Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 24 Apr 2023 13:07:50 +0200
From:   Oleg Nesterov <oleg@redhat.com>
To:     Tiezhu Yang <yangtiezhu@loongson.cn>
Cc:     Bagas Sanjaya <bagasdotme@gmail.com>,
        Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
        linux-mips@vger.kernel.org, linux-kernel@vger.kernel.org,
        loongson-kernel@lists.loongnix.cn,
        Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Subject: Re: [PATCH] MIPS: uprobes: Restore thread.trap_nr
Message-ID: <20230424110750.GA11631@redhat.com>
References: <1682213883-3654-1-git-send-email-yangtiezhu@loongson.cn>
 <ZEShNUil0gsVlrDo@debian.me>
 <f99fe71d-cd97-e641-d637-5e7efbcad5f5@loongson.cn>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <f99fe71d-cd97-e641-d637-5e7efbcad5f5@loongson.cn>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

On 04/23, Tiezhu Yang wrote:
>
> Cc:
> Oleg Nesterov <oleg@redhat.com>
> Srikar Dronamraju <srikar@linux.vnet.ibm.com>
>
> On 04/23/2023 11:08 AM, Bagas Sanjaya wrote:
> >On Sun, Apr 23, 2023 at 09:38:03AM +0800, Tiezhu Yang wrote:
> >>thread.trap_nr is saved in arch_uprobe_pre_xol(), it should be restored
> >>in arch_uprobe_{post,abort}_xol() accordingly, actually it was only done
> >>in the post function, just do it in the abort function too, this change
> >>is similar with x86 and powerpc.
> >
> >I'm confused (please fix up grammar, spelling, and punctuation). Can you
> >explain why thread.trap_nr should be restored somewhere else? Also, what
> >x86/powerpc changes as reference?
> >
>
> Here is the related first commit for x86 in 2012:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=0326f5a94dde
>
> When xol insn itself triggers the signal, restart the original insn,
> in this case, UTASK_SSTEP_TRAPPED is set [1], it does *abort_xol()
> instead of *post_xol() [2], then should do the restore operations.

Yes... for example, if the uprobed task was killed abort() should restore
the state and (in particular) change ->trap_nr from UPROBE_TRAP_NR back to
->saved_trap_nr.

So the patch looks fine to me.

Oleg.