linux-kernel - Re: [syzbot] WARNING in kthread_is_per

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 for Android: free password hash cracker in your pocket

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

Date:   Tue, 20 Apr 2021 11:17:15 +0100
From:   Valentin Schneider <valentin.schneider@....com>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     syzbot <syzbot+9362b31a2e0cad8b749d@...kaller.appspotmail.com>,
        bp@...en8.de, dwmw@...zon.co.uk, hpa@...or.com,
        linux-kernel@...r.kernel.org, luto@...nel.org, mingo@...hat.com,
        syzkaller-bugs@...glegroups.com, tglx@...utronix.de, x86@...nel.org
Subject: Re: [syzbot] WARNING in kthread_is_per_cpu

On 20/04/21 12:11, Peter Zijlstra wrote:
> On Tue, Apr 20, 2021 at 10:43:43AM +0100, Valentin Schneider wrote:
>> On 20/04/21 10:51, Peter Zijlstra wrote:
>
>> > I think free_kthread_struct() is ok, because a task at that point in its
>> > lifetime cannot be also doing exec().
>> >
>>
>> What if it's one of those kthreads created by directly invoking
>> kernel_thread()? AFAICT right now it's only umh, and that one does execve()
>> so it ends up stripped of PF_KTHREAD. It could however go through an error
>> path, i.e. not call exec, and exit, giving us:
>>
>>   put_task_struct(p)
>>   `\
>>     free_task(p)
>>     `\
>>       if (tsk->flags & PF_KTHREAD)
>>           free_kthread_struct(tsk);
>>           `\
>>             to_kthread(p)
>
> I'm not following, at the point we hit free_task() it had better be dead
> and p->flags had better be stable. Either it will, or will not, have
> PF_KTHREAD.

Bah, don't mind me, for some reason I was obsessed by that umh thing of
having

  (p->flags & PF_KTHREAD) && !p->set_child_tid

but that's not a problem there. Sorry about that.