DRM-related audio dropouts when playing unprotected content, even in external processes
Categories
(Core :: Audio/Video: Playback, defect, P3)
Tracking
()
People
(Reporter: mdl, Unassigned)
Details
Attachments
(1 file)
|
43.78 KB,
application/json
|
Details |
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:93.0) Gecko/20100101 Firefox/93.0
Steps to reproduce:
OS: Windows 10 1607 x64
Go to https://www.youtube.com/watch?v=wZ1gEDquuyQ and play the video at 1080p60 (though the resolution shouldn't matter).
Actual results:
Audio will intermittently drop out not only in Firefox (and can be correlated as interrupt to DPC latency spikes in LatencyMon), but also, it seems that even having Firefox open at all will cause audio dropouts in other processes.
Bisected:
app_name: firefox
build_date: 2021-05-03 14:54:42.234000
build_file: C:\Users\Administrator.mozilla\mozregression\persist\4153e02c1d0e-shippable--mozilla-release--target.zip
build_type: integration
build_url: https://firefox-ci-tc.services.mozilla.com/api/queue/v1/task/GuZNORx3RDuLtiEEz8XUDg/runs/0/artifacts/public%2Fbuild%2Ftarget.zip
changeset: 4153e02c1d0e97a623518d558580b9c4739a036e
pushlog_url: https://hg.mozilla.org/releases/mozilla-release/pushloghtml?fromchange=b9c4b17c51c460c426d5ed3ec5c11efdff50a482&tochange=4153e02c1d0e97a623518d558580b9c4739a036e
repo_name: mozilla-release
repo_url: https://hg.mozilla.org/releases/mozilla-release
task_id: GuZNORx3RDuLtiEEz8XUDg
Bug 1707559 - Stub out QueryOutputProtectionStatus. r=jbauman a=RyanVM
Differential Revision: https://phabricator.services.mozilla.com/D113361
2021-09-26T19:05:58.906000: DEBUG : Did not find a branch, checking all integration branches
2021-09-26T19:05:58.907000: INFO : The bisection is done.
2021-09-26T19:05:58.908000: INFO : Stopped
Expected results:
Not having wasted multiple days of my time, unpaid, for DRM I don't use in an "open web" browser would be ideal. Failing that, I'd just like the bug to be fixed.
Bonus: you can trigger even more frequent dropouts if you go to about:processes while the video is playing.
|
||
Comment 2•4 years ago
|
||
The Bugbug bot thinks this bug should belong to the 'Core::Audio/Video: Playback' component, and is moving the bug to that component. Please revert this change in case you think the bot is wrong.
Reporter, could you capture another profile and upload it using the button in the profiler? I don't think the one in comment 3 is publicly accessible.
Could you navigate to about:config and set the media.eme.enabled pref to false, and see if the problem still reproduces?
Oops, I'm an idiot; didn't realize I hadn't actually uploaded the profile. Here you go: https://share.firefox.dev/3CVJsQI
The issue still occurs with media.eme.enabled set to false.
Thanks. For other Mozillians, I'm not sure about the original symbolification on that profile, there's a lot of stuff marked as workerlz4_compress which seems odd. https://share.firefox.dev/39NOj9U is a re-symbolicated version.
Could you also upload the information from your about:support onto this bug?
Setting media.eme.enabled to false should prevent the code in https://phabricator.services.mozilla.com/D113361 being called. It may be worth doing another bisection, if you have time, to identify if something else is triggering the issue.
Does the issue take place if you run Firefox with a fresh profile?
:kinetik, any ideas on how we could debug this further, given the system wide audio specific symptom?
Triaging based on current info and that we've been unable to repro so far.
|
||
Comment 10•4 years ago
|
||
Which driver or process does LatencyMon attribute the DPC spikes to? What's the highest DPC time you're seeing in this situation?
If you could please try disabling hardware acceleration in about:preferences (disable "Use recommended performance settings" and "Use hardware acceleration when available" under "Performance") then restart and see if the issue is still present, that'd also be helpful.
(In reply to Bryce Seager van Dyk (:bryce) from comment #7)
Setting
media.eme.enabledto false should prevent the code in https://phabricator.services.mozilla.com/D113361 being called. It may be worth doing another bisection, if you have time, to identify if something else is triggering the issue.
Also, if you have a chance to do this, it'd be very useful.
|
Reporter | |
Comment 11•4 years ago
|
||
LatencyMon was attributing it to the nvidia driver, but this problem did not exist before I upgraded Firefox versions, so I don't think it has to do with it.
I did try to bisect this and got inconclusive results, so in an attempt to remove other factors, I uninstalled a recent Windows update, which immediately resulted in being unable to boot that Windows installation on bare metal--typical Windows living, why should I have trusted that would work--so I'm not sure I can triage this much farther; I've been trying to get it booting outside VMware since then but no luck.
Setting media.eme.enabled to false did not help, so it's possible DRM is not the cause. I apologize for being sassy in my initial report. I would like to look into this farther but it might be a dead end at this point.
|
|
Reporter | |
Comment 12•4 years ago
|
||
Scratch that, after a week of fiddling I made it boot again.
The issue I'm seeing isn't high ISR or DPC execution time, but rather "measured interrupt to process latency". The LatencyMon documentation suggests that this is a meaningless measure on modern Windows, but it correlates perfectly with audio drop-outs, so I'm not sure what to make of that. (I emailed the LatencyMon people inquiring about this and got no response.)
Unfortunately disabling hardware acceleration and the recommended performance settings doesn't seem to help.
Description
•