block: unify request timeout handling

Right now SCSI and others do their own command timeout handling.
Move those bits to the block layer.

Instead of having a timer per command, we try to be a bit more clever
and simply have one per-queue. This avoids the overhead of having to
tear down and setup a timer for each command, so it will result in a lot
less timer fiddling.

Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

7 files changed, 226 insertions, 14 deletions

diff --git a/block/Makefile b/block/Makefile
index 0da976ce67d..bfe73049f93 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -4,8 +4,8 @@
 
 obj-$(CONFIG_BLOCK) := elevator.o blk-core.o blk-tag.o blk-sysfs.o \
 			blk-barrier.o blk-settings.o blk-ioc.o blk-map.o \
-			blk-exec.o blk-merge.o blk-softirq.o ioctl.o genhd.o \
-			scsi_ioctl.o cmd-filter.o
+			blk-exec.o blk-merge.o blk-softirq.o blk-timeout.o \
+			ioctl.o genhd.o scsi_ioctl.o cmd-filter.o
 
 obj-$(CONFIG_BLK_DEV_BSG)	+= bsg.o
 obj-$(CONFIG_IOSCHED_NOOP)	+= noop-iosched.o
diff --git a/block/blk-core.c b/block/blk-core.c
index f25eb9786d9..d768a8ddc17 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -110,6 +110,7 @@ void blk_rq_init(struct request_queue *q, struct request *rq)
 	memset(rq, 0, sizeof(*rq));
 
 	INIT_LIST_HEAD(&rq->queuelist);
+	INIT_LIST_HEAD(&rq->timeout_list);
 	rq->cpu = -1;
 	rq->q = q;
 	rq->sector = rq->hard_sector = (sector_t) -1;
@@ -490,6 +491,8 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 	}
 
 	init_timer(&q->unplug_timer);
+	setup_timer(&q->timeout, blk_rq_timed_out_timer, (unsigned long) q);
+	INIT_LIST_HEAD(&q->timeout_list);
 
 	kobject_init(&q->kobj, &blk_queue_ktype);
 
@@ -897,6 +900,8 @@ EXPORT_SYMBOL(blk_start_queueing);
  */
 void blk_requeue_request(struct request_queue *q, struct request *rq)
 {
+	blk_delete_timer(rq);
+	blk_clear_rq_complete(rq);
 	blk_add_trace_rq(q, rq, BLK_TA_REQUEUE);
 
 	if (blk_rq_tagged(rq))
@@ -1650,6 +1655,8 @@ static void end_that_request_last(struct request *req, int error)
 {
 	struct gendisk *disk = req->rq_disk;
 
+	blk_delete_timer(req);
+
 	if (blk_rq_tagged(req))
 		blk_queue_end_tag(req->q, req);
 
diff --git a/block/blk-settings.c b/block/blk-settings.c
index d70692badcd..1d0330d0b40 100644
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -77,6 +77,18 @@ void blk_queue_softirq_done(struct request_queue *q, softirq_done_fn *fn)
 }
 EXPORT_SYMBOL(blk_queue_softirq_done);
 
+void blk_queue_rq_timeout(struct request_queue *q, unsigned int timeout)
+{
+	q->rq_timeout = timeout;
+}
+EXPORT_SYMBOL_GPL(blk_queue_rq_timeout);
+
+void blk_queue_rq_timed_out(struct request_queue *q, rq_timed_out_fn *fn)
+{
+	q->rq_timed_out_fn = fn;
+}
+EXPORT_SYMBOL_GPL(blk_queue_rq_timed_out);
+
 /**
  * blk_queue_make_request - define an alternate make_request function for a device
  * @q:  the request queue for the device to be affected
diff --git a/block/blk-softirq.c b/block/blk-softirq.c
index 3a1af551191..7ab344afb16 100644
--- a/block/blk-softirq.c
+++ b/block/blk-softirq.c
@@ -101,18 +101,7 @@ static struct notifier_block __cpuinitdata blk_cpu_notifier = {
 	.notifier_call	= blk_cpu_notify,
 };
 
