mmc: block: Add write packing control

The write packing control will ensure that read requests latency is
not increased due to long write packed commands.

The trigger for enabling the write packing is managing to pack several
write requests. The number of potential packed requests that will trigger
the packing can be configured via sysfs by writing the required value to:
/sys/block/<block_dev_name>/num_wr_reqs_to_start_packing.
The trigger for disabling the write packing is fetching a read request.

Change-Id: I22e8ab04cd9aca220678784c39306068a0996790
Signed-off-by: Maya Erez <merez@codeaurora.org>
Signed-off-by: Tatyana Brokhman <tlinder@codeaurora.org>
diff --git a/Documentation/mmc/mmc-dev-attrs.txt b/Documentation/mmc/mmc-dev-attrs.txt
index 22ae844..08f7312 100644
--- a/Documentation/mmc/mmc-dev-attrs.txt
+++ b/Documentation/mmc/mmc-dev-attrs.txt
@@ -8,6 +8,23 @@
 
 	force_ro		Enforce read-only access even if write protect switch is off.
 
+	num_wr_reqs_to_start_packing 	This attribute is used to determine
+	the trigger for activating the write packing, in case the write
+	packing control feature is enabled.
+
+	When the MMC manages to reach a point where num_wr_reqs_to_start_packing
+	write requests could be packed, it enables the write packing feature.
+	This allows us to start the write packing only when it is beneficial
+	and has minimum affect on the read latency.
+
+	The number of potential packed requests that will trigger the packing
+	can be configured via sysfs by writing the required value to:
+	/sys/block/<block_dev_name>/num_wr_reqs_to_start_packing.
+
+	The default value of num_wr_reqs_to_start_packing was determined by
+	running parallel lmdd write and lmdd read operations and calculating
+	the max number of packed writes requests.
+
 SD and MMC Device Attributes
 ============================
 
diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c
index a9f1b53..989c53a 100644
--- a/drivers/mmc/card/block.c
+++ b/drivers/mmc/card/block.c
@@ -115,6 +115,7 @@
 	unsigned int	part_curr;
 	struct device_attribute force_ro;
 	struct device_attribute power_ro_lock;
+	struct device_attribute num_wr_reqs_to_start_packing;
 	int	area_type;
 };
 
@@ -279,6 +280,38 @@
 	return ret;
 }
 
+static ssize_t
+num_wr_reqs_to_start_packing_show(struct device *dev,
+				  struct device_attribute *attr, char *buf)
+{
+	struct mmc_blk_data *md = mmc_blk_get(dev_to_disk(dev));
+	int num_wr_reqs_to_start_packing;
+	int ret;
+
+	num_wr_reqs_to_start_packing = md->queue.num_wr_reqs_to_start_packing;
+
+	ret = snprintf(buf, PAGE_SIZE, "%d\n", num_wr_reqs_to_start_packing);
+
+	mmc_blk_put(md);
+	return ret;
+}
+
+static ssize_t
+num_wr_reqs_to_start_packing_store(struct device *dev,
+				 struct device_attribute *attr,
+				 const char *buf, size_t count)
+{
+	int value;
+	struct mmc_blk_data *md = mmc_blk_get(dev_to_disk(dev));
+
+	sscanf(buf, "%d", &value);
+	if (value >= 0)
+		md->queue.num_wr_reqs_to_start_packing = value;
+
+	mmc_blk_put(md);
+	return count;
+}
+
 static int mmc_blk_open(struct block_device *bdev, fmode_t mode)
 {
 	struct mmc_blk_data *md = mmc_blk_get(bdev->bd_disk);
@@ -1324,6 +1357,48 @@
 	mmc_queue_bounce_pre(mqrq);
 }
 
+static void mmc_blk_write_packing_control(struct mmc_queue *mq,
+					  struct request *req)
+{
+	struct mmc_host *host = mq->card->host;
+	int data_dir;
+
+	if (!(host->caps2 & MMC_CAP2_PACKED_WR))
+		return;
+
+	/*
+	 * In case the packing control is not supported by the host, it should
+	 * not have an effect on the write packing. Therefore we have to enable
+	 * the write packing
+	 */
+	if (!(host->caps2 & MMC_CAP2_PACKED_WR_CONTROL)) {
+		mq->wr_packing_enabled = true;
+		return;
+	}
+
+	if (!req || (req && (req->cmd_flags & REQ_FLUSH))) {
+		if (mq->num_of_potential_packed_wr_reqs >
+				mq->num_wr_reqs_to_start_packing)
+			mq->wr_packing_enabled = true;
+		return;
+	}
+
+	data_dir = rq_data_dir(req);
+
+	if (data_dir == READ) {
+		mq->num_of_potential_packed_wr_reqs = 0;
+		mq->wr_packing_enabled = false;
+		return;
+	} else if (data_dir == WRITE) {
+		mq->num_of_potential_packed_wr_reqs++;
+	}
+
+	if (mq->num_of_potential_packed_wr_reqs >
+			mq->num_wr_reqs_to_start_packing)
+		mq->wr_packing_enabled = true;
+
+}
+
 static u8 mmc_blk_prep_packed_list(struct mmc_queue *mq, struct request *req)
 {
 	struct request_queue *q = mq->queue;
@@ -1343,6 +1418,9 @@
 			!card->ext_csd.packed_event_en)
 		goto no_packed;
 
