Received-SPF: pass (google.com: domain of linux-kernel+bounces-150366-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <362c3c63-6fd9-401c-9281-e2e0c7efc14a@kernel.dk>
Date: Thu, 18 Apr 2024 09:09:07 -0600
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH 1/2] completion: move blk_wait_io to
 kernel/sched/completion.c
Content-Language: en-US
To: Christoph Hellwig <hch@infradead.org>
Cc: Mikulas Patocka <mpatocka@redhat.com>,
 Peter Zijlstra <peterz@infradead.org>, Mike Snitzer <msnitzer@redhat.com>,
 Damien Le Moal <dlemoal@kernel.org>, Ingo Molnar <mingo@redhat.com>,
 Will Deacon <will@kernel.org>, Waiman Long <longman@redhat.com>,
 Guangwu Zhang <guazhang@redhat.com>, dm-devel@lists.linux.dev,
 linux-block@vger.kernel.org, linux-kernel@vger.kernel.org
References: <31b118f3-bc8d-b18b-c4b9-e57d74a73f@redhat.com>
 <20240417175538.GP40213@noisy.programming.kicks-ass.net>
 <546473fd-ca4b-3c64-349d-cc739088b748@redhat.com>
 <ZiCoIHFLAzCva2lU@infradead.org>
 <5f3d434b-e05c-445f-bee5-2bb1f11a5946@kernel.dk>
 <ZiEyLvL6Pq_RB-Eh@infradead.org>
From: Jens Axboe <axboe@kernel.dk>
In-Reply-To: <ZiEyLvL6Pq_RB-Eh@infradead.org>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

On 4/18/24 8:46 AM, Christoph Hellwig wrote:
> On Thu, Apr 18, 2024 at 08:30:14AM -0600, Jens Axboe wrote:
>> It certainly is a hack/work-around, but unless there are a lot more that
>> should be using something like this, I don't think adding extra core
>> complexity in terms of a special task state (or per-task flag, at least
>> that would be easier) is really warranted.
> 
> Basically any kernel thread doing on-demand work has the same problem.
> It just has an easier workaround hack, as the kernel threads can simply
> claim to do an interruptible sleep to not trigger the softlockup
> warnings.

A kernel thread can just use TASK_INTERRUPTIBLE, as it doesn't take
signals anyway. But yeah, I guess you could view that as a work-around
as well.

Outside of that, mostly only a block problem, where our sleep is always
uninterruptible. Unless there are similar hacks elsewhere in the kernel
that I'm not aware of?

-- 
Jens Axboe