Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Bug]: Segfault with VAAPI using mpv on Meteor Lake #1887

Closed
q234rty opened this issue Dec 4, 2024 · 19 comments
Closed

[Bug]: Segfault with VAAPI using mpv on Meteor Lake #1887

q234rty opened this issue Dec 4, 2024 · 19 comments
Assignees
@MicroYY
Labels
VP Video Processing

Comments

@q234rty
Copy link

q234rty commented Dec 4, 2024

Which component impacted?

Decode

Is it regression? Good in old configuration?

Yes, it's good in old version

What happened?

  1. Use mpv --no-config --hwdec=vaapi to play a video with VA-API
  2. Observe that mpv crashes.

OS is Arch Linux, with intel-media-driver 24.4.3 and intel-gmmlib 22.5.2.
24.4.2 seems to be fine and my bisection points to 2c51f70

What's the usage scenario when you are seeing the problem?

Playback

What impacted?

No response

Debug Information

gdb bt:

Program terminated with signal SIGSEGV, Segmentation fault.
#0  vp::VpAllocator::DestroyVpSurface (this=0x7e27542f3b90, surface=@0x7e27830c8138: 0x7e275471d520, deferredDestroyed=false, flags=...)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/bufferMgr/vp_allocator.cpp:508
508	       int64_t currentSize = static_cast<int64_t>(surface->osSurface->OsResource.pGmmResInfo ? surface->osSurface->OsResource.pGmmResInfo->GetSizeAllocation() : 0);
[Current thread is 1 (Thread 0x7e27830ca6c0 (LWP 30826))]
(gdb) bt
#0  vp::VpAllocator::DestroyVpSurface (this=0x7e27542f3b90, surface=@0x7e27830c8138: 0x7e275471d520, deferredDestroyed=false, flags=...)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/bufferMgr/vp_allocator.cpp:508
#1  0x00007e2771f93313 in vp::SwFilterPipe::Clean (this=<optimized out>)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/feature_manager/sw_filter_pipe.cpp:409
#2  0x00007e2772350e43 in vp::VpObjAllocator<vp::SwFilterPipe>::Destory (this=0x7e27547066a0, obj=@0x7e2754493240: 0x7e27546d2080)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/feature_manager/vp_obj_factories.h:84
#3  vp::SwFilterPipeFactory::Destory (this=0x7e2754706698, swFilterPipe=@0x7e2754493240: 0x7e27546d2080)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/feature_manager/vp_obj_factories.cpp:298
#4  vp::VpPipeline::ExecuteSingleswFilterPipe(vp::VpSinglePipeContext*, vp::SwFilterPipe*&, vp::PacketPipe*, vp::VpFeatureManagerNext*) [clone .constprop.0]
    (this=this@entry=0x7e2754493270, singlePipeCtx=<optimized out>, pipe=@0x7e2754493240: 0x7e27546d2080, featureManagerNext=featureManagerNext@entry=0x7e27546ffa00, pPacketPipe=<optimized out>) at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/pipeline/vp_pipeline.cpp:692
#5  0x00007e2771fe7e21 in vp::VpPipeline::ExecuteVpPipeline (this=0x7e2754493270)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/pipeline/vp_pipeline.cpp:589
#6  0x00007e277200f035 in virtual thunk to VpPipelineAdapterXe_Lpm_Plus::Execute(VPHAL_RENDER_PARAMS*) ()
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/Xe_M_plus/Xe_LPM_plus/vp/hal/pipeline/vp_pipeline_adapter_xe_lpm_plus.h:73
#7  0x00007e2771fe8538 in VpPipelineAdapter::Render (this=0x7e2754484968, pcRenderParams=0x7e27546d1400)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/common/vp/hal/pipeline/vp_pipeline_adapter.cpp:177
#8  0x00007e2772014d1b in virtual thunk to VpPipelineAdapterXe_Lpm_Plus::Render(VPHAL_RENDER_PARAMS const*) ()
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/agnostic/Xe_M_plus/Xe_LPM_plus/vp/hal/pipeline/vp_pipeline_adapter_xe_lpm_plus.h:53
#9  0x00007e2771c8bc45 in DdiVpFunctions::EndPicture (this=<optimized out>, ctx=0x7e275432b8d0, context=<optimized out>)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/linux/common/vp/ddi/ddi_vp_functions.cpp:534
#10 0x00007e2771c90201 in DdiVpFunctions::ProcessPipeline
    (this=0x7e27544499f0, vaDrvCtx=0x7e275432b8d0, ctxID=3489660928, srcSurface=2, srcRect=0x7e27830c8ba8, dstSurface=1, dstRect=0x7e27830c8bb0)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/linux/common/vp/ddi/ddi_vp_functions.cpp:4472
#11 0x00007e2771c49b4a in MediaLibvaInterfaceNext::PutImage
    (ctx=0x7e275432b8d0, surface=1, image=<optimized out>, srcX=0, srcY=0, srcWidth=2, srcHeight=2, destX=0, destY=0, destWidth=2, destHeight=2)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_softlet/linux/common/ddi/media_libva_interface_next.cpp:1867
#12 MediaLibvaInterface::PutImage
    (ctx=0x7e275432b8d0, surface=1, image=<optimized out>, src_x=0, src_y=0, src_width=2, src_height=2, dest_x=0, dest_y=0, dest_width=2, dest_height=2)
    at /usr/src/debug/intel-media-driver/media-driver-intel-media-24.4.3/media_driver/linux/common/ddi/media_libva_interface.cpp:384
#13 0x00007e27a9abd83f in vaPutImage
    (dpy=0x7e275432b760, surface=surface@entry=1, image=0, src_x=src_x@entry=0, src_y=src_y@entry=0, src_width=2, src_height=2, dest_x=0, dest_y=0, dest_width=2, dest_height=2)
    at ../libva/va/va.c:1893
#14 0x00007e27aa63c37d in vaapi_unmap_frame (hwfc=0x7e27546e0c80, hwmap=<optimized out>) at libavutil/hwcontext_vaapi.c:782
#15 0x00007e27aa631dc3 in ff_hwframe_unmap (opaque=<optimized out>, data=0x7e2754450c00 "@PET'~") at libavutil/hwcontext.c:717
#16 0x00007e27aa61c9b7 in buffer_replace (dst=<optimized out>, src=0x0) at libavutil/buffer.c:133
#17 av_buffer_unref (buf=<optimized out>) at libavutil/buffer.c:144
#18 av_buffer_unref (buf=<optimized out>) at libavutil/buffer.c:139
#19 0x00007e27aa62bdbd in av_frame_unref (frame=0x7e2754451dc0) at libavutil/frame.c:618
#20 av_frame_unref (frame=0x7e2754451dc0) at libavutil/frame.c:610
#21 0x00007e27aa62bf1b in av_frame_free (frame=0x7e27830c8d50) at libavutil/frame.c:167
#22 0x00007e27aa63fbbe in vaapi_transfer_data_to (hwfc=<optimized out>, dst=0x7e275444c100, src=0x7e27544518c0) at libavutil/hwcontext_vaapi.c:1005
#23 0x00007e27aa63201a in av_hwframe_transfer_data (dst=0x7e275444c100, src=0x7e27544518c0, flags=flags@entry=0) at libavutil/hwcontext.c:481
#24 0x0000618063ee8db2 in mp_image_hw_upload (hw_img=0x7e2754450dd0, src=0x7e275476a000) at ../mpv/video/mp_image_pool.c:350
#25 mp_image_hw_upload (hw_img=0x7e2754450dd0, src=0x7e275476a000) at ../mpv/video/mp_image_pool.c:327
#26 mp_av_pool_image_hw_upload (hw_frames_ctx=<optimized out>, src=0x7e275476a000) at ../mpv/video/mp_image_pool.c:438
#27 0x0000618063f6a07f in try_format_upload (hw=hw@entry=0x7e27542c13f0, pixfmt=<optimized out>) at ../mpv/video/out/hwdec/hwdec_vaapi.c:436
#28 0x0000618063f6ab0c in try_format_config (hw=0x7e27542c13f0, hwconfig=0x7e27543424c0) at ../mpv/video/out/hwdec/hwdec_vaapi.c:504
#29 determine_working_formats (hw=0x7e27542c13f0) at ../mpv/video/out/hwdec/hwdec_vaapi.c:572
--Type <RET> for more, q to quit, c to continue without paging--c
#30 init (hw=0x7e27542c13f0) at ../mpv/video/out/hwdec/hwdec_vaapi.c:177
#31 0x0000618063ee2937 in ra_hwdec_load_driver
    (ra_ctx=<optimized out>, log=0x618080da77b0, global=<optimized out>, devs=0x7e27542c4a90, drv=0x618064071a80 <ra_hwdec_vaapi>, is_auto=false)
    at ../mpv/video/out/gpu/hwdec.c:104
#32 load_add_hwdec (ctx=0x7e27542bcad8, devs=0x7e27542c4a90, drv=0x618064071a80 <ra_hwdec_vaapi>, is_auto=false) at ../mpv/video/out/gpu/hwdec.c:236
#33 load_add_hwdec (ctx=0x7e27542bcad8, devs=0x7e27542c4a90, drv=0x618064071a80 <ra_hwdec_vaapi>, is_auto=<optimized out>) at ../mpv/video/out/gpu/hwdec.c:226
#34 0x0000618063eeaa62 in ra_hwdec_ctx_load_fmt (ctx=0x7e27542bcad8, devs=0x7e27542c4a90, params=0x7fff3baf5be0) at ../mpv/video/out/gpu/hwdec.c:332
#35 0x0000618063f05945 in gl_video_load_hwdecs_for_img_fmt (p=<optimized out>, devs=<optimized out>, params=0x7fff3baf5be0) at ../mpv/video/out/gpu/video.c:4369
#36 request_hwdec_api (vo=0x618080dc8960, data=0x7fff3baf5be0) at ../mpv/video/out/vo_gpu.c:134
#37 control (vo=0x618080dc8960, request=<optimized out>, data=0x7fff3baf5be0) at ../mpv/video/out/vo_gpu.c:203
#38 0x0000618063efadde in run_control (p=0x7fff3baf5b30) at ../mpv/video/out/vo.c:652
#39 0x0000618063e75cc2 in mp_dispatch_queue_process (queue=0x618080da30c0, timeout=<optimized out>) at ../mpv/misc/dispatch.c:300
#40 0x0000618063f0420e in vo_thread (ptr=<optimized out>) at ../mpv/video/out/vo.c:1100
#41 0x00007e27a990d39d in start_thread (arg=<optimized out>) at pthread_create.c:447
#42 0x00007e27a999249c in __GI___clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:78

