Update bcachefs sources to e7015d0df31c bcachefs: Simplify bch2_xattr_emit() implementation

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

.bcachefs_revision

@@ -1 +1 @@
-070f7d6a382a0d67628756727751c24f70eb48ea
+e7015d0df31c410789666b4c8ad0683aae87b4da

include/linux/xattr.h

@@ -47,6 +47,12 @@ struct xattr_handler {
 		   size_t size, int flags);
 };
 
+static inline bool xattr_handler_can_list(const struct xattr_handler *handler,
+					  struct dentry *dentry)
+{
+	return handler && (!handler->list || handler->list(dentry));
+}
+
 const char *xattr_full_name(const struct xattr_handler *, const char *);
 
 struct xattr {

libbcachefs/alloc_background.c

@@ -196,75 +196,71 @@ static unsigned bch_alloc_v1_val_u64s(const struct bch_alloc *a)
 	return DIV_ROUND_UP(bytes, sizeof(u64));
 }
 
-int bch2_alloc_v1_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_alloc_v1_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_s_c_alloc a = bkey_s_c_to_alloc(k);
 	int ret = 0;
 
 	/* allow for unknown fields */
-	bkey_fsck_err_on(bkey_val_u64s(a.k) < bch_alloc_v1_val_u64s(a.v), c, err,
+	bkey_fsck_err_on(bkey_val_u64s(a.k) < bch_alloc_v1_val_u64s(a.v),
-			 alloc_v1_val_size_bad,
+			 c, alloc_v1_val_size_bad,
 			 "incorrect value size (%zu < %u)",
 			 bkey_val_u64s(a.k), bch_alloc_v1_val_u64s(a.v));
 fsck_err:
 	return ret;
 }
 
-int bch2_alloc_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_alloc_v2_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_alloc_unpacked u;
 	int ret = 0;
 
-	bkey_fsck_err_on(bch2_alloc_unpack_v2(&u, k), c, err,
+	bkey_fsck_err_on(bch2_alloc_unpack_v2(&u, k),
-			 alloc_v2_unpack_error,
+			 c, alloc_v2_unpack_error,
 			 "unpack error");
 fsck_err:
 	return ret;
 }
 
-int bch2_alloc_v3_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_alloc_v3_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_alloc_unpacked u;
 	int ret = 0;
 
-	bkey_fsck_err_on(bch2_alloc_unpack_v3(&u, k), c, err,
+	bkey_fsck_err_on(bch2_alloc_unpack_v3(&u, k),
-			 alloc_v2_unpack_error,
+			 c, alloc_v2_unpack_error,
 			 "unpack error");
 fsck_err:
 	return ret;
 }
 
-int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_alloc_v4_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags, struct printbuf *err)
+			   enum bch_validate_flags flags)
 {
 	struct bkey_s_c_alloc_v4 a = bkey_s_c_to_alloc_v4(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(alloc_v4_u64s_noerror(a.v) > bkey_val_u64s(k.k), c, err,
+	bkey_fsck_err_on(alloc_v4_u64s_noerror(a.v) > bkey_val_u64s(k.k),
-			 alloc_v4_val_size_bad,
+			 c, alloc_v4_val_size_bad,
 			 "bad val size (%u > %zu)",
 			 alloc_v4_u64s_noerror(a.v), bkey_val_u64s(k.k));
 
 	bkey_fsck_err_on(!BCH_ALLOC_V4_BACKPOINTERS_START(a.v) &&
-			 BCH_ALLOC_V4_NR_BACKPOINTERS(a.v), c, err,
+			 BCH_ALLOC_V4_NR_BACKPOINTERS(a.v),
-			 alloc_v4_backpointers_start_bad,
+			 c, alloc_v4_backpointers_start_bad,
 			 "invalid backpointers_start");
 
-	bkey_fsck_err_on(alloc_data_type(*a.v, a.v->data_type) != a.v->data_type, c, err,
+	bkey_fsck_err_on(alloc_data_type(*a.v, a.v->data_type) != a.v->data_type,
-			 alloc_key_data_type_bad,
+			 c, alloc_key_data_type_bad,
 			 "invalid data type (got %u should be %u)",
 			 a.v->data_type, alloc_data_type(*a.v, a.v->data_type));
 
 	for (unsigned i = 0; i < 2; i++)
 		bkey_fsck_err_on(a.v->io_time[i] > LRU_TIME_MAX,
-				 c, err,
+				 c, alloc_key_io_time_bad,
-				 alloc_key_io_time_bad,
 				 "invalid io_time[%s]: %llu, max %llu",
 				 i == READ ? "read" : "write",
 				 a.v->io_time[i], LRU_TIME_MAX);
@@ -282,7 +278,7 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 				 a.v->dirty_sectors ||
 				 a.v->cached_sectors ||
 				 a.v->stripe,
-				 c, err, alloc_key_empty_but_have_data,
+				 c, alloc_key_empty_but_have_data,
 				 "empty data type free but have data %u.%u.%u %u",
 				 stripe_sectors,
 				 a.v->dirty_sectors,
@@ -296,7 +292,7 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 	case BCH_DATA_parity:
 		bkey_fsck_err_on(!a.v->dirty_sectors &&
 				 !stripe_sectors,
-				 c, err, alloc_key_dirty_sectors_0,
+				 c, alloc_key_dirty_sectors_0,
 				 "data_type %s but dirty_sectors==0",
 				 bch2_data_type_str(a.v->data_type));
 		break;
@@ -305,12 +301,12 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 				 a.v->dirty_sectors ||
 				 stripe_sectors ||
 				 a.v->stripe,
-				 c, err, alloc_key_cached_inconsistency,
+				 c, alloc_key_cached_inconsistency,
 				 "data type inconsistency");
 
 		bkey_fsck_err_on(!a.v->io_time[READ] &&
 				 c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_to_lru_refs,
-				 c, err, alloc_key_cached_but_read_time_zero,
+				 c, alloc_key_cached_but_read_time_zero,
 				 "cached bucket with read_time == 0");
 		break;
 	case BCH_DATA_stripe:
@@ -513,14 +509,13 @@ static unsigned alloc_gen(struct bkey_s_c k, unsigned offset)
 		: 0;
 }
 
-int bch2_bucket_gens_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_bucket_gens_validate(struct bch_fs *c, struct bkey_s_c k,
-			     enum bch_validate_flags flags,
+			     enum bch_validate_flags flags)
-			     struct printbuf *err)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(bkey_val_bytes(k.k) != sizeof(struct bch_bucket_gens), c, err,
+	bkey_fsck_err_on(bkey_val_bytes(k.k) != sizeof(struct bch_bucket_gens),
-			 bucket_gens_val_size_bad,
+			 c, bucket_gens_val_size_bad,
 			 "bad val size (%zu != %zu)",
 			 bkey_val_bytes(k.k), sizeof(struct bch_bucket_gens));
 fsck_err:
@@ -829,7 +824,19 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 
 	struct bch_alloc_v4 old_a_convert;
 	const struct bch_alloc_v4 *old_a = bch2_alloc_to_v4(old, &old_a_convert);
-	struct bch_alloc_v4 *new_a = bkey_s_to_alloc_v4(new).v;
+
+	struct bch_alloc_v4 *new_a;
+	if (likely(new.k->type == KEY_TYPE_alloc_v4)) {
+		new_a = bkey_s_to_alloc_v4(new).v;
+	} else {
+		BUG_ON(!(flags & BTREE_TRIGGER_gc));
+
+		struct bkey_i_alloc_v4 *new_ka = bch2_alloc_to_v4_mut_inlined(trans, new.s_c);
+		ret = PTR_ERR_OR_ZERO(new_ka);
+		if (unlikely(ret))
+			goto err;
+		new_a = &new_ka->v;
+	}
 
 	if (flags & BTREE_TRIGGER_transactional) {
 		alloc_data_type_set(new_a, new_a->data_type);

libbcachefs/alloc_background.h

@@ -150,7 +150,9 @@ static inline void alloc_data_type_set(struct bch_alloc_v4 *a, enum bch_data_typ
 
 static inline u64 alloc_lru_idx_read(struct bch_alloc_v4 a)
 {
-	return a.data_type == BCH_DATA_cached ? a.io_time[READ] : 0;
+	return a.data_type == BCH_DATA_cached
+		? a.io_time[READ] & LRU_TIME_MAX
+		: 0;
 }
 
 #define DATA_TYPES_MOVABLE		\
@@ -240,52 +242,48 @@ struct bkey_i_alloc_v4 *bch2_alloc_to_v4_mut(struct btree_trans *, struct bkey_s
 
 int bch2_bucket_io_time_reset(struct btree_trans *, unsigned, size_t, int);
 
-int bch2_alloc_v1_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_alloc_v1_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			  enum bch_validate_flags, struct printbuf *);
+int bch2_alloc_v2_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-int bch2_alloc_v2_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_alloc_v3_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			  enum bch_validate_flags, struct printbuf *);
+int bch2_alloc_v4_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-int bch2_alloc_v3_invalid(struct bch_fs *, struct bkey_s_c,
-			  enum bch_validate_flags, struct printbuf *);
-int bch2_alloc_v4_invalid(struct bch_fs *, struct bkey_s_c,
-			  enum bch_validate_flags, struct printbuf *);
 void bch2_alloc_v4_swab(struct bkey_s);
 void bch2_alloc_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_alloc ((struct bkey_ops) {	\
-	.key_invalid	= bch2_alloc_v1_invalid,	\
+	.key_validate	= bch2_alloc_v1_validate,	\
 	.val_to_text	= bch2_alloc_to_text,		\
 	.trigger	= bch2_trigger_alloc,		\
 	.min_val_size	= 8,				\
 })
 
 #define bch2_bkey_ops_alloc_v2 ((struct bkey_ops) {	\
-	.key_invalid	= bch2_alloc_v2_invalid,	\
+	.key_validate	= bch2_alloc_v2_validate,	\
 	.val_to_text	= bch2_alloc_to_text,		\
 	.trigger	= bch2_trigger_alloc,		\
 	.min_val_size	= 8,				\
 })
 
 #define bch2_bkey_ops_alloc_v3 ((struct bkey_ops) {	\
-	.key_invalid	= bch2_alloc_v3_invalid,	\
+	.key_validate	= bch2_alloc_v3_validate,	\
 	.val_to_text	= bch2_alloc_to_text,		\
 	.trigger	= bch2_trigger_alloc,		\
 	.min_val_size	= 16,				\
 })
 
 #define bch2_bkey_ops_alloc_v4 ((struct bkey_ops) {	\
-	.key_invalid	= bch2_alloc_v4_invalid,	\
+	.key_validate	= bch2_alloc_v4_validate,	\
 	.val_to_text	= bch2_alloc_to_text,		\
 	.swab		= bch2_alloc_v4_swab,		\
 	.trigger	= bch2_trigger_alloc,		\
 	.min_val_size	= 48,				\
 })
 
-int bch2_bucket_gens_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_bucket_gens_validate(struct bch_fs *, struct bkey_s_c,
-			     enum bch_validate_flags, struct printbuf *);
+			     enum bch_validate_flags);
 void bch2_bucket_gens_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_bucket_gens ((struct bkey_ops) {	\
-	.key_invalid	= bch2_bucket_gens_invalid,	\
+	.key_validate	= bch2_bucket_gens_validate,	\
 	.val_to_text	= bch2_bucket_gens_to_text,	\
 })
 

libbcachefs/backpointers.c

@@ -47,9 +47,8 @@ static bool extent_matches_bp(struct bch_fs *c,
 	return false;
 }
 
-int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_backpointer_validate(struct bch_fs *c, struct bkey_s_c k,
-			     enum bch_validate_flags flags,
+			      enum bch_validate_flags flags)
-			     struct printbuf *err)
 {
 	struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(k);
 
@@ -68,8 +67,7 @@ int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
 
 	bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||
 			 !bpos_eq(bp.k->p, bp_pos),
-			 c, err,
+			 c, backpointer_bucket_offset_wrong,
-			 backpointer_bucket_offset_wrong,
 			 "backpointer bucket_offset wrong");
 fsck_err:
 	return ret;

libbcachefs/backpointers.h

@@ -18,14 +18,13 @@ static inline u64 swab40(u64 x)
 		((x & 0xff00000000ULL) >> 32));
 }
 
-int bch2_backpointer_invalid(struct bch_fs *, struct bkey_s_c k,
+int bch2_backpointer_validate(struct bch_fs *, struct bkey_s_c k, enum bch_validate_flags);
-			     enum bch_validate_flags, struct printbuf *);
 void bch2_backpointer_to_text(struct printbuf *, const struct bch_backpointer *);
 void bch2_backpointer_k_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 void bch2_backpointer_swab(struct bkey_s);
 
 #define bch2_bkey_ops_backpointer ((struct bkey_ops) {	\
-	.key_invalid	= bch2_backpointer_invalid,	\
+	.key_validate	= bch2_backpointer_validate,	\
 	.val_to_text	= bch2_backpointer_k_to_text,	\
 	.swab		= bch2_backpointer_swab,	\
 	.min_val_size	= 32,				\

libbcachefs/bcachefs.h

@@ -447,6 +447,7 @@ BCH_DEBUG_PARAMS_DEBUG()
 	x(blocked_journal_low_on_space)		\
 	x(blocked_journal_low_on_pin)		\
 	x(blocked_journal_max_in_flight)	\
+	x(blocked_key_cache_flush)		\
 	x(blocked_allocate)			\
 	x(blocked_allocate_open_bucket)		\
 	x(blocked_write_buffer_full)		\

libbcachefs/bcachefs_format.h

@@ -676,7 +676,8 @@ struct bch_sb_field_ext {
 	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_v3,		BCH_VERSION(1, 10))
+	x(disk_accounting_v3,		BCH_VERSION(1, 10))		\
+	x(disk_accounting_inum,		BCH_VERSION(1, 11))
 
 enum bcachefs_metadata_version {
 	bcachefs_metadata_version_min = 9,

libbcachefs/bkey.h

@@ -10,9 +10,10 @@
 #include "vstructs.h"
 
 enum bch_validate_flags {
-	BCH_VALIDATE_write		= (1U << 0),
+	BCH_VALIDATE_write		= BIT(0),
-	BCH_VALIDATE_commit		= (1U << 1),
+	BCH_VALIDATE_commit		= BIT(1),
-	BCH_VALIDATE_journal		= (1U << 2),
+	BCH_VALIDATE_journal		= BIT(2),
+	BCH_VALIDATE_silent		= BIT(3),
 };
 
 #if 0

libbcachefs/bkey_methods.c

@@ -27,27 +27,27 @@ const char * const bch2_bkey_types[] = {
 	NULL
 };
 
-static int deleted_key_invalid(struct bch_fs *c, struct bkey_s_c k,
+static int deleted_key_validate(struct bch_fs *c, struct bkey_s_c k,
-			       enum bch_validate_flags flags, struct printbuf *err)
+				enum bch_validate_flags flags)
 {
 	return 0;
 }
 
 #define bch2_bkey_ops_deleted ((struct bkey_ops) {	\
-	.key_invalid = deleted_key_invalid,		\
+	.key_validate	= deleted_key_validate,		\
 })
 
 #define bch2_bkey_ops_whiteout ((struct bkey_ops) {	\
-	.key_invalid = deleted_key_invalid,		\
+	.key_validate	= deleted_key_validate,		\
 })
 
-static int empty_val_key_invalid(struct bch_fs *c, struct bkey_s_c k,
+static int empty_val_key_validate(struct bch_fs *c, struct bkey_s_c k,
-				 enum bch_validate_flags flags, struct printbuf *err)
+				 enum bch_validate_flags flags)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(bkey_val_bytes(k.k), c, err,
+	bkey_fsck_err_on(bkey_val_bytes(k.k),
-			 bkey_val_size_nonzero,
+			 c, bkey_val_size_nonzero,
 			 "incorrect value size (%zu != 0)",
 			 bkey_val_bytes(k.k));
 fsck_err:
@@ -55,11 +55,11 @@ fsck_err:
 }
 
 #define bch2_bkey_ops_error ((struct bkey_ops) {	\
-	.key_invalid = empty_val_key_invalid,		\
+	.key_validate = empty_val_key_validate,		\
 })
 
-static int key_type_cookie_invalid(struct bch_fs *c, struct bkey_s_c k,
+static int key_type_cookie_validate(struct bch_fs *c, struct bkey_s_c k,
-				   enum bch_validate_flags flags, struct printbuf *err)
+				    enum bch_validate_flags flags)
 {
 	return 0;
 }
@@ -73,17 +73,17 @@ static void key_type_cookie_to_text(struct printbuf *out, struct bch_fs *c,
 }
 
 #define bch2_bkey_ops_cookie ((struct bkey_ops) {	\
-	.key_invalid	= key_type_cookie_invalid,	\
+	.key_validate	= key_type_cookie_validate,	\
 	.val_to_text	= key_type_cookie_to_text,	\
 	.min_val_size	= 8,				\
 })
 
 #define bch2_bkey_ops_hash_whiteout ((struct bkey_ops) {\
-	.key_invalid = empty_val_key_invalid,		\
+	.key_validate	= empty_val_key_validate,	\
 })
 
-static int key_type_inline_data_invalid(struct bch_fs *c, struct bkey_s_c k,
+static int key_type_inline_data_validate(struct bch_fs *c, struct bkey_s_c k,
-					enum bch_validate_flags flags, struct printbuf *err)
+					 enum bch_validate_flags flags)
 {
 	return 0;
 }
@@ -98,9 +98,9 @@ static void key_type_inline_data_to_text(struct printbuf *out, struct bch_fs *c,
 	       datalen, min(datalen, 32U), d.v->data);
 }
 
-#define bch2_bkey_ops_inline_data ((struct bkey_ops) {	\
+#define bch2_bkey_ops_inline_data ((struct bkey_ops) {		\
-	.key_invalid	= key_type_inline_data_invalid,	\
+	.key_validate	= key_type_inline_data_validate,	\
-	.val_to_text	= key_type_inline_data_to_text,	\
+	.val_to_text	= key_type_inline_data_to_text,		\
 })
 
 static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
@@ -110,7 +110,7 @@ static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_
 }
 
 #define bch2_bkey_ops_set ((struct bkey_ops) {		\
-	.key_invalid	= empty_val_key_invalid,	\
+	.key_validate	= empty_val_key_validate,	\
 	.key_merge	= key_type_set_merge,		\
 })
 
@@ -123,9 +123,8 @@ const struct bkey_ops bch2_bkey_ops[] = {
 const struct bkey_ops bch2_bkey_null_ops = {
 };
 
-int bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_bkey_val_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	if (test_bit(BCH_FS_no_invalid_checks, &c->flags))
 		return 0;
@@ -133,15 +132,15 @@ int bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k,
 	const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
 	int ret = 0;
 
-	bkey_fsck_err_on(bkey_val_bytes(k.k) < ops->min_val_size, c, err,
+	bkey_fsck_err_on(bkey_val_bytes(k.k) < ops->min_val_size,
-			 bkey_val_size_too_small,
+			 c, bkey_val_size_too_small,
 			 "bad val size (%zu < %u)",
 			 bkey_val_bytes(k.k), ops->min_val_size);
 
-	if (!ops->key_invalid)
+	if (!ops->key_validate)
 		return 0;
 
-	ret = ops->key_invalid(c, k, flags, err);
+	ret = ops->key_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
@@ -161,18 +160,17 @@ const char *bch2_btree_node_type_str(enum btree_node_type type)
 	return type == BKEY_TYPE_btree ? "internal btree node" : bch2_btree_id_str(type - 1);
 }
 
-int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
+int __bch2_bkey_validate(struct bch_fs *c, struct bkey_s_c k,
-			enum btree_node_type type,
+			 enum btree_node_type type,
-			enum bch_validate_flags flags,
+			 enum bch_validate_flags flags)
-			struct printbuf *err)
 {
 	if (test_bit(BCH_FS_no_invalid_checks, &c->flags))
 		return 0;
 
 	int ret = 0;
 
-	bkey_fsck_err_on(k.k->u64s < BKEY_U64s, c, err,
+	bkey_fsck_err_on(k.k->u64s < BKEY_U64s,
-			 bkey_u64s_too_small,
+			 c, bkey_u64s_too_small,
 			 "u64s too small (%u < %zu)", k.k->u64s, BKEY_U64s);
 
 	if (type >= BKEY_TYPE_NR)
@@ -180,8 +178,8 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 
 	bkey_fsck_err_on(k.k->type < KEY_TYPE_MAX &&
 			 (type == BKEY_TYPE_btree || (flags & BCH_VALIDATE_commit)) &&
-			 !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, err,
+			 !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)),
-			 bkey_invalid_type_for_btree,
+			 c, bkey_invalid_type_for_btree,
 			 "invalid key type for btree %s (%s)",
 			 bch2_btree_node_type_str(type),
 			 k.k->type < KEY_TYPE_MAX
@@ -189,17 +187,17 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 			 : "(unknown)");
 
 	if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) {
-		bkey_fsck_err_on(k.k->size == 0, c, err,
+		bkey_fsck_err_on(k.k->size == 0,
-				 bkey_extent_size_zero,
+				 c, bkey_extent_size_zero,
 				 "size == 0");
 
-		bkey_fsck_err_on(k.k->size > k.k->p.offset, c, err,
+		bkey_fsck_err_on(k.k->size > k.k->p.offset,
-				 bkey_extent_size_greater_than_offset,
+				 c, bkey_extent_size_greater_than_offset,
 				 "size greater than offset (%u > %llu)",
 				 k.k->size, k.k->p.offset);
 	} else {
-		bkey_fsck_err_on(k.k->size, c, err,
+		bkey_fsck_err_on(k.k->size,
-				 bkey_size_nonzero,
+				 c, bkey_size_nonzero,
 				 "size != 0");
 	}
 
@@ -207,12 +205,12 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 		enum btree_id btree = type - 1;
 
 		if (btree_type_has_snapshots(btree)) {
-			bkey_fsck_err_on(!k.k->p.snapshot, c, err,
+			bkey_fsck_err_on(!k.k->p.snapshot,
-					 bkey_snapshot_zero,
+					 c, bkey_snapshot_zero,
 					 "snapshot == 0");
 		} else if (!btree_type_has_snapshot_field(btree)) {
-			bkey_fsck_err_on(k.k->p.snapshot, c, err,
+			bkey_fsck_err_on(k.k->p.snapshot,
-					 bkey_snapshot_nonzero,
+					 c, bkey_snapshot_nonzero,
 					 "nonzero snapshot");
 		} else {
 			/*
@@ -221,34 +219,33 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
 			 */
 		}
 
-		bkey_fsck_err_on(bkey_eq(k.k->p, POS_MAX), c, err,
+		bkey_fsck_err_on(bkey_eq(k.k->p, POS_MAX),
-				 bkey_at_pos_max,
+				 c, bkey_at_pos_max,
 				 "key at POS_MAX");
 	}
 fsck_err:
 	return ret;
 }
 
-int bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_bkey_validate(struct bch_fs *c, struct bkey_s_c k,
 		      enum btree_node_type type,
-		      enum bch_validate_flags flags,
+		      enum bch_validate_flags flags)
-		      struct printbuf *err)
 {
-	return __bch2_bkey_invalid(c, k, type, flags, err) ?:
+	return __bch2_bkey_validate(c, k, type, flags) ?:
-		bch2_bkey_val_invalid(c, k, flags, err);
+		bch2_bkey_val_validate(c, k, flags);
 }
 
 int bch2_bkey_in_btree_node(struct bch_fs *c, struct btree *b,
-			    struct bkey_s_c k, struct printbuf *err)
+			    struct bkey_s_c k, enum bch_validate_flags flags)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(bpos_lt(k.k->p, b->data->min_key), c, err,
+	bkey_fsck_err_on(bpos_lt(k.k->p, b->data->min_key),
-			 bkey_before_start_of_btree_node,
+			 c, bkey_before_start_of_btree_node,
 			 "key before start of btree node");
 
-	bkey_fsck_err_on(bpos_gt(k.k->p, b->data->max_key), c, err,
+	bkey_fsck_err_on(bpos_gt(k.k->p, b->data->max_key),
-			 bkey_after_end_of_btree_node,
+			 c, bkey_after_end_of_btree_node,
 			 "key past end of btree node");
 fsck_err:
 	return ret;

libbcachefs/bkey_methods.h

@@ -14,15 +14,15 @@ extern const char * const bch2_bkey_types[];
 extern const struct bkey_ops bch2_bkey_null_ops;
 
 /*
- * key_invalid: checks validity of @k, returns 0 if good or -EINVAL if bad. If
+ * key_validate: checks validity of @k, returns 0 if good or -EINVAL if bad. If
  * invalid, entire key will be deleted.
  *
  * When invalid, error string is returned via @err. @rw indicates whether key is
  * being read or written; more aggressive checks can be enabled when rw == WRITE.
  */
 struct bkey_ops {
-	int		(*key_invalid)(struct bch_fs *c, struct bkey_s_c k,
+	int		(*key_validate)(struct bch_fs *c, struct bkey_s_c k,
-				       enum bch_validate_flags flags, struct printbuf *err);
+					enum bch_validate_flags flags);
 	void		(*val_to_text)(struct printbuf *, struct bch_fs *,
 				       struct bkey_s_c);
 	void		(*swab)(struct bkey_s);
@@ -48,14 +48,13 @@ static inline const struct bkey_ops *bch2_bkey_type_ops(enum bch_bkey_type type)
 		: &bch2_bkey_null_ops;
 }
 
-int bch2_bkey_val_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_bkey_val_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			  enum bch_validate_flags, struct printbuf *);
+int __bch2_bkey_validate(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
-int __bch2_bkey_invalid(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
+			 enum bch_validate_flags);
-			enum bch_validate_flags, struct printbuf *);
+int bch2_bkey_validate(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
-int bch2_bkey_invalid(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
+		       enum bch_validate_flags);
-		      enum bch_validate_flags, struct printbuf *);
+int bch2_bkey_in_btree_node(struct bch_fs *, struct btree *, struct bkey_s_c,
-int bch2_bkey_in_btree_node(struct bch_fs *, struct btree *,
+			    enum bch_validate_flags);
-			    struct bkey_s_c, struct printbuf *);
 
 void bch2_bpos_to_text(struct printbuf *, struct bpos);
 void bch2_bkey_to_text(struct printbuf *, const struct bkey *);

libbcachefs/bset.c

@@ -304,11 +304,6 @@ struct bkey_float {
 };
 #define BKEY_MANTISSA_BITS	16
 
-static unsigned bkey_float_byte_offset(unsigned idx)
-{
-	return idx * sizeof(struct bkey_float);
-}
-
 struct ro_aux_tree {
 	u8			nothing[0];
 	struct bkey_float	f[];
@@ -328,8 +323,7 @@ static unsigned bset_aux_tree_buf_end(const struct bset_tree *t)
 		return t->aux_data_offset;
 	case BSET_RO_AUX_TREE:
 		return t->aux_data_offset +
-			DIV_ROUND_UP(t->size * sizeof(struct bkey_float) +
+			DIV_ROUND_UP(t->size * sizeof(struct bkey_float), 8);
-				     t->size * sizeof(u8), 8);
 	case BSET_RW_AUX_TREE:
 		return t->aux_data_offset +
 			DIV_ROUND_UP(sizeof(struct rw_aux_tree) * t->size, 8);
@@ -360,14 +354,6 @@ static struct ro_aux_tree *ro_aux_tree_base(const struct btree *b,
 	return __aux_tree_base(b, t);
 }
 
-static u8 *ro_aux_tree_prev(const struct btree *b,
-			    const struct bset_tree *t)
-{
-	EBUG_ON(bset_aux_tree_type(t) != BSET_RO_AUX_TREE);
-
-	return __aux_tree_base(b, t) + bkey_float_byte_offset(t->size);
-}
-
 static struct bkey_float *bkey_float(const struct btree *b,
 				     const struct bset_tree *t,
 				     unsigned idx)
@@ -479,15 +465,6 @@ static inline struct bkey_packed *tree_to_bkey(const struct btree *b,
 			bkey_float(b, t, j)->key_offset);
 }
 
-static struct bkey_packed *tree_to_prev_bkey(const struct btree *b,
-					     const struct bset_tree *t,
-					     unsigned j)
-{
-	unsigned prev_u64s = ro_aux_tree_prev(b, t)[j];
-
-	return (void *) ((u64 *) tree_to_bkey(b, t, j)->_data - prev_u64s);
-}
-
 static struct rw_aux_tree *rw_aux_tree(const struct btree *b,
 				       const struct bset_tree *t)
 {
@@ -585,8 +562,7 @@ static unsigned rw_aux_tree_bsearch(struct btree *b,
 }
 
 static inline unsigned bkey_mantissa(const struct bkey_packed *k,
-				     const struct bkey_float *f,
+				     const struct bkey_float *f)
-				     unsigned idx)
 {
 	u64 v;
 
@@ -617,7 +593,7 @@ static __always_inline void make_bfloat(struct btree *b, struct bset_tree *t,
 	struct bkey_packed *m = tree_to_bkey(b, t, j);
 	struct bkey_packed *l = is_power_of_2(j)
 		? min_key
-		: tree_to_prev_bkey(b, t, j >> ffs(j));
+		: tree_to_bkey(b, t, j >> ffs(j));
 	struct bkey_packed *r = is_power_of_2(j + 1)
 		? max_key
 		: tree_to_bkey(b, t, j >> (ffz(j) + 1));
@@ -668,7 +644,7 @@ static __always_inline void make_bfloat(struct btree *b, struct bset_tree *t,
 	EBUG_ON(shift < 0 || shift >= BFLOAT_FAILED);
 
 	f->exponent = shift;
-	mantissa = bkey_mantissa(m, f, j);
+	mantissa = bkey_mantissa(m, f);
 
