Skip to contentSkip to conversation

- piglit test - streaming-texture-leak deadlock GPU

PATCH In review Shawn Starr 1 participant 4 messages Oct 27, 2016 – Nov 12, 2016

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 - piglit test - streaming-texture-leak deadlock GPU

Hello AMD folks,

Using latest LLVM trunk/Mesa master and latest changes from drm-next-4.10-wip available, I am unable to run piglit concurrently, causes GPU graphics pipe to get pegged 100%.

[   86.777305] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c28770c
[   86.777306] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00100061
[   86.777308] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0807700C
[   86.777309] VM fault (0x0c, vmid 4) at page 1048673, read from 'SDM0' (0x53444d30) (119)
[   86.785159] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c44770c
[   86.785163] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00100062
[   86.785164] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0407700C
[   86.785166] VM fault (0x0c, vmid 2) at page 1048674, read from 'SDM0' (0x53444d30) (119)
[   86.785173] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c42770c
[   86.785174] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00100062
[   86.785175] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0207700C
[   86.785177] VM fault (0x0c, vmid 1) at page 1048674, read from 'SDM0' (0x53444d30) (119)

Running: streaming-texture-leak -auto 3 times, wedges GPU.

Thanks,
Shawn
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#1 lore raw

More info kernel boot options (some are unused like DAL)

BOOT_IMAGE=/boot/vmlinuz-4.9.0-rc2+ root=/dev/sda2 ro rhgb slub_debug=- cgroup_disable=memory console=tty0 console=ttyUSB0,9600n8 nmi_watchdog=0 audit=0 amdgpu.gartsize=8192 amdgpu.vm_size=16 amdgpu.exp_hw_support=1 rd.driver.blacklist=radeon amdgpu.dal=1 amdgpu.moverate=-1 video=LVDS-1:d video=VGA-0:e resume=/dev/sda7 pcie_aspm=force

Thanks,
Shawn

Quoted 24 lines · EDT Shawn Starr

Hello AMD folks,

Using latest LLVM trunk/Mesa master and latest changes from drm-next-4.10-wip available, I am unable to run piglit concurrently, causes GPU graphics pipe to get pegged 100%.

[   86.777305] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c28770c
[   86.777306] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR  

0x00100061 [ 86.777308] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0807700C [ 86.777309] VM fault
(0x0c, vmid 4) at page 1048673, read from 'SDM0' (0x53444d30) (119) [ 86.785159] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c44770c [
86.785163] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785164] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0407700C [ 86.785166] VM fault
(0x0c, vmid 2) at page 1048674, read from 'SDM0' (0x53444d30) (119) [ 86.785173] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c42770c [
86.785174] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785175] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0207700C [ 86.785177] VM fault
(0x0c, vmid 1) at page 1048674, read from 'SDM0' (0x53444d30) (119)

Quoted 8 lines

Running: streaming-texture-leak -auto 3 times, wedges GPU.

Thanks,
Shawn
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#2 lore raw

The following processes are left running:

spstarr   9830  0.0  0.1 2318428 39476 ?       Dsl  04:44   0:03 /home/spstarr/Sources/piglit/bin/max-texture-size -auto
spstarr  11082  0.0  0.1 295716 43668 ?        Ssl  04:44   0:00 /home/spstarr/Sources/piglit/bin/streaming-texture-leak -auto
spstarr  11960  0.0  0.0 217888 25112 ?        Ssl  04:44   0:00 /home/spstarr/Sources/piglit/bin/texsubimage-unpack -auto -fbo
spstarr  11961  0.0  0.0 108172  4780 ?        Ss   04:44   0:00 /home/spstarr/Sources/piglit/bin/texsubimage-depth-formats -auto
spstarr  11962  0.0  0.0 108172  4852 ?        Ss   04:44   0:00 /home/spstarr/Sources/piglit/bin/texture-al -auto
spstarr  11963  0.0  0.0 108168  4636 ?        Ss   04:44   0:00 /home/spstarr/Sources/piglit/bin/triangle-guardband-viewport -auto -fbo
spstarr  11964  0.0  0.0 108172  4636 ?        Ss   04:44   0:00 /home/spstarr/Sources/piglit/bin/getteximage-targets 1D -auto -fbo
spstarr  12474  0.0  0.0 117140   940 pts/2    S+   15:38   0:00 grep --color=auto piglit
spstarr  22856  0.0  0.1 292360 44112 ?        Ssl  04:42   0:19 /home/spstarr/Sources/piglit/bin/glsl-vs-raytrace-bug26691 -auto -fbo
spstarr  24043  0.8  0.5 1480452 166248 pts/1  Sl+  04:11   5:31 python3 ./piglit run tests/gpu summary/amdgpu-cik-gl -c

This is with Linus master (-rc9) and latest drm-next-4.10-wip.

