From mboxrd@z Thu Jan  1 00:00:00 1970
From: axboe@fb.com (Jens Axboe)
Date: Thu, 29 Jan 2015 09:24:21 -0800
Subject: [PATCH v2] NVMe: avoid kmalloc/kfree for smaller IO
In-Reply-To: <alpine.LNX.2.00.1501291714300.15481@localhost.lm.intel.com>
References: <20150122190758.GA28666@kernel.dk>
 <alpine.LNX.2.00.1501291714300.15481@localhost.lm.intel.com>
Message-ID: <54CA6CC5.7090602@fb.com>

On 01/29/2015 09:21 AM, Keith Busch wrote:
> On Thu, 22 Jan 2015, Jens Axboe wrote:
>> Currently we allocate an nvme_iod for each IO, which holds the
>> sg list, prps, and other IO related info. Set a threshold of
>> 2 pages and/or 8KB of data, below which we can just embed this
>> in the per-command pdu in blk-mq. For any IO at or below
>> NVME_INT_PAGES and NVME_INT_BYTES, we save a kmalloc and kfree.
>>
>> For higher IOPS, this saves up to 1% of CPU time.
>
> Looks good to me, tests well on my machines. I had to make one minor
> modification to the diff for it to apply to your latest tree, but it's
> trivial and you'll see it (just below nvme_set_info()).
>
> Reviewed-by: Keith Busch <keith.busch at intel.com>

Thanks! I'll apply it for 3.20/drivers.

-- 
Jens Axboe