linux-kernel - Re: [PATCH RFC 09/19] slab: add optimized sheaf refill from partial list

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: <2a95b2db-c487-440c-b95c-35549c8f5ba6@suse.cz>
Date: Thu, 30 Oct 2025 14:18:46 +0100
From: Vlastimil Babka <vbabka@...e.cz>
To: Harry Yoo <harry.yoo@...cle.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
 Christoph Lameter <cl@...two.org>, David Rientjes <rientjes@...gle.com>,
 Roman Gushchin <roman.gushchin@...ux.dev>,
 Uladzislau Rezki <urezki@...il.com>,
 "Liam R. Howlett" <Liam.Howlett@...cle.com>,
 Suren Baghdasaryan <surenb@...gle.com>,
 Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
 Alexei Starovoitov <ast@...nel.org>, linux-mm@...ck.org,
 linux-kernel@...r.kernel.org, linux-rt-devel@...ts.linux.dev,
 bpf@...r.kernel.org, kasan-dev@...glegroups.com
Subject: Re: [PATCH RFC 09/19] slab: add optimized sheaf refill from partial
 list

On 10/30/25 01:07, Harry Yoo wrote:
> On Wed, Oct 29, 2025 at 09:48:27PM +0100, Vlastimil Babka wrote:
>> (side note: gfpflags_allow_blocking() might be too conservative now that
>> sheafs will be the only caching layer, that condition could be perhaps
>> changed to gfpflags_allow_spinning() to allow some cheap refill).
> 
> Sounds good to me.

Hm now I realized the gfpflags_allow_blocking() check is there to make sure
we can take the local lock without trylock after obtaining a full sheaf, so
we can install it - because it should mean we're not in an interrupt
context. The fact we already succeeded trylock earlier should be enough, but
we'd run again into inventing ugly tricks to make lockdep happy.

Or we use trylock and have failure paths that are only possible to hit on RT
in practice...