-/**
- * blk_complete_request - end I/O on a request
- * @req:      the request being processed
- *
- * Description:
- *     Ends all I/O on a request. It does not handle partial completions,
- *     unless the driver actually implements this in its completion callback
- *     through requeueing. The actual completion happens out-of-order,
- *     through a softirq handler. The user must have registered a completion
- *     callback through blk_queue_softirq_done().
- **/
-void blk_complete_request(struct request *req)
+void __blk_complete_request(struct request *req)
 {
 	struct request_queue *q = req->q;
 	unsigned long flags;
@@ -151,6 +140,23 @@ do_local:
 
 	local_irq_restore(flags);
 }
+
+/**
+ * blk_complete_request - end I/O on a request
+ * @req:      the request being processed
+ *
+ * Description:
+ *     Ends all I/O on a request. It does not handle partial completions,
+ *     unless the driver actually implements this in its completion callback
+ *     through requeueing. The actual completion happens out-of-order,
+ *     through a softirq handler. The user must have registered a completion
+ *     callback through blk_queue_softirq_done().
+ **/
+void blk_complete_request(struct request *req)
+{
+	if (!blk_mark_rq_complete(req))
+		__blk_complete_request(req);
+}
 EXPORT_SYMBOL(blk_complete_request);
 
 __init int blk_softirq_init(void)
diff --git a/block/blk-timeout.c b/block/blk-timeout.c
new file mode 100644
index 00000000000..b36d07bf0af
--- /dev/null
+++ b/block/blk-timeout.c
@@ -0,0 +1,155 @@
+/*
+ * Functions related to generic timeout handling of requests.
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/blkdev.h>
+
+#include "blk.h"
+
+/*
+ * blk_delete_timer - Delete/cancel timer for a given function.
+ * @req:	request that we are canceling timer for
+ *
+ */
+void blk_delete_timer(struct request *req)
+{
+	struct request_queue *q = req->q;
+
+	/*
+	 * Nothing to detach
+	 */
+	if (!q->rq_timed_out_fn || !req->deadline)
+		return;
+
+	list_del_init(&req->timeout_list);
+
+	if (list_empty(&q->timeout_list))
+		del_timer(&q->timeout);
+}
+
+static void blk_rq_timed_out(struct request *req)
+{
+	struct request_queue *q = req->q;
+	enum blk_eh_timer_return ret;
+
+	ret = q->rq_timed_out_fn(req);
+	switch (ret) {
+	case BLK_EH_HANDLED:
+		__blk_complete_request(req);
+		break;
+	case BLK_EH_RESET_TIMER:
+		blk_clear_rq_complete(req);
+		blk_add_timer(req);
+		break;
+	case BLK_EH_NOT_HANDLED:
+		/*
+		 * LLD handles this for now but in the future
+		 * we can send a request msg to abort the command
+		 * and we can move more of the generic scsi eh code to
+		 * the blk layer.
+		 */
+		break;
+	default:
+		printk(KERN_ERR "block: bad eh return: %d\n", ret);
+		break;
+	}
+}
+
+void blk_rq_timed_out_timer(unsigned long data)
+{
+	struct request_queue *q = (struct request_queue *) data;
+	unsigned long flags, uninitialized_var(next), next_set = 0;
+	struct request *rq, *tmp;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+
+	list_for_each_entry_safe(rq, tmp, &q->timeout_list, timeout_list) {
+		if (time_after_eq(jiffies, rq->deadline)) {
+			list_del_init(&rq->timeout_list);
+
+			/*
+			 * Check if we raced with end io completion
+			 */
+			if (blk_mark_rq_complete(rq))
+				continue;
+			blk_rq_timed_out(rq);
+		}
+		if (!next_set) {
+			next = rq->deadline;
+			next_set = 1;
+		} else if (time_after(next, rq->deadline))
+			next = rq->deadline;
+	}
+
+	if (next_set && !list_empty(&q->timeout_list))
+		mod_timer(&q->timeout, round_jiffies(next));
+
+	spin_unlock_irqrestore(q->queue_lock, flags);
+}
+
+/**
+ * blk_abort_request -- Request request recovery for the specified command
+ * @req:	pointer to the request of interest
+ *
+ * This function requests that the block layer start recovery for the
+ * request by deleting the timer and calling the q's timeout function.
+ * LLDDs who implement their own error recovery MAY ignore the timeout
+ * event if they generated blk_abort_req. Must hold queue lock.
+ */
+void blk_abort_request(struct request *req)
+{
+	blk_delete_timer(req);
+	blk_rq_timed_out(req);
+}
+EXPORT_SYMBOL_GPL(blk_abort_request);
+
+/**
+ * blk_add_timer - Start timeout timer for a single request
+ * @req:	request that is about to start running.
+ *
+ * Notes:
+ *    Each request has its own timer, and as it is added to the queue, we
+ *    set up the timer. When the request completes, we cancel the timer.
+ */
+void blk_add_timer(struct request *req)
+{
+	struct request_queue *q = req->q;
+	unsigned long expiry;
+
+	if (!q->rq_timed_out_fn)
+		return;
+
+	BUG_ON(!list_empty(&req->timeout_list));
+	BUG_ON(test_bit(REQ_ATOM_COMPLETE, &req->atomic_flags));
+
+	if (req->timeout)
+		req->deadline = jiffies + req->timeout;
+	else {
+		req->deadline = jiffies + q->rq_timeout;
+		/*
+		 * Some LLDs, like scsi, peek at the timeout to prevent
+		 * a command from being retried forever.
+		 */
+		req->timeout = q->rq_timeout;
+	}
+	list_add_tail(&req->timeout_list, &q->timeout_list);
+
+	/*
+	 * If the timer isn't already pending or this timeout is earlier
+	 * than an existing one, modify the timer. Round to next nearest
+	 * second.
+	 */
+	expiry = round_jiffies(req->deadline);
+
+	/*
+	 * We use ->deadline == 0 to detect whether a timer was added or
+	 * not, so just increase to next jiffy for that specific case
+	 */
+	if (unlikely(!req->deadline))
+		req->deadline = 1;
+
+	if (!timer_pending(&q->timeout) ||
+	    time_before(expiry, q->timeout.expires))
+		mod_timer(&q->timeout, expiry);
+}
diff --git a/block/blk.h b/block/blk.h
index de74254cb91..a4f4a50aefa 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -17,6 +17,30 @@ void __blk_queue_free_tags(struct request_queue *q);
 
 void blk_unplug_work(struct work_struct *work);
 void blk_unplug_timeout(unsigned long data);
+void blk_rq_timed_out_timer(unsigned long data);
+void blk_delete_timer(struct request *);
+void blk_add_timer(struct request *);
+
+/*
+ * Internal atomic flags for request handling
+ */
+enum rq_atomic_flags {
+	REQ_ATOM_COMPLETE = 0,
+};
+
+/*
+ * EH timer and IO completion will both attempt to 'grab' the request, make
+ * sure that only one of them suceeds
+ */
+static inline int blk_mark_rq_complete(struct request *rq)
+{
+	return test_and_set_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
+}
+
+static inline void blk_clear_rq_complete(struct request *rq)
+{
+	clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
+}
 
 struct io_context *current_io_context(gfp_t gfp_flags, int node);
 
diff --git a/block/elevator.c b/block/elevator.c
index 8e3fc3afc77..a91fc59edd0 100644
--- a/block/elevator.c
+++ b/block/elevator.c
@@ -36,6 +36,8 @@
 #include <linux/hash.h>
 #include <linux/uaccess.h>
 
+#include "blk.h"
+
 static DEFINE_SPINLOCK(elv_list_lock);
 static LIST_HEAD(elv_list);
 
@@ -771,6 +773,12 @@ struct request *elv_next_request(struct request_queue *q)
 			 */
 			rq->cmd_flags |= REQ_STARTED;
 			blk_add_trace_rq(q, rq, BLK_TA_ISSUE);
+
+			/*
+			 * We are now handing the request to the hardware,
+			 * add the timeout handler
+			 */
+			blk_add_timer(rq);
 		}
 
 		if (!q->boundary_rq || q->boundary_rq == rq) {