Merge tag 'bcachefs-2024-10-05' of git://evilpiepirate.org/bcachefs

Pull bcachefs fixes from Kent Overstreet:
 "A lot of little fixes, bigger ones include:

   - bcachefs's __wait_on_freeing_inode() was broken in rc1 due to vfs
     changes, now fixed along with another lost wakeup

   - fragmentation LRU fixes; fsck now repairs successfully (this is the
     data structure copygc uses); along with some nice simplification.

   - Rework logged op error handling, so that if logged op replay errors
     (due to another filesystem error) we delete the logged op instead
     of going into an infinite loop)

   - Various small filesystem connectivitity repair fixes"

* tag 'bcachefs-2024-10-05' of git://evilpiepirate.org/bcachefs:
  bcachefs: Rework logged op error handling
  bcachefs: Add warn param to subvol_get_snapshot, peek_inode
  bcachefs: Kill snapshot arg to fsck_write_inode()
  bcachefs: Check for unlinked, non-empty dirs in check_inode()
  bcachefs: Check for unlinked inodes with dirents
  bcachefs: Check for directories with no backpointers
  bcachefs: Kill alloc_v4.fragmentation_lru
  bcachefs: minor lru fsck fixes
  bcachefs: Mark more errors AUTOFIX
  bcachefs: Make sure we print error that causes fsck to bail out
  bcachefs: bkey errors are only AUTOFIX during read
  bcachefs: Create lost+found in correct snapshot
  bcachefs: Fix reattach_inode()
  bcachefs: Add missing wakeup to bch2_inode_hash_remove()
  bcachefs: Fix trans_commit disk accounting revert
  bcachefs: Fix bch2_inode_is_open() check
  bcachefs: Fix return type of dirent_points_to_inode_nowarn()
  bcachefs: Fix bad shift in bch2_read_flag_list()

fs/bcachefs/alloc_background.c

@@ -332,7 +332,6 @@ void bch2_alloc_v4_swab(struct bkey_s k)
 	a->io_time[1]		= swab64(a->io_time[1]);
 	a->stripe		= swab32(a->stripe);
 	a->nr_external_backpointers = swab32(a->nr_external_backpointers);
-	a->fragmentation_lru	= swab64(a->fragmentation_lru);
 	a->stripe_sectors	= swab32(a->stripe_sectors);
 
 	bps = alloc_v4_backpointers(a);
@@ -347,6 +346,7 @@ void bch2_alloc_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c
 {
 	struct bch_alloc_v4 _a;
 	const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &_a);
+	struct bch_dev *ca = c ? bch2_dev_bucket_tryget_noerror(c, k.k->p) : NULL;
 
 	prt_newline(out);
 	printbuf_indent_add(out, 2);
@@ -364,9 +364,13 @@ void bch2_alloc_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c
 	prt_printf(out, "stripe_redundancy %u\n",	a->stripe_redundancy);
 	prt_printf(out, "io_time[READ]     %llu\n",	a->io_time[READ]);
 	prt_printf(out, "io_time[WRITE]    %llu\n",	a->io_time[WRITE]);
-	prt_printf(out, "fragmentation     %llu\n",	a->fragmentation_lru);
+
+	if (ca)
+		prt_printf(out, "fragmentation     %llu\n",	alloc_lru_idx_fragmentation(*a, ca));
 	prt_printf(out, "bp_start          %llu\n", BCH_ALLOC_V4_BACKPOINTERS_START(a));
 	printbuf_indent_sub(out, 2);
+
+	bch2_dev_put(ca);
 }
 
 void __bch2_alloc_to_v4(struct bkey_s_c k, struct bch_alloc_v4 *out)
@@ -882,12 +886,13 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 				goto err;
 		}
 
-		new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a, ca);
-		if (old_a->fragmentation_lru != new_a->fragmentation_lru) {
+		old_lru = alloc_lru_idx_fragmentation(*old_a, ca);
+		new_lru = alloc_lru_idx_fragmentation(*new_a, ca);
+		if (old_lru != new_lru) {
 			ret = bch2_lru_change(trans,
 					BCH_LRU_FRAGMENTATION_START,
 					bucket_to_u64(new.k->p),
-					old_a->fragmentation_lru, new_a->fragmentation_lru);
+					old_lru, new_lru);
 			if (ret)
 				goto err;
 		}
@@ -1629,18 +1634,22 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 	if (ret)
 		return ret;
 
+	struct bch_dev *ca = bch2_dev_tryget_noerror(c, alloc_k.k->p.inode);
+	if (!ca)
+		return 0;
+
 	a = bch2_alloc_to_v4(alloc_k, &a_convert);
 
-	if (a->fragmentation_lru) {
+	u64 lru_idx = alloc_lru_idx_fragmentation(*a, ca);
+	if (lru_idx) {
 		ret = bch2_lru_check_set(trans, BCH_LRU_FRAGMENTATION_START,
-					 a->fragmentation_lru,
-					 alloc_k, last_flushed);
+					 lru_idx, alloc_k, last_flushed);
 		if (ret)
-			return ret;
+			goto err;
 	}
 
 	if (a->data_type != BCH_DATA_cached)
-		return 0;
+		goto err;
 
 	if (fsck_err_on(!a->io_time[READ],
 			trans, alloc_key_cached_but_read_time_zero,
@@ -1669,6 +1678,7 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 		goto err;
 err:
 fsck_err:
+	bch2_dev_put(ca);
 	printbuf_exit(&buf);
 	return ret;
 }

fs/bcachefs/alloc_background_format.h

@@ -70,7 +70,7 @@ struct bch_alloc_v4 {
 	__u32			stripe;
 	__u32			nr_external_backpointers;
 	/* end of fields in original version of alloc_v4 */
-	__u64			fragmentation_lru;
+	__u64			_fragmentation_lru; /* obsolete */
 	__u32			stripe_sectors;
 	__u32			pad;
 } __packed __aligned(8);

fs/bcachefs/btree_gc.c

@@ -828,8 +828,6 @@ 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"
@@ -857,7 +855,6 @@ static int bch2_alloc_write_key(struct btree_trans *trans,
 	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))

fs/bcachefs/btree_trans_commit.c

@@ -832,7 +832,8 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 	for (struct jset_entry *entry2 = trans->journal_entries;
 	     entry2 != entry;
 	     entry2 = vstruct_next(entry2))
-		if (jset_entry_is_key(entry2) && entry2->start->k.type == KEY_TYPE_accounting) {
+		if (entry2->type == BCH_JSET_ENTRY_write_buffer_keys &&
+		    entry2->start->k.type == KEY_TYPE_accounting) {
 			struct bkey_s_accounting a = bkey_i_to_s_accounting(entry2->start);
 
 			bch2_accounting_neg(a);

fs/bcachefs/error.c

@@ -393,6 +393,14 @@ int __bch2_fsck_err(struct bch_fs *c,
 	     !(flags & FSCK_CAN_IGNORE)))
 		ret = -BCH_ERR_fsck_errors_not_fixed;
 
+	bool exiting =
+		test_bit(BCH_FS_fsck_running, &c->flags) &&
+		(ret != -BCH_ERR_fsck_fix &&
+		 ret != -BCH_ERR_fsck_ignore);
+
+	if (exiting)
+		print = true;
+
 	if (print) {
 		if (bch2_fs_stdio_redirect(c))
 			bch2_print(c, "%s\n", out->buf);
@@ -400,9 +408,7 @@ int __bch2_fsck_err(struct bch_fs *c,
 			bch2_print_string_as_lines(KERN_ERR, out->buf);
 	}
 
-	if (test_bit(BCH_FS_fsck_running, &c->flags) &&
-	    (ret != -BCH_ERR_fsck_fix &&
-	     ret != -BCH_ERR_fsck_ignore))
+	if (exiting)
 		bch_err(c, "Unable to continue, halting");
 	else if (suppressing)
 		bch_err(c, "Ratelimiting new instances of previous error");
@@ -430,10 +436,17 @@ int __bch2_fsck_err(struct bch_fs *c,
 
 int __bch2_bkey_fsck_err(struct bch_fs *c,
 			 struct bkey_s_c k,
-			 enum bch_fsck_flags flags,
+			 enum bch_validate_flags validate_flags,
 			 enum bch_sb_error_id err,
 			 const char *fmt, ...)
 {
+	if (validate_flags & BCH_VALIDATE_silent)
+		return -BCH_ERR_fsck_delete_bkey;
+
+	unsigned fsck_flags = 0;
+	if (!(validate_flags & (BCH_VALIDATE_write|BCH_VALIDATE_commit)))
+		fsck_flags |= FSCK_AUTOFIX|FSCK_CAN_FIX;
+
 	struct printbuf buf = PRINTBUF;
 	va_list args;
 
@@ -445,7 +458,7 @@ int __bch2_bkey_fsck_err(struct bch_fs *c,
 	va_end(args);
 	prt_str(&buf, ": delete?");
 
-	int ret = __bch2_fsck_err(c, NULL, flags, err, "%s", buf.buf);
+	int ret = __bch2_fsck_err(c, NULL, fsck_flags, err, "%s", buf.buf);
 	printbuf_exit(&buf);
 	return ret;
 }

fs/bcachefs/error.h

@@ -167,10 +167,11 @@ void bch2_flush_fsck_errs(struct bch_fs *);
 #define fsck_err_on(cond, c, _err_type, ...)				\
 	__fsck_err_on(cond, c, FSCK_CAN_FIX|FSCK_CAN_IGNORE, _err_type, __VA_ARGS__)
 
+enum bch_validate_flags;
 __printf(5, 6)
 int __bch2_bkey_fsck_err(struct bch_fs *,
 			 struct bkey_s_c,
-			 enum bch_fsck_flags,
+			 enum bch_validate_flags,
 			 enum bch_sb_error_id,
 			 const char *, ...);
 
@@ -180,11 +181,7 @@ int __bch2_bkey_fsck_err(struct bch_fs *,
  */
 #define bkey_fsck_err(c, _err_type, _err_msg, ...)			\
 do {									\
-	if ((flags & BCH_VALIDATE_silent)) {				\
-		ret = -BCH_ERR_fsck_delete_bkey;			\
-		goto fsck_err;						\
-	}								\
-	int _ret = __bch2_bkey_fsck_err(c, k, FSCK_CAN_FIX|FSCK_AUTOFIX,\
+	int _ret = __bch2_bkey_fsck_err(c, k, flags,			\
 				BCH_FSCK_ERR_##_err_type,		\
 				_err_msg, ##__VA_ARGS__);		\
 	if (_ret != -BCH_ERR_fsck_fix &&				\

fs/bcachefs/fs.c

@@ -174,20 +174,24 @@ static const struct rhashtable_params bch2_vfs_inodes_params = {
 	.automatic_shrinking	= true,
 };
 
-static void __wait_on_freeing_inode(struct inode *inode)
+struct bch_inode_info *__bch2_inode_hash_find(struct bch_fs *c, subvol_inum inum)
 {
-	wait_queue_head_t *wq;
-	DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
-	wq = bit_waitqueue(&inode->i_state, __I_NEW);
-	prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
-	spin_unlock(&inode->i_lock);
-	schedule();
-	finish_wait(wq, &wait.wq_entry);
+	return rhashtable_lookup_fast(&c->vfs_inodes_table, &inum, bch2_vfs_inodes_params);
 }
 
-struct bch_inode_info *__bch2_inode_hash_find(struct bch_fs *c, subvol_inum inum)
+static void __wait_on_freeing_inode(struct bch_fs *c,
+				    struct bch_inode_info *inode,
+				    subvol_inum inum)
 {
-	return rhashtable_lookup_fast(&c->vfs_inodes_table, &inum, bch2_vfs_inodes_params);
+	wait_queue_head_t *wq;
+	DEFINE_WAIT_BIT(wait, &inode->v.i_state, __I_NEW);
+	wq = inode_bit_waitqueue(&wait, &inode->v, __I_NEW);
+	prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
+	spin_unlock(&inode->v.i_lock);
+
+	if (__bch2_inode_hash_find(c, inum) == inode)
+		schedule_timeout(HZ * 10);
+	finish_wait(wq, &wait.wq_entry);
 }
 
 static struct bch_inode_info *bch2_inode_hash_find(struct bch_fs *c, struct btree_trans *trans,
@@ -204,10 +208,10 @@ static struct bch_inode_info *bch2_inode_hash_find(struct bch_fs *c, struct btre
 		}
 		if ((inode->v.i_state & (I_FREEING|I_WILL_FREE))) {
 			if (!trans) {
-				__wait_on_freeing_inode(&inode->v);
+				__wait_on_freeing_inode(c, inode, inum);
 			} else {
 				bch2_trans_unlock(trans);
-				__wait_on_freeing_inode(&inode->v);
+				__wait_on_freeing_inode(c, inode, inum);
 				int ret = bch2_trans_relock(trans);
 				if (ret)
 					return ERR_PTR(ret);
@@ -232,6 +236,11 @@ static void bch2_inode_hash_remove(struct bch_fs *c, struct bch_inode_info *inod
 					&inode->hash, bch2_vfs_inodes_params);
 		BUG_ON(ret);
 		inode->v.i_hash.pprev = NULL;
+		/*
+		 * This pairs with the bch2_inode_hash_find() ->
+		 * __wait_on_freeing_inode() path
+		 */
+		inode_wake_up_bit(&inode->v, __I_NEW);
 	}
 }
 

fs/bcachefs/inode.c

@@ -327,22 +327,20 @@ int bch2_inode_unpack(struct bkey_s_c k,
 		: bch2_inode_unpack_slowpath(k, unpacked);
 }
 
-int bch2_inode_peek_nowarn(struct btree_trans *trans,
-		    struct btree_iter *iter,
-		    struct bch_inode_unpacked *inode,
-		    subvol_inum inum, unsigned flags)
+int __bch2_inode_peek(struct btree_trans *trans,
+		      struct btree_iter *iter,
+		      struct bch_inode_unpacked *inode,
+		      subvol_inum inum, unsigned flags,
+		      bool warn)
 {
-	struct bkey_s_c k;
 	u32 snapshot;
-	int ret;
-
-	ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+	int ret = __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot, warn);
 	if (ret)
 		return ret;
 
-	k = bch2_bkey_get_iter(trans, iter, BTREE_ID_inodes,
-			       SPOS(0, inum.inum, snapshot),
-			       flags|BTREE_ITER_cached);
+	struct bkey_s_c k = bch2_bkey_get_iter(trans, iter, BTREE_ID_inodes,
+					       SPOS(0, inum.inum, snapshot),
+					       flags|BTREE_ITER_cached);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -357,20 +355,12 @@ int bch2_inode_peek_nowarn(struct btree_trans *trans,
 
 	return 0;
 err:
+	if (warn)
+		bch_err_msg(trans->c, ret, "looking up inum %llu:%llu:", inum.subvol, inum.inum);
 	bch2_trans_iter_exit(trans, iter);
 	return ret;
 }
 
-int bch2_inode_peek(struct btree_trans *trans,
-		    struct btree_iter *iter,
-		    struct bch_inode_unpacked *inode,
-		    subvol_inum inum, unsigned flags)
-{
-	int ret = bch2_inode_peek_nowarn(trans, iter, inode, inum, flags);
-	bch_err_msg(trans->c, ret, "looking up inum %llu:%llu:", inum.subvol, inum.inum);
-	return ret;
-}
-
 int bch2_inode_write_flags(struct btree_trans *trans,
 		     struct btree_iter *iter,
 		     struct bch_inode_unpacked *inode,
@@ -387,9 +377,7 @@ int bch2_inode_write_flags(struct btree_trans *trans,
 	return bch2_trans_update(trans, iter, &inode_p->inode.k_i, flags);
 }
 
-int __bch2_fsck_write_inode(struct btree_trans *trans,
-			 struct bch_inode_unpacked *inode,
-			 u32 snapshot)
+int __bch2_fsck_write_inode(struct btree_trans *trans, struct bch_inode_unpacked *inode)
 {
 	struct bkey_inode_buf *inode_p =
 		bch2_trans_kmalloc(trans, sizeof(*inode_p));
@@ -398,19 +386,17 @@ int __bch2_fsck_write_inode(struct btree_trans *trans,
 		return PTR_ERR(inode_p);
 
 	bch2_inode_pack(inode_p, inode);
-	inode_p->inode.k.p.snapshot = snapshot;
+	inode_p->inode.k.p.snapshot = inode->bi_snapshot;
 
 	return bch2_btree_insert_nonextent(trans, BTREE_ID_inodes,
 				&inode_p->inode.k_i,
 				BTREE_UPDATE_internal_snapshot_node);
 }
 
