1 files changed, 116 insertions, 124 deletions

diff --git a/drivers/md/bcache/alloc.c b/drivers/md/bcache/alloc.c
index 4c9852d92b0..443d03fbac4 100644
--- a/drivers/md/bcache/alloc.c
+++ b/drivers/md/bcache/alloc.c
@@ -78,12 +78,6 @@ uint8_t bch_inc_gen(struct cache *ca, struct bucket *b)
 	ca->set->need_gc = max(ca->set->need_gc, bucket_gc_gen(b));
 	WARN_ON_ONCE(ca->set->need_gc > BUCKET_GC_GEN_MAX);
 
-	if (CACHE_SYNC(&ca->set->sb)) {
-		ca->need_save_prio = max(ca->need_save_prio,
-					 bucket_disk_gen(b));
-		WARN_ON_ONCE(ca->need_save_prio > BUCKET_DISK_GEN_MAX);
-	}
-
 	return ret;
 }
 
@@ -120,50 +114,63 @@ void bch_rescale_priorities(struct cache_set *c, int sectors)
 	mutex_unlock(&c->bucket_lock);
 }
 
-/* Allocation */
+/*
+ * Background allocation thread: scans for buckets to be invalidated,
+ * invalidates them, rewrites prios/gens (marking them as invalidated on disk),
+ * then optionally issues discard commands to the newly free buckets, then puts
+ * them on the various freelists.
+ */
 
 static inline bool can_inc_bucket_gen(struct bucket *b)
 {
-	return bucket_gc_gen(b) < BUCKET_GC_GEN_MAX &&
-		bucket_disk_gen(b) < BUCKET_DISK_GEN_MAX;
+	return bucket_gc_gen(b) < BUCKET_GC_GEN_MAX;
 }
 
-bool bch_bucket_add_unused(struct cache *ca, struct bucket *b)
+bool bch_can_invalidate_bucket(struct cache *ca, struct bucket *b)
 {
-	BUG_ON(GC_MARK(b) || GC_SECTORS_USED(b));
-
-	if (fifo_used(&ca->free) > ca->watermark[WATERMARK_MOVINGGC] &&
-	    CACHE_REPLACEMENT(&ca->sb) == CACHE_REPLACEMENT_FIFO)
-		return false;
-
-	b->prio = 0;
-
-	if (can_inc_bucket_gen(b) &&
-	    fifo_push(&ca->unused, b - ca->buckets)) {
-		atomic_inc(&b->pin);
-		return true;
-	}
+	BUG_ON(!ca->set->gc_mark_valid);
 
-	return false;
-}
-
-static bool can_invalidate_bucket(struct cache *ca, struct bucket *b)
-{
-	return GC_MARK(b) == GC_MARK_RECLAIMABLE &&
+	return (!GC_MARK(b) ||
+		GC_MARK(b) == GC_MARK_RECLAIMABLE) &&
 		!atomic_read(&b->pin) &&
 		can_inc_bucket_gen(b);
 }
 
-static void invalidate_one_bucket(struct cache *ca, struct bucket *b)
+void __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
 {
+	lockdep_assert_held(&ca->set->bucket_lock);
+	BUG_ON(GC_MARK(b) && GC_MARK(b) != GC_MARK_RECLAIMABLE);
+
+	if (GC_SECTORS_USED(b))
+		trace_bcache_invalidate(ca, b - ca->buckets);
+
 	bch_inc_gen(ca, b);
 	b->prio = INITIAL_PRIO;
 	atomic_inc(&b->pin);
+}
+
+static void bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
+{
+	__bch_invalidate_one_bucket(ca, b);
+
 	fifo_push(&ca->free_inc, b - ca->buckets);
 }
 
-#define bucket_prio(b)				\
-	(((unsigned) (b->prio - ca->set->min_prio)) * GC_SECTORS_USED(b))
+/*
+ * Determines what order we're going to reuse buckets, smallest bucket_prio()
+ * first: we also take into account the number of sectors of live data in that
+ * bucket, and in order for that multiply to make sense we have to scale bucket
+ *
+ * Thus, we scale the bucket priorities so that the bucket with the smallest
+ * prio is worth 1/8th of what INITIAL_PRIO is worth.
+ */
+
+#define bucket_prio(b)							\
+({									\
+	unsigned min_prio = (INITIAL_PRIO - ca->set->min_prio) / 8;	\
+									\
+	(b->prio - ca->set->min_prio + min_prio) * GC_SECTORS_USED(b);	\
+})
 
 #define bucket_max_cmp(l, r)	(bucket_prio(l) < bucket_prio(r))
 #define bucket_min_cmp(l, r)	(bucket_prio(l) > bucket_prio(r))
@@ -176,20 +183,7 @@ static void invalidate_buckets_lru(struct cache *ca)
 	ca->heap.used = 0;
 
 	for_each_bucket(b, ca) {
-		/*
-		 * If we fill up the unused list, if we then return before
-		 * adding anything to the free_inc list we'll skip writing
-		 * prios/gens and just go back to allocating from the unused
-		 * list:
-		 */
-		if (fifo_full(&ca->unused))
-			return;
-
-		if (!can_invalidate_bucket(ca, b))
-			continue;
-
-		if (!GC_SECTORS_USED(b) &&
-		    bch_bucket_add_unused(ca, b))
+		if (!bch_can_invalidate_bucket(ca, b))
 			continue;
 
 		if (!heap_full(&ca->heap))
@@ -214,7 +208,7 @@ static void invalidate_buckets_lru(struct cache *ca)
 			return;
 		}
 
-		invalidate_one_bucket(ca, b);
+		bch_invalidate_one_bucket(ca, b);
 	}
 }
 
@@ -230,8 +224,8 @@ static void invalidate_buckets_fifo(struct cache *ca)
 
 		b = ca->buckets + ca->fifo_last_bucket++;
 
-		if (can_invalidate_bucket(ca, b))
-			invalidate_one_bucket(ca, b);
+		if (bch_can_invalidate_bucket(ca, b))
+			bch_invalidate_one_bucket(ca, b);
 
 		if (++checked >= ca->sb.nbuckets) {
 			ca->invalidate_needs_gc = 1;
@@ -255,8 +249,8 @@ static void invalidate_buckets_random(struct cache *ca)
 
 		b = ca->buckets + n;
 
-		if (can_invalidate_bucket(ca, b))
-			invalidate_one_bucket(ca, b);
+		if (bch_can_invalidate_bucket(ca, b))
+			bch_invalidate_one_bucket(ca, b);
 
 		if (++checked >= ca->sb.nbuckets / 2) {
 			ca->invalidate_needs_gc = 1;
@@ -268,8 +262,7 @@ static void invalidate_buckets_random(struct cache *ca)
 
 static void invalidate_buckets(struct cache *ca)
 {
-	if (ca->invalidate_needs_gc)
-		return;
+	BUG_ON(ca->invalidate_needs_gc);
 
 	switch (CACHE_REPLACEMENT(&ca->sb)) {
 	case CACHE_REPLACEMENT_LRU:
@@ -282,8 +275,6 @@ static void invalidate_buckets(struct cache *ca)
 		invalidate_buckets_random(ca);
 		break;
 	}
-
-	trace_bcache_alloc_invalidate(ca);
 }
 
 #define allocator_wait(ca, cond)					\
@@ -304,6 +295,21 @@ do {									\
 	__set_current_state(TASK_RUNNING);				\
 } while (0)
 
+static int bch_allocator_push(struct cache *ca, long bucket)
+{
+	unsigned i;
+
+	/* Prios/gens are actually the most important reserve */
+	if (fifo_push(&ca->free[RESERVE_PRIO], bucket))
+		return true;
+
+	for (i = 0; i < RESERVE_NR; i++)
+		if (fifo_push(&ca->free[i], bucket))
+			return true;
+
+	return false;
+}
+
 static int bch_allocator_thread(void *arg)
 {
 	struct cache *ca = arg;
@@ -316,17 +322,10 @@ static int bch_allocator_thread(void *arg)
 		 * possibly issue discards to them, then we add the bucket to
 		 * the free list:
 		 */
-		while (1) {
+		while (!fifo_empty(&ca->free_inc)) {
 			long bucket;
 
-			if ((!atomic_read(&ca->set->prio_blocked) ||
-			     !CACHE_SYNC(&ca->set->sb)) &&
-			    !fifo_empty(&ca->unused))
-				fifo_pop(&ca->unused, bucket);
-			else if (!fifo_empty(&ca->free_inc))
-				fifo_pop(&ca->free_inc, bucket);
-			else
-				break;
+			fifo_pop(&ca->free_inc, bucket);
 
 			if (ca->discard) {
 				mutex_unlock(&ca->set->bucket_lock);
@@ -336,9 +335,8 @@ static int bch_allocator_thread(void *arg)
 				mutex_lock(&ca->set->bucket_lock);
 			}
 
-			allocator_wait(ca, !fifo_full(&ca->free));
-
-			fifo_push(&ca->free, bucket);
+			allocator_wait(ca, bch_allocator_push(ca, bucket));
+			wake_up(&ca->set->btree_cache_wait);
 			wake_up(&ca->set->bucket_wait);
 		}
 
@@ -348,9 +346,9 @@ static int bch_allocator_thread(void *arg)
 		 * them to the free_inc list:
 		 */
 
+retry_invalidate:
 		allocator_wait(ca, ca->set->gc_mark_valid &&
-			       (ca->need_save_prio > 64 ||
-				!ca->invalidate_needs_gc));
+			       !ca->invalidate_needs_gc);
 		invalidate_buckets(ca);
 
 		/*
@@ -358,59 +356,73 @@ static int bch_allocator_thread(void *arg)
 		 * new stuff to them:
 		 */
 		allocator_wait(ca, !atomic_read(&ca->set->prio_blocked));
-		if (CACHE_SYNC(&ca->set->sb) &&
-		    (!fifo_empty(&ca->free_inc) ||
-		     ca->need_save_prio > 64))
+		if (CACHE_SYNC(&ca->set->sb)) {
+			/*
+			 * This could deadlock if an allocation with a btree
+			 * node locked ever blocked - having the btree node
+			 * locked would block garbage collection, but here we're
+			 * waiting on garbage collection before we invalidate
+			 * and free anything.
+			 *
+			 * But this should be safe since the btree code always
+			 * uses btree_check_reserve() before allocating now, and
+			 * if it fails it blocks without btree nodes locked.
+			 */
+			if (!fifo_full(&ca->free_inc))
+				goto retry_invalidate;
+
 			bch_prio_write(ca);
+		}
 	}
 }
 
-long bch_bucket_alloc(struct cache *ca, unsigned watermark, bool wait)
+/* Allocation */
+
+long bch_bucket_alloc(struct cache *ca, unsigned reserve, bool wait)
 {
 	DEFINE_WAIT(w);
 	struct bucket *b;
 	long r;
 
 	/* fastpath */
-	if (fifo_used(&ca->free) > ca->watermark[watermark]) {
-		fifo_pop(&ca->free, r);
+	if (fifo_pop(&ca->free[RESERVE_NONE], r) ||
+	    fifo_pop(&ca->free[reserve], r))
 		goto out;
-	}
 
-	if (!wait)
+	if (!wait) {
+		trace_bcache_alloc_fail(ca, reserve);
 		return -1;
+	}
 
-	while (1) {
-		if (fifo_used(&ca->free) > ca->watermark[watermark]) {
-			fifo_pop(&ca->free, r);
-			break;
-		}
-
+	do {
 		prepare_to_wait(&ca->set->bucket_wait, &w,
 				TASK_UNINTERRUPTIBLE);
 
 		mutex_unlock(&ca->set->bucket_lock);
 		schedule();
 		mutex_lock(&ca->set->bucket_lock);
-	}
+	} while (!fifo_pop(&ca->free[RESERVE_NONE], r) &&
+		 !fifo_pop(&ca->free[reserve], r));
 
 	finish_wait(&ca->set->bucket_wait, &w);
 out:
 	wake_up_process(ca->alloc_thread);
 
+	trace_bcache_alloc(ca, reserve);
+
 	if (expensive_debug_checks(ca->set)) {
 		size_t iter;
 		long i;
+		unsigned j;
 
 		for (iter = 0; iter < prio_buckets(ca) * 2; iter++)
 			BUG_ON(ca->prio_buckets[iter] == (uint64_t) r);
 
