Merge tag 'bcachefs-2024-07-10' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:

 - Switch some asserts to WARN()

 - Fix a few "transaction not locked" asserts in the data read retry
   paths and backpointers gc

 - Fix a race that would cause the journal to get stuck on a flush
   commit

 - Add missing fsck checks for the fragmentation LRU

 - The usual assorted ssorted syzbot fixes

* tag 'bcachefs-2024-07-10' of https://evilpiepirate.org/git/bcachefs: (22 commits)
  bcachefs: Add missing bch2_trans_begin()
  bcachefs: Fix missing error check in journal_entry_btree_keys_validate()
  bcachefs: Warn on attempting a move with no replicas
  bcachefs: bch2_data_update_to_text()
  bcachefs: Log mount failure error code
  bcachefs: Fix undefined behaviour in eytzinger1_first()
  bcachefs: Mark bch_inode_info as SLAB_ACCOUNT
  bcachefs: Fix bch2_inode_insert() race path for tmpfiles
  closures: fix closure_sync + closure debugging
  bcachefs: Fix journal getting stuck on a flush commit
  bcachefs: io clock: run timer fns under clock lock
  bcachefs: Repair fragmentation_lru in alloc_write_key()
  bcachefs: add check for missing fragmentation in check_alloc_to_lru_ref()
  bcachefs: bch2_btree_write_buffer_maybe_flush()
  bcachefs: Add missing printbuf_tabstops_reset() calls
  bcachefs: Fix loop restart in bch2_btree_transactions_read()
  bcachefs: Fix bch2_read_retry_nodecode()
  bcachefs: Don't use the new_fs() bucket alloc path on an initialized fs
  bcachefs: Fix shift greater than integer size
  bcachefs: Change bch2_fs_journal_stop() BUG_ON() to warning
  ...

23 files changed, 256 insertions, 142 deletions

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 1de9fac3bcf4..658f11aebda1 100644
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/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"
@@ -1553,13 +1554,13 @@ 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;
 
@@ -1573,6 +1574,14 @@ 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;
 
@@ -1597,41 +1606,30 @@ 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, c,
-			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/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c
index 9d3d64746a5b..27d97c22ae27 100644
--- a/fs/bcachefs/alloc_foreground.c
+++ b/fs/bcachefs/alloc_foreground.c
@@ -1703,6 +1703,7 @@ 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);
 
 	percpu_down_read(&c->mark_lock);
@@ -1736,6 +1737,7 @@ 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/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c
index 4321f9fb73bd..6d8b1bc90be0 100644
--- a/fs/bcachefs/backpointers.c
+++ b/fs/bcachefs/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, c,
-			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(c, 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/fs/bcachefs/bkey.c b/fs/bcachefs/bkey.c
index 94a1d1982fa8..587d7318a2e8 100644
--- a/fs/bcachefs/bkey.c
+++ b/fs/bcachefs/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/fs/bcachefs/bkey.h b/fs/bcachefs/bkey.h
index fcd43915df07..936357149cf0 100644
--- a/fs/bcachefs/bkey.h
+++ b/fs/bcachefs/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/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 0e477a926579..7c72a9e6f511 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -903,6 +903,8 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 		bch2_dev_usage_update(c, ca, &old_gc, &gc, 0, true);
 	percpu_up_read(&c->mark_lock);
 
+	gc.fragmentation_lru = alloc_lru_idx_fragmentation(gc, ca);
+
 	if (fsck_err_on(new.data_type != gc.data_type, c,
 			alloc_key_data_type_wrong,
 			"bucket %llu:%llu gen %u has wrong data_type"
@@ -916,23 +918,19 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 #define copy_bucket_field(_errtype, _f)					\
 	if (fsck_err_on(new._f != gc._f, c, _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_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_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))
@@ -946,7 +944,7 @@ 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/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c
index 75c8a196b3f6..d0e92d948002 100644
--- a/fs/bcachefs/btree_write_buffer.c
+++ b/fs/bcachefs/btree_write_buffer.c
@@ -1,11 +1,13 @@
 // 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 "error.h"
+#include "extents.h"
 #include "journal.h"
 #include "journal_io.h"
 #include "journal_reclaim.h"
@@ -492,6 +494,41 @@ 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/fs/bcachefs/btree_write_buffer.h b/fs/bcachefs/btree_write_buffer.h
index eebcd2b15249..dd5e64218b50 100644
--- a/fs/bcachefs/btree_write_buffer.h
+++ b/fs/bcachefs/btree_write_buffer.h
@@ -23,6 +23,9 @@ 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/fs/bcachefs/clock.c b/fs/bcachefs/clock.c
index 363644451106..0f40b585ce2b 100644
--- a/fs/bcachefs/clock.c
+++ b/fs/bcachefs/clock.c
@@ -132,14 +132,9 @@ static struct io_timer *get_expired_timer(struct io_clock *clock,
 {
 	struct io_timer *ret = NULL;
 
-	spin_lock(&clock->timer_lock);
-
 	if (clock->timers.used &&
 	    time_after_eq(now, clock->timers.data[0]->expire))
 		heap_pop(&clock->timers, ret, io_timer_cmp, NULL);
-
-	spin_unlock(&clock->timer_lock);
-
 	return ret;
 }
 
@@ -148,8 +143,10 @@ void __bch2_increment_clock(struct io_clock *clock, unsigned sectors)
 	struct io_timer *timer;
 	unsigned long 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)
diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c
index 1a0072eef109..0087b8555ead 100644
--- a/fs/bcachefs/data_update.c
+++ b/fs/bcachefs/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/fs/bcachefs/data_update.h b/fs/bcachefs/data_update.h
index 991095bbd469..8d36365bdea8 100644
--- a/fs/bcachefs/data_update.h
+++ b/fs/bcachefs/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/fs/bcachefs/debug.c b/fs/bcachefs/debug.c
index f0d4727c4dc2..ebabab171fe5 100644
--- a/fs/bcachefs/debug.c
+++ b/fs/bcachefs/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/fs/bcachefs/eytzinger.h b/fs/bcachefs/eytzinger.h
index 24840aee335c..795f4fc0bab1 100644
--- a/fs/bcachefs/eytzinger.h
+++ b/fs/bcachefs/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/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index f9c9a95d7d4c..74a1e12a7a14 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/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 {
@@ -2026,6 +2032,8 @@ err_put_super:
 	__bch2_fs_stop(c);
 	deactivate_locked_super(sb);
 err:
+	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
@@ -2065,7 +2073,8 @@ 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/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c
index c97fa7002b06..ebf39ef72fb2 100644
--- a/fs/bcachefs/io_read.c
+++ b/fs/bcachefs/io_read.c
@@ -389,7 +389,6 @@ retry:
 
 	bch2_bkey_buf_reassemble(&sk, c, k);
 	k = bkey_i_to_s_c(sk.k);
-	bch2_trans_unlock(trans);
 
 	if (!bch2_bkey_matches_ptr(c, k,
 				   rbio->pick.ptr,
@@ -1004,6 +1003,9 @@ 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/fs/bcachefs/journal.c b/fs/bcachefs/journal.c
index 13669dd0e375..10b19791ec98 100644
--- a/fs/bcachefs/journal.c
+++ b/fs/bcachefs/journal.c
@@ -1095,7 +1095,7 @@ 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;
@@ -1117,7 +1117,7 @@ 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;
@@ -1129,7 +1129,7 @@ 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;
@@ -1184,9 +1184,11 @@ 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);
@@ -1418,8 +1420,8 @@ 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/fs/bcachefs/journal.h b/fs/bcachefs/journal.h
index fd1f7cdaa8bc..bc6b9c39dcb4 100644
--- a/fs/bcachefs/journal.h
+++ b/fs/bcachefs/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/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c
index db24ce21b2ac..2326e2cb9cd2 100644
--- a/fs/bcachefs/journal_io.c
+++ b/fs/bcachefs/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);
 	}
@@ -1762,11 +1764,13 @@ 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/fs/bcachefs/lru.c b/fs/bcachefs/lru.c
index a40d116224ed..b12894ef44f3 100644
--- a/fs/bcachefs/lru.c
+++ b/fs/bcachefs/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(c, 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/fs/bcachefs/lru.h b/fs/bcachefs/lru.h
index bd71ba77de07..ed75bcf59d47 100644
--- a/fs/bcachefs/lru.h
+++ b/fs/bcachefs/lru.h
@@ -61,6 +61,9 @@ 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/fs/bcachefs/move.c b/fs/bcachefs/move.c
index 6e477fadaa2a..e714e3bd5bbb 100644
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/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/fs/bcachefs/sb-errors_format.h b/fs/bcachefs/sb-errors_format.h
index d6f35a99c429..d54121ec093f 100644
--- a/fs/bcachefs/sb-errors_format.h
+++ b/fs/bcachefs/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/fs/bcachefs/super.c b/fs/bcachefs/super.c
index fb906467201e..da735608d47c 100644
--- a/fs/bcachefs/super.c
+++ b/fs/bcachefs/super.c
@@ -563,8 +563,11 @@ static void __bch2_fs_free(struct bch_fs *c)
 	BUG_ON(atomic_read(&c->journal_keys.ref));
 	bch2_fs_btree_write_buffer_exit(c);
 	percpu_free_rwsem(&c->mark_lock);
-	EBUG_ON(c->online_reserved && percpu_u64_get(c->online_reserved));
-	free_percpu(c->online_reserved);
+	if (c->online_reserved) {
+		u64 v = percpu_u64_get(c->online_reserved);
+		WARN(v, "online_reserved not 0 at shutdown: %lli", v);
+		free_percpu(c->online_reserved);
+	}
 
 	darray_exit(&c->btree_roots_extra);
 	free_percpu(c->pcpu);
@@ -1769,7 +1772,7 @@ 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;
@@ -1929,7 +1932,7 @@ 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;