From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-fsdevel-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:39512 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1751049AbcFFXqI (ORCPT <rfc822;linux-fsdevel@vger.kernel.org>);
	Mon, 6 Jun 2016 19:46:08 -0400
Date: Tue, 7 Jun 2016 00:43:36 +0200
From: Oleg Nesterov <oleg@redhat.com>
To: Dmitry Safonov <dsafonov@virtuozzo.com>
Cc: linux-kernel@vger.kernel.org, mingo@redhat.com,
	luto@amacapital.net, tglx@linutronix.de, hpa@zytor.com,
	x86@kernel.org, 0x7f454c46@gmail.com, gorcunov@openvz.org,
	xemul@virtuozzo.com, khorenko@virtuozzo.com,
	Andy Lutomirski <luto@kernel.org>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 4/6] x86/coredump: use core regs, rather that TIF_IA32
	flag
Message-ID: <20160606224336.GA30331@redhat.com>
References: <1464786697-20639-1-git-send-email-dsafonov@virtuozzo.com> <1464786697-20639-5-git-send-email-dsafonov@virtuozzo.com> <20160606204548.GA23681@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20160606204548.GA23681@redhat.com>
Sender: linux-fsdevel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

On 06/06, Oleg Nesterov wrote:
>
> On 06/01, Dmitry Safonov wrote:
> >
> >  static int fill_thread_core_info(struct elf_thread_core_info *t,
> >  				 const struct user_regset_view *view,
> > -				 long signr, size_t *total)
> > +				 long signr, size_t *total,
> > +				 struct pt_regs *regs __maybe_unused)
> >  {
> >  	unsigned int i;
> >
> > @@ -1652,11 +1653,11 @@ static int fill_thread_core_info(struct elf_thread_core_info *t,
> >  	 */
> >  	fill_prstatus(&t->prstatus, t->task, signr);
> >  	(void) view->regsets[0].get(t->task, &view->regsets[0],
> > -				    0, PR_REG_SIZE(t->prstatus.pr_reg),
> > +				    0, PR_REG_SIZE(t->prstatus.pr_reg, regs),
>
> Hmm. I don't understand this... Note that this "regs" argument has nothing
> to do with t->task if the process is multithreaded,
>
> > @@ -1772,7 +1773,8 @@ static int fill_note_info(struct elfhdr *elf, int phdrs,
> >  	 * Now fill in each thread's information.
> >  	 */
> >  	for (t = info->thread; t != NULL; t = t->next)
> > -		if (!fill_thread_core_info(t, view, siginfo->si_signo, &info->size))
> > +		if (!fill_thread_core_info(t, view, siginfo->si_signo,
> > +					&info->size, regs))
>
> fill_note_info(..., args) is called with args = task_pt_regs(dumper_thread).

forgot to mention... yes, this matches the fact we use a single "view"
for all threads, and we get it via task_user_regset_view(dump_task).

But this change (imo) adds even more confusion, and without the next patch
the logic looks "obviously wrong", becauase PR_REG_SIZE/etc look at
dumper_thread->cs while task_user_regset_view() checks thread flags.

Anyway I fail to understand these macros... Say, PR_REG_SIZE(S). Can't we
kill it and use regsets[0].n * regsets[0].size instead ? These numbers
should match whatever we do, if we call ->get().

Oleg.