block, drivers: add REQ_OP_FLUSH operation

This adds a REQ_OP_FLUSH operation that is sent to request_fn
based drivers by the block layer's flush code, instead of
sending requests with the request->cmd_flags REQ_FLUSH bit set.

Signed-off-by: Mike Christie <mchristi@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

Documentation/block/writeback_cache_control.txt

@@ -73,9 +73,9 @@ doing:
 
 	blk_queue_write_cache(sdkp->disk->queue, true, false);
 
-and handle empty REQ_FLUSH requests in its prep_fn/request_fn.  Note that
+and handle empty REQ_OP_FLUSH requests in its prep_fn/request_fn.  Note that
 REQ_FLUSH requests with a payload are automatically turned into a sequence
-of an empty REQ_FLUSH request followed by the actual write by the block
+of an empty REQ_OP_FLUSH request followed by the actual write by the block
 layer.  For devices that also support the FUA bit the block layer needs
 to be told to pass through the REQ_FUA bit using:
 
@@ -83,4 +83,4 @@ to be told to pass through the REQ_FUA bit using:
 
 and the driver must handle write requests that have the REQ_FUA bit set
 in prep_fn/request_fn.  If the FUA bit is not natively supported the block
-layer turns it into an empty REQ_FLUSH request after the actual write.
+layer turns it into an empty REQ_OP_FLUSH request after the actual write.

arch/um/drivers/ubd_kern.c

@@ -1286,7 +1286,7 @@ static void do_ubd_request(struct request_queue *q)
 
 		req = dev->request;
 
-		if (req->cmd_flags & REQ_FLUSH) {
+		if (req_op(req) == REQ_OP_FLUSH) {
 			io_req = kmalloc(sizeof(struct io_thread_req),
 					 GFP_ATOMIC);
 			if (io_req == NULL) {

block/blk-flush.c

@@ -29,7 +29,7 @@
  * The actual execution of flush is double buffered.  Whenever a request
  * needs to execute PRE or POSTFLUSH, it queues at
  * fq->flush_queue[fq->flush_pending_idx].  Once certain criteria are met, a
- * flush is issued and the pending_idx is toggled.  When the flush
+ * REQ_OP_FLUSH is issued and the pending_idx is toggled.  When the flush
  * completes, all the requests which were pending are proceeded to the next
  * step.  This allows arbitrary merging of different types of FLUSH/FUA
  * requests.
@@ -330,7 +330,7 @@ static bool blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq)
 	}
 
 	flush_rq->cmd_type = REQ_TYPE_FS;
-	flush_rq->cmd_flags = WRITE_FLUSH | REQ_FLUSH_SEQ;
+	req_set_op_attrs(flush_rq, REQ_OP_FLUSH, WRITE_FLUSH | REQ_FLUSH_SEQ);
 	flush_rq->rq_disk = first_rq->rq_disk;
 	flush_rq->end_io = flush_end_io;
 

drivers/block/loop.c

@@ -542,7 +542,7 @@ static int do_req_filebacked(struct loop_device *lo, struct request *rq)
 	pos = ((loff_t) blk_rq_pos(rq) << 9) + lo->lo_offset;
 
 	if (op_is_write(req_op(rq))) {
-		if (rq->cmd_flags & REQ_FLUSH)
+		if (req_op(rq) == REQ_OP_FLUSH)
 			ret = lo_req_flush(lo, rq);
 		else if (req_op(rq) == REQ_OP_DISCARD)
 			ret = lo_discard(lo, rq, pos);
@@ -1659,7 +1659,7 @@ static int loop_queue_rq(struct blk_mq_hw_ctx *hctx,
 	if (lo->lo_state != Lo_bound)
 		return -EIO;
 