+	if (!mq->wr_packing_enabled)
+		goto no_packed;
+
 	if ((rq_data_dir(cur) == WRITE) &&
 			(card->host->caps2 & MMC_CAP2_PACKED_WR))
 		max_packed_rw = card->ext_csd.max_packed_writes;
@@ -1416,6 +1494,8 @@
 			break;
 		}
 
+		if (rq_data_dir(next) == WRITE)
+			mq->num_of_potential_packed_wr_reqs++;
 		list_add_tail(&next->queuelist, &mq->mqrq_cur->packed_list);
 		cur = next;
 		reqs++;
@@ -1797,6 +1877,8 @@
 		goto out;
 	}
 
+	mmc_blk_write_packing_control(mq, req);
+
 	if (req && req->cmd_flags & REQ_SANITIZE) {
 		/* complete ongoing async transfer before issuing sanitize */
 		if (card->host && card->host->areq)
@@ -2028,6 +2110,8 @@
 
 	if (md) {
 		card = md->queue.card;
+		device_remove_file(disk_to_dev(md->disk),
+				   &md->num_wr_reqs_to_start_packing);
 		if (md->disk->flags & GENHD_FL_UP) {
 			device_remove_file(disk_to_dev(md->disk), &md->force_ro);
 			if ((md->area_type & MMC_BLK_DATA_AREA_BOOT) &&
@@ -2095,6 +2179,16 @@
 			goto power_ro_lock_fail;
 	}
 
+	md->num_wr_reqs_to_start_packing.show =
+		num_wr_reqs_to_start_packing_show;
+	md->num_wr_reqs_to_start_packing.store =
+		num_wr_reqs_to_start_packing_store;
+	sysfs_attr_init(&md->num_wr_reqs_to_start_packing.attr);
+	md->num_wr_reqs_to_start_packing.attr.name =
+		"num_wr_reqs_to_start_packing";
+	md->num_wr_reqs_to_start_packing.attr.mode = S_IRUGO | S_IWUSR;
+	ret = device_create_file(disk_to_dev(md->disk),
+				 &md->num_wr_reqs_to_start_packing);
 	if (ret)
 		goto power_ro_lock_fail;
 
diff --git a/drivers/mmc/card/queue.c b/drivers/mmc/card/queue.c
index d818fc4..f3692a9 100644
--- a/drivers/mmc/card/queue.c
+++ b/drivers/mmc/card/queue.c
@@ -25,6 +25,13 @@
 #define MMC_QUEUE_SUSPENDED	(1 << 0)
 
 /*
+ * Based on benchmark tests the default num of requests to trigger the write
+ * packing was determined, to keep the read latency as low as possible and
+ * manage to keep the high write throughput.
+ */
+#define DEFAULT_NUM_REQS_TO_START_PACK 17
+
+/*
  * Prepare a MMC request. This just filters out odd stuff.
  */
 static int mmc_prep_request(struct request_queue *q, struct request *req)
@@ -189,6 +196,7 @@
 	mq->mqrq_cur = mqrq_cur;
 	mq->mqrq_prev = mqrq_prev;
 	mq->queue->queuedata = mq;
+	mq->num_wr_reqs_to_start_packing = DEFAULT_NUM_REQS_TO_START_PACK;
 
 	blk_queue_prep_rq(mq->queue, mmc_prep_request);
 	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, mq->queue);
diff --git a/drivers/mmc/card/queue.h b/drivers/mmc/card/queue.h
index 5e04938..93e4b59 100644
--- a/drivers/mmc/card/queue.h
+++ b/drivers/mmc/card/queue.h
@@ -45,6 +45,9 @@
 	struct mmc_queue_req	mqrq[2];
 	struct mmc_queue_req	*mqrq_cur;
 	struct mmc_queue_req	*mqrq_prev;
+	bool			wr_packing_enabled;
+	int			num_of_potential_packed_wr_reqs;
+	int			num_wr_reqs_to_start_packing;
 };
 
 extern int mmc_init_queue(struct mmc_queue *, struct mmc_card *, spinlock_t *,
diff --git a/include/linux/mmc/host.h b/include/linux/mmc/host.h
index 714cc76..9e536be 100644
--- a/include/linux/mmc/host.h
+++ b/include/linux/mmc/host.h
@@ -245,6 +245,8 @@
 #define MMC_CAP2_PACKED_WR	(1 << 11)	/* Allow packed write */
 #define MMC_CAP2_PACKED_CMD	(MMC_CAP2_PACKED_RD | \
 				 MMC_CAP2_PACKED_WR) /* Allow packed commands */
+#define MMC_CAP2_PACKED_WR_CONTROL (1 << 12) /* Allow write packing control */
+
 #define MMC_CAP2_SANITIZE	(1 << 13)		/* Support Sanitize */
 #define MMC_CAP2_INIT_BKOPS	    (1 << 15)	/* Need to set BKOPS_EN */
 #define MMC_CAP2_POWER_OFF_VCCQ_DURING_SUSPEND	(1 << 16)