linux-kernel - Re: [syzbot] INFO: rcu detected stall in syscall_exit_to_user

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: <87ee9qb2p5.ffs@tglx>
Date:   Wed, 15 Sep 2021 11:36:22 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     paulmck@...nel.org, Dmitry Vyukov <dvyukov@...gle.com>
Cc:     Hillf Danton <hdanton@...a.com>,
        syzbot <syzbot+0e964fad69a9c462bc1e@...kaller.appspotmail.com>,
        linux-kernel@...r.kernel.org, syzkaller-bugs@...glegroups.com,
        Peter Zijlstra <peterz@...radead.org>,
        kasan-dev <kasan-dev@...glegroups.com>
Subject: Re: [syzbot] INFO: rcu detected stall in syscall_exit_to_user_mode

On Tue, Sep 14 2021 at 11:31, Paul E. McKenney wrote:
> On Tue, Sep 14, 2021 at 08:00:04PM +0200, Dmitry Vyukov wrote:
>> If I understand it correctly the timer is not actually set up as
>> periodic, but rather each callback invocation arms it again. Setting
>> up a timer for 1 ns _once_ (or few times) is probably fine (right?),
>> so the check needs to be somewhat more elaborate and detect "infinite"
>> rearming.
>
> If it were practical, I would suggest checking for a CPU never actually
> executing any instructions in the interrupted context.  The old-school
> way of doing this was to check the amount of time spent interrupted,
> perhaps adding some guess at interrupt entry/exit overhead.  Is there
> a better new-school way?

Set NR_CPUS=0 and if then any executed instruction is observed the bug
is pretty obvious, isn't it?