linux-kernel - Re: [PATCH] mm/memcg: reorder retry checks for clarity in try_charge

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20251218072543.5071-1-kdipendra88@gmail.com>
Date: Thu, 18 Dec 2025 07:25:41 +0000
From: Dipendra Khadka <kdipendra88@...il.com>
To: hannes@...xchg.org
Cc: akpm@...ux-foundation.org,
	mhocko@...nel.org,
	roman.gushchin@...ux.dev,
	shakeel.butt@...ux.dev,
	muchun.song@...ux.dev,
	cgroups@...r.kernel.org,
	linux-mm@...ck.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm/memcg: reorder retry checks for clarity in try_charge_memcg

Hi Johannes,

Thank you for the feedback. Let me clarify the scenario this patch addresses.

The task_is_dying() check in try_charge_memcg() identifies when the
CURRENT task (the caller) is the OOM victim - not when some other 
process was killed.

Two scenarios:

1. Normal allocator triggers OOM:
  - Process A allocates → triggers OOM
  - Process B is killed (victim)
  - Process A continues with reset retries - task_is_dying() = false for A
  → Unchanged by my patch

2. Victim tries to allocate:
 - Process B (victim, TIF_MEMDIE set) tries to allocate
  - task_is_dying() = true
  - Current code: wastes retries on hopeless reclaims
  - My patch: exits immediately
  → Optimization for this case

The victim has three safety mechanisms that make the retries unnecessary:
1. oom_reaper proactively frees its memory
2. __alloc_pages_slowpath() grants reserves via oom_reserves_allowed()
3. Critical allocations with __GFP_NOFAIL still reach force: label

The retry loop for a dying victim is futile because:
- Reclaim won't help (victim is being killed to free memory!)
- Victim will exit regardless
- Just wastes CPU cycles

Would you like me to provide evidence showing the unnecessary retries,
or run specific tests to verify the safety mechanisms are sufficient?

Best Regards,
Dipendra