044700
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6095]=0x0528
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6096]=0x0080
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6097]=0x2000
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6098]=0x8000
> oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6099]=0x
> 
> while trying Rocket League with this patch.

That's strange... the patch isn't touching any command stream packets. Just a
guess in the blue: did the GPU hang/reset usually happen before the broken
rendering point? And why is the command steam.desktop? I'd expect the RL
executable there...

So I just bought RL and I'll take a look.

Also, at least for the always lockups, an apitrace would be awesome :)

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #31 from Frederic Romagne  ---
(In reply to Frederic Romagne from comment #30)
> I can confirm it removes the lockup in Octodad, however it breaks the
> rendering in Rocket League... Recompiling mesa without the patch and Rocket
> League renders properly again...

It was a reply to comment 28,

I had a bunch of 

oct. 02 10:06:22 kiss-desktop kernel: [drm:evergreen_packet3_check.isra.14
[radeon]] *ERROR* bad EVENT_WRITE
oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_ioctl [radeon]] *ERROR*
Invalid command stream !
oct. 02 10:06:22 kiss-desktop steam.desktop[1150]: radeon: The kernel rejected
CS, see dmesg for more information (-22).
oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_packet_next_reloc
[radeon]] *ERROR* No packet3 for relocation for packet at 6094.
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6094]=0xC0044700
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6095]=0x0528
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6096]=0x0080
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6097]=0x2000
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6098]=0x8000
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6099]=0x
oct. 02 10:06:22 kiss-desktop kernel: [drm:evergreen_packet3_check.isra.14
[radeon]] *ERROR* bad EVENT_WRITE
oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_ioctl [radeon]] *ERROR*
Invalid command stream !
oct. 02 10:06:22 kiss-desktop steam.desktop[1150]: radeon: The kernel rejected
CS, see dmesg for more information (-22).
oct. 02 10:06:22 kiss-desktop kernel: [drm:radeon_cs_packet_next_reloc
[radeon]] *ERROR* No packet3 for relocation for packet at 6094.
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6094]=0xC0044700
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6095]=0x0528
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6096]=0x0080
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6097]=0x2000
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6098]=0x8000
oct. 02 10:06:22 kiss-desktop kernel: [drm] ib[6099]=0x


while trying Rocket League with this patch.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #30 from Frederic Romagne  ---
I can confirm it removes the lockup in Octodad, however it breaks the rendering
in Rocket League... Recompiling mesa without the patch and Rocket League
renders properly again...

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #29 from Marek Olšák  ---
(In reply to Heiko from comment #28)
> @Marek, what's the best/usual way to test for performance/instruction count
> in mesa changes. I noticed those 'helped'/'hurt'/'+-%' infos and some
> runtime numbers in commit messages, but I don't know how they are produced :/

There is no way to get stats for r600g easily. Intel and radeonsi developers
use their private shader-db repositories and generate their own stats. Intel
have "helped/hurt +-%" stats, while we have much more detailed stats for
radeonsi.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

Heiko  changed:

   What|Removed |Added

 Attachment #125271|0   |1
is obsolete||

--- Comment #28 from Heiko  ---
Created attachment 126832
  --> https://bugs.freedesktop.org/attachment.cgi?id=126832=edit
Possible fix for the lockups

The more I look at the sb code the more I dislike it :/ Anyhow, looks like the
GCM pass is b0rked and doesn't like unused ops at all.

The problem with that octodad trace is that with a pass through fold_assoc() an
ADD_INT op becomes unused, but isn't removed prior to GCM. GCM then moves it up
to the front of the shader (because there are no users), where the op's src
values aren't defined (in that particular case the loop counter variable). GCM
also moves ops up, if the usage count isn't fulfilled yet. Well that's when
things get really broken, since it seems to move the loop counter -- or at
least the initializer -- to fulfill the usage count. And well, then the GPU
finally locks up on the shader (or if mesa is compiled in debug mode, sb shows
unset registers), probably due to endlessly looping.

I tried to fix GCM, but everytime I thought I've did the right thing, I got
either unscheduled ops or wrong levels for the basic blocks. Also, the
DONT_HOIST stuff doesn't really seem to work that well either.

So I decided to fix the input feeded into the GCM pass, by iteratively removing
all unused ops in dce_cleanup. This could also be reducing amount of
instructions, that weren't actually removed before. Also optimized valtable's
use_count(), which gets called 1500+ times for the octodad trace and did
iterate over the whole use_info list every time...