Do you want to contribute a patch to fix the issue?

None
@database64128
Copy link

Same as #1879 (comment).

@q234rty
Copy link
Author

q234rty commented Dec 4, 2024

libva_trace:
libva_trace.log.153023.thd-0x00008296.log

@Jexu Jexu assigned MicroYY and unassigned Jexu and XinfengZhang Dec 5, 2024
@Jexu Jexu added the VP Video Processing label Dec 5, 2024
@intel-mediadev
Copy link
Contributor

Auto Created VSMGWL-77367 for further analysis.

@jiafengy1
Copy link
Contributor

@q234rty Could you share the complete cmd and video?

@q234rty
Copy link
Author

q234rty commented Dec 5, 2024

mpv --no-config --hwdec=vaapi spbtv_sample_bipbop_av1_960x540_25fps.mp4 where the mp4 is the one from https://github.com/SPBTV/video_av1_samples/blob/master/spbtv_sample_bipbop_av1_960x540_25fps.mp4

@jiafengy1
Copy link
Contributor

jiafengy1 commented Dec 6, 2024 •
edited
Loading

Hi @q234rty We can not reproduce this issue on our side with the same driver version and gmmlib version. I have some questions:

  1. What's the mpv version?
  2. what's the libva version?
  3. Is this issue always reproduced? or run to run reproduce?

@q234rty
Copy link
Author

q234rty commented Dec 6, 2024

mpv is on 0.39.0 and libva is on 2.22.0. For me, the issue is always reproducible.

@jiafengy1
Copy link
Contributor

mpv is on 0.39.0 and libva is on 2.22.0. For me, the issue is always reproducible.

Could you share the repro environment?

@q234rty
Copy link
Author

q234rty commented Dec 6, 2024

Could you share the repro environment?

Not sure what more can I share... The DE in use is KDE wayland if that's relevant. Have you tried reproducing using Arch Linux?

@database64128
Copy link

database64128 commented Dec 6, 2024 •
edited
Loading

For me, the issue is always reproducible.

Same here. And I'm using GNOME wayland on Arch Linux.

@nyanmisaka
Copy link
Contributor

One tip is that Arch linux has enabled LTO by default for the packages they build, which may not always be stable.

https://rfc.archlinux.page/0004-lto-by-default/

@jiafengy1
Copy link
Contributor

Hi, @q234rty We have root caused this issue, and our fix patch is in the formal process. @q234rty @nyanmisaka @database64128 Thanks for your support.

@kasper93 kasper93 mentioned this issue Dec 11, 2024
Closed
6 tasks
@jiafengy1
Copy link
Contributor

@q234rty patch checked in, you can have a try.

@nyanmisaka
Copy link
Contributor

Can you also point out which commit contains the fix? This way users can patch it themselves before a new tag appears.

@q234rty
Copy link
Author

q234rty commented Dec 14, 2024

I confirm the current master works though I'm not sure which commit fixed this...

@nyanmisaka
Copy link
Contributor

Great, then at least 25.1.0 will be safe.

@MicroYY
Copy link
Contributor

MicroYY commented Dec 16, 2024

7faa3bf is the fix commit.

@jiafengy1
Copy link
Contributor

@q234rty Now this issue has been resolved. Could you help close this issue?

@q234rty
Copy link
Author

q234rty commented Dec 16, 2024

I confirm patching 24.4.4 with just 7faa3bf also fixes this issue.

@q234rty q234rty closed this as completed Dec 16, 2024
@database64128 database64128 mentioned this issue Jan 6, 2025
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@MicroYY MicroYY

Labels
VP Video Processing
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
8 participants
@q234rty @Jexu @MicroYY @nyanmisaka @database64128 @XinfengZhang @intel-mediadev @jiafengy1