linux-kernel - Re: 5.8-rc*: kernel BUG at kernel/signal.c:1917

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]

Date:   Mon, 20 Jul 2020 08:56:33 +0200
From:   Oleg Nesterov <oleg@...hat.com>
To:     Jiri Slaby <jirislaby@...nel.org>
Cc:     christian@...uner.io, "Eric W. Biederman" <ebiederm@...ssion.com>,
        Linux kernel mailing list <linux-kernel@...r.kernel.org>
Subject: Re: 5.8-rc*: kernel BUG at kernel/signal.c:1917

On 07/20, Jiri Slaby wrote:
>
> On 18. 07. 20, 19:14, Oleg Nesterov wrote:
> >
> > This is already wrong. But
> >
> > 	Where does this __might_sleep() come from ??? I ses no blocking calls
> > 	in ptrace_stop(). Not to mention it is called with ->siglock held and
> > 	right after this lock is dropped we take tasklist_lock.
>
> Decoded stacktrace:
>
> > ptrace_stop (include/linux/freezer.h:57 include/linux/freezer.h:67 include/linux/freezer.h:128 include/linux/freezer.h:173 kernel/signal.c:2217)
> > ptrace_do_notify (kernel/signal.c:2272)
> > ptrace_notify (arch/x86/include/asm/paravirt.h:656 arch/x86/include/asm/qspinlock.h:55 include/linux/spinlock.h:211 include/linux/spinlock_api_smp.h:167 include/linux/spinlock.h:403 kernel/signal.c:2282)
> > syscall_trace_enter (include/linux/tracehook.h:73 include/linux/tracehook.h:104 arch/x86/entry/common.c:159)
> > do_syscall_64 (arch/x86/entry/common.c:380)
> > entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:124)
>
> It is try_to_freeze_unsafe in try_to_freeze in freezable_schedule in
> ptrace_stop.

Aha, thanks a lot!

try_to_freeze_unsafe() is called after schedule() which must return with
->state = RUNNING, so this matches another WARN_ON(current->state) added
by debugging patch after freezable_schedule().

Somehow I decided __might_sleep() was called before read_unlock/schedule.

> >
> >       How this connects to the debugging patch I sent? Did you see this warning
> >       without that patch?
>
> I suppose this made it appear:
> +CONFIG_PREEMPT_COUNT=y
> -# CONFIG_DEBUG_ATOMIC_SLEEP is not set
> +CONFIG_DEBUG_ATOMIC_SLEEP=y

I see,

> Please see my other e-mail, all this is with dbfb089d360b applied. Maybe
> it makes more sense now?

Yes. Thanks Jiri!

Oleg.