[SCSI] mpt2sas : Fix for memory allocation error for large host credits

The amount of memory required for tracking chain buffers is rather
large, and when the host credit count is big, memory allocation
failure occurs inside __get_free_pages.

The fix is to limit the number of chains to 100,000.  In addition,
the number of host credits is limited to 30,000 IOs. However this
limitation can be overridden this using the command line option
max_queue_depth.  The algorithm for calculating the
reply_post_queue_depth is changed so that it is equal to
(reply_free_queue_depth + 16), previously it was (reply_free_queue_depth * 2).

Signed-off-by: Nagalakshmi Nandigama <nagalakshmi.nandigama@lsi.com>
Cc: stable@kernel.org
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
diff --git a/drivers/scsi/mpt2sas/mpt2sas_base.c b/drivers/scsi/mpt2sas/mpt2sas_base.c
index c052687..961b11e 100644
--- a/drivers/scsi/mpt2sas/mpt2sas_base.c
+++ b/drivers/scsi/mpt2sas/mpt2sas_base.c
@@ -66,6 +66,8 @@
 
 #define FAULT_POLLING_INTERVAL 1000 /* in milliseconds */
 
+#define MAX_HBA_QUEUE_DEPTH	30000
+#define MAX_CHAIN_DEPTH		100000
 static int max_queue_depth = -1;
 module_param(max_queue_depth, int, 0);
 MODULE_PARM_DESC(max_queue_depth, " max controller queue depth ");
@@ -2390,8 +2392,6 @@
 		}
 		if (ioc->chain_dma_pool)
 			pci_pool_destroy(ioc->chain_dma_pool);
-	}
-	if (ioc->chain_lookup) {
 		free_pages((ulong)ioc->chain_lookup, ioc->chain_pages);
 		ioc->chain_lookup = NULL;
 	}
@@ -2409,9 +2409,7 @@
 _base_allocate_memory_pools(struct MPT2SAS_ADAPTER *ioc,  int sleep_flag)
 {
 	struct mpt2sas_facts *facts;
-	u32 queue_size, queue_diff;
 	u16 max_sge_elements;
-	u16 num_of_reply_frames;
 	u16 chains_needed_per_io;
 	u32 sz, total_sz, reply_post_free_sz;
 	u32 retry_sz;
@@ -2438,7 +2436,8 @@
 		max_request_credit = (max_queue_depth < facts->RequestCredit)
 		    ? max_queue_depth : facts->RequestCredit;
 	else
-		max_request_credit = facts->RequestCredit;
+		max_request_credit = min_t(u16, facts->RequestCredit,
+		    MAX_HBA_QUEUE_DEPTH);
 
 	ioc->hba_queue_depth = max_request_credit;
 	ioc->hi_priority_depth = facts->HighPriorityCredit;
@@ -2479,50 +2478,25 @@
 	}
 	ioc->chains_needed_per_io = chains_needed_per_io;
 
-	/* reply free queue sizing - taking into account for events */
-	num_of_reply_frames = ioc->hba_queue_depth + 32;
+	/* reply free queue sizing - taking into account for 64 FW events */
+	ioc->reply_free_queue_depth = ioc->hba_queue_depth + 64;
 
-	/* number of replies frames can't be a multiple of 16 */
-	/* decrease number of reply frames by 1 */
-	if (!(num_of_reply_frames % 16))
-		num_of_reply_frames--;
-
-	/* calculate number of reply free queue entries
-	 *  (must be multiple of 16)
-	 */
-
-	/* (we know reply_free_queue_depth is not a multiple of 16) */
-	queue_size = num_of_reply_frames;
-	queue_size += 16 - (queue_size % 16);
-	ioc->reply_free_queue_depth = queue_size;
-
-	/* reply descriptor post queue sizing */
-	/* this size should be the number of request frames + number of reply
-	 * frames
-	 */
-
-	queue_size = ioc->hba_queue_depth + num_of_reply_frames + 1;
-	/* round up to 16 byte boundary */
-	if (queue_size % 16)
-		queue_size += 16 - (queue_size % 16);
-
-	/* check against IOC maximum reply post queue depth */
-	if (queue_size > facts->MaxReplyDescriptorPostQueueDepth) {
-		queue_diff = queue_size -
-		    facts->MaxReplyDescriptorPostQueueDepth;
-
-		/* round queue_diff up to multiple of 16 */
-		if (queue_diff % 16)
-			queue_diff += 16 - (queue_diff % 16);
-
-		/* adjust hba_queue_depth, reply_free_queue_depth,
-		 * and queue_size
-		 */
-		ioc->hba_queue_depth -= (queue_diff / 2);
-		ioc->reply_free_queue_depth -= (queue_diff / 2);
-		queue_size = facts->MaxReplyDescriptorPostQueueDepth;
+	/* align the reply post queue on the next 16 count boundary */
+	if (!ioc->reply_free_queue_depth % 16)
+		ioc->reply_post_queue_depth = ioc->reply_free_queue_depth + 16;
+	else
+		ioc->reply_post_queue_depth = ioc->reply_free_queue_depth +
+				32 - (ioc->reply_free_queue_depth % 16);
+	if (ioc->reply_post_queue_depth >
+	    facts->MaxReplyDescriptorPostQueueDepth) {
+		ioc->reply_post_queue_depth = min_t(u16,
+		    (facts->MaxReplyDescriptorPostQueueDepth -
+		    (facts->MaxReplyDescriptorPostQueueDepth % 16)),
+		    (ioc->hba_queue_depth - (ioc->hba_queue_depth % 16)));
+		ioc->reply_free_queue_depth = ioc->reply_post_queue_depth - 16;
+		ioc->hba_queue_depth = ioc->reply_free_queue_depth - 64;
 	}
-	ioc->reply_post_queue_depth = queue_size;
+
 
 	dinitprintk(ioc, printk(MPT2SAS_INFO_FMT "scatter gather: "
 	    "sge_in_main_msg(%d), sge_per_chain(%d), sge_per_io(%d), "
@@ -2608,15 +2582,12 @@
 	    "depth(%d)\n", ioc->name, ioc->request,
 	    ioc->scsiio_depth));
 
-	/* loop till the allocation succeeds */
-	do {
-		sz = ioc->chain_depth * sizeof(struct chain_tracker);
-		ioc->chain_pages = get_order(sz);
-		ioc->chain_lookup = (struct chain_tracker *)__get_free_pages(
-		    GFP_KERNEL, ioc->chain_pages);
-		if (ioc->chain_lookup == NULL)
-			ioc->chain_depth -= 100;
-	} while (ioc->chain_lookup == NULL);
+	ioc->chain_depth = min_t(u32, ioc->chain_depth, MAX_CHAIN_DEPTH);
+	sz = ioc->chain_depth * sizeof(struct chain_tracker);
+	ioc->chain_pages = get_order(sz);
+
+	ioc->chain_lookup = (struct chain_tracker *)__get_free_pages(
+	    GFP_KERNEL, ioc->chain_pages);
 	ioc->chain_dma_pool = pci_pool_create("chain pool", ioc->pdev,
 	    ioc->request_sz, 16, 0);
 	if (!ioc->chain_dma_pool) {
diff --git a/drivers/scsi/mpt2sas/mpt2sas_scsih.c b/drivers/scsi/mpt2sas/mpt2sas_scsih.c
index fd1ba74..8cd4ebb 100644
--- a/drivers/scsi/mpt2sas/mpt2sas_scsih.c
+++ b/drivers/scsi/mpt2sas/mpt2sas_scsih.c
@@ -1011,8 +1011,8 @@
 	spin_lock_irqsave(&ioc->scsi_lookup_lock, flags);
 	if (list_empty(&ioc->free_chain_list)) {
 		spin_unlock_irqrestore(&ioc->scsi_lookup_lock, flags);
-		printk(MPT2SAS_WARN_FMT "chain buffers not available\n",
-		    ioc->name);
+		dfailprintk(ioc, printk(MPT2SAS_WARN_FMT "chain buffers not "
+			"available\n", ioc->name));
 		return NULL;
 	}
 	chain_req = list_entry(ioc->free_chain_list.next,