-	if (lo->use_dio && (!(cmd->rq->cmd_flags & REQ_FLUSH) ||
+	if (lo->use_dio && (req_op(cmd->rq) != REQ_OP_FLUSH ||
 	    req_op(cmd->rq) == REQ_OP_DISCARD))
 		cmd->use_aio = true;
 	else

drivers/block/nbd.c

@@ -284,7 +284,7 @@ static int nbd_send_req(struct nbd_device *nbd, struct request *req)
 		type = NBD_CMD_DISC;
 	else if (req_op(req) == REQ_OP_DISCARD)
 		type = NBD_CMD_TRIM;
-	else if (req->cmd_flags & REQ_FLUSH)
+	else if (req_op(req) == REQ_OP_FLUSH)
 		type = NBD_CMD_FLUSH;
 	else if (rq_data_dir(req) == WRITE)
 		type = NBD_CMD_WRITE;

drivers/block/osdblk.c

@@ -321,7 +321,7 @@ static void osdblk_rq_fn(struct request_queue *q)
 		 * driver-specific, etc.
 		 */
 
-		do_flush = rq->cmd_flags & REQ_FLUSH;
+		do_flush = (req_op(rq) == REQ_OP_FLUSH);
 		do_write = (rq_data_dir(rq) == WRITE);
 
 		if (!do_flush) { /* osd_flush does not use a bio */

drivers/block/ps3disk.c

@@ -196,7 +196,7 @@ static void ps3disk_do_request(struct ps3_storage_device *dev,
 	dev_dbg(&dev->sbd.core, "%s:%u\n", __func__, __LINE__);
 
 	while ((req = blk_fetch_request(q))) {
-		if (req->cmd_flags & REQ_FLUSH) {
+		if (req_op(req) == REQ_OP_FLUSH) {
 			if (ps3disk_submit_flush_request(dev, req))
 				break;
 		} else if (req->cmd_type == REQ_TYPE_FS) {
@@ -256,7 +256,7 @@ static irqreturn_t ps3disk_interrupt(int irq, void *data)
 		return IRQ_HANDLED;
 	}
 
-	if (req->cmd_flags & REQ_FLUSH) {
+	if (req_op(req) == REQ_OP_FLUSH) {
 		read = 0;
 		op = "flush";
 	} else {

drivers/block/skd_main.c

@@ -597,7 +597,7 @@ static void skd_request_fn(struct request_queue *q)
 		data_dir = rq_data_dir(req);
 		io_flags = req->cmd_flags;
 
-		if (io_flags & REQ_FLUSH)
+		if (req_op(req) == REQ_OP_FLUSH)
 			flush++;
 
 		if (io_flags & REQ_FUA)

drivers/block/virtio_blk.c

@@ -172,7 +172,7 @@ static int virtio_queue_rq(struct blk_mq_hw_ctx *hctx,
 	BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems);
 
 	vbr->req = req;
-	if (req->cmd_flags & REQ_FLUSH) {
+	if (req_op(req) == REQ_OP_FLUSH) {
 		vbr->out_hdr.type = cpu_to_virtio32(vblk->vdev, VIRTIO_BLK_T_FLUSH);
 		vbr->out_hdr.sector = 0;
 		vbr->out_hdr.ioprio = cpu_to_virtio32(vblk->vdev, req_get_ioprio(vbr->req));

drivers/block/xen-blkfront.c

@@ -743,7 +743,7 @@ static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *ri
 		 * The indirect operation can only be a BLKIF_OP_READ or
 		 * BLKIF_OP_WRITE
 		 */
-		BUG_ON(req->cmd_flags & (REQ_FLUSH | REQ_FUA));
+		BUG_ON(req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA);
 		ring_req->operation = BLKIF_OP_INDIRECT;
 		ring_req->u.indirect.indirect_op = rq_data_dir(req) ?
 			BLKIF_OP_WRITE : BLKIF_OP_READ;
@@ -755,7 +755,7 @@ static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *ri
 		ring_req->u.rw.handle = info->handle;
 		ring_req->operation = rq_data_dir(req) ?
 			BLKIF_OP_WRITE : BLKIF_OP_READ;
-		if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) {
+		if (req_op(req) == REQ_OP_FLUSH || req->cmd_flags & REQ_FUA) {
 			/*
 			 * Ideally we can do an unordered flush-to-disk.
 			 * In case the backend onlysupports barriers, use that.
@@ -865,7 +865,7 @@ static inline bool blkif_request_flush_invalid(struct request *req,
 					       struct blkfront_info *info)
 {
 	return ((req->cmd_type != REQ_TYPE_FS) ||
-		((req->cmd_flags & REQ_FLUSH) &&
+		((req_op(req) == REQ_OP_FLUSH) &&
 		 !(info->feature_flush & REQ_FLUSH)) ||
 		((req->cmd_flags & REQ_FUA) &&
 		 !(info->feature_flush & REQ_FUA)));
@@ -2055,7 +2055,7 @@ static int blkif_recover(struct blkfront_info *info)
 			/*
 			 * Get the bios in the request so we can re-queue them.
 			 */
-			if (copy[i].request->cmd_flags & REQ_FLUSH ||
+			if (req_op(copy[i].request) == REQ_OP_FLUSH ||
 			    req_op(copy[i].request) == REQ_OP_DISCARD ||
 			    copy[i].request->cmd_flags & (REQ_FUA | REQ_SECURE)) {
 				/*

drivers/ide/ide-disk.c

@@ -431,7 +431,7 @@ static int idedisk_prep_fn(struct request_queue *q, struct request *rq)
 	ide_drive_t *drive = q->queuedata;
 	struct ide_cmd *cmd;
 
-	if (!(rq->cmd_flags & REQ_FLUSH))
+	if (req_op(rq) != REQ_OP_FLUSH)
 		return BLKPREP_OK;
 
 	if (rq->special) {

drivers/md/dm.c

@@ -2171,7 +2171,7 @@ static void dm_request_fn(struct request_queue *q)
 
 		/* always use block 0 to find the target for flushes for now */
 		pos = 0;
-		if (!(rq->cmd_flags & REQ_FLUSH))
+		if (req_op(rq) != REQ_OP_FLUSH)
 			pos = blk_rq_pos(rq);
 
 		if ((dm_request_peeked_before_merge_deadline(md) &&

drivers/mmc/card/block.c

@@ -1722,7 +1722,8 @@ static u8 mmc_blk_prep_packed_list(struct mmc_queue *mq, struct request *req)
 		    !IS_ALIGNED(blk_rq_sectors(next), 8))
 			break;
 
-		if (req_op(next) == REQ_OP_DISCARD || next->cmd_flags & REQ_FLUSH)
+		if (req_op(next) == REQ_OP_DISCARD ||
+		    req_op(next) == REQ_OP_FLUSH)
 			break;
 
 		if (rq_data_dir(cur) != rq_data_dir(next))
@@ -2147,7 +2148,6 @@ static int mmc_blk_issue_rq(struct mmc_queue *mq, struct request *req)
 	struct mmc_card *card = md->queue.card;
 	struct mmc_host *host = card->host;
 	unsigned long flags;
-	unsigned int cmd_flags = req ? req->cmd_flags : 0;
 
 	if (req && !mq->mqrq_prev->req)
 		/* claim host only for the first request */
@@ -2171,7 +2171,7 @@ static int mmc_blk_issue_rq(struct mmc_queue *mq, struct request *req)
 			ret = mmc_blk_issue_secdiscard_rq(mq, req);
 		else
 			ret = mmc_blk_issue_discard_rq(mq, req);
-	} else if (cmd_flags & REQ_FLUSH) {
+	} else if (req && req_op(req) == REQ_OP_FLUSH) {
 		/* complete ongoing async transfer before issuing flush */
 		if (card->host->areq)
 			mmc_blk_issue_rw_rq(mq, NULL);

drivers/mmc/card/queue.h

@@ -3,7 +3,8 @@
 
 static inline bool mmc_req_is_special(struct request *req)
 {
-	return req && (req->cmd_flags & REQ_FLUSH || req_op(req) == REQ_OP_DISCARD);
+	return req &&
+		(req_op(req) == REQ_OP_FLUSH || req_op(req) == REQ_OP_DISCARD);
 }
 
 struct request;

drivers/mtd/mtd_blkdevs.c

@@ -87,7 +87,7 @@ static int do_blktrans_request(struct mtd_blktrans_ops *tr,
 	if (req->cmd_type != REQ_TYPE_FS)
 		return -EIO;
 
