dm cache: fix truncation bug when mapping I/O to >2TB fast device
authorHeinz Mauelshagen <heinzm@redhat.com>
Thu, 27 Feb 2014 21:46:48 +0000 (22:46 +0100)
committerMike Snitzer <snitzer@redhat.com>
Fri, 28 Feb 2014 14:23:02 +0000 (09:23 -0500)
When remapping a block to the cache's fast device that is larger than
2TB we must not truncate the destination sector to 32bits.  The 32bit
temporary result of from_cblock() was being overflowed in
remap_to_cache() due to the logical left shift.

Use an intermediate 64bit type to store the 32bit from_cblock() result
to fix the overflow.

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
drivers/md/dm-cache-target.c

index db0944465127d7d8eea09185e004d1beaa375445..1af70145fab9bcee990dd54c08224f9e16924851 100644 (file)
@@ -671,15 +671,16 @@ static void remap_to_cache(struct cache *cache, struct bio *bio,
                           dm_cblock_t cblock)
 {
        sector_t bi_sector = bio->bi_iter.bi_sector;
+       sector_t block = from_cblock(cblock);
 
        bio->bi_bdev = cache->cache_dev->bdev;
        if (!block_size_is_power_of_two(cache))
                bio->bi_iter.bi_sector =
-                       (from_cblock(cblock) * cache->sectors_per_block) +
+                       (block * cache->sectors_per_block) +
                        sector_div(bi_sector, cache->sectors_per_block);
        else
                bio->bi_iter.bi_sector =
-                       (from_cblock(cblock) << cache->sectors_per_block_shift) |
+                       (block << cache->sectors_per_block_shift) |
                        (bi_sector & (cache->sectors_per_block - 1));
 }