xen/blkback: Add support for BLKIF_OP_FLUSH_DISKCACHE and drop BLKIF_OP_WRITE_BARRIER.

We drop the support for 'feature-barrier' and add in the support
for the 'feature-flush-cache' if the real backend storage supports
flushing.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>

3 files changed, 34 insertions, 25 deletions

diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
index eb068d0..72ede0b 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -46,8 +46,6 @@
 #include <asm/xen/hypercall.h>
 #include "common.h"
 
-#define WRITE_BARRIER	(REQ_WRITE | REQ_FLUSH | REQ_FUA)
-
 /*
  * These are rather arbitrary. They are fairly large because adjacent requests
  * pulled from a communication ring are quite likely to end up being part of
@@ -256,9 +254,9 @@ irqreturn_t xen_blkif_be_int(int irq, void *dev_id)
 
 static void print_stats(struct blkif_st *blkif)
 {
-	printk(KERN_DEBUG "%s: oo %3d  |  rd %4d  |  wr %4d  |  br %4d\n",
+	printk(KERN_DEBUG "%s: oo %3d  |  rd %4d  |  wr %4d  |  f %4d\n",
 	       current->comm, blkif->st_oo_req,
-	       blkif->st_rd_req, blkif->st_wr_req, blkif->st_br_req);
+	       blkif->st_rd_req, blkif->st_wr_req, blkif->st_f_req);
 	blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000);
 	blkif->st_rd_req = 0;
 	blkif->st_wr_req = 0;
@@ -414,10 +412,10 @@ static int xen_blkbk_map(struct blkif_request *req, struct pending_req *pending_
 static void __end_block_io_op(struct pending_req *pending_req, int error)
 {
 	/* An error fails the entire request. */
-	if ((pending_req->operation == BLKIF_OP_WRITE_BARRIER) &&
+	if ((pending_req->operation == BLKIF_OP_FLUSH_DISKCACHE) &&
 	    (error == -EOPNOTSUPP)) {
-		DPRINTK("blkback: write barrier op failed, not supported\n");
-		xen_blkbk_barrier(XBT_NIL, pending_req->blkif->be, 0);
+		DPRINTK("blkback: flush diskcache op failed, not supported\n");
+		xen_blkbk_flush_diskcache(XBT_NIL, pending_req->blkif->be, 0);
 		pending_req->status = BLKIF_RSP_EOPNOTSUPP;
 	} else if (error) {
 		DPRINTK("Buffer not up-to-date at end of operation, "
@@ -506,13 +504,14 @@ static int do_block_io_op(struct blkif_st *blkif)
 			blkif->st_rd_req++;
 			dispatch_rw_block_io(blkif, &req, pending_req);
 			break;
-		case BLKIF_OP_WRITE_BARRIER:
-			blkif->st_br_req++;
+		case BLKIF_OP_FLUSH_DISKCACHE:
+			blkif->st_f_req++;
 			/* fall through */
 		case BLKIF_OP_WRITE:
 			blkif->st_wr_req++;
 			dispatch_rw_block_io(blkif, &req, pending_req);
 			break;
+		case BLKIF_OP_WRITE_BARRIER:
 		default:
 			/* A good sign something is wrong: sleep for a while to
 			 * avoid excessive CPU consumption by a bad guest. */
@@ -556,9 +555,14 @@ static void dispatch_rw_block_io(struct blkif_st *blkif,
 	case BLKIF_OP_WRITE:
 		operation = WRITE_ODIRECT;
 		break;
-	case BLKIF_OP_WRITE_BARRIER:
-		operation = WRITE_BARRIER;
+	case BLKIF_OP_FLUSH_DISKCACHE:
+		operation = WRITE_FLUSH;
+		/* The frontend likes to set this to -1, which vbd_translate
+		 * is alergic too. */
+		req->u.rw.sector_number = 0;
 		break;
+	case BLKIF_OP_WRITE_BARRIER:
+		/* Should never get here. */
 	default:
 		operation = 0; /* make gcc happy */
 		BUG();
@@ -566,7 +570,7 @@ static void dispatch_rw_block_io(struct blkif_st *blkif,
 
 	/* Check that the number of segments is sane. */
 	nseg = req->nr_segments;
-	if (unlikely(nseg == 0 && operation != WRITE_BARRIER) ||
+	if (unlikely(nseg == 0 && operation != WRITE_FLUSH) ||
 	    unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
 		DPRINTK("Bad number of segments in request (%d)\n", nseg);
 		/* Haven't submitted any bio's yet. */
@@ -643,7 +647,7 @@ static void dispatch_rw_block_io(struct blkif_st *blkif,
 
 	/* This will be hit if the operation was a barrier. */
 	if (!bio) {
-		BUG_ON(operation != WRITE_BARRIER);
+		BUG_ON(operation != WRITE_FLUSH);
 		bio = biolist[nbio++] = bio_alloc(GFP_KERNEL, 0);
 		if (unlikely(bio == NULL))
 			goto fail_put_bio;
@@ -651,7 +655,6 @@ static void dispatch_rw_block_io(struct blkif_st *blkif,
 		bio->bi_bdev    = preq.bdev;
 		bio->bi_private = pending_req;
 		bio->bi_end_io  = end_block_io_op;
-		bio->bi_sector  = -1;
 	}
 
