bcachefs: BCH_FEATURE_no_alloc_info

If a filesystem is going to only be used read-only, and will be a
deployable image, we can strip out alloc info for a substantial
reduction in metadata size - around half, due to backpointers.

Alloc info will be regenerated on first read-write mount.

Remounting RW is disallowed for now, since we don't yet have
check_allocations running in RW mode.

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

fs/bcachefs/bcachefs_format.h

@@ -923,7 +923,8 @@ static inline void SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE(struct bch_sb *sb, __u
 	x(alloc_v2,			17)	\
 	x(extents_across_btree_nodes,	18)	\
 	x(incompat_version_field,	19)	\
-	x(casefolding,			20)
+	x(casefolding,			20)	\
+	x(no_alloc_info,		21)
 
 #define BCH_SB_FEATURES_ALWAYS				\
 	(BIT_ULL(BCH_FEATURE_new_extent_overwrite)|	\

fs/bcachefs/errcode.h

@@ -221,6 +221,7 @@
 	x(EROFS,			erofs_unfixed_errors)			\
 	x(EROFS,			erofs_norecovery)			\
 	x(EROFS,			erofs_nochanges)			\
+	x(EROFS,			erofs_no_alloc_info)			\
 	x(EROFS,			insufficient_devices)			\
 	x(0,				operation_blocked)			\
 	x(BCH_ERR_operation_blocked,	btree_cache_cannibalize_lock_blocked)	\

fs/bcachefs/recovery.c

@@ -33,7 +33,6 @@
 #include <linux/sort.h>
 #include <linux/stat.h>
 
-
 int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
 {
 	u64 b = BIT_ULL(btree);
@@ -114,11 +113,8 @@ static void kill_btree(struct bch_fs *c, enum btree_id btree)
 }
 
 /* for -o reconstruct_alloc: */
-static void bch2_reconstruct_alloc(struct bch_fs *c)
+void bch2_reconstruct_alloc(struct bch_fs *c)
 {
-	bch2_journal_log_msg(c, "dropping alloc info");
-	bch_info(c, "dropping and reconstructing all alloc info");
-
 	mutex_lock(&c->sb_lock);
 	struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 
@@ -160,6 +156,8 @@ static void bch2_reconstruct_alloc(struct bch_fs *c)
 
 	c->opts.recovery_passes |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
 
+	c->disk_sb.sb->features[0] &= ~cpu_to_le64(BIT_ULL(BCH_FEATURE_no_alloc_info));
+
 	bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
 
@@ -889,8 +887,26 @@ int bch2_fs_recovery(struct bch_fs *c)
 	if (ret)
 		goto err;
 
-	if (c->opts.reconstruct_alloc)
+	if (!c->opts.read_only &&
+	    (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info))) {
+		bch_info(c, "mounting a filesystem with no alloc info read-write; will recreate");
+
 		bch2_reconstruct_alloc(c);
+	} else if (c->opts.reconstruct_alloc) {
+		bch2_journal_log_msg(c, "dropping alloc info");
+		bch_info(c, "dropping and reconstructing all alloc info");
+
+		bch2_reconstruct_alloc(c);
+	}
+
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)) {
+		/* We can't go RW to fix errors without alloc info */
+		if (c->opts.fix_errors == FSCK_FIX_yes ||
+		    c->opts.fix_errors == FSCK_FIX_ask)
+			c->opts.fix_errors = FSCK_FIX_no;
+		if (c->opts.errors == BCH_ON_ERROR_fix_safe)
+			c->opts.errors = BCH_ON_ERROR_continue;
+	}
 
 	/*
 	 * After an unclean shutdown, skip then next few journal sequence

fs/bcachefs/recovery.h

@@ -3,6 +3,7 @@
 #define _BCACHEFS_RECOVERY_H
 
 int bch2_btree_lost_data(struct bch_fs *, enum btree_id);
+void bch2_reconstruct_alloc(struct bch_fs *);
 
 int bch2_journal_replay(struct bch_fs *);
 

fs/bcachefs/recovery_passes.c

@@ -47,8 +47,18 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
 
 	set_bit(BCH_FS_may_go_rw, &c->flags);
 
-	if (keys->nr || !c->opts.read_only || c->opts.fsck || !c->sb.clean || c->opts.recovery_passes)
+	if (keys->nr ||
+	    !c->opts.read_only ||
+	    !c->sb.clean ||
+	    c->opts.recovery_passes ||
+	    (c->opts.fsck && !(c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))) {
+		if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)) {
+			bch_info(c, "mounting a filesystem with no alloc info read-write; will recreate");
+			bch2_reconstruct_alloc(c);
+		}
+
 		return bch2_fs_read_write_early(c);
+	}
 	return 0;
 }
 
