Merge tag 'bcachefs-2024-08-10' of git://evilpiepirate.org/bcachefs

Pull more bcachefs fixes from Kent Overstreet:
 "A couple last minute fixes for the new disk accounting

   - fix a bug that was causing ACLs to seemingly "disappear"

   - new on disk format version, bcachefs_metadata_version_disk_accounting_v3

     bcachefs_metadata_version_disk_accounting_v2 accidentally included
     padding in disk_accounting_key; fortunately, 6.11 isn't out yet so
     we can fix this with another version bump"

* tag 'bcachefs-2024-08-10' of git://evilpiepirate.org/bcachefs:
  bcachefs: bcachefs_metadata_version_disk_accounting_v3
  bcachefs: improve bch2_dev_usage_to_text()
  bcachefs: bch2_accounting_invalid()
  bcachefs: Switch to .get_inode_acl()

fs/bcachefs/acl.c

@@ -272,16 +272,19 @@ bch2_acl_to_xattr(struct btree_trans *trans,
 	return xattr;
 }
 
-struct posix_acl *bch2_get_acl(struct mnt_idmap *idmap,
-			       struct dentry *dentry, int type)
+struct posix_acl *bch2_get_acl(struct inode *vinode, int type, bool rcu)
 {
-	struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
+	struct bch_inode_info *inode = to_bch_ei(vinode);
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	struct bch_hash_info hash = bch2_hash_info_init(c, &inode->ei_inode);
 	struct xattr_search_key search = X_SEARCH(acl_to_xattr_type(type), "", 0);
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter iter = { NULL };
 	struct posix_acl *acl = NULL;
+
+	if (rcu)
+		return ERR_PTR(-ECHILD);
+
+	struct btree_trans *trans = bch2_trans_get(c);
 retry:
 	bch2_trans_begin(trans);
 

fs/bcachefs/acl.h

@@ -28,7 +28,7 @@ void bch2_acl_to_text(struct printbuf *, const void *, size_t);
 
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
 
-struct posix_acl *bch2_get_acl(struct mnt_idmap *, struct dentry *, int);
+struct posix_acl *bch2_get_acl(struct inode *, int, bool);
 
 int bch2_set_acl_trans(struct btree_trans *, subvol_inum,
 		       struct bch_inode_unpacked *,

fs/bcachefs/alloc_foreground.c

@@ -1740,7 +1740,7 @@ void bch2_dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 	printbuf_tabstop_push(out, 16);
 	printbuf_tabstop_push(out, 16);
 
-	bch2_dev_usage_to_text(out, &stats);
+	bch2_dev_usage_to_text(out, ca, &stats);
 
 	prt_newline(out);
 

fs/bcachefs/bcachefs_format.h

@@ -675,7 +675,8 @@ struct bch_sb_field_ext {
 	x(btree_subvolume_children,	BCH_VERSION(1,  6))		\
 	x(mi_btree_bitmap,		BCH_VERSION(1,  7))		\
 	x(bucket_stripe_sectors,	BCH_VERSION(1,  8))		\
-	x(disk_accounting_v2,		BCH_VERSION(1,  9))
+	x(disk_accounting_v2,		BCH_VERSION(1,  9))		\
+	x(disk_accounting_v3,		BCH_VERSION(1, 10))
 
 enum bcachefs_metadata_version {
 	bcachefs_metadata_version_min = 9,

fs/bcachefs/buckets.c

@@ -71,7 +71,9 @@ bch2_fs_usage_read_short(struct bch_fs *c)
 	return ret;
 }
 
-void bch2_dev_usage_to_text(struct printbuf *out, struct bch_dev_usage *usage)
+void bch2_dev_usage_to_text(struct printbuf *out,
+			    struct bch_dev *ca,
+			    struct bch_dev_usage *usage)
 {
 	prt_printf(out, "\tbuckets\rsectors\rfragmented\r\n");
 
@@ -82,6 +84,8 @@ void bch2_dev_usage_to_text(struct printbuf *out, struct bch_dev_usage *usage)
 			   usage->d[i].sectors,
 			   usage->d[i].fragmented);
 	}
+
+	prt_printf(out, "capacity\t%llu\r\n", ca->mi.nbuckets);
 }
 
 static int bch2_check_fix_ptr(struct btree_trans *trans,

fs/bcachefs/buckets.h

@@ -212,7 +212,7 @@ static inline struct bch_dev_usage bch2_dev_usage_read(struct bch_dev *ca)
 	return ret;
 }
 
-void bch2_dev_usage_to_text(struct printbuf *, struct bch_dev_usage *);
+void bch2_dev_usage_to_text(struct printbuf *, struct bch_dev *, struct bch_dev_usage *);
 
 static inline u64 bch2_dev_buckets_reserved(struct bch_dev *ca, enum bch_watermark watermark)
 {

fs/bcachefs/disk_accounting.c

@@ -114,11 +114,74 @@ int bch2_mod_dev_cached_sectors(struct btree_trans *trans,
 	return bch2_disk_accounting_mod(trans, &acc, &sectors, 1, gc);
 }
 
