blk-mq: call blk_mq_start_request from ->queue_rq
When we call blk_mq_start_request from the core blk-mq code before calling into
->queue_rq there is a racy window where the timeout handler can hit before we've
fully set up the driver specific part of the command.
Move the call to blk_mq_start_request into the driver so the driver can start
the request only once it is fully set up.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 32b4797..141f2e0 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -384,7 +384,7 @@
}
EXPORT_SYMBOL(blk_mq_complete_request);
-static void blk_mq_start_request(struct request *rq)
+void blk_mq_start_request(struct request *rq)
{
struct request_queue *q = rq->q;
@@ -422,16 +422,18 @@
rq->nr_phys_segments++;
}
}
+EXPORT_SYMBOL(blk_mq_start_request);
static void __blk_mq_requeue_request(struct request *rq)
{
struct request_queue *q = rq->q;
trace_block_rq_requeue(q, rq);
- clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
- if (q->dma_drain_size && blk_rq_bytes(rq))
- rq->nr_phys_segments--;
+ if (test_and_clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags)) {
+ if (q->dma_drain_size && blk_rq_bytes(rq))
+ rq->nr_phys_segments--;
+ }
}
void blk_mq_requeue_request(struct request *rq)
@@ -743,8 +745,6 @@
rq = list_first_entry(&rq_list, struct request, queuelist);
list_del_init(&rq->queuelist);
- blk_mq_start_request(rq);
-
ret = q->mq_ops->queue_rq(hctx, rq, list_empty(&rq_list));
switch (ret) {
case BLK_MQ_RQ_QUEUE_OK:
@@ -1186,7 +1186,6 @@
int ret;
blk_mq_bio_to_request(rq, bio);
- blk_mq_start_request(rq);
/*
* For OK queue, we are done. For error, kill it. Any other
diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c
index 0e2084f..4042440 100644
--- a/drivers/block/mtip32xx/mtip32xx.c
+++ b/drivers/block/mtip32xx/mtip32xx.c
@@ -3783,6 +3783,8 @@
if (unlikely(mtip_check_unal_depth(hctx, rq)))
return BLK_MQ_RQ_QUEUE_BUSY;
+ blk_mq_start_request(rq);
+
ret = mtip_submit_request(hctx, rq);
if (likely(!ret))
return BLK_MQ_RQ_QUEUE_OK;
diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index c5b7315..332ce20 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -321,6 +321,8 @@
cmd->rq = rq;
cmd->nq = hctx->driver_data;
+ blk_mq_start_request(rq);
+
null_handle_cmd(cmd);
return BLK_MQ_RQ_QUEUE_OK;
}
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 13756e0..83816bf 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -205,6 +205,8 @@
}
}
+ blk_mq_start_request(req);
+
num = blk_rq_map_sg(hctx->queue, vbr->req, vbr->sg);
if (num) {
if (rq_data_dir(vbr->req) == WRITE)
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index f1df411..2dcd907 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1890,6 +1890,7 @@
scsi_init_cmd_errh(cmd);
cmd->scsi_done = scsi_mq_done;
+ blk_mq_start_request(req);
reason = scsi_dispatch_cmd(cmd);
if (reason) {
scsi_set_blocked(cmd, reason);
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 9c4e306..878b6f7 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -159,6 +159,7 @@
struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *, const int ctx_index);
struct blk_mq_hw_ctx *blk_mq_alloc_single_hw_queue(struct blk_mq_tag_set *, unsigned int, int);
+void blk_mq_start_request(struct request *rq);
void blk_mq_end_io(struct request *rq, int error);
void __blk_mq_end_io(struct request *rq, int error);