1863885 - High heap-unclassified memory when looking at Cacti real-time graphs

Status: UNCONFIRMED

(Core :: Graphics: CanvasWebGL, defect, P2)

Description

[majedzouhairy]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/119.0

Steps to reproduce:

open 25 normal tabs and 48 pop up windows with java script which are real time graphing. this is on linux mint Victoria. This is on firefox 119.0.1 (64-bit) and all earlier versions.

Actual results:

firefox-bin uses 18% of memory of 32GB and typing, scrolling switching tabs , basically anything done in firefox is extremely slow.

Expected results:

use less cpu (4%) and less memory and scroll, switch, type a lot faster

Comment 1

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

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

Comment 2

[Jon Coppeard (:jonco)]

Can you provide instructions to reproduce this, such as the URLs you were looking at? This is not really actionable otherwise.

Also can you go to about:memory and press 'Measure' and provide the output here.

Comment 3

[majedzouhairy]

Attachment: memory-report.json.gz

to manually reproduce the problem, go to cacti.net, download and install a cacti server, add some network devices to monitor, then open real time graphs popups.

Comment 4

[Jon Coppeard (:jonco)]

Thanks.

The main process is showing nearly 5GB of heap-unclassified, so there's clearly a problem somewhere.

This is unlikely to be GC related so reclassifying this bug.

Are you able to get a DMD report for this? You would need to reproduce this in Firefox nightly run with the environment variable DMD=1 and use 'Save DMD output' in about:memory. This is described more here: https://firefox-source-docs.mozilla.org/performance/memory/dmd.html#building-and-running

Comment 5

[majedzouhairy]

i downloaded firefox nightly, but the save dmd report is greyed out, and it is unclear how to enable the environmental variable for linux.

Comment 6

[Jon Coppeard (:jonco)]

(In reply to majedzouhairy from comment #5)
You can run the binary from command line with: DMD=1 path/to/firefox

Comment 7

[majedzouhairy]

Attachment: unified-memory-report-1699595240-5254.json.gz

here is the dmd report

Comment 8

[Jon Coppeard (:jonco)]

(In reply to majedzouhairy from comment #7)
Thanks. Can you files named like dmd-XXXXXXXX-XXXXXX.json.gz?

Comment 9

[majedzouhairy]

Attachment: dmd-1699595240-5513.json.gz

Comment 10

[majedzouhairy]

Attachment: dmd-1699595240-5527.json.gz

Comment 11

[majedzouhairy]

Attachment: dmd-1699595240-5343.json.gz

Comment 12

[majedzouhairy]

Attachment: dmd-1699595240-5548.json.gz

Comment 13

[majedzouhairy]

Attachment: dmd-1699595240-5659.json.gz

Comment 14

[majedzouhairy]

Attachment: dmd-1699595240-5706.json.gz

Comment 15

[majedzouhairy]

Attachment: dmd-1699595240-9335.json.gz

Comment 16

[majedzouhairy]

Attachment: dmd-1699595240-9365.json.gz

Comment 17

[majedzouhairy]

Attachment: dmd-1699595240-5301.json.gz

Comment 18

[majedzouhairy]

Attachment: dmd-1699595240-5699.json.gz

Comment 19

[majedzouhairy]

Attachment: dmd-1699595240-5254.json.gz

Comment 20

[Jon Coppeard (:jonco)]

Thanks for uploading these.

Looking at dmd-1699595240-5254.json.gz with dmd.py, the largest unreported section is:

Unreported {
  603 blocks in heap block record 1 of 10,037
  3,730,087,936 bytes (3,727,631,917 requested / 2,456,019 slop)
  Individual block sizes: 16,781,312 x 158; 4,980,736; 4,198,400 x 120; 2,101,248 x 245; 1,052,672 x 40; 331,776 x 39
  82.85% of the heap (82.85% cumulative)
  91.43% of unreported (91.43% cumulative)
  Allocated at {
    #01: ???[/home/ameen/Downloads/firefox/firefox-bin +0x92125]
    #02: ???[/home/ameen/Downloads/firefox/firefox-bin +0x91c69]
    #03: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xd8ae25]
    #04: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xe128c1]
    #05: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xd7a9a1]
    #06: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xeb9f59]
    #07: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xebadf6]
    #08: ???[/lib/x86_64-linux-gnu/libnvidia-eglcore.so.470.223.02 +0xb668c1]
  }
}

Given the presence of libnvidia-eglcore.so in the stack I'm going to move this to Graphics.

Comment 21

[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 BugBot documentation.

Comment 22

[Bob Hood [:bhood]]

Does this behavior still reproduce with the current Nightly (Fx123)? Can we also please get your about:support output attached using a current Nightly build?

The STR are a bit difficult, but because of GL references in the heap list, dropping this into CanvasWebGL for an initial look.

Comment 23

[majedzouhairy]

Attachment: about:support

it's much faster in the nightly build!

Comment 24

[majedzouhairy]

it's much faster in the nightly build!