From mboxrd@z Thu Jan 1 00:00:00 1970 From: Don Brace Subject: [PATCH 31/48] hpsa: optimize cmd_alloc function by remembering last allocation Date: Wed, 14 Jan 2015 16:02:21 -0600 Message-ID: <20150114220221.21325.80254.stgit@brunhilda> References: <20150114215756.21325.41198.stgit@brunhilda> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Return-path: Received: from g9t1613g.houston.hp.com ([15.240.0.71]:58567 "EHLO g9t1613g.houston.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754316AbbANWDJ (ORCPT ); Wed, 14 Jan 2015 17:03:09 -0500 Received: from g5t1627.atlanta.hp.com (g5t1627.atlanta.hp.com [15.192.137.10]) (using TLSv1.2 with cipher DHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by g9t1613g.houston.hp.com (Postfix) with ESMTPS id 2A07061452 for ; Wed, 14 Jan 2015 22:03:09 +0000 (UTC) In-Reply-To: <20150114215756.21325.41198.stgit@brunhilda> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: scott.teel@pmcs.com, Kevin.Barnett@pmcs.com, james.bottomley@parallels.com, hch@infradead.org, Justin.Lindley@pmcs.com, brace@pmcs.com Cc: linux-scsi@vger.kernel.org From: Robert Elliott Empirically, this improves performance slightly (~2% max IOPS) by allowing cmd_alloc to remember where it left off searching for free commands between calls instead of always starting its search at command 0. Reviewed-by: Scott Teel Signed-off-by: Robert Elliott Signed-off-by: Don Brace --- drivers/scsi/hpsa.c | 7 +++++-- drivers/scsi/hpsa.h | 1 + 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c index c95a20c..72abcf3 100644 --- a/drivers/scsi/hpsa.c +++ b/drivers/scsi/hpsa.c @@ -4649,9 +4649,10 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h) union u64bit temp64; dma_addr_t cmd_dma_handle, err_dma_handle; int refcount; - unsigned long offset = 0; + unsigned long offset; - /* There is some *extremely* small but non-zero chance that that + /* + * There is some *extremely* small but non-zero chance that that * multiple threads could get in here, and one thread could * be scanning through the list of bits looking for a free * one, but the free ones are always behind him, and other @@ -4662,6 +4663,7 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h) * infrequently as to be indistinguishable from never. */ + offset = h->last_allocation; /* benignly racy */ for (;;) { i = find_next_zero_bit(h->cmd_pool_bits, h->nr_cmds, offset); if (unlikely(i == h->nr_cmds)) { @@ -4679,6 +4681,7 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h) h->cmd_pool_bits + (i / BITS_PER_LONG)); break; /* it's ours now. */ } + h->last_allocation = i; /* benignly racy */ /* Zero out all of commandlist except the last field, refcount */ memset(c, 0, offsetof(struct CommandList, refcount)); diff --git a/drivers/scsi/hpsa.h b/drivers/scsi/hpsa.h index 679e4d2..981479a 100644 --- a/drivers/scsi/hpsa.h +++ b/drivers/scsi/hpsa.h @@ -133,6 +133,7 @@ struct ctlr_info { struct CfgTable __iomem *cfgtable; int interrupts_enabled; int max_commands; + int last_allocation; atomic_t commands_outstanding; # define PERF_MODE_INT 0 # define DOORBELL_INT 1