Message-ID: <1458105532.5809.10.camel@kernel.org>
Subject: Re: [PATCH RFC 0/2] mempool based chained scatterlist alloc/free
 api api
From: Ming Lin <mlin@kernel.org>
To: James Bottomley <James.Bottomley@HansenPartnership.com>,
        linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org
Cc: Christoph Hellwig <hch@lst.de>
Date: Tue, 15 Mar 2016 22:18:52 -0700
In-Reply-To: <1458083523.2375.120.camel@HansenPartnership.com>
References: <1458081569-30953-1-git-send-email-mlin@kernel.org>
	 <1458083523.2375.120.camel@HansenPartnership.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1843
Lines: 58

On Tue, 2016-03-15 at 16:12 -0700, James Bottomley wrote:
> On Tue, 2016-03-15 at 15:39 -0700, Ming Lin wrote:
> > From: Ming Lin <ming.l@ssi.samsung.com>
> > 
> > Hi list,
> > 
> > This moves the mempool based chained scatterlist alloc/free code
> > from
> > scsi_lib.c to lib/scatterlist.c.
> > 
> > So other drivers(for example, the under development NVMe over
> > fabric 
> > drivers) can also use it.
> > 
> > Ming Lin (2):
> >   scatterlist: add mempool based chained SG alloc/free api
> >   scsi: use the new chained SG api
> > 
> >  drivers/scsi/scsi_lib.c     | 129 ++----------------------------
> > ------
> >  include/linux/scatterlist.h |  12 ++++
> >  lib/scatterlist.c           | 156
> > ++++++++++++++++++++++++++++++++++++++++++++
> >  3 files changed, 175 insertions(+), 122 deletions(-)
> 
> I'd really rather this were a single patch so git can tell us the
> code
> motion.  If you add in one patch and remove in another the code
> motion
> trackers don't see it.
> 
> Secondly, you said "This copied code from scsi_lib.c to scatterlist.c
> and modified it a bit" could you move in one patch and modify in
> another, so we can see exactly what you're changing.

The modification is mostly about structure names and function names
changes.

I can do it in a single patch.

> 
> Thirdly, are you sure the pool structure for NVMe should be the same
> as
> for SCSI?  We don't do buddy pools for 1,2 or 4 entry transactions in
> SCSI just basically because of heuristics, but the packetised io
> characteristics of NVMe make single entry lists more likely for it,
> don't they?

Not sure about this, but the nvme-pci driver may not use this api,
because it also has a PRP lists except for the SG lists.

But nvme-over-rdma/nvme-over-fiber-channel driver is good to use this
api.

> 
> James
> 
>