 	/*
 	 * If we've got garbage bits, set them to all 1s - it's legal for the
@@ -690,8 +666,7 @@ static unsigned __bset_tree_capacity(struct btree *b, const struct bset_tree *t)
 
 static unsigned bset_ro_tree_capacity(struct btree *b, const struct bset_tree *t)
 {
-	return __bset_tree_capacity(b, t) /
+	return __bset_tree_capacity(b, t) / sizeof(struct bkey_float);
-		(sizeof(struct bkey_float) + sizeof(u8));
 }
 
 static unsigned bset_rw_tree_capacity(struct btree *b, const struct bset_tree *t)
@@ -720,7 +695,7 @@ static noinline void __build_rw_aux_tree(struct btree *b, struct bset_tree *t)
 
 static noinline void __build_ro_aux_tree(struct btree *b, struct bset_tree *t)
 {
-	struct bkey_packed *prev = NULL, *k = btree_bkey_first(b, t);
+	struct bkey_packed *k = btree_bkey_first(b, t);
 	struct bkey_i min_key, max_key;
 	unsigned cacheline = 1;
 
@@ -733,12 +708,12 @@ retry:
 		return;
 	}
 
-	t->extra = (t->size - rounddown_pow_of_two(t->size - 1)) << 1;
+	t->extra = eytzinger1_extra(t->size - 1);
 
 	/* First we figure out where the first key in each cacheline is */
 	eytzinger1_for_each(j, t->size - 1) {
 		while (bkey_to_cacheline(b, t, k) < cacheline)
-			prev = k, k = bkey_p_next(k);
+			k = bkey_p_next(k);
 
 		if (k >= btree_bkey_last(b, t)) {
 			/* XXX: this path sucks */
@@ -746,17 +721,12 @@ retry:
 			goto retry;
 		}
 
-		ro_aux_tree_prev(b, t)[j] = prev->u64s;
 		bkey_float(b, t, j)->key_offset =
 			bkey_to_cacheline_offset(b, t, cacheline++, k);
 
-		EBUG_ON(tree_to_prev_bkey(b, t, j) != prev);
 		EBUG_ON(tree_to_bkey(b, t, j) != k);
 	}
 
-	while (k != btree_bkey_last(b, t))
-		prev = k, k = bkey_p_next(k);
-
 	if (!bkey_pack_pos(bkey_to_packed(&min_key), b->data->min_key, b)) {
 		bkey_init(&min_key.k);
 		min_key.k.p = b->data->min_key;
@@ -999,7 +969,6 @@ static void bch2_bset_fix_lookup_table(struct btree *b,
 }
 
 void bch2_bset_insert(struct btree *b,
-		      struct btree_node_iter *iter,
 		      struct bkey_packed *where,
 		      struct bkey_i *insert,
 		      unsigned clobber_u64s)
@@ -1098,8 +1067,7 @@ static inline void prefetch_four_cachelines(void *p)
 }
 
 static inline bool bkey_mantissa_bits_dropped(const struct btree *b,
-					      const struct bkey_float *f,
+					      const struct bkey_float *f)
-					      unsigned idx)
 {
 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
 	unsigned key_bits_start = b->format.key_u64s * 64 - b->nr_key_bits;
@@ -1133,9 +1101,9 @@ static struct bkey_packed *bset_search_tree(const struct btree *b,
 			goto slowpath;
 
 		l = f->mantissa;
-		r = bkey_mantissa(packed_search, f, n);
+		r = bkey_mantissa(packed_search, f);
 
-		if (unlikely(l == r) && bkey_mantissa_bits_dropped(b, f, n))
+		if (unlikely(l == r) && bkey_mantissa_bits_dropped(b, f))
 			goto slowpath;
 
 		n = n * 2 + (l < r);

libbcachefs/bset.h

@@ -270,8 +270,8 @@ void bch2_bset_init_first(struct btree *, struct bset *);
 void bch2_bset_init_next(struct btree *, struct btree_node_entry *);
 void bch2_bset_build_aux_tree(struct btree *, struct bset_tree *, bool);
 
-void bch2_bset_insert(struct btree *, struct btree_node_iter *,
+void bch2_bset_insert(struct btree *, struct bkey_packed *, struct bkey_i *,
-		     struct bkey_packed *, struct bkey_i *, unsigned);
+		      unsigned);
 void bch2_bset_delete(struct btree *, struct bkey_packed *, unsigned);
 
 /* Bkey utility code */

libbcachefs/btree_gc.c

@@ -741,12 +741,9 @@ fsck_err:
 
 static int bch2_mark_superblocks(struct bch_fs *c)
 {
-	mutex_lock(&c->sb_lock);
 	gc_pos_set(c, gc_phase(GC_PHASE_sb));
 
-	int ret = bch2_trans_mark_dev_sbs_flags(c, BTREE_TRIGGER_gc);
+	return bch2_trans_mark_dev_sbs_flags(c, BTREE_TRIGGER_gc);
-	mutex_unlock(&c->sb_lock);
-	return ret;
 }
 
 static void bch2_gc_free(struct bch_fs *c)

libbcachefs/btree_io.c

@@ -836,14 +836,13 @@ fsck_err:
 	return ret;
 }
 
-static int bset_key_invalid(struct bch_fs *c, struct btree *b,
+static int bset_key_validate(struct bch_fs *c, struct btree *b,
-			    struct bkey_s_c k,
+			     struct bkey_s_c k,
-			    bool updated_range, int rw,
+			     bool updated_range, int rw)
-			    struct printbuf *err)
 {
-	return __bch2_bkey_invalid(c, k, btree_node_type(b), READ, err) ?:
+	return __bch2_bkey_validate(c, k, btree_node_type(b), 0) ?:
-		(!updated_range ? bch2_bkey_in_btree_node(c, b, k, err) : 0) ?:
+		(!updated_range ? bch2_bkey_in_btree_node(c, b, k, 0) : 0) ?:
-		(rw == WRITE ? bch2_bkey_val_invalid(c, k, READ, err) : 0);
+		(rw == WRITE ? bch2_bkey_val_validate(c, k, 0) : 0);
 }
 
 static bool bkey_packed_valid(struct bch_fs *c, struct btree *b,
@@ -858,12 +857,9 @@ static bool bkey_packed_valid(struct bch_fs *c, struct btree *b,
 	if (!bkeyp_u64s_valid(&b->format, k))
 		return false;
 
-	struct printbuf buf = PRINTBUF;
 	struct bkey tmp;
 	struct bkey_s u = __bkey_disassemble(b, k, &tmp);
-	bool ret = __bch2_bkey_invalid(c, u.s_c, btree_node_type(b), READ, &buf);
+	return !__bch2_bkey_validate(c, u.s_c, btree_node_type(b), BCH_VALIDATE_silent);
-	printbuf_exit(&buf);
-	return ret;
 }
 
 static int validate_bset_keys(struct bch_fs *c, struct btree *b,
@@ -915,19 +911,11 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b,
 
 		u = __bkey_disassemble(b, k, &tmp);
 
-		printbuf_reset(&buf);
+		ret = bset_key_validate(c, b, u.s_c, updated_range, write);
-		if (bset_key_invalid(c, b, u.s_c, updated_range, write, &buf)) {
+		if (ret == -BCH_ERR_fsck_delete_bkey)
-			printbuf_reset(&buf);
-			bset_key_invalid(c, b, u.s_c, updated_range, write, &buf);
-			prt_printf(&buf, "\n  ");
-			bch2_bkey_val_to_text(&buf, c, u.s_c);
-
-			btree_err(-BCH_ERR_btree_node_read_err_fixable,
-				  c, NULL, b, i, k,
-				  btree_node_bad_bkey,
-				  "invalid bkey: %s", buf.buf);
 			goto drop_this_key;
-		}
+		if (ret)
+			goto fsck_err;
 
 		if (write)
 			bch2_bkey_compat(b->c.level, b->c.btree_id, version,
@@ -1228,23 +1216,10 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
 		struct bkey tmp;
 		struct bkey_s u = __bkey_disassemble(b, k, &tmp);
 
-		printbuf_reset(&buf);
+		ret = bch2_bkey_val_validate(c, u.s_c, READ);
-
+		if (ret == -BCH_ERR_fsck_delete_bkey ||
-		if (bch2_bkey_val_invalid(c, u.s_c, READ, &buf) ||
 		    (bch2_inject_invalid_keys &&
 		     !bversion_cmp(u.k->version, MAX_VERSION))) {
-			printbuf_reset(&buf);
-
-			prt_printf(&buf, "invalid bkey: ");
-			bch2_bkey_val_invalid(c, u.s_c, READ, &buf);
-			prt_printf(&buf, "\n  ");
-			bch2_bkey_val_to_text(&buf, c, u.s_c);
-
-			btree_err(-BCH_ERR_btree_node_read_err_fixable,
-				  c, NULL, b, i, k,
-				  btree_node_bad_bkey,
-				  "%s", buf.buf);
-
 			btree_keys_account_key_drop(&b->nr, 0, k);
 
 			i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - k->u64s);
@@ -1253,6 +1228,8 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
 			set_btree_bset_end(b, b->set);
 			continue;
 		}
+		if (ret)
+			goto fsck_err;
 
 		if (u.k->type == KEY_TYPE_btree_ptr_v2) {
 			struct bkey_s_btree_ptr_v2 bp = bkey_s_to_btree_ptr_v2(u);
@@ -1767,6 +1744,8 @@ static int __bch2_btree_root_read(struct btree_trans *trans, enum btree_id id,
 
 	set_btree_node_read_in_flight(b);
 
+	/* we can't pass the trans to read_done() for fsck errors, so it must be unlocked */
+	bch2_trans_unlock(trans);
 	bch2_btree_node_read(trans, b, true);
 
 	if (btree_node_read_error(b)) {
@@ -1952,18 +1931,14 @@ static void btree_node_write_endio(struct bio *bio)
 static int validate_bset_for_write(struct bch_fs *c, struct btree *b,
 				   struct bset *i, unsigned sectors)
 {
-	struct printbuf buf = PRINTBUF;
 	bool saw_error;
-	int ret;
 
-	ret = bch2_bkey_invalid(c, bkey_i_to_s_c(&b->key),
+	int ret = bch2_bkey_validate(c, bkey_i_to_s_c(&b->key),
-				BKEY_TYPE_btree, WRITE, &buf);
+				     BKEY_TYPE_btree, WRITE);
-
+	if (ret) {
-	if (ret)
+		bch2_fs_inconsistent(c, "invalid btree node key before write");
-		bch2_fs_inconsistent(c, "invalid btree node key before write: %s", buf.buf);
-	printbuf_exit(&buf);
-	if (ret)
 		return ret;
+	}
 
 	ret = validate_bset_keys(c, b, i, WRITE, false, &saw_error) ?:
 		validate_bset(c, NULL, b, i, b->written, sectors, WRITE, false, &saw_error);

libbcachefs/btree_key_cache.c

@@ -497,11 +497,6 @@ int bch2_btree_path_traverse_cached(struct btree_trans *trans, struct btree_path
 
 	path->l[1].b = NULL;
 
-	if (bch2_btree_node_relock_notrace(trans, path, 0)) {
-		path->uptodate = BTREE_ITER_UPTODATE;
-		return 0;
-	}
-
 	int ret;
 	do {
 		ret = btree_path_traverse_cached_fast(trans, path);

libbcachefs/btree_key_cache.h

@@ -11,13 +11,27 @@ static inline size_t bch2_nr_btree_keys_need_flush(struct bch_fs *c)
 	return max_t(ssize_t, 0, nr_dirty - max_dirty);
 }
 
-static inline bool bch2_btree_key_cache_must_wait(struct bch_fs *c)
+static inline ssize_t __bch2_btree_key_cache_must_wait(struct bch_fs *c)
 {
 	size_t nr_dirty = atomic_long_read(&c->btree_key_cache.nr_dirty);
 	size_t nr_keys = atomic_long_read(&c->btree_key_cache.nr_keys);
 	size_t max_dirty = 4096 + (nr_keys * 3) / 4;
 
-	return nr_dirty > max_dirty;
+	return nr_dirty - max_dirty;
+}
+
+static inline bool bch2_btree_key_cache_must_wait(struct bch_fs *c)
+{
+	return __bch2_btree_key_cache_must_wait(c) > 0;
+}
+
+static inline bool bch2_btree_key_cache_wait_done(struct bch_fs *c)
+{
+	size_t nr_dirty = atomic_long_read(&c->btree_key_cache.nr_dirty);
+	size_t nr_keys = atomic_long_read(&c->btree_key_cache.nr_keys);
+	size_t max_dirty = 2048 + (nr_keys * 5) / 8;
+
+	return nr_dirty <= max_dirty;
 }
 
 int bch2_btree_key_cache_journal_flush(struct journal *,

libbcachefs/btree_node_scan.c

@@ -530,7 +530,7 @@ int bch2_get_scanned_nodes(struct bch_fs *c, enum btree_id btree,
 		bch_verbose(c, "%s(): recovering %s", __func__, buf.buf);
 		printbuf_exit(&buf);
 
-		BUG_ON(bch2_bkey_invalid(c, bkey_i_to_s_c(&tmp.k), BKEY_TYPE_btree, 0, NULL));
+		BUG_ON(bch2_bkey_validate(c, bkey_i_to_s_c(&tmp.k), BKEY_TYPE_btree, 0));
 
 		ret = bch2_journal_key_insert(c, btree, level + 1, &tmp.k);
 		if (ret)

libbcachefs/btree_trans_commit.c

@@ -214,7 +214,7 @@ bool bch2_btree_bset_insert_key(struct btree_trans *trans,
 
 	k = bch2_btree_node_iter_bset_pos(node_iter, b, bset_tree_last(b));
 overwrite:
-	bch2_bset_insert(b, node_iter, k, insert, clobber_u64s);
+	bch2_bset_insert(b, k, insert, clobber_u64s);
 	new_u64s = k->u64s;
 fix_iter:
 	if (clobber_u64s != new_u64s)
@@ -712,7 +712,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
 				a->k.version = journal_pos_to_bversion(&trans->journal_res,
 								(u64 *) entry - (u64 *) trans->journal_entries);
 				BUG_ON(bversion_zero(a->k.version));
-				ret = bch2_accounting_mem_mod_locked(trans, accounting_i_to_s_c(a), false);
+				ret = bch2_accounting_mem_mod_locked(trans, accounting_i_to_s_c(a), false, false);
 				if (ret)
 					goto revert_fs_usage;
 			}
@@ -798,7 +798,7 @@ revert_fs_usage:
 			struct bkey_s_accounting a = bkey_i_to_s_accounting(entry2->start);
 
 			bch2_accounting_neg(a);
-			bch2_accounting_mem_mod_locked(trans, a.c, false);
+			bch2_accounting_mem_mod_locked(trans, a.c, false, false);
 			bch2_accounting_neg(a);
 		}
 	percpu_up_read(&c->mark_lock);
@@ -818,50 +818,6 @@ static noinline void bch2_drop_overwrites_from_journal(struct btree_trans *trans
 			bch2_journal_key_overwritten(trans->c, i->btree_id, i->level, i->k->k.p);
 }
 
-static noinline int bch2_trans_commit_bkey_invalid(struct btree_trans *trans,
-						   enum bch_validate_flags flags,
-						   struct btree_insert_entry *i,
-						   struct printbuf *err)
-{
-	struct bch_fs *c = trans->c;
-
-	printbuf_reset(err);
-	prt_printf(err, "invalid bkey on insert from %s -> %ps\n",
-		   trans->fn, (void *) i->ip_allocated);
-	printbuf_indent_add(err, 2);
-
-	bch2_bkey_val_to_text(err, c, bkey_i_to_s_c(i->k));
-	prt_newline(err);
-
-	bch2_bkey_invalid(c, bkey_i_to_s_c(i->k), i->bkey_type, flags, err);
-	bch2_print_string_as_lines(KERN_ERR, err->buf);
-
-	bch2_inconsistent_error(c);
-	bch2_dump_trans_updates(trans);
-
-	return -EINVAL;
-}
-
-static noinline int bch2_trans_commit_journal_entry_invalid(struct btree_trans *trans,
-						   struct jset_entry *i)
-{
-	struct bch_fs *c = trans->c;
-	struct printbuf buf = PRINTBUF;
-
-	prt_printf(&buf, "invalid bkey on insert from %s\n", trans->fn);
-	printbuf_indent_add(&buf, 2);
-
-	bch2_journal_entry_to_text(&buf, c, i);
-	prt_newline(&buf);
-
-	bch2_print_string_as_lines(KERN_ERR, buf.buf);
-
-	bch2_inconsistent_error(c);
-	bch2_dump_trans_updates(trans);
-
-	return -EINVAL;
-}
-
 static int bch2_trans_commit_journal_pin_flush(struct journal *j,
 				struct journal_entry_pin *_pin, u64 seq)
 {
@@ -927,7 +883,7 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans, unsigned flags
 static int journal_reclaim_wait_done(struct bch_fs *c)
 {
 	int ret = bch2_journal_error(&c->journal) ?:
-		!bch2_btree_key_cache_must_wait(c);
+		bch2_btree_key_cache_wait_done(c);
 
 	if (!ret)
 		journal_reclaim_kick(&c->journal);
@@ -973,9 +929,13 @@ int bch2_trans_commit_error(struct btree_trans *trans, unsigned flags,
 		bch2_trans_unlock(trans);
 
 		trace_and_count(c, trans_blocked_journal_reclaim, trans, trace_ip);
+		track_event_change(&c->times[BCH_TIME_blocked_key_cache_flush], true);
 
 		wait_event_freezable(c->journal.reclaim_wait,
 				     (ret = journal_reclaim_wait_done(c)));
+
+		track_event_change(&c->times[BCH_TIME_blocked_key_cache_flush], false);
+
 		if (ret < 0)
 			break;
 
@@ -1060,20 +1020,19 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
 		goto out_reset;
 
 	trans_for_each_update(trans, i) {
-		struct printbuf buf = PRINTBUF;
 		enum bch_validate_flags invalid_flags = 0;
 
 		if (!(flags & BCH_TRANS_COMMIT_no_journal_res))
 			invalid_flags |= BCH_VALIDATE_write|BCH_VALIDATE_commit;
 
-		if (unlikely(bch2_bkey_invalid(c, bkey_i_to_s_c(i->k),
+		ret = bch2_bkey_validate(c, bkey_i_to_s_c(i->k),
-					       i->bkey_type, invalid_flags, &buf)))
+					 i->bkey_type, invalid_flags);
-			ret = bch2_trans_commit_bkey_invalid(trans, invalid_flags, i, &buf);
+		if (unlikely(ret)){
-		btree_insert_entry_checks(trans, i);
+			bch2_trans_inconsistent(trans, "invalid bkey on insert from %s -> %ps\n",
-		printbuf_exit(&buf);
+						trans->fn, (void *) i->ip_allocated);
-
-		if (ret)
 			return ret;
+		}
+		btree_insert_entry_checks(trans, i);
 	}
 
 	for (struct jset_entry *i = trans->journal_entries;
@@ -1084,13 +1043,14 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
 		if (!(flags & BCH_TRANS_COMMIT_no_journal_res))
 			invalid_flags |= BCH_VALIDATE_write|BCH_VALIDATE_commit;
 
-		if (unlikely(bch2_journal_entry_validate(c, NULL, i,
+		ret = bch2_journal_entry_validate(c, NULL, i,
-					bcachefs_metadata_version_current,
+						  bcachefs_metadata_version_current,
-					CPU_BIG_ENDIAN, invalid_flags)))
+						  CPU_BIG_ENDIAN, invalid_flags);
-			ret = bch2_trans_commit_journal_entry_invalid(trans, i);
+		if (unlikely(ret)) {
-
+			bch2_trans_inconsistent(trans, "invalid journal entry on insert from %s\n",
-		if (ret)
+						trans->fn);
 			return ret;
+		}
 	}
 
 	if (unlikely(!test_bit(BCH_FS_may_go_rw, &c->flags))) {

libbcachefs/btree_update_interior.c

@@ -1364,18 +1364,10 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as,
 	if (unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)))
 		bch2_journal_key_overwritten(c, b->c.btree_id, b->c.level, insert->k.p);
 
-	if (bch2_bkey_invalid(c, bkey_i_to_s_c(insert),
+	if (bch2_bkey_validate(c, bkey_i_to_s_c(insert),
-			      btree_node_type(b), WRITE, &buf) ?:
+			      btree_node_type(b), BCH_VALIDATE_write) ?:
-	    bch2_bkey_in_btree_node(c, b, bkey_i_to_s_c(insert), &buf)) {
+	    bch2_bkey_in_btree_node(c, b, bkey_i_to_s_c(insert), BCH_VALIDATE_write)) {
-		printbuf_reset(&buf);
+		bch2_fs_inconsistent(c, "%s: inserting invalid bkey", __func__);
-		prt_printf(&buf, "inserting invalid bkey\n  ");
-		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(insert));
-		prt_printf(&buf, "\n  ");
-		bch2_bkey_invalid(c, bkey_i_to_s_c(insert),
-				  btree_node_type(b), WRITE, &buf);
-		bch2_bkey_in_btree_node(c, b, bkey_i_to_s_c(insert), &buf);
-
-		bch2_fs_inconsistent(c, "%s", buf.buf);
 		dump_stack();
 	}
 

libbcachefs/buckets.c

@@ -810,6 +810,20 @@ static int __trigger_extent(struct btree_trans *trans,
 		ret = bch2_disk_accounting_mod(trans, &acc_btree_key, &replicas_sectors, 1, gc);
 		if (ret)
 			return ret;
+	} else {
+		bool insert = !(flags & BTREE_TRIGGER_overwrite);
+		struct disk_accounting_pos acc_inum_key = {
+			.type		= BCH_DISK_ACCOUNTING_inum,
+			.inum.inum	= k.k->p.inode,
+		};
+		s64 v[3] = {
+			insert ? 1 : -1,
+			insert ? k.k->size : -((s64) k.k->size),
+			replicas_sectors,
+		};
+		ret = bch2_disk_accounting_mod(trans, &acc_inum_key, v, ARRAY_SIZE(v), gc);
+		if (ret)
+			return ret;
 	}
 
 	if (bch2_bkey_rebalance_opts(k)) {
@@ -901,7 +915,6 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 				    enum bch_data_type type,
 				    unsigned sectors)
 {
-	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	int ret = 0;
 
@@ -911,7 +924,7 @@ static int __bch2_trans_mark_metadata_bucket(struct btree_trans *trans,
 		return PTR_ERR(a);
 
 	if (a->v.data_type && type && a->v.data_type != type) {
-		bch2_fsck_err(c, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
+		bch2_fsck_err(trans, FSCK_CAN_IGNORE|FSCK_NEED_FSCK,
 			      bucket_metadata_type_mismatch,
 			"bucket %llu:%llu gen %u different types of data in same bucket: %s, %s\n"
 			"while marking %s",
@@ -1032,13 +1045,18 @@ static int bch2_trans_mark_metadata_sectors(struct btree_trans *trans,
 static int __bch2_trans_mark_dev_sb(struct btree_trans *trans, struct bch_dev *ca,
 			enum btree_iter_update_trigger_flags flags)
 {
-	struct bch_sb_layout *layout = &ca->disk_sb.sb->layout;
+	struct bch_fs *c = trans->c;
+
+	mutex_lock(&c->sb_lock);
+	struct bch_sb_layout layout = ca->disk_sb.sb->layout;
+	mutex_unlock(&c->sb_lock);
+
 	u64 bucket = 0;
 	unsigned i, bucket_sectors = 0;
 	int ret;
 
-	for (i = 0; i < layout->nr_superblocks; i++) {
+	for (i = 0; i < layout.nr_superblocks; i++) {
-		u64 offset = le64_to_cpu(layout->sb_offset[i]);
+		u64 offset = le64_to_cpu(layout.sb_offset[i]);
 
 		if (offset == BCH_SB_SECTOR) {
 			ret = bch2_trans_mark_metadata_sectors(trans, ca,
@@ -1049,7 +1067,7 @@ static int __bch2_trans_mark_dev_sb(struct btree_trans *trans, struct bch_dev *c
 		}
 
 		ret = bch2_trans_mark_metadata_sectors(trans, ca, offset,
-				      offset + (1 << layout->sb_max_size_bits),
+				      offset + (1 << layout.sb_max_size_bits),
 				      BCH_DATA_sb, &bucket, &bucket_sectors, flags);
 		if (ret)
 			return ret;

libbcachefs/buckets_waiting_for_journal.c

@@ -93,7 +93,7 @@ int bch2_set_bucket_needs_journal_commit(struct buckets_waiting_for_journal *b,
 		.dev_bucket	= (u64) dev << 56 | bucket,
 		.journal_seq	= journal_seq,
 	};
-	size_t i, size, new_bits, nr_elements = 1, nr_rehashes = 0;
+	size_t i, size, new_bits, nr_elements = 1, nr_rehashes = 0, nr_rehashes_this_size = 0;
 	int ret = 0;
 
 	mutex_lock(&b->lock);
@@ -106,7 +106,7 @@ int bch2_set_bucket_needs_journal_commit(struct buckets_waiting_for_journal *b,
 	for (i = 0; i < size; i++)
 		nr_elements += t->d[i].journal_seq > flushed_seq;
 
-	new_bits = t->bits + (nr_elements * 3 > size);
+	new_bits = ilog2(roundup_pow_of_two(nr_elements * 3));
 
 	n = kvmalloc(sizeof(*n) + (sizeof(n->d[0]) << new_bits), GFP_KERNEL);
 	if (!n) {
@@ -115,7 +115,14 @@ int bch2_set_bucket_needs_journal_commit(struct buckets_waiting_for_journal *b,
 	}
 
 retry_rehash:
+	if (nr_rehashes_this_size == 3) {
+		new_bits++;
+		nr_rehashes_this_size = 0;
+	}
+
 	nr_rehashes++;
+	nr_rehashes_this_size++;
+
 	bucket_table_init(n, new_bits);
 
 	tmp = new;

libbcachefs/data_update.c

@@ -250,10 +250,8 @@ restart_drop_extra_replicas:
 		 * it's been hard to reproduce, so this should give us some more
 		 * information when it does occur:
 		 */
-		struct printbuf err = PRINTBUF;
+		int invalid = bch2_bkey_validate(c, bkey_i_to_s_c(insert), __btree_node_type(0, m->btree_id),
-		int invalid = bch2_bkey_invalid(c, bkey_i_to_s_c(insert), __btree_node_type(0, m->btree_id), 0, &err);
+						 BCH_VALIDATE_commit);
-		printbuf_exit(&err);
-
 		if (invalid) {
 			struct printbuf buf = PRINTBUF;
 

libbcachefs/dirent.c

@@ -100,20 +100,19 @@ const struct bch_hash_desc bch2_dirent_hash_desc = {
 	.is_visible	= dirent_is_visible,
 };
 
-int bch2_dirent_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_dirent_validate(struct bch_fs *c, struct bkey_s_c k,
-			enum bch_validate_flags flags,
+			 enum bch_validate_flags flags)
-			struct printbuf *err)
 {
 	struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k);
 	struct qstr d_name = bch2_dirent_get_name(d);
 	int ret = 0;
 
-	bkey_fsck_err_on(!d_name.len, c, err,
+	bkey_fsck_err_on(!d_name.len,
-			 dirent_empty_name,
+			 c, dirent_empty_name,
 			 "empty name");
 
-	bkey_fsck_err_on(bkey_val_u64s(k.k) > dirent_val_u64s(d_name.len), c, err,
+	bkey_fsck_err_on(bkey_val_u64s(k.k) > dirent_val_u64s(d_name.len),
-			 dirent_val_too_big,
+			 c, dirent_val_too_big,
 			 "value too big (%zu > %u)",
 			 bkey_val_u64s(k.k), dirent_val_u64s(d_name.len));
 
@@ -121,27 +120,27 @@ int bch2_dirent_invalid(struct bch_fs *c, struct bkey_s_c k,
 	 * Check new keys don't exceed the max length
 	 * (older keys may be larger.)
 	 */
-	bkey_fsck_err_on((flags & BCH_VALIDATE_commit) && d_name.len > BCH_NAME_MAX, c, err,
+	bkey_fsck_err_on((flags & BCH_VALIDATE_commit) && d_name.len > BCH_NAME_MAX,
-			 dirent_name_too_long,
+			 c, dirent_name_too_long,
 			 "dirent name too big (%u > %u)",
 			 d_name.len, BCH_NAME_MAX);
 
-	bkey_fsck_err_on(d_name.len != strnlen(d_name.name, d_name.len), c, err,
+	bkey_fsck_err_on(d_name.len != strnlen(d_name.name, d_name.len),
-			 dirent_name_embedded_nul,
+			 c, dirent_name_embedded_nul,
 			 "dirent has stray data after name's NUL");
 
 	bkey_fsck_err_on((d_name.len == 1 && !memcmp(d_name.name, ".", 1)) ||
-			 (d_name.len == 2 && !memcmp(d_name.name, "..", 2)), c, err,
+			 (d_name.len == 2 && !memcmp(d_name.name, "..", 2)),
-			 dirent_name_dot_or_dotdot,
+			 c, dirent_name_dot_or_dotdot,
 			 "invalid name");
 
