Message ID | 1446653619-22880-1-git-send-email-jack@suse.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
> -----Original Message----- > From: linux-kernel-owner@vger.kernel.org [mailto:linux-kernel- > owner@vger.kernel.org] On Behalf Of Jan Kara > Sent: Wednesday, November 4, 2015 10:14 AM > To: axboe@kernel.dk > Cc: LKML <linux-kernel@vger.kernel.org>; linux-fsdevel@vger.kernel.org; > Christoph Hellwig <hch@infradead.org>; Jan Kara <jack@suse.com> > Subject: [PATCH] brd: Refuse improperly aligned discard requests > > Currently when improperly aligned discard request is submitted, we just > silently discard more / less data which results in filesystem corruption > in some cases. Refuse such misaligned requests. I agree discarding more than requested is very bad. If they are routed to SCSI or ATA devices, though, the discard commands (SCSI UNMAP or ATA DATA SET MANAGEMENT/TRIM) are just hints, so there is no guarantee the discard will do anything. Are you finding filesystems that still don't understand that? dm-raid held that mistaken assumption for a long time. --- Robert Elliott, HP Server Storage -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu 05-11-15 04:18:49, Elliott, Robert (Persistent Memory) wrote: > > > -----Original Message----- > > From: linux-kernel-owner@vger.kernel.org [mailto:linux-kernel- > > owner@vger.kernel.org] On Behalf Of Jan Kara > > Sent: Wednesday, November 4, 2015 10:14 AM > > To: axboe@kernel.dk > > Cc: LKML <linux-kernel@vger.kernel.org>; linux-fsdevel@vger.kernel.org; > > Christoph Hellwig <hch@infradead.org>; Jan Kara <jack@suse.com> > > Subject: [PATCH] brd: Refuse improperly aligned discard requests > > > > Currently when improperly aligned discard request is submitted, we just > > silently discard more / less data which results in filesystem corruption > > in some cases. Refuse such misaligned requests. > > I agree discarding more than requested is very bad. > > If they are routed to SCSI or ATA devices, though, the discard commands > (SCSI UNMAP or ATA DATA SET MANAGEMENT/TRIM) are just hints, so there > is no guarantee the discard will do anything. Are you finding > filesystems that still don't understand that? dm-raid held that > mistaken assumption for a long time. So there is blkdev_issue_zeroout() which uses blkdev_issue_discard() if 'discard_zeroes_data' is set. I'd hope that in that case zeroing really happens as submitted or error is returned... Filesystems definitely depend on blkdev_issue_zeroout() doing what it is told. Honza
On 11/04/2015 09:13 AM, Jan Kara wrote: > Currently when improperly aligned discard request is submitted, we just > silently discard more / less data which results in filesystem corruption > in some cases. Refuse such misaligned requests. Applied, thanks Jan.
diff --git a/drivers/block/brd.c b/drivers/block/brd.c index b9794aeeb878..4ef4cdf67ede 100644 --- a/drivers/block/brd.c +++ b/drivers/block/brd.c @@ -337,6 +337,9 @@ static void brd_make_request(struct request_queue *q, struct bio *bio) goto io_error; if (unlikely(bio->bi_rw & REQ_DISCARD)) { + if (sector & ((PAGE_SIZE >> SECTOR_SHIFT) - 1) || + bio->bi_iter.bi_size & PAGE_MASK) + goto io_error; discard_from_brd(brd, sector, bio->bi_iter.bi_size); goto out; }
Currently when improperly aligned discard request is submitted, we just silently discard more / less data which results in filesystem corruption in some cases. Refuse such misaligned requests. Signed-off-by: Jan Kara <jack@suse.com> --- drivers/block/brd.c | 3 +++ 1 file changed, 3 insertions(+)