-		fifo_for_each(i, &ca->free, iter)
-			BUG_ON(i == r);
+		for (j = 0; j < RESERVE_NR; j++)
+			fifo_for_each(i, &ca->free[j], iter)
+				BUG_ON(i == r);
 		fifo_for_each(i, &ca->free_inc, iter)
 			BUG_ON(i == r);
-		fifo_for_each(i, &ca->unused, iter)
-			BUG_ON(i == r);
 	}
 
 	b = ca->buckets + r;
@@ -419,7 +431,7 @@ out:
 
 	SET_GC_SECTORS_USED(b, ca->sb.bucket_size);
 
-	if (watermark <= WATERMARK_METADATA) {
+	if (reserve <= RESERVE_PRIO) {
 		SET_GC_MARK(b, GC_MARK_METADATA);
 		SET_GC_MOVE(b, 0);
 		b->prio = BTREE_PRIO;
@@ -432,20 +444,22 @@ out:
 	return r;
 }
 
+void __bch_bucket_free(struct cache *ca, struct bucket *b)
+{
+	SET_GC_MARK(b, 0);
+	SET_GC_SECTORS_USED(b, 0);
+}
+
 void bch_bucket_free(struct cache_set *c, struct bkey *k)
 {
 	unsigned i;
 
-	for (i = 0; i < KEY_PTRS(k); i++) {
-		struct bucket *b = PTR_BUCKET(c, k, i);
-
-		SET_GC_MARK(b, GC_MARK_RECLAIMABLE);
-		SET_GC_SECTORS_USED(b, 0);
-		bch_bucket_add_unused(PTR_CACHE(c, k, i), b);
-	}
+	for (i = 0; i < KEY_PTRS(k); i++)
+		__bch_bucket_free(PTR_CACHE(c, k, i),
+				  PTR_BUCKET(c, k, i));
 }
 
-int __bch_bucket_alloc_set(struct cache_set *c, unsigned watermark,
+int __bch_bucket_alloc_set(struct cache_set *c, unsigned reserve,
 			   struct bkey *k, int n, bool wait)
 {
 	int i;
@@ -459,7 +473,7 @@ int __bch_bucket_alloc_set(struct cache_set *c, unsigned watermark,
 
 	for (i = 0; i < n; i++) {
 		struct cache *ca = c->cache_by_alloc[i];
-		long b = bch_bucket_alloc(ca, watermark, wait);
+		long b = bch_bucket_alloc(ca, reserve, wait);
 
 		if (b == -1)
 			goto err;
@@ -478,12 +492,12 @@ err:
 	return -1;
 }
 
-int bch_bucket_alloc_set(struct cache_set *c, unsigned watermark,
+int bch_bucket_alloc_set(struct cache_set *c, unsigned reserve,
 			 struct bkey *k, int n, bool wait)
 {
 	int ret;
 	mutex_lock(&c->bucket_lock);
-	ret = __bch_bucket_alloc_set(c, watermark, k, n, wait);
+	ret = __bch_bucket_alloc_set(c, reserve, k, n, wait);
 	mutex_unlock(&c->bucket_lock);
 	return ret;
 }
@@ -573,8 +587,8 @@ bool bch_alloc_sectors(struct cache_set *c, struct bkey *k, unsigned sectors,
 
 	while (!(b = pick_data_bucket(c, k, write_point, &alloc.key))) {
 		unsigned watermark = write_prio
-			? WATERMARK_MOVINGGC
-			: WATERMARK_NONE;
+			? RESERVE_MOVINGGC
+			: RESERVE_NONE;
 
 		spin_unlock(&c->data_bucket_lock);
 
@@ -680,25 +694,3 @@ int bch_cache_allocator_start(struct cache *ca)
 	ca->alloc_thread = k;
 	return 0;
 }
-
-int bch_cache_allocator_init(struct cache *ca)
-{
-	/*
-	 * Reserve:
-	 * Prio/gen writes first
-	 * Then 8 for btree allocations
-	 * Then half for the moving garbage collector
-	 */
-
-	ca->watermark[WATERMARK_PRIO] = 0;
-
-	ca->watermark[WATERMARK_METADATA] = prio_buckets(ca);
-
-	ca->watermark[WATERMARK_MOVINGGC] = 8 +
-		ca->watermark[WATERMARK_METADATA];
-
-	ca->watermark[WATERMARK_NONE] = ca->free.size / 2 +
-		ca->watermark[WATERMARK_MOVINGGC];
-
-	return 0;
-}