Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <1615296712-175334-1-git-send-email-zhaoqianligood@gmail.com>
 <20210309182657.GA1408@redhat.com> <CAPx_LQEQto2fget=kCnfjSCdAp9XCwc2AWgzwxpbxdyCgNs5Mg@mail.gmail.com>
 <m1lfavufep.fsf@fess.ebiederm.org>
In-Reply-To: <m1lfavufep.fsf@fess.ebiederm.org>
From:   qianli zhao <zhaoqianligood@gmail.com>
Date:   Thu, 11 Mar 2021 12:40:49 +0800
Message-ID: <CAPx_LQF90joRVLCjR9M44uZ0829Ktw3GAFStdOWEBANGLfDvdQ@mail.gmail.com>
Subject: Re: [PATCH] exit: trigger panic when init process is set to SIGNAL_GROUP_EXIT
To:     "Eric W. Biederman" <ebiederm@xmission.com>
Cc:     Oleg Nesterov <oleg@redhat.com>, christian@brauner.io,
        axboe@kernel.dk, Thomas Gleixner <tglx@linutronix.de>,
        Peter Collingbourne <pcc@google.com>,
        linux-kernel@vger.kernel.org, Qianli Zhao <zhaoqianli@xiaomi.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi, Eric
Thank you for your suggestion

> At the start of your changelog and your patch subject you describe what
> you are doing but not why.  For the next revision of the patch please
> lead with the why it makes what you are trying to do much easier to
> understand.

got it.

>
> It does not work to use SIGNAL_UNKILLABLE for this.  Normally init
> has SIGNAL_UNKILLABLE set.  The only case that clears SIGNAL_UNKILLABLE
> is force_sig_info_to_task.  If the init process exits with exit(2)
> SIGNAL_UNKILLABLE will already be set.  Which means testing
> SIGNAL_UNKILLABLE as your patch does will prevent the panic.
>

Ok,using SIGNAL_UNKILLABLE is incorrect.

> Further simply calling panic is sufficient to guarantee that the other
> threads don't exit, and that whichever thread calls panic first
> will be the reporting thread.  The rest of the threads will be caught
> in panic_smp_self_stop(), if they happen to be running on other cpus.
>
> So I would make the whole thing just be:
>
>         /* If global init has exited,
>          * panic immediately to get a useable coredump.
>          */
>         if (unlikely(is_global_init(tsk) &&
>             (thread_group_empty(tsk) ||
>             (tsk->signal->flags & SIGNAL_GROUP_EXIT)))) {
>                 panic("Attempted to kill init!  exitcode=0x%08x\n",
>                         tsk->signal->group_exit_code ?: (int)code);
>         }
>
> The thread_group_empty test is needed to handle single threaded
> inits.
>
> Do you think you can respin your patch as something like that?
>

Ok.it's a very good change,other CPUs calls to panic() will be caught
and execute panic_smp_self_stop(),
there is no need to deal with this situation separately when other CPUs exit().