Contributors: 22
Author Tokens Token Proportion Commits Commit Proportion
Kent Overstreet 2457 53.36% 24 26.09%
Coly Li 1200 26.06% 27 29.35%
Michael Lyle 258 5.60% 6 6.52%
tang.junhui 244 5.30% 6 6.52%
dongdong tao 177 3.84% 1 1.09%
ZouMingzhe 106 2.30% 5 5.43%
Dongsheng Yang 48 1.04% 1 1.09%
Christoph Hellwig 29 0.63% 7 7.61%
Konstantin Khlebnikov 24 0.52% 1 1.09%
Gustavo A. R. Silva 17 0.37% 2 2.17%
Slava Pestov 10 0.22% 1 1.09%
Matthew Mirvish 7 0.15% 1 1.09%
Michael Christie 6 0.13% 1 1.09%
Lei Ming 5 0.11% 1 1.09%
Guoqing Jiang 4 0.09% 1 1.09%
Ingo Molnar 3 0.07% 1 1.09%
Davidlohr Bueso A 3 0.07% 1 1.09%
Joe Perches 2 0.04% 1 1.09%
Shan Hai 2 0.04% 1 1.09%
Greg Kroah-Hartman 1 0.02% 1 1.09%
Ming Lei 1 0.02% 1 1.09%
Wang Sheng-Hui 1 0.02% 1 1.09%
Total 4605 92 


// SPDX-License-Identifier: GPL-2.0
/*
 * background writeback - scan btree for dirty data and write it to the backing
 * device
 *
 * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
 * Copyright 2012 Google, Inc.
 */

#include "bcache.h"
#include "btree.h"
#include "debug.h"
#include "writeback.h"

#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/sched/clock.h>
#include <trace/events/bcache.h>

static void update_gc_after_writeback(struct cache_set *c)
{
	if (c->gc_after_writeback != (BCH_ENABLE_AUTO_GC) ||
	    c->gc_stats.in_use < BCH_AUTO_GC_DIRTY_THRESHOLD)
		return;

	c->gc_after_writeback |= BCH_DO_AUTO_GC;
}

/* Rate limiting */
static uint64_t __calc_target_rate(struct cached_dev *dc)
{
	struct cache_set *c = dc->disk.c;

	/*
	 * This is the size of the cache, minus the amount used for
	 * flash-only devices
	 */
	uint64_t cache_sectors = c->nbuckets * c->cache->sb.bucket_size -
				atomic_long_read(&c->flash_dev_dirty_sectors);

	/*
	 * Unfortunately there is no control of global dirty data.  If the
	 * user states that they want 10% dirty data in the cache, and has,
	 * e.g., 5 backing volumes of equal size, we try and ensure each
	 * backing volume uses about 2% of the cache for dirty data.
	 */
	uint32_t bdev_share =
		div64_u64(bdev_nr_sectors(dc->bdev) << WRITEBACK_SHARE_SHIFT,
				c->cached_dev_sectors);

	uint64_t cache_dirty_target =
		div_u64(cache_sectors * dc->writeback_percent, 100);

	/* Ensure each backing dev gets at least one dirty share */
	if (bdev_share < 1)
		bdev_share = 1;

	return (cache_dirty_target * bdev_share) >> WRITEBACK_SHARE_SHIFT;
}

static void __update_writeback_rate(struct cached_dev *dc)
{
	/*
	 * PI controller:
	 * Figures out the amount that should be written per second.
	 *
	 * First, the error (number of sectors that are dirty beyond our
	 * target) is calculated.  The error is accumulated (numerically
	 * integrated).
	 *
	 * Then, the proportional value and integral value are scaled
	 * based on configured values.  These are stored as inverses to
	 * avoid fixed point math and to make configuration easy-- e.g.
	 * the default value of 40 for writeback_rate_p_term_inverse
	 * attempts to write at a rate that would retire all the dirty
	 * blocks in 40 seconds.
	 *
	 * The writeback_rate_i_inverse value of 10000 means that 1/10000th
	 * of the error is accumulated in the integral term per second.
	 * This acts as a slow, long-term average that is not subject to
	 * variations in usage like the p term.
	 */
	int64_t target = __calc_target_rate(dc);
	int64_t dirty = bcache_dev_sectors_dirty(&dc->disk);
	int64_t error = dirty - target;
	int64_t proportional_scaled =
		div_s64(error, dc->writeback_rate_p_term_inverse);
	int64_t integral_scaled;
	uint32_t new_rate;

	/*
	 * We need to consider the number of dirty buckets as well
	 * when calculating the proportional_scaled, Otherwise we might
	 * have an unreasonable small writeback rate at a highly fragmented situation
	 * when very few dirty sectors consumed a lot dirty buckets, the
	 * worst case is when dirty buckets reached cutoff_writeback_sync and
	 * dirty data is still not even reached to writeback percent, so the rate
	 * still will be at the minimum value, which will cause the write
	 * stuck at a non-writeback mode.
	 */
	struct cache_set *c = dc->disk.c;

	int64_t dirty_buckets = c->nbuckets - c->avail_nbuckets;

	if (dc->writeback_consider_fragment &&
		c->gc_stats.in_use > BCH_WRITEBACK_FRAGMENT_THRESHOLD_LOW && dirty > 0) {
		int64_t fragment =
			div_s64((dirty_buckets *  c->cache->sb.bucket_size), dirty);
		int64_t fp_term;
		int64_t fps;

		if (c->gc_stats.in_use <= BCH_WRITEBACK_FRAGMENT_THRESHOLD_MID) {
			fp_term = (int64_t)dc->writeback_rate_fp_term_low *
			(c->gc_stats.in_use - BCH_WRITEBACK_FRAGMENT_THRESHOLD_LOW);
		} else if (c->gc_stats.in_use <= BCH_WRITEBACK_FRAGMENT_THRESHOLD_HIGH) {
			fp_term = (int64_t)dc->writeback_rate_fp_term_mid *
			(c->gc_stats.in_use - BCH_WRITEBACK_FRAGMENT_THRESHOLD_MID);
		} else {
			fp_term = (int64_t)dc->writeback_rate_fp_term_high *
			(c->gc_stats.in_use - BCH_WRITEBACK_FRAGMENT_THRESHOLD_HIGH);
		}
		fps = div_s64(dirty, dirty_buckets) * fp_term;
		if (fragment > 3 && fps > proportional_scaled) {
			/* Only overrite the p when fragment > 3 */
			proportional_scaled = fps;
		}
	}

	if ((error < 0 && dc->writeback_rate_integral > 0) ||
	    (error > 0 && time_before64(local_clock(),
			 dc->writeback_rate.next + NSEC_PER_MSEC))) {
		/*
		 * Only decrease the integral term if it's more than
		 * zero.  Only increase the integral term if the device
		 * is keeping up.  (Don't wind up the integral
		 * ineffectively in either case).
		 *
		 * It's necessary to scale this by
		 * writeback_rate_update_seconds to keep the integral
		 * term dimensioned properly.
		 */
		dc->writeback_rate_integral += error *
			dc->writeback_rate_update_seconds;
	}

	integral_scaled = div_s64(dc->writeback_rate_integral,
			dc->writeback_rate_i_term_inverse);

	new_rate = clamp_t(int32_t, (proportional_scaled + integral_scaled),
			dc->writeback_rate_minimum, NSEC_PER_SEC);

	dc->writeback_rate_proportional = proportional_scaled;
	dc->writeback_rate_integral_scaled = integral_scaled;
	dc->writeback_rate_change = new_rate -
			atomic_long_read(&dc->writeback_rate.rate);
	atomic_long_set(&dc->writeback_rate.rate, new_rate);
	dc->writeback_rate_target = target;
}

static bool idle_counter_exceeded(struct cache_set *c)
{
	int counter, dev_nr;

	/*
	 * If c->idle_counter is overflow (idel for really long time),
	 * reset as 0 and not set maximum rate this time for code
	 * simplicity.
	 */
	counter = atomic_inc_return(&c->idle_counter);
	if (counter <= 0) {
		atomic_set(&c->idle_counter, 0);
		return false;
	}

	dev_nr = atomic_read(&c->attached_dev_nr);
	if (dev_nr == 0)
		return false;

	/*
	 * c->idle_counter is increased by writeback thread of all
	 * attached backing devices, in order to represent a rough
	 * time period, counter should be divided by dev_nr.
	 * Otherwise the idle time cannot be larger with more backing
	 * device attached.
	 * The following calculation equals to checking
	 *	(counter / dev_nr) < (dev_nr * 6)
	 */
	if (counter < (dev_nr * dev_nr * 6))
		return false;

	return true;
}

/*
 * Idle_counter is increased every time when update_writeback_rate() is
 * called. If all backing devices attached to the same cache set have
 * identical dc->writeback_rate_update_seconds values, it is about 6
 * rounds of update_writeback_rate() on each backing device before
 * c->at_max_writeback_rate is set to 1, and then max wrteback rate set
 * to each dc->writeback_rate.rate.
 * In order to avoid extra locking cost for counting exact dirty cached
 * devices number, c->attached_dev_nr is used to calculate the idle
 * throushold. It might be bigger if not all cached device are in write-
 * back mode, but it still works well with limited extra rounds of
 * update_writeback_rate().
 */
static bool set_at_max_writeback_rate(struct cache_set *c,
				       struct cached_dev *dc)
{
	/* Don't sst max writeback rate if it is disabled */
	if (!c->idle_max_writeback_rate_enabled)
		return false;

	/* Don't set max writeback rate if gc is running */
	if (!c->gc_mark_valid)
		return false;

	if (!idle_counter_exceeded(c))
		return false;

	if (atomic_read(&c->at_max_writeback_rate) != 1)
		atomic_set(&c->at_max_writeback_rate, 1);

	atomic_long_set(&dc->writeback_rate.rate, INT_MAX);

	/* keep writeback_rate_target as existing value */
	dc->writeback_rate_proportional = 0;
	dc->writeback_rate_integral_scaled = 0;
	dc->writeback_rate_change = 0;

	/*
	 * In case new I/O arrives during before
	 * set_at_max_writeback_rate() returns.
	 */
	if (!idle_counter_exceeded(c) ||
	    !atomic_read(&c->at_max_writeback_rate))
		return false;

	return true;
}

static void update_writeback_rate(struct work_struct *work)
{
	struct cached_dev *dc = container_of(to_delayed_work(work),
					     struct cached_dev,
					     writeback_rate_update);
	struct cache_set *c = dc->disk.c;

	/*
	 * should check BCACHE_DEV_RATE_DW_RUNNING before calling
	 * cancel_delayed_work_sync().
	 */
	set_bit(BCACHE_DEV_RATE_DW_RUNNING, &dc->disk.flags);
	/* paired with where BCACHE_DEV_RATE_DW_RUNNING is tested */
	smp_mb__after_atomic();

	/*
	 * CACHE_SET_IO_DISABLE might be set via sysfs interface,
	 * check it here too.
	 */
	if (!test_bit(BCACHE_DEV_WB_RUNNING, &dc->disk.flags) ||
	    test_bit(CACHE_SET_IO_DISABLE, &c->flags)) {
		clear_bit(BCACHE_DEV_RATE_DW_RUNNING, &dc->disk.flags);
		/* paired with where BCACHE_DEV_RATE_DW_RUNNING is tested */
		smp_mb__after_atomic();
		return;
	}

	/*
	 * If the whole cache set is idle, set_at_max_writeback_rate()
	 * will set writeback rate to a max number. Then it is
	 * unncessary to update writeback rate for an idle cache set
	 * in maximum writeback rate number(s).
	 */
	if (atomic_read(&dc->has_dirty) && dc->writeback_percent &&
	    !set_at_max_writeback_rate(c, dc)) {
		do {
			if (!down_read_trylock((&dc->writeback_lock))) {
				dc->rate_update_retry++;
				if (dc->rate_update_retry <=
				    BCH_WBRATE_UPDATE_MAX_SKIPS)
					break;
				down_read(&dc->writeback_lock);
				dc->rate_update_retry = 0;
			}
			__update_writeback_rate(dc);
			update_gc_after_writeback(c);
			up_read(&dc->writeback_lock);
		} while (0);
	}


	/*
	 * CACHE_SET_IO_DISABLE might be set via sysfs interface,
	 * check it here too.
	 */
	if (test_bit(BCACHE_DEV_WB_RUNNING, &dc->disk.flags) &&
	    !test_bit(CACHE_SET_IO_DISABLE, &c->flags)) {
		schedule_delayed_work(&dc->writeback_rate_update,
			      dc->writeback_rate_update_seconds * HZ);
	}

	/*
	 * should check BCACHE_DEV_RATE_DW_RUNNING before calling
	 * cancel_delayed_work_sync().
	 */
	clear_bit(BCACHE_DEV_RATE_DW_RUNNING, &dc->disk.flags);
	/* paired with where BCACHE_DEV_RATE_DW_RUNNING is tested */
	smp_mb__after_atomic();
}

static unsigned int writeback_delay(struct cached_dev *dc,
				    unsigned int sectors)
{
	if (test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) ||
	    !dc->writeback_percent)
		return 0;

	return bch_next_delay(&dc->writeback_rate, sectors);
}

struct dirty_io {
	struct closure		cl;
	struct cached_dev	*dc;
	uint16_t		sequence;
	struct bio		bio;
};

static void dirty_init(struct keybuf_key *w)
{
	struct dirty_io *io = w->private;
	struct bio *bio = &io->bio;

	bio_init(bio, NULL, bio->bi_inline_vecs,
		 DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS), 0);
	if (!io->dc->writeback_percent)
		bio_set_prio(bio, IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0));

	bio->bi_iter.bi_size	= KEY_SIZE(&w->key) << 9;
	bio->bi_private		= w;
	bch_bio_map(bio, NULL);
}

static CLOSURE_CALLBACK(dirty_io_destructor)
{
	closure_type(io, struct dirty_io, cl);

	kfree(io);
}

static CLOSURE_CALLBACK(write_dirty_finish)
{
	closure_type(io, struct dirty_io, cl);
	struct keybuf_key *w = io->bio.bi_private;
	struct cached_dev *dc = io->dc;

	bio_free_pages(&io->bio);

	/* This is kind of a dumb way of signalling errors. */
	if (KEY_DIRTY(&w->key)) {
		int ret;
		unsigned int i;
		struct keylist keys;

		bch_keylist_init(&keys);

		bkey_copy(keys.top, &w->key);
		SET_KEY_DIRTY(keys.top, false);
		bch_keylist_push(&keys);

		for (i = 0; i < KEY_PTRS(&w->key); i++)
			atomic_inc(&PTR_BUCKET(dc->disk.c, &w->key, i)->pin);

		ret = bch_btree_insert(dc->disk.c, &keys, NULL, &w->key);

		if (ret)
			trace_bcache_writeback_collision(&w->key);

		atomic_long_inc(ret
				? &dc->disk.c->writeback_keys_failed
				: &dc->disk.c->writeback_keys_done);
	}

	bch_keybuf_del(&dc->writeback_keys, w);
	up(&dc->in_flight);

	closure_return_with_destructor(cl, dirty_io_destructor);
}

static void dirty_endio(struct bio *bio)
{
	struct keybuf_key *w = bio->bi_private;
	struct dirty_io *io = w->private;

	if (bio->bi_status) {
		SET_KEY_DIRTY(&w->key, false);
		bch_count_backing_io_errors(io->dc, bio);
	}

	closure_put(&io->cl);
}

static CLOSURE_CALLBACK(write_dirty)
{
	closure_type(io, struct dirty_io, cl);
	struct keybuf_key *w = io->bio.bi_private;
	struct cached_dev *dc = io->dc;

	uint16_t next_sequence;

	if (atomic_read(&dc->writeback_sequence_next) != io->sequence) {
		/* Not our turn to write; wait for a write to complete */
		closure_wait(&dc->writeback_ordering_wait, cl);

		if (atomic_read(&dc->writeback_sequence_next) == io->sequence) {
			/*
			 * Edge case-- it happened in indeterminate order
			 * relative to when we were added to wait list..
			 */
			closure_wake_up(&dc->writeback_ordering_wait);
		}

		continue_at(cl, write_dirty, io->dc->writeback_write_wq);
		return;
	}

	next_sequence = io->sequence + 1;

	/*
	 * IO errors are signalled using the dirty bit on the key.
	 * If we failed to read, we should not attempt to write to the
	 * backing device.  Instead, immediately go to write_dirty_finish
	 * to clean up.
	 */
	if (KEY_DIRTY(&w->key)) {
		dirty_init(w);
		io->bio.bi_opf = REQ_OP_WRITE;
		io->bio.bi_iter.bi_sector = KEY_START(&w->key);
		bio_set_dev(&io->bio, io->dc->bdev);
		io->bio.bi_end_io	= dirty_endio;

		/* I/O request sent to backing device */
		closure_bio_submit(io->dc->disk.c, &io->bio, cl);
	}

	atomic_set(&dc->writeback_sequence_next, next_sequence);
	closure_wake_up(&dc->writeback_ordering_wait);

	continue_at(cl, write_dirty_finish, io->dc->writeback_write_wq);
}

static void read_dirty_endio(struct bio *bio)
{
	struct keybuf_key *w = bio->bi_private;
	struct dirty_io *io = w->private;

	/* is_read = 1 */
	bch_count_io_errors(io->dc->disk.c->cache,
			    bio->bi_status, 1,
			    "reading dirty data from cache");

	dirty_endio(bio);
}

static CLOSURE_CALLBACK(read_dirty_submit)
{
	closure_type(io, struct dirty_io, cl);

	closure_bio_submit(io->dc->disk.c, &io->bio, cl);

	continue_at(cl, write_dirty, io->dc->writeback_write_wq);
}

static void read_dirty(struct cached_dev *dc)
{
	unsigned int delay = 0;
	struct keybuf_key *next, *keys[MAX_WRITEBACKS_IN_PASS], *w;
	size_t size;
	int nk, i;
	struct dirty_io *io;
	struct closure cl;
	uint16_t sequence = 0;

	BUG_ON(!llist_empty(&dc->writeback_ordering_wait.list));
	atomic_set(&dc->writeback_sequence_next, sequence);
	closure_init_stack(&cl);

	/*
	 * XXX: if we error, background writeback just spins. Should use some
	 * mempools.
	 */

	next = bch_keybuf_next(&dc->writeback_keys);

	while (!kthread_should_stop() &&
	       !test_bit(CACHE_SET_IO_DISABLE, &dc->disk.c->flags) &&
	       next) {
		size = 0;
		nk = 0;

		do {
			BUG_ON(ptr_stale(dc->disk.c, &next->key, 0));

			/*
			 * Don't combine too many operations, even if they
			 * are all small.
			 */
			if (nk >= MAX_WRITEBACKS_IN_PASS)
				break;

			/*
			 * If the current operation is very large, don't
			 * further combine operations.
			 */
			if (size >= MAX_WRITESIZE_IN_PASS)
				break;

			/*
			 * Operations are only eligible to be combined
			 * if they are contiguous.
			 *
			 * TODO: add a heuristic willing to fire a
			 * certain amount of non-contiguous IO per pass,
			 * so that we can benefit from backing device
			 * command queueing.
			 */
			if ((nk != 0) && bkey_cmp(&keys[nk-1]->key,
						&START_KEY(&next->key)))
				break;

			size += KEY_SIZE(&next->key);
			keys[nk++] = next;
		} while ((next = bch_keybuf_next(&dc->writeback_keys)));

		/* Now we have gathered a set of 1..5 keys to write back. */
		for (i = 0; i < nk; i++) {
			w = keys[i];

			io = kzalloc(struct_size(io, bio.bi_inline_vecs,
						DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS)),
				     GFP_KERNEL);
			if (!io)
				goto err;

			w->private	= io;
			io->dc		= dc;
			io->sequence    = sequence++;

			dirty_init(w);
			io->bio.bi_opf = REQ_OP_READ;
			io->bio.bi_iter.bi_sector = PTR_OFFSET(&w->key, 0);
			bio_set_dev(&io->bio, dc->disk.c->cache->bdev);
			io->bio.bi_end_io	= read_dirty_endio;

			if (bch_bio_alloc_pages(&io->bio, GFP_KERNEL))
				goto err_free;

			trace_bcache_writeback(&w->key);

			down(&dc->in_flight);

			/*
			 * We've acquired a semaphore for the maximum
			 * simultaneous number of writebacks; from here
			 * everything happens asynchronously.
			 */
			closure_call(&io->cl, read_dirty_submit, NULL, &cl);
		}

		delay = writeback_delay(dc, size);

		while (!kthread_should_stop() &&
		       !test_bit(CACHE_SET_IO_DISABLE, &dc->disk.c->flags) &&
		       delay) {
			schedule_timeout_interruptible(delay);
			delay = writeback_delay(dc, 0);
		}
	}

	if (0) {
err_free:
		kfree(w->private);
err:
		bch_keybuf_del(&dc->writeback_keys, w);
	}

	/*
	 * Wait for outstanding writeback IOs to finish (and keybuf slots to be
	 * freed) before refilling again
	 */
	closure_sync(&cl);
}

/* Scan for dirty data */

void bcache_dev_sectors_dirty_add(struct cache_set *c, unsigned int inode,
				  uint64_t offset, int nr_sectors)
{
	struct bcache_device *d = c->devices[inode];
	unsigned int stripe_offset, sectors_dirty;
	int stripe;

	if (!d)
		return;

	stripe = offset_to_stripe(d, offset);
	if (stripe < 0)
		return;

	if (UUID_FLASH_ONLY(&c->uuids[inode]))
		atomic_long_add(nr_sectors, &c->flash_dev_dirty_sectors);

	stripe_offset = offset & (d->stripe_size - 1);

	while (nr_sectors) {
		int s = min_t(unsigned int, abs(nr_sectors),
			      d->stripe_size - stripe_offset);

		if (nr_sectors < 0)
			s = -s;

		if (stripe >= d->nr_stripes)
			return;

		sectors_dirty = atomic_add_return(s,
					d->stripe_sectors_dirty + stripe);
		if (sectors_dirty == d->stripe_size) {
			if (!test_bit(stripe, d->full_dirty_stripes))
				set_bit(stripe, d->full_dirty_stripes);
		} else {
			if (test_bit(stripe, d->full_dirty_stripes))
				clear_bit(stripe, d->full_dirty_stripes);
		}

		nr_sectors -= s;
		stripe_offset = 0;
		stripe++;
	}
}

static bool dirty_pred(struct keybuf *buf, struct bkey *k)
{
	struct cached_dev *dc = container_of(buf,
					     struct cached_dev,
					     writeback_keys);

	BUG_ON(KEY_INODE(k) != dc->disk.id);

	return KEY_DIRTY(k);
}

static void refill_full_stripes(struct cached_dev *dc)
{
	struct keybuf *buf = &dc->writeback_keys;
	unsigned int start_stripe, next_stripe;
	int stripe;
	bool wrapped = false;

	stripe = offset_to_stripe(&dc->disk, KEY_OFFSET(&buf->last_scanned));
	if (stripe < 0)
		stripe = 0;

	start_stripe = stripe;

	while (1) {
		stripe = find_next_bit(dc->disk.full_dirty_stripes,
				       dc->disk.nr_stripes, stripe);

		if (stripe == dc->disk.nr_stripes)
			goto next;

		next_stripe = find_next_zero_bit(dc->disk.full_dirty_stripes,
						 dc->disk.nr_stripes, stripe);

		buf->last_scanned = KEY(dc->disk.id,
					stripe * dc->disk.stripe_size, 0);

		bch_refill_keybuf(dc->disk.c, buf,
				  &KEY(dc->disk.id,
				       next_stripe * dc->disk.stripe_size, 0),
				  dirty_pred);

		if (array_freelist_empty(&buf->freelist))
			return;

		stripe = next_stripe;
next:
		if (wrapped && stripe > start_stripe)
			return;

		if (stripe == dc->disk.nr_stripes) {
			stripe = 0;
			wrapped = true;
		}
	}
}

/*
 * Returns true if we scanned the entire disk
 */
static bool refill_dirty(struct cached_dev *dc)
{
	struct keybuf *buf = &dc->writeback_keys;
	struct bkey start = KEY(dc->disk.id, 0, 0);
	struct bkey end = KEY(dc->disk.id, MAX_KEY_OFFSET, 0);
	struct bkey start_pos;

	/*
	 * make sure keybuf pos is inside the range for this disk - at bringup
	 * we might not be attached yet so this disk's inode nr isn't
	 * initialized then
	 */
	if (bkey_cmp(&buf->last_scanned, &start) < 0 ||
	    bkey_cmp(&buf->last_scanned, &end) > 0)
		buf->last_scanned = start;

	if (dc->partial_stripes_expensive) {
		refill_full_stripes(dc);
		if (array_freelist_empty(&buf->freelist))
			return false;
	}

	start_pos = buf->last_scanned;
	bch_refill_keybuf(dc->disk.c, buf, &end, dirty_pred);

	if (bkey_cmp(&buf->last_scanned, &end) < 0)
		return false;

	/*
	 * If we get to the end start scanning again from the beginning, and
	 * only scan up to where we initially started scanning from:
	 */
	buf->last_scanned = start;
	bch_refill_keybuf(dc->disk.c, buf, &start_pos, dirty_pred);

	return bkey_cmp(&buf->last_scanned, &start_pos) >= 0;
}

static int bch_writeback_thread(void *arg)
{
	struct cached_dev *dc = arg;
	struct cache_set *c = dc->disk.c;
	bool searched_full_index;

	bch_ratelimit_reset(&dc->writeback_rate);

	while (!kthread_should_stop() &&
	       !test_bit(CACHE_SET_IO_DISABLE, &c->flags)) {
		down_write(&dc->writeback_lock);
		set_current_state(TASK_INTERRUPTIBLE);
		/*
		 * If the bache device is detaching, skip here and continue
		 * to perform writeback. Otherwise, if no dirty data on cache,
		 * or there is dirty data on cache but writeback is disabled,
		 * the writeback thread should sleep here and wait for others
		 * to wake up it.
		 */
		if (!test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) &&
		    (!atomic_read(&dc->has_dirty) || !dc->writeback_running)) {
			up_write(&dc->writeback_lock);

			if (kthread_should_stop() ||
			    test_bit(CACHE_SET_IO_DISABLE, &c->flags)) {
				set_current_state(TASK_RUNNING);
				break;
			}

			schedule();
			continue;
		}
		set_current_state(TASK_RUNNING);

		searched_full_index = refill_dirty(dc);

		if (searched_full_index &&
		    RB_EMPTY_ROOT(&dc->writeback_keys.keys)) {
			atomic_set(&dc->has_dirty, 0);
			SET_BDEV_STATE(&dc->sb, BDEV_STATE_CLEAN);
			bch_write_bdev_super(dc, NULL);
			/*
			 * If bcache device is detaching via sysfs interface,
			 * writeback thread should stop after there is no dirty
			 * data on cache. BCACHE_DEV_DETACHING flag is set in
			 * bch_cached_dev_detach().
			 */
			if (test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags)) {
				struct closure cl;

				closure_init_stack(&cl);
				memset(&dc->sb.set_uuid, 0, 16);
				SET_BDEV_STATE(&dc->sb, BDEV_STATE_NONE);

				bch_write_bdev_super(dc, &cl);
				closure_sync(&cl);

				up_write(&dc->writeback_lock);
				break;
			}

			/*
			 * When dirty data rate is high (e.g. 50%+), there might
			 * be heavy buckets fragmentation after writeback
			 * finished, which hurts following write performance.
			 * If users really care about write performance they
			 * may set BCH_ENABLE_AUTO_GC via sysfs, then when
			 * BCH_DO_AUTO_GC is set, garbage collection thread
			 * will be wake up here. After moving gc, the shrunk
			 * btree and discarded free buckets SSD space may be
			 * helpful for following write requests.
			 */
			if (c->gc_after_writeback ==
			    (BCH_ENABLE_AUTO_GC|BCH_DO_AUTO_GC)) {
				c->gc_after_writeback &= ~BCH_DO_AUTO_GC;
				force_wake_up_gc(c);
			}
		}

		up_write(&dc->writeback_lock);

		read_dirty(dc);

		if (searched_full_index) {
			unsigned int delay = dc->writeback_delay * HZ;

			while (delay &&
			       !kthread_should_stop() &&
			       !test_bit(CACHE_SET_IO_DISABLE, &c->flags) &&
			       !test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags))
				delay = schedule_timeout_interruptible(delay);

			bch_ratelimit_reset(&dc->writeback_rate);
		}
	}

	if (dc->writeback_write_wq)
		destroy_workqueue(dc->writeback_write_wq);

	cached_dev_put(dc);
	wait_for_kthread_stop();

	return 0;
}

/* Init */
#define INIT_KEYS_EACH_TIME	500000

struct sectors_dirty_init {
	struct btree_op	op;
	unsigned int	inode;
	size_t		count;
};

static int sectors_dirty_init_fn(struct btree_op *_op, struct btree *b,
				 struct bkey *k)
{
	struct sectors_dirty_init *op = container_of(_op,
						struct sectors_dirty_init, op);
	if (KEY_INODE(k) > op->inode)
		return MAP_DONE;

	if (KEY_DIRTY(k))
		bcache_dev_sectors_dirty_add(b->c, KEY_INODE(k),
					     KEY_START(k), KEY_SIZE(k));

	op->count++;
	if (!(op->count % INIT_KEYS_EACH_TIME))
		cond_resched();

	return MAP_CONTINUE;
}

static int bch_root_node_dirty_init(struct cache_set *c,
				     struct bcache_device *d,
				     struct bkey *k)
{
	struct sectors_dirty_init op;
	int ret;

	bch_btree_op_init(&op.op, -1);
	op.inode = d->id;
	op.count = 0;

	ret = bcache_btree(map_keys_recurse,
			   k,
			   c->root,
			   &op.op,
			   &KEY(op.inode, 0, 0),
			   sectors_dirty_init_fn,
			   0);
	if (ret < 0)
		pr_warn("sectors dirty init failed, ret=%d!\n", ret);

	/*
	 * The op may be added to cache_set's btree_cache_wait
	 * in mca_cannibalize(), must ensure it is removed from
	 * the list and release btree_cache_alloc_lock before
	 * free op memory.
	 * Otherwise, the btree_cache_wait will be damaged.
	 */
	bch_cannibalize_unlock(c);
	finish_wait(&c->btree_cache_wait, &(&op.op)->wait);

	return ret;
}

static int bch_dirty_init_thread(void *arg)
{
	struct dirty_init_thrd_info *info = arg;
	struct bch_dirty_init_state *state = info->state;
	struct cache_set *c = state->c;
	struct btree_iter_stack iter;
	struct bkey *k, *p;
	int cur_idx, prev_idx, skip_nr;

	k = p = NULL;
	prev_idx = 0;

	bch_btree_iter_stack_init(&c->root->keys, &iter, NULL);
	k = bch_btree_iter_next_filter(&iter.iter, &c->root->keys, bch_ptr_bad);
	BUG_ON(!k);

	p = k;

	while (k) {
		spin_lock(&state->idx_lock);
		cur_idx = state->key_idx;
		state->key_idx++;
		spin_unlock(&state->idx_lock);

		skip_nr = cur_idx - prev_idx;

		while (skip_nr) {
			k = bch_btree_iter_next_filter(&iter.iter,
						       &c->root->keys,
						       bch_ptr_bad);
			if (k)
				p = k;
			else {
				atomic_set(&state->enough, 1);
				/* Update state->enough earlier */
				smp_mb__after_atomic();
				goto out;
			}
			skip_nr--;
		}

		if (p) {
			if (bch_root_node_dirty_init(c, state->d, p) < 0)
				goto out;
		}

		p = NULL;
		prev_idx = cur_idx;
	}

out:
	/* In order to wake up state->wait in time */
	smp_mb__before_atomic();
	if (atomic_dec_and_test(&state->started))
		wake_up(&state->wait);

	return 0;
}

static int bch_btre_dirty_init_thread_nr(void)
{
	int n = num_online_cpus()/2;

	if (n == 0)
		n = 1;
	else if (n > BCH_DIRTY_INIT_THRD_MAX)
		n = BCH_DIRTY_INIT_THRD_MAX;

	return n;
}

void bch_sectors_dirty_init(struct bcache_device *d)
{
	int i;
	struct btree *b = NULL;
	struct bkey *k = NULL;
	struct btree_iter_stack iter;
	struct sectors_dirty_init op;
	struct cache_set *c = d->c;
	struct bch_dirty_init_state state;

retry_lock:
	b = c->root;
	rw_lock(0, b, b->level);
	if (b != c->root) {
		rw_unlock(0, b);
		goto retry_lock;
	}

	/* Just count root keys if no leaf node */
	if (c->root->level == 0) {
		bch_btree_op_init(&op.op, -1);
		op.inode = d->id;
		op.count = 0;

		for_each_key_filter(&c->root->keys,
				    k, &iter, bch_ptr_invalid) {
			if (KEY_INODE(k) != op.inode)
				continue;
			sectors_dirty_init_fn(&op.op, c->root, k);
		}

		rw_unlock(0, b);
		return;
	}

	memset(&state, 0, sizeof(struct bch_dirty_init_state));
	state.c = c;
	state.d = d;
	state.total_threads = bch_btre_dirty_init_thread_nr();
	state.key_idx = 0;
	spin_lock_init(&state.idx_lock);
	atomic_set(&state.started, 0);
	atomic_set(&state.enough, 0);
	init_waitqueue_head(&state.wait);

	for (i = 0; i < state.total_threads; i++) {
		/* Fetch latest state.enough earlier */
		smp_mb__before_atomic();
		if (atomic_read(&state.enough))
			break;

		atomic_inc(&state.started);
		state.infos[i].state = &state;
		state.infos[i].thread =
			kthread_run(bch_dirty_init_thread, &state.infos[i],
				    "bch_dirtcnt[%d]", i);
		if (IS_ERR(state.infos[i].thread)) {
			pr_err("fails to run thread bch_dirty_init[%d]\n", i);
			atomic_dec(&state.started);
			for (--i; i >= 0; i--)
				kthread_stop(state.infos[i].thread);
			goto out;
		}
	}

out:
	/* Must wait for all threads to stop. */
	wait_event(state.wait, atomic_read(&state.started) == 0);
	rw_unlock(0, b);
}

void bch_cached_dev_writeback_init(struct cached_dev *dc)
{
	sema_init(&dc->in_flight, 64);
	init_rwsem(&dc->writeback_lock);
	bch_keybuf_init(&dc->writeback_keys);

	dc->writeback_metadata		= true;
	dc->writeback_running		= false;
	dc->writeback_consider_fragment = true;
	dc->writeback_percent		= 10;
	dc->writeback_delay		= 30;
	atomic_long_set(&dc->writeback_rate.rate, 1024);
	dc->writeback_rate_minimum	= 8;

	dc->writeback_rate_update_seconds = WRITEBACK_RATE_UPDATE_SECS_DEFAULT;
	dc->writeback_rate_p_term_inverse = 40;
	dc->writeback_rate_fp_term_low = 1;
	dc->writeback_rate_fp_term_mid = 10;
	dc->writeback_rate_fp_term_high = 1000;
	dc->writeback_rate_i_term_inverse = 10000;

	/* For dc->writeback_lock contention in update_writeback_rate() */
	dc->rate_update_retry = 0;

	WARN_ON(test_and_clear_bit(BCACHE_DEV_WB_RUNNING, &dc->disk.flags));
	INIT_DELAYED_WORK(&dc->writeback_rate_update, update_writeback_rate);
}

int bch_cached_dev_writeback_start(struct cached_dev *dc)
{
	dc->writeback_write_wq = alloc_workqueue("bcache_writeback_wq",
						WQ_MEM_RECLAIM, 0);
	if (!dc->writeback_write_wq)
		return -ENOMEM;

	cached_dev_get(dc);
	dc->writeback_thread = kthread_create(bch_writeback_thread, dc,
					      "bcache_writeback");
	if (IS_ERR(dc->writeback_thread)) {
		cached_dev_put(dc);
		destroy_workqueue(dc->writeback_write_wq);
		return PTR_ERR(dc->writeback_thread);
	}
	dc->writeback_running = true;

	WARN_ON(test_and_set_bit(BCACHE_DEV_WB_RUNNING, &dc->disk.flags));
	schedule_delayed_work(&dc->writeback_rate_update,
			      dc->writeback_rate_update_seconds * HZ);

	bch_writeback_queue(dc);

	return 0;
}