block: get rid of struct blk_issue_stat

struct blk_issue_stat squashes three things into one u64:

- The time the driver started working on a request
- The original size of the request (for the io.low controller)
- Flags for writeback throttling

It turns out that on x86_64, we have a 4 byte hole in struct request
which we can fill with the non-timestamp fields from blk_issue_stat,
simplifying things quite a bit.

Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/block/blk-core.c b/block/blk-core.c
index fe2f457..33d5c7d 100644 (file)
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -2991,7 +2991,10 @@ void blk_start_request(struct request *req)
        blk_dequeue_request(req);
 
        if (test_bit(QUEUE_FLAG_STATS, &req->q->queue_flags)) {
-               blk_stat_set_issue(&req->issue_stat, blk_rq_sectors(req));
+               req->io_start_time_ns = ktime_get_ns();
+#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
+               req->throtl_size = blk_rq_sectors(req);
+#endif
                req->rq_flags |= RQF_STATS;
                wbt_issue(req->q->rq_wb, req);
        }

diff --git a/block/blk-mq.c b/block/blk-mq.c
index deb85b5..17612e0 100644 (file)
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -310,6 +310,7 @@ static struct request *blk_mq_rq_ctx_init(struct blk_mq_alloc_data *data,
        rq->rq_disk = NULL;
        rq->part = NULL;
        rq->start_time = jiffies;
+       rq->io_start_time_ns = 0;
        rq->nr_phys_segments = 0;
 #if defined(CONFIG_BLK_DEV_INTEGRITY)
        rq->nr_integrity_segments = 0;
@@ -329,7 +330,7 @@ static struct request *blk_mq_rq_ctx_init(struct blk_mq_alloc_data *data,
 #ifdef CONFIG_BLK_CGROUP
        rq->rl = NULL;
        set_start_time_ns(rq);
-       rq->io_start_time_ns = 0;
+       rq->cgroup_io_start_time_ns = 0;
 #endif
 
        data->ctx->rq_dispatched[op_is_sync(op)]++;
@@ -669,7 +670,10 @@ void blk_mq_start_request(struct request *rq)
        trace_block_rq_issue(q, rq);
 
        if (test_bit(QUEUE_FLAG_STATS, &q->queue_flags)) {
-               blk_stat_set_issue(&rq->issue_stat, blk_rq_sectors(rq));
+               rq->io_start_time_ns = ktime_get_ns();
+#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
+               rq->throtl_size = blk_rq_sectors(rq);
+#endif
                rq->rq_flags |= RQF_STATS;
                wbt_issue(q->rq_wb, rq);
        }

diff --git a/block/blk-stat.c b/block/blk-stat.c
index bd365a9..725a881 100644 (file)
--- a/block/blk-stat.c
+++ b/block/blk-stat.c
@@ -55,11 +55,8 @@ void blk_stat_add(struct request *rq)
        int bucket;
        u64 now, value;
 
-       now = __blk_stat_time(ktime_to_ns(ktime_get()));
-       if (now < blk_stat_time(&rq->issue_stat))
-               return;
-
-       value = now - blk_stat_time(&rq->issue_stat);
+       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 c22049a..17c812d 100644 (file)
--- a/block/blk-stat.h
+++ b/block/blk-stat.h
@@ -8,21 +8,6 @@
 #include <linux/rcupdate.h>
 #include <linux/timer.h>
 
-/*
- * from upper:
- * 4 bits: reserved for other usage
- * 12 bits: size
- * 48 bits: time
- */
-#define BLK_STAT_RES_BITS      4
-#define BLK_STAT_SIZE_BITS     12
-#define BLK_STAT_RES_SHIFT     (64 - BLK_STAT_RES_BITS)
-#define BLK_STAT_SIZE_SHIFT    (BLK_STAT_RES_SHIFT - BLK_STAT_SIZE_BITS)
-#define BLK_STAT_TIME_MASK     ((1ULL << BLK_STAT_SIZE_SHIFT) - 1)
-#define BLK_STAT_SIZE_MASK     \
-       (((1ULL << BLK_STAT_SIZE_BITS) - 1) << BLK_STAT_SIZE_SHIFT)
-#define BLK_STAT_RES_MASK      (~((1ULL << BLK_STAT_RES_SHIFT) - 1))
-
 /**
  * struct blk_stat_callback - Block statistics callback.
  *
@@ -82,34 +67,6 @@ void blk_free_queue_stats(struct blk_queue_stats *);
 
 void blk_stat_add(struct request *);
 
-static inline u64 __blk_stat_time(u64 time)
-{
-       return time & BLK_STAT_TIME_MASK;
-}
-
-static inline u64 blk_stat_time(struct blk_issue_stat *stat)
-{
-       return __blk_stat_time(stat->stat);
-}
-
-static inline sector_t blk_capped_size(sector_t size)
-{
-       return size & ((1ULL << BLK_STAT_SIZE_BITS) - 1);
-}
-
-static inline sector_t blk_stat_size(struct blk_issue_stat *stat)
-{
-       return (stat->stat & BLK_STAT_SIZE_MASK) >> BLK_STAT_SIZE_SHIFT;
-}
-
-static inline void blk_stat_set_issue(struct blk_issue_stat *stat,
-       sector_t size)
-{
-       stat->stat = (stat->stat & BLK_STAT_RES_MASK) |
-               (ktime_to_ns(ktime_get()) & BLK_STAT_TIME_MASK) |
-               (((u64)blk_capped_size(size)) << BLK_STAT_SIZE_SHIFT);
-}
-
 /* 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-throttle.c b/block/blk-throttle.c
index 35f9b8f..f63d88c 100644 (file)
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -2279,8 +2279,7 @@ void blk_throtl_stat_add(struct request *rq, u64 time_ns)
        struct request_queue *q = rq->q;
        struct throtl_data *td = q->td;
 
-       throtl_track_latency(td, blk_stat_size(&rq->issue_stat),
-               req_op(rq), time_ns >> 10);
+       throtl_track_latency(td, rq->throtl_size, req_op(rq), time_ns >> 10);
 }
 
 void blk_throtl_bio_endio(struct bio *bio)

diff --git a/block/blk-wbt.c b/block/blk-wbt.c
index 592e914..4f89b28 100644 (file)
--- a/block/blk-wbt.c
+++ b/block/blk-wbt.c
@@ -31,22 +31,22 @@
 
 static inline void wbt_clear_state(struct request *rq)
 {
-       rq->issue_stat.stat &= ~BLK_STAT_RES_MASK;
+       rq->wbt_flags = 0;
 }
 
 static inline enum wbt_flags wbt_flags(struct request *rq)
 {
-       return (rq->issue_stat.stat & BLK_STAT_RES_MASK) >> BLK_STAT_RES_SHIFT;
+       return rq->wbt_flags;
 }
 
 static inline bool wbt_is_tracked(struct request *rq)
 {
-       return (rq->issue_stat.stat >> BLK_STAT_RES_SHIFT) & WBT_TRACKED;
+       return rq->wbt_flags & WBT_TRACKED;
 }
 
 static inline bool wbt_is_read(struct request *rq)
 {
-       return (rq->issue_stat.stat >> BLK_STAT_RES_SHIFT) & WBT_READ;
+       return rq->wbt_flags & WBT_READ;
 }
 
 enum {
@@ -657,7 +657,7 @@ void wbt_issue(struct rq_wb *rwb, struct request *rq)
         */
        if (wbt_is_read(rq) && !rwb->sync_issue) {
                rwb->sync_cookie = rq;
-               rwb->sync_issue = blk_stat_time(&rq->issue_stat);
+               rwb->sync_issue = rq->io_start_time_ns;
        }
 }
 
@@ -746,8 +746,6 @@ int wbt_init(struct request_queue *q)
        struct rq_wb *rwb;
        int i;
 
-       BUILD_BUG_ON(WBT_NR_BITS > BLK_STAT_RES_BITS);
-
        rwb = kzalloc(sizeof(*rwb), GFP_KERNEL);
        if (!rwb)
                return -ENOMEM;

diff --git a/block/blk-wbt.h b/block/blk-wbt.h
index 85fbccc..300df53 100644 (file)
--- a/block/blk-wbt.h
+++ b/block/blk-wbt.h
@@ -63,7 +63,7 @@ struct rq_wb {
 
        struct blk_stat_callback *cb;
 
-       s64 sync_issue;
+       u64 sync_issue;
        void *sync_cookie;
 
        unsigned int wc;
@@ -90,7 +90,7 @@ static inline unsigned int wbt_inflight(struct rq_wb *rwb)
 
 static inline void wbt_track(struct request *rq, enum wbt_flags flags)
 {
-       rq->issue_stat.stat |= ((u64)flags) << BLK_STAT_RES_SHIFT;
+       rq->wbt_flags |= flags;
 }
 
 void __wbt_done(struct rq_wb *, enum wbt_flags);

diff --git a/block/kyber-iosched.c b/block/kyber-iosched.c
index 0d6d25e..564967f 100644 (file)
--- a/block/kyber-iosched.c
+++ b/block/kyber-iosched.c
@@ -485,11 +485,11 @@ static void kyber_completed_request(struct request *rq)
        if (blk_stat_is_active(kqd->cb))
                return;
 
-       now = __blk_stat_time(ktime_to_ns(ktime_get()));
-       if (now < blk_stat_time(&rq->issue_stat))
+       now = ktime_get_ns();
+       if (now < rq->io_start_time_ns)
                return;
 
-       latency = now - blk_stat_time(&rq->issue_stat);
+       latency = now - rq->io_start_time_ns;
 
        if (latency > target)
                blk_stat_activate_msecs(kqd->cb, 10);

diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index b6f1d53..4cb970c 100644 (file)
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -91,10 +91,6 @@ static inline bool blk_path_error(blk_status_t error)
        return true;
 }
 
-struct blk_issue_stat {
-       u64 stat;
-};
-
 /*
  * From most significant bit:
  * 1 bit: reserved for other usage, see below

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 5c4eee0..f2c2fc0 100644 (file)
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -206,8 +206,18 @@ struct request {
        struct gendisk *rq_disk;
        struct hd_struct *part;
        unsigned long start_time;
-       struct blk_issue_stat issue_stat;
-       /* Number of scatter-gather DMA addr+len pairs after
+       /* Time that I/O was submitted to the device. */
+       u64 io_start_time_ns;
+
+#ifdef CONFIG_BLK_WBT
+       unsigned short wbt_flags;
+#endif
+#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
+       unsigned short throtl_size;
+#endif
+
+       /*
+        * Number of scatter-gather DMA addr+len pairs after
         * physical address coalescing is performed.
         */
        unsigned short nr_phys_segments;
@@ -267,8 +277,8 @@ struct request {
 
 #ifdef CONFIG_BLK_CGROUP
        struct request_list *rl;                /* rl this rq is alloced from */
-       unsigned long long start_time_ns;
-       unsigned long long io_start_time_ns;    /* when passed to hardware */
+       unsigned long long cgroup_start_time_ns;
+       unsigned long long cgroup_io_start_time_ns;    /* when passed to hardware */
 #endif
 };
 
@@ -1797,25 +1807,25 @@ int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned lo
 static inline void set_start_time_ns(struct request *req)
 {
        preempt_disable();
-       req->start_time_ns = sched_clock();
+       req->cgroup_start_time_ns = sched_clock();
        preempt_enable();
 }
 
 static inline void set_io_start_time_ns(struct request *req)
 {
        preempt_disable();
-       req->io_start_time_ns = sched_clock();
+       req->cgroup_io_start_time_ns = sched_clock();
        preempt_enable();
 }
 
 static inline uint64_t rq_start_time_ns(struct request *req)
 {
-        return req->start_time_ns;
+       return req->cgroup_start_time_ns;
 }
 
 static inline uint64_t rq_io_start_time_ns(struct request *req)
 {
-        return req->io_start_time_ns;
+       return req->cgroup_io_start_time_ns;
 }
 #else
 static inline void set_start_time_ns(struct request *req) {}