The (untidied) patch fixes octodad for me. Would be nice, if someone could test
the other problematic games (be sure to test with a debug build, to get an
exception rather than a lockup). If it works there as well, I'd clean things
up.

@Marek, what's the best/usual way to test for performance/instruction count in
mesa changes. I noticed those 'helped'/'hurt'/'+-%' infos and some runtime
numbers in commit messages, but I don't know how they are produced :/

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #27 from Heiko  ---
Well, tiling and hyz were disabled per default as well, until bugs were ironed
out. So I'd go with the overall stability and use sbsafemath as default for now
(or just disable the calls to fold_assoc()).

And you're still able to revert with R600_DEBUG settings, if there is need for
that particular performance boost. Are there benchmark results to show the
gains for sb[no]safemath?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #26 from Gregor Münch  ---
(In reply to Marek Olšák from comment #24)
> How about we make R600_DEBUG=sbsafemath be the default?

If there are no resources inside AMD to fix those bugs, this will be the best
option. Those stalls are way more annoying than loosing some fps.

At least this should be the default in stable releases.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #25 from russianneuromancer at ya.ru ---
But how this will affect performance with other games that doesn't freeze?
There is no other way to fix this?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #24 from Marek Olšák  ---
How about we make R600_DEBUG=sbsafemath be the default?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #23 from Alex  ---
This also happens 100% of the time under Faeria and Rocket League.

The bug that appeared during a Rocket League session :
[36447.794626] BUG: unable to handle kernel paging request at c90401d40ffc
[36447.794656] IP: [] radeon_ring_backup+0xda/0x190 [radeon]
[36447.794697] PGD 40e099067 PUD 0
[36447.794710] Oops:  [#1] SMP
[36447.794723] Modules linked in: pci_stub vboxpci(OE) vboxnetadp(OE)
vboxnetflt(OE) snd_hrtimer vboxdrv(OE) binfmt_misc intel_rapl x86_pkg_te
mp_thermal intel_powerclamp kvm_intel kvm irqbypass crct10dif_pclmul
crc32_pclmul aesni_intel aes_x86_64 lrw gf128mul glue_helper ablk_helper
cryptd serio_raw hid_logitech_hidpp lpc_ich joydev input_leds
snd_hda_codec_realtek arc4 snd_hda_codec_hdmi snd_hda_codec_generic ath9k
ath9k_
common snd_usb_audio ath9k_hw snd_usbmidi_lib ath mac80211 cfg80211
snd_hda_intel snd_hda_codec snd_ctxfi snd_hda_core snd_hwdep snd_seq_midi
snd_seq_midi_event snd_pcm snd_rawmidi snd_seq snd_seq_device snd_timer snd
mei_me mei soundcore tpm_infineon shpchp mac_hid it87 hwmon_vid co
retemp parport_pc ppdev lp parport autofs4 hid_logitech_dj uas usb_storage
wacom hid_generic usbhid
[36447.794999]  hid mxm_wmi amdkfd amd_iommu_v2 radeon psmouse firewire_ohci
firewire_core crc_itu_t atl1c i2c_algo_bit ttm ahci drm_kms_helpe
r libahci syscopyarea sysfillrect sysimgblt fb_sys_fops e1000e drm ptp pps_core
video fjes wmi
[36447.795083] CPU: 2 PID: 3667 Comm: Xorg Tainted: G   OE  
4.4.0-36-generic #55-Ubuntu
[36447.795109] Hardware name: Gigabyte Technology Co., Ltd. To be filled by
O.E.M./Z77X-UD5H, BIOS F4 03/09/2012
[36447.795137] task: 8804072c5280 ti: 8800d9ed task.ti:
8800d9ed
[36447.795158] RIP: 0010:[]  []
radeon_ring_backup+0xda/0x190 [radeon]
[36447.795197] RSP: :8800d9ed3c48  EFLAGS: 00010202
[36447.795213] RAX: c90002313000 RBX:  RCX:

