[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 1 Jan 2014 12:57:18 +0800
From: Ming Lei <tom.leiming@...il.com>
To: Jens Axboe <axboe@...nel.dk>
Cc: Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Christoph Hellwig <hch@...radead.org>
Subject: Re: [PATCH v1 1/5] block: blk-mq: avoid initializing request during
its completion
Hi Jens,
On Wed, Jan 1, 2014 at 12:38 AM, Jens Axboe <axboe@...nel.dk> wrote:
> On Thu, Dec 26 2013, Ming Lei wrote:
>> One problem is that request->start_time/start_time_ns could be
>> set as wrong.
>>
>> Also it is normal to intialize one data structure just after its
>> allocation.
>>
>> So move the initialization out of its completion path.
>
> It's done that way because of presumed cache hotness on completion,
> since we just touched a lot of the members. Lets just fix the start time
> issue by itself.
But some members of request are already touched in allocation
path(blk-mq core) too, and lots of members will be touched by
driver for starting the transfer.
Also I didn't observe obvious effect on L1 dcache load/store miss rate
after applying the patch when reading/writing on null_blk device.
Considered the theoretic advantage of reinitializing in free path, I
will just fix the start time issue by itself.
Thanks,
--
Ming Lei
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists