From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bart Van Assche Subject: Re: [PATCH 05/13] IB/core: add a simple MR pool Date: Wed, 2 Mar 2016 07:22:43 -0800 Message-ID: <56D70543.1000506@sandisk.com> References: <1456596631-19418-1-git-send-email-hch@lst.de> <1456596631-19418-6-git-send-email-hch@lst.de> <20160302091554.GC1308@lst.de> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20160302091554.GC1308@lst.de> Sender: target-devel-owner@vger.kernel.org To: Christoph Hellwig , Parav Pandit Cc: linux-rdma@vger.kernel.org, swise@opengridcomputing.com, Sagi Grimberg , target-devel@vger.kernel.org List-Id: linux-rdma@vger.kernel.org On 03/02/16 01:15, Christoph Hellwig wrote: > On Wed, Mar 02, 2016 at 08:18:14AM +0530, Parav Pandit wrote: >>> struct list_head xrcd_list; >>> + >>> + spinlock_t mr_lock; >>> + int mrs_used; >>> + >> >> Can you please add the comment for mr_lock as requested by the >> checkpatch script? > > No. checkpath is asking for silly things (and apparently this one > is so silly it doesn't even ask for it by default). > >> Also you might want to consider adding this field after recv_cq so >> that we find mr_lock and used count in single cache line along with >> other data for the qp? > > That sounds useful, I'll look into it. Hello Christoph, With the approach of V2 of this patch series mr_lock and the MR pool list head exist in different structures which is unfortunate. How about introducing a new structure for the MR pool list head and mr_lock? An additional advantage of this approach is that it would allow to move the initialization of both structure members into ib_mr_pool_init(). Bart.