1820416 - VA-API with ffvpx doesn't work anymore

Status: VERIFIED FIXED

(Core :: Audio/Video: Playback, defect, P2)

Description

[tgn-ff]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/112.0

Steps to reproduce:

Play a VP9 video (on supported hardware).

Actual results:

VA-API doesn't work. I also saw a higher than usual memory usage in the RDD process after a long session, even after closing all media tabs. Almost all if it was under heap-unclassified, though.

After setting media.ffvpx.enabled:false, VA-API works.

Expected results:

This is a regression caused by Bug 1819374.

https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=0681bdf9275feb82dc9c082a7aeac76096335786&tochange=b59cfb24a83c1a9988a8271e6a7b926f7cd770d9

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

:padenot, since you are the author of the regressor, bug 1819374, could you take a look? Also, could you set the severity field?

For more information, please visit auto_nag documentation.

Comment 2

[Paul Adenot (:padenot)]

Reporter, thanks for filing and sorry for having broken this, we don't have CI for VAAPI yet. Would you mind getting us logs:

MOZ_LOG=PlatformMediaDecoder:5 ./firefox

after having closed Firefox (so that it starts a new instance), and attempting to decode something with ffvpx enabled?

Comment 3

[tgn-ff]

Of course. You mean with PlatformDecoderModule, right? PlatformMediaDecoder doesn't output anything. I will attach logs from both the release and the nightly channels.

Comment 4

[tgn-ff]

Attachment: nightly_with_ffvpx_enabled_broken.txt

Comment 5

[tgn-ff]

Attachment: release_with_ffvpx_enabled_working.txt

Comment 6

[Paul Adenot (:padenot)]

Yes, of course, thanks for doing the right thing. I'm having a look at your logs.

Comment 7

[Paul Adenot (:padenot)]

Stransky, I've updated to ffmpeg 6.0 in preparation for other work, and I'm not seeing https://searchfox.org/mozilla-central/source/media/ffvpx/libavutil/hwcontext_vaapi.c#418 in the logs, which is weird, do you have an idea, or a machine to debug this? Maybe you can do an rr recording or something that I could then debug?

I'm currently in the process of refreshing my hardware, but I don't have the right machine at the moment.

Comment 8

[Martin Stránský [:stransky] (ni? me)]

I do see that too. No idea yet.

Comment 9

[Alastor Wu [:alwu]]

Would you mind help us capture the log with Dmabuf:5 as well? Thanks!

Comment 10

[tgn-ff]

Attachment: dmabuf_5_release_working.txt

Not at all. I'm attaching the log from the release since it's long, but the log from nightly is just the following lines:

[Parent 18948: Main Thread]: D/Dmabuf Using DRM device /dev/dri/renderD128
[Parent 18948: Main Thread]: D/Dmabuf nsDMABufDevice::Configure()
[Parent 18948: Main Thread]: D/Dmabuf Loading DMABuf system library libgbm.so.1 ...
[Parent 18948: Main Thread]: D/Dmabuf DMABuf is enabled
[ERROR glean_core] Error setting metrics feature config: Json(Error("EOF while parsing a value", line: 1, column: 0))
[RDD 19325: MediaPDecoder #1]: D/Dmabuf Using DRM device /dev/dri/renderD128

Only the last line was printed after I started the video.

Comment 11

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1819374

Comment 12

[Martin Stránský [:stransky] (ni? me)]

Attachment: Bug 1820416 Use correct FFVPX headers from ffmpeg 6.0 r?alwu

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/4e3d5ec4e2b1
Use correct FFVPX headers from ffmpeg 6.0 r=padenot

Comment 15

[Natalia Csoregi [:nataliaCs]]

https://hg.mozilla.org/mozilla-central/rev/4e3d5ec4e2b1