+static inline bool is_zero(char *start, char *end)
+{
+	BUG_ON(start > end);
+
+	for (; start < end; start++)
+		if (*start)
+			return false;
+	return true;
+}
+
+#define field_end(p, member)	(((void *) (&p.member)) + sizeof(p.member))
+
 int bch2_accounting_invalid(struct bch_fs *c, struct bkey_s_c k,
 			    enum bch_validate_flags flags,
 			    struct printbuf *err)
 {
-	return 0;
+	struct disk_accounting_pos acc_k;
+	bpos_to_disk_accounting_pos(&acc_k, k.k->p);
+	void *end = &acc_k + 1;
+	int ret = 0;
+
+	switch (acc_k.type) {
+	case BCH_DISK_ACCOUNTING_nr_inodes:
+		end = field_end(acc_k, nr_inodes);
+		break;
+	case BCH_DISK_ACCOUNTING_persistent_reserved:
+		end = field_end(acc_k, persistent_reserved);
+		break;
+	case BCH_DISK_ACCOUNTING_replicas:
+		bkey_fsck_err_on(!acc_k.replicas.nr_devs,
+				 c, err, accounting_key_replicas_nr_devs_0,
+				 "accounting key replicas entry with nr_devs=0");
+
+		bkey_fsck_err_on(acc_k.replicas.nr_required > acc_k.replicas.nr_devs ||
+				 (acc_k.replicas.nr_required > 1 &&
+				  acc_k.replicas.nr_required == acc_k.replicas.nr_devs),
+				 c, err, accounting_key_replicas_nr_required_bad,
+				 "accounting key replicas entry with bad nr_required");
+
+		for (unsigned i = 0; i + 1 < acc_k.replicas.nr_devs; i++)
+			bkey_fsck_err_on(acc_k.replicas.devs[i] > acc_k.replicas.devs[i + 1],
+					 c, err, accounting_key_replicas_devs_unsorted,
+					 "accounting key replicas entry with unsorted devs");
+
+		end = (void *) &acc_k.replicas + replicas_entry_bytes(&acc_k.replicas);
+		break;
+	case BCH_DISK_ACCOUNTING_dev_data_type:
+		end = field_end(acc_k, dev_data_type);
+		break;
+	case BCH_DISK_ACCOUNTING_compression:
+		end = field_end(acc_k, compression);
+		break;
+	case BCH_DISK_ACCOUNTING_snapshot:
+		end = field_end(acc_k, snapshot);
+		break;
+	case BCH_DISK_ACCOUNTING_btree:
+		end = field_end(acc_k, btree);
+		break;
+	case BCH_DISK_ACCOUNTING_rebalance_work:
+		end = field_end(acc_k, rebalance_work);
+		break;
+	}
+
+	bkey_fsck_err_on(!is_zero(end, (void *) (&acc_k + 1)),
+			 c, err, accounting_key_junk_at_end,
+			 "junk at end of accounting key");
+fsck_err:
+	return ret;
 }
 
 void bch2_accounting_key_to_text(struct printbuf *out, struct disk_accounting_pos *k)

fs/bcachefs/disk_accounting_format.h

@@ -124,20 +124,19 @@ struct bch_dev_data_type {
 	__u8			data_type;
 };
 
-struct bch_dev_stripe_buckets {
-	__u8			dev;
-};
-
 struct bch_acct_compression {
 	__u8			type;
 };
 
 struct bch_acct_snapshot {
 	__u32			id;
-};
+} __packed;
 
 struct bch_acct_btree {
 	__u32			id;
+} __packed;
+
+struct bch_acct_rebalance_work {
 };
 
 struct disk_accounting_pos {
@@ -149,12 +148,12 @@ struct disk_accounting_pos {
 		struct bch_persistent_reserved	persistent_reserved;
 		struct bch_replicas_entry_v1	replicas;
 		struct bch_dev_data_type	dev_data_type;
-		struct bch_dev_stripe_buckets	dev_stripe_buckets;
 		struct bch_acct_compression	compression;
 		struct bch_acct_snapshot	snapshot;
 		struct bch_acct_btree		btree;
-		};
-	};
+		struct bch_acct_rebalance_work	rebalance_work;
+		} __packed;
+	} __packed;
 		struct bpos			_pad;
 	};
 };

fs/bcachefs/fs.c

@@ -1199,7 +1199,7 @@ static const struct inode_operations bch_file_inode_operations = {
 	.fiemap		= bch2_fiemap,
 	.listxattr	= bch2_xattr_list,
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
-	.get_acl	= bch2_get_acl,
+	.get_inode_acl	= bch2_get_acl,
 	.set_acl	= bch2_set_acl,
 #endif
 };
@@ -1219,7 +1219,7 @@ static const struct inode_operations bch_dir_inode_operations = {
 	.tmpfile	= bch2_tmpfile,
 	.listxattr	= bch2_xattr_list,
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
-	.get_acl	= bch2_get_acl,
+	.get_inode_acl	= bch2_get_acl,
 	.set_acl	= bch2_set_acl,
 #endif
 };
@@ -1241,7 +1241,7 @@ static const struct inode_operations bch_symlink_inode_operations = {
 	.setattr	= bch2_setattr,
 	.listxattr	= bch2_xattr_list,
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
-	.get_acl	= bch2_get_acl,
+	.get_inode_acl	= bch2_get_acl,
 	.set_acl	= bch2_set_acl,
 #endif
 };
@@ -1251,7 +1251,7 @@ static const struct inode_operations bch_special_inode_operations = {
 	.setattr	= bch2_setattr,
 	.listxattr	= bch2_xattr_list,
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
-	.get_acl	= bch2_get_acl,
+	.get_inode_acl	= bch2_get_acl,
 	.set_acl	= bch2_set_acl,
 #endif
 };