 
@@ -671,7 +674,7 @@ static void dispatch_rw_block_io(struct blkif_st *blkif,
 
 	if (operation == READ)
 		blkif->st_rd_sect += preq.nr_sects;
-	else if (operation == WRITE || operation == WRITE_BARRIER)
+	else if (operation == WRITE || operation == WRITE_FLUSH)
 		blkif->st_wr_sect += preq.nr_sects;
 
 	return;
diff --git a/drivers/block/xen-blkback/common.h b/drivers/block/xen-blkback/common.h
index 16af388..af93837 100644
--- a/drivers/block/xen-blkback/common.h
+++ b/drivers/block/xen-blkback/common.h
@@ -53,6 +53,7 @@ struct vbd {
 	u32            pdevice;     /* phys device that this vbd maps to */
 	struct block_device *bdev;
 	sector_t       size;        /* Cached size parameter */
+	bool           flush_support;
 };
 
 struct backend_info;
@@ -85,7 +86,7 @@ struct blkif_st {
 	int                 st_rd_req;
 	int                 st_wr_req;
 	int                 st_oo_req;
-	int                 st_br_req;
+	int                 st_f_req;
 	int                 st_rd_sect;
 	int                 st_wr_sect;
 
@@ -120,8 +121,8 @@ int xen_blkif_xenbus_init(void);
 irqreturn_t xen_blkif_be_int(int irq, void *dev_id);
 int xen_blkif_schedule(void *arg);
 
-int xen_blkbk_barrier(struct xenbus_transaction xbt,
-			struct backend_info *be, int state);
+int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
+			      struct backend_info *be, int state);
 
 struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
 
diff --git a/drivers/block/xen-blkback/xenbus.c b/drivers/block/xen-blkback/xenbus.c
index 64b0a1c..9adcf80 100644
--- a/drivers/block/xen-blkback/xenbus.c
+++ b/drivers/block/xen-blkback/xenbus.c
@@ -276,7 +276,7 @@ int __init xen_blkif_interface_init(void)
 VBD_SHOW(oo_req,  "%d\n", be->blkif->st_oo_req);
 VBD_SHOW(rd_req,  "%d\n", be->blkif->st_rd_req);
 VBD_SHOW(wr_req,  "%d\n", be->blkif->st_wr_req);
-VBD_SHOW(br_req,  "%d\n", be->blkif->st_br_req);
+VBD_SHOW(f_req,  "%d\n", be->blkif->st_f_req);
 VBD_SHOW(rd_sect, "%d\n", be->blkif->st_rd_sect);
 VBD_SHOW(wr_sect, "%d\n", be->blkif->st_wr_sect);
 
@@ -284,7 +284,7 @@ static struct attribute *vbdstat_attrs[] = {
 	&dev_attr_oo_req.attr,
 	&dev_attr_rd_req.attr,
 	&dev_attr_wr_req.attr,
-	&dev_attr_br_req.attr,
+	&dev_attr_f_req.attr,
 	&dev_attr_rd_sect.attr,
 	&dev_attr_wr_sect.attr,
 	NULL
@@ -343,6 +343,7 @@ static int vbd_create(struct blkif_st *blkif, blkif_vdev_t handle,
 {
 	struct vbd *vbd;
 	struct block_device *bdev;
+	struct request_queue *q;
 
 	vbd = &blkif->vbd;
 	vbd->handle   = handle;
@@ -375,6 +376,10 @@ static int vbd_create(struct blkif_st *blkif, blkif_vdev_t handle,
 	if (vbd->bdev->bd_disk->flags & GENHD_FL_REMOVABLE)
 		vbd->type |= VDISK_REMOVABLE;
 
+	q = bdev_get_queue(bdev);
+	if (q && q->flush_flags)
+		vbd->flush_support = true;
+
 	DPRINTK("Successful creation of handle=%04x (dom=%u)\n",
 		handle, blkif->domid);
 	return 0;
@@ -406,16 +411,16 @@ static int xen_blkbk_remove(struct xenbus_device *dev)
 	return 0;
 }
 
-int xen_blkbk_barrier(struct xenbus_transaction xbt,
-		      struct backend_info *be, int state)
+int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
+			      struct backend_info *be, int state)
 {
 	struct xenbus_device *dev = be->dev;
 	int err;
 
-	err = xenbus_printf(xbt, dev->nodename, "feature-barrier",
+	err = xenbus_printf(xbt, dev->nodename, "feature-flush-cache",
 			    "%d", state);
 	if (err)
-		xenbus_dev_fatal(dev, err, "writing feature-barrier");
+		xenbus_dev_fatal(dev, err, "writing feature-flush-cache");
 
 	return err;
 }
@@ -642,7 +647,7 @@ again:
 		return;
 	}
 
-	err = xen_blkbk_barrier(xbt, be, 1);
+	err = xen_blkbk_flush_diskcache(xbt, be, be->blkif->vbd.flush_support);
 	if (err)
 		goto abort;