Re: [PATCH] io_thread/x86: don't reset 'cs', 'ss', 'ds' and 'es' registers for io_threads

From: Jens Axboe <axboe@kernel.dk>
To: Linus Torvalds <torvalds@linux-foundation.org>,
	Olivier Langlois <olivier@trillion01.com>
Cc: Stefan Metzmacher <metze@samba.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Andy Lutomirski <luto@kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	io-uring <io-uring@vger.kernel.org>,
	the arch/x86 maintainers <x86@kernel.org>
Subject: Re: [PATCH] io_thread/x86: don't reset 'cs', 'ss', 'ds' and 'es' registers for io_threads
Date: Wed, 12 May 2021 14:55:02 -0600	[thread overview]
Message-ID: <3df541c3-728c-c63d-eaeb-a4c382e01f0b@kernel.dk> (raw)
In-Reply-To: <CAHk-=whoJCocFsQ7+Sqq=dkuzHE+RXxvRdd4ZvyYqnsKBqsKAA@mail.gmail.com>

On 5/12/21 11:44 AM, Linus Torvalds wrote:
> On Tue, May 11, 2021 at 9:24 PM Olivier Langlois <olivier@trillion01.com> wrote:
>>
>> I have reported an issue that I have with a user process using io_uring
>> where when it core dumps, the dump fails to be generated.
>> https://github.com/axboe/liburing/issues/346
> 
> I suspect most kernel developers don't have github notifications
> enabled. I know I have them disabled because it would be *way* too
> noisy not to.
> 
> But maybe Jens does for that libiouring part.
> 
>> Pavel did comment to my report and he did point out this thread as
>> possibly a related issue.
> 
> I don't think this is related. The gdb confusion wouldn't affect core
> dump generation.
> 
> I don't see why a core-dump shouldn't work from an IO thread these
> days - the signal struct and synchronization should all be the same as
> for a regular user thread.
> 
> That said, I do wonder if we should avoid generating core dumps from
> the IO worker thread itself. The IO thread itself should never get a
> SIGSEGV/SIGBUS anyway, it should have been turned into -EFAULT.
> 
> So maybe the
> 
>                 if (current->flags & PF_IO_WORKER)
>                         goto out;
> 
> in kernel/signal.c should be moved up above the do_coredump() logic regardless.

I actually think that's how I originally had it, but Eric had some comment
on that and we moved it. IIRC. I'll dig out the conversation.

> Jens, have you played with core-dumping when there are active io_uring
> threads? There's a test-program in that github issue report..

Yes, I also did that again after the report, and did so again right now
just to verify. I'm not seeing any issues with coredumps being generated
if the app crashes, or if I send it SIGILL, for example... I also just
now tried Olivier's test case, and it seems to dump just fine for me.

I then tried backing out the patch from Stefan, and it works fine with
that reverted too. So a bit puzzled as to what is going on here...

Anyway, I'll check in on that github thread and see if we can narrow
this down.

-- 
Jens Axboe