@@ -240,6 +250,8 @@ static bool should_run_recovery_pass(struct bch_fs *c, enum bch_recovery_pass pa
 {
 	struct recovery_pass_fn *p = recovery_pass_fns + pass;
 
+	if ((p->when & PASS_ALLOC) && (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))
+		return false;
 	if (c->opts.recovery_passes_exclude & BIT_ULL(pass))
 		return false;
 	if (c->opts.recovery_passes & BIT_ULL(pass))

fs/bcachefs/recovery_passes_types.h

@@ -7,6 +7,8 @@
 #define PASS_UNCLEAN		BIT(2)
 #define PASS_ALWAYS		BIT(3)
 #define PASS_ONLINE		BIT(4)
+#define PASS_ALLOC		BIT(5)
+#define PASS_FSCK_ALLOC		(PASS_FSCK|PASS_ALLOC)
 
 #ifdef CONFIG_BCACHEFS_DEBUG
 #define PASS_FSCK_DEBUG		BIT(1)
@@ -27,17 +29,17 @@
 	x(stripes_read,				 1, 0)					\
 	x(initialize_subvolumes,		 2, 0)					\
 	x(snapshots_read,			 3, PASS_ALWAYS)			\
-	x(check_allocations,			 5, PASS_FSCK)				\
-	x(trans_mark_dev_sbs,			 6, PASS_ALWAYS|PASS_SILENT)		\
-	x(fs_journal_alloc,			 7, PASS_ALWAYS|PASS_SILENT)		\
+	x(check_allocations,			 5, PASS_FSCK_ALLOC)			\
+	x(trans_mark_dev_sbs,			 6, PASS_ALWAYS|PASS_SILENT|PASS_ALLOC)	\
+	x(fs_journal_alloc,			 7, PASS_ALWAYS|PASS_SILENT|PASS_ALLOC)	\
 	x(set_may_go_rw,			 8, PASS_ALWAYS|PASS_SILENT)		\
 	x(journal_replay,			 9, PASS_ALWAYS)			\
-	x(check_alloc_info,			10, PASS_ONLINE|PASS_FSCK)		\
-	x(check_lrus,				11, PASS_ONLINE|PASS_FSCK)		\
-	x(check_btree_backpointers,		12, PASS_ONLINE|PASS_FSCK)		\
+	x(check_alloc_info,			10, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_lrus,				11, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_btree_backpointers,		12, PASS_ONLINE|PASS_FSCK_ALLOC)	\
 	x(check_backpointers_to_extents,	13, PASS_ONLINE|PASS_FSCK_DEBUG)	\
-	x(check_extents_to_backpointers,	14, PASS_ONLINE|PASS_FSCK)		\
-	x(check_alloc_to_lru_refs,		15, PASS_ONLINE|PASS_FSCK)		\
+	x(check_extents_to_backpointers,	14, PASS_ONLINE|PASS_FSCK_ALLOC)	\
+	x(check_alloc_to_lru_refs,		15, PASS_ONLINE|PASS_FSCK_ALLOC)	\
 	x(fs_freespace_init,			16, PASS_ALWAYS|PASS_SILENT)		\
 	x(bucket_gens_init,			17, 0)					\
 	x(reconstruct_snapshots,		38, 0)					\

fs/bcachefs/sb-members.c

@@ -190,6 +190,12 @@ static int validate_member(struct printbuf *err,
 		return -BCH_ERR_invalid_sb_members;
 	}
 
+	if (BCH_MEMBER_FREESPACE_INITIALIZED(&m) &&
+	    sb->features[0] & cpu_to_le64(BIT_ULL(BCH_FEATURE_no_alloc_info))) {
+		prt_printf(err, "device %u: freespace initialized but fs has no alloc info", i);
+		return -BCH_ERR_invalid_sb_members;
+	}
+
 	return 0;
 }
 

fs/bcachefs/super.c

@@ -443,6 +443,9 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early)
 
 	BUG_ON(!test_bit(BCH_FS_may_go_rw, &c->flags));
 
+	if (WARN_ON(c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info)))
+		return -BCH_ERR_erofs_no_alloc_info;
+
 	if (test_bit(BCH_FS_initial_gc_unfixed, &c->flags)) {
 		bch_err(c, "cannot go rw, unfixed btree errors");
 		return -BCH_ERR_erofs_unfixed_errors;
@@ -535,6 +538,9 @@ int bch2_fs_read_write(struct bch_fs *c)
 	if (c->opts.nochanges)
 		return -BCH_ERR_erofs_nochanges;
 
+	if (c->sb.features & BIT_ULL(BCH_FEATURE_no_alloc_info))
+		return -BCH_ERR_erofs_no_alloc_info;
+
 	return __bch2_fs_read_write(c, false);
 }