1792738 - Memory leak with the latest version of Firefox

Status: RESOLVED INCOMPLETE

(Core :: Graphics, defect, P3)

Description

[susreeni]

Attachment: memory-report.json.gz

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:105.0) Gecko/20100101 Firefox/105.0

Steps to reproduce:

Pretty much nothing different

Actual results:

Memory usage keeps creeping up

Expected results:

Memory usage shouldn't keep ramping up until all RAM is depleted

Comment 1

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

The Bugbug bot thinks this bug should belong to the 'Core::Performance' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[Bas Schouten (:bas.schouten)]

The only significant memory usage I see in the attached memory report is inside graphics. Moving component although I'm not sure there's enough information in here for this to be actionable.

Comment 3

[alfonsmcbobr]

Attachment: memory-report.json.gz

Memory report after opening a single tab and playing few 10min videos. Process manager is reporting 7+GB used by GPU process.

Comment 4

[alfonsmcbobr]

Attachment: About:support

Comment 5

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

The severity field is not set for this bug.
:bhood, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 6

[Bob Hood [:bhood]]

Sotaro, any chance you can reproduce this?

Comment 7

[Sotaro Ikeda [:sotaro]]

The symptom seemed similar to Bug 1799716.

Comment 8

[Sotaro Ikeda [:sotaro]]

Hi susreeni and alfonsmcbobr, can you check if the problem happens with "pref gfx.canvas.remote = false"? The prf could be changed from about:config. Firefox needs to restart after changing the pref. Thank you.

Comment 9

[Sotaro Ikeda [:sotaro]]

From Bug 1798206 Comment 8, there is a case that YouTube uses canvas.

Comment 10

[alfonsmcbobr]

(In reply to Sotaro Ikeda [:sotaro] from comment #8)

Hi susreeni and alfonsmcbobr, can you check if the problem happens with "pref gfx.canvas.remote = false"? The prf could be changed from about:config. Firefox needs to restart after changing the pref. Thank you.

Yes, it happens even with gfx.canvas.remote = false.

Comment 11

[tcbbdddd]

I have met this issue. I think it is not related to canvas. To trigger it, just view some page with a lot of images, then you'll find that the GPU process eats >5G mem. Most of the mem are reported as gpu-committed. Here's an example:

      0.00 MB ── d3d11-shared-textures
      0.00 MB ── gfx-d2d-vram-draw-target
      0.00 MB ── gfx-d2d-vram-source-surface
  6,351.98 MB ── gpu-committed
  1,540.23 MB ── gpu-dedicated
     13.35 MB ── gpu-shared
  1,156.41 MB ── heap-allocated
      1.00 MB ── heap-chunksize
  1,483.00 MB ── heap-mapped
  7,676.11 MB ── private
  5,411.42 MB ── resident
  5,230.66 MB ── resident-unique
      0.00 MB ── shmem-allocated
    159.27 MB ── shmem-mapped
     36.47 MB ── system-heap-allocated
            0 ── unresolved-ipc-responses

2,110,981.74 MB ── vsize
129,700,696.98 MB ── vsize-max-contiguous

Comment 12

[tcbbdddd]

Attachment: tcbbd-about-support.txt

Comment 13

[tcbbdddd]

Attachment: tcbbd-memory-report.json.gz

Comment 14

[tcbbdddd]

Well, now I find that the "resident" or "resident-unique" is the amout of memory actually consumed by GPU process (at least it is what is shown in Task Manager). So gpu-committed is what is sent to GPU mem? Anyway, I've uploaded the memory report so that you can anaylze this bug.

Comment 15

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

A needinfo is requested from the reporter, however, the reporter is inactive on Bugzilla. Given that the bug is still UNCONFIRMED, closing the bug as incomplete.

For more information, please visit BugBot documentation.