virtio: block: dynamic maximum segments

Enhance the driver to handle whatever maximum segment number the host
tells us to handle.  Do to this, we need to allocate the scatterlist
dynamically.

We set max_phys_segments and max_hw_segments to the same value (1 if
the host doesn't tell us, since that's safest and all known hosts do
tell us).

Note that kmalloc'ing the structure for large sg_elems might be
problematic: the fix for this is sg_table, but that requires more
work.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

1 files changed, 23 insertions, 13 deletions

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 0e2ade648aff..7b9b38a12d25 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -6,7 +6,6 @@
 #include <linux/virtio_blk.h>
 #include <linux/scatterlist.h>
 
-#define VIRTIO_MAX_SG	(3+MAX_PHYS_SEGMENTS)
 #define PART_BITS 4
 
 static int major, index;
@@ -26,8 +25,11 @@ struct virtio_blk
 
 	mempool_t *pool;
 
+	/* What host tells us, plus 2 for header & tailer. */
+	unsigned int sg_elems;
+
 	/* Scatterlist: can be too big for stack. */
-	struct scatterlist sg[VIRTIO_MAX_SG];
+	struct scatterlist sg[/*sg_elems*/];
 };
 
 struct virtblk_req
@@ -97,8 +99,6 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
 	if (blk_barrier_rq(vbr->req))
 		vbr->out_hdr.type |= VIRTIO_BLK_T_BARRIER;
 
-	/* This init could be done at vblk creation time */
-	sg_init_table(vblk->sg, VIRTIO_MAX_SG);
 	sg_set_buf(&vblk->sg[0], &vbr->out_hdr, sizeof(vbr->out_hdr));
 	num = blk_rq_map_sg(q, vbr->req, vblk->sg+1);
 	sg_set_buf(&vblk->sg[num+1], &vbr->status, sizeof(vbr->status));
@@ -130,7 +130,7 @@ static void do_virtblk_request(struct request_queue *q)
 
 	while ((req = elv_next_request(q)) != NULL) {
 		vblk = req->rq_disk->private_data;
-		BUG_ON(req->nr_phys_segments > ARRAY_SIZE(vblk->sg));
+		BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems);
 
 		/* If this request fails, stop queue and wait for something to
 		   finish to restart it. */
@@ -196,12 +196,22 @@ static int virtblk_probe(struct virtio_device *vdev)
 	int err;
 	u64 cap;
 	u32 v;
-	u32 blk_size;
+	u32 blk_size, sg_elems;
 
 	if (index_to_minor(index) >= 1 << MINORBITS)
 		return -ENOSPC;
 
-	vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL);
+	/* We need to know how many segments before we allocate. */
+	err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
+				offsetof(struct virtio_blk_config, seg_max),
+				&sg_elems);
+	if (err)
+		sg_elems = 1;
+
+	/* We need an extra sg elements at head and tail. */
+	sg_elems += 2;
+	vdev->priv = vblk = kmalloc(sizeof(*vblk) +
+				    sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL);
 	if (!vblk) {
 		err = -ENOMEM;
 		goto out;
@@ -210,6 +220,8 @@ static int virtblk_probe(struct virtio_device *vdev)
 	INIT_LIST_HEAD(&vblk->reqs);
 	spin_lock_init(&vblk->lock);
 	vblk->vdev = vdev;
+	vblk->sg_elems = sg_elems;
+	sg_init_table(vblk->sg, vblk->sg_elems);
 
 	/* We expect one virtqueue, for output. */
 	vblk->vq = vdev->config->find_vq(vdev, 0, blk_done);
@@ -277,6 +289,10 @@ static int virtblk_probe(struct virtio_device *vdev)
 	}
 	set_capacity(vblk->disk, cap);
 
+	/* We can handle whatever the host told us to handle. */
+	blk_queue_max_phys_segments(vblk->disk->queue, vblk->sg_elems-2);
+	blk_queue_max_hw_segments(vblk->disk->queue, vblk->sg_elems-2);
+
 	/* No real sector limit. */
 	blk_queue_max_sectors(vblk->disk->queue, -1U);
 
@@ -290,12 +306,6 @@ static int virtblk_probe(struct virtio_device *vdev)
 	else
 		blk_queue_max_segment_size(vblk->disk->queue, -1UL);
 
-	err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
-				offsetof(struct virtio_blk_config, seg_max),
-				&v);
-	if (!err)
-		blk_queue_max_hw_segments(vblk->disk->queue, v);
-
 	/* Host can optionally specify the block size of the device */
 	err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE,
 				offsetof(struct virtio_blk_config, blk_size),