Merge tag 'bcachefs-2023-12-27' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:
 "Just a few fixes: besides a few one liners, we have a fix for
  snapshots + compression where the extent update path didn't account
  for the fact that with snapshots, we might split an existing extent
  into three, not just two; and a small fixup for promotes which were
  broken by the recent changes in the data update path to correctly take
  into account device durability"

* tag 'bcachefs-2023-12-27' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: Fix promotes
  bcachefs: Fix leakage of internal error code
  bcachefs: Fix insufficient disk reservation with compression + snapshots
  bcachefs: fix BCH_FSCK_ERR enum

Author: torvalds

fs/bcachefs/btree_update.c

@@ -186,19 +186,21 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 	enum btree_id btree_id = iter->btree_id;
 	struct bkey_i *update;
 	struct bpos new_start = bkey_start_pos(new.k);
-	bool front_split = bkey_lt(bkey_start_pos(old.k), new_start);
-	bool back_split  = bkey_gt(old.k->p, new.k->p);
+	unsigned front_split = bkey_lt(bkey_start_pos(old.k), new_start);
+	unsigned back_split  = bkey_gt(old.k->p, new.k->p);
+	unsigned middle_split = (front_split || back_split) &&
+		old.k->p.snapshot != new.k->p.snapshot;
+	unsigned nr_splits = front_split + back_split + middle_split;
 	int ret = 0, compressed_sectors;
 
 	/*
 	 * If we're going to be splitting a compressed extent, note it
 	 * so that __bch2_trans_commit() can increase our disk
 	 * reservation:
 	 */
-	if (((front_split && back_split) ||
-	     ((front_split || back_split) && old.k->p.snapshot != new.k->p.snapshot)) &&
+	if (nr_splits > 1 &&
 	    (compressed_sectors = bch2_bkey_sectors_compressed(old)))
-		trans->extra_journal_res += compressed_sectors;
+		trans->extra_journal_res += compressed_sectors * (nr_splits - 1);
 
 	if (front_split) {
 		update = bch2_bkey_make_mut_noupdate(trans, old);
@@ -216,8 +218,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
 	}
 
 	/* If we're overwriting in a different snapshot - middle split: */
-	if (old.k->p.snapshot != new.k->p.snapshot &&
-	    (front_split || back_split)) {
+	if (middle_split) {
 		update = bch2_bkey_make_mut_noupdate(trans, old);
 		if ((ret = PTR_ERR_OR_ZERO(update)))
 			return ret;

fs/bcachefs/data_update.c

@@ -587,7 +587,8 @@ int bch2_data_update_init(struct btree_trans *trans,
 	 * Increasing replication is an explicit operation triggered by
 	 * rereplicate, currently, so that users don't get an unexpected -ENOSPC
 	 */
-	if (durability_have >= io_opts.data_replicas) {
+	if (!(m->data_opts.write_flags & BCH_WRITE_CACHED) &&
+	    durability_have >= io_opts.data_replicas) {
 		m->data_opts.kill_ptrs |= m->data_opts.rewrite_ptrs;
 		m->data_opts.rewrite_ptrs = 0;
 		/* if iter == NULL, it's just a promote */

fs/bcachefs/error.h

@@ -157,6 +157,7 @@ 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__)
 
+__printf(4, 0)
 static inline void bch2_bkey_fsck_err(struct bch_fs *c,
 				     struct printbuf *err_msg,
 				     enum bch_sb_error_id err_type,
@@ -167,7 +168,6 @@ static inline void bch2_bkey_fsck_err(struct bch_fs *c,
 	va_start(args, fmt);
 	prt_vprintf(err_msg, fmt, args);
 	va_end(args);
-
 }
 
 #define bkey_fsck_err(c, _err_msg, _err_type, ...)			\

fs/bcachefs/journal_io.c

@@ -408,8 +408,10 @@ static int journal_entry_btree_root_validate(struct bch_fs *c,
 		return 0;
 	}
 
-	return journal_validate_key(c, jset, entry, 1, entry->btree_id, k,
-				    version, big_endian, flags);
+	ret = journal_validate_key(c, jset, entry, 1, entry->btree_id, k,
+				   version, big_endian, flags);
+	if (ret == FSCK_DELETED_KEY)
+		ret = 0;
 fsck_err:
 	return ret;
 }

fs/bcachefs/sb-errors.h

@@ -65,7 +65,7 @@
 	x(btree_node_bkey_out_of_order,				57)	\
 	x(btree_root_bkey_invalid,				58)	\
 	x(btree_root_read_error,				59)	\
-	x(btree_root_bad_min_key,				50)	\
+	x(btree_root_bad_min_key,				60)	\
 	x(btree_root_bad_max_key,				61)	\
 	x(btree_node_read_error,				62)	\
 	x(btree_node_topology_bad_min_key,			63)	\