From mboxrd@z Thu Jan 1 00:00:00 1970 From: james.smart@broadcom.com (James Smart) Date: Mon, 22 Aug 2016 10:28:39 -0700 Subject: [PATCH 4/5] nvme-fabrics: Add target FC transport support In-Reply-To: <20160811212238.GG18013@infradead.org> References: <5792b91e.51+4Ind0/2MxGqHU%james.smart@broadcom.com> <20160811212238.GG18013@infradead.org> Message-ID: <9641de6d-1bfb-394c-d4e0-712c8ef6d740@broadcom.com> On 8/11/2016 2:22 PM, Christoph Hellwig wrote: > >> + queue->work_q = alloc_workqueue("ntfc%d.%d.%d", 0, 0, >> + assoc->tgtport->fc_target_port.port_num, >> + assoc->a_id, qid); > Do we really need a workqueue per queue? I thought we did - so there was always an execution context for the nvmet layer when a command is received. My understanding was the nvmet layer processes the cmd synchronously. I avoided per cpu - as I assumed the synchronous flows would end up blocking other connections to other subsystems/controllers. If that's not the case - can you describe the nvmet layer better so I understand what we need ? > > No reference counting? > yes - will make a pass at it. -- james From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Smart Subject: Re: [PATCH 4/5] nvme-fabrics: Add target FC transport support Date: Mon, 22 Aug 2016 10:28:39 -0700 Message-ID: <9641de6d-1bfb-394c-d4e0-712c8ef6d740@broadcom.com> References: <5792b91e.51+4Ind0/2MxGqHU%james.smart@broadcom.com> <20160811212238.GG18013@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail-pf0-f171.google.com ([209.85.192.171]:33589 "EHLO mail-pf0-f171.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755101AbcHVR2u (ORCPT ); Mon, 22 Aug 2016 13:28:50 -0400 Received: by mail-pf0-f171.google.com with SMTP id y134so34386994pfg.0 for ; Mon, 22 Aug 2016 10:28:41 -0700 (PDT) In-Reply-To: <20160811212238.GG18013@infradead.org> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Christoph Hellwig Cc: linux-nvme@lists.infradead.org, linux-scsi@vger.kernel.org On 8/11/2016 2:22 PM, Christoph Hellwig wrote: > >> + queue->work_q = alloc_workqueue("ntfc%d.%d.%d", 0, 0, >> + assoc->tgtport->fc_target_port.port_num, >> + assoc->a_id, qid); > Do we really need a workqueue per queue? I thought we did - so there was always an execution context for the nvmet layer when a command is received. My understanding was the nvmet layer processes the cmd synchronously. I avoided per cpu - as I assumed the synchronous flows would end up blocking other connections to other subsystems/controllers. If that's not the case - can you describe the nvmet layer better so I understand what we need ? > > No reference counting? > yes - will make a pass at it. -- james