Open Bug 1346222 Opened 7 years ago Updated 2 years ago

ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died, (Intel Haswell on Linux)

Categories

(Core :: Graphics, defect, P3)

54 Branch
defect

Tracking

()

UNCONFIRMED
Tracking Status
platform-rel --- -

People

(Reporter: me, Unassigned)

References

Details

(Whiteboard: [platform-rel-Intel] [gfx-noted])

Attachments

(4 files)

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/57.0.2987.98 Safari/537.36

Steps to reproduce:

After Updating Firefox Developer Edition to 54.0a2 (2017-03-10) (64-bit)

None of the pages are shown, they load, the title is even updated, but the page is not rendered.
I tried safe-mode, it did not work the same white screen. 

Update:
Refreshing the Browser works:
firefox --safe-mode > refresh developer edition


Actual results:

These are the errors on commandline:
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.
[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died.



Expected results:

Render the page.
Component: Untriaged → Graphics
Product: Firefox → Core
Attached file about.support
To have more information available
Summary: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died → ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died, (Intel Hasswell on Linux)
Summary: ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died, (Intel Hasswell on Linux) → ClientLayerManager::BeginTransaction with IPC channel down. GPU process may have died, (Intel Haswell on Linux)
platform-rel: --- → ?
Whiteboard: [platform-rel-Intel]
I've noticed this happening on my Ubuntu 16.04.2 LTS x64 VM several times today while I was running a compiled m-c asan build. I usually run into these errors when I initially launch the browser and quickly open several other tabs at the same time. Unfortunately I don't really have solid STR.
See Also: → 1352545
The messages are a red herring; David, did we clear those up?  We're not trying to run accelerated, there is no GPU process, so these messages sometimes confuse the triage.

With the safe mode still running into problems, it should not be platform specific.
platform-rel: ? → -
Flags: needinfo?(dvander)
(In reply to Milan Sreckovic [:milan] from comment #3)
> The messages are a red herring; David, did we clear those up?  We're not
> trying to run accelerated, there is no GPU process, so these messages
> sometimes confuse the triage.
> 
> With the safe mode still running into problems, it should not be platform
> specific.

I think Kats may have fixed the red-herring message with bug 1351384 already. The problem was webextensions accidentally uploading layers to a disconnected compositor.
Flags: needinfo?(dvander)
A reddit user [1] seems to have run into this issue with the Fx 54 release:

> Just leaving a feedback here: I'm seeing a lot of messages in 
> the "Failure Log" section of about:support. Those are: 
> CP+[GFX1-]: ClientLayerManager::BeginTransaction with IPC channel
> down. GPU process may have died.
> 
> Anyone with the same problem? It's a fresh profile of v54 and the 
> dom.ipc.processCount is set to 1.

[1] https://www.reddit.com/r/firefox/comments/6h3cof/psa_not_all_firefox_54_users_will_have_multie10s/divuh27/
Attached file about:support
I'm seeing the same issue too, Firefox 54 on MacbookPro 11,2, macOS 10.12.5.

Attached output of about:support. Let me know what other information you need.
Kats, your patch from bug 1351384 seems to be in 54, but we're getting more reports - does that make sense, or is it something else?
Flags: needinfo?(bugmail)
I think this is something else. Seems like the IPC channel between the content process and the compositor thread in the UI process (since this is OS X, there's no GPU process) is down. WebRender is not in the build, and it's properly trying to use a ClientLayerManager (as opposed to a WebRenderLayerManager) so I don't think this is WR-related.
Flags: needinfo?(bugmail)
Attached file about:support2
Very strange.

I have went into Settings, then disable "Use hardware acceleration when available", restart Firefox, then reenable "Use hardware acceleration when available". No more GPU error messages.

Attached new about:support output as about:support2.
I'm seeing this issue here two: My console gets flooded with this message: ~5 messages per Minute.

FF54, Linux Arch on Lenovo T460s (i5-6300U CPU, Intel HD Graphics 520 Skylake GT2)
(In reply to Billy K from comment #9)
> Created attachment 8878379 [details]
> about:support2
> 
> Very strange.
> 
> I have went into Settings, then disable "Use hardware acceleration when
> available", restart Firefox, then reenable "Use hardware acceleration when
> available". No more GPU error messages.
> 
> Attached new about:support output as about:support2.

This solved the issue for me too.
Attached file about:support
Having the same problem on radeon. 
I have also tried to forcefully disable hw acceleration (I mean, not that by default there's a lot enabled) and e10s - still happens.
Whiteboard: [platform-rel-Intel] → [platform-rel-Intel] [gfx-noted]
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: