Kernel 4.14 Memory Managment Issue

Looks like with 4.14.32-1 running firefox or firefox beta(& others?) will consume all memory till freezing & cordumping occurs. Switch to 4.4.126-1 and notice the difference in memory stability. This may be exhibited on stable but I haven't currently checked.
Running P4 with 2GB memory.

I doubt this is kernel related
Anyway you can check for kernel memory leaks here

Just confirmed this issue CAN be recreated on stable also.

Looking for someone to confirm or refute this symptom finding if possible.
I just run firefox or firefox beta with gnome system monitor & monitor memory consumption.

It looks like both 4.9 & 4.14 run in a high memory usage mode but only 4.14 falters with memory management. 4.4 runs using less memory & doesn't have any issues.

Maybe this sheds some light?
https://bugs.archlinux.org/task/56574

1 Like

The patch mentioned in the thread you linked, and here: https://bugzilla.netfilter.org/show_bug.cgi?id=1212
applies cleanly to current 4.14.32.

I haven't tested it yet though. Also, I'm on 64-bit.

@jonathon What does this mean for i686 if anything? Have you tested this? I remember @philm built 4.14 for i686 because of a bfq scheduler problem/issue. I think.

So first things first. The linked bug report specifically mentions an issue when running ipset. Are you running ipset? If not, it's a different issue.

I don't have any issues running Firefox with 4.14 in a 32-bit VM or on an EeePC 901 so it's a local hardware, configuration, or software issue. IIRC the 17.1.7-rc1 installer image uses kernel 4.14 so you could try replicating using that to narrow it down somewhat.

1 Like

The original post doesn't reference ipset..so do what you will & I'll run a kernel that doesn't have an issue.

The author of the bug report mentions it.

https://bugs.archlinux.org/task/56574#comment164096

I have systemd core dumps from firefox & firefox beta showing sig(11) which seems to be a stack overflow so...?

Have you tried with a clean browser profile?

1 Like

No but I will try recreating the issue with chrominum to rule out firefox though!

I also don't see this problem with either 4.14 or 4.9. Maybe those kernels don't do well with your hardware. Stick with an older one. That's what they're for.

Seems to be happening on testing with 4.14:

[dad@dad3000 ~]$ coredumpctl list
TIME PID UID GID SIG COREFILE EXE
Wed 2018-04-04 20:10:14 EDT 877 1000 1000 11 present /usr/lib/firefox-beta/firefox
[dad@dad3000 ~]$ coredumpctl info 877
PID: 877 (Web Content)
UID: 1000 (dad)
GID: 1000 (dad)
Signal: 11 (SEGV)
Timestamp: Wed 2018-04-04 20:05:34 EDT (1h 48min ago)
Command Line: /usr/lib/firefox-beta/firefox -contentproc -childID 5 -isForBrowser -intPrefs 6:10|4>
Executable: /usr/lib/firefox-beta/firefox
Control Group: /user.slice/user-1000.slice/session-c1.scope
Unit: session-c1.scope
Slice: user-1000.slice
Session: c1
Owner UID: 1000 (dad)
Boot ID: 540fd93834344aeba1c5398d9f989eec
Machine ID: 314b7397f9b948e6859a08db799cd657
Hostname: dad3000
Storage: /var/lib/systemd/coredump/core.Web\x20Content.1000.540fd93834344aeba1c5398d9f989eec.>
Message: Process 877 (Web Content) of user 1000 dumped core.

I can't launch chromium in testing due to icu mismatch.

[dad@dad3000 ~]$ chromium
/usr/lib/chromium/chromium: error while loading shared libraries: libicui18n.so.60: cannot open shared object file: No such file or directory
[dad@dad3000 ~]$

Is that on x32-stable or x32-testing?

Have you tried with a clean browser profile?

If you're serious about troubleshooting you need to limit the number of variables at one time.

1 Like

testing

New profile created with firefox --ProfileManager on 4.14.32-1 unstable:

       PID: 3681 (firefox)
       UID: 1000 (dad)
       GID: 1000 (dad)
    Signal: 11 (SEGV)
 Timestamp: Fri 2018-04-06 00:28:05 EDT (2min 2s ago)

Command Line: /usr/lib/firefox/firefox
Executable: /usr/lib/firefox/firefox
Control Group: /user.slice/user-1000.slice/session-c1.scope
Unit: session-c1.scope
Slice: user-1000.slice
Session: c1
Owner UID: 1000 (dad)
Boot ID: 35e6109a4fa046bebb82d1e47be148c0
Machine ID: 314b7397f9b948e6859a08db799cd657
Hostname: dad3000
Storage: /var/lib/systemd/coredump/core.firefox.1000.35e6109a4fa046bebb82d1e47be148c0.3681.15>
Message: Process 3681 (firefox) of user 1000 dumped core.

Forum kindly sponsored by