From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from e37.co.us.ibm.com (e37.co.us.ibm.com [32.97.110.158]) (using TLSv1.2 with cipher CAMELLIA256-SHA (256/256 bits)) (No client certificate requested) by lists.ozlabs.org (Postfix) with ESMTPS id 3qWtdv00VczDqT4 for ; Sat, 26 Mar 2016 06:29:06 +1100 (AEDT) Received: from localhost by e37.co.us.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Fri, 25 Mar 2016 13:29:05 -0600 Received: from b03cxnp07029.gho.boulder.ibm.com (b03cxnp07029.gho.boulder.ibm.com [9.17.130.16]) by d03dlp01.boulder.ibm.com (Postfix) with ESMTP id 9456A1FF0021 for ; Fri, 25 Mar 2016 13:17:10 -0600 (MDT) Received: from d03av01.boulder.ibm.com (d03av01.boulder.ibm.com [9.17.195.167]) by b03cxnp07029.gho.boulder.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id u2PJT1x533554674 for ; Fri, 25 Mar 2016 12:29:01 -0700 Received: from d03av01.boulder.ibm.com (localhost [127.0.0.1]) by d03av01.boulder.ibm.com (8.14.4/8.14.4/NCO v10.0 AVout) with ESMTP id u2PJT1Qa002106 for ; Fri, 25 Mar 2016 13:29:01 -0600 From: Uma Krishnan To: linux-scsi@vger.kernel.org, James Bottomley , "Martin K. Petersen" , "Matthew R. Ochs" , "Manoj N. Kumar" Cc: Brian King , linuxppc-dev@lists.ozlabs.org, Ian Munsie , Andrew Donnellan , Frederic Barrat , Christophe Lombard Subject: [PATCH v2 2/2] cxlflash: Move to exponential back-off when cmd_room is not available Date: Fri, 25 Mar 2016 14:26:49 -0500 Message-Id: <1458934009-28861-1-git-send-email-ukrishn@linux.vnet.ibm.com> In-Reply-To: <1458933920-28788-1-git-send-email-ukrishn@linux.vnet.ibm.com> References: <1458933920-28788-1-git-send-email-ukrishn@linux.vnet.ibm.com> List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , From: "Manoj N. Kumar" While profiling the cxlflash_queuecommand() path under a heavy load it was found that number of retries to find cmd_room was fairly high. There are two problems with the current back-off: a) It starts with a udelay of 0 b) It backs-off linearly Tried several approaches (a higher multiple 10*n, 100*n, as well as n^2, 2^n) and found that the exponential back-off(2^n) approach had the least overall cost. Cost as being defined as overall time spent waiting. The fix is to change the linear back-off to an exponential back-off. This solution also takes care of the problem with the initial delay (starts with 1 usec). Signed-off-by: Manoj N. Kumar Acked-by: Matthew R. Ochs Reviewed-by: Johannes Thumshirn Signed-off-by: Uma Krishnan --- drivers/scsi/cxlflash/main.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index e6d56ac..8fb9643 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -289,7 +289,7 @@ static void context_reset(struct afu_cmd *cmd) atomic64_set(&afu->room, room); if (room) goto write_rrin; - udelay(nretry); + udelay(1 << nretry); } while (nretry++ < MC_ROOM_RETRY_CNT); pr_err("%s: no cmd_room to send reset\n", __func__); @@ -303,7 +303,7 @@ write_rrin: if (rrin != 0x1) break; /* Double delay each time */ - udelay(2 << nretry); + udelay(1 << nretry); } while (nretry++ < MC_ROOM_RETRY_CNT); } @@ -338,7 +338,7 @@ retry: atomic64_set(&afu->room, room); if (room) goto write_ioarrin; - udelay(nretry); + udelay(1 << nretry); } while (nretry++ < MC_ROOM_RETRY_CNT); dev_err(dev, "%s: no cmd_room to send 0x%X\n", @@ -352,7 +352,7 @@ retry: * afu->room. */ if (nretry++ < MC_ROOM_RETRY_CNT) { - udelay(nretry); + udelay(1 << nretry); goto retry; } -- 2.1.0