[36447.795233] RDX:  RSI: c90401d40ffc RDI:
00096440
[36447.795253] RBP: 8800d9ed3c78 R08: 88010c011e00 R09:
8163
[36447.795274] R10:  R11: 81ccf5ea R12:
8800358494e0
[36447.795294] R13: 8800358494b8 R14: 00025911 R15:
8800d9ed3cc0
[36447.795314] FS:  7fc9397a1a00() GS:88041ed0()
knlGS:
[36447.795337] CS:  0010 DS:  ES:  CR0: 80050033
[36447.795354] CR2: c90401d40ffc CR3: 00040a177000 CR4:
000406e0
[36447.795374] Stack:
[36447.795380]  880035848000 880035848000 8800358494e0
8800d9ed3cc0
[36447.795406]  8800358494e0  8800d9ed3d30
c020acdd
[36447.795430]  880035848740 00ff88040001 880035848018
8803216e3000
[36447.795455] Call Trace:
[36447.795472]  [] radeon_gpu_reset+0xcd/0x330 [radeon]
[36447.795492]  [] ? fence_wait_timeout+0x7d/0x160
[36447.795521]  [] radeon_gem_handle_lockup.part.3+0xe/0x20
[radeon]
[36447.795553]  [] radeon_gem_wait_idle_ioctl+0xdf/0x130
[radeon]
[36447.795584]  [] drm_ioctl+0x152/0x540 [drm]
[36447.795611]  [] ? radeon_gem_busy_ioctl+0xe0/0xe0 [radeon]
[36447.795639]  [] radeon_drm_ioctl+0x4c/0x80 [radeon]
[36447.795659]  [] do_vfs_ioctl+0x29f/0x490
[36447.795677]  [] ? SyS_futex+0x81/0x180
[36447.795693]  [] SyS_ioctl+0x79/0x90
[36447.795709]  [] entry_SYSCALL_64_fastpath+0x16/0x71
[36447.795727] Code: fb c0 48 85 c0 49 89 07 74 6c 41 8d 7e ff 31 d2 48 c1 e7
02 eb 07 49 8b 07 48 83 c2 04 49 8b 74 24 08 8d 4b 01 89 db 48 8
d 34 9e <8b> 36 89 34 10 41 23 4c 24 54 48 39 d7 89 cb 75 da 4c 89 ef e8
[36447.795850] RIP  [] radeon_ring_backup+0xda/0x190 [radeon]
[36447.795882]  RSP 
[36447.795893] CR2: c90401d40ffc
[36447.809449] ---[ end trace ffefc2413ee46759 ]---

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #22 from Alex  ---
It seems I'm esperiencing the same kind of crash, but with an HD7950 (and
usually under the 'Cities: skylines' game).

I'm using kubuntu 16.04 with the padoka ppa.
> uname -a
Linux AciD 4.4.0-36-generic #55-Ubuntu SMP Thu Aug 11 18:01:55 UTC 2016 x86_64
x86_64 x86_64 GNU/Linux

> lspci -v
09:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI]
Tahiti PRO [Radeon HD 7950/8950 OEM / R9 280] (prog-if 00 [VGA controller])
Subsystem: PC Partner Limited / Sapphire Technology Tahiti PRO [Radeon
HD 7950/8950 OEM / R9 280]
Flags: bus master, fast devsel, latency 0, IRQ 27
Memory at e000 (64-bit, prefetchable) [size=256M]
Memory at f430 (64-bit, non-prefetchable) [size=256K]
I/O ports at a000 [size=256]
Expansion ROM at f434 [disabled] [size=128K]
Capabilities: 
Kernel driver in use: radeon
Kernel modules: radeon


the crash info :
> dmesg
[87732.677658] radeon :09:00.0: ring 3 stalled for more than 1msec  
[87732.677664] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x0028462e on ring 3)   
[87732.993670] radeon :09:00.0: ring 0 stalled for more than 1msec  
[87732.993676] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc701 on ring 0)   
[87733.177659] radeon :09:00.0: ring 3 stalled for more than 10500msec  
[87733.177663] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284634 on ring 3)   
[87733.493683] radeon :09:00.0: ring 0 stalled for more than 10500msec
[87733.177663] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284634 on ring 3)
[87733.493683] radeon :09:00.0: ring 0 stalled for more than 10500msec  
[87733.493689] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc701 on ring 0)   
[87733.677671] radeon :09:00.0: ring 3 stalled for more than 11000msec  
[87733.677678] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284634 on ring 3)   
[87733.993674] radeon :09:00.0: ring 0 stalled for more than 11000msec  
[87733.993680] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc702 on ring 0)   
[87734.177683] radeon :09:00.0: ring 3 stalled for more than 11500msec  
[87734.177689] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x0028463a on ring 3)   
[87734.493685] radeon :09:00.0: ring 0 stalled for more than 11500msec  
[87734.493691] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc702 on ring 0)   
[87734.677675] radeon :09:00.0: ring 3 stalled for more than 12000msec  
[87734.677681] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x0028463a on ring 3)   
[87734.993694] radeon :09:00.0: ring 0 stalled for more than 12000msec  
[87734.993698] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc703 on ring 0)   
[87735.177730] radeon :09:00.0: ring 3 stalled for more than 12500msec  
[87735.177736] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284640 on ring 3)   
[87735.493705] radeon :09:00.0: ring 0 stalled for more than 12500msec  
[87735.493711] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc703 on ring 0)   
[87735.677711] radeon :09:00.0: ring 3 stalled for more than 13000msec  
[87735.677718] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284640 on ring 3)   
[87735.993724] radeon :09:00.0: ring 0 stalled for more than 13000msec  
[87735.993731] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc704 on ring 0)   
[87736.177724] radeon :09:00.0: ring 3 stalled for more than 13500msec  
[87736.177730] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 0x00284648 on ring 3)   
[87736.493725] radeon :09:00.0: ring 0 stalled for more than 13500msec  
[87736.493732] radeon :09:00.0: GPU lockup (current fence id
0x005cc6d9 last fence id 0x005cc705 on ring 0)   
[87736.677765] radeon :09:00.0: ring 3 stalled for more than 14000msec  
[87736.61] radeon :09:00.0: GPU lockup (current fence id
0x0028455c last fence id 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