-	if (req->cmd_flags & REQ_FLUSH)
+	if (req_op(req) == REQ_OP_FLUSH)
 		return tr->flush(dev);
 
 	if (blk_rq_pos(req) + blk_rq_cur_sectors(req) >

drivers/nvme/host/core.c

@@ -290,7 +290,7 @@ int nvme_setup_cmd(struct nvme_ns *ns, struct request *req,
 
 	if (req->cmd_type == REQ_TYPE_DRV_PRIV)
 		memcpy(cmd, req->cmd, sizeof(*cmd));
-	else if (req->cmd_flags & REQ_FLUSH)
+	else if (req_op(req) == REQ_OP_FLUSH)
 		nvme_setup_flush(ns, cmd);
 	else if (req_op(req) == REQ_OP_DISCARD)
 		ret = nvme_setup_discard(ns, req, cmd);

drivers/scsi/sd.c

@@ -1143,12 +1143,11 @@ static int sd_init_command(struct scsi_cmnd *cmd)
 		return sd_setup_discard_cmnd(cmd);
 	case REQ_OP_WRITE_SAME:
 		return sd_setup_write_same_cmnd(cmd);
+	case REQ_OP_FLUSH:
+		return sd_setup_flush_cmnd(cmd);
 	case REQ_OP_READ:
 	case REQ_OP_WRITE:
-		if (rq->cmd_flags & REQ_FLUSH)
-			return sd_setup_flush_cmnd(cmd);
-		else
-			return sd_setup_read_write_cmnd(cmd);
+		return sd_setup_read_write_cmnd(cmd);
 	default:
 		BUG();
 	}

include/linux/blk_types.h

@@ -249,9 +249,10 @@ enum req_op {
 	REQ_OP_WRITE,
 	REQ_OP_DISCARD,		/* request to discard sectors */
 	REQ_OP_WRITE_SAME,	/* write same block many times */
+	REQ_OP_FLUSH,		/* request for cache flush */
 };
 
-#define REQ_OP_BITS 2
+#define REQ_OP_BITS 3
 
 typedef unsigned int blk_qc_t;
 #define BLK_QC_T_NONE	-1U

include/linux/blkdev.h

@@ -666,6 +666,9 @@ static inline bool rq_mergeable(struct request *rq)
 	if (rq->cmd_type != REQ_TYPE_FS)
 		return false;
 
+	if (req_op(rq) == REQ_OP_FLUSH)
+		return false;
+
 	if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
 		return false;
 

kernel/trace/blktrace.c

@@ -223,6 +223,8 @@ static void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes,
 	what |= MASK_TC_BIT(op_flags, FUA);
 	if (op == REQ_OP_DISCARD)
 		what |= BLK_TC_ACT(BLK_TC_DISCARD);
+	if (op == REQ_OP_FLUSH)
+		what |= BLK_TC_ACT(BLK_TC_FLUSH);
 
 	pid = tsk->pid;
 	if (act_log_check(bt, what, sector, pid))
@@ -1788,6 +1790,9 @@ void blk_fill_rwbs(char *rwbs, int op, u32 rw, int bytes)
 	case REQ_OP_DISCARD:
 		rwbs[i++] = 'D';
 		break;
+	case REQ_OP_FLUSH:
+		rwbs[i++] = 'F';
+		break;
 	case REQ_OP_READ:
 		rwbs[i++] = 'R';
 		break;