NVMe: Fix discontiguous accesses

When we submit subsequent portions of the I/O, we need to access the
updated block, not start reading again from the original position.
This was showing up as miscompares in the XFS randholes testcase.

Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>

diff --git a/drivers/block/nvme.c b/drivers/block/nvme.c
index 2409227..562d75a 100644 (file)
--- a/drivers/block/nvme.c
+++ b/drivers/block/nvme.c
@@ -529,6 +529,8 @@ static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
        cmnd->rw.control = cpu_to_le16(control);
        cmnd->rw.dsmgmt = cpu_to_le32(dsmgmt);
 
+       bio->bi_sector += length >> 9;
+
        if (++nvmeq->sq_tail == nvmeq->q_depth)
                nvmeq->sq_tail = 0;
        writel(nvmeq->sq_tail, nvmeq->q_db);