blk-wbt: add general throttling mechanism

We can hook this up to the block layer, to help throttle buffered
writes.

wbt registers a few trace points that can be used to track what is
happening in the system:

wbt_lat: 259:0: latency 2446318
wbt_stat: 259:0: rmean=2446318, rmin=2446318, rmax=2446318, rsamples=1,
               wmean=518866, wmin=15522, wmax=5330353, wsamples=57
wbt_step: 259:0: step down: step=1, window=72727272, background=8, normal=16, max=32

This shows a sync issue event (wbt_lat) that exceeded it's time. wbt_stat
dumps the current read/write stats for that window, and wbt_step shows a
step down event where we now scale back writes. Each trace includes the
device, 259:0 in this case.

Signed-off-by: Jens Axboe <axboe@fb.com>

1 files changed, 735 insertions, 0 deletions

diff --git a/block/blk-wbt.c b/block/blk-wbt.c
new file mode 100644
index 000000000000..889c17ff8503
--- /dev/null
+++ b/block/blk-wbt.c
@@ -0,0 +1,735 @@
+/*
+ * buffered writeback throttling. loosely based on CoDel. We can't drop
+ * packets for IO scheduling, so the logic is something like this:
+ *
+ * - Monitor latencies in a defined window of time.
+ * - If the minimum latency in the above window exceeds some target, increment
+ *   scaling step and scale down queue depth by a factor of 2x. The monitoring
+ *   window is then shrunk to 100 / sqrt(scaling step + 1).
+ * - For any window where we don't have solid data on what the latencies
+ *   look like, retain status quo.
+ * - If latencies look good, decrement scaling step.
+ * - If we're only doing writes, allow the scaling step to go negative. This
+ *   will temporarily boost write performance, snapping back to a stable
+ *   scaling step of 0 if reads show up or the heavy writers finish. Unlike
+ *   positive scaling steps where we shrink the monitoring window, a negative
+ *   scaling step retains the default step==0 window size.
+ *
+ * Copyright (C) 2016 Jens Axboe
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/blk_types.h>
+#include <linux/slab.h>
+#include <linux/backing-dev.h>
+#include <linux/swap.h>
+
+#include "blk-wbt.h"
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/wbt.h>
+
+enum {
+        /*
+         * Default setting, we'll scale up (to 75% of QD max) or down (min 1)
+         * from here depending on device stats
+         */
+        RWB_DEF_DEPTH   = 16,
+
+        /*
+         * 100msec window
+         */
+        RWB_WINDOW_NSEC         = 100 * 1000 * 1000ULL,
+
+        /*
+         * Disregard stats, if we don't meet this minimum
+         */
+        RWB_MIN_WRITE_SAMPLES   = 3,
+
+        /*
+         * If we have this number of consecutive windows with not enough
+         * information to scale up or down, scale up.
+         */
+        RWB_UNKNOWN_BUMP        = 5,
+};
+
+static inline bool rwb_enabled(struct rq_wb *rwb)
+{
+        return rwb && rwb->wb_normal != 0;
+}
+
+/*
+ * Increment 'v', if 'v' is below 'below'. Returns true if we succeeded,
+ * false if 'v' + 1 would be bigger than 'below'.
+ */
+static bool atomic_inc_below(atomic_t *v, int below)
+{
+        int cur = atomic_read(v);
+
+        for (;;) {
+                int old;
+
+                if (cur >= below)
+                        return false;
+                old = atomic_cmpxchg(v, cur, cur + 1);
+                if (old == cur)
+                        break;
+                cur = old;
+        }
+
+        return true;
+}
+
+static void wb_timestamp(struct rq_wb *rwb, unsigned long *var)
+{
+        if (rwb_enabled(rwb)) {
+                const unsigned long cur = jiffies;
+
+                if (cur != *var)
+                        *var = cur;
+        }
+}
+
+/*
+ * If a task was rate throttled in balance_dirty_pages() within the last
+ * second or so, use that to indicate a higher cleaning rate.
+ */
+static bool wb_recent_wait(struct rq_wb *rwb)
+{
+        struct bdi_writeback *wb = &rwb->bdi->wb;
+
+        return time_before(jiffies, wb->dirty_sleep + HZ);
+}
+
+static inline struct rq_wait *get_rq_wait(struct rq_wb *rwb, bool is_kswapd)
+{
+        return &rwb->rq_wait[is_kswapd];
+}
+
+static void rwb_wake_all(struct rq_wb *rwb)
+{
+        int i;
+
+        for (i = 0; i < WBT_NUM_RWQ; i++) {
+                struct rq_wait *rqw = &rwb->rq_wait[i];
+
+                if (waitqueue_active(&rqw->wait))
+                        wake_up_all(&rqw->wait);
+        }
+}
+
+void __wbt_done(struct rq_wb *rwb, enum wbt_flags wb_acct)
+{
+        struct rq_wait *rqw;
+        int inflight, limit;
+
+        if (!(wb_acct & WBT_TRACKED))
+                return;
+
+        rqw = get_rq_wait(rwb, wb_acct & WBT_KSWAPD);
+        inflight = atomic_dec_return(&rqw->inflight);
+
+        /*
+         * wbt got disabled with IO in flight. Wake up any potential
+         * waiters, we don't have to do more than that.
+         */
+        if (unlikely(!rwb_enabled(rwb))) {
+                rwb_wake_all(rwb);
+                return;
+        }
+
+        /*
+         * If the device does write back caching, drop further down
+         * before we wake people up.
+         */
+        if (rwb->wc && !wb_recent_wait(rwb))
+                limit = 0;
+        else
+                limit = rwb->wb_normal;
+
+        /*
+         * Don't wake anyone up if we are above the normal limit.
+         */
+        if (inflight && inflight >= limit)
+                return;
+
+        if (waitqueue_active(&rqw->wait)) {
+                int diff = limit - inflight;
+
+                if (!inflight || diff >= rwb->wb_background / 2)
+                        wake_up_all(&rqw->wait);
+        }
+}
+
+/*
+ * Called on completion of a request. Note that it's also called when
+ * a request is merged, when the request gets freed.
+ */
+void wbt_done(struct rq_wb *rwb, struct blk_issue_stat *stat)
+{
+        if (!rwb)
+                return;
+
+        if (!wbt_is_tracked(stat)) {
+                if (rwb->sync_cookie == stat) {
+                        rwb->sync_issue = 0;
+                        rwb->sync_cookie = NULL;
+                }
+
+                if (wbt_is_read(stat))
+                        wb_timestamp(rwb, &rwb->last_comp);
+                wbt_clear_state(stat);
+        } else {
+                WARN_ON_ONCE(stat == rwb->sync_cookie);
+                __wbt_done(rwb, wbt_stat_to_mask(stat));
+                wbt_clear_state(stat);
+        }
+}
+
+/*
+ * Return true, if we can't increase the depth further by scaling
+ */
+static bool calc_wb_limits(struct rq_wb *rwb)
+{
+        unsigned int depth;
+        bool ret = false;
+
+        if (!rwb->min_lat_nsec) {
+                rwb->wb_max = rwb->wb_normal = rwb->wb_background = 0;
+                return false;
+        }
+
+        /*
+         * For QD=1 devices, this is a special case. It's important for those
+         * to have one request ready when one completes, so force a depth of
+         * 2 for those devices. On the backend, it'll be a depth of 1 anyway,
+         * since the device can't have more than that in flight. If we're
+         * scaling down, then keep a setting of 1/1/1.
+         */
+        if (rwb->queue_depth == 1) {
+                if (rwb->scale_step > 0)
+                        rwb->wb_max = rwb->wb_normal = 1;
+                else {
+                        rwb->wb_max = rwb->wb_normal = 2;
+                        ret = true;
+                }
+                rwb->wb_background = 1;
+        } else {
+                /*
+                 * scale_step == 0 is our default state. If we have suffered
+                 * latency spikes, step will be > 0, and we shrink the
+                 * allowed write depths. If step is < 0, we're only doing
+                 * writes, and we allow a temporarily higher depth to
+                 * increase performance.
+                 */
+                depth = min_t(unsigned int, RWB_DEF_DEPTH, rwb->queue_depth);
+                if (rwb->scale_step > 0)
+                        depth = 1 + ((depth - 1) >> min(31, rwb->scale_step));
+                else if (rwb->scale_step < 0) {
+                        unsigned int maxd = 3 * rwb->queue_depth / 4;
+
+                        depth = 1 + ((depth - 1) << -rwb->scale_step);
+                        if (depth > maxd) {
+                                depth = maxd;
+                                ret = true;
+                        }
+                }
+
+                /*
+                 * Set our max/normal/bg queue depths based on how far
+                 * we have scaled down (->scale_step).
+                 */
+                rwb->wb_max = depth;
+                rwb->wb_normal = (rwb->wb_max + 1) / 2;
+                rwb->wb_background = (rwb->wb_max + 3) / 4;
+        }
+
+        return ret;
+}
+
+static bool inline stat_sample_valid(struct blk_rq_stat *stat)
+{
+        /*
+         * We need at least one read sample, and a minimum of
+         * RWB_MIN_WRITE_SAMPLES. We require some write samples to know
+         * that it's writes impacting us, and not just some sole read on
+         * a device that is in a lower power state.
+         */
+        return stat[0].nr_samples >= 1 &&
+                stat[1].nr_samples >= RWB_MIN_WRITE_SAMPLES;
+}
+
+static u64 rwb_sync_issue_lat(struct rq_wb *rwb)
+{
+        u64 now, issue = ACCESS_ONCE(rwb->sync_issue);
+
+        if (!issue || !rwb->sync_cookie)
+                return 0;
+
+        now = ktime_to_ns(ktime_get());
+        return now - issue;
+}
+
+enum {
+        LAT_OK = 1,
+        LAT_UNKNOWN,
+        LAT_UNKNOWN_WRITES,
+        LAT_EXCEEDED,
+};
+
+static int __latency_exceeded(struct rq_wb *rwb, struct blk_rq_stat *stat)
+{
+        u64 thislat;
+
+        /*
+         * If our stored sync issue exceeds the window size, or it
+         * exceeds our min target AND we haven't logged any entries,
+         * flag the latency as exceeded. wbt works off completion latencies,
+         * but for a flooded device, a single sync IO can take a long time
+         * to complete after being issued. If this time exceeds our
+         * monitoring window AND we didn't see any other completions in that
+         * window, then count that sync IO as a violation of the latency.
+         */
+        thislat = rwb_sync_issue_lat(rwb);
+        if (thislat > rwb->cur_win_nsec ||
+            (thislat > rwb->min_lat_nsec && !stat[0].nr_samples)) {
+                trace_wbt_lat(rwb->bdi, thislat);
+                return LAT_EXCEEDED;
+        }
+
+        /*
+         * No read/write mix, if stat isn't valid
+         */
+        if (!stat_sample_valid(stat)) {
+                /*
+                 * If we had writes in this stat window and the window is
+                 * current, we're only doing writes. If a task recently
+                 * waited or still has writes in flights, consider us doing
+                 * just writes as well.
+                 */
+                if ((stat[1].nr_samples && rwb->stat_ops->is_current(stat)) ||
+                    wb_recent_wait(rwb) || wbt_inflight(rwb))
+                        return LAT_UNKNOWN_WRITES;
+                return LAT_UNKNOWN;
+        }
+
+        /*
+         * If the 'min' latency exceeds our target, step down.
+         */
+        if (stat[0].min > rwb->min_lat_nsec) {
+                trace_wbt_lat(rwb->bdi, stat[0].min);
+                trace_wbt_stat(rwb->bdi, stat);
+                return LAT_EXCEEDED;
+        }
+
+        if (rwb->scale_step)
+                trace_wbt_stat(rwb->bdi, stat);
+
+        return LAT_OK;
+}
+
+static int latency_exceeded(struct rq_wb *rwb)
+{
+        struct blk_rq_stat stat[2];
+
+        rwb->stat_ops->get(rwb->ops_data, stat);
+        return __latency_exceeded(rwb, stat);
+}
+
+static void rwb_trace_step(struct rq_wb *rwb, const char *msg)
+{
+        trace_wbt_step(rwb->bdi, msg, rwb->scale_step, rwb->cur_win_nsec,
+                        rwb->wb_background, rwb->wb_normal, rwb->wb_max);
+}
+
+static void scale_up(struct rq_wb *rwb)
+{
+        /*
+         * Hit max in previous round, stop here
+         */
+        if (rwb->scaled_max)
+                return;
+
+        rwb->scale_step--;
+        rwb->unknown_cnt = 0;
+        rwb->stat_ops->clear(rwb->ops_data);
+
+        rwb->scaled_max = calc_wb_limits(rwb);
+
+        rwb_wake_all(rwb);
+
+        rwb_trace_step(rwb, "step up");
+}
+
+/*
+ * Scale rwb down. If 'hard_throttle' is set, do it quicker, since we
+ * had a latency violation.
+ */
+static void scale_down(struct rq_wb *rwb, bool hard_throttle)
+{
+        /*
+         * Stop scaling down when we've hit the limit. This also prevents
+         * ->scale_step from going to crazy values, if the device can't
+         * keep up.
+         */
+        if (rwb->wb_max == 1)
+                return;
+
+        if (rwb->scale_step < 0 && hard_throttle)
+                rwb->scale_step = 0;
+        else
+                rwb->scale_step++;
+
+        rwb->scaled_max = false;
+        rwb->unknown_cnt = 0;
+        rwb->stat_ops->clear(rwb->ops_data);
+        calc_wb_limits(rwb);
+        rwb_trace_step(rwb, "step down");
+}
+
+static void rwb_arm_timer(struct rq_wb *rwb)
+{
+        unsigned long expires;
+
+        if (rwb->scale_step > 0) {
+                /*
+                 * We should speed this up, using some variant of a fast
+                 * integer inverse square root calculation. Since we only do
+                 * this for every window expiration, it's not a huge deal,
+                 * though.
+                 */
+                rwb->cur_win_nsec = div_u64(rwb->win_nsec << 4,
+                                        int_sqrt((rwb->scale_step + 1) << 8));
+        } else {
+                /*
+                 * For step < 0, we don't want to increase/decrease the
+                 * window size.
+                 */
+                rwb->cur_win_nsec = rwb->win_nsec;
+        }
+
+        expires = jiffies + nsecs_to_jiffies(rwb->cur_win_nsec);
+        mod_timer(&rwb->window_timer, expires);
+}
+
+static void wb_timer_fn(unsigned long data)
+{
+        struct rq_wb *rwb = (struct rq_wb *) data;
+        unsigned int inflight = wbt_inflight(rwb);
+        int status;
+
+        status = latency_exceeded(rwb);
+
+        trace_wbt_timer(rwb->bdi, status, rwb->scale_step, inflight);
+
+        /*
+         * If we exceeded the latency target, step down. If we did not,
+         * step one level up. If we don't know enough to say either exceeded
+         * or ok, then don't do anything.
+         */
+        switch (status) {
+        case LAT_EXCEEDED:
+                scale_down(rwb, true);
+                break;
+        case LAT_OK:
+                scale_up(rwb);
+                break;
+        case LAT_UNKNOWN_WRITES:
+                /*
+                 * We started a the center step, but don't have a valid
+                 * read/write sample, but we do have writes going on.
+                 * Allow step to go negative, to increase write perf.
+                 */
+                scale_up(rwb);
+                break;
+        case LAT_UNKNOWN:
+                if (++rwb->unknown_cnt < RWB_UNKNOWN_BUMP)
+                        break;
+                /*
+                 * We get here when previously scaled reduced depth, and we
+                 * currently don't have a valid read/write sample. For that
+                 * case, slowly return to center state (step == 0).
+                 */
+                if (rwb->scale_step > 0)
+                        scale_up(rwb);
+                else if (rwb->scale_step < 0)
+                        scale_down(rwb, false);
+                break;
+        default:
+                break;
+        }
+
+        /*
+         * Re-arm timer, if we have IO in flight
+         */
+        if (rwb->scale_step || inflight)
+                rwb_arm_timer(rwb);
+}
+
+void wbt_update_limits(struct rq_wb *rwb)
+{
+        rwb->scale_step = 0;
+        rwb->scaled_max = false;
+        calc_wb_limits(rwb);
+
+        rwb_wake_all(rwb);
+}
+
+static bool close_io(struct rq_wb *rwb)
+{
+        const unsigned long now = jiffies;
+
+        return time_before(now, rwb->last_issue + HZ / 10) ||
+                time_before(now, rwb->last_comp + HZ / 10);
+}
+
+#define REQ_HIPRIO      (REQ_SYNC | REQ_META | REQ_PRIO)
+
+static inline unsigned int get_limit(struct rq_wb *rwb, unsigned long rw)
+{
+        unsigned int limit;
+
+        /*
+         * At this point we know it's a buffered write. If this is
+         * kswapd trying to free memory, or REQ_SYNC is set, set, then
+         * it's WB_SYNC_ALL writeback, and we'll use the max limit for
+         * that. If the write is marked as a background write, then use
+         * the idle limit, or go to normal if we haven't had competing
+         * IO for a bit.
+         */
+        if ((rw & REQ_HIPRIO) || wb_recent_wait(rwb) || current_is_kswapd())
+                limit = rwb->wb_max;
+        else if ((rw & REQ_BACKGROUND) || close_io(rwb)) {
+                /*
+                 * If less than 100ms since we completed unrelated IO,
+                 * limit us to half the depth for background writeback.
+                 */
+                limit = rwb->wb_background;
+        } else
+                limit = rwb->wb_normal;
+
+        return limit;
+}
+
+static inline bool may_queue(struct rq_wb *rwb, struct rq_wait *rqw,
+                             wait_queue_t *wait, unsigned long rw)
+{
+        /*
+         * inc it here even if disabled, since we'll dec it at completion.
+         * this only happens if the task was sleeping in __wbt_wait(),
+         * and someone turned it off at the same time.
+         */
+        if (!rwb_enabled(rwb)) {
+                atomic_inc(&rqw->inflight);
+                return true;
+        }
+
+        /*
+         * If the waitqueue is already active and we are not the next
+         * in line to be woken up, wait for our turn.
+         */
+        if (waitqueue_active(&rqw->wait) &&
+            rqw->wait.task_list.next != &wait->task_list)
+                return false;
+
+        return atomic_inc_below(&rqw->inflight, get_limit(rwb, rw));
+}
+
+/*
+ * Block if we will exceed our limit, or if we are currently waiting for
+ * the timer to kick off queuing again.
+ */
+static void __wbt_wait(struct rq_wb *rwb, unsigned long rw, spinlock_t *lock)
+{
+        struct rq_wait *rqw = get_rq_wait(rwb, current_is_kswapd());
+        DEFINE_WAIT(wait);
+
+        if (may_queue(rwb, rqw, &wait, rw))
+                return;
+
+        do {
+                prepare_to_wait_exclusive(&rqw->wait, &wait,
+                                                TASK_UNINTERRUPTIBLE);
+
+                if (may_queue(rwb, rqw, &wait, rw))
+                        break;
+
+                if (lock)
+                        spin_unlock_irq(lock);
+
+                io_schedule();
+
+                if (lock)
+                        spin_lock_irq(lock);
+        } while (1);
+
+        finish_wait(&rqw->wait, &wait);
+}
+
+static inline bool wbt_should_throttle(struct rq_wb *rwb, struct bio *bio)
+{
+        const int op = bio_op(bio);
+
+        /*
+         * If not a WRITE (or a discard), do nothing
+         */
+        if (!(op == REQ_OP_WRITE || op == REQ_OP_DISCARD))
+                return false;
+
+        /*
+         * Don't throttle WRITE_ODIRECT
+         */
+        if ((bio->bi_opf & (REQ_SYNC | REQ_IDLE)) == (REQ_SYNC | REQ_IDLE))
+                return false;
+
+        return true;
+}
+
+/*
+ * Returns true if the IO request should be accounted, false if not.
+ * May sleep, if we have exceeded the writeback limits. Caller can pass
+ * in an irq held spinlock, if it holds one when calling this function.
+ * If we do sleep, we'll release and re-grab it.
+ */
+unsigned int wbt_wait(struct rq_wb *rwb, struct bio *bio, spinlock_t *lock)
+{
+        unsigned int ret = 0;
+
+        if (!rwb_enabled(rwb))
+                return 0;
+
+        if (bio_op(bio) == REQ_OP_READ)
+                ret = WBT_READ;
+
+        if (!wbt_should_throttle(rwb, bio)) {
+                if (ret & WBT_READ)
+                        wb_timestamp(rwb, &rwb->last_issue);
+                return ret;
+        }
+
+        __wbt_wait(rwb, bio->bi_opf, lock);
+
+        if (!timer_pending(&rwb->window_timer))
+                rwb_arm_timer(rwb);
+
+        if (current_is_kswapd())
+                ret |= WBT_KSWAPD;
+
+        return ret | WBT_TRACKED;
+}
+
+void wbt_issue(struct rq_wb *rwb, struct blk_issue_stat *stat)
+{
+        if (!rwb_enabled(rwb))
+                return;
+
+        /*
+         * Track sync issue, in case it takes a long time to complete. Allows
+         * us to react quicker, if a sync IO takes a long time to complete.
+         * Note that this is just a hint. 'stat' can go away when the
+         * request completes, so it's important we never dereference it. We
+         * only use the address to compare with, which is why we store the
+         * sync_issue time locally.
+         */
+        if (wbt_is_read(stat) && !rwb->sync_issue) {
+                rwb->sync_cookie = stat;
+                rwb->sync_issue = blk_stat_time(stat);
+        }
+}
+
+void wbt_requeue(struct rq_wb *rwb, struct blk_issue_stat *stat)
+{
+        if (!rwb_enabled(rwb))
+                return;
+        if (stat == rwb->sync_cookie) {
+                rwb->sync_issue = 0;
+                rwb->sync_cookie = NULL;
+        }
+}
+
+void wbt_set_queue_depth(struct rq_wb *rwb, unsigned int depth)
+{
+        if (rwb) {
+                rwb->queue_depth = depth;
+                wbt_update_limits(rwb);
+        }
+}
+
+void wbt_set_write_cache(struct rq_wb *rwb, bool write_cache_on)
+{
+        if (rwb)
+                rwb->wc = write_cache_on;
+}
+
+void wbt_disable(struct rq_wb *rwb)
+{
+        if (rwb) {
+                del_timer_sync(&rwb->window_timer);
+                rwb->win_nsec = rwb->min_lat_nsec = 0;
+                wbt_update_limits(rwb);
+        }
+}
+EXPORT_SYMBOL_GPL(wbt_disable);
+
+int wbt_init(struct request_queue *q, struct wb_stat_ops *ops)
+{
+        struct rq_wb *rwb;
+        int i;
+
+        /*
+         * For now, we depend on the stats window being larger than
+         * our monitoring window. Ensure that this isn't inadvertently
+         * violated.
+         */
+        BUILD_BUG_ON(RWB_WINDOW_NSEC > BLK_STAT_NSEC);
+        BUILD_BUG_ON(WBT_NR_BITS > BLK_STAT_RES_BITS);
+
+        if (!ops->get || !ops->is_current || !ops->clear)
+                return -EINVAL;
+
+        rwb = kzalloc(sizeof(*rwb), GFP_KERNEL);
+        if (!rwb)
+                return -ENOMEM;
+
+        for (i = 0; i < WBT_NUM_RWQ; i++) {
+                atomic_set(&rwb->rq_wait[i].inflight, 0);
+                init_waitqueue_head(&rwb->rq_wait[i].wait);
+        }
+
+        setup_timer(&rwb->window_timer, wb_timer_fn, (unsigned long) rwb);
+        rwb->wc = 1;
+        rwb->queue_depth = RWB_DEF_DEPTH;
+        rwb->last_comp = rwb->last_issue = jiffies;
+        rwb->bdi = &q->backing_dev_info;
+        rwb->win_nsec = RWB_WINDOW_NSEC;
+        rwb->stat_ops = ops;
+        rwb->ops_data = q;
+        wbt_update_limits(rwb);
+
+        /*
+         * Assign rwb, and turn on stats tracking for this queue
+         */
+        q->rq_wb = rwb;
+        blk_stat_enable(q);
+
+        if (blk_queue_nonrot(q))
+                rwb->min_lat_nsec = 2000000ULL;
+        else
+                rwb->min_lat_nsec = 75000000ULL;
+
+        wbt_set_queue_depth(rwb, blk_queue_depth(q));
+        wbt_set_write_cache(rwb, test_bit(QUEUE_FLAG_WC, &q->queue_flags));
+
+        return 0;
+}
+
+void wbt_exit(struct request_queue *q)
+{
+        struct rq_wb *rwb = q->rq_wb;
+
+        if (rwb) {
+                del_timer_sync(&rwb->window_timer);
+                q->rq_wb = NULL;
+                kfree(rwb);
+        }
+}