Message-ID: <55A7C7A0.9050907@kernel.dk>
Date: Thu, 16 Jul 2015 09:02:56 -0600
From: Jens Axboe <axboe@kernel.dk>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.7.0
MIME-Version: 1.0
To: Tejun Heo <tj@kernel.org>
CC: Ming Lei <tom.leiming@gmail.com>, linux-kernel@vger.kernel.org,
        Mike Snitzer <snitzer@redhat.com>
Subject: Re: [PATCH 0/4] block: account io: kill atomic operations
References: <1437016607-375-1-git-send-email-tom.leiming@gmail.com> <55A7C449.4030406@kernel.dk> <20150716145919.GV15934@mtj.duckdns.org>
In-Reply-To: <20150716145919.GV15934@mtj.duckdns.org>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1188
Lines: 25

On 07/16/2015 08:59 AM, Tejun Heo wrote:
> On Thu, Jul 16, 2015 at 08:48:41AM -0600, Jens Axboe wrote:
>> I've played with this before, but always ran into the hurdle of making
>> part_in_flight() too expensive ended up hurting results in the end. Making
>> the inc/dec parts of accounting percpu is a no-brainer, unfortunately the
>> summing then becomes pretty expensive. I'll run this through some testing
>> and see what kind of results I get.
>
> The only place which could be a problem is part_round_stats() and we
> can do that *way* lazier.  I don't think we're using that internally.
> Why are we even invoking it from IO issue / completion path?

Right, that's where it's called in the fast path. Right now, it's 
per-jiffy lazy. As to how/when it's called, I don't believe that has 
changed since those stats were introduced. If we can move that out of 
the fast path, then the percpu changes are straight forward.

-- 
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/