-int bch2_fsck_write_inode(struct btree_trans *trans,
-			    struct bch_inode_unpacked *inode,
-			    u32 snapshot)
+int bch2_fsck_write_inode(struct btree_trans *trans, struct bch_inode_unpacked *inode)
 {
 	int ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			    __bch2_fsck_write_inode(trans, inode, snapshot));
+			    __bch2_fsck_write_inode(trans, inode));
 	bch_err_fn(trans->c, ret);
 	return ret;
 }

fs/bcachefs/inode.h

@@ -97,10 +97,26 @@ struct bkey_i *bch2_inode_to_v3(struct btree_trans *, struct bkey_i *);
 
 void bch2_inode_unpacked_to_text(struct printbuf *, struct bch_inode_unpacked *);
 
-int bch2_inode_peek_nowarn(struct btree_trans *, struct btree_iter *,
-		    struct bch_inode_unpacked *, subvol_inum, unsigned);
-int bch2_inode_peek(struct btree_trans *, struct btree_iter *,
-		    struct bch_inode_unpacked *, subvol_inum, unsigned);
+int __bch2_inode_peek(struct btree_trans *, struct btree_iter *,
+		      struct bch_inode_unpacked *, subvol_inum, unsigned, bool);
+
+static inline int bch2_inode_peek_nowarn(struct btree_trans *trans,
+					 struct btree_iter *iter,
+					 struct bch_inode_unpacked *inode,
+					 subvol_inum inum, unsigned flags)
+{
+	return __bch2_inode_peek(trans, iter, inode, inum, flags, false);
+}
+
+static inline int bch2_inode_peek(struct btree_trans *trans,
+				  struct btree_iter *iter,
+				  struct bch_inode_unpacked *inode,
+				  subvol_inum inum, unsigned flags)
+{
+	return __bch2_inode_peek(trans, iter, inode, inum, flags, true);
+	int ret = bch2_inode_peek_nowarn(trans, iter, inode, inum, flags);
+	return ret;
+}
 
 int bch2_inode_write_flags(struct btree_trans *, struct btree_iter *,
 		     struct bch_inode_unpacked *, enum btree_iter_update_trigger_flags);
@@ -112,8 +128,8 @@ static inline int bch2_inode_write(struct btree_trans *trans,
 	return bch2_inode_write_flags(trans, iter, inode, 0);
 }
 
-int __bch2_fsck_write_inode(struct btree_trans *, struct bch_inode_unpacked *, u32);
-int bch2_fsck_write_inode(struct btree_trans *, struct bch_inode_unpacked *, u32);
+int __bch2_fsck_write_inode(struct btree_trans *, struct bch_inode_unpacked *);
+int bch2_fsck_write_inode(struct btree_trans *, struct bch_inode_unpacked *);
 
 void bch2_inode_init_early(struct bch_fs *,
 			   struct bch_inode_unpacked *);

fs/bcachefs/io_misc.c

@@ -224,13 +224,14 @@ void bch2_logged_op_truncate_to_text(struct printbuf *out, struct bch_fs *c, str
 
 static int truncate_set_isize(struct btree_trans *trans,
 			      subvol_inum inum,
-			      u64 new_i_size)
+			      u64 new_i_size,
+			      bool warn)
 {
 	struct btree_iter iter = { NULL };
 	struct bch_inode_unpacked inode_u;
 	int ret;
 
-	ret   = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent) ?:
+	ret   = __bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent, warn) ?:
 		(inode_u.bi_size = new_i_size, 0) ?:
 		bch2_inode_write(trans, &iter, &inode_u);
 
@@ -247,10 +248,11 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans,
 	struct bkey_i_logged_op_truncate *op = bkey_i_to_logged_op_truncate(op_k);
 	subvol_inum inum = { le32_to_cpu(op->v.subvol), le64_to_cpu(op->v.inum) };
 	u64 new_i_size = le64_to_cpu(op->v.new_i_size);
+	bool warn_errors = i_sectors_delta != NULL;
 	int ret;
 
 	ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			truncate_set_isize(trans, inum, new_i_size));
+			truncate_set_isize(trans, inum, new_i_size, i_sectors_delta != NULL));
 	if (ret)
 		goto err;
 
@@ -263,8 +265,8 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans,
 	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		ret = 0;
 err:
-	bch2_logged_op_finish(trans, op_k);
-	bch_err_fn(c, ret);
+	if (warn_errors)
+		bch_err_fn(c, ret);
 	return ret;
 }
 
@@ -288,9 +290,14 @@ int bch2_truncate(struct bch_fs *c, subvol_inum inum, u64 new_i_size, u64 *i_sec
 	 * resume only proceeding in one of the snapshots
 	 */
 	down_read(&c->snapshot_create_lock);
-	int ret = bch2_trans_run(c,
-		bch2_logged_op_start(trans, &op.k_i) ?:
-		__bch2_resume_logged_op_truncate(trans, &op.k_i, i_sectors_delta));
+	struct btree_trans *trans = bch2_trans_get(c);
+	int ret = bch2_logged_op_start(trans, &op.k_i);
+	if (ret)
+		goto out;
+	ret = __bch2_resume_logged_op_truncate(trans, &op.k_i, i_sectors_delta);
+	ret = bch2_logged_op_finish(trans, &op.k_i) ?: ret;
+out:
+	bch2_trans_put(trans);
 	up_read(&c->snapshot_create_lock);
 
 	return ret;
@@ -308,7 +315,8 @@ void bch2_logged_op_finsert_to_text(struct printbuf *out, struct bch_fs *c, stru
 	prt_printf(out, " src_offset=%llu",	le64_to_cpu(op.v->src_offset));
 }
 
-static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset, s64 len)
+static int adjust_i_size(struct btree_trans *trans, subvol_inum inum,
+			 u64 offset, s64 len, bool warn)
 {
 	struct btree_iter iter;
 	struct bch_inode_unpacked inode_u;
@@ -317,7 +325,7 @@ static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset
 	offset	<<= 9;
 	len	<<= 9;
 
-	ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent);
+	ret = __bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent, warn);
 	if (ret)
 		return ret;
 
@@ -357,12 +365,22 @@ static int __bch2_resume_logged_op_finsert(struct btree_trans *trans,
 	u64 len = abs(shift);
 	u64 pos = le64_to_cpu(op->v.pos);
 	bool insert = shift > 0;
+	u32 snapshot;
+	bool warn_errors = i_sectors_delta != NULL;
 	int ret = 0;
 
 	ret = bch2_inum_opts_get(trans, inum, &opts);
 	if (ret)
 		return ret;
 
+	/*
+	 * check for missing subvolume before fpunch, as in resume we don't want
+	 * it to be a fatal error
+	 */
+	ret = __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot, warn_errors);
+	if (ret)
+		return ret;
+
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(inum.inum, 0),
 			     BTREE_ITER_intent);
@@ -373,7 +391,7 @@ case LOGGED_OP_FINSERT_start:
 
 	if (insert) {
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, src_offset, len) ?:
+				adjust_i_size(trans, inum, src_offset, len, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 		if (ret)
 			goto err;
@@ -396,11 +414,11 @@ case LOGGED_OP_FINSERT_shift_extents:
 		struct bkey_i delete, *copy;
 		struct bkey_s_c k;
 		struct bpos src_pos = POS(inum.inum, src_offset);
-		u32 snapshot;
 
 		bch2_trans_begin(trans);
 
-		ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+		ret = __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot,
+						    warn_errors);
 		if (ret)
 			goto btree_err;
 
@@ -463,12 +481,12 @@ case LOGGED_OP_FINSERT_shift_extents:
 
 	if (!insert) {
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, src_offset, shift) ?:
+				adjust_i_size(trans, inum, src_offset, shift, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 	} else {
 		/* We need an inode update to update bi_journal_seq for fsync: */
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, 0, 0) ?:
+				adjust_i_size(trans, inum, 0, 0, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 	}
 
@@ -477,9 +495,9 @@ case LOGGED_OP_FINSERT_finish:
 	break;
 	}
 err:
-	bch_err_fn(c, ret);
-	bch2_logged_op_finish(trans, op_k);
 	bch2_trans_iter_exit(trans, &iter);
+	if (warn_errors)
+		bch_err_fn(c, ret);
 	return ret;
 }
 
@@ -508,9 +526,14 @@ int bch2_fcollapse_finsert(struct bch_fs *c, subvol_inum inum,
 	 * resume only proceeding in one of the snapshots
 	 */
 	down_read(&c->snapshot_create_lock);
-	int ret = bch2_trans_run(c,
-		bch2_logged_op_start(trans, &op.k_i) ?:
-		__bch2_resume_logged_op_finsert(trans, &op.k_i, i_sectors_delta));
+	struct btree_trans *trans = bch2_trans_get(c);
+	int ret = bch2_logged_op_start(trans, &op.k_i);
+	if (ret)
+		goto out;
+	ret = __bch2_resume_logged_op_finsert(trans, &op.k_i, i_sectors_delta);
+	ret = bch2_logged_op_finish(trans, &op.k_i) ?: ret;
+out:
+	bch2_trans_put(trans);
 	up_read(&c->snapshot_create_lock);
 
 	return ret;

fs/bcachefs/logged_ops.c

@@ -34,8 +34,6 @@ static int resume_logged_op(struct btree_trans *trans, struct btree_iter *iter,
 			    struct bkey_s_c k)
 {
 	struct bch_fs *c = trans->c;
-	const struct bch_logged_op_fn *fn = logged_op_fn(k.k->type);
-	struct bkey_buf sk;
 	u32 restart_count = trans->restart_count;
 	struct printbuf buf = PRINTBUF;
 	int ret = 0;
@@ -46,13 +44,15 @@ static int resume_logged_op(struct btree_trans *trans, struct btree_iter *iter,
 		    (bch2_bkey_val_to_text(&buf, c, k),
 		     buf.buf));
 
-	if (!fn)
-		return 0;
-
+	struct bkey_buf sk;
 	bch2_bkey_buf_init(&sk);
 	bch2_bkey_buf_reassemble(&sk, c, k);
 
-	fn->resume(trans, sk.k);
+	const struct bch_logged_op_fn *fn = logged_op_fn(sk.k->k.type);
+	if (fn)
+		fn->resume(trans, sk.k);
+
+	ret = bch2_logged_op_finish(trans, sk.k);
 
 	bch2_bkey_buf_exit(&sk, c);
 fsck_err:
@@ -93,7 +93,7 @@ int bch2_logged_op_start(struct btree_trans *trans, struct bkey_i *k)
 			 __bch2_logged_op_start(trans, k));
 }
 
-void bch2_logged_op_finish(struct btree_trans *trans, struct bkey_i *k)
+int bch2_logged_op_finish(struct btree_trans *trans, struct bkey_i *k)
 {
 	int ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
 			    bch2_btree_delete(trans, BTREE_ID_logged_ops, k->k.p, 0));
@@ -113,4 +113,6 @@ void bch2_logged_op_finish(struct btree_trans *trans, struct bkey_i *k)
 				    buf.buf, bch2_err_str(ret));
 		printbuf_exit(&buf);
 	}
+
+	return ret;
 }

fs/bcachefs/logged_ops.h

@@ -15,6 +15,6 @@ static inline int bch2_logged_op_update(struct btree_trans *trans, struct bkey_i
 
 int bch2_resume_logged_ops(struct bch_fs *);
 int bch2_logged_op_start(struct btree_trans *, struct bkey_i *);
-void bch2_logged_op_finish(struct btree_trans *, struct bkey_i *);
+int bch2_logged_op_finish(struct btree_trans *, struct bkey_i *);
 
 #endif /* _BCACHEFS_LOGGED_OPS_H */

fs/bcachefs/lru.c

@@ -2,6 +2,7 @@
 
 #include "bcachefs.h"
 #include "alloc_background.h"
+#include "bkey_buf.h"
 #include "btree_iter.h"
 #include "btree_update.h"
 #include "btree_write_buffer.h"
@@ -118,7 +119,7 @@ int bch2_lru_check_set(struct btree_trans *trans,
 static int bch2_check_lru_key(struct btree_trans *trans,
 			      struct btree_iter *lru_iter,
 			      struct bkey_s_c lru_k,
-			      struct bpos *last_flushed_pos)
+			      struct bkey_buf *last_flushed)
 {
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
@@ -132,11 +133,13 @@ static int bch2_check_lru_key(struct btree_trans *trans,
 	u64 idx;
 	int ret;
 
-	if (fsck_err_on(!bch2_dev_bucket_exists(c, alloc_pos),
+	struct bch_dev *ca = bch2_dev_bucket_tryget_noerror(c, alloc_pos);
+
+	if (fsck_err_on(!ca,
 			trans, lru_entry_to_invalid_bucket,
 			"lru key points to nonexistent device:bucket %llu:%llu",
 			alloc_pos.inode, alloc_pos.offset))
-		return bch2_btree_delete_at(trans, lru_iter, 0);
+		return bch2_btree_bit_mod_buffered(trans, BTREE_ID_lru, lru_iter->pos, false);
 
 	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc, alloc_pos, 0);
 	ret = bkey_err(k);
@@ -150,18 +153,15 @@ static int bch2_check_lru_key(struct btree_trans *trans,
 		idx = alloc_lru_idx_read(*a);
 		break;
 	case BCH_LRU_fragmentation:
-		idx = a->fragmentation_lru;
+		idx = alloc_lru_idx_fragmentation(*a, ca);
 		break;
 	}
 
 	if (lru_k.k->type != KEY_TYPE_set ||
 	    lru_pos_time(lru_k.k->p) != idx) {
-		if (!bpos_eq(*last_flushed_pos, lru_k.k->p)) {
-			*last_flushed_pos = lru_k.k->p;
-			ret = bch2_btree_write_buffer_flush_sync(trans) ?:
-				-BCH_ERR_transaction_restart_write_buffer_flush;
-			goto out;
-		}
+		ret = bch2_btree_write_buffer_maybe_flush(trans, lru_k, last_flushed);
+		if (ret)
+			goto err;
 
 		if (fsck_err(trans, lru_entry_bad,
 			     "incorrect lru entry: lru %s time %llu\n"
@@ -171,12 +171,12 @@ static int bch2_check_lru_key(struct btree_trans *trans,
 			     lru_pos_time(lru_k.k->p),
 			     (bch2_bkey_val_to_text(&buf1, c, lru_k), buf1.buf),
 			     (bch2_bkey_val_to_text(&buf2, c, k), buf2.buf)))
-			ret = bch2_btree_delete_at(trans, lru_iter, 0);
+			ret = bch2_btree_bit_mod_buffered(trans, BTREE_ID_lru, lru_iter->pos, false);
 	}
-out:
 err:
 fsck_err:
 	bch2_trans_iter_exit(trans, &iter);
+	bch2_dev_put(ca);
 	printbuf_exit(&buf2);
 	printbuf_exit(&buf1);
 	return ret;
@@ -184,12 +184,18 @@ static int bch2_check_lru_key(struct btree_trans *trans,
 
 int bch2_check_lrus(struct bch_fs *c)
 {
-	struct bpos last_flushed_pos = POS_MIN;
+	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_lru, POS_MIN, BTREE_ITER_prefetch, k,
 				NULL, NULL, BCH_TRANS_COMMIT_no_enospc|BCH_TRANS_COMMIT_lazy_rw,
-			bch2_check_lru_key(trans, &iter, k, &last_flushed_pos)));
+			bch2_check_lru_key(trans, &iter, k, &last_flushed)));
+
+	bch2_bkey_buf_exit(&last_flushed, c);
 	bch_err_fn(c, ret);
 	return ret;
 

fs/bcachefs/move.c

@@ -692,7 +692,7 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
 	a = bch2_alloc_to_v4(k, &a_convert);
 	dirty_sectors = bch2_bucket_sectors_dirty(*a);
 	bucket_size = ca->mi.bucket_size;
-	fragmentation = a->fragmentation_lru;
+	fragmentation = alloc_lru_idx_fragmentation(*a, ca);
 
 	ret = bch2_btree_write_buffer_tryflush(trans);
 	bch_err_msg(c, ret, "flushing btree write buffer");

fs/bcachefs/movinggc.c

@@ -73,6 +73,7 @@ move_bucket_in_flight_add(struct buckets_in_flight *list, struct move_bucket b)
 static int bch2_bucket_is_movable(struct btree_trans *trans,
 				  struct move_bucket *b, u64 time)
 {
+	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	struct bkey_s_c k;
 	struct bch_alloc_v4 _a;
@@ -90,14 +91,19 @@ static int bch2_bucket_is_movable(struct btree_trans *trans,
 	if (ret)
 		return ret;
 
+	struct bch_dev *ca = bch2_dev_tryget(c, k.k->p.inode);
+	if (!ca)
+		goto out;
+
 	a = bch2_alloc_to_v4(k, &_a);
 	b->k.gen	= a->gen;
 	b->sectors	= bch2_bucket_sectors_dirty(*a);
+	u64 lru_idx	= alloc_lru_idx_fragmentation(*a, ca);
 
-	ret = data_type_movable(a->data_type) &&
-		a->fragmentation_lru &&
-		a->fragmentation_lru <= time;
+	ret = lru_idx && lru_idx <= time;
 
+	bch2_dev_put(ca);
+out:
 	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }

fs/bcachefs/sb-errors_format.h

@@ -115,8 +115,8 @@ enum bch_fsck_flags {
 	x(alloc_key_data_type_inconsistency,			101,	0)		\
 	x(alloc_key_to_missing_dev_bucket,			102,	0)		\
 	x(alloc_key_cached_inconsistency,			103,	0)		\
-	x(alloc_key_cached_but_read_time_zero,			104,	0)		\
-	x(alloc_key_to_missing_lru_entry,			105,	0)		\
+	x(alloc_key_cached_but_read_time_zero,			104,	FSCK_AUTOFIX)	\
+	x(alloc_key_to_missing_lru_entry,			105,	FSCK_AUTOFIX)	\
 	x(alloc_key_data_type_wrong,				106,	FSCK_AUTOFIX)	\
 	x(alloc_key_gen_wrong,					107,	FSCK_AUTOFIX)	\
 	x(alloc_key_dirty_sectors_wrong,			108,	FSCK_AUTOFIX)	\
@@ -129,20 +129,20 @@ enum bch_fsck_flags {
 	x(freespace_key_wrong,					115,	0)		\
 	x(freespace_hole_missing,				116,	0)		\
 	x(bucket_gens_val_size_bad,				117,	0)		\
-	x(bucket_gens_key_wrong,				118,	0)		\
-	x(bucket_gens_hole_wrong,				119,	0)		\
-	x(bucket_gens_to_invalid_dev,				120,	0)		\
-	x(bucket_gens_to_invalid_buckets,			121,	0)		\
-	x(bucket_gens_nonzero_for_invalid_buckets,		122,	0)		\
+	x(bucket_gens_key_wrong,				118,	FSCK_AUTOFIX)	\
+	x(bucket_gens_hole_wrong,				119,	FSCK_AUTOFIX)	\
+	x(bucket_gens_to_invalid_dev,				120,	FSCK_AUTOFIX)	\
+	x(bucket_gens_to_invalid_buckets,			121,	FSCK_AUTOFIX)	\
+	x(bucket_gens_nonzero_for_invalid_buckets,		122,	FSCK_AUTOFIX)	\
 	x(need_discard_freespace_key_to_invalid_dev_bucket,	123,	0)		\
 	x(need_discard_freespace_key_bad,			124,	0)		\
 	x(backpointer_bucket_offset_wrong,			125,	0)		\
 	x(backpointer_to_missing_device,			126,	0)		\
 	x(backpointer_to_missing_alloc,				127,	0)		\
 	x(backpointer_to_missing_ptr,				128,	0)		\
-	x(lru_entry_at_time_0,					129,	0)		\
-	x(lru_entry_to_invalid_bucket,				130,	0)		\
-	x(lru_entry_bad,					131,	0)		\
+	x(lru_entry_at_time_0,					129,	FSCK_AUTOFIX)	\
+	x(lru_entry_to_invalid_bucket,				130,	FSCK_AUTOFIX)	\
+	x(lru_entry_bad,					131,	FSCK_AUTOFIX)	\
 	x(btree_ptr_val_too_big,				132,	0)		\
 	x(btree_ptr_v2_val_too_big,				133,	0)		\
 	x(btree_ptr_has_non_ptr,				134,	0)		\
@@ -158,9 +158,9 @@ enum bch_fsck_flags {
 	x(ptr_after_last_bucket,				144,	0)		\
 	x(ptr_before_first_bucket,				145,	0)		\
 	x(ptr_spans_multiple_buckets,				146,	0)		\
-	x(ptr_to_missing_backpointer,				147,	0)		\
-	x(ptr_to_missing_alloc_key,				148,	0)		\
-	x(ptr_to_missing_replicas_entry,			149,	0)		\
+	x(ptr_to_missing_backpointer,				147,	FSCK_AUTOFIX)	\
+	x(ptr_to_missing_alloc_key,				148,	FSCK_AUTOFIX)	\
+	x(ptr_to_missing_replicas_entry,			149,	FSCK_AUTOFIX)	\
 	x(ptr_to_missing_stripe,				150,	0)		\
 	x(ptr_to_incorrect_stripe,				151,	0)		\
 	x(ptr_gen_newer_than_bucket_gen,			152,	0)		\
@@ -194,7 +194,7 @@ enum bch_fsck_flags {
 	x(snapshot_skiplist_not_normalized,			180,	0)		\
 	x(snapshot_skiplist_bad,				181,	0)		\
 	x(snapshot_should_not_have_subvol,			182,	0)		\
-	x(snapshot_to_bad_snapshot_tree,			183,	0)		\
+	x(snapshot_to_bad_snapshot_tree,			183,	FSCK_AUTOFIX)	\
 	x(snapshot_bad_depth,					184,	0)		\
 	x(snapshot_bad_skiplist,				185,	0)		\
 	x(subvol_pos_bad,					186,	0)		\
@@ -211,6 +211,7 @@ enum bch_fsck_flags {
 	x(inode_unlinked_but_clean,				197,	0)		\
 	x(inode_unlinked_but_nlink_nonzero,			198,	0)		\
 	x(inode_unlinked_and_not_open,				281,	0)		\
+	x(inode_unlinked_but_has_dirent,			285,	0)		\
 	x(inode_checksum_type_invalid,				199,	0)		\
 	x(inode_compression_type_invalid,			200,	0)		\
 	x(inode_subvol_root_but_not_dir,			201,	0)		\
@@ -219,6 +220,8 @@ enum bch_fsck_flags {
 	x(inode_i_sectors_wrong,				204,	FSCK_AUTOFIX)	\
 	x(inode_dir_wrong_nlink,				205,	FSCK_AUTOFIX)	\
 	x(inode_dir_multiple_links,				206,	FSCK_AUTOFIX)	\
+	x(inode_dir_missing_backpointer,			284,	FSCK_AUTOFIX)	\
+	x(inode_dir_unlinked_but_not_empty,			286,	FSCK_AUTOFIX)	\
 	x(inode_multiple_links_but_nlink_0,			207,	FSCK_AUTOFIX)	\
 	x(inode_wrong_backpointer,				208,	FSCK_AUTOFIX)	\
 	x(inode_wrong_nlink,					209,	FSCK_AUTOFIX)	\
@@ -295,7 +298,7 @@ enum bch_fsck_flags {
 	x(accounting_key_replicas_devs_unsorted,		280,	FSCK_AUTOFIX)	\
 	x(accounting_key_version_0,				282,	FSCK_AUTOFIX)	\
 	x(logged_op_but_clean,					283,	FSCK_AUTOFIX)	\
-	x(MAX,							284,	0)
+	x(MAX,							287,	0)
 
 enum bch_sb_error_id {
 #define x(t, n, ...) BCH_FSCK_ERR_##t = n,

fs/bcachefs/subvolume.c

@@ -102,7 +102,8 @@ static int check_subvol(struct btree_trans *trans,
 				inode.bi_inum, inode.bi_snapshot,
 				inode.bi_subvol, subvol.k->p.offset)) {
 			inode.bi_subvol = subvol.k->p.offset;
-			ret = __bch2_fsck_write_inode(trans, &inode, le32_to_cpu(subvol.v->snapshot));
+			inode.bi_snapshot = le32_to_cpu(subvol.v->snapshot);
+			ret = __bch2_fsck_write_inode(trans, &inode);
 			if (ret)
 				goto err;
 		}
@@ -331,8 +332,8 @@ int bch2_snapshot_get_subvol(struct btree_trans *trans, u32 snapshot,
 		bch2_subvolume_get(trans, le32_to_cpu(snap.subvol), true, 0, subvol);
 }
 
-int bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
-				u32 *snapid)
+int __bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
+				  u32 *snapid, bool warn)
 {
 	struct btree_iter iter;
 	struct bkey_s_c_subvolume subvol;
@@ -343,7 +344,8 @@ int bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
 					  BTREE_ITER_cached|BTREE_ITER_with_updates,
 					  subvolume);
 	ret = bkey_err(subvol);
-	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT), trans->c,
+
+	bch2_fs_inconsistent_on(warn && bch2_err_matches(ret, ENOENT), trans->c,
 				"missing subvolume %u", subvolid);
 
 	if (likely(!ret))
@@ -352,6 +354,12 @@ int bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
 	return ret;
 }
 
+int bch2_subvolume_get_snapshot(struct btree_trans *trans, u32 subvolid,
+				u32 *snapid)
+{
+	return __bch2_subvolume_get_snapshot(trans, subvolid, snapid, true);
+}
+
 static int bch2_subvolume_reparent(struct btree_trans *trans,
 				   struct btree_iter *iter,
 				   struct bkey_s_c k,

fs/bcachefs/subvolume.h

@@ -26,6 +26,8 @@ int bch2_subvolume_trigger(struct btree_trans *, enum btree_id, unsigned,
 int bch2_subvol_has_children(struct btree_trans *, u32);
 int bch2_subvolume_get(struct btree_trans *, unsigned,
 		       bool, int, struct bch_subvolume *);
+int __bch2_subvolume_get_snapshot(struct btree_trans *, u32,
+				  u32 *, bool);
 int bch2_subvolume_get_snapshot(struct btree_trans *, u32, u32 *);
 
 int bch2_subvol_is_ro_trans(struct btree_trans *, u32);

fs/bcachefs/util.c

@@ -222,7 +222,7 @@ u64 bch2_read_flag_list(const char *opt, const char * const list[])
 			break;
 		}
 
-		ret |= 1 << flag;
+		ret |= BIT_ULL(flag);
 	}
 
 	kfree(d);