summaryrefslogtreecommitdiff
path: root/include
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2012-12-19 07:18:35 -0800
committerLinus Torvalds <torvalds@linux-foundation.org>2012-12-19 07:18:35 -0800
commit59771079c18c44e39106f0f30054025acafadb41 (patch)
tree9463781cf1d6f3055bc87840190cc322b59daa67 /include
parent752451f01c4567b506bf4343082682dbb8fb30dd (diff)
downloadlwn-59771079c18c44e39106f0f30054025acafadb41.tar.gz
lwn-59771079c18c44e39106f0f30054025acafadb41.zip
blk: avoid divide-by-zero with zero discard granularity
Commit 8dd2cb7e880d ("block: discard granularity might not be power of 2") changed a couple of 'binary and' operations into modulus operations. Which turned the harmless case of a zero discard_granularity into a possible divide-by-zero. The code also had a much more subtle bug: it was doing the modulus of a value in bytes using 'sector_t'. That was always conceptually wrong, but didn't actually matter back when the code assumed a power-of-two granularity: we only looked at the low bits anyway. But with potentially arbitrary sector numbers, using a 'sector_t' to express bytes is very very wrong: depending on configuration it limits the starting offset of the device to just 32 bits, and any overflow would result in a wrong value if the modulus wasn't a power-of-two. So re-write the code to not only protect against the divide-by-zero, but to do the starting sector arithmetic in sectors, and using the proper types. [ For any mathematicians out there: it also looks monumentally stupid to do the 'modulo granularity' operation *twice*, never mind having a "+ granularity" in the second modulus op. But that's the easiest way to avoid negative values or overflow, and it is how the original code was done. ] Reported-by: Ingo Molnar <mingo@kernel.org> Reported-by: Doug Anderson <dianders@chromium.org> Cc: Neil Brown <neilb@suse.de> Cc: Shaohua Li <shli@fusionio.com> Acked-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'include')
-rw-r--r--include/linux/blkdev.h19
1 files changed, 15 insertions, 4 deletions
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index acb4f7bbbd32..f94bc83011ed 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1188,14 +1188,25 @@ static inline int queue_discard_alignment(struct request_queue *q)
static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector)
{
- sector_t alignment = sector << 9;
- alignment = sector_div(alignment, lim->discard_granularity);
+ unsigned int alignment, granularity, offset;
if (!lim->max_discard_sectors)
return 0;
- alignment = lim->discard_granularity + lim->discard_alignment - alignment;
- return sector_div(alignment, lim->discard_granularity);
+ /* Why are these in bytes, not sectors? */
+ alignment = lim->discard_alignment >> 9;
+ granularity = lim->discard_granularity >> 9;
+ if (!granularity)
+ return 0;
+
+ /* Offset of the partition start in 'granularity' sectors */
+ offset = sector_div(sector, granularity);
+
+ /* And why do we do this modulus *again* in blkdev_issue_discard()? */
+ offset = (granularity + alignment - offset) % granularity;
+
+ /* Turn it back into bytes, gaah */
+ return offset << 9;
}
static inline int bdev_discard_alignment(struct block_device *bdev)