From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932959AbbLRQsN (ORCPT ); Fri, 18 Dec 2015 11:48:13 -0500 Received: from bedivere.hansenpartnership.com ([66.63.167.143]:49280 "EHLO bedivere.hansenpartnership.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753927AbbLRQsL (ORCPT ); Fri, 18 Dec 2015 11:48:11 -0500 Message-ID: <1450457288.2439.7.camel@HansenPartnership.com> Subject: Re: [PATCH V3 4/4] scsi: storvsc: Tighten up the interrupt path From: James Bottomley To: KY Srinivasan , Hannes Reinecke , "gregkh@linuxfoundation.org" , "linux-kernel@vger.kernel.org" , "devel@linuxdriverproject.org" , "ohering@suse.com" , "jbottomley@parallels.com" , "hch@infradead.org" , "linux-scsi@vger.kernel.org" , "apw@canonical.com" , "vkuznets@redhat.com" , "jasowang@redhat.com" , "martin.petersen@oracle.com" Date: Fri, 18 Dec 2015 08:48:08 -0800 In-Reply-To: References: <1450038486-19211-1-git-send-email-kys@microsoft.com> <1450038512-19252-1-git-send-email-kys@microsoft.com> <1450038512-19252-4-git-send-email-kys@microsoft.com> <5673C92B.5070602@suse.de> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.16.5 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2015-12-18 at 16:20 +0000, KY Srinivasan wrote: > > > -----Original Message----- > > From: Hannes Reinecke [mailto:hare@suse.de] > > Sent: Friday, December 18, 2015 12:52 AM > > To: KY Srinivasan ; gregkh@linuxfoundation.org; > > linux- > > kernel@vger.kernel.org; devel@linuxdriverproject.org; > > ohering@suse.com; > > jbottomley@parallels.com; hch@infradead.org; > > linux-scsi@vger.kernel.org; > > apw@canonical.com; vkuznets@redhat.com; jasowang@redhat.com; > > martin.petersen@oracle.com > > Subject: Re: [PATCH V3 4/4] scsi: storvsc: Tighten up the interrupt > > path > > > > On 12/13/2015 09:28 PM, K. Y. Srinivasan wrote: > > > On the interrupt path, we repeatedly establish the pointer to the > > > storvsc_device. Fix this. > > > > > > Signed-off-by: K. Y. Srinivasan > > > Reviewed-by: Long Li > > > Reviewed-by: Johannes Thumshirn > > > Tested-by: Alex Ng > > > --- > > > drivers/scsi/storvsc_drv.c | 23 ++++++++--------------- > > > 1 files changed, 8 insertions(+), 15 deletions(-) > > > > > > diff --git a/drivers/scsi/storvsc_drv.c > > > b/drivers/scsi/storvsc_drv.c > > > index d6ca4f2..b68aebe 100644 > > > --- a/drivers/scsi/storvsc_drv.c > > > +++ b/drivers/scsi/storvsc_drv.c > > > @@ -945,19 +945,16 @@ static void storvsc_handle_error(struct > > vmscsi_request *vm_srb, > > > } > > > > > > > > > -static void storvsc_command_completion(struct > > > storvsc_cmd_request > > *cmd_request) > > > +static void storvsc_command_completion(struct > > > storvsc_cmd_request > > *cmd_request, > > > + struct storvsc_device > > > *stor_dev) > > > { > > > struct scsi_cmnd *scmnd = cmd_request->cmd; > > > - struct hv_host_device *host_dev = shost_priv(scmnd > > > ->device- > > > host); > > > struct scsi_sense_hdr sense_hdr; > > > struct vmscsi_request *vm_srb; > > > struct Scsi_Host *host; > > > - struct storvsc_device *stor_dev; > > > - struct hv_device *dev = host_dev->dev; > > > u32 payload_sz = cmd_request->payload_sz; > > > void *payload = cmd_request->payload; > > > > > > - stor_dev = get_in_stor_device(dev); > > > host = stor_dev->host; > > > > > > vm_srb = &cmd_request->vstor_packet.vm_srb; > > > @@ -987,14 +984,13 @@ static void > > > storvsc_command_completion(struct > > storvsc_cmd_request *cmd_request) > > > kfree(payload); > > > } > > > > > > -static void storvsc_on_io_completion(struct hv_device *device, > > > +static void storvsc_on_io_completion(struct