Thanks,
Shawn

Quoted 8 lines · EDT Shawn Starr

More info kernel boot options (some are unused like DAL)

BOOT_IMAGE=/boot/vmlinuz-4.9.0-rc2+ root=/dev/sda2 ro rhgb slub_debug=- cgroup_disable=memory console=tty0 console=ttyUSB0,9600n8 nmi_watchdog=0 audit=0 amdgpu.gartsize=8192 amdgpu.vm_size=16 amdgpu.exp_hw_support=1 rd.driver.blacklist=radeon amdgpu.dal=1 amdgpu.moverate=-1 video=LVDS-1:d video=VGA-0:e resume=/dev/sda7 pcie_aspm=force

Quoted 38 lines

Thanks,
Shawn

Quoted 33 lines · EDT Shawn Starr

Hello AMD folks,

Using latest LLVM trunk/Mesa master and latest changes from drm-next-4.10-wip available, I am unable to run piglit concurrently, causes
GPU graphics pipe to get pegged 100%.

[   86.777305] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c28770c
[   86.777306] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR

0x00100061 [ 86.777308] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0807700C [ 86.777309] VM fault
(0x0c, vmid 4) at page 1048673, read from 'SDM0' (0x53444d30) (119) [ 86.785159] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c44770c [
86.785163] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785164] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0407700C [ 86.785166] VM fault
(0x0c, vmid 2) at page 1048674, read from 'SDM0' (0x53444d30) (119) [ 86.785173] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c42770c [
86.785174] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785175] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0207700C [ 86.785177] VM fault
(0x0c, vmid 1) at page 1048674, read from 'SDM0' (0x53444d30) (119)

Running: streaming-texture-leak -auto 3 times, wedges GPU.

Thanks,
Shawn
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw

[drm-next-4.10-wip (rc4)][AMDGPU CIK] - piglit tests still deadlock GPU

To continue testing piglit concurrency, the following processes remain stuck running:

spstarr  30217  1.5  0.1 486232 43848 ?        Ssl  15:41   0:02 /home/spstarr/Sources/piglit/bin/copypixels-sync -auto
spstarr  30285  1.2  0.1 486200 42708 ?        Ssl  15:41   0:01 /home/spstarr/Sources/piglit/bin/draw-sync -auto -fbo
spstarr  30538  0.4  0.1 552200 43220 ?        Ssl  15:41   0:00 /home/spstarr/Sources/piglit/bin/getteximage-formats -auto
spstarr  30561  0.3  0.1 485476 42116 ?        Ssl  15:41   0:00 /home/spstarr/Sources/piglit/bin/getteximage-depth -auto -fbo
spstarr  30573  0.6  0.1 1532200 42344 ?       Ssl  15:41   0:00 /home/spstarr/Sources/piglit/bin/max-texture-size -auto
spstarr  30597  0.3  0.1 570096 46616 ?        Dsl  15:41   0:00 /home/spstarr/Sources/piglit/bin/streaming-texture-leak -auto
spstarr  30614  0.1  0.1 550032 43220 ?        Ssl  15:41   0:00 /home/spstarr/Sources/piglit/bin/texsubimage -auto -fbo
spstarr  30620  0.1  0.1 549600 42568 ?        Ssl  15:41   0:00 /home/spstarr/Sources/piglit/bin/texsubimage-unpack -auto -fbo

The following kernel messages show from the test up to this point:

[ 1415.402790] glx-visuals-dep[29653]: segfault at 1c ip 00007f0a26951f22 sp 00007ffce31eef80 error 4 in radeonsi_dri.so[7f0a266bd000+d0a000]
[ 1417.649705] glx-visuals-ste[29671]: segfault at 1c ip 00007f857193ff22 sp 00007ffe66fba660 error 4 in radeonsi_dri.so[7f85716ab000+d0a000]
[ 1424.814581] amdgpu 0000:01:00.0: GPU fault detected: 146 0x058a760c
[ 1424.814586] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x0010002C
[ 1424.814587] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A07600C
[ 1424.814588] VM fault (0x0c, vmid 5) at page 1048620, read from 'CPF' (0x43504600) (118)
[ 1424.814653] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0724770c
[ 1424.814655] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x001005B9
[ 1424.814656] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0407700C
[ 1424.814660] VM fault (0x0c, vmid 2) at page 1050041, read from 'SDM0' (0x53444d30) (119)

Thanks,
Shawn

Quoted 22 lines · EST Shawn Starr

The following processes are left running:

spstarr 9830 0.0 0.1 2318428 39476 ? Dsl 04:44 0:03
/home/spstarr/Sources/piglit/bin/max-texture-size -auto spstarr 11082 0.0
0.1 295716 43668 ? Ssl 04:44 0:00
/home/spstarr/Sources/piglit/bin/streaming-texture-leak -auto spstarr
11960 0.0 0.0 217888 25112 ? Ssl 04:44 0:00
/home/spstarr/Sources/piglit/bin/texsubimage-unpack -auto -fbo spstarr
11961 0.0 0.0 108172 4780 ? Ss 04:44 0:00
/home/spstarr/Sources/piglit/bin/texsubimage-depth-formats -auto spstarr
11962 0.0 0.0 108172 4852 ? Ss 04:44 0:00
/home/spstarr/Sources/piglit/bin/texture-al -auto spstarr 11963 0.0 0.0
108168 4636 ? Ss 04:44 0:00
/home/spstarr/Sources/piglit/bin/triangle-guardband-viewport -auto -fbo
spstarr 11964 0.0 0.0 108172 4636 ? Ss 04:44 0:00
/home/spstarr/Sources/piglit/bin/getteximage-targets 1D -auto -fbo spstarr
12474 0.0 0.0 117140 940 pts/2 S+ 15:38 0:00 grep --color=auto
piglit spstarr 22856 0.0 0.1 292360 44112 ? Ssl 04:42 0:19
/home/spstarr/Sources/piglit/bin/glsl-vs-raytrace-bug26691 -auto -fbo
spstarr 24043 0.8 0.5 1480452 166248 pts/1 Sl+ 04:11 5:31 python3
./piglit run tests/gpu summary/amdgpu-cik-gl -c

Quoted 51 lines

This is with Linus master (-rc9) and latest drm-next-4.10-wip.

Thanks,
Shawn

Quoted 45 lines · EDT Shawn Starr

More info kernel boot options (some are unused like DAL)

BOOT_IMAGE=/boot/vmlinuz-4.9.0-rc2+ root=/dev/sda2 ro rhgb slub_debug=- cgroup_disable=memory console=tty0 console=ttyUSB0,9600n8 nmi_watchdog=0 audit=0 amdgpu.gartsize=8192 amdgpu.vm_size=16 amdgpu.exp_hw_support=1 rd.driver.blacklist=radeon amdgpu.dal=1 amdgpu.moverate=-1 video=LVDS-1:d video=VGA-0:e resume=/dev/sda7 pcie_aspm=force

Thanks,
Shawn

Quoted 33 lines · EDT Shawn Starr

Hello AMD folks,

Using latest LLVM trunk/Mesa master and latest changes from drm-next-4.10-wip available, I am unable to run piglit concurrently, causes
GPU graphics pipe to get pegged 100%.

[   86.777305] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c28770c
[   86.777306] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR

0x00100061 [ 86.777308] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0807700C [ 86.777309] VM fault
(0x0c, vmid 4) at page 1048673, read from 'SDM0' (0x53444d30) (119) [ 86.785159] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c44770c [
86.785163] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785164] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0407700C [ 86.785166] VM fault
(0x0c, vmid 2) at page 1048674, read from 'SDM0' (0x53444d30) (119) [ 86.785173] amdgpu 0000:01:00.0: GPU fault detected: 146 0x0c42770c [
86.785174] amdgpu 0000:01:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR
0x00100062 [ 86.785175] amdgpu 0000:01:00.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0207700C [ 86.785177] VM fault
(0x0c, vmid 1) at page 1048674, read from 'SDM0' (0x53444d30) (119)

Running: streaming-texture-leak -auto 3 times, wedges GPU.

Thanks,
Shawn
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx