1 files changed, 213 insertions, 72 deletions

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index fd3e175d83..534ba2b02b 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -29,6 +29,8 @@
 #include <linux/sched/task.h>
 #include <linux/sort.h>
 
+static void bch2_discard_one_bucket_fast(struct bch_fs *c, struct bpos bucket);
+
 /* Persistent alloc info: */
 
 static const unsigned BCH_ALLOC_V1_FIELD_BYTES[] = {
@@ -242,10 +244,10 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 	struct bkey_s_c_alloc_v4 a = bkey_s_c_to_alloc_v4(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(alloc_v4_u64s(a.v) > bkey_val_u64s(k.k), c, err,
+	bkey_fsck_err_on(alloc_v4_u64s_noerror(a.v) > bkey_val_u64s(k.k), c, err,
 			 alloc_v4_val_size_bad,
 			 "bad val size (%u > %zu)",
-			 alloc_v4_u64s(a.v), bkey_val_u64s(k.k));
+			 alloc_v4_u64s_noerror(a.v), bkey_val_u64s(k.k));
 
 	bkey_fsck_err_on(!BCH_ALLOC_V4_BACKPOINTERS_START(a.v) &&
 			 BCH_ALLOC_V4_NR_BACKPOINTERS(a.v), c, err,
@@ -530,13 +532,13 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 		u8 gen = bch2_alloc_to_v4(k, &a)->gen;
 		unsigned offset;
 		struct bpos pos = alloc_gens_pos(iter.pos, &offset);
+		int ret2 = 0;
 
 		if (have_bucket_gens_key && bkey_cmp(iter.pos, pos)) {
-			ret = commit_do(trans, NULL, NULL,
-					BCH_TRANS_COMMIT_no_enospc,
-				bch2_btree_insert_trans(trans, BTREE_ID_bucket_gens, &g.k_i, 0));
-			if (ret)
-				break;
+			ret2 =  bch2_btree_insert_trans(trans, BTREE_ID_bucket_gens, &g.k_i, 0) ?:
+				bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
+			if (ret2)
+				goto iter_err;
 			have_bucket_gens_key = false;
 		}
 
@@ -547,7 +549,8 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 		}
 
 		g.v.gens[offset] = gen;
-		0;
+iter_err:
+		ret2;
 	}));
 
 	if (have_bucket_gens_key && !ret)
@@ -850,7 +853,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 					bucket_journal_seq);
 			if (ret) {
 				bch2_fs_fatal_error(c,
-					"error setting bucket_needs_journal_commit: %i", ret);
+					"setting bucket_needs_journal_commit: %s", bch2_err_str(ret));
 				return ret;
 			}
 		}
@@ -860,23 +863,28 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 			*bucket_gen(ca, new.k->p.offset) = new_a->gen;
 
 		bch2_dev_usage_update(c, ca, old_a, new_a, journal_seq, false);
+		percpu_up_read(&c->mark_lock);
+
+#define eval_state(_a, expr)		({ const struct bch_alloc_v4 *a = _a; expr; })
+#define statechange(expr)		!eval_state(old_a, expr) && eval_state(new_a, expr)
+#define bucket_flushed(a)		(!a->journal_seq || a->journal_seq <= c->journal.flushed_seq_ondisk)
 
-		if (new_a->data_type == BCH_DATA_free &&
-		    (!new_a->journal_seq || new_a->journal_seq < c->journal.flushed_seq_ondisk))
+		if (statechange(a->data_type == BCH_DATA_free) &&
+		    bucket_flushed(new_a))
 			closure_wake_up(&c->freelist_wait);
 
-		if (new_a->data_type == BCH_DATA_need_discard &&
-		    (!bucket_journal_seq || bucket_journal_seq < c->journal.flushed_seq_ondisk))
-			bch2_do_discards(c);
+		if (statechange(a->data_type == BCH_DATA_need_discard) &&
+		    !bch2_bucket_is_open(c, new.k->p.inode, new.k->p.offset) &&
+		    bucket_flushed(new_a))
+			bch2_discard_one_bucket_fast(c, new.k->p);
 
-		if (old_a->data_type != BCH_DATA_cached &&
-		    new_a->data_type == BCH_DATA_cached &&
+		if (statechange(a->data_type == BCH_DATA_cached) &&
+		    !bch2_bucket_is_open(c, new.k->p.inode, new.k->p.offset) &&
 		    should_invalidate_buckets(ca, bch2_dev_usage_read(ca)))
 			bch2_do_invalidates(c);
 
-		if (new_a->data_type == BCH_DATA_need_gc_gens)
+		if (statechange(a->data_type == BCH_DATA_need_gc_gens))
 			bch2_do_gc_gens(c);
-		percpu_up_read(&c->mark_lock);
 	}
 
 	if ((flags & BTREE_TRIGGER_GC) &&
@@ -1045,14 +1053,13 @@ int bch2_check_alloc_key(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	if (k.k->type != discard_key_type &&
-	    (c->opts.reconstruct_alloc ||
-	     fsck_err(c, need_discard_key_wrong,
-		      "incorrect key in need_discard btree (got %s should be %s)\n"
-		      "  %s",
-		      bch2_bkey_types[k.k->type],
-		      bch2_bkey_types[discard_key_type],
-		      (bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf)))) {
+	if (fsck_err_on(k.k->type != discard_key_type,
+			c, need_discard_key_wrong,
+			"incorrect key in need_discard btree (got %s should be %s)\n"
+			"  %s",
+			bch2_bkey_types[k.k->type],
+			bch2_bkey_types[discard_key_type],
+			(bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
 		struct bkey_i *update =
 			bch2_trans_kmalloc(trans, sizeof(*update));
 
@@ -1076,15 +1083,14 @@ int bch2_check_alloc_key(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	if (k.k->type != freespace_key_type &&
-	    (c->opts.reconstruct_alloc ||
-	     fsck_err(c, freespace_key_wrong,
-		      "incorrect key in freespace btree (got %s should be %s)\n"
-		      "  %s",
-		      bch2_bkey_types[k.k->type],
-		      bch2_bkey_types[freespace_key_type],
-		      (printbuf_reset(&buf),
-		       bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf)))) {
+	if (fsck_err_on(k.k->type != freespace_key_type,
+			c, freespace_key_wrong,
+			"incorrect key in freespace btree (got %s should be %s)\n"
+			"  %s",
+			bch2_bkey_types[k.k->type],
+			bch2_bkey_types[freespace_key_type],
+			(printbuf_reset(&buf),
+			 bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
 		struct bkey_i *update =
 			bch2_trans_kmalloc(trans, sizeof(*update));
 
@@ -1108,14 +1114,13 @@ int bch2_check_alloc_key(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	if (a->gen != alloc_gen(k, gens_offset) &&
-	    (c->opts.reconstruct_alloc ||
-	     fsck_err(c, bucket_gens_key_wrong,
-		      "incorrect gen in bucket_gens btree (got %u should be %u)\n"
-		      "  %s",
-		      alloc_gen(k, gens_offset), a->gen,
-		      (printbuf_reset(&buf),
-		       bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf)))) {
+	if (fsck_err_on(a->gen != alloc_gen(k, gens_offset),
+			c, bucket_gens_key_wrong,
+			"incorrect gen in bucket_gens btree (got %u should be %u)\n"
+			"  %s",
+			alloc_gen(k, gens_offset), a->gen,
+			(printbuf_reset(&buf),
+			 bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
 		struct bkey_i_bucket_gens *g =
 			bch2_trans_kmalloc(trans, sizeof(*g));
 
@@ -1167,14 +1172,13 @@ int bch2_check_alloc_hole_freespace(struct btree_trans *trans,
 
 	*end = bkey_min(k.k->p, *end);
 
-	if (k.k->type != KEY_TYPE_set &&
-	    (c->opts.reconstruct_alloc ||
-	     fsck_err(c, freespace_hole_missing,
-		      "hole in alloc btree missing in freespace btree\n"
-		      "  device %llu buckets %llu-%llu",
-		      freespace_iter->pos.inode,
-		      freespace_iter->pos.offset,
-		      end->offset))) {
+	if (fsck_err_on(k.k->type != KEY_TYPE_set,
+			c, freespace_hole_missing,
+			"hole in alloc btree missing in freespace btree\n"
+			"  device %llu buckets %llu-%llu",
+			freespace_iter->pos.inode,
+			freespace_iter->pos.offset,
+			end->offset)) {
 		struct bkey_i *update =
 			bch2_trans_kmalloc(trans, sizeof(*update));
 
