block: introduce BLKDEV_DISCARD_ZERO to fix zeroout

Currently blkdev_issue_zeroout cascades down from discards (if the driver
guarantees that discards zero data), to WRITE SAME and then to a loop
writing zeroes.  Unfortunately we ignore run-time EOPNOTSUPP errors in the
block layer blkdev_issue_discard helper to work around DM volumes that
may have mixed discard support underneath.

This patch intoroduces a new BLKDEV_DISCARD_ZERO flag to
blkdev_issue_discard that indicates we are called for zeroing operation.
This allows both to ignore the EOPNOTSUPP hack and actually consolidating
the discard_zeroes_data check into the function.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

diff --git a/block/blk-lib.c b/block/blk-lib.c
index 78626c2fde3397f20ae1cea7bb46ccf67c297f8e..45b35b15496f20d24feff007e2149cf5dfe26560 100644 (file)
--- a/block/blk-lib.c
+++ b/block/blk-lib.c
@@ -36,12 +36,17 @@ int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
                return -ENXIO;
 
        if (flags & BLKDEV_DISCARD_SECURE) {
+               if (flags & BLKDEV_DISCARD_ZERO)
+                       return -EOPNOTSUPP;
                if (!blk_queue_secure_erase(q))
                        return -EOPNOTSUPP;
                op = REQ_OP_SECURE_ERASE;
        } else {
                if (!blk_queue_discard(q))
                        return -EOPNOTSUPP;
+               if ((flags & BLKDEV_DISCARD_ZERO) &&
+                   !q->limits.discard_zeroes_data)
+                       return -EOPNOTSUPP;
                op = REQ_OP_DISCARD;
        }
 
@@ -116,7 +121,7 @@ int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
                        &bio);
        if (!ret && bio) {
                ret = submit_bio_wait(bio);
-               if (ret == -EOPNOTSUPP)
+               if (ret == -EOPNOTSUPP && !(flags & BLKDEV_DISCARD_ZERO))
                        ret = 0;
        }
        blk_finish_plug(&plug);
@@ -241,11 +246,11 @@ static int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
 int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
                         sector_t nr_sects, gfp_t gfp_mask, bool discard)
 {
-       struct request_queue *q = bdev_get_queue(bdev);
-
-       if (discard && blk_queue_discard(q) && q->limits.discard_zeroes_data &&
-           blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask, 0) == 0)
-               return 0;
+       if (discard) {
+               if (!blkdev_issue_discard(bdev, sector, nr_sects, gfp_mask,
+                               BLKDEV_DISCARD_ZERO))
+                       return 0;
+       }
 
        if (bdev_write_same(bdev) &&
            blkdev_issue_write_same(bdev, sector, nr_sects, gfp_mask,

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 53fee6123893733aeab67d5012653b742fae01ec..156455cb07ad00f68b66af4c5e5f7cd1222597ed 100644 (file)
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1137,7 +1137,9 @@ static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
        return bqt->tag_index[tag];
 }
 
-#define BLKDEV_DISCARD_SECURE  0x01    /* secure discard */
+
+#define BLKDEV_DISCARD_SECURE  (1 << 0)        /* issue a secure erase */
+#define BLKDEV_DISCARD_ZERO    (1 << 1)        /* must reliably zero data */
 
 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *);
 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector,