block: consolidate struct request timestamp fields

Currently, struct request has four timestamp fields:

- A start time, set at get_request time, in jiffies, used for iostats
- An I/O start time, set at start_request time, in ktime nanoseconds,
  used for blk-stats (i.e., wbt, kyber, hybrid polling)
- Another start time and another I/O start time, used for cfq and bfq

These can all be consolidated into one start time and one I/O start
time, both in ktime nanoseconds, shaving off up to 16 bytes from struct
request depending on the kernel config.

Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 771ae97..ebc264c 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -4778,8 +4778,8 @@
 
 	if (rq->rq_flags & RQF_STARTED)
 		bfqg_stats_update_completion(bfqq_group(bfqq),
-					     rq_start_time_ns(rq),
-					     rq_io_start_time_ns(rq),
+					     rq->start_time_ns,
+					     rq->io_start_time_ns,
 					     rq->cmd_flags);
 
 	if (likely(rq->rq_flags & RQF_STARTED)) {
diff --git a/block/blk-core.c b/block/blk-core.c
index 33d5c7d..1418a1c 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -196,8 +196,7 @@
 	RB_CLEAR_NODE(&rq->rb_node);
 	rq->tag = -1;
 	rq->internal_tag = -1;
-	rq->start_time = jiffies;
-	set_start_time_ns(rq);
+	rq->start_time_ns = ktime_get_ns();
 	rq->part = NULL;
 	seqcount_init(&rq->gstate_seq);
 	u64_stats_init(&rq->aborted_gstate_sync);
@@ -2726,7 +2725,7 @@
 	}
 }
 
-void blk_account_io_done(struct request *req)
+void blk_account_io_done(struct request *req, u64 now)
 {
 	/*
 	 * Account IO completion.  flush_rq isn't accounted as a
@@ -2734,11 +2733,12 @@
 	 * containing request is enough.
 	 */
 	if (blk_do_io_stat(req) && !(req->rq_flags & RQF_FLUSH_SEQ)) {
-		unsigned long duration = jiffies - req->start_time;
+		unsigned long duration;
 		const int rw = rq_data_dir(req);
 		struct hd_struct *part;
 		int cpu;
 
+		duration = nsecs_to_jiffies(now - req->start_time_ns);
 		cpu = part_stat_lock();
 		part = req->part;
 
@@ -2969,10 +2969,8 @@
 	 * and to it is freed is accounted as io that is in progress at
 	 * the driver side.
 	 */
-	if (blk_account_rq(rq)) {
+	if (blk_account_rq(rq))
 		q->in_flight[rq_is_sync(rq)]++;
-		set_io_start_time_ns(rq);
-	}
 }
 
 /**
@@ -3192,12 +3190,13 @@
 void blk_finish_request(struct request *req, blk_status_t error)
 {
 	struct request_queue *q = req->q;
+	u64 now = ktime_get_ns();
 
 	lockdep_assert_held(req->q->queue_lock);
 	WARN_ON_ONCE(q->mq_ops);
 
 	if (req->rq_flags & RQF_STATS)
-		blk_stat_add(req);
+		blk_stat_add(req, now);
 
 	if (req->rq_flags & RQF_QUEUED)
 		blk_queue_end_tag(q, req);
@@ -3212,7 +3211,7 @@
 	if (req->rq_flags & RQF_DONTPREP)
 		blk_unprep_request(req);
 
-	blk_account_io_done(req);
+	blk_account_io_done(req, now);
 
 	if (req->end_io) {
 		wbt_done(req->q->rq_wb, req);
diff --git a/block/blk-merge.c b/block/blk-merge.c
index 782940c..5573d0f 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -724,13 +724,12 @@
 	}
 
 	/*
-	 * At this point we have either done a back merge
-	 * or front merge. We need the smaller start_time of
-	 * the merged requests to be the current request
-	 * for accounting purposes.
+	 * At this point we have either done a back merge or front merge. We
+	 * need the smaller start_time_ns of the merged requests to be the
+	 * current request for accounting purposes.
 	 */
-	if (time_after(req->start_time, next->start_time))
-		req->start_time = next->start_time;
+	if (next->start_time_ns < req->start_time_ns)
+		req->start_time_ns = next->start_time_ns;
 
 	req->biotail->bi_next = next->bio;
 	req->biotail = next->biotail;
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 39b4e98..4e9d835 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -309,7 +309,7 @@
 	RB_CLEAR_NODE(&rq->rb_node);
 	rq->rq_disk = NULL;
 	rq->part = NULL;
-	rq->start_time = jiffies;
+	rq->start_time_ns = ktime_get_ns();
 	rq->io_start_time_ns = 0;
 	rq->nr_phys_segments = 0;
 #if defined(CONFIG_BLK_DEV_INTEGRITY)
@@ -329,8 +329,6 @@
 
 #ifdef CONFIG_BLK_CGROUP
 	rq->rl = NULL;
-	set_start_time_ns(rq);
-	rq->cgroup_io_start_time_ns = 0;
 #endif
 
 	data->ctx->rq_dispatched[op_is_sync(op)]++;
@@ -506,12 +504,14 @@
 
 inline void __blk_mq_end_request(struct request *rq, blk_status_t error)
 {
+	u64 now = ktime_get_ns();
+
 	if (rq->rq_flags & RQF_STATS) {
 		blk_mq_poll_stats_start(rq->q);
-		blk_stat_add(rq);
+		blk_stat_add(rq, now);
 	}
 
-	blk_account_io_done(rq);
+	blk_account_io_done(rq, now);
 
 	if (rq->end_io) {
 		wbt_done(rq->q->rq_wb, rq);
diff --git a/block/blk-stat.c b/block/blk-stat.c
index 725a881..175c143 100644
--- a/block/blk-stat.c
+++ b/block/blk-stat.c
@@ -47,15 +47,14 @@
 	stat->nr_samples++;
 }
 
-void blk_stat_add(struct request *rq)
+void blk_stat_add(struct request *rq, u64 now)
 {
 	struct request_queue *q = rq->q;
 	struct blk_stat_callback *cb;
 	struct blk_rq_stat *stat;
 	int bucket;
-	u64 now, value;
+	u64 value;
 
-	now = ktime_get_ns();
 	value = (now >= rq->io_start_time_ns) ? now - rq->io_start_time_ns : 0;
 
 	blk_throtl_stat_add(rq, value);
diff --git a/block/blk-stat.h b/block/blk-stat.h
index 17c812d..78399cd 100644
--- a/block/blk-stat.h
+++ b/block/blk-stat.h
@@ -65,7 +65,7 @@
 struct blk_queue_stats *blk_alloc_queue_stats(void);
 void blk_free_queue_stats(struct blk_queue_stats *);
 
-void blk_stat_add(struct request *);
+void blk_stat_add(struct request *rq, u64 now);
 
 /* record time/size info in request but not add a callback */
 void blk_stat_enable_accounting(struct request_queue *q);
diff --git a/block/blk.h b/block/blk.h
index b034fd2..eaf1a8e 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -186,7 +186,7 @@
 
 void blk_account_io_start(struct request *req, bool new_io);
 void blk_account_io_completion(struct request *req, unsigned int bytes);
-void blk_account_io_done(struct request *req);
+void blk_account_io_done(struct request *req, u64 now);
 
 /*
  * EH timer and IO completion will both attempt to 'grab' the request, make
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 652ca06..6b9f6b1 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -4228,8 +4228,8 @@
 	cfqd->rq_in_driver--;
 	cfqq->dispatched--;
 	(RQ_CFQG(rq))->dispatched--;
-	cfqg_stats_update_completion(cfqq->cfqg, rq_start_time_ns(rq),
-				     rq_io_start_time_ns(rq), rq->cmd_flags);
+	cfqg_stats_update_completion(cfqq->cfqg, rq->start_time_ns,
+				     rq->io_start_time_ns, rq->cmd_flags);
 
 	cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--;
 
@@ -4245,16 +4245,7 @@
 					cfqq_type(cfqq));
 
 		st->ttime.last_end_request = now;
-		/*
-		 * We have to do this check in jiffies since start_time is in
-		 * jiffies and it is not trivial to convert to ns. If
-		 * cfq_fifo_expire[1] ever comes close to 1 jiffie, this test
-		 * will become problematic but so far we are fine (the default
-		 * is 128 ms).
-		 */
-		if (!time_after(rq->start_time +
-				  nsecs_to_jiffies(cfqd->cfq_fifo_expire[1]),
-				jiffies))
+		if (rq->start_time_ns + cfqd->cfq_fifo_expire[1] <= now)
 			cfqd->last_delayed_sync = now;
 	}
 
diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c
index bf0b840..1c18f33 100644
--- a/drivers/md/dm-rq.c
+++ b/drivers/md/dm-rq.c
@@ -406,7 +406,7 @@
 	if (blk_queue_io_stat(clone->q))
 		clone->rq_flags |= RQF_IO_STAT;
 
-	clone->start_time = jiffies;
+	clone->start_time_ns = ktime_get_ns();
 	r = blk_insert_cloned_request(clone->q, clone);
 	if (r != BLK_STS_OK && r != BLK_STS_RESOURCE && r != BLK_STS_DEV_RESOURCE)
 		/* must complete clone in terms of original request */
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 9ef4126..e42d510 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -205,7 +205,8 @@
 
 	struct gendisk *rq_disk;
 	struct hd_struct *part;
-	unsigned long start_time;
+	/* Time that I/O was submitted to the kernel. */
+	u64 start_time_ns;
 	/* Time that I/O was submitted to the device. */
 	u64 io_start_time_ns;
 
@@ -277,8 +278,6 @@
 
 #ifdef CONFIG_BLK_CGROUP
 	struct request_list *rl;		/* rl this rq is alloced from */
-	unsigned long long cgroup_start_time_ns;
-	unsigned long long cgroup_io_start_time_ns;    /* when passed to hardware */
 #endif
 };
 
@@ -1798,39 +1797,6 @@
 int kblockd_schedule_work_on(int cpu, struct work_struct *work);
 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
 
-#ifdef CONFIG_BLK_CGROUP
-static inline void set_start_time_ns(struct request *req)
-{
-	req->cgroup_start_time_ns = ktime_get_ns();
-}
-
-static inline void set_io_start_time_ns(struct request *req)
-{
-	req->cgroup_io_start_time_ns = ktime_get_ns();
-}
-
-static inline u64 rq_start_time_ns(struct request *req)
-{
-	return req->cgroup_start_time_ns;
-}
-
-static inline u64 rq_io_start_time_ns(struct request *req)
-{
-	return req->cgroup_io_start_time_ns;
-}
-#else
-static inline void set_start_time_ns(struct request *req) {}
-static inline void set_io_start_time_ns(struct request *req) {}
-static inline u64 rq_start_time_ns(struct request *req)
-{
-	return 0;
-}
-static inline u64 rq_io_start_time_ns(struct request *req)
-{
-	return 0;
-}
-#endif
-
 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \