This is a note to let you know that I've just added the patch titled
block: don't let passthrough IO go into .make_request_fn()
to the 4.14-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git%3Ba=su...
The filename of the patch is: block-don-t-let-passthrough-io-go-into-.make_request_fn.patch and it can be found in the queue-4.14 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree, please let stable@vger.kernel.org know about it.
From 14cb0dc6479dc5ebc63b3a459a5d89a2f1b39fed Mon Sep 17 00:00:00 2001
From: Ming Lei ming.lei@redhat.com Date: Mon, 18 Dec 2017 15:40:43 +0800 Subject: block: don't let passthrough IO go into .make_request_fn()
From: Ming Lei ming.lei@redhat.com
commit 14cb0dc6479dc5ebc63b3a459a5d89a2f1b39fed upstream.
Commit a8821f3f3("block: Improvements to bounce-buffer handling") tries to make sure that the bio to .make_request_fn won't exceed BIO_MAX_PAGES, but ignores that passthrough I/O can use blk_queue_bounce() too. Especially, passthrough IO may not be sector-aligned, and the check of 'sectors < bio_sectors(*bio_orig)' inside __blk_queue_bounce() may become true even though the max bvec number doesn't exceed BIO_MAX_PAGES, then cause the bio splitted, and the original passthrough bio is submited to generic_make_request().
This patch fixes this issue by checking if the bio is passthrough IO, and use bio_kmalloc() to allocate the cloned passthrough bio.
Cc: NeilBrown neilb@suse.com Fixes: a8821f3f3("block: Improvements to bounce-buffer handling") Tested-by: Michele Ballabio barra_cuda@katamail.com Signed-off-by: Ming Lei ming.lei@redhat.com Signed-off-by: Jens Axboe axboe@kernel.dk Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- block/bounce.c | 6 ++++-- include/linux/blkdev.h | 21 +++++++++++++++++++-- 2 files changed, 23 insertions(+), 4 deletions(-)
--- a/block/bounce.c +++ b/block/bounce.c @@ -200,6 +200,7 @@ static void __blk_queue_bounce(struct re unsigned i = 0; bool bounce = false; int sectors = 0; + bool passthrough = bio_is_passthrough(*bio_orig);
bio_for_each_segment(from, *bio_orig, iter) { if (i++ < BIO_MAX_PAGES) @@ -210,13 +211,14 @@ static void __blk_queue_bounce(struct re if (!bounce) return;
- if (sectors < bio_sectors(*bio_orig)) { + if (!passthrough && sectors < bio_sectors(*bio_orig)) { bio = bio_split(*bio_orig, sectors, GFP_NOIO, bounce_bio_split); bio_chain(bio, *bio_orig); generic_make_request(*bio_orig); *bio_orig = bio; } - bio = bio_clone_bioset(*bio_orig, GFP_NOIO, bounce_bio_set); + bio = bio_clone_bioset(*bio_orig, GFP_NOIO, passthrough ? NULL : + bounce_bio_set);
bio_for_each_segment_all(to, bio, i) { struct page *page = to->bv_page; --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -241,14 +241,24 @@ struct request { struct request *next_rq; };
+static inline bool blk_op_is_scsi(unsigned int op) +{ + return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; +} + +static inline bool blk_op_is_private(unsigned int op) +{ + return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; +} + static inline bool blk_rq_is_scsi(struct request *rq) { - return req_op(rq) == REQ_OP_SCSI_IN || req_op(rq) == REQ_OP_SCSI_OUT; + return blk_op_is_scsi(req_op(rq)); }
static inline bool blk_rq_is_private(struct request *rq) { - return req_op(rq) == REQ_OP_DRV_IN || req_op(rq) == REQ_OP_DRV_OUT; + return blk_op_is_private(req_op(rq)); }
static inline bool blk_rq_is_passthrough(struct request *rq) @@ -256,6 +266,13 @@ static inline bool blk_rq_is_passthrough return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); }
+static inline bool bio_is_passthrough(struct bio *bio) +{ + unsigned op = bio_op(bio); + + return blk_op_is_scsi(op) || blk_op_is_private(op); +} + static inline unsigned short req_get_ioprio(struct request *req) { return req->ioprio;
Patches currently in stable-queue which might be from ming.lei@redhat.com are
queue-4.14/block-don-t-let-passthrough-io-go-into-.make_request_fn.patch queue-4.14/block-fix-blk_rq_append_bio.patch