Arch Linux

Xelvet

Member

Registered: 2014-02-20

Posts: 82

Slab MemLeak & SUnreclaim

related to these threads -- https://bbs.archlinux.org/search.php?se … 1318316554

                total        used        free      shared  buff/cache   available
Mem:            15Gi        12Gi       2.2Gi       123Mi       731Mi       2.6Gi
Swap:           15Gi       1.8Gi        14Gi

yeah -- zen kernel & zswap & vega7i amdgpu & argon2id pbkdf & lvmcache..

cat /proc/meminfo
SUnreclaim:      8470632 kB

yet in my case slabtop reports -- dmaengine-unmap-2 being a culprit huh?

132540608 132419775  99%    0.06K 2070947       64   8283788K dmaengine-unmap-2

uptime -- 12d. what should be the next investig-step?

Last edited by Xelvet (2023-12-09 17:30:14)

vindicator

Member

Registered: 2016-02-25

Posts: 18

Re: Slab MemLeak & SUnreclaim

First off, that link of yours comes back with "Your search returned no hits."
I too have this issue (just had a reboot after

22158848 22142749  99%    0.06K 346232       64   1384928K dmaengine-unmap-2

) Yeah, 1.4GB.
I came across (https://www.alibabacloud.com/help/en/al … cb9144dm1k) that looks like it would help narrowing down the issue... BUT, it doesn't look like it's applicable to "dmaengine-unmap*" as there's no such entry within "/sys/kernel/tracing/events".

EDIT0: That was from a 47 day uptime. I'm currently at 9 hours 41 minutes with:

65216  65216 100%    0.06K   1019       64      4076K dmaengine-unmap-2

EDIT1: I wonder how much it has to do with USB... specifically a TV tuner that may flood with errors from a poor OTA signal ("Transport error indicator (total 119)"; (can be in the (tens-/)thousands)).

EDIT2: Nope, unrelated. I came back after sleeping and the tuner is capturing now and producing a bunch of "Transport error indicator (total 2274)", but the dmaengine active amount hasn't changed from 64303 which had been 65274 before.

Last edited by vindicator (2024-11-12 02:47:44)