@@ -1604,6 +1608,36 @@ int bch2_check_alloc_to_lru_refs(struct bch_fs *c)
 	return ret;
 }
 
+static int discard_in_flight_add(struct bch_fs *c, struct bpos bucket)
+{
+	int ret;
+
+	mutex_lock(&c->discard_buckets_in_flight_lock);
+	darray_for_each(c->discard_buckets_in_flight, i)
+		if (bkey_eq(*i, bucket)) {
+			ret = -EEXIST;
+			goto out;
+		}
+
+	ret = darray_push(&c->discard_buckets_in_flight, bucket);
+out:
+	mutex_unlock(&c->discard_buckets_in_flight_lock);
+	return ret;
+}
+
+static void discard_in_flight_remove(struct bch_fs *c, struct bpos bucket)
+{
+	mutex_lock(&c->discard_buckets_in_flight_lock);
+	darray_for_each(c->discard_buckets_in_flight, i)
+		if (bkey_eq(*i, bucket)) {
+			darray_remove_item(&c->discard_buckets_in_flight, i);
+			goto found;
+		}
+	BUG();
+found:
+	mutex_unlock(&c->discard_buckets_in_flight_lock);
+}
+
 struct discard_buckets_state {
 	u64		seen;
 	u64		open;
@@ -1642,6 +1676,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	struct bch_dev *ca;
 	struct bkey_i_alloc_v4 *a;
 	struct printbuf buf = PRINTBUF;
+	bool discard_locked = false;
 	int ret = 0;
 
 	ca = bch_dev_bkey_exists(c, pos.inode);
@@ -1678,37 +1713,45 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	if (ret)
 		goto out;
 
-	if (BCH_ALLOC_V4_NEED_INC_GEN(&a->v)) {
-		a->v.gen++;
-		SET_BCH_ALLOC_V4_NEED_INC_GEN(&a->v, false);
-		goto write;
-	}
-
-	if (a->v.journal_seq > c->journal.flushed_seq_ondisk) {
-		if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info) {
-			bch2_trans_inconsistent(trans,
-				"clearing need_discard but journal_seq %llu > flushed_seq %llu\n"
-				"%s",
-				a->v.journal_seq,
-				c->journal.flushed_seq_ondisk,
-				(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
+	if (a->v.dirty_sectors) {
+		if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info,
+					       trans, "attempting to discard bucket with dirty data\n%s",
+					       (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
 			ret = -EIO;
-		}
 		goto out;
 	}
 
 	if (a->v.data_type != BCH_DATA_need_discard) {
-		if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info) {
-			bch2_trans_inconsistent(trans,
-				"bucket incorrectly set in need_discard btree\n"
-				"%s",
-				(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
-			ret = -EIO;
+		if (data_type_is_empty(a->v.data_type) &&
+		    BCH_ALLOC_V4_NEED_INC_GEN(&a->v)) {
+			a->v.gen++;
+			SET_BCH_ALLOC_V4_NEED_INC_GEN(&a->v, false);
+			goto write;
 		}
 
+		if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info,
+					       trans, "bucket incorrectly set in need_discard btree\n"
+					       "%s",
+					       (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+			ret = -EIO;
 		goto out;
 	}
 
+	if (a->v.journal_seq > c->journal.flushed_seq_ondisk) {
+		if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info,
+					       trans, "clearing need_discard but journal_seq %llu > flushed_seq %llu\n%s",
+					       a->v.journal_seq,
+					       c->journal.flushed_seq_ondisk,
+					       (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+			ret = -EIO;
+		goto out;
+	}
+
+	if (discard_in_flight_add(c, SPOS(iter.pos.inode, iter.pos.offset, true)))
+		goto out;
+
+	discard_locked = true;
+
 	if (!bkey_eq(*discard_pos_done, iter.pos) &&
 	    ca->mi.discard && !c->opts.nochanges) {
 		/*
@@ -1740,6 +1783,8 @@ write:
 	count_event(c, bucket_discard);
 	s->discarded++;
 out:
+	if (discard_locked)
+		discard_in_flight_remove(c, iter.pos);
 	s->seen++;
 	bch2_trans_iter_exit(trans, &iter);
 	percpu_ref_put(&ca->io_ref);
@@ -1779,6 +1824,94 @@ void bch2_do_discards(struct bch_fs *c)
 		bch2_write_ref_put(c, BCH_WRITE_REF_discard);
 }
 
+static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpos bucket)
+{
+	struct btree_iter iter;
+	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, BTREE_ITER_INTENT);
+	struct bkey_s_c k = bch2_btree_iter_peek_slot(&iter);
+	int ret = bkey_err(k);
+	if (ret)
+		goto err;
+
+	struct bkey_i_alloc_v4 *a = bch2_alloc_to_v4_mut(trans, k);
+	ret = PTR_ERR_OR_ZERO(a);
+	if (ret)
+		goto err;
+
+	BUG_ON(a->v.dirty_sectors);
+	SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
+	a->v.data_type = alloc_data_type(a->v, a->v.data_type);
+
+	ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
+err:
+	bch2_trans_iter_exit(trans, &iter);
+	return ret;
+}
+
+static void bch2_do_discards_fast_work(struct work_struct *work)
+{
+	struct bch_fs *c = container_of(work, struct bch_fs, discard_fast_work);
+
+	while (1) {
+		bool got_bucket = false;
+		struct bpos bucket;
+		struct bch_dev *ca;
+
+		mutex_lock(&c->discard_buckets_in_flight_lock);
+		darray_for_each(c->discard_buckets_in_flight, i) {
+			if (i->snapshot)
+				continue;
+
+			ca = bch_dev_bkey_exists(c, i->inode);
+
+			if (!percpu_ref_tryget(&ca->io_ref)) {
+				darray_remove_item(&c->discard_buckets_in_flight, i);
+				continue;
+			}
+
+			got_bucket = true;
+			bucket = *i;
+			i->snapshot = true;
+			break;
+		}
+		mutex_unlock(&c->discard_buckets_in_flight_lock);
+
+		if (!got_bucket)
+			break;
+
+		if (ca->mi.discard && !c->opts.nochanges)
+			blkdev_issue_discard(ca->disk_sb.bdev,
+					     bucket.offset * ca->mi.bucket_size,
+					     ca->mi.bucket_size,
+					     GFP_KERNEL);
+
+		int ret = bch2_trans_do(c, NULL, NULL,
+					BCH_WATERMARK_btree|
+					BCH_TRANS_COMMIT_no_enospc,
+					bch2_clear_bucket_needs_discard(trans, bucket));
+		bch_err_fn(c, ret);
+
+		percpu_ref_put(&ca->io_ref);
+		discard_in_flight_remove(c, bucket);
+
+		if (ret)
+			break;
+	}
+
+	bch2_write_ref_put(c, BCH_WRITE_REF_discard_fast);
+}
+
+static void bch2_discard_one_bucket_fast(struct bch_fs *c, struct bpos bucket)
+{
+	struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
+
+	if (!percpu_ref_is_dying(&ca->io_ref) &&
+	    !discard_in_flight_add(c, bucket) &&
+	    bch2_write_ref_tryget(c, BCH_WRITE_REF_discard_fast) &&
+	    !queue_work(c->write_ref_wq, &c->discard_fast_work))
+		bch2_write_ref_put(c, BCH_WRITE_REF_discard_fast);
+}
+
 static int invalidate_one_bucket(struct btree_trans *trans,
 				 struct btree_iter *lru_iter,
 				 struct bkey_s_c lru_k,
@@ -1813,6 +1946,7 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 		goto out;
 
 	BUG_ON(a->v.data_type != BCH_DATA_cached);
+	BUG_ON(a->v.dirty_sectors);
 
 	if (!a->v.cached_sectors)
 		bch_err(c, "invalidating empty bucket, confused");
@@ -2210,9 +2344,16 @@ void bch2_dev_allocator_add(struct bch_fs *c, struct bch_dev *ca)
 			set_bit(ca->dev_idx, c->rw_devs[i].d);
 }
 
+void bch2_fs_allocator_background_exit(struct bch_fs *c)
+{
+	darray_exit(&c->discard_buckets_in_flight);
+}
+
 void bch2_fs_allocator_background_init(struct bch_fs *c)
 {
 	spin_lock_init(&c->freelist_lock);
+	mutex_init(&c->discard_buckets_in_flight_lock);
 	INIT_WORK(&c->discard_work, bch2_do_discards_work);
+	INIT_WORK(&c->discard_fast_work, bch2_do_discards_fast_work);
 	INIT_WORK(&c->invalidate_work, bch2_do_invalidates_work);
 }