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;
Subject: Re: [PATCH 3/6] kernel: split syscall restart from signal handling
To:     Oleg Nesterov <oleg@redhat.com>
Cc:     linux-kernel@vger.kernel.org, io-uring@vger.kernel.org,
        peterz@infradead.org, tglx@linutronix.de
References: <20201005150438.6628-1-axboe@kernel.dk>
 <20201005150438.6628-4-axboe@kernel.dk> <20201008142135.GH9995@redhat.com>
 <de00f13d-9ff0-6955-5d37-557f044ce2aa@kernel.dk>
 <20201008144539.GJ9995@redhat.com>
From:   Jens Axboe <axboe@kernel.dk>
Message-ID: <28abe1ad-f9ef-75e1-a7aa-06543b2cf8ca@kernel.dk>
Date:   Thu, 8 Oct 2020 08:47:58 -0600
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <20201008144539.GJ9995@redhat.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 10/8/20 8:45 AM, Oleg Nesterov wrote:
> On 10/08, Jens Axboe wrote:
>>
>> On 10/8/20 8:21 AM, Oleg Nesterov wrote:
>>>
>>> Can't we avoid this patch and the and simplify the change in
>>> exit_to_user_mode_loop() from the next patch? Can't the much more simple
>>> patch below work?
>>>
>>> Then later we can even change arch_do_signal() to accept the additional
>>> argument, ti_work, so that it can use ti_work & TIF_NOTIFY_SIGNAL/SIGPENDING
>>> instead of test_thread_flag/task_sigpending.
>>
>> Yeah I guess that would be a bit simpler, maybe I'm too focused on
>> decoupling the two. But if we go this route, and avoid sighand->lock for
>> just having TIF_NOTIFY_SIGNAL set, then that should be functionally
>> equivalent as far as I'm concerned.
> 
> Not sure I understand... I think that the change I propose is functionally
> equivalent or I missed something.

Sorry, maybe my phrasing wasn't good, I'm totally agreeing with you :-)
Was just noting that the task_sigpending() is key for not calling
get_signal(), to avoid hitting the sighand->lock again.

-- 
Jens Axboe