From: Tomas Henzl <thenzl@redhat.com>
Date: Wed, 19 Dec 2007 16:16:29 +0100
Subject: [scsi] cciss: move READ_AHEAD to block layer
Message-id: 476935CD.80109@redhat.com
O-Subject: [RHEL5.2 PATCH 4/8] RHEL 5.2: Cciss driver update remove READ_AHEAD from driver and use block layer
Bugzilla: 424371

This patch resolves the bz#424371: RH 5.2: Cciss driver update: Remove READ_AHEAD from driver and use block layer

Comment from HP:
We have found that under certain workloads our READ_AHEAD = 1024 can cause
systems to hang. We do not believe this is a cciss bug, per say. We have found
that by either setting the READ_AHEAD to 256 or just using the block layer
default of 256 the hangs do not occur. Very preliminary benchmark testing shows
the difference in performance is within the margin of error. Random read ops
improved slightly, while sequential reads dropped slightly. The delta is <2% in
our preliminary tests.

diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index 8fd1abc..d25ae75 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -149,7 +149,6 @@ static struct board_type products[] = {
 /*define how many times we will try a command because of bus resets */
 #define MAX_CMD_RETRIES 3
 
-#define READ_AHEAD 	 1024
 #define MAX_CTLR	32
 
 /* Originally cciss driver only supports 8 major numbers */
@@ -1346,7 +1345,6 @@ static void cciss_update_drive_info(int ctlr, int drv_index)
 		disk->private_data = &h->drv[drv_index];
 
 		/* Set up queue information */
-		disk->queue->backing_dev_info.ra_pages = READ_AHEAD;
 		blk_queue_bounce_limit(disk->queue, hba[ctlr]->pdev->dma_mask);
 
 		/* This is a hardware imposed limit. */
@@ -3362,7 +3360,6 @@ static int __devinit cciss_init_one(struct pci_dev *pdev,
 		}
 		drv->queue = q;
 
-		q->backing_dev_info.ra_pages = READ_AHEAD;
 		blk_queue_bounce_limit(q, hba[i]->pdev->dma_mask);
 
 		/* This is a hardware imposed limit. */