fs/bcachefs/replicas.c

@@ -24,7 +24,6 @@ static int bch2_memcmp(const void *l, const void *r,  const void *priv)
 static void verify_replicas_entry(struct bch_replicas_entry_v1 *e)
 {
 #ifdef CONFIG_BCACHEFS_DEBUG
-	BUG_ON(e->data_type >= BCH_DATA_NR);
 	BUG_ON(!e->nr_devs);
 	BUG_ON(e->nr_required > 1 &&
 	       e->nr_required >= e->nr_devs);

fs/bcachefs/sb-downgrade.c

@@ -61,12 +61,37 @@
 	  BCH_FSCK_ERR_dev_usage_buckets_wrong,			\
 	  BCH_FSCK_ERR_dev_usage_sectors_wrong,			\
 	  BCH_FSCK_ERR_dev_usage_fragmented_wrong,		\
-	  BCH_FSCK_ERR_accounting_mismatch)
+	  BCH_FSCK_ERR_accounting_mismatch)			\
+	x(disk_accounting_v3,					\
+	  BIT_ULL(BCH_RECOVERY_PASS_check_allocations),		\
+	  BCH_FSCK_ERR_bkey_version_in_future,			\
+	  BCH_FSCK_ERR_dev_usage_buckets_wrong,			\
+	  BCH_FSCK_ERR_dev_usage_sectors_wrong,			\
+	  BCH_FSCK_ERR_dev_usage_fragmented_wrong,		\
+	  BCH_FSCK_ERR_accounting_mismatch,			\
+	  BCH_FSCK_ERR_accounting_key_replicas_nr_devs_0,	\
+	  BCH_FSCK_ERR_accounting_key_replicas_nr_required_bad,	\
+	  BCH_FSCK_ERR_accounting_key_replicas_devs_unsorted,	\
+	  BCH_FSCK_ERR_accounting_key_junk_at_end)
 
 #define DOWNGRADE_TABLE()					\
 	x(bucket_stripe_sectors,				\
 	  0)							\
 	x(disk_accounting_v2,					\
+	  BIT_ULL(BCH_RECOVERY_PASS_check_allocations),		\
+	  BCH_FSCK_ERR_dev_usage_buckets_wrong,			\
+	  BCH_FSCK_ERR_dev_usage_sectors_wrong,			\
+	  BCH_FSCK_ERR_dev_usage_fragmented_wrong,		\
+	  BCH_FSCK_ERR_fs_usage_hidden_wrong,			\
+	  BCH_FSCK_ERR_fs_usage_btree_wrong,			\
+	  BCH_FSCK_ERR_fs_usage_data_wrong,			\
+	  BCH_FSCK_ERR_fs_usage_cached_wrong,			\
+	  BCH_FSCK_ERR_fs_usage_reserved_wrong,			\
+	  BCH_FSCK_ERR_fs_usage_nr_inodes_wrong,		\
+	  BCH_FSCK_ERR_fs_usage_persistent_reserved_wrong,	\
+	  BCH_FSCK_ERR_fs_usage_replicas_wrong,			\
+	  BCH_FSCK_ERR_bkey_version_in_future)			\
+	x(disk_accounting_v3,					\
 	  BIT_ULL(BCH_RECOVERY_PASS_check_allocations),		\
 	  BCH_FSCK_ERR_dev_usage_buckets_wrong,			\
 	  BCH_FSCK_ERR_dev_usage_sectors_wrong,			\

fs/bcachefs/sb-errors_format.h

@@ -287,7 +287,11 @@ enum bch_fsck_flags {
 	x(accounting_replicas_not_marked,			273,	0)		\
 	x(invalid_btree_id,					274,	0)		\
 	x(alloc_key_io_time_bad,				275,	0)		\
-	x(alloc_key_fragmentation_lru_wrong,			276,	FSCK_AUTOFIX)
+	x(alloc_key_fragmentation_lru_wrong,			276,	FSCK_AUTOFIX)	\
+	x(accounting_key_junk_at_end,				277,	0)		\
+	x(accounting_key_replicas_nr_devs_0,			278,	0)		\
+	x(accounting_key_replicas_nr_required_bad,		279,	0)		\
+	x(accounting_key_replicas_devs_unsorted,		280,	0)		\
 
 enum bch_sb_error_id {
 #define x(t, n, ...) BCH_FSCK_ERR_##t = n,