linux-kernel - Re: [PATCH] mm/swap: annotate data races for lru_rotate

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-next>] [day] [month] [year] [list]

Message-Id: <463BBB2A-8F9A-4CF1-80AE-677ACD21A3C6@lca.pw>
Date:   Fri, 28 Feb 2020 06:30:26 -0500
From:   Qian Cai <cai@....pw>
To:     Marco Elver <elver@...gle.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Linux Memory Management List <linux-mm@...ck.org>,
        LKML <linux-kernel@...r.kernel.org>,
        kasan-dev <kasan-dev@...glegroups.com>
Subject: Re: [PATCH] mm/swap: annotate data races for lru_rotate_pvecs

> On Feb 28, 2020, at 5:49 AM, Marco Elver <elver@...gle.com> wrote:
> 
> Note that, the fact that the writer has local interrupts disabled for
> the write is irrelevant because it's the interrupt that triggered
> while the read was happening that led to the concurrent write.

I was just to explain that concurrent writers are rather unlikely as people may ask.

> 
> I assume you ran this with CONFIG_KCSAN_INTERRUPT_WATCHER=y?  The
> option is disabled by default (see its help-text). I don't know if we
> want to deal with data races due to interrupts right now, especially
> those that just result in 'data_race' annotations. Thoughts?

Yes, I somehow got quite a bit clean runs lately thanks to the fix/annotations efforts for the last a few weeks (still struggling with the flags things a bit), so I am naturally expanding the testing coverage here.

Right now the bottleneck is rather some subsystem maintainers are not so keen to deal with data races (looking forward to seeing more education opportunities for all), but the MM subsystem is not one of them.