Re: mm: LTP/memcg testcase regression induced by 8cd7c588decf..66ce520bb7c2 series

From: Mike Galbraith
Date: Mon Nov 22 2021 - 01:47:14 EST

On Sun, 2021-11-21 at 14:51 +0100, Thorsten Leemhuis wrote:
> Hi, this is your Linux kernel regression tracker speaking.

This isn't necessarily a kernel regression, though the fact that
bystander tasks like ps can get snagged for ages as the testcase runs
at an amazingly glacial pace doesn't look particularly wonderful.

homer:/root # time ps lax
real 11m33.258s
user 0m0.004s
sys 0m0.008s
homer:/root #

The other memcg tests complete fine, though taking more time to do so.
Running all tests except memcg_regression, (and memcg_stress because
it's a 30 minute swap storm from hell as delivered), runtime increased
from 0m37s to 4m43s.

The problematic testcase is root intentionally shooting its own foot,
then checking for battle damage that occurred in days of yore, so it
now hurting in a different way isn't necessarily grounds to gripe at
the kernel.. even with the rather ugly side effects.. maybe <shrug>.

# Bug: The bug was, while forking mass processes, trigger memcgroup OOM,
# then NULL pointer dereference may be hit.
# Kernel: 2.6.25-rcX
# Links:
# Fix: commit e115f2d89253490fb2dbf304b627f8d908df26f1
mkdir memcg/0/
echo 0 > memcg/0/memory.limit_in_bytes


rmdir memcg/0/

if [ $? -eq 1 ]; then
tst_resm TPASS "no kernel bug was found"

I'm gonna just comment the little bugger out as obsolete and walk away.