imx8 gpu crash when playing video

[jiangyaqiang]

HI  anyone:

I face a problem of gpu crash when playing video (the testing video is 1080p ,hdmi output).

platform: mx8mq , fsl-yocto-L4.9.51_mx8mq-ga (xwaylan).

the phenomenon happens about every one hour, at most time it had no log, just system hang up quietly .

here is part log(complete log in the attachment):

====== AIUR: 4.3.4 build on Aug 23 2018 15:19:16. ======
Core: MPEG4PARSER_06.13.06 build on Feb 6 2018 07:43:39
file: /usr/lib/imx-mm/parser/lib_mp4_parser_arm_elinux.so.3.2
------------------------
Track 00 [video_0] Enabled
Duration: 0:01:00.143411000
Language: eng
Mime:
video/x-h264, parsed=(boolean)true, alignment=(string)au, stream-format=(string)avc, width=(int)1920, height=(int)1080, framerate=(fraction)24000/1001, codec_data=(buffer)014d0033ffe10031674d4033965280f0044fcb80b4400000fa40002ee03840000f03c0001b775fc638c20000781e0000dbbafe31c3b430632c01000568e9093520
------------------------
====== VPUDEC: 4.3.4 build on Aug 23 2018 15:19:29. ======
wrapper: 3.0.0 (VPUWRAPPER_ARM64_LINUX Build on Aug 23 2018 14:27:18)
vpulib: 1.1.1
firmware: 1.1.1.65535
------------------------
Track 01 [audio_0] Enabled
Duration: 0:01:00.186122000
Language: eng
Mime:
audio/mpeg, mpegversion=(int)4, channels=(int)2, rate=(int)44100, bitrate=(int)61615, stream-format=(string)raw, codec_data=(buffer)12100000
------------------------
[12926.162132] function:imx_hifi_hw_params line:85 params_rate(params):44100
[12926.169197] sample_rate:44100 sample_format:2
[12926.175787] function:imx_hifi_hw_params line:120 priv->clk_frequency :24576000
[12926.183453] rt5640 0-001c: bclk is 1411200Hz and lrck is 44100Hz
[12926.189478] rt5640 0-001c: bclk_ms is 0 and pre_div is 1 for iis 0

====== BEEP: 4.3.4 build on Aug 23 2018 15:19:21. ======
Core: AAC decoder Wrapper build on Dec 7 2017 18:13:51
file: /usr/lib/imx-mm/audio-codec/wrap/lib_aacd_wrap_arm_elinux.so.3
CODEC: BLN_MAD-MMCODECS_AACD_ARM_03.09.00_ARMV8 build on Sep 20 2017 15:02:50.
Redistribute latency...
Pipeline is PREROLLED ...
Setting pipeline to PLAYING ...
New clock: GstAudioSinkClock
Got EOS from element "playbin0".
Execution ended after 0:00:55.297515425
Setting pipeline to PAUSED ...
Setting pipeline to READY ...
Setting pipeline to NULL ...
Freeing pipeline ...
Setting pipeline to PAUSED ...
[13009.799393] **************************
[13009.803159] *** GPU DRV CONFIG ***
[13009.807159] **************************
[13009.811072] Galcore version 6.2.4.150331
[13009.815103] Galcore options:
[13009.818113] irqLine = 49
[13009.821759] registerMemBase = 0x38000000
[13009.826102] registerMemSize = 0x00040000
[13009.830440] contiguousSize = 0x08000000
[13009.834781] contiguousBase = 0x00000000
[13009.839120] externalSize = 0x00000000
[13009.843463] externalBase = 0x00000000
[13009.847802] bankSize = 0x00000000
[13009.852143] fastClear = -1
[13009.855788] compression = 13
[13009.859434] signal = 48
[13009.863010] powerManagement = 1
[13009.866582] baseAddress = 0x40000000
[13009.871058] physSize = 0xC0000000
[13009.875419] logFileSize = 0 KB
[13009.879270] recovery = 0
[13009.882843] stuckDump = 0
[13009.886412] gpuProfiler = 0
[13009.889991] irqs = [13009.893289] -1,
-1, [13009.895561] -1,
-1, [13009.897763] -1,
-1, [13009.900032] -1,
-1, [13009.902233] -1,

[r_trivedi123]

Hi, Even this issues still persist in kernel v5.15.71 in NXP BSP - kirkstone branch.

We faced it with one of our customer for sway compositor.

[   27.464790] etnaviv-gpu 38000000.gpu: AXI bus error
[   27.469745] etnaviv-gpu 38000000.gpu: MMU fault status 0x00000000
[   27.970450] etnaviv-gpu 38000000.gpu: recover hung GPU!

However we were able to resolve the issue and now its running smooth with the weston and sway compositor with 4GiB RAM usage.

As this was done for the customer we are unable to share the details, in case of support please reach-out to us on rutvij.trivedi@siliconsignals.io or visit www.siliconsignals.io.

--
BR

[Bio_TICFSL]

Hi Jiang,

There must be a memory leak somewhere, Do this happens with video only?

Did you try to change the gpu memory?

Regards

[jiangyaqiang]

HI Bio:

Yes, currently , it happened only when playing video.

GPU memory ? you means CMA size ? I did change it, but it happened again(It did could not duplicate less).

The device DDR is 2GB, and CMA is setted to 512MB.

江亚强

软件工程师

Shenzhen Huameishi Technology Co., Ltd

深圳市华美视科技有限公司

深圳市南山区科苑路6号科技园工业大厦605

Tel：0755-26037882-616

Mail:yaqiang.jiang@huameishi.com