storvsc_device > > > *stor_device, > > > struct vstor_packet > > > *vstor_packet, > > > struct storvsc_cmd_request > > > *request) > > > { > > > - struct storvsc_device *stor_device; > > > struct vstor_packet *stor_pkt; > > > + struct hv_device *device = stor_device->device; > > > > > > - stor_device = hv_get_drvdata(device); > > > stor_pkt = &request->vstor_packet; > > > > > > /* > > > @@ -1049,7 +1045,7 @@ static void storvsc_on_io_completion(struct > > hv_device *device, > > > stor_pkt->vm_srb.data_transfer_length = > > > vstor_packet->vm_srb.data_transfer_length; > > > > > > - storvsc_command_completion(request); > > > + storvsc_command_completion(request, stor_device); > > > > > > if (atomic_dec_and_test(&stor_device > > > ->num_outstanding_req) && > > > stor_device->drain_notify) > > > @@ -1058,21 +1054,19 @@ static void > > > storvsc_on_io_completion(struct > > hv_device *device, > > > > > > } > > > > > > -static void storvsc_on_receive(struct hv_device *device, > > > +static void storvsc_on_receive(struct storvsc_device > > > *stor_device, > > > struct vstor_packet *vstor_packet, > > > struct storvsc_cmd_request > > > *request) > > > { > > > struct storvsc_scan_work *work; > > > - struct storvsc_device *stor_device; > > > > > > switch (vstor_packet->operation) { > > > case VSTOR_OPERATION_COMPLETE_IO: > > > - storvsc_on_io_completion(device, vstor_packet, > > > request); > > > + storvsc_on_io_completion(stor_device, > > > vstor_packet, > > request); > > > break; > > > > > > case VSTOR_OPERATION_REMOVE_DEVICE: > > > case VSTOR_OPERATION_ENUMERATE_BUS: > > > - stor_device = get_in_stor_device(device); > > > work = kmalloc(sizeof(struct > > > storvsc_scan_work), > > GFP_ATOMIC); > > > if (!work) > > > return; > > > @@ -1083,7 +1077,6 @@ static void storvsc_on_receive(struct > > > hv_device > > *device, > > > break; > > > > > > case VSTOR_OPERATION_FCHBA_DATA: > > > - stor_device = get_in_stor_device(device); > > > cache_wwn(stor_device, vstor_packet); > > > #ifdef CONFIG_SCSI_FC_ATTRS > > > fc_host_node_name(stor_device->host) = > > > stor_device- > > > node_name; > > > @@ -1133,7 +1126,7 @@ static void > > > storvsc_on_channel_callback(void > > *context) > > > vmscsi_size_delta)); > > > complete(&request->wait_event); > > > } else { > > > - storvsc_on_receive(device, > > > + storvsc_on_receive(stor_device, > > > (struct > > > vstor_packet > > *)packet, > > > request); > > > } > > > > > Hmm. I would've thought the compiler optimizes this away. Have you > > checked whether it actually makes a difference in the assembler > > output? > > I have not checked the assembler output. It was easy enough to fix > the source. Could you? You're making what you describe as an optimisation but there are two reasons why this might not be so. The first is that the compiler is entitled to inline static functions. If it did, likely it picked up the optmisation anyway as Hannes suggested. However, the other reason this might not be an optimisation (assuming the compiler doesn't inline the function) is you're passing an argument which can be offset computed. On all architectures, you have a fixed number of registers for passing function arguments, then we have to use the stack. Using the stack comes in far more expensive than computing an offset to an existing pointer. Even if you're still in registers, the offset now has to be computed and stored and the compiler loses track of the relation. The bottom line is that adding an extra argument for a value which can be offset computed is rarely a win. James