Update bcachefs sources to 2be6fc9b111c bcachefs: bch2_gc_btree() should not use btree_root_lock

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

40 files changed, 477 insertions, 263 deletions

diff --git a/.bcachefs_revision b/.bcachefs_revision
index 50da14dd..2a12d1f7 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-9404a01d3dc5553b106fa590602f4771b8e0b8ae
+2be6fc9b111cad37da8838e39c66244767bc7d0a
diff --git a/include/linux/closure.h b/include/linux/closure.h
index 59b8c06b..2af44427 100644
--- a/include/linux/closure.h
+++ b/include/linux/closure.h
@@ -159,6 +159,7 @@ struct closure {
 #ifdef CONFIG_DEBUG_CLOSURES
 #define CLOSURE_MAGIC_DEAD	0xc054dead
 #define CLOSURE_MAGIC_ALIVE	0xc054a11e
+#define CLOSURE_MAGIC_STACK	0xc05451cc
 
 	unsigned int		magic;
 	struct list_head	all;
@@ -323,12 +324,18 @@ static inline void closure_init_stack(struct closure *cl)
 {
 	memset(cl, 0, sizeof(struct closure));
 	atomic_set(&cl->remaining, CLOSURE_REMAINING_INITIALIZER);
+#ifdef CONFIG_DEBUG_CLOSURES
+	cl->magic = CLOSURE_MAGIC_STACK;
+#endif
 }
 
 static inline void closure_init_stack_release(struct closure *cl)
 {
 	memset(cl, 0, sizeof(struct closure));
 	atomic_set_release(&cl->remaining, CLOSURE_REMAINING_INITIALIZER);
+#ifdef CONFIG_DEBUG_CLOSURES
+	cl->magic = CLOSURE_MAGIC_STACK;
+#endif
 }
 
 /**
diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index 3831ef2d..27bf6915 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -5,6 +5,8 @@ struct lock_class_key {};
 struct task_struct;
 
 # define lock_acquire(l, s, t, r, c, n, i)	do { } while (0)
+# define lock_acquire_exclusive(...)		do { } while (0)
+# define lockdep_set_notrack_class(...)		do { } while (0)
 # define lock_release(l, i)			do { } while (0)
 # define lock_set_class(l, n, k, s, i)		do { } while (0)
 # define lock_set_subclass(l, s, i)		do { } while (0)
diff --git a/libbcachefs/acl.c b/libbcachefs/acl.c
index 250d6c6d..a7b425d3 100644
--- a/libbcachefs/acl.c
+++ b/libbcachefs/acl.c
@@ -346,7 +346,6 @@ int bch2_set_acl(struct mnt_idmap *idmap,
 {
 	struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter inode_iter = { NULL };
 	struct bch_inode_unpacked inode_u;
 	struct posix_acl *acl;
@@ -354,6 +353,7 @@ int bch2_set_acl(struct mnt_idmap *idmap,
 	int ret;
 
 	mutex_lock(&inode->ei_update_lock);
+	struct btree_trans *trans = bch2_trans_get(c);
 retry:
 	bch2_trans_begin(trans);
 	acl = _acl;
@@ -394,8 +394,8 @@ btree_err:
 
 	set_cached_acl(&inode->v, type, acl);
 err:
-	mutex_unlock(&inode->ei_update_lock);
 	bch2_trans_put(trans);
+	mutex_unlock(&inode->ei_update_lock);
 
 	return ret;
 }
diff --git a/libbcachefs/alloc_background.c b/libbcachefs/alloc_background.c
index 8e8aed2a..9ff822e4 100644
--- a/libbcachefs/alloc_background.c
+++ b/libbcachefs/alloc_background.c
@@ -3,6 +3,7 @@
 #include "alloc_background.h"
 #include "alloc_foreground.h"
 #include "backpointers.h"
+#include "bkey_buf.h"
 #include "btree_cache.h"
 #include "btree_io.h"
 #include "btree_key_cache.h"
@@ -1600,13 +1601,13 @@ err:
 }
 
 static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
-				       struct btree_iter *alloc_iter)
+				       struct btree_iter *alloc_iter,
+				       struct bkey_buf *last_flushed)
 {
 	struct bch_fs *c = trans->c;
-	struct btree_iter lru_iter;
 	struct bch_alloc_v4 a_convert;
 	const struct bch_alloc_v4 *a;
-	struct bkey_s_c alloc_k, lru_k;
+	struct bkey_s_c alloc_k;
 	struct printbuf buf = PRINTBUF;
 	int ret;
 
@@ -1620,6 +1621,14 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 
 	a = bch2_alloc_to_v4(alloc_k, &a_convert);
 
+	if (a->fragmentation_lru) {
+		ret = bch2_lru_check_set(trans, BCH_LRU_FRAGMENTATION_START,
+					 a->fragmentation_lru,
+					 alloc_k, last_flushed);
+		if (ret)
+			return ret;
+	}
+
 	if (a->data_type != BCH_DATA_cached)
 		return 0;
 
@@ -1644,41 +1653,30 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 		a = &a_mut->v;
 	}
 
-	lru_k = bch2_bkey_get_iter(trans, &lru_iter, BTREE_ID_lru,
-			     lru_pos(alloc_k.k->p.inode,
-				     bucket_to_u64(alloc_k.k->p),
-				     a->io_time[READ]), 0);
-	ret = bkey_err(lru_k);
+	ret = bch2_lru_check_set(trans, alloc_k.k->p.inode, a->io_time[READ],
+				 alloc_k, last_flushed);
 	if (ret)
-		return ret;
-
-	if (fsck_err_on(lru_k.k->type != KEY_TYPE_set,
-			trans, alloc_key_to_missing_lru_entry,
-			"missing lru entry\n"
-			"  %s",
-			(printbuf_reset(&buf),
-			 bch2_bkey_val_to_text(&buf, c, alloc_k), buf.buf))) {
-		ret = bch2_lru_set(trans,
-				   alloc_k.k->p.inode,
-				   bucket_to_u64(alloc_k.k->p),
-				   a->io_time[READ]);
-		if (ret)
-			goto err;
-	}
+		goto err;
 err:
 fsck_err:
-	bch2_trans_iter_exit(trans, &lru_iter);
 	printbuf_exit(&buf);
 	return ret;
 }
 
 int bch2_check_alloc_to_lru_refs(struct bch_fs *c)
 {
+	struct bkey_buf last_flushed;
+
+	bch2_bkey_buf_init(&last_flushed);
+	bkey_init(&last_flushed.k->k);
+
 	int ret = bch2_trans_run(c,
 		for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
 				POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			bch2_check_alloc_to_lru_ref(trans, &iter)));
+			bch2_check_alloc_to_lru_ref(trans, &iter, &last_flushed)));
+
+	bch2_bkey_buf_exit(&last_flushed, c);
 	bch_err_fn(c, ret);
 	return ret;
 }
diff --git a/libbcachefs/alloc_foreground.c b/libbcachefs/alloc_foreground.c
index 73228b25..49a5789d 100644
--- a/libbcachefs/alloc_foreground.c
+++ b/libbcachefs/alloc_foreground.c
@@ -626,8 +626,14 @@ again:
 	if (usage->d[BCH_DATA_need_gc_gens].buckets > avail)
 		bch2_gc_gens_async(c);
 
-	if (should_invalidate_buckets(ca, *usage))
+	if (should_invalidate_buckets(ca, *usage)) {
 		bch2_dev_do_invalidates(ca);
+		rcu_read_lock();
+		struct task_struct *t = rcu_dereference(c->copygc_thread);
+		if (t)
+			wake_up_process(t);
+		rcu_read_unlock();
+	}
 
 	if (!avail) {
 		if (cl && !waiting) {
@@ -1703,6 +1709,7 @@ void bch2_fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 		nr[c->open_buckets[i].data_type]++;
 
+	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 24);
 
 	prt_printf(out, "hidden\t%llu\n",		percpu_u64_get(&c->usage->hidden));
@@ -1734,6 +1741,7 @@ void bch2_dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 	for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 		nr[c->open_buckets[i].data_type]++;
 
+	printbuf_tabstops_reset(out);
 	printbuf_tabstop_push(out, 12);
 	printbuf_tabstop_push(out, 16);
 	printbuf_tabstop_push(out, 16);
diff --git a/libbcachefs/backpointers.c b/libbcachefs/backpointers.c
index ca16fa5d..96217e8d 100644
--- a/libbcachefs/backpointers.c
+++ b/libbcachefs/backpointers.c
@@ -434,13 +434,6 @@ int bch2_check_btree_backpointers(struct bch_fs *c)
 	return ret;
 }
 
-static inline bool bkey_and_val_eq(struct bkey_s_c l, struct bkey_s_c r)
-{
-	return bpos_eq(l.k->p, r.k->p) &&
-		bkey_bytes(l.k) == bkey_bytes(r.k) &&
-		!memcmp(l.v, r.v, bkey_val_bytes(l.k));
-}
-
 struct extents_to_bp_state {
 	struct bpos	bucket_start;
 	struct bpos	bucket_end;
@@ -536,11 +529,8 @@ static int check_bp_exists(struct btree_trans *trans,
 	struct btree_iter other_extent_iter = {};
 	struct printbuf buf = PRINTBUF;
 	struct bkey_s_c bp_k;
-	struct bkey_buf tmp;
 	int ret = 0;
 
-	bch2_bkey_buf_init(&tmp);
-
 	struct bch_dev *ca = bch2_dev_bucket_tryget(c, bucket);
 	if (!ca) {
 		prt_str(&buf, "extent for nonexistent device:bucket ");
@@ -565,22 +555,9 @@ static int check_bp_exists(struct btree_trans *trans,
 
 	if (bp_k.k->type != KEY_TYPE_backpointer ||
 	    memcmp(bkey_s_c_to_backpointer(bp_k).v, &bp, sizeof(bp))) {
-		bch2_bkey_buf_reassemble(&tmp, c, orig_k);
-
-		if (!bkey_and_val_eq(orig_k, bkey_i_to_s_c(s->last_flushed.k))) {
-			if (bp.level) {
-				bch2_trans_unlock(trans);
-				bch2_btree_interior_updates_flush(c);
-			}
-
-			ret = bch2_btree_write_buffer_flush_sync(trans);
-			if (ret)
-				goto err;
-
-			bch2_bkey_buf_copy(&s->last_flushed, c, tmp.k);
-			ret = -BCH_ERR_transaction_restart_write_buffer_flush;
-			goto out;
-		}
+		ret = bch2_btree_write_buffer_maybe_flush(trans, orig_k, &s->last_flushed);
+		if (ret)
+			goto err;
 
 		goto check_existing_bp;
 	}
@@ -589,7 +566,6 @@ err:
 fsck_err:
 	bch2_trans_iter_exit(trans, &other_extent_iter);
 	bch2_trans_iter_exit(trans, &bp_iter);
-	bch2_bkey_buf_exit(&tmp, c);
 	bch2_dev_put(ca);
 	printbuf_exit(&buf);
 	return ret;
@@ -794,6 +770,8 @@ static int bch2_get_btree_in_memory_pos(struct btree_trans *trans,
 		    !((1U << btree) & btree_interior_mask))
 			continue;
 
+		bch2_trans_begin(trans);
+
 		__for_each_btree_node(trans, iter, btree,
 				      btree == start.btree ? start.pos : POS_MIN,
 				      0, depth, BTREE_ITER_prefetch, b, ret) {
@@ -905,7 +883,7 @@ static int check_one_backpointer(struct btree_trans *trans,
 				 struct bbpos start,
 				 struct bbpos end,
 				 struct bkey_s_c_backpointer bp,
-				 struct bpos *last_flushed_pos)
+				 struct bkey_buf *last_flushed)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
@@ -925,20 +903,18 @@ static int check_one_backpointer(struct btree_trans *trans,
 	if (ret)
 		return ret;
 
-	if (!k.k && !bpos_eq(*last_flushed_pos, bp.k->p)) {
-		*last_flushed_pos = bp.k->p;
-		ret = bch2_btree_write_buffer_flush_sync(trans) ?:
-			-BCH_ERR_transaction_restart_write_buffer_flush;
-		goto out;
-	}
+	if (!k.k) {
+		ret = bch2_btree_write_buffer_maybe_flush(trans, bp.s_c, last_flushed);
+		if (ret)
+			goto out;
 
-	if (fsck_err_on(!k.k,
-			trans, backpointer_to_missing_ptr,
-			"backpointer for missing %s\n  %s",
-			bp.v->level ? "btree node" : "extent",
-			(bch2_bkey_val_to_text(&buf, c, bp.s_c), buf.buf))) {
-		ret = bch2_btree_delete_at_buffered(trans, BTREE_ID_backpointers, bp.k->p);
-		goto out;
+		if (fsck_err(trans, backpointer_to_missing_ptr,
+			     "backpointer for missing %s\n  %s",
+			     bp.v->level ? "btree node" : "extent",
+			     (bch2_bkey_val_to_text(&buf, c, bp.s_c), buf.buf))) {
+			ret = bch2_btree_delete_at_buffered(trans, BTREE_ID_backpointers, bp.k->p);
+			goto out;
+		}
 	}
 out:
 fsck_err:
@@ -951,14 +927,20 @@ static int bch2_check_backpointers_to_extents_pass(struct btree_trans *trans,
 						   struct bbpos start,
 						   struct bbpos end)
 {
-	struct bpos last_flushed_pos = SPOS_MAX;
+	struct bkey_buf last_flushed;
 
-	return for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
+	bch2_bkey_buf_init(&last_flushed);
+	bkey_init(&last_flushed.k->k);
+
+	int ret = for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
 				  POS_MIN, BTREE_ITER_prefetch, k,
 				  NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 		check_one_backpointer(trans, start, end,
 				      bkey_s_c_to_backpointer(k),
-				      &last_flushed_pos));
+				      &last_flushed));
+
+	bch2_bkey_buf_exit(&last_flushed, trans->c);
+	return ret;
 }
 
 int bch2_check_backpointers_to_extents(struct bch_fs *c)
diff --git a/libbcachefs/bcachefs.h b/libbcachefs/bcachefs.h
index 372bc339..65e46225 100644
--- a/libbcachefs/bcachefs.h
+++ b/libbcachefs/bcachefs.h
@@ -980,7 +980,7 @@ struct bch_fs {
 	struct bch_fs_rebalance	rebalance;
 
 	/* COPYGC */
-	struct task_struct	*copygc_thread;
+	struct task_struct __rcu *copygc_thread;
 	struct write_point	copygc_write_point;
 	s64			copygc_wait_at;
 	s64			copygc_wait;
diff --git a/libbcachefs/bkey.c b/libbcachefs/bkey.c
index 94a1d198..587d7318 100644
--- a/libbcachefs/bkey.c
+++ b/libbcachefs/bkey.c
@@ -660,8 +660,9 @@ int bch2_bkey_format_invalid(struct bch_fs *c,
 		    bch2_bkey_format_field_overflows(f, i)) {
 			unsigned unpacked_bits = bch2_bkey_format_current.bits_per_field[i];
 			u64 unpacked_max = ~((~0ULL << 1) << (unpacked_bits - 1));
-			u64 packed_max = f->bits_per_field[i]
-				? ~((~0ULL << 1) << (f->bits_per_field[i] - 1))
+			unsigned packed_bits = min(64, f->bits_per_field[i]);
+			u64 packed_max = packed_bits
+				? ~((~0ULL << 1) << (packed_bits - 1))
 				: 0;
 
 			prt_printf(err, "field %u too large: %llu + %llu > %llu",
diff --git a/libbcachefs/bkey.h b/libbcachefs/bkey.h
index fcd43915..93635714 100644
--- a/libbcachefs/bkey.h
+++ b/libbcachefs/bkey.h
@@ -194,6 +194,13 @@ static inline struct bpos bkey_max(struct bpos l, struct bpos r)
 	return bkey_gt(l, r) ? l : r;
 }
 
+static inline bool bkey_and_val_eq(struct bkey_s_c l, struct bkey_s_c r)
+{
+	return bpos_eq(l.k->p, r.k->p) &&
+		bkey_bytes(l.k) == bkey_bytes(r.k) &&
+		!memcmp(l.v, r.v, bkey_val_bytes(l.k));
+}
+
 void bch2_bpos_swab(struct bpos *);
 void bch2_bkey_swab_key(const struct bkey_format *, struct bkey_packed *);
 
diff --git a/libbcachefs/btree_gc.c b/libbcachefs/btree_gc.c
index 2e9ccb20..1e4c3e30 100644
--- a/libbcachefs/btree_gc.c
+++ b/libbcachefs/btree_gc.c
@@ -675,15 +675,26 @@ static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree, bool in
 	}
 
 	/* root */
-	mutex_lock(&c->btree_root_lock);
-	struct btree *b = bch2_btree_id_root(c, btree)->b;
-	if (!btree_node_fake(b)) {
+	do {
+		struct btree_iter iter;
+		bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN,
+					  0, bch2_btree_id_root(c, btree)->b->c.level, 0);
+		struct btree *b = bch2_btree_iter_peek_node(&iter);
+		ret = PTR_ERR_OR_ZERO(b);
+		if (ret)
+			goto err_root;
+
+		if (b != btree_node_root(c, b)) {
+			bch2_trans_iter_exit(trans, &iter);
+			continue;
+		}
+
 		gc_pos_set(c, gc_pos_btree(btree, b->c.level + 1, SPOS_MAX));
-		ret = lockrestart_do(trans,
-			bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1,
-					 NULL, NULL, bkey_i_to_s_c(&b->key), initial));
-	}
-	mutex_unlock(&c->btree_root_lock);
+		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
+		ret = bch2_gc_mark_key(trans, btree, b->c.level + 1, NULL, NULL, k, initial);
+err_root:
+		bch2_trans_iter_exit(trans, &iter);
+	} while (bch2_err_matches(ret, BCH_ERR_transaction_restart));
 err:
 	bch_err_fn(c, ret);
 	return ret;
@@ -818,6 +829,8 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 			return ret;
 	}
 
+	gc.fragmentation_lru = alloc_lru_idx_fragmentation(gc, ca);
+
 	if (fsck_err_on(new.data_type != gc.data_type,
 			trans, alloc_key_data_type_wrong,
 			"bucket %llu:%llu gen %u has wrong data_type"
@@ -832,25 +845,20 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	if (fsck_err_on(new._f != gc._f,				\
 			trans, _errtype,				\
 			"bucket %llu:%llu gen %u data type %s has wrong " #_f	\
-			": got %u, should be %u",			\
+			": got %llu, should be %llu",			\
 			iter->pos.inode, iter->pos.offset,		\
 			gc.gen,						\
 			bch2_data_type_str(gc.data_type),		\
-			new._f, gc._f))					\
+			(u64) new._f, (u64) gc._f))				\
 		new._f = gc._f;						\
 
-	copy_bucket_field(alloc_key_gen_wrong,
-			  gen);
-	copy_bucket_field(alloc_key_dirty_sectors_wrong,
-			  dirty_sectors);
-	copy_bucket_field(alloc_key_stripe_sectors_wrong,
-			  stripe_sectors);
-	copy_bucket_field(alloc_key_cached_sectors_wrong,
-			  cached_sectors);
-	copy_bucket_field(alloc_key_stripe_wrong,
-			  stripe);
-	copy_bucket_field(alloc_key_stripe_redundancy_wrong,
-			  stripe_redundancy);
+	copy_bucket_field(alloc_key_gen_wrong,			gen);
+	copy_bucket_field(alloc_key_dirty_sectors_wrong,	dirty_sectors);
+	copy_bucket_field(alloc_key_stripe_sectors_wrong,	stripe_sectors);
+	copy_bucket_field(alloc_key_cached_sectors_wrong,	cached_sectors);
+	copy_bucket_field(alloc_key_stripe_wrong,		stripe);
+	copy_bucket_field(alloc_key_stripe_redundancy_wrong,	stripe_redundancy);
+	copy_bucket_field(alloc_key_fragmentation_lru_wrong,	fragmentation_lru);
 #undef copy_bucket_field
 
 	if (!bch2_alloc_v4_cmp(*old, new))
@@ -864,7 +872,7 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	a->v = new;
 
 	/*
-	 * The trigger normally makes sure this is set, but we're not running
+	 * The trigger normally makes sure these are set, but we're not running
 	 * triggers:
 	 */
 	if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ])
diff --git a/libbcachefs/btree_iter.c b/libbcachefs/btree_iter.c
index 80f4a395..803cc58f 100644
--- a/libbcachefs/btree_iter.c
+++ b/libbcachefs/btree_iter.c
@@ -987,8 +987,7 @@ retry_all:
 
 	bch2_trans_unlock(trans);
 	cond_resched();
-	trans->locked = true;
-	trans->last_unlock_ip = 0;
+	trans_set_locked(trans);
 
 	if (unlikely(trans->memory_allocation_failure)) {
 		struct closure cl;
@@ -3088,8 +3087,8 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 		bch2_trans_srcu_unlock(trans);
 
 	trans->last_begin_ip = _RET_IP_;
-	trans->locked  = true;
-	trans->last_unlock_ip = 0;
+
+	trans_set_locked(trans);
 
 	if (trans->restarted) {
 		bch2_btree_path_traverse_all(trans);
@@ -3159,7 +3158,6 @@ got_trans:
 	trans->last_begin_time	= local_clock();
 	trans->fn_idx		= fn_idx;
 	trans->locking_wait.task = current;
-	trans->locked		= true;
 	trans->journal_replay_not_finished =
 		unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)) &&
 		atomic_inc_not_zero(&c->journal_keys.ref);
@@ -3173,6 +3171,9 @@ got_trans:
 
 	trans->paths_allocated[0] = 1;
 
+	static struct lock_class_key lockdep_key;
+	lockdep_init_map(&trans->dep_map, "bcachefs_btree", &lockdep_key, 0);
+
 	if (fn_idx < BCH_TRANSACTIONS_NR) {
 		trans->fn = bch2_btree_transaction_fns[fn_idx];
 
@@ -3193,6 +3194,7 @@ got_trans:
 	trans->srcu_idx		= srcu_read_lock(&c->btree_trans_barrier);
 	trans->srcu_lock_time	= jiffies;
 	trans->srcu_held	= true;
+	trans_set_locked(trans);
 
 	closure_init_stack_release(&trans->ref);
 	return trans;
@@ -3441,7 +3443,22 @@ int bch2_fs_btree_iter_init(struct bch_fs *c)
 		mempool_init_kmalloc_pool(&c->btree_trans_mem_pool, 1,
 					  BTREE_TRANS_MEM_MAX) ?:
 		init_srcu_struct(&c->btree_trans_barrier);
-	if (!ret)
-		c->btree_trans_barrier_initialized = true;
-	return ret;
+	if (ret)
+		return ret;
+
+	/*
+	 * static annotation (hackily done) for lock ordering of reclaim vs.
+	 * btree node locks:
+	 */
+#ifdef CONFIG_LOCKDEP
+	fs_reclaim_acquire(GFP_KERNEL);
+	struct btree_trans *trans = bch2_trans_get(c);
+	trans_set_locked(trans);
+	bch2_trans_put(trans);
+	fs_reclaim_release(GFP_KERNEL);
+#endif
+
+	c->btree_trans_barrier_initialized = true;
+	return 0;
+
 }
diff --git a/libbcachefs/btree_locking.c b/libbcachefs/btree_locking.c
index 79057cde..dd52372a 100644
--- a/libbcachefs/btree_locking.c
+++ b/libbcachefs/btree_locking.c
@@ -10,7 +10,7 @@ void bch2_btree_lock_init(struct btree_bkey_cached_common *b,
 			  enum six_lock_init_flags flags)
 {
 	__six_lock_init(&b->lock, "b->c.lock", &bch2_btree_node_lock_key, flags);
-	lockdep_set_novalidate_class(&b->lock);
+	lockdep_set_notrack_class(&b->lock);
 }
 
 #ifdef CONFIG_LOCKDEP
@@ -792,8 +792,7 @@ static inline int __bch2_trans_relock(struct btree_trans *trans, bool trace)
 			return bch2_trans_relock_fail(trans, path, &f, trace);
 	}
 
-	trans->locked = true;
-	trans->last_unlock_ip = 0;
+	trans_set_locked(trans);
 out:
 	bch2_trans_verify_locks(trans);
 	return 0;
@@ -813,16 +812,14 @@ void bch2_trans_unlock_noassert(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock_long(struct btree_trans *trans)
diff --git a/libbcachefs/btree_locking.h b/libbcachefs/btree_locking.h
index 2530fdb2..b9fd14e3 100644
--- a/libbcachefs/btree_locking.h
+++ b/libbcachefs/btree_locking.h
@@ -13,6 +13,8 @@
 #include "btree_iter.h"
 #include "six.h"
 
+#include <linux/sched/mm.h>
+
 void bch2_btree_lock_init(struct btree_bkey_cached_common *, enum six_lock_init_flags);
 
 #ifdef CONFIG_LOCKDEP
@@ -194,6 +196,30 @@ int bch2_six_check_for_deadlock(struct six_lock *lock, void *p);
 
 /* lock: */
 
+static inline void trans_set_locked(struct btree_trans *trans)
+{
+	if (!trans->locked) {
+		lock_acquire_exclusive(&trans->dep_map, 0, 0, NULL, _THIS_IP_);
+		trans->locked = true;
+		trans->last_unlock_ip = 0;
+
+		trans->pf_memalloc_nofs = (current->flags & PF_MEMALLOC_NOFS) != 0;
+		current->flags |= PF_MEMALLOC_NOFS;
+	}
+}
+
+static inline void trans_set_unlocked(struct btree_trans *trans)
+{
+	if (trans->locked) {
+		lock_release(&trans->dep_map, _THIS_IP_);
+		trans->locked = false;
+		trans->last_unlock_ip = _RET_IP_;
+
+		if (!trans->pf_memalloc_nofs)
+			current->flags &= ~PF_MEMALLOC_NOFS;
+	}
+}
+
 static inline int __btree_node_lock_nopath(struct btree_trans *trans,
 					 struct btree_bkey_cached_common *b,
 					 enum six_lock_type type,
diff --git a/libbcachefs/btree_types.h b/libbcachefs/btree_types.h
index c9c9864a..79898f68 100644
--- a/libbcachefs/btree_types.h
+++ b/libbcachefs/btree_types.h
@@ -483,6 +483,7 @@ struct btree_trans {
 	bool			lock_may_not_fail:1;
 	bool			srcu_held:1;
 	bool			locked:1;
+	bool			pf_memalloc_nofs:1;
 	bool			write_locked:1;
 	bool			used_mempool:1;
 	bool			in_traverse_all:1;
@@ -522,6 +523,9 @@ struct btree_trans {
 	unsigned		journal_u64s;
 	unsigned		extra_disk_res; /* XXX kill */
 
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map	dep_map;
+#endif
 	/* Entries before this are zeroed out on every bch2_trans_get() call */
 
 	struct list_head	list;
diff --git a/libbcachefs/btree_write_buffer.c b/libbcachefs/btree_write_buffer.c
index 5e1262a5..b9fe7368 100644
--- a/libbcachefs/btree_write_buffer.c
+++ b/libbcachefs/btree_write_buffer.c
@@ -1,12 +1,14 @@
 // SPDX-License-Identifier: GPL-2.0
 
 #include "bcachefs.h"
+#include "bkey_buf.h"
 #include "btree_locking.h"
 #include "btree_update.h"
 #include "btree_update_interior.h"
 #include "btree_write_buffer.h"
 #include "disk_accounting.h"
 #include "error.h"
+#include "extents.h"
 #include "journal.h"
 #include "journal_io.h"
 #include "journal_reclaim.h"
@@ -558,6 +560,41 @@ int bch2_btree_write_buffer_tryflush(struct btree_trans *trans)
 	return ret;
 }
 
+/**
+ * In check and repair code, when checking references to write buffer btrees we
+ * need to issue a flush before we have a definitive error: this issues a flush
+ * if this is a key we haven't yet checked.
+ */
+int bch2_btree_write_buffer_maybe_flush(struct btree_trans *trans,
+					struct bkey_s_c referring_k,
+					struct bkey_buf *last_flushed)
+{
+	struct bch_fs *c = trans->c;
+	struct bkey_buf tmp;
+	int ret = 0;
+
+	bch2_bkey_buf_init(&tmp);
+
+	if (!bkey_and_val_eq(referring_k, bkey_i_to_s_c(last_flushed->k))) {
+		bch2_bkey_buf_reassemble(&tmp, c, referring_k);
+
+		if (bkey_is_btree_ptr(referring_k.k)) {
+			bch2_trans_unlock(trans);
+			bch2_btree_interior_updates_flush(c);
+		}
+
+		ret = bch2_btree_write_buffer_flush_sync(trans);
+		if (ret)
+			goto err;
+
+		bch2_bkey_buf_copy(last_flushed, c, tmp.k);
+		ret = -BCH_ERR_transaction_restart_write_buffer_flush;
+	}
+err:
+	bch2_bkey_buf_exit(&tmp, c);
+	return ret;
+}
+
 static void bch2_btree_write_buffer_flush_work(struct work_struct *work)
 {
 	struct bch_fs *c = container_of(work, struct bch_fs, btree_write_buffer.flush_work);
diff --git a/libbcachefs/btree_write_buffer.h b/libbcachefs/btree_write_buffer.h
index ee3710ea..725e7965 100644
--- a/libbcachefs/btree_write_buffer.h
+++ b/libbcachefs/btree_write_buffer.h
@@ -24,6 +24,9 @@ int bch2_btree_write_buffer_flush_sync(struct btree_trans *);
 int bch2_btree_write_buffer_flush_nocheck_rw(struct btree_trans *);
 int bch2_btree_write_buffer_tryflush(struct btree_trans *);
 
+struct bkey_buf;
+int bch2_btree_write_buffer_maybe_flush(struct btree_trans *, struct bkey_s_c, struct bkey_buf *);
+
 struct journal_keys_to_wb {
 	struct btree_write_buffer_keys	*wb;
 	size_t				room;
diff --git a/libbcachefs/clock.c b/libbcachefs/clock.c
index 36364445..df3763c1 100644
--- a/libbcachefs/clock.c
+++ b/libbcachefs/clock.c
@@ -15,18 +15,15 @@ static inline long io_timer_cmp(io_timer_heap *h,
 
 void bch2_io_timer_add(struct io_clock *clock, struct io_timer *timer)
 {
-	size_t i;
-
 	spin_lock(&clock->timer_lock);
 
-	if (time_after_eq((unsigned long) atomic64_read(&clock->now),
-			  timer->expire)) {
+	if (time_after_eq64((u64) atomic64_read(&clock->now), timer->expire)) {
 		spin_unlock(&clock->timer_lock);
 		timer->fn(timer);
 		return;
 	}
 
-	for (i = 0; i < clock->timers.used; i++)
+	for (size_t i = 0; i < clock->timers.used; i++)
 		if (clock->timers.data[i] == timer)
 			goto out;
 
@@ -37,11 +34,9 @@ out:
 
 void bch2_io_timer_del(struct io_clock *clock, struct io_timer *timer)
 {
-	size_t i;
-
 	spin_lock(&clock->timer_lock);
 
-	for (i = 0; i < clock->timers.used; i++)
+	for (size_t i = 0; i < clock->timers.used; i++)
 		if (clock->timers.data[i] == timer) {
 			heap_del(&clock->timers, i, io_timer_cmp, NULL);
 			break;
@@ -75,33 +70,31 @@ static void io_clock_cpu_timeout(struct timer_list *timer)
 	wake_up_process(wait->task);
 }
 
-void bch2_io_clock_schedule_timeout(struct io_clock *clock, unsigned long until)
+void bch2_io_clock_schedule_timeout(struct io_clock *clock, u64 until)
 {
-	struct io_clock_wait wait;
+	struct io_clock_wait wait = {
+		.io_timer.expire	= until,
+		.io_timer.fn		= io_clock_wait_fn,
+		.io_timer.fn2		= (void *) _RET_IP_,
+		.task			= current,
+	};
 
-	/* XXX: calculate sleep time rigorously */
-	wait.io_timer.expire	= until;
-	wait.io_timer.fn	= io_clock_wait_fn;
-	wait.task		= current;
-	wait.expired		= 0;
 	bch2_io_timer_add(clock, &wait.io_timer);
-
 	schedule();
-
 	bch2_io_timer_del(clock, &wait.io_timer);
 }
 
 void bch2_kthread_io_clock_wait(struct io_clock *clock,
-				unsigned long io_until,
-				unsigned long cpu_timeout)
+				u64 io_until, unsigned long cpu_timeout)
 {
 	bool kthread = (current->flags & PF_KTHREAD) != 0;
-	struct io_clock_wait wait;
+	struct io_clock_wait wait = {
+		.io_timer.expire	= io_until,
+		.io_timer.fn		= io_clock_wait_fn,
+		.io_timer.fn2		= (void *) _RET_IP_,
+		.task			= current,
+	};
 
-	wait.io_timer.expire	= io_until;
-	wait.io_timer.fn	= io_clock_wait_fn;
-	wait.task		= current;
-	wait.expired		= 0;
 	bch2_io_timer_add(clock, &wait.io_timer);
 
 	timer_setup_on_stack(&wait.cpu_timer, io_clock_cpu_timeout, 0);
@@ -127,44 +120,41 @@ void bch2_kthread_io_clock_wait(struct io_clock *clock,
 	bch2_io_timer_del(clock, &wait.io_timer);
 }
 
-static struct io_timer *get_expired_timer(struct io_clock *clock,
-					  unsigned long now)
+static struct io_timer *get_expired_timer(struct io_clock *clock, u64 now)
 {
 	struct io_timer *ret = NULL;
 
-	spin_lock(&clock->timer_lock);
-
 	if (clock->timers.used &&
-	    time_after_eq(now, clock->timers.data[0]->expire))
+	    time_after_eq64(now, clock->timers.data[0]->expire))
 		heap_pop(&clock->timers, ret, io_timer_cmp, NULL);
-
-	spin_unlock(&clock->timer_lock);
-
 	return ret;
 }
 
-void __bch2_increment_clock(struct io_clock *clock, unsigned sectors)
+void __bch2_increment_clock(struct io_clock *clock, u64 sectors)
 {
 	struct io_timer *timer;
-	unsigned long now = atomic64_add_return(sectors, &clock->now);
+	u64 now = atomic64_add_return(sectors, &clock->now);
 
+	spin_lock(&clock->timer_lock);
 	while ((timer = get_expired_timer(clock, now)))
 		timer->fn(timer);
+	spin_unlock(&clock->timer_lock);
 }
 
 void bch2_io_timers_to_text(struct printbuf *out, struct io_clock *clock)
 {
-	unsigned long now;
-	unsigned i;
-
 	out->atomic++;
 	spin_lock(&clock->timer_lock);
-	now = atomic64_read(&clock->now);
+	u64 now = atomic64_read(&clock->now);
+
+	printbuf_tabstop_push(out, 40);
+	prt_printf(out, "current time:\t%llu\n", now);
 
-	for (i = 0; i < clock->timers.used; i++)
-		prt_printf(out, "%ps:\t%li\n",
+	for (unsigned i = 0; i < clock->timers.used; i++)
+		prt_printf(out, "%ps %ps:\t%llu\n",
 		       clock->timers.data[i]->fn,
-		       clock->timers.data[i]->expire - now);
+		       clock->timers.data[i]->fn2,
+		       clock->timers.data[i]->expire);
 	spin_unlock(&clock->timer_lock);
 	--out->atomic;
 }
diff --git a/libbcachefs/clock.h b/libbcachefs/clock.h
index 70a0f743..85c975df 100644
--- a/libbcachefs/clock.h
+++ b/libbcachefs/clock.h
@@ -4,12 +4,11 @@
 
 void bch2_io_timer_add(struct io_clock *, struct io_timer *);
 void bch2_io_timer_del(struct io_clock *, struct io_timer *);
-void bch2_kthread_io_clock_wait(struct io_clock *, unsigned long,
-				unsigned long);
+void bch2_kthread_io_clock_wait(struct io_clock *, u64, unsigned long);
 
-void __bch2_increment_clock(struct io_clock *, unsigned);
+void __bch2_increment_clock(struct io_clock *, u64);
 
-static inline void bch2_increment_clock(struct bch_fs *c, unsigned sectors,
+static inline void bch2_increment_clock(struct bch_fs *c, u64 sectors,
 					int rw)
 {
 	struct io_clock *clock = &c->io_clock[rw];
@@ -19,7 +18,7 @@ static inline void bch2_increment_clock(struct bch_fs *c, unsigned sectors,
 		__bch2_increment_clock(clock, this_cpu_xchg(*clock->pcpu_buf, 0));
 }
 
-void bch2_io_clock_schedule_timeout(struct io_clock *, unsigned long);
+void bch2_io_clock_schedule_timeout(struct io_clock *, u64);
 
 #define bch2_kthread_wait_event_ioclock_timeout(condition, clock, timeout)\
 ({									\
diff --git a/libbcachefs/clock_types.h b/libbcachefs/clock_types.h
index 5fae0012..9c25d0fc 100644
--- a/libbcachefs/clock_types.h
+++ b/libbcachefs/clock_types.h
@@ -17,7 +17,8 @@ typedef void (*io_timer_fn)(struct io_timer *);
 
 struct io_timer {
 	io_timer_fn		fn;
-	unsigned long		expire;
+	void			*fn2;
+	u64			expire;
 };
 
 /* Amount to buffer up on a percpu counter */
diff --git a/libbcachefs/data_update.c b/libbcachefs/data_update.c
index 1a0072ee..0087b855 100644
--- a/libbcachefs/data_update.c
+++ b/libbcachefs/data_update.c
@@ -5,7 +5,9 @@
 #include "bkey_buf.h"
 #include "btree_update.h"
 #include "buckets.h"
+#include "compress.h"
 #include "data_update.h"
+#include "disk_groups.h"
 #include "ec.h"
 #include "error.h"
 #include "extents.h"
@@ -454,6 +456,38 @@ static void bch2_update_unwritten_extent(struct btree_trans *trans,
 	}
 }
 
+void bch2_data_update_opts_to_text(struct printbuf *out, struct bch_fs *c,
+				   struct bch_io_opts *io_opts,
+				   struct data_update_opts *data_opts)
+{
+	printbuf_tabstop_push(out, 20);
+	prt_str(out, "rewrite ptrs:\t");
+	bch2_prt_u64_base2(out, data_opts->rewrite_ptrs);
+	prt_newline(out);
+
+	prt_str(out, "kill ptrs:\t");
+	bch2_prt_u64_base2(out, data_opts->kill_ptrs);
+	prt_newline(out);
+
+	prt_str(out, "target:\t");
+	bch2_target_to_text(out, c, data_opts->target);
+	prt_newline(out);
+
+	prt_str(out, "compression:\t");
+	bch2_compression_opt_to_text(out, background_compression(*io_opts));
+	prt_newline(out);
+
+	prt_str(out, "extra replicas:\t");
+	prt_u64(out, data_opts->extra_replicas);
+}
+
+void bch2_data_update_to_text(struct printbuf *out, struct data_update *m)
+{
+	bch2_bkey_val_to_text(out, m->op.c, bkey_i_to_s_c(m->k.k));
+	prt_newline(out);
+	bch2_data_update_opts_to_text(out, m->op.c, &m->op.opts, &m->data_opts);
+}
+
 int bch2_extent_drop_ptrs(struct btree_trans *trans,
 			  struct btree_iter *iter,
 			  struct bkey_s_c k,
@@ -643,6 +677,16 @@ int bch2_data_update_init(struct btree_trans *trans,
 	if (!(durability_have + durability_removing))
 		m->op.nr_replicas = max((unsigned) m->op.nr_replicas, 1);
 
+	if (!m->op.nr_replicas) {
+		struct printbuf buf = PRINTBUF;
+
+		bch2_data_update_to_text(&buf, m);
+		WARN(1, "trying to move an extent, but nr_replicas=0\n%s", buf.buf);
+		printbuf_exit(&buf);
+		ret = -BCH_ERR_data_update_done;
+		goto done;
+	}
+
 	m->op.nr_replicas_required = m->op.nr_replicas;
 
 	if (reserve_sectors) {
diff --git a/libbcachefs/data_update.h b/libbcachefs/data_update.h
index 991095bb..8d36365b 100644
--- a/libbcachefs/data_update.h
+++ b/libbcachefs/data_update.h
@@ -17,6 +17,9 @@ struct data_update_opts {
 	unsigned	write_flags;
 };
 
+void bch2_data_update_opts_to_text(struct printbuf *, struct bch_fs *,
+				   struct bch_io_opts *, struct data_update_opts *);
+
 struct data_update {
 	/* extent being updated: */
 	enum btree_id		btree_id;
@@ -27,6 +30,8 @@ struct data_update {
 	struct bch_write_op	op;
 };
 
+void bch2_data_update_to_text(struct printbuf *, struct data_update *);
+
 int bch2_data_update_index_update(struct bch_write_op *);
 
 void bch2_data_update_read_done(struct data_update *,
diff --git a/libbcachefs/debug.c b/libbcachefs/debug.c
index f0d4727c..ebabab17 100644
--- a/libbcachefs/debug.c
+++ b/libbcachefs/debug.c
@@ -610,7 +610,7 @@ restart:
 	list_sort(&c->btree_trans_list, list_ptr_order_cmp);
 
 	list_for_each_entry(trans, &c->btree_trans_list, list) {
-		if ((ulong) trans < i->iter)
+		if ((ulong) trans <= i->iter)
 			continue;
 
 		i->iter = (ulong) trans;
@@ -832,16 +832,16 @@ static const struct file_operations btree_transaction_stats_op = {
 static void btree_deadlock_to_text(struct printbuf *out, struct bch_fs *c)
 {
 	struct btree_trans *trans;
-	pid_t iter = 0;
+	ulong iter = 0;
 restart:
 	seqmutex_lock(&c->btree_trans_lock);
-	list_for_each_entry(trans, &c->btree_trans_list, list) {
-		struct task_struct *task = READ_ONCE(trans->locking_wait.task);
+	list_sort(&c->btree_trans_list, list_ptr_order_cmp);
 
-		if (!task || task->pid <= iter)
+	list_for_each_entry(trans, &c->btree_trans_list, list) {
+		if ((ulong) trans <= iter)
 			continue;
 
-		iter = task->pid;
+		iter = (ulong) trans;
 
 		if (!closure_get_not_zero(&trans->ref))
 			continue;
diff --git a/libbcachefs/extents.c b/libbcachefs/extents.c
index 057df38f..07973198 100644
--- a/libbcachefs/extents.c
+++ b/libbcachefs/extents.c
@@ -37,8 +37,8 @@ static void bch2_extent_crc_pack(union bch_extent_crc *,
 				 struct bch_extent_crc_unpacked,
 				 enum bch_extent_entry_type);
 
-static struct bch_dev_io_failures *dev_io_failures(struct bch_io_failures *f,
-						   unsigned dev)
+struct bch_dev_io_failures *bch2_dev_io_failures(struct bch_io_failures *f,
+						 unsigned dev)
 {
 	struct bch_dev_io_failures *i;
 
@@ -52,7 +52,7 @@ static struct bch_dev_io_failures *dev_io_failures(struct bch_io_failures *f,
 void bch2_mark_io_failure(struct bch_io_failures *failed,
 			  struct extent_ptr_decoded *p)
 {
-	struct bch_dev_io_failures *f = dev_io_failures(failed, p->ptr.dev);
+	struct bch_dev_io_failures *f = bch2_dev_io_failures(failed, p->ptr.dev);
 
 	if (!f) {
 		BUG_ON(failed->nr >= ARRAY_SIZE(failed->devs));
@@ -140,7 +140,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
 		if (p.ptr.cached && (!ca || dev_ptr_stale_rcu(ca, &p.ptr)))
 			continue;
 
-		f = failed ? dev_io_failures(failed, p.ptr.dev) : NULL;
+		f = failed ? bch2_dev_io_failures(failed, p.ptr.dev) : NULL;
 		if (f)
 			p.idx = f->nr_failed < f->nr_retries
 				? f->idx
diff --git a/libbcachefs/extents.h b/libbcachefs/extents.h
index 530686aa..facdb8a8 100644
--- a/libbcachefs/extents.h
+++ b/libbcachefs/extents.h
@@ -399,6 +399,8 @@ out:									\
 
 /* utility code common to all keys with pointers: */
 
+struct bch_dev_io_failures *bch2_dev_io_failures(struct bch_io_failures *,
+						 unsigned);
 void bch2_mark_io_failure(struct bch_io_failures *,
 			  struct extent_ptr_decoded *);
 int bch2_bkey_pick_read_device(struct bch_fs *, struct bkey_s_c,
diff --git a/libbcachefs/eytzinger.h b/libbcachefs/eytzinger.h
index c6e078f2..0541192d 100644
--- a/libbcachefs/eytzinger.h
+++ b/libbcachefs/eytzinger.h
@@ -48,7 +48,7 @@ static inline unsigned eytzinger1_right_child(unsigned i)
 
 static inline unsigned eytzinger1_first(unsigned size)
 {
-	return rounddown_pow_of_two(size);
+	return size ? rounddown_pow_of_two(size) : 0;
 }
 
 static inline unsigned eytzinger1_last(unsigned size)
@@ -101,7 +101,9 @@ static inline unsigned eytzinger1_prev(unsigned i, unsigned size)
 
 static inline unsigned eytzinger1_extra(unsigned size)
 {
-	return (size + 1 - rounddown_pow_of_two(size)) << 1;
+	return size
+		? (size + 1 - rounddown_pow_of_two(size)) << 1
+		: 0;
 }
 
 static inline unsigned __eytzinger1_to_inorder(unsigned i, unsigned size,
diff --git a/libbcachefs/fs-io.c b/libbcachefs/fs-io.c
index 189d408e..77b85da3 100644
--- a/libbcachefs/fs-io.c
+++ b/libbcachefs/fs-io.c
@@ -515,7 +515,7 @@ static int inode_update_times_fn(struct btree_trans *trans,
 	return 0;
 }
 
-static long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len)
+static noinline long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	u64 end		= offset + len;
@@ -554,7 +554,7 @@ err:
 	return ret;
 }
 
-static long bchfs_fcollapse_finsert(struct bch_inode_info *inode,
+static noinline long bchfs_fcollapse_finsert(struct bch_inode_info *inode,
 				   loff_t offset, loff_t len,
 				   bool insert)
 {
@@ -590,7 +590,7 @@ static long bchfs_fcollapse_finsert(struct bch_inode_info *inode,
 	return ret;
 }
 
-static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
+static noinline int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
 			     u64 start_sector, u64 end_sector)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
@@ -711,7 +711,7 @@ bkey_err:
 	return ret;
 }
 
-static long bchfs_fallocate(struct bch_inode_info *inode, int mode,
+static noinline long bchfs_fallocate(struct bch_inode_info *inode, int mode,
 			    loff_t offset, loff_t len)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
diff --git a/libbcachefs/fs.c b/libbcachefs/fs.c
index d34d628f..3ea8dbc4 100644
--- a/libbcachefs/fs.c
+++ b/libbcachefs/fs.c
@@ -194,6 +194,12 @@ static struct bch_inode_info *bch2_inode_insert(struct bch_fs *c, struct bch_ino
 		 * discard_new_inode() expects it to be set...
 		 */
 		inode->v.i_flags |= I_NEW;
+		/*
+		 * We don't want bch2_evict_inode() to delete the inode on disk,
+		 * we just raced and had another inode in cache. Normally new
+		 * inodes don't have nlink == 0 - except tmpfiles do...
+		 */
+		set_nlink(&inode->v, 1);
 		discard_new_inode(&inode->v);
 		inode = old;
 	} else {
@@ -511,11 +517,11 @@ static int __bch2_link(struct bch_fs *c,
 		       struct bch_inode_info *dir,
 		       struct dentry *dentry)
 {
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct bch_inode_unpacked dir_u, inode_u;
 	int ret;
 
 	mutex_lock(&inode->ei_update_lock);
+	struct btree_trans *trans = bch2_trans_get(c);
 
 	ret = commit_do(trans, NULL, NULL, 0,
 			bch2_link_trans(trans,
@@ -562,11 +568,12 @@ int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
 	struct bch_inode_info *dir = to_bch_ei(vdir);
 	struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
 	struct bch_inode_unpacked dir_u, inode_u;
-	struct btree_trans *trans = bch2_trans_get(c);
 	int ret;
 
 	bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
 
+	struct btree_trans *trans = bch2_trans_get(c);
+
 	ret = commit_do(trans, NULL, NULL,
 			BCH_TRANS_COMMIT_no_enospc,
 		bch2_unlink_trans(trans,
@@ -589,8 +596,8 @@ int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
 		set_nlink(&inode->v, 0);
 	}
 err:
-	bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
 	bch2_trans_put(trans);
+	bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
 
 	return ret;
 }
@@ -675,14 +682,14 @@ static int bch2_rename2(struct mnt_idmap *idmap,
 			return ret;
 	}
 
-	trans = bch2_trans_get(c);
-
 	bch2_lock_inodes(INODE_UPDATE_LOCK,
 			 src_dir,
 			 dst_dir,
 			 src_inode,
 			 dst_inode);
 
+	trans = bch2_trans_get(c);
+
 	ret   = bch2_subvol_is_ro_trans(trans, src_dir->ei_subvol) ?:
 		bch2_subvol_is_ro_trans(trans, dst_dir->ei_subvol);
 	if (ret)
@@ -2019,6 +2026,8 @@ out:
 err:
 	darray_exit(&devs_to_fs);
 	bch2_darray_str_exit(&devs);
+	if (ret)
+		pr_err("error: %s", bch2_err_str(ret));
 	/*
 	 * On an inconsistency error in recovery we might see an -EROFS derived
 	 * errorcode (from the journal), but we don't want to return that to
@@ -2131,7 +2140,8 @@ int __init bch2_vfs_init(void)
 {
 	int ret = -ENOMEM;
 
-	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
+	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT |
+				      SLAB_ACCOUNT);
 	if (!bch2_inode_cache)
 		goto err;
 
diff --git a/libbcachefs/io_read.c b/libbcachefs/io_read.c
index 2a5c4371..8b484c75 100644
--- a/libbcachefs/io_read.c
+++ b/libbcachefs/io_read.c
@@ -93,21 +93,24 @@ static const struct rhashtable_params bch_promote_params = {
 static inline int should_promote(struct bch_fs *c, struct bkey_s_c k,
 				  struct bpos pos,
 				  struct bch_io_opts opts,
-				  unsigned flags)
+				  unsigned flags,
+				  struct bch_io_failures *failed)
 {
-	BUG_ON(!opts.promote_target);
+	if (!failed) {
+		BUG_ON(!opts.promote_target);
 
-	if (!(flags & BCH_READ_MAY_PROMOTE))
-		return -BCH_ERR_nopromote_may_not;
+		if (!(flags & BCH_READ_MAY_PROMOTE))
+			return -BCH_ERR_nopromote_may_not;
 
-	if (bch2_bkey_has_target(c, k, opts.promote_target))
-		return -BCH_ERR_nopromote_already_promoted;
+		if (bch2_bkey_has_target(c, k, opts.promote_target))
+			return -BCH_ERR_nopromote_already_promoted;
 
-	if (bkey_extent_is_unwritten(k))
-		return -BCH_ERR_nopromote_unwritten;
+		if (bkey_extent_is_unwritten(k))
+			return -BCH_ERR_nopromote_unwritten;
 
-	if (bch2_target_congested(c, opts.promote_target))
-		return -BCH_ERR_nopromote_congested;
+		if (bch2_target_congested(c, opts.promote_target))
+			return -BCH_ERR_nopromote_congested;
+	}
 
 	if (rhashtable_lookup_fast(&c->promote_table, &pos,
 				   bch_promote_params))
@@ -164,7 +167,8 @@ static struct promote_op *__promote_alloc(struct btree_trans *trans,
 					  struct extent_ptr_decoded *pick,
 					  struct bch_io_opts opts,
 					  unsigned sectors,
-					  struct bch_read_bio **rbio)
+					  struct bch_read_bio **rbio,
+					  struct bch_io_failures *failed)
 {
 	struct bch_fs *c = trans->c;
 	struct promote_op *op = NULL;
@@ -217,14 +221,28 @@ static struct promote_op *__promote_alloc(struct btree_trans *trans,
 	bio = &op->write.op.wbio.bio;
 	bio_init(bio, NULL, bio->bi_inline_vecs, pages, 0);
 
+	struct data_update_opts update_opts = {};
+
+	if (!failed) {
+		update_opts.target = opts.promote_target;
+		update_opts.extra_replicas = 1;
+		update_opts.write_flags = BCH_WRITE_ALLOC_NOWAIT|BCH_WRITE_CACHED;
+	} else {
+		update_opts.target = opts.foreground_target;
+
+		struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+		unsigned i = 0;
+		bkey_for_each_ptr(ptrs, ptr) {
+			if (bch2_dev_io_failures(failed, ptr->dev))
+				update_opts.rewrite_ptrs |= BIT(i);
+			i++;
+		}
+	}
+
 	ret = bch2_data_update_init(trans, NULL, NULL, &op->write,
 			writepoint_hashed((unsigned long) current),
 			opts,
-			(struct data_update_opts) {
-				.target		= opts.promote_target,
-				.extra_replicas	= 1,
-				.write_flags	= BCH_WRITE_ALLOC_NOWAIT|BCH_WRITE_CACHED,
-			},
+			update_opts,
 			btree_id, k);
 	/*
 	 * possible errors: -BCH_ERR_nocow_lock_blocked,
@@ -258,10 +276,17 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
 					unsigned flags,
 					struct bch_read_bio **rbio,
 					bool *bounce,
-					bool *read_full)
+					bool *read_full,
+					struct bch_io_failures *failed)
 {
 	struct bch_fs *c = trans->c;
-	bool promote_full = *read_full || READ_ONCE(c->promote_whole_extents);
+	/*
+	 * if failed != NULL we're not actually doing a promote, we're
+	 * recovering from an io/checksum error
+	 */
+	bool promote_full = (failed ||
+			     *read_full ||
+			     READ_ONCE(c->promote_whole_extents));
 	/* data might have to be decompressed in the write path: */
 	unsigned sectors = promote_full
 		? max(pick->crc.compressed_size, pick->crc.live_size)
@@ -272,7 +297,7 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
 	struct promote_op *promote;
 	int ret;
 
-	ret = should_promote(c, k, pos, opts, flags);
+	ret = should_promote(c, k, pos, opts, flags, failed);
 	if (ret)
 		goto nopromote;
 
@@ -280,7 +305,7 @@ static struct promote_op *promote_alloc(struct btree_trans *trans,
 				  k.k->type == KEY_TYPE_reflink_v
 				  ? BTREE_ID_reflink
 				  : BTREE_ID_extents,
-				  k, pos, pick, opts, sectors, rbio);
+				  k, pos, pick, opts, sectors, rbio, failed);
 	ret = PTR_ERR_OR_ZERO(promote);
 	if (ret)
 		goto nopromote;
@@ -910,9 +935,9 @@ retry_pick:
 		bounce = true;
 	}
 
-	if (orig->opts.promote_target)
+	if (orig->opts.promote_target)// || failed)
 		promote = promote_alloc(trans, iter, k, &pick, orig->opts, flags,
-					&rbio, &bounce, &read_full);
+					&rbio, &bounce, &read_full, failed);
 
 	if (!read_full) {
 		EBUG_ON(crc_is_compressed(pick.crc));
@@ -1003,6 +1028,9 @@ get_bio:
 	rbio->promote		= promote;
 	INIT_WORK(&rbio->work, NULL);
 
+	if (flags & BCH_READ_NODECODE)
+		orig->pick = pick;
+
 	rbio->bio.bi_opf	= orig->bio.bi_opf;
 	rbio->bio.bi_iter.bi_sector = pick.ptr.offset;
 	rbio->bio.bi_end_io	= bch2_read_endio;
diff --git a/libbcachefs/journal.c b/libbcachefs/journal.c
index 6209d778..649e3a01 100644
--- a/libbcachefs/journal.c
+++ b/libbcachefs/journal.c
@@ -1094,7 +1094,7 @@ unlock:
 	return ret;
 }
 
-int bch2_dev_journal_alloc(struct bch_dev *ca)
+int bch2_dev_journal_alloc(struct bch_dev *ca, bool new_fs)
 {
 	unsigned nr;
 	int ret;
@@ -1116,7 +1116,7 @@ int bch2_dev_journal_alloc(struct bch_dev *ca)
 		     min(1 << 13,
 			 (1 << 24) / ca->mi.bucket_size));
 
-	ret = __bch2_set_nr_journal_buckets(ca, nr, true, NULL);
+	ret = __bch2_set_nr_journal_buckets(ca, nr, new_fs, NULL);
 err:
 	bch_err_fn(ca, ret);
 	return ret;
@@ -1128,7 +1128,7 @@ int bch2_fs_journal_alloc(struct bch_fs *c)
 		if (ca->journal.nr)
 			continue;
 
-		int ret = bch2_dev_journal_alloc(ca);
+		int ret = bch2_dev_journal_alloc(ca, true);
 		if (ret) {
 			percpu_ref_put(&ca->io_ref);
 			return ret;
@@ -1183,9 +1183,11 @@ void bch2_fs_journal_stop(struct journal *j)
 	journal_quiesce(j);
 	cancel_delayed_work_sync(&j->write_work);
 
-	BUG_ON(!bch2_journal_error(j) &&
-	       test_bit(JOURNAL_replay_done, &j->flags) &&
-	       j->last_empty_seq != journal_cur_seq(j));
+	WARN(!bch2_journal_error(j) &&
+	     test_bit(JOURNAL_replay_done, &j->flags) &&
+	     j->last_empty_seq != journal_cur_seq(j),
+	     "journal shutdown error: cur seq %llu but last empty seq %llu",
+	     journal_cur_seq(j), j->last_empty_seq);
 
 	if (!bch2_journal_error(j))
 		clear_bit(JOURNAL_running, &j->flags);
@@ -1417,8 +1419,8 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
 	unsigned long now = jiffies;
 	u64 nr_writes = j->nr_flush_writes + j->nr_noflush_writes;
 
-	if (!out->nr_tabstops)
-		printbuf_tabstop_push(out, 28);
+	printbuf_tabstops_reset(out);
+	printbuf_tabstop_push(out, 28);
 	out->atomic++;
 
 	rcu_read_lock();
diff --git a/libbcachefs/journal.h b/libbcachefs/journal.h
index 00d0f4fc..377a3750 100644
--- a/libbcachefs/journal.h
+++ b/libbcachefs/journal.h
@@ -433,7 +433,7 @@ bool bch2_journal_seq_pins_to_text(struct printbuf *, struct journal *, u64 *);
 
 int bch2_set_nr_journal_buckets(struct bch_fs *, struct bch_dev *,
 				unsigned nr);
-int bch2_dev_journal_alloc(struct bch_dev *);
+int bch2_dev_journal_alloc(struct bch_dev *, bool);
 int bch2_fs_journal_alloc(struct bch_fs *);
 
 void bch2_dev_journal_stop(struct journal *, struct bch_dev *);
diff --git a/libbcachefs/journal_io.c b/libbcachefs/journal_io.c
index ff832d20..7a833a3f 100644
--- a/libbcachefs/journal_io.c
+++ b/libbcachefs/journal_io.c
@@ -415,6 +415,8 @@ static int journal_entry_btree_keys_validate(struct bch_fs *c,
 					       flags|BCH_VALIDATE_journal);
 		if (ret == FSCK_DELETED_KEY)
 			continue;
+		else if (ret)
+			return ret;
 
 		k = bkey_next(k);
 	}
@@ -1766,11 +1768,13 @@ static CLOSURE_CALLBACK(journal_write_preflush)
 
 	if (j->seq_ondisk + 1 != le64_to_cpu(w->data->seq)) {
 		spin_lock(&j->lock);
-		closure_wait(&j->async_wait, cl);
+		if (j->seq_ondisk + 1 != le64_to_cpu(w->data->seq)) {
+			closure_wait(&j->async_wait, cl);
+			spin_unlock(&j->lock);
+			continue_at(cl, journal_write_preflush, j->wq);
+			return;
+		}
 		spin_unlock(&j->lock);
-
-		continue_at(cl, journal_write_preflush, j->wq);
-		return;
 	}
 
 	if (w->separate_flush) {
diff --git a/libbcachefs/lru.c b/libbcachefs/lru.c
index c535ba27..83b1586c 100644
--- a/libbcachefs/lru.c
+++ b/libbcachefs/lru.c
@@ -77,6 +77,45 @@ static const char * const bch2_lru_types[] = {
 	NULL
 };
 
+int bch2_lru_check_set(struct btree_trans *trans,
+		       u16 lru_id, u64 time,
+		       struct bkey_s_c referring_k,
+		       struct bkey_buf *last_flushed)
+{
+	struct bch_fs *c = trans->c;
+	struct printbuf buf = PRINTBUF;
+	struct btree_iter lru_iter;
+	struct bkey_s_c lru_k =
+		bch2_bkey_get_iter(trans, &lru_iter, BTREE_ID_lru,
+				   lru_pos(lru_id,
+					   bucket_to_u64(referring_k.k->p),
+					   time), 0);
+	int ret = bkey_err(lru_k);
+	if (ret)
+		return ret;
+
+	if (lru_k.k->type != KEY_TYPE_set) {
+		ret = bch2_btree_write_buffer_maybe_flush(trans, referring_k, last_flushed);
+		if (ret)
+			goto err;
+
+		if (fsck_err(trans, alloc_key_to_missing_lru_entry,
+			     "missing %s lru entry\n"
+			     "  %s",
+			     bch2_lru_types[lru_type(lru_k)],
+			     (bch2_bkey_val_to_text(&buf, c, referring_k), buf.buf))) {
+			ret = bch2_lru_set(trans, lru_id, bucket_to_u64(referring_k.k->p), time);
+			if (ret)
+				goto err;
+		}
+	}
+err:
+fsck_err:
+	bch2_trans_iter_exit(trans, &lru_iter);
+	printbuf_exit(&buf);
+	return ret;
+}
+
 static int bch2_check_lru_key(struct btree_trans *trans,
 			      struct btree_iter *lru_iter,
 			      struct bkey_s_c lru_k,
diff --git a/libbcachefs/lru.h b/libbcachefs/lru.h
index 425ba732..5bd8974a 100644
--- a/libbcachefs/lru.h
+++ b/libbcachefs/lru.h
@@ -49,6 +49,9 @@ int bch2_lru_del(struct btree_trans *, u16, u64, u64);
 int bch2_lru_set(struct btree_trans *, u16, u64, u64);
 int bch2_lru_change(struct btree_trans *, u16, u64, u64, u64);
 
+struct bkey_buf;
+int bch2_lru_check_set(struct btree_trans *, u16, u64, struct bkey_s_c, struct bkey_buf *);
+
 int bch2_check_lrus(struct bch_fs *);
 
 #endif /* _BCACHEFS_LRU_H */
diff --git a/libbcachefs/move.c b/libbcachefs/move.c
index 60e4ef94..7d3920e0 100644
--- a/libbcachefs/move.c
+++ b/libbcachefs/move.c
@@ -36,31 +36,6 @@ const char * const bch2_data_ops_strs[] = {
 	NULL
 };
 
-static void bch2_data_update_opts_to_text(struct printbuf *out, struct bch_fs *c,
-					  struct bch_io_opts *io_opts,
-					  struct data_update_opts *data_opts)
-{
-	printbuf_tabstop_push(out, 20);
-	prt_str(out, "rewrite ptrs:\t");
-	bch2_prt_u64_base2(out, data_opts->rewrite_ptrs);
-	prt_newline(out);
-
-	prt_str(out, "kill ptrs:\t");
-	bch2_prt_u64_base2(out, data_opts->kill_ptrs);
-	prt_newline(out);
-
-	prt_str(out, "target:\t");
-	bch2_target_to_text(out, c, data_opts->target);
-	prt_newline(out);
-
-	prt_str(out, "compression:\t");
-	bch2_compression_opt_to_text(out, background_compression(*io_opts));
-	prt_newline(out);
-
-	prt_str(out, "extra replicas:\t");
-	prt_u64(out, data_opts->extra_replicas);
-}
-
 static void trace_move_extent2(struct bch_fs *c, struct bkey_s_c k,
 			       struct bch_io_opts *io_opts,
 			       struct data_update_opts *data_opts)
diff --git a/libbcachefs/movinggc.c b/libbcachefs/movinggc.c
index eb49dd04..f7aa835d 100644
--- a/libbcachefs/movinggc.c
+++ b/libbcachefs/movinggc.c
@@ -290,18 +290,23 @@ unsigned long bch2_copygc_wait_amount(struct bch_fs *c)
 
 void bch2_copygc_wait_to_text(struct printbuf *out, struct bch_fs *c)
 {
-	prt_printf(out, "Currently waiting for:     ");
+	printbuf_tabstop_push(out, 32);
+	prt_printf(out, "running:\t%u\n",		c->copygc_running);
+	prt_printf(out, "copygc_wait:\t%llu\n",		c->copygc_wait);
+	prt_printf(out, "copygc_wait_at:\t%llu\n",	c->copygc_wait_at);
+
+	prt_printf(out, "Currently waiting for:\t");
 	prt_human_readable_u64(out, max(0LL, c->copygc_wait -
 					atomic64_read(&c->io_clock[WRITE].now)) << 9);
 	prt_newline(out);
 
-	prt_printf(out, "Currently waiting since:   ");
+	prt_printf(out, "Currently waiting since:\t");
 	prt_human_readable_u64(out, max(0LL,
 					atomic64_read(&c->io_clock[WRITE].now) -
 					c->copygc_wait_at) << 9);
 	prt_newline(out);
 
-	prt_printf(out, "Currently calculated wait: ");
+	prt_printf(out, "Currently calculated wait:\t");
 	prt_human_readable_u64(out, bch2_copygc_wait_amount(c));
 	prt_newline(out);
 }
@@ -352,19 +357,18 @@ static int bch2_copygc_thread(void *arg)
 		}
 
 		last = atomic64_read(&clock->now);
-		wait = bch2_copygc_wait_amount(c);
+		wait = max_t(long, 0, bch2_copygc_wait_amount(c) - clock->max_slop);
 
-		if (wait > clock->max_slop) {
+		if (wait > 0) {
 			c->copygc_wait_at = last;
 			c->copygc_wait = last + wait;
 			move_buckets_wait(&ctxt, buckets, true);
-			trace_and_count(c, copygc_wait, c, wait, last + wait);
-			bch2_kthread_io_clock_wait(clock, last + wait,
-					MAX_SCHEDULE_TIMEOUT);
+			trace_and_count(c, copygc_wait, c, wait, c->copygc_wait);
+			bch2_io_clock_schedule_timeout(clock, c->copygc_wait);
 			continue;
 		}
 
-		c->copygc_wait = 0;
+		c->copygc_wait = c->copygc_wait_at = 0;
 
 		c->copygc_running = true;
 		ret = bch2_copygc(&ctxt, buckets, &did_work);
@@ -379,8 +383,7 @@ static int bch2_copygc_thread(void *arg)
 				min_member_capacity = 128 * 2048;
 
 			bch2_trans_unlock_long(ctxt.trans);
-			bch2_kthread_io_clock_wait(clock, last + (min_member_capacity >> 6),
-					MAX_SCHEDULE_TIMEOUT);
+			bch2_io_clock_schedule_timeout(clock, last + (min_member_capacity >> 8));
 		}
 	}
 
@@ -396,9 +399,10 @@ static int bch2_copygc_thread(void *arg)
 
 void bch2_copygc_stop(struct bch_fs *c)
 {
-	if (c->copygc_thread) {
-		kthread_stop(c->copygc_thread);
-		put_task_struct(c->copygc_thread);
+	struct task_struct *t = rcu_dereference_protected(c->copygc_thread, true);
+	if (t) {
+		kthread_stop(t);
+		put_task_struct(t);
 	}
 	c->copygc_thread = NULL;
 }
@@ -425,8 +429,8 @@ int bch2_copygc_start(struct bch_fs *c)
 
 	get_task_struct(t);
 
-	c->copygc_thread = t;
-	wake_up_process(c->copygc_thread);
+	rcu_assign_pointer(c->copygc_thread, t);
+	wake_up_process(t);
 
 	return 0;
 }
diff --git a/libbcachefs/sb-errors_format.h b/libbcachefs/sb-errors_format.h
index 67648b77..d1b2f2aa 100644
--- a/libbcachefs/sb-errors_format.h
+++ b/libbcachefs/sb-errors_format.h
@@ -286,7 +286,8 @@ enum bch_fsck_flags {
 	x(accounting_mismatch,					272,	0)		\
 	x(accounting_replicas_not_marked,			273,	0)		\
 	x(invalid_btree_id,					274,	0)		\
-	x(alloc_key_io_time_bad,				275,	0)
+	x(alloc_key_io_time_bad,				275,	0)		\
+	x(alloc_key_fragmentation_lru_wrong,			276,	FSCK_AUTOFIX)
 
 enum bch_sb_error_id {
 #define x(t, n, ...) BCH_FSCK_ERR_##t = n,
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index ced63397..0455a100 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -968,7 +968,7 @@ static void print_mount_opts(struct bch_fs *c)
 	struct printbuf p = PRINTBUF;
 	bool first = true;
 
-	prt_str(&p, "mounting version ");
+	prt_str(&p, "starting version ");
 	bch2_version_to_text(&p, c->sb.version);
 
 	if (c->opts.read_only) {
@@ -1766,7 +1766,7 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
 	if (ret)
 		goto err;
 
-	ret = bch2_dev_journal_alloc(ca);
+	ret = bch2_dev_journal_alloc(ca, true);
 	bch_err_msg(c, ret, "allocating journal");
 	if (ret)
 		goto err;
@@ -1928,7 +1928,7 @@ int bch2_dev_online(struct bch_fs *c, const char *path)
 	}
 
 	if (!ca->journal.nr) {
-		ret = bch2_dev_journal_alloc(ca);
+		ret = bch2_dev_journal_alloc(ca, false);
 		bch_err_msg(ca, ret, "allocating journal");
 		if (ret)
 			goto err;
diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c
index 8df52a23..7c9cafdc 100644
--- a/libbcachefs/sysfs.c
+++ b/libbcachefs/sysfs.c
@@ -141,6 +141,7 @@ write_attribute(trigger_gc);
 write_attribute(trigger_discards);
 write_attribute(trigger_invalidates);
 write_attribute(trigger_journal_flush);
+write_attribute(trigger_journal_writes);
 write_attribute(trigger_btree_cache_shrink);
 write_attribute(trigger_btree_key_cache_shrink);
 rw_attribute(gc_gens_pos);
@@ -471,6 +472,9 @@ STORE(bch2_fs)
 		bch2_journal_meta(&c->journal);
 	}
 
+	if (attr == &sysfs_trigger_journal_writes)
+		bch2_journal_do_writes(&c->journal);
+
 #ifdef CONFIG_BCACHEFS_TESTS
 	if (attr == &sysfs_perf_test) {
 		char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
@@ -589,6 +593,7 @@ struct attribute *bch2_fs_internal_files[] = {
 	&sysfs_trigger_discards,
 	&sysfs_trigger_invalidates,
 	&sysfs_trigger_journal_flush,
+	&sysfs_trigger_journal_writes,
 	&sysfs_trigger_btree_cache_shrink,
 	&sysfs_trigger_btree_key_cache_shrink,
 
diff --git a/linux/closure.c b/linux/closure.c
index c971216d..116afae2 100644
--- a/linux/closure.c
+++ b/linux/closure.c
@@ -244,6 +244,9 @@ void closure_debug_destroy(struct closure *cl)
 {
 	unsigned long flags;
 
+	if (cl->magic == CLOSURE_MAGIC_STACK)
+		return;
+
 	BUG_ON(cl->magic != CLOSURE_MAGIC_ALIVE);
 	cl->magic = CLOSURE_MAGIC_DEAD;