Heiko  changed:

   What|Removed |Added

 CC||lil_tux at web.de

--- Comment #21 from Heiko  ---
Created attachment 125271
  --> https://bugs.freedesktop.org/attachment.cgi?id=125271=edit
Simple workaround

'Fixes' the octodad trace for me. That stops sb from using fold_assoc to fold
ADD_INT(ADD_INT(x, 1), 2) for scalar registers.

Problem is, that sb currently optimizes away the loop counter node, due to
optimizing its reference out of another node and marking it as unused/dead,
finally nuking the counter increment as well. Thus the loop break condition is
never met and the gpu hangs due to an endless loop, and finally, gets reset.

If mesa is compiled in debug mode, sb checks the shader it would put out and
fails assertions with unset registers... just being the loop counter of the
three loops in the octodad case (shader 70).

On a side note, this scenario might be already known for phi nodes, as the
comment in sb_expr.cpp:expr_handler::fold() mentions similar issues...

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #20 from Daniel T.  ---
I can confirm that setting R600_DEBUG=sbsafemath allows atleast octodad to be
launched and played without problems.

I currently do not have the other games installed, but I will start the
downloads and test ASAP.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

--- Comment #19 from iive at yahoo.com ---
Created attachment 125130
  --> https://bugs.freedesktop.org/attachment.cgi?id=125130=edit
Octrodad crashing pixel shader - TGSI, disassembly, sbdump of IR code and crash
backtrace

I've discovered few more things.

1. When retracing with mesa3d debug enabled, the shader backend internal
checker is run and it does detect errors (and crashes at failed assert). It
seems to error out at shader 70 (a draw command before the one that hangs).
I've attached a log that includes the output of `sbdump` too.

2. `R600_DEBUG=sbsafemath` could also be used to workaround the above crash and
the original lockup. (Please test if this works with the other titles too.)

Looking at the source, the 'safe math' option skips few optimization. The one
that seems to cause problem is the call to `fold_assoc()` in
`expr_handler::fold_alu_op2()` somewhere around `sb_expr.cpp:740`.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

iive at yahoo.com changed:

   What|Removed |Added

 CC||iive at yahoo.com

--- Comment #18 from iive at yahoo.com ---
Created attachment 125121
  --> https://bugs.freedesktop.org/attachment.cgi?id=125121=edit
Octrodad crashing pixel shader - TGSI, disassembly of normal and SB optimised
versions

I used SB_DSKIP_* variables to narrow down the exact shader that causes the
crash.

The shader is big and complicated. It does contain loops, breaks, if/then/else
statements, jumps. 

Just a note. `glretrace` outputs 2-3 more shaders after the dump point, so it
won't always be the last shader. Still the correct 2 shaders are output right
before the dump.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: 

[Bug 94900] HD6950 GPU lockup loop with various steam games (octodad[always], saints row 4[always], dead island[always], grid autosport[sometimes])

[bugzilla-dae...@freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94900

Daniel T.  changed:

   What|Removed |Added

Summary|HD6950 GPU lockup loop with |HD6950 GPU lockup loop with
   |various steam games |various steam games
   |(octodad, saints row 4, |(octodad[always], saints
   |dead island, grid   |row 4[always], dead
   |autosport)  |island[always], grid
   ||autosport[sometimes])

-- 
You are receiving this mail because:
You are the assignee for the bug.
-- next part --
An HTML attachment was scrubbed...
URL: