1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * bcache journalling code, for btree insertions
   4 *
   5 * Copyright 2012 Google, Inc.
   6 */
   7
   8#include "bcache.h"
   9#include "btree.h"
  10#include "debug.h"
  11#include "extents.h"
  12
  13#include <trace/events/bcache.h>
  14
  15/*
  16 * Journal replay/recovery:
  17 *
  18 * This code is all driven from run_cache_set(); we first read the journal
  19 * entries, do some other stuff, then we mark all the keys in the journal
  20 * entries (same as garbage collection would), then we replay them - reinserting
  21 * them into the cache in precisely the same order as they appear in the
  22 * journal.
  23 *
  24 * We only journal keys that go in leaf nodes, which simplifies things quite a
  25 * bit.
  26 */
  27
  28static void journal_read_endio(struct bio *bio)
  29{
  30	struct closure *cl = bio->bi_private;
  31
  32	closure_put(cl);
  33}
  34
  35static int journal_read_bucket(struct cache *ca, struct list_head *list,
  36			       unsigned int bucket_index)
  37{
  38	struct journal_device *ja = &ca->journal;
  39	struct bio *bio = &ja->bio;
  40
  41	struct journal_replay *i;
  42	struct jset *j, *data = ca->set->journal.w[0].data;
  43	struct closure cl;
  44	unsigned int len, left, offset = 0;
  45	int ret = 0;
  46	sector_t bucket = bucket_to_sector(ca->set, ca->sb.d[bucket_index]);
  47
  48	closure_init_stack(&cl);
  49
  50	pr_debug("reading %u\n", bucket_index);
  51
  52	while (offset < ca->sb.bucket_size) {
  53reread:		left = ca->sb.bucket_size - offset;
  54		len = min_t(unsigned int, left, PAGE_SECTORS << JSET_BITS);
  55
  56		bio_reset(bio, ca->bdev, REQ_OP_READ);
  57		bio->bi_iter.bi_sector	= bucket + offset;
  58		bio->bi_iter.bi_size	= len << 9;
  59
  60		bio->bi_end_io	= journal_read_endio;
  61		bio->bi_private = &cl;
  62		bch_bio_map(bio, data);
  63
  64		closure_bio_submit(ca->set, bio, &cl);
  65		closure_sync(&cl);
  66
  67		/* This function could be simpler now since we no longer write
  68		 * journal entries that overlap bucket boundaries; this means
  69		 * the start of a bucket will always have a valid journal entry
  70		 * if it has any journal entries at all.
  71		 */
  72
  73		j = data;
  74		while (len) {
  75			struct list_head *where;
  76			size_t blocks, bytes = set_bytes(j);
  77
  78			if (j->magic != jset_magic(&ca->sb)) {
  79				pr_debug("%u: bad magic\n", bucket_index);
  80				return ret;
  81			}
  82
  83			if (bytes > left << 9 ||
  84			    bytes > PAGE_SIZE << JSET_BITS) {
  85				pr_info("%u: too big, %zu bytes, offset %u\n",
  86					bucket_index, bytes, offset);
  87				return ret;
  88			}
  89
  90			if (bytes > len << 9)
  91				goto reread;
  92
  93			if (j->csum != csum_set(j)) {
  94				pr_info("%u: bad csum, %zu bytes, offset %u\n",
  95					bucket_index, bytes, offset);
  96				return ret;
  97			}
  98
  99			blocks = set_blocks(j, block_bytes(ca));
 100
 101			/*
 102			 * Nodes in 'list' are in linear increasing order of
 103			 * i->j.seq, the node on head has the smallest (oldest)
 104			 * journal seq, the node on tail has the biggest
 105			 * (latest) journal seq.
 106			 */
 107
 108			/*
 109			 * Check from the oldest jset for last_seq. If
 110			 * i->j.seq < j->last_seq, it means the oldest jset
 111			 * in list is expired and useless, remove it from
 112			 * this list. Otherwise, j is a candidate jset for
 113			 * further following checks.
 114			 */
 115			while (!list_empty(list)) {
 116				i = list_first_entry(list,
 117					struct journal_replay, list);
 118				if (i->j.seq >= j->last_seq)
 119					break;
 120				list_del(&i->list);
 121				kfree(i);
 122			}
 123
 124			/* iterate list in reverse order (from latest jset) */
 125			list_for_each_entry_reverse(i, list, list) {
 126				if (j->seq == i->j.seq)
 127					goto next_set;
 128
 129				/*
 130				 * if j->seq is less than any i->j.last_seq
 131				 * in list, j is an expired and useless jset.
 132				 */
 133				if (j->seq < i->j.last_seq)
 134					goto next_set;
 135
 136				/*
 137				 * 'where' points to first jset in list which
 138				 * is elder then j.
 139				 */
 140				if (j->seq > i->j.seq) {
 141					where = &i->list;
 142					goto add;
 143				}
 144			}
 145
 146			where = list;
 147add:
 148			i = kmalloc(offsetof(struct journal_replay, j) +
 149				    bytes, GFP_KERNEL);
 150			if (!i)
 151				return -ENOMEM;
 152			unsafe_memcpy(&i->j, j, bytes,
 153				/* "bytes" was calculated by set_bytes() above */);
 154			/* Add to the location after 'where' points to */
 155			list_add(&i->list, where);
 156			ret = 1;
 157
 158			if (j->seq > ja->seq[bucket_index])
 159				ja->seq[bucket_index] = j->seq;
 160next_set:
 161			offset	+= blocks * ca->sb.block_size;
 162			len	-= blocks * ca->sb.block_size;
 163			j = ((void *) j) + blocks * block_bytes(ca);
 164		}
 165	}
 166
 167	return ret;
 168}
 169
 170int bch_journal_read(struct cache_set *c, struct list_head *list)
 171{
 172#define read_bucket(b)							\
 173	({								\
 174		ret = journal_read_bucket(ca, list, b);			\
 175		__set_bit(b, bitmap);					\
 176		if (ret < 0)						\
 177			return ret;					\
 178		ret;							\
 179	})
 180
 181	struct cache *ca = c->cache;
 182	int ret = 0;
 183	struct journal_device *ja = &ca->journal;
 184	DECLARE_BITMAP(bitmap, SB_JOURNAL_BUCKETS);
 185	unsigned int i, l, r, m;
 186	uint64_t seq;
 187
 188	bitmap_zero(bitmap, SB_JOURNAL_BUCKETS);
 189	pr_debug("%u journal buckets\n", ca->sb.njournal_buckets);
 190
 191	/*
 192	 * Read journal buckets ordered by golden ratio hash to quickly
 193	 * find a sequence of buckets with valid journal entries
 194	 */
 195	for (i = 0; i < ca->sb.njournal_buckets; i++) {
 196		/*
 197		 * We must try the index l with ZERO first for
 198		 * correctness due to the scenario that the journal
 199		 * bucket is circular buffer which might have wrapped
 200		 */
 201		l = (i * 2654435769U) % ca->sb.njournal_buckets;
 202
 203		if (test_bit(l, bitmap))
 204			break;
 205
 206		if (read_bucket(l))
 207			goto bsearch;
 208	}
 209
 210	/*
 211	 * If that fails, check all the buckets we haven't checked
 212	 * already
 213	 */
 214	pr_debug("falling back to linear search\n");
 215
 216	for_each_clear_bit(l, bitmap, ca->sb.njournal_buckets)
 217		if (read_bucket(l))
 218			goto bsearch;
 219
 220	/* no journal entries on this device? */
 221	if (l == ca->sb.njournal_buckets)
 222		goto out;
 223bsearch:
 224	BUG_ON(list_empty(list));
 225
 226	/* Binary search */
 227	m = l;
 228	r = find_next_bit(bitmap, ca->sb.njournal_buckets, l + 1);
 229	pr_debug("starting binary search, l %u r %u\n", l, r);
 230
 231	while (l + 1 < r) {
 232		seq = list_entry(list->prev, struct journal_replay,
 233				 list)->j.seq;
 234
 235		m = (l + r) >> 1;
 236		read_bucket(m);
 237
 238		if (seq != list_entry(list->prev, struct journal_replay,
 239				      list)->j.seq)
 240			l = m;
 241		else
 242			r = m;
 243	}
 244
 245	/*
 246	 * Read buckets in reverse order until we stop finding more
 247	 * journal entries
 248	 */
 249	pr_debug("finishing up: m %u njournal_buckets %u\n",
 250		 m, ca->sb.njournal_buckets);
 251	l = m;
 252
 253	while (1) {
 254		if (!l--)
 255			l = ca->sb.njournal_buckets - 1;
 256
 257		if (l == m)
 258			break;
 259
 260		if (test_bit(l, bitmap))
 261			continue;
 262
 263		if (!read_bucket(l))
 264			break;
 265	}
 266
 267	seq = 0;
 268
 269	for (i = 0; i < ca->sb.njournal_buckets; i++)
 270		if (ja->seq[i] > seq) {
 271			seq = ja->seq[i];
 272			/*
 273			 * When journal_reclaim() goes to allocate for
 274			 * the first time, it'll use the bucket after
 275			 * ja->cur_idx
 276			 */
 277			ja->cur_idx = i;
 278			ja->last_idx = ja->discard_idx = (i + 1) %
 279				ca->sb.njournal_buckets;
 280
 281		}
 282
 283out:
 284	if (!list_empty(list))
 285		c->journal.seq = list_entry(list->prev,
 286					    struct journal_replay,
 287					    list)->j.seq;
 288
 289	return 0;
 290#undef read_bucket
 291}
 292
 293void bch_journal_mark(struct cache_set *c, struct list_head *list)
 294{
 295	atomic_t p = { 0 };
 296	struct bkey *k;
 297	struct journal_replay *i;
 298	struct journal *j = &c->journal;
 299	uint64_t last = j->seq;
 300
 301	/*
 302	 * journal.pin should never fill up - we never write a journal
 303	 * entry when it would fill up. But if for some reason it does, we
 304	 * iterate over the list in reverse order so that we can just skip that
 305	 * refcount instead of bugging.
 306	 */
 307
 308	list_for_each_entry_reverse(i, list, list) {
 309		BUG_ON(last < i->j.seq);
 310		i->pin = NULL;
 311
 312		while (last-- != i->j.seq)
 313			if (fifo_free(&j->pin) > 1) {
 314				fifo_push_front(&j->pin, p);
 315				atomic_set(&fifo_front(&j->pin), 0);
 316			}
 317
 318		if (fifo_free(&j->pin) > 1) {
 319			fifo_push_front(&j->pin, p);
 320			i->pin = &fifo_front(&j->pin);
 321			atomic_set(i->pin, 1);
 322		}
 323
 324		for (k = i->j.start;
 325		     k < bset_bkey_last(&i->j);
 326		     k = bkey_next(k))
 327			if (!__bch_extent_invalid(c, k)) {
 328				unsigned int j;
 329
 330				for (j = 0; j < KEY_PTRS(k); j++)
 331					if (ptr_available(c, k, j))
 332						atomic_inc(&PTR_BUCKET(c, k, j)->pin);
 333
 334				bch_initial_mark_key(c, 0, k);
 335			}
 336	}
 337}
 338
 339static bool is_discard_enabled(struct cache_set *s)
 340{
 341	struct cache *ca = s->cache;
 342
 343	if (ca->discard)
 344		return true;
 345
 346	return false;
 347}
 348
 349int bch_journal_replay(struct cache_set *s, struct list_head *list)
 350{
 351	int ret = 0, keys = 0, entries = 0;
 352	struct bkey *k;
 353	struct journal_replay *i =
 354		list_entry(list->prev, struct journal_replay, list);
 355
 356	uint64_t start = i->j.last_seq, end = i->j.seq, n = start;
 357	struct keylist keylist;
 358
 359	list_for_each_entry(i, list, list) {
 360		BUG_ON(i->pin && atomic_read(i->pin) != 1);
 361
 362		if (n != i->j.seq) {
 363			if (n == start && is_discard_enabled(s))
 364				pr_info("journal entries %llu-%llu may be discarded! (replaying %llu-%llu)\n",
 365					n, i->j.seq - 1, start, end);
 366			else {
 367				pr_err("journal entries %llu-%llu missing! (replaying %llu-%llu)\n",
 368					n, i->j.seq - 1, start, end);
 369				ret = -EIO;
 370				goto err;
 371			}
 372		}
 373
 374		for (k = i->j.start;
 375		     k < bset_bkey_last(&i->j);
 376		     k = bkey_next(k)) {
 377			trace_bcache_journal_replay_key(k);
 378
 379			bch_keylist_init_single(&keylist, k);
 380
 381			ret = bch_btree_insert(s, &keylist, i->pin, NULL);
 382			if (ret)
 383				goto err;
 384
 385			BUG_ON(!bch_keylist_empty(&keylist));
 386			keys++;
 387
 388			cond_resched();
 389		}
 390
 391		if (i->pin)
 392			atomic_dec(i->pin);
 393		n = i->j.seq + 1;
 394		entries++;
 395	}
 396
 397	pr_info("journal replay done, %i keys in %i entries, seq %llu\n",
 398		keys, entries, end);
 399err:
 400	while (!list_empty(list)) {
 401		i = list_first_entry(list, struct journal_replay, list);
 402		list_del(&i->list);
 403		kfree(i);
 404	}
 405
 406	return ret;
 407}
 408
 409void bch_journal_space_reserve(struct journal *j)
 410{
 411	j->do_reserve = true;
 412}
 413
 414/* Journalling */
 415
 416static void btree_flush_write(struct cache_set *c)
 417{
 418	struct btree *b, *t, *btree_nodes[BTREE_FLUSH_NR];
 419	unsigned int i, nr;
 420	int ref_nr;
 421	atomic_t *fifo_front_p, *now_fifo_front_p;
 422	size_t mask;
 423
 424	if (c->journal.btree_flushing)
 425		return;
 426
 427	spin_lock(&c->journal.flush_write_lock);
 428	if (c->journal.btree_flushing) {
 429		spin_unlock(&c->journal.flush_write_lock);
 430		return;
 431	}
 432	c->journal.btree_flushing = true;
 433	spin_unlock(&c->journal.flush_write_lock);
 434
 435	/* get the oldest journal entry and check its refcount */
 436	spin_lock(&c->journal.lock);
 437	fifo_front_p = &fifo_front(&c->journal.pin);
 438	ref_nr = atomic_read(fifo_front_p);
 439	if (ref_nr <= 0) {
 440		/*
 441		 * do nothing if no btree node references
 442		 * the oldest journal entry
 443		 */
 444		spin_unlock(&c->journal.lock);
 445		goto out;
 446	}
 447	spin_unlock(&c->journal.lock);
 448
 449	mask = c->journal.pin.mask;
 450	nr = 0;
 451	atomic_long_inc(&c->flush_write);
 452	memset(btree_nodes, 0, sizeof(btree_nodes));
 453
 454	mutex_lock(&c->bucket_lock);
 455	list_for_each_entry_safe_reverse(b, t, &c->btree_cache, list) {
 456		/*
 457		 * It is safe to get now_fifo_front_p without holding
 458		 * c->journal.lock here, because we don't need to know
 459		 * the exactly accurate value, just check whether the
 460		 * front pointer of c->journal.pin is changed.
 461		 */
 462		now_fifo_front_p = &fifo_front(&c->journal.pin);
 463		/*
 464		 * If the oldest journal entry is reclaimed and front
 465		 * pointer of c->journal.pin changes, it is unnecessary
 466		 * to scan c->btree_cache anymore, just quit the loop and
 467		 * flush out what we have already.
 468		 */
 469		if (now_fifo_front_p != fifo_front_p)
 470			break;
 471		/*
 472		 * quit this loop if all matching btree nodes are
 473		 * scanned and record in btree_nodes[] already.
 474		 */
 475		ref_nr = atomic_read(fifo_front_p);
 476		if (nr >= ref_nr)
 477			break;
 478
 479		if (btree_node_journal_flush(b))
 480			pr_err("BUG: flush_write bit should not be set here!\n");
 481
 482		mutex_lock(&b->write_lock);
 483
 484		if (!btree_node_dirty(b)) {
 485			mutex_unlock(&b->write_lock);
 486			continue;
 487		}
 488
 489		if (!btree_current_write(b)->journal) {
 490			mutex_unlock(&b->write_lock);
 491			continue;
 492		}
 493
 494		/*
 495		 * Only select the btree node which exactly references
 496		 * the oldest journal entry.
 497		 *
 498		 * If the journal entry pointed by fifo_front_p is
 499		 * reclaimed in parallel, don't worry:
 500		 * - the list_for_each_xxx loop will quit when checking
 501		 *   next now_fifo_front_p.
 502		 * - If there are matched nodes recorded in btree_nodes[],
 503		 *   they are clean now (this is why and how the oldest
 504		 *   journal entry can be reclaimed). These selected nodes
 505		 *   will be ignored and skipped in the following for-loop.
 506		 */
 507		if (((btree_current_write(b)->journal - fifo_front_p) &
 508		     mask) != 0) {
 509			mutex_unlock(&b->write_lock);
 510			continue;
 511		}
 512
 513		set_btree_node_journal_flush(b);
 514
 515		mutex_unlock(&b->write_lock);
 516
 517		btree_nodes[nr++] = b;
 518		/*
 519		 * To avoid holding c->bucket_lock too long time,
 520		 * only scan for BTREE_FLUSH_NR matched btree nodes
 521		 * at most. If there are more btree nodes reference
 522		 * the oldest journal entry, try to flush them next
 523		 * time when btree_flush_write() is called.
 524		 */
 525		if (nr == BTREE_FLUSH_NR)
 526			break;
 527	}
 528	mutex_unlock(&c->bucket_lock);
 529
 530	for (i = 0; i < nr; i++) {
 531		b = btree_nodes[i];
 532		if (!b) {
 533			pr_err("BUG: btree_nodes[%d] is NULL\n", i);
 534			continue;
 535		}
 536
 537		/* safe to check without holding b->write_lock */
 538		if (!btree_node_journal_flush(b)) {
 539			pr_err("BUG: bnode %p: journal_flush bit cleaned\n", b);
 540			continue;
 541		}
 542
 543		mutex_lock(&b->write_lock);
 544		if (!btree_current_write(b)->journal) {
 545			clear_bit(BTREE_NODE_journal_flush, &b->flags);
 546			mutex_unlock(&b->write_lock);
 547			pr_debug("bnode %p: written by others\n", b);
 548			continue;
 549		}
 550
 551		if (!btree_node_dirty(b)) {
 552			clear_bit(BTREE_NODE_journal_flush, &b->flags);
 553			mutex_unlock(&b->write_lock);
 554			pr_debug("bnode %p: dirty bit cleaned by others\n", b);
 555			continue;
 556		}
 557
 558		__bch_btree_node_write(b, NULL);
 559		clear_bit(BTREE_NODE_journal_flush, &b->flags);
 560		mutex_unlock(&b->write_lock);
 561	}
 562
 563out:
 564	spin_lock(&c->journal.flush_write_lock);
 565	c->journal.btree_flushing = false;
 566	spin_unlock(&c->journal.flush_write_lock);
 567}
 568
 569#define last_seq(j)	((j)->seq - fifo_used(&(j)->pin) + 1)
 570
 571static void journal_discard_endio(struct bio *bio)
 572{
 573	struct journal_device *ja =
 574		container_of(bio, struct journal_device, discard_bio);
 575	struct cache *ca = container_of(ja, struct cache, journal);
 576
 577	atomic_set(&ja->discard_in_flight, DISCARD_DONE);
 578
 579	closure_wake_up(&ca->set->journal.wait);
 580	closure_put(&ca->set->cl);
 581}
 582
 583static void journal_discard_work(struct work_struct *work)
 584{
 585	struct journal_device *ja =
 586		container_of(work, struct journal_device, discard_work);
 587
 588	submit_bio(&ja->discard_bio);
 589}
 590
 591static void do_journal_discard(struct cache *ca)
 592{
 593	struct journal_device *ja = &ca->journal;
 594	struct bio *bio = &ja->discard_bio;
 595
 596	if (!ca->discard) {
 597		ja->discard_idx = ja->last_idx;
 598		return;
 599	}
 600
 601	switch (atomic_read(&ja->discard_in_flight)) {
 602	case DISCARD_IN_FLIGHT:
 603		return;
 604
 605	case DISCARD_DONE:
 606		ja->discard_idx = (ja->discard_idx + 1) %
 607			ca->sb.njournal_buckets;
 608
 609		atomic_set(&ja->discard_in_flight, DISCARD_READY);
 610		fallthrough;
 611
 612	case DISCARD_READY:
 613		if (ja->discard_idx == ja->last_idx)
 614			return;
 615
 616		atomic_set(&ja->discard_in_flight, DISCARD_IN_FLIGHT);
 617
 618		bio_init(bio, ca->bdev, bio->bi_inline_vecs, 1, REQ_OP_DISCARD);
 619		bio->bi_iter.bi_sector	= bucket_to_sector(ca->set,
 620						ca->sb.d[ja->discard_idx]);
 621		bio->bi_iter.bi_size	= bucket_bytes(ca);
 622		bio->bi_end_io		= journal_discard_endio;
 623
 624		closure_get(&ca->set->cl);
 625		INIT_WORK(&ja->discard_work, journal_discard_work);
 626		queue_work(bch_journal_wq, &ja->discard_work);
 627	}
 628}
 629
 630static unsigned int free_journal_buckets(struct cache_set *c)
 631{
 632	struct journal *j = &c->journal;
 633	struct cache *ca = c->cache;
 634	struct journal_device *ja = &c->cache->journal;
 635	unsigned int n;
 636
 637	/* In case njournal_buckets is not power of 2 */
 638	if (ja->cur_idx >= ja->discard_idx)
 639		n = ca->sb.njournal_buckets +  ja->discard_idx - ja->cur_idx;
 640	else
 641		n = ja->discard_idx - ja->cur_idx;
 642
 643	if (n > (1 + j->do_reserve))
 644		return n - (1 + j->do_reserve);
 645
 646	return 0;
 647}
 648
 649static void journal_reclaim(struct cache_set *c)
 650{
 651	struct bkey *k = &c->journal.key;
 652	struct cache *ca = c->cache;
 653	uint64_t last_seq;
 654	struct journal_device *ja = &ca->journal;
 655	atomic_t p __maybe_unused;
 656
 657	atomic_long_inc(&c->reclaim);
 658
 659	while (!atomic_read(&fifo_front(&c->journal.pin)))
 660		fifo_pop(&c->journal.pin, p);
 661
 662	last_seq = last_seq(&c->journal);
 663
 664	/* Update last_idx */
 665
 666	while (ja->last_idx != ja->cur_idx &&
 667	       ja->seq[ja->last_idx] < last_seq)
 668		ja->last_idx = (ja->last_idx + 1) %
 669			ca->sb.njournal_buckets;
 670
 671	do_journal_discard(ca);
 672
 673	if (c->journal.blocks_free)
 674		goto out;
 675
 676	if (!free_journal_buckets(c))
 677		goto out;
 678
 679	ja->cur_idx = (ja->cur_idx + 1) % ca->sb.njournal_buckets;
 680	k->ptr[0] = MAKE_PTR(0,
 681			     bucket_to_sector(c, ca->sb.d[ja->cur_idx]),
 682			     ca->sb.nr_this_dev);
 683	atomic_long_inc(&c->reclaimed_journal_buckets);
 684
 685	bkey_init(k);
 686	SET_KEY_PTRS(k, 1);
 687	c->journal.blocks_free = ca->sb.bucket_size >> c->block_bits;
 688
 689out:
 690	if (!journal_full(&c->journal))
 691		__closure_wake_up(&c->journal.wait);
 692}
 693
 694void bch_journal_next(struct journal *j)
 695{
 696	atomic_t p = { 1 };
 697
 698	j->cur = (j->cur == j->w)
 699		? &j->w[1]
 700		: &j->w[0];
 701
 702	/*
 703	 * The fifo_push() needs to happen at the same time as j->seq is
 704	 * incremented for last_seq() to be calculated correctly
 705	 */
 706	BUG_ON(!fifo_push(&j->pin, p));
 707	atomic_set(&fifo_back(&j->pin), 1);
 708
 709	j->cur->data->seq	= ++j->seq;
 710	j->cur->dirty		= false;
 711	j->cur->need_write	= false;
 712	j->cur->data->keys	= 0;
 713
 714	if (fifo_full(&j->pin))
 715		pr_debug("journal_pin full (%zu)\n", fifo_used(&j->pin));
 716}
 717
 718static void journal_write_endio(struct bio *bio)
 719{
 720	struct journal_write *w = bio->bi_private;
 721
 722	cache_set_err_on(bio->bi_status, w->c, "journal io error");
 723	closure_put(&w->c->journal.io);
 724}
 725
 726static void journal_write(struct closure *cl);
 727
 728static void journal_write_done(struct closure *cl)
 729{
 730	struct journal *j = container_of(cl, struct journal, io);
 731	struct journal_write *w = (j->cur == j->w)
 732		? &j->w[1]
 733		: &j->w[0];
 734
 735	__closure_wake_up(&w->wait);
 736	continue_at_nobarrier(cl, journal_write, bch_journal_wq);
 737}
 738
 739static void journal_write_unlock(struct closure *cl)
 740	__releases(&c->journal.lock)
 741{
 742	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
 743
 744	c->journal.io_in_flight = 0;
 745	spin_unlock(&c->journal.lock);
 746}
 747
 748static void journal_write_unlocked(struct closure *cl)
 749	__releases(c->journal.lock)
 750{
 751	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
 752	struct cache *ca = c->cache;
 753	struct journal_write *w = c->journal.cur;
 754	struct bkey *k = &c->journal.key;
 755	unsigned int i, sectors = set_blocks(w->data, block_bytes(ca)) *
 756		ca->sb.block_size;
 757
 758	struct bio *bio;
 759	struct bio_list list;
 760
 761	bio_list_init(&list);
 762
 763	if (!w->need_write) {
 764		closure_return_with_destructor(cl, journal_write_unlock);
 765		return;
 766	} else if (journal_full(&c->journal)) {
 767		journal_reclaim(c);
 768		spin_unlock(&c->journal.lock);
 769
 770		btree_flush_write(c);
 771		continue_at(cl, journal_write, bch_journal_wq);
 772		return;
 773	}
 774
 775	c->journal.blocks_free -= set_blocks(w->data, block_bytes(ca));
 776
 777	w->data->btree_level = c->root->level;
 778
 779	bkey_copy(&w->data->btree_root, &c->root->key);
 780	bkey_copy(&w->data->uuid_bucket, &c->uuid_bucket);
 781
 782	w->data->prio_bucket[ca->sb.nr_this_dev] = ca->prio_buckets[0];
 783	w->data->magic		= jset_magic(&ca->sb);
 784	w->data->version	= BCACHE_JSET_VERSION;
 785	w->data->last_seq	= last_seq(&c->journal);
 786	w->data->csum		= csum_set(w->data);
 787
 788	for (i = 0; i < KEY_PTRS(k); i++) {
 789		ca = c->cache;
 790		bio = &ca->journal.bio;
 791
 792		atomic_long_add(sectors, &ca->meta_sectors_written);
 793
 794		bio_reset(bio, ca->bdev, REQ_OP_WRITE | 
 795			  REQ_SYNC | REQ_META | REQ_PREFLUSH | REQ_FUA);
 796		bio->bi_iter.bi_sector	= PTR_OFFSET(k, i);
 797		bio->bi_iter.bi_size = sectors << 9;
 798
 799		bio->bi_end_io	= journal_write_endio;
 800		bio->bi_private = w;
 801		bch_bio_map(bio, w->data);
 802
 803		trace_bcache_journal_write(bio, w->data->keys);
 804		bio_list_add(&list, bio);
 805
 806		SET_PTR_OFFSET(k, i, PTR_OFFSET(k, i) + sectors);
 807
 808		ca->journal.seq[ca->journal.cur_idx] = w->data->seq;
 809	}
 810
 811	/* If KEY_PTRS(k) == 0, this jset gets lost in air */
 812	BUG_ON(i == 0);
 813
 814	atomic_dec_bug(&fifo_back(&c->journal.pin));
 815	bch_journal_next(&c->journal);
 816	journal_reclaim(c);
 817
 818	spin_unlock(&c->journal.lock);
 819
 820	while ((bio = bio_list_pop(&list)))
 821		closure_bio_submit(c, bio, cl);
 822
 823	continue_at(cl, journal_write_done, NULL);
 824}
 825
 826static void journal_write(struct closure *cl)
 827{
 828	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
 829
 830	spin_lock(&c->journal.lock);
 831	journal_write_unlocked(cl);
 832}
 833
 834static void journal_try_write(struct cache_set *c)
 835	__releases(c->journal.lock)
 836{
 837	struct closure *cl = &c->journal.io;
 838	struct journal_write *w = c->journal.cur;
 839
 840	w->need_write = true;
 841
 842	if (!c->journal.io_in_flight) {
 843		c->journal.io_in_flight = 1;
 844		closure_call(cl, journal_write_unlocked, NULL, &c->cl);
 845	} else {
 846		spin_unlock(&c->journal.lock);
 847	}
 848}
 849
 850static struct journal_write *journal_wait_for_write(struct cache_set *c,
 851						    unsigned int nkeys)
 852	__acquires(&c->journal.lock)
 853{
 854	size_t sectors;
 855	struct closure cl;
 856	bool wait = false;
 857	struct cache *ca = c->cache;
 858
 859	closure_init_stack(&cl);
 860
 861	spin_lock(&c->journal.lock);
 862
 863	while (1) {
 864		struct journal_write *w = c->journal.cur;
 865
 866		sectors = __set_blocks(w->data, w->data->keys + nkeys,
 867				       block_bytes(ca)) * ca->sb.block_size;
 868
 869		if (sectors <= min_t(size_t,
 870				     c->journal.blocks_free * ca->sb.block_size,
 871				     PAGE_SECTORS << JSET_BITS))
 872			return w;
 873
 874		if (wait)
 875			closure_wait(&c->journal.wait, &cl);
 876
 877		if (!journal_full(&c->journal)) {
 878			if (wait)
 879				trace_bcache_journal_entry_full(c);
 880
 881			/*
 882			 * XXX: If we were inserting so many keys that they
 883			 * won't fit in an _empty_ journal write, we'll
 884			 * deadlock. For now, handle this in
 885			 * bch_keylist_realloc() - but something to think about.
 886			 */
 887			BUG_ON(!w->data->keys);
 888
 889			journal_try_write(c); /* unlocks */
 890		} else {
 891			if (wait)
 892				trace_bcache_journal_full(c);
 893
 894			journal_reclaim(c);
 895			spin_unlock(&c->journal.lock);
 896
 897			btree_flush_write(c);
 898		}
 899
 900		closure_sync(&cl);
 901		spin_lock(&c->journal.lock);
 902		wait = true;
 903	}
 904}
 905
 906static void journal_write_work(struct work_struct *work)
 907{
 908	struct cache_set *c = container_of(to_delayed_work(work),
 909					   struct cache_set,
 910					   journal.work);
 911	spin_lock(&c->journal.lock);
 912	if (c->journal.cur->dirty)
 913		journal_try_write(c);
 914	else
 915		spin_unlock(&c->journal.lock);
 916}
 917
 918/*
 919 * Entry point to the journalling code - bio_insert() and btree_invalidate()
 920 * pass bch_journal() a list of keys to be journalled, and then
 921 * bch_journal() hands those same keys off to btree_insert_async()
 922 */
 923
 924atomic_t *bch_journal(struct cache_set *c,
 925		      struct keylist *keys,
 926		      struct closure *parent)
 927{
 928	struct journal_write *w;
 929	atomic_t *ret;
 930
 931	/* No journaling if CACHE_SET_IO_DISABLE set already */
 932	if (unlikely(test_bit(CACHE_SET_IO_DISABLE, &c->flags)))
 933		return NULL;
 934
 935	if (!CACHE_SYNC(&c->cache->sb))
 936		return NULL;
 937
 938	w = journal_wait_for_write(c, bch_keylist_nkeys(keys));
 939
 940	memcpy(bset_bkey_last(w->data), keys->keys, bch_keylist_bytes(keys));
 941	w->data->keys += bch_keylist_nkeys(keys);
 942
 943	ret = &fifo_back(&c->journal.pin);
 944	atomic_inc(ret);
 945
 946	if (parent) {
 947		closure_wait(&w->wait, parent);
 948		journal_try_write(c);
 949	} else if (!w->dirty) {
 950		w->dirty = true;
 951		queue_delayed_work(bch_flush_wq, &c->journal.work,
 952				   msecs_to_jiffies(c->journal_delay_ms));
 953		spin_unlock(&c->journal.lock);
 954	} else {
 955		spin_unlock(&c->journal.lock);
 956	}
 957
 958
 959	return ret;
 960}
 961
 962void bch_journal_meta(struct cache_set *c, struct closure *cl)
 963{
 964	struct keylist keys;
 965	atomic_t *ref;
 966
 967	bch_keylist_init(&keys);
 968
 969	ref = bch_journal(c, &keys, cl);
 970	if (ref)
 971		atomic_dec_bug(ref);
 972}
 973
 974void bch_journal_free(struct cache_set *c)
 975{
 976	free_pages((unsigned long) c->journal.w[1].data, JSET_BITS);
 977	free_pages((unsigned long) c->journal.w[0].data, JSET_BITS);
 978	free_fifo(&c->journal.pin);
 979}
 980
 981int bch_journal_alloc(struct cache_set *c)
 982{
 983	struct journal *j = &c->journal;
 984
 985	spin_lock_init(&j->lock);
 986	spin_lock_init(&j->flush_write_lock);
 987	INIT_DELAYED_WORK(&j->work, journal_write_work);
 988
 989	c->journal_delay_ms = 100;
 990
 991	j->w[0].c = c;
 992	j->w[1].c = c;
 993
 994	if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)) ||
 995	    !(j->w[0].data = (void *) __get_free_pages(GFP_KERNEL|__GFP_COMP, JSET_BITS)) ||
 996	    !(j->w[1].data = (void *) __get_free_pages(GFP_KERNEL|__GFP_COMP, JSET_BITS)))
 997		return -ENOMEM;
 998
 999	return 0;
1000}