-	bkey_fsck_err_on(memchr(d_name.name, '/', d_name.len), c, err,
+	bkey_fsck_err_on(memchr(d_name.name, '/', d_name.len),
-			 dirent_name_has_slash,
+			 c, dirent_name_has_slash,
 			 "name with /");
 
 	bkey_fsck_err_on(d.v->d_type != DT_SUBVOL &&
-			 le64_to_cpu(d.v->d_inum) == d.k->p.inode, c, err,
+			 le64_to_cpu(d.v->d_inum) == d.k->p.inode,
-			 dirent_to_itself,
+			 c, dirent_to_itself,
 			 "dirent points to own directory");
 fsck_err:
 	return ret;

libbcachefs/dirent.h

@@ -7,12 +7,11 @@
 enum bch_validate_flags;
 extern const struct bch_hash_desc bch2_dirent_hash_desc;
 
-int bch2_dirent_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_dirent_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			enum bch_validate_flags, struct printbuf *);
 void bch2_dirent_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_dirent ((struct bkey_ops) {	\
-	.key_invalid	= bch2_dirent_invalid,		\
+	.key_validate	= bch2_dirent_validate,		\
 	.val_to_text	= bch2_dirent_to_text,		\
 	.min_val_size	= 16,				\
 })

libbcachefs/disk_accounting.c

@@ -126,9 +126,8 @@ static inline bool is_zero(char *start, char *end)
 
 #define field_end(p, member)	(((void *) (&p.member)) + sizeof(p.member))
 
-int bch2_accounting_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_accounting_validate(struct bch_fs *c, struct bkey_s_c k,
-			    enum bch_validate_flags flags,
+			     enum bch_validate_flags flags)
-			    struct printbuf *err)
 {
 	struct disk_accounting_pos acc_k;
 	bpos_to_disk_accounting_pos(&acc_k, k.k->p);
@@ -144,18 +143,18 @@ int bch2_accounting_invalid(struct bch_fs *c, struct bkey_s_c k,
 		break;
 	case BCH_DISK_ACCOUNTING_replicas:
 		bkey_fsck_err_on(!acc_k.replicas.nr_devs,
-				 c, err, accounting_key_replicas_nr_devs_0,
+				 c, 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,
+				 c, 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],
+			bkey_fsck_err_on(acc_k.replicas.devs[i] >= acc_k.replicas.devs[i + 1],
-					 c, err, accounting_key_replicas_devs_unsorted,
+					 c, accounting_key_replicas_devs_unsorted,
 					 "accounting key replicas entry with unsorted devs");
 
 		end = (void *) &acc_k.replicas + replicas_entry_bytes(&acc_k.replicas);
@@ -178,7 +177,7 @@ int bch2_accounting_invalid(struct bch_fs *c, struct bkey_s_c k,
 	}
 
 	bkey_fsck_err_on(!is_zero(end, (void *) (&acc_k + 1)),
-			 c, err, accounting_key_junk_at_end,
+			 c, accounting_key_junk_at_end,
 			 "junk at end of accounting key");
 fsck_err:
 	return ret;
@@ -528,6 +527,9 @@ int bch2_gc_accounting_done(struct bch_fs *c)
 		struct disk_accounting_pos acc_k;
 		bpos_to_disk_accounting_pos(&acc_k, e->pos);
 
+		if (acc_k.type >= BCH_DISK_ACCOUNTING_TYPE_NR)
+			continue;
+
 		u64 src_v[BCH_ACCOUNTING_MAX_COUNTERS];
 		u64 dst_v[BCH_ACCOUNTING_MAX_COUNTERS];
 
@@ -564,7 +566,7 @@ int bch2_gc_accounting_done(struct bch_fs *c)
 					struct { __BKEY_PADDED(k, BCH_ACCOUNTING_MAX_COUNTERS); } k_i;
 
 					accounting_key_init(&k_i.k, &acc_k, src_v, nr);
-					bch2_accounting_mem_mod_locked(trans, bkey_i_to_s_c_accounting(&k_i.k), false);
+					bch2_accounting_mem_mod_locked(trans, bkey_i_to_s_c_accounting(&k_i.k), false, false);
 
 					preempt_disable();
 					struct bch_fs_usage_base *dst = this_cpu_ptr(c->usage);
@@ -593,7 +595,7 @@ static int accounting_read_key(struct btree_trans *trans, struct bkey_s_c k)
 		return 0;
 
 	percpu_down_read(&c->mark_lock);
-	int ret = __bch2_accounting_mem_mod(c, bkey_s_c_to_accounting(k), false);
+	int ret = bch2_accounting_mem_mod_locked(trans, bkey_s_c_to_accounting(k), false, true);
 	percpu_up_read(&c->mark_lock);
 
 	if (bch2_accounting_key_is_zero(bkey_s_c_to_accounting(k)) &&
@@ -760,6 +762,15 @@ void bch2_verify_accounting_clean(struct bch_fs *c)
 			struct bkey_s_c_accounting a = bkey_s_c_to_accounting(k);
 			unsigned nr = bch2_accounting_counters(k.k);
 
+			struct disk_accounting_pos acc_k;
+			bpos_to_disk_accounting_pos(&acc_k, k.k->p);
+
+			if (acc_k.type >= BCH_DISK_ACCOUNTING_TYPE_NR)
+				continue;
+
+			if (acc_k.type == BCH_DISK_ACCOUNTING_inum)
+				continue;
+
 			bch2_accounting_mem_read(c, k.k->p, v, nr);
 
 			if (memcmp(a.v->d, v, nr * sizeof(u64))) {
@@ -775,9 +786,6 @@ void bch2_verify_accounting_clean(struct bch_fs *c)
 				mismatch = true;
 			}
 
-			struct disk_accounting_pos acc_k;
-			bpos_to_disk_accounting_pos(&acc_k, a.k->p);
-
 			switch (acc_k.type) {
 			case BCH_DISK_ACCOUNTING_persistent_reserved:
 				base.reserved += acc_k.persistent_reserved.nr_replicas * a.v->d[0];

libbcachefs/disk_accounting.h

@@ -82,14 +82,13 @@ int bch2_disk_accounting_mod(struct btree_trans *, struct disk_accounting_pos *,
 			     s64 *, unsigned, bool);
 int bch2_mod_dev_cached_sectors(struct btree_trans *, unsigned, s64, bool);
 
-int bch2_accounting_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_accounting_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			    enum bch_validate_flags, struct printbuf *);
 void bch2_accounting_key_to_text(struct printbuf *, struct disk_accounting_pos *);
 void bch2_accounting_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 void bch2_accounting_swab(struct bkey_s);
 
 #define bch2_bkey_ops_accounting ((struct bkey_ops) {	\
-	.key_invalid	= bch2_accounting_invalid,	\
+	.key_validate	= bch2_accounting_validate,	\
 	.val_to_text	= bch2_accounting_to_text,	\
 	.swab		= bch2_accounting_swab,		\
 	.min_val_size	= 8,				\
@@ -107,8 +106,40 @@ static inline int accounting_pos_cmp(const void *_l, const void *_r)
 int bch2_accounting_mem_insert(struct bch_fs *, struct bkey_s_c_accounting, bool);
 void bch2_accounting_mem_gc(struct bch_fs *);
 
-static inline int __bch2_accounting_mem_mod(struct bch_fs *c, struct bkey_s_c_accounting a, bool gc)
+/*
+ * Update in memory counters so they match the btree update we're doing; called
+ * from transaction commit path
+ */
+static inline int bch2_accounting_mem_mod_locked(struct btree_trans *trans, struct bkey_s_c_accounting a, bool gc, bool read)
 {
+	struct bch_fs *c = trans->c;
+	struct disk_accounting_pos acc_k;
+	bpos_to_disk_accounting_pos(&acc_k, a.k->p);
+
+	if (acc_k.type == BCH_DISK_ACCOUNTING_inum)
+		return 0;
+
+	if (!gc && !read) {
+		switch (acc_k.type) {
+		case BCH_DISK_ACCOUNTING_persistent_reserved:
+			trans->fs_usage_delta.reserved += acc_k.persistent_reserved.nr_replicas * a.v->d[0];
+			break;
+		case BCH_DISK_ACCOUNTING_replicas:
+			fs_usage_data_type_to_base(&trans->fs_usage_delta, acc_k.replicas.data_type, a.v->d[0]);
+			break;
+		case BCH_DISK_ACCOUNTING_dev_data_type:
+			rcu_read_lock();
+			struct bch_dev *ca = bch2_dev_rcu(c, acc_k.dev_data_type.dev);
+			if (ca) {
+				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].buckets, a.v->d[0]);
+				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].sectors, a.v->d[1]);
+				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].fragmented, a.v->d[2]);
+			}
+			rcu_read_unlock();
+			break;
+		}
+	}
+
 	struct bch_accounting_mem *acc = &c->accounting;
 	unsigned idx;
 
@@ -130,45 +161,10 @@ static inline int __bch2_accounting_mem_mod(struct bch_fs *c, struct bkey_s_c_ac
 	return 0;
 }
 
-/*
- * Update in memory counters so they match the btree update we're doing; called
- * from transaction commit path
- */
-static inline int bch2_accounting_mem_mod_locked(struct btree_trans *trans, struct bkey_s_c_accounting a, bool gc)
-{
-	struct bch_fs *c = trans->c;
-
-	if (!gc) {
-		struct disk_accounting_pos acc_k;
-		bpos_to_disk_accounting_pos(&acc_k, a.k->p);
-
-		switch (acc_k.type) {
-		case BCH_DISK_ACCOUNTING_persistent_reserved:
-			trans->fs_usage_delta.reserved += acc_k.persistent_reserved.nr_replicas * a.v->d[0];
-			break;
-		case BCH_DISK_ACCOUNTING_replicas:
-			fs_usage_data_type_to_base(&trans->fs_usage_delta, acc_k.replicas.data_type, a.v->d[0]);
-			break;
-		case BCH_DISK_ACCOUNTING_dev_data_type:
-			rcu_read_lock();
-			struct bch_dev *ca = bch2_dev_rcu(c, acc_k.dev_data_type.dev);
-			if (ca) {
-				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].buckets, a.v->d[0]);
-				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].sectors, a.v->d[1]);
-				this_cpu_add(ca->usage->d[acc_k.dev_data_type.data_type].fragmented, a.v->d[2]);
-			}
-			rcu_read_unlock();
-			break;
-		}
-	}
-
-	return __bch2_accounting_mem_mod(c, a, gc);
-}
-
 static inline int bch2_accounting_mem_add(struct btree_trans *trans, struct bkey_s_c_accounting a, bool gc)
 {
 	percpu_down_read(&trans->c->mark_lock);
-	int ret = bch2_accounting_mem_mod_locked(trans, a, gc);
+	int ret = bch2_accounting_mem_mod_locked(trans, a, gc, false);
 	percpu_up_read(&trans->c->mark_lock);
 	return ret;
 }

libbcachefs/disk_accounting_format.h

@@ -103,7 +103,8 @@ static inline bool data_type_is_hidden(enum bch_data_type type)
 	x(compression,		4)		\
 	x(snapshot,		5)		\
 	x(btree,		6)		\
-	x(rebalance_work,	7)
+	x(rebalance_work,	7)		\
+	x(inum,			8)
 
 enum disk_accounting_type {
 #define x(f, nr)	BCH_DISK_ACCOUNTING_##f	= nr,
@@ -136,6 +137,10 @@ struct bch_acct_btree {
 	__u32			id;
 } __packed;
 
+struct bch_acct_inum {
+	__u64			inum;
+} __packed;
+
 struct bch_acct_rebalance_work {
 };
 
@@ -152,6 +157,7 @@ struct disk_accounting_pos {
 		struct bch_acct_snapshot	snapshot;
 		struct bch_acct_btree		btree;
 		struct bch_acct_rebalance_work	rebalance_work;
+		struct bch_acct_inum		inum;
 		} __packed;
 	} __packed;
 		struct bpos			_pad;

libbcachefs/ec.c

@@ -107,24 +107,23 @@ struct ec_bio {
 
 /* Stripes btree keys: */
 
-int bch2_stripe_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_stripe_validate(struct bch_fs *c, struct bkey_s_c k,
-			enum bch_validate_flags flags,
+			 enum bch_validate_flags flags)
-			struct printbuf *err)
 {
 	const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
 	int ret = 0;
 
 	bkey_fsck_err_on(bkey_eq(k.k->p, POS_MIN) ||
-			 bpos_gt(k.k->p, POS(0, U32_MAX)), c, err,
+			 bpos_gt(k.k->p, POS(0, U32_MAX)),
-			 stripe_pos_bad,
+			 c, stripe_pos_bad,
 			 "stripe at bad pos");
 
-	bkey_fsck_err_on(bkey_val_u64s(k.k) < stripe_val_u64s(s), c, err,
+	bkey_fsck_err_on(bkey_val_u64s(k.k) < stripe_val_u64s(s),
-			 stripe_val_size_bad,
+			 c, stripe_val_size_bad,
 			 "incorrect value size (%zu < %u)",
 			 bkey_val_u64s(k.k), stripe_val_u64s(s));
 
-	ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
+	ret = bch2_bkey_ptrs_validate(c, k, flags);
 fsck_err:
 	return ret;
 }

libbcachefs/ec.h

@@ -8,8 +8,7 @@
 
 enum bch_validate_flags;
 
-int bch2_stripe_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_stripe_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			enum bch_validate_flags, struct printbuf *);
 void bch2_stripe_to_text(struct printbuf *, struct bch_fs *,
 			 struct bkey_s_c);
 int bch2_trigger_stripe(struct btree_trans *, enum btree_id, unsigned,
@@ -17,7 +16,7 @@ int bch2_trigger_stripe(struct btree_trans *, enum btree_id, unsigned,
 			enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_stripe ((struct bkey_ops) {	\
-	.key_invalid	= bch2_stripe_invalid,		\
+	.key_validate	= bch2_stripe_validate,		\
 	.val_to_text	= bch2_stripe_to_text,		\
 	.swab		= bch2_ptr_swab,		\
 	.trigger	= bch2_trigger_stripe,		\

libbcachefs/errcode.h

@@ -145,6 +145,7 @@
 	x(BCH_ERR_transaction_restart,	transaction_restart_journal_preres_get)	\
 	x(BCH_ERR_transaction_restart,	transaction_restart_split_race)		\
 	x(BCH_ERR_transaction_restart,	transaction_restart_write_buffer_flush)	\
+	x(BCH_ERR_transaction_restart,	transaction_restart_freeing_inode)		\
 	x(BCH_ERR_transaction_restart,	transaction_restart_nested)		\
 	x(0,				no_btree_node)				\
 	x(BCH_ERR_no_btree_node,	no_btree_node_relock)			\
@@ -166,6 +167,7 @@
 	x(0,				journal_reclaim_would_deadlock)		\
 	x(EINVAL,			fsck)					\
 	x(BCH_ERR_fsck,			fsck_fix)				\
+	x(BCH_ERR_fsck,			fsck_delete_bkey)			\
 	x(BCH_ERR_fsck,			fsck_ignore)				\
 	x(BCH_ERR_fsck,			fsck_errors_not_fixed)			\
 	x(BCH_ERR_fsck,			fsck_repair_unimplemented)		\

libbcachefs/error.c

@@ -416,6 +416,28 @@ err:
 	return ret;
 }
 
+int __bch2_bkey_fsck_err(struct bch_fs *c,
+			 struct bkey_s_c k,
+			 enum bch_fsck_flags flags,
+			 enum bch_sb_error_id err,
+			 const char *fmt, ...)
+{
+	struct printbuf buf = PRINTBUF;
+	va_list args;
+
+	prt_str(&buf, "invalid bkey ");
+	bch2_bkey_val_to_text(&buf, c, k);
+	prt_str(&buf, "\n  ");
+	va_start(args, fmt);
+	prt_vprintf(&buf, fmt, args);
+	va_end(args);
+	prt_str(&buf, ": delete?");
+
+	int ret = __bch2_fsck_err(c, NULL, flags, err, "%s", buf.buf);
+	printbuf_exit(&buf);
+	return ret;
+}
+
 void bch2_flush_fsck_errs(struct bch_fs *c)
 {
 	struct fsck_err_state *s, *n;

libbcachefs/error.h

@@ -4,6 +4,7 @@
 
 #include <linux/list.h>
 #include <linux/printk.h>
+#include "bkey_types.h"
 #include "sb-errors.h"
 
 struct bch_dev;
@@ -166,24 +167,30 @@ 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)
+__printf(5, 6)
-static inline void bch2_bkey_fsck_err(struct bch_fs *c,
+int __bch2_bkey_fsck_err(struct bch_fs *,
-				     struct printbuf *err_msg,
+			 struct bkey_s_c,
-				     enum bch_sb_error_id err_type,
+			 enum bch_fsck_flags,
-				     const char *fmt, ...)
+			 enum bch_sb_error_id,
-{
+			 const char *, ...);
-	va_list args;
 
-	va_start(args, fmt);
+/*
-	prt_vprintf(err_msg, fmt, args);
+ * for now, bkey fsck errors are always handled by deleting the entire key -
-	va_end(args);
+ * this will change at some point
-}
+ */
-
+#define bkey_fsck_err(c, _err_type, _err_msg, ...)			\
-#define bkey_fsck_err(c, _err_msg, _err_type, ...)			\
 do {									\
-	prt_printf(_err_msg, __VA_ARGS__);				\
+	if ((flags & BCH_VALIDATE_silent)) {				\
-	bch2_sb_error_count(c, BCH_FSCK_ERR_##_err_type);		\
+		ret = -BCH_ERR_fsck_delete_bkey;			\
-	ret = -BCH_ERR_invalid_bkey;					\
+		goto fsck_err;						\
+	}								\
+	int _ret = __bch2_bkey_fsck_err(c, k, FSCK_CAN_FIX,		\
+				BCH_FSCK_ERR_##_err_type,		\
+				_err_msg, ##__VA_ARGS__);		\
+	if (_ret != -BCH_ERR_fsck_fix &&				\
+	    _ret != -BCH_ERR_fsck_ignore)				\
+		ret = _ret;						\
+	ret = -BCH_ERR_fsck_delete_bkey;				\
 	goto fsck_err;							\
 } while (0)
 

libbcachefs/extents.c

@@ -171,17 +171,16 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
 
 /* KEY_TYPE_btree_ptr: */
 
-int bch2_btree_ptr_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_btree_ptr_validate(struct bch_fs *c, struct bkey_s_c k,
-			   enum bch_validate_flags flags,
+			    enum bch_validate_flags flags)
-			   struct printbuf *err)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(bkey_val_u64s(k.k) > BCH_REPLICAS_MAX, c, err,
+	bkey_fsck_err_on(bkey_val_u64s(k.k) > BCH_REPLICAS_MAX,
-			 btree_ptr_val_too_big,
+			 c, btree_ptr_val_too_big,
 			 "value too big (%zu > %u)", bkey_val_u64s(k.k), BCH_REPLICAS_MAX);
 
-	ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
+	ret = bch2_bkey_ptrs_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
@@ -192,28 +191,27 @@ void bch2_btree_ptr_to_text(struct printbuf *out, struct bch_fs *c,
 	bch2_bkey_ptrs_to_text(out, c, k);
 }
 
-int bch2_btree_ptr_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_btree_ptr_v2_validate(struct bch_fs *c, struct bkey_s_c k,
-			      enum bch_validate_flags flags,
+			       enum bch_validate_flags flags)
-			      struct printbuf *err)
 {
 	struct bkey_s_c_btree_ptr_v2 bp = bkey_s_c_to_btree_ptr_v2(k);
 	int ret = 0;
 
 	bkey_fsck_err_on(bkey_val_u64s(k.k) > BKEY_BTREE_PTR_VAL_U64s_MAX,
-			 c, err, btree_ptr_v2_val_too_big,
+			 c, btree_ptr_v2_val_too_big,
 			 "value too big (%zu > %zu)",
 			 bkey_val_u64s(k.k), BKEY_BTREE_PTR_VAL_U64s_MAX);
 
 	bkey_fsck_err_on(bpos_ge(bp.v->min_key, bp.k->p),
-			 c, err, btree_ptr_v2_min_key_bad,
+			 c, btree_ptr_v2_min_key_bad,
 			 "min_key > key");
 
 	if (flags & BCH_VALIDATE_write)
 		bkey_fsck_err_on(!bp.v->sectors_written,
-				 c, err, btree_ptr_v2_written_0,
+				 c, btree_ptr_v2_written_0,
 				 "sectors_written == 0");
 
-	ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
+	ret = bch2_bkey_ptrs_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
@@ -399,15 +397,14 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
 
 /* KEY_TYPE_reservation: */
 
-int bch2_reservation_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_reservation_validate(struct bch_fs *c, struct bkey_s_c k,
-			     enum bch_validate_flags flags,
+			      enum bch_validate_flags flags)
-			     struct printbuf *err)
 {
 	struct bkey_s_c_reservation r = bkey_s_c_to_reservation(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(!r.v->nr_replicas || r.v->nr_replicas > BCH_REPLICAS_MAX, c, err,
+	bkey_fsck_err_on(!r.v->nr_replicas || r.v->nr_replicas > BCH_REPLICAS_MAX,
-			 reservation_key_nr_replicas_invalid,
+			 c, reservation_key_nr_replicas_invalid,
 			 "invalid nr_replicas (%u)", r.v->nr_replicas);
 fsck_err:
 	return ret;
@@ -1102,14 +1099,12 @@ void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
 	}
 }
 
-
+static int extent_ptr_validate(struct bch_fs *c,
-static int extent_ptr_invalid(struct bch_fs *c,
+			       struct bkey_s_c k,
-			      struct bkey_s_c k,
+			       enum bch_validate_flags flags,
-			      enum bch_validate_flags flags,
+			       const struct bch_extent_ptr *ptr,
-			      const struct bch_extent_ptr *ptr,
+			       unsigned size_ondisk,
-			      unsigned size_ondisk,
+			       bool metadata)
-			      bool metadata,
-			      struct printbuf *err)
 {
 	int ret = 0;
 
@@ -1128,28 +1123,27 @@ static int extent_ptr_invalid(struct bch_fs *c,
 
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 	bkey_for_each_ptr(ptrs, ptr2)
-		bkey_fsck_err_on(ptr != ptr2 && ptr->dev == ptr2->dev, c, err,
+		bkey_fsck_err_on(ptr != ptr2 && ptr->dev == ptr2->dev,
-				 ptr_to_duplicate_device,
+				 c, ptr_to_duplicate_device,
 				 "multiple pointers to same device (%u)", ptr->dev);
 
 
-	bkey_fsck_err_on(bucket >= nbuckets, c, err,
+	bkey_fsck_err_on(bucket >= nbuckets,
-			 ptr_after_last_bucket,
+			 c, ptr_after_last_bucket,
 			 "pointer past last bucket (%llu > %llu)", bucket, nbuckets);
-	bkey_fsck_err_on(bucket < first_bucket, c, err,
+	bkey_fsck_err_on(bucket < first_bucket,
-			 ptr_before_first_bucket,
+			 c, ptr_before_first_bucket,
 			 "pointer before first bucket (%llu < %u)", bucket, first_bucket);
-	bkey_fsck_err_on(bucket_offset + size_ondisk > bucket_size, c, err,
+	bkey_fsck_err_on(bucket_offset + size_ondisk > bucket_size,
-			 ptr_spans_multiple_buckets,
+			 c, ptr_spans_multiple_buckets,
 			 "pointer spans multiple buckets (%u + %u > %u)",
 		       bucket_offset, size_ondisk, bucket_size);
 fsck_err:
 	return ret;
 }
 
-int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_bkey_ptrs_validate(struct bch_fs *c, struct bkey_s_c k,
-			   enum bch_validate_flags flags,
+			    enum bch_validate_flags flags)
-			   struct printbuf *err)
 {
 	struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 	const union bch_extent_entry *entry;
@@ -1164,25 +1158,24 @@ int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
 		size_ondisk = btree_sectors(c);
 
 	bkey_extent_entry_for_each(ptrs, entry) {
-		bkey_fsck_err_on(__extent_entry_type(entry) >= BCH_EXTENT_ENTRY_MAX, c, err,
+		bkey_fsck_err_on(__extent_entry_type(entry) >= BCH_EXTENT_ENTRY_MAX,
-			extent_ptrs_invalid_entry,
+				 c, extent_ptrs_invalid_entry,
-			"invalid extent entry type (got %u, max %u)",
+				 "invalid extent entry type (got %u, max %u)",
-			__extent_entry_type(entry), BCH_EXTENT_ENTRY_MAX);
+				 __extent_entry_type(entry), BCH_EXTENT_ENTRY_MAX);
 
 		bkey_fsck_err_on(bkey_is_btree_ptr(k.k) &&
-				 !extent_entry_is_ptr(entry), c, err,
+				 !extent_entry_is_ptr(entry),
-				 btree_ptr_has_non_ptr,
+				 c, btree_ptr_has_non_ptr,
 				 "has non ptr field");
 
 		switch (extent_entry_type(entry)) {
 		case BCH_EXTENT_ENTRY_ptr:
-			ret = extent_ptr_invalid(c, k, flags, &entry->ptr,
+			ret = extent_ptr_validate(c, k, flags, &entry->ptr, size_ondisk, false);
-						 size_ondisk, false, err);
 			if (ret)
 				return ret;
 
-			bkey_fsck_err_on(entry->ptr.cached && have_ec, c, err,
+			bkey_fsck_err_on(entry->ptr.cached && have_ec,
-					 ptr_cached_and_erasure_coded,
+					 c, ptr_cached_and_erasure_coded,
 					 "cached, erasure coded ptr");
 
 			if (!entry->ptr.unwritten)
@@ -1199,44 +1192,50 @@ int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
 		case BCH_EXTENT_ENTRY_crc128:
 			crc = bch2_extent_crc_unpack(k.k, entry_to_crc(entry));
 
-			bkey_fsck_err_on(crc.offset + crc.live_size > crc.uncompressed_size, c, err,
+			bkey_fsck_err_on(crc.offset + crc.live_size > crc.uncompressed_size,
-					 ptr_crc_uncompressed_size_too_small,
+					 c, ptr_crc_uncompressed_size_too_small,
 					 "checksum offset + key size > uncompressed size");
-			bkey_fsck_err_on(!bch2_checksum_type_valid(c, crc.csum_type), c, err,
+			bkey_fsck_err_on(!bch2_checksum_type_valid(c, crc.csum_type),
-					 ptr_crc_csum_type_unknown,
+					 c, ptr_crc_csum_type_unknown,
 					 "invalid checksum type");
-			bkey_fsck_err_on(crc.compression_type >= BCH_COMPRESSION_TYPE_NR, c, err,
+			bkey_fsck_err_on(crc.compression_type >= BCH_COMPRESSION_TYPE_NR,
-					 ptr_crc_compression_type_unknown,
+					 c, ptr_crc_compression_type_unknown,
 					 "invalid compression type");
 
 			if (bch2_csum_type_is_encryption(crc.csum_type)) {
 				if (nonce == UINT_MAX)
 					nonce = crc.offset + crc.nonce;
 				else if (nonce != crc.offset + crc.nonce)
-					bkey_fsck_err(c, err, ptr_crc_nonce_mismatch,
+					bkey_fsck_err(c, ptr_crc_nonce_mismatch,
 						      "incorrect nonce");
 			}
 
-			bkey_fsck_err_on(crc_since_last_ptr, c, err,
+			bkey_fsck_err_on(crc_since_last_ptr,
-					 ptr_crc_redundant,
+					 c, ptr_crc_redundant,
 					 "redundant crc entry");
 			crc_since_last_ptr = true;
 
 			bkey_fsck_err_on(crc_is_encoded(crc) &&
 					 (crc.uncompressed_size > c->opts.encoded_extent_max >> 9) &&
-					 (flags & (BCH_VALIDATE_write|BCH_VALIDATE_commit)), c, err,
+					 (flags & (BCH_VALIDATE_write|BCH_VALIDATE_commit)),
-					 ptr_crc_uncompressed_size_too_big,
+					 c, ptr_crc_uncompressed_size_too_big,
 					 "too large encoded extent");
 
 			size_ondisk = crc.compressed_size;
 			break;
 		case BCH_EXTENT_ENTRY_stripe_ptr:
-			bkey_fsck_err_on(have_ec, c, err,
+			bkey_fsck_err_on(have_ec,
-					 ptr_stripe_redundant,
+					 c, ptr_stripe_redundant,
 					 "redundant stripe entry");
 			have_ec = true;
 			break;
 		case BCH_EXTENT_ENTRY_rebalance: {
+			/*
+			 * this shouldn't be a fsck error, for forward
+			 * compatibility; the rebalance code should just refetch
+			 * the compression opt if it's unknown
+			 */
+#if 0
 			const struct bch_extent_rebalance *r = &entry->rebalance;
 
 			if (!bch2_compression_opt_valid(r->compression)) {
@@ -1245,28 +1244,29 @@ int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
 					   opt.type, opt.level);
 				return -BCH_ERR_invalid_bkey;
 			}
+#endif
 			break;
 		}
 		}
 	}
 
-	bkey_fsck_err_on(!nr_ptrs, c, err,
+	bkey_fsck_err_on(!nr_ptrs,
-			 extent_ptrs_no_ptrs,
+			 c, extent_ptrs_no_ptrs,
 			 "no ptrs");
-	bkey_fsck_err_on(nr_ptrs > BCH_BKEY_PTRS_MAX, c, err,
+	bkey_fsck_err_on(nr_ptrs > BCH_BKEY_PTRS_MAX,
-			 extent_ptrs_too_many_ptrs,
+			 c, extent_ptrs_too_many_ptrs,
 			 "too many ptrs: %u > %u", nr_ptrs, BCH_BKEY_PTRS_MAX);
-	bkey_fsck_err_on(have_written && have_unwritten, c, err,
+	bkey_fsck_err_on(have_written && have_unwritten,
-			 extent_ptrs_written_and_unwritten,
+			 c, extent_ptrs_written_and_unwritten,
 			 "extent with unwritten and written ptrs");
-	bkey_fsck_err_on(k.k->type != KEY_TYPE_extent && have_unwritten, c, err,
+	bkey_fsck_err_on(k.k->type != KEY_TYPE_extent && have_unwritten,
-			 extent_ptrs_unwritten,
+			 c, extent_ptrs_unwritten,
 			 "has unwritten ptrs");
-	bkey_fsck_err_on(crc_since_last_ptr, c, err,
+	bkey_fsck_err_on(crc_since_last_ptr,
-			 extent_ptrs_redundant_crc,
+			 c, extent_ptrs_redundant_crc,
 			 "redundant crc entry");
-	bkey_fsck_err_on(have_ec, c, err,
+	bkey_fsck_err_on(have_ec,
-			 extent_ptrs_redundant_stripe,
+			 c, extent_ptrs_redundant_stripe,
 			 "redundant stripe entry");
 fsck_err:
 	return ret;

libbcachefs/extents.h

@@ -409,26 +409,26 @@ int bch2_bkey_pick_read_device(struct bch_fs *, struct bkey_s_c,
 
 /* KEY_TYPE_btree_ptr: */
 
-int bch2_btree_ptr_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_btree_ptr_validate(struct bch_fs *, struct bkey_s_c,
-			   enum bch_validate_flags, struct printbuf *);
+			    enum bch_validate_flags);
 void bch2_btree_ptr_to_text(struct printbuf *, struct bch_fs *,
 			    struct bkey_s_c);
 
-int bch2_btree_ptr_v2_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_btree_ptr_v2_validate(struct bch_fs *, struct bkey_s_c,
-			      enum bch_validate_flags, struct printbuf *);
+			       enum bch_validate_flags);
 void bch2_btree_ptr_v2_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 void bch2_btree_ptr_v2_compat(enum btree_id, unsigned, unsigned,
 			      int, struct bkey_s);
 
 #define bch2_bkey_ops_btree_ptr ((struct bkey_ops) {		\
-	.key_invalid	= bch2_btree_ptr_invalid,		\
+	.key_validate	= bch2_btree_ptr_validate,		\
 	.val_to_text	= bch2_btree_ptr_to_text,		\
 	.swab		= bch2_ptr_swab,			\
 	.trigger	= bch2_trigger_extent,			\
 })
 
 #define bch2_bkey_ops_btree_ptr_v2 ((struct bkey_ops) {		\
-	.key_invalid	= bch2_btree_ptr_v2_invalid,		\
+	.key_validate	= bch2_btree_ptr_v2_validate,		\
 	.val_to_text	= bch2_btree_ptr_v2_to_text,		\
 	.swab		= bch2_ptr_swab,			\
 	.compat		= bch2_btree_ptr_v2_compat,		\
@@ -441,7 +441,7 @@ void bch2_btree_ptr_v2_compat(enum btree_id, unsigned, unsigned,
 bool bch2_extent_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 
 #define bch2_bkey_ops_extent ((struct bkey_ops) {		\
-	.key_invalid	= bch2_bkey_ptrs_invalid,		\
+	.key_validate	= bch2_bkey_ptrs_validate,		\
 	.val_to_text	= bch2_bkey_ptrs_to_text,		\
 	.swab		= bch2_ptr_swab,			\
 	.key_normalize	= bch2_extent_normalize,		\
@@ -451,13 +451,13 @@ bool bch2_extent_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 
 /* KEY_TYPE_reservation: */
 
-int bch2_reservation_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_reservation_validate(struct bch_fs *, struct bkey_s_c,
-			     enum bch_validate_flags, struct printbuf *);
+			      enum bch_validate_flags);
 void bch2_reservation_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 bool bch2_reservation_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 
 #define bch2_bkey_ops_reservation ((struct bkey_ops) {		\
-	.key_invalid	= bch2_reservation_invalid,		\
+	.key_validate	= bch2_reservation_validate,		\
 	.val_to_text	= bch2_reservation_to_text,		\
 	.key_merge	= bch2_reservation_merge,		\
 	.trigger	= bch2_trigger_reservation,		\
@@ -683,8 +683,8 @@ bool bch2_extent_normalize(struct bch_fs *, struct bkey_s);
 void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *, const struct bch_extent_ptr *);
 void bch2_bkey_ptrs_to_text(struct printbuf *, struct bch_fs *,
 			    struct bkey_s_c);
-int bch2_bkey_ptrs_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_bkey_ptrs_validate(struct bch_fs *, struct bkey_s_c,
-			   enum bch_validate_flags, struct printbuf *);
+			    enum bch_validate_flags);
 
 void bch2_ptr_swab(struct bkey_s);
 

libbcachefs/fs.c

@@ -108,7 +108,7 @@ retry:
 		goto retry;
 
 	bch2_fs_fatal_err_on(bch2_err_matches(ret, ENOENT), c,
-			     "%s: inode %u:%llu not found when updating",
+			     "%s: inode %llu:%llu not found when updating",
 			     bch2_err_str(ret),
 			     inode_inum(inode).subvol,
 			     inode_inum(inode).inum);
@@ -185,16 +185,22 @@ static void __wait_on_freeing_inode(struct inode *inode)
 	finish_wait(wq, &wait.wq_entry);
 }
 
-static struct bch_inode_info *bch2_inode_hash_find(struct bch_fs *c, subvol_inum inum)
+static struct bch_inode_info *bch2_inode_hash_find(struct bch_fs *c, struct btree_trans *trans,
+						   subvol_inum inum)
 {
 	struct bch_inode_info *inode;
 repeat:
-	inode = rhashtable_lookup_fast(&c->vfs_inodes_table, &inum,
+	inode = rhashtable_lookup_fast(&c->vfs_inodes_table, &inum, bch2_vfs_inodes_params);
-				       bch2_vfs_inodes_params);
 	if (inode) {
 		spin_lock(&inode->v.i_lock);
 		if ((inode->v.i_state & (I_FREEING|I_WILL_FREE))) {
+			if (trans)
+				bch2_trans_unlock(trans);
 			__wait_on_freeing_inode(&inode->v);
+			if (trans) {
+				trace_and_count(c, trans_restart_freeing_inode, trans, _THIS_IP_);
+				return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_freeing_inode));
+			}
 			goto repeat;
 		}
 		__iget(&inode->v);
@@ -215,23 +221,27 @@ static void bch2_inode_hash_remove(struct bch_fs *c, struct bch_inode_info *inod
 	}
 }
 
-static struct bch_inode_info *bch2_inode_hash_insert(struct bch_fs *c, struct bch_inode_info *inode)
+static struct bch_inode_info *bch2_inode_hash_insert(struct bch_fs *c,
+						     struct btree_trans *trans,
+						     struct bch_inode_info *inode)
 {
 	struct bch_inode_info *old = inode;
 retry:
 	if (unlikely(rhashtable_lookup_insert_fast(&c->vfs_inodes_table,
 					&inode->hash,
 					bch2_vfs_inodes_params))) {
-		old = bch2_inode_hash_find(c, inode->ei_inum);
+		old = bch2_inode_hash_find(c, trans, inode->ei_inum);
 		if (!old)
 			goto retry;
+		if (IS_ERR(old))
+			return old;
 
 		/*
 		 * bcachefs doesn't use I_NEW; we have no use for it since we
 		 * only insert fully created inodes in the inode hash table. But
 		 * discard_new_inode() expects it to be set...
 		 */
-		inode->v.i_flags |= I_NEW;
+		inode->v.i_state |= 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
@@ -312,9 +322,31 @@ static struct bch_inode_info *bch2_new_inode(struct btree_trans *trans)
 	return inode;
 }
 
+static struct bch_inode_info *bch2_inode_hash_init_insert(struct btree_trans *trans,
+							  subvol_inum inum,
+							  struct bch_inode_unpacked *bi,
+							  struct bch_subvolume *subvol)
+{
+	struct bch_inode_info *inode = bch2_new_inode(trans);
+	if (IS_ERR(inode))
+		return inode;
+
+	bch2_vfs_inode_init(trans, inum, inode, bi, subvol);
+
+	struct bch_inode_info *ret = bch2_inode_hash_insert(trans->c, trans, inode);
+	if (IS_ERR(ret)) {
+		inode->v.i_state |= I_NEW;
+		set_nlink(&inode->v, 1);
+		discard_new_inode(&inode->v);
+	}
+
+	return ret;
+
+}
+
 struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
 {
-	struct bch_inode_info *inode = bch2_inode_hash_find(c, inum);
+	struct bch_inode_info *inode = bch2_inode_hash_find(c, NULL, inum);
 	if (inode)
 		return &inode->v;
 
@@ -325,11 +357,7 @@ struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
 	int ret = lockrestart_do(trans,
 		bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
 		bch2_inode_find_by_inum_trans(trans, inum, &inode_u)) ?:
-		PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
+		PTR_ERR_OR_ZERO(inode = bch2_inode_hash_init_insert(trans, inum, &inode_u, &subvol));
-	if (!ret) {
-		bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
-		inode = bch2_inode_hash_insert(c, inode);
-	}
 	bch2_trans_put(trans);
 
 	return ret ? ERR_PTR(ret) : &inode->v;
@@ -420,8 +448,16 @@ err_before_quota:
 	 * we must insert the new inode into the inode cache before calling
 	 * bch2_trans_exit() and dropping locks, else we could race with another
 	 * thread pulling the inode in and modifying it:
+	 *
+	 * also, calling bch2_inode_hash_insert() without passing in the
+	 * transaction object is sketchy - if we could ever end up in
+	 * __wait_on_freeing_inode(), we'd risk deadlock.
+	 *
+	 * But that shouldn't be possible, since we still have the inode locked
+	 * that we just created, and we _really_ can't take a transaction
+	 * restart here.
 	 */
-	inode = bch2_inode_hash_insert(c, inode);
+	inode = bch2_inode_hash_insert(c, NULL, inode);
 	bch2_trans_put(trans);
 err:
 	posix_acl_release(default_acl);
@@ -461,7 +497,7 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	struct bch_inode_info *inode = bch2_inode_hash_find(c, inum);
+	struct bch_inode_info *inode = bch2_inode_hash_find(c, trans, inum);
 	if (inode)
 		goto out;
 
@@ -469,7 +505,7 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
 	struct bch_inode_unpacked inode_u;
 	ret =   bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
 		bch2_inode_find_by_inum_nowarn_trans(trans, inum, &inode_u) ?:
-		PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
+		PTR_ERR_OR_ZERO(inode = bch2_inode_hash_init_insert(trans, inum, &inode_u, &subvol));
 
 	bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT),
 				c, "dirent to missing inode:\n  %s",
@@ -489,9 +525,6 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
 		ret = -ENOENT;
 		goto err;
 	}
-
-	bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
-	inode = bch2_inode_hash_insert(c, inode);
 out:
 	bch2_trans_iter_exit(trans, &dirent_iter);
 	printbuf_exit(&buf);
@@ -1536,7 +1569,8 @@ static const struct export_operations bch_export_ops = {
 	.get_name	= bch2_get_name,
 };
 
-static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
+static void bch2_vfs_inode_init(struct btree_trans *trans,
+				subvol_inum inum,
 				struct bch_inode_info *inode,
 				struct bch_inode_unpacked *bi,
 				struct bch_subvolume *subvol)

libbcachefs/inode.c

@@ -365,7 +365,7 @@ int bch2_inode_peek(struct btree_trans *trans,
 		    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 %u:%llu:", inum.subvol, inum.inum);
+	bch_err_msg(trans->c, ret, "looking up inum %llu:%llu:", inum.subvol, inum.inum);
 	return ret;
 }
 
@@ -434,100 +434,98 @@ struct bkey_i *bch2_inode_to_v3(struct btree_trans *trans, struct bkey_i *k)
 	return &inode_p->inode.k_i;
 }
 
-static int __bch2_inode_invalid(struct bch_fs *c, struct bkey_s_c k, struct printbuf *err)
+static int __bch2_inode_validate(struct bch_fs *c, struct bkey_s_c k,
+				 enum bch_validate_flags flags)
 {
 	struct bch_inode_unpacked unpacked;
 	int ret = 0;
 
-	bkey_fsck_err_on(k.k->p.inode, c, err,
+	bkey_fsck_err_on(k.k->p.inode,
-			 inode_pos_inode_nonzero,
+			 c, inode_pos_inode_nonzero,
 			 "nonzero k.p.inode");
 
-	bkey_fsck_err_on(k.k->p.offset < BLOCKDEV_INODE_MAX, c, err,
+	bkey_fsck_err_on(k.k->p.offset < BLOCKDEV_INODE_MAX,
-			 inode_pos_blockdev_range,
+			 c, inode_pos_blockdev_range,
 			 "fs inode in blockdev range");
 
-	bkey_fsck_err_on(bch2_inode_unpack(k, &unpacked), c, err,
+	bkey_fsck_err_on(bch2_inode_unpack(k, &unpacked),
-			 inode_unpack_error,
+			 c, inode_unpack_error,
 			 "invalid variable length fields");
 
-	bkey_fsck_err_on(unpacked.bi_data_checksum >= BCH_CSUM_OPT_NR + 1, c, err,
+	bkey_fsck_err_on(unpacked.bi_data_checksum >= BCH_CSUM_OPT_NR + 1,
-			 inode_checksum_type_invalid,
+			 c, inode_checksum_type_invalid,
 			 "invalid data checksum type (%u >= %u",
 			 unpacked.bi_data_checksum, BCH_CSUM_OPT_NR + 1);
 
 	bkey_fsck_err_on(unpacked.bi_compression &&
-			 !bch2_compression_opt_valid(unpacked.bi_compression - 1), c, err,
+			 !bch2_compression_opt_valid(unpacked.bi_compression - 1),
-			 inode_compression_type_invalid,
+			 c, inode_compression_type_invalid,
 			 "invalid compression opt %u", unpacked.bi_compression - 1);
 
 	bkey_fsck_err_on((unpacked.bi_flags & BCH_INODE_unlinked) &&
-			 unpacked.bi_nlink != 0, c, err,
+			 unpacked.bi_nlink != 0,
-			 inode_unlinked_but_nlink_nonzero,
+			 c, inode_unlinked_but_nlink_nonzero,
 			 "flagged as unlinked but bi_nlink != 0");
 
-	bkey_fsck_err_on(unpacked.bi_subvol && !S_ISDIR(unpacked.bi_mode), c, err,
+	bkey_fsck_err_on(unpacked.bi_subvol && !S_ISDIR(unpacked.bi_mode),
-			 inode_subvol_root_but_not_dir,
+			 c, inode_subvol_root_but_not_dir,
 			 "subvolume root but not a directory");
 fsck_err:
 	return ret;
 }
 
-int bch2_inode_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_inode_validate(struct bch_fs *c, struct bkey_s_c k,
-		       enum bch_validate_flags flags,
+			enum bch_validate_flags flags)
-		       struct printbuf *err)
 {
 	struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR, c, err,
+	bkey_fsck_err_on(INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR,
-			 inode_str_hash_invalid,
+			 c, inode_str_hash_invalid,
 			 "invalid str hash type (%llu >= %u)",
 			 INODE_STR_HASH(inode.v), BCH_STR_HASH_NR);
 
-	ret = __bch2_inode_invalid(c, k, err);
+	ret = __bch2_inode_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
 
-int bch2_inode_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_inode_v2_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_s_c_inode_v2 inode = bkey_s_c_to_inode_v2(k);
 	int ret = 0;
 
-	bkey_fsck_err_on(INODEv2_STR_HASH(inode.v) >= BCH_STR_HASH_NR, c, err,
+	bkey_fsck_err_on(INODEv2_STR_HASH(inode.v) >= BCH_STR_HASH_NR,
-			 inode_str_hash_invalid,
+			 c, inode_str_hash_invalid,
 			 "invalid str hash type (%llu >= %u)",
 			 INODEv2_STR_HASH(inode.v), BCH_STR_HASH_NR);
 
-	ret = __bch2_inode_invalid(c, k, err);
+	ret = __bch2_inode_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
 
-int bch2_inode_v3_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_inode_v3_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			   enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_s_c_inode_v3 inode = bkey_s_c_to_inode_v3(k);
 	int ret = 0;
 
 	bkey_fsck_err_on(INODEv3_FIELDS_START(inode.v) < INODEv3_FIELDS_START_INITIAL ||
-			 INODEv3_FIELDS_START(inode.v) > bkey_val_u64s(inode.k), c, err,
+			 INODEv3_FIELDS_START(inode.v) > bkey_val_u64s(inode.k),
-			 inode_v3_fields_start_bad,
+			 c, inode_v3_fields_start_bad,
 			 "invalid fields_start (got %llu, min %u max %zu)",
 			 INODEv3_FIELDS_START(inode.v),
 			 INODEv3_FIELDS_START_INITIAL,
 			 bkey_val_u64s(inode.k));
 
-	bkey_fsck_err_on(INODEv3_STR_HASH(inode.v) >= BCH_STR_HASH_NR, c, err,
+	bkey_fsck_err_on(INODEv3_STR_HASH(inode.v) >= BCH_STR_HASH_NR,
-			 inode_str_hash_invalid,
+			 c, inode_str_hash_invalid,
 			 "invalid str hash type (%llu >= %u)",
 			 INODEv3_STR_HASH(inode.v), BCH_STR_HASH_NR);
 
-	ret = __bch2_inode_invalid(c, k, err);
+	ret = __bch2_inode_validate(c, k, flags);
 fsck_err:
 	return ret;
 }
@@ -625,14 +623,13 @@ int bch2_trigger_inode(struct btree_trans *trans,
 	return 0;
 }
 
-int bch2_inode_generation_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_inode_generation_validate(struct bch_fs *c, struct bkey_s_c k,
-				  enum bch_validate_flags flags,
+				   enum bch_validate_flags flags)
-				  struct printbuf *err)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(k.k->p.inode, c, err,
+	bkey_fsck_err_on(k.k->p.inode,
-			 inode_pos_inode_nonzero,
+			 c, inode_pos_inode_nonzero,
 			 "nonzero k.p.inode");
 fsck_err:
 	return ret;

libbcachefs/inode.h

@@ -9,12 +9,12 @@
 enum bch_validate_flags;
 extern const char * const bch2_inode_opts[];
 
-int bch2_inode_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_inode_validate(struct bch_fs *, struct bkey_s_c,
-		       enum bch_validate_flags, struct printbuf *);
+		       enum bch_validate_flags);
-int bch2_inode_v2_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_inode_v2_validate(struct bch_fs *, struct bkey_s_c,
-			  enum bch_validate_flags, struct printbuf *);
+			  enum bch_validate_flags);
-int bch2_inode_v3_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_inode_v3_validate(struct bch_fs *, struct bkey_s_c,
-			  enum bch_validate_flags, struct printbuf *);
+			  enum bch_validate_flags);
 void bch2_inode_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 int bch2_trigger_inode(struct btree_trans *, enum btree_id, unsigned,
@@ -22,21 +22,21 @@ int bch2_trigger_inode(struct btree_trans *, enum btree_id, unsigned,
 		       enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_inode ((struct bkey_ops) {	\
-	.key_invalid	= bch2_inode_invalid,		\
+	.key_validate	= bch2_inode_validate,		\
 	.val_to_text	= bch2_inode_to_text,		\
 	.trigger	= bch2_trigger_inode,		\
 	.min_val_size	= 16,				\
 })
 
 #define bch2_bkey_ops_inode_v2 ((struct bkey_ops) {	\
-	.key_invalid	= bch2_inode_v2_invalid,	\
+	.key_validate	= bch2_inode_v2_validate,	\
 	.val_to_text	= bch2_inode_to_text,		\
 	.trigger	= bch2_trigger_inode,		\
 	.min_val_size	= 32,				\
 })
 
 #define bch2_bkey_ops_inode_v3 ((struct bkey_ops) {	\
-	.key_invalid	= bch2_inode_v3_invalid,	\
+	.key_validate	= bch2_inode_v3_validate,	\
 	.val_to_text	= bch2_inode_to_text,		\
 	.trigger	= bch2_trigger_inode,		\
 	.min_val_size	= 48,				\
@@ -49,12 +49,12 @@ static inline bool bkey_is_inode(const struct bkey *k)
 		k->type == KEY_TYPE_inode_v3;
 }
 
-int bch2_inode_generation_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_inode_generation_validate(struct bch_fs *, struct bkey_s_c,
-				  enum bch_validate_flags, struct printbuf *);
+				  enum bch_validate_flags);
 void bch2_inode_generation_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_inode_generation ((struct bkey_ops) {	\
-	.key_invalid	= bch2_inode_generation_invalid,	\
+	.key_validate	= bch2_inode_generation_validate,	\
 	.val_to_text	= bch2_inode_generation_to_text,	\
 	.min_val_size	= 8,					\
 })

libbcachefs/journal_io.c

@@ -332,7 +332,6 @@ static int journal_validate_key(struct bch_fs *c,
 {
 	int write = flags & BCH_VALIDATE_write;
 	void *next = vstruct_next(entry);
-	struct printbuf buf = PRINTBUF;
 	int ret = 0;
 
 	if (journal_entry_err_on(!k->k.u64s,
@@ -368,34 +367,21 @@ static int journal_validate_key(struct bch_fs *c,
 		bch2_bkey_compat(level, btree_id, version, big_endian,
 				 write, NULL, bkey_to_packed(k));
 
-	if (bch2_bkey_invalid(c, bkey_i_to_s_c(k),
+	ret = bch2_bkey_validate(c, bkey_i_to_s_c(k),
-			      __btree_node_type(level, btree_id), write, &buf)) {
+				 __btree_node_type(level, btree_id), write);
-		printbuf_reset(&buf);
+	if (ret == -BCH_ERR_fsck_delete_bkey) {
-		journal_entry_err_msg(&buf, version, jset, entry);
-		prt_newline(&buf);
-		printbuf_indent_add(&buf, 2);
-
-		bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(k));
-		prt_newline(&buf);
-		bch2_bkey_invalid(c, bkey_i_to_s_c(k),
-				  __btree_node_type(level, btree_id), write, &buf);
-
-		mustfix_fsck_err(c, journal_entry_bkey_invalid,
-				 "%s", buf.buf);
-
 		le16_add_cpu(&entry->u64s, -((u16) k->k.u64s));
 		memmove(k, bkey_next(k), next - (void *) bkey_next(k));
 		journal_entry_null_range(vstruct_next(entry), next);
-
-		printbuf_exit(&buf);
 		return FSCK_DELETED_KEY;
 	}
+	if (ret)
+		goto fsck_err;
 
 	if (write)
 		bch2_bkey_compat(level, btree_id, version, big_endian,
 				 write, NULL, bkey_to_packed(k));
 fsck_err:
-	printbuf_exit(&buf);
 	return ret;
 }
 

libbcachefs/lru.c

@@ -10,14 +10,13 @@
 #include "recovery.h"
 
 /* KEY_TYPE_lru is obsolete: */
-int bch2_lru_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_lru_validate(struct bch_fs *c, struct bkey_s_c k,
-		     enum bch_validate_flags flags,
+		     enum bch_validate_flags flags)
-		     struct printbuf *err)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(!lru_pos_time(k.k->p), c, err,
+	bkey_fsck_err_on(!lru_pos_time(k.k->p),
-			 lru_entry_at_time_0,
+			 c, lru_entry_at_time_0,
 			 "lru entry at time=0");
 fsck_err:
 	return ret;

libbcachefs/lru.h

@@ -33,14 +33,13 @@ static inline enum bch_lru_type lru_type(struct bkey_s_c l)
 	return BCH_LRU_read;
 }
 
-int bch2_lru_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_lru_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-		     enum bch_validate_flags, struct printbuf *);
 void bch2_lru_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 void bch2_lru_pos_to_text(struct printbuf *, struct bpos);
 
 #define bch2_bkey_ops_lru ((struct bkey_ops) {	\
-	.key_invalid	= bch2_lru_invalid,	\
+	.key_validate	= bch2_lru_validate,	\
 	.val_to_text	= bch2_lru_to_text,	\
 	.min_val_size	= 8,			\
 })

libbcachefs/quota.c

@@ -59,13 +59,13 @@ const struct bch_sb_field_ops bch_sb_field_ops_quota = {
 	.to_text	= bch2_sb_quota_to_text,
 };
 
-int bch2_quota_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_quota_validate(struct bch_fs *c, struct bkey_s_c k,
-		       enum bch_validate_flags flags, struct printbuf *err)
+			enum bch_validate_flags flags)
 {
 	int ret = 0;
 
-	bkey_fsck_err_on(k.k->p.inode >= QTYP_NR, c, err,
+	bkey_fsck_err_on(k.k->p.inode >= QTYP_NR,
-			 quota_type_invalid,
+			 c, quota_type_invalid,
 			 "invalid quota type (%llu >= %u)",
 			 k.k->p.inode, QTYP_NR);
 fsck_err:

libbcachefs/quota.h

@@ -8,12 +8,11 @@
 enum bch_validate_flags;
 extern const struct bch_sb_field_ops bch_sb_field_ops_quota;
 
-int bch2_quota_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_quota_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-		       enum bch_validate_flags, struct printbuf *);
 void bch2_quota_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_quota ((struct bkey_ops) {	\
-	.key_invalid	= bch2_quota_invalid,		\
+	.key_validate	= bch2_quota_validate,		\
 	.val_to_text	= bch2_quota_to_text,		\
 	.min_val_size	= 32,				\
 })

libbcachefs/reflink.c

@@ -29,15 +29,14 @@ static inline unsigned bkey_type_to_indirect(const struct bkey *k)
 
 /* reflink pointers */
 
-int bch2_reflink_p_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_reflink_p_validate(struct bch_fs *c, struct bkey_s_c k,
-			   enum bch_validate_flags flags,
+			    enum bch_validate_flags flags)
-			   struct printbuf *err)
 {
 	struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
 	int ret = 0;
 
 	bkey_fsck_err_on(le64_to_cpu(p.v->idx) < le32_to_cpu(p.v->front_pad),
-			 c, err, reflink_p_front_pad_bad,
+			 c, reflink_p_front_pad_bad,
 			 "idx < front_pad (%llu < %u)",
 			 le64_to_cpu(p.v->idx), le32_to_cpu(p.v->front_pad));
 fsck_err:
@@ -256,11 +255,10 @@ int bch2_trigger_reflink_p(struct btree_trans *trans,
 
 /* indirect extents */
 
-int bch2_reflink_v_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_reflink_v_validate(struct bch_fs *c, struct bkey_s_c k,
-			   enum bch_validate_flags flags,
+			    enum bch_validate_flags flags)
-			   struct printbuf *err)
 {
-	return bch2_bkey_ptrs_invalid(c, k, flags, err);
+	return bch2_bkey_ptrs_validate(c, k, flags);
 }
 
 void bch2_reflink_v_to_text(struct printbuf *out, struct bch_fs *c,
@@ -311,9 +309,8 @@ int bch2_trigger_reflink_v(struct btree_trans *trans,
 
 /* indirect inline data */
 
-int bch2_indirect_inline_data_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_indirect_inline_data_validate(struct bch_fs *c, struct bkey_s_c k,
-				      enum bch_validate_flags flags,
+				      enum bch_validate_flags flags)
-				      struct printbuf *err)
 {
 	return 0;
 }

libbcachefs/reflink.h

@@ -4,41 +4,37 @@
 
 enum bch_validate_flags;
 
-int bch2_reflink_p_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_reflink_p_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			   enum bch_validate_flags, struct printbuf *);
+void bch2_reflink_p_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
-void bch2_reflink_p_to_text(struct printbuf *, struct bch_fs *,
-			    struct bkey_s_c);
 bool bch2_reflink_p_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
 int bch2_trigger_reflink_p(struct btree_trans *, enum btree_id, unsigned,
 			   struct bkey_s_c, struct bkey_s,
 			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_reflink_p ((struct bkey_ops) {		\
-	.key_invalid	= bch2_reflink_p_invalid,		\
+	.key_validate	= bch2_reflink_p_validate,		\
 	.val_to_text	= bch2_reflink_p_to_text,		\
 	.key_merge	= bch2_reflink_p_merge,			\
 	.trigger	= bch2_trigger_reflink_p,		\
 	.min_val_size	= 16,					\
 })
 
-int bch2_reflink_v_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_reflink_v_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			   enum bch_validate_flags, struct printbuf *);
+void bch2_reflink_v_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
-void bch2_reflink_v_to_text(struct printbuf *, struct bch_fs *,
-			    struct bkey_s_c);
 int bch2_trigger_reflink_v(struct btree_trans *, enum btree_id, unsigned,
 			   struct bkey_s_c, struct bkey_s,
 			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_reflink_v ((struct bkey_ops) {		\
-	.key_invalid	= bch2_reflink_v_invalid,		\
+	.key_validate	= bch2_reflink_v_validate,		\
 	.val_to_text	= bch2_reflink_v_to_text,		\
 	.swab		= bch2_ptr_swab,			\
 	.trigger	= bch2_trigger_reflink_v,		\
 	.min_val_size	= 8,					\
 })
 
-int bch2_indirect_inline_data_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_indirect_inline_data_validate(struct bch_fs *, struct bkey_s_c,
-				      enum bch_validate_flags, struct printbuf *);
+				      enum bch_validate_flags);
 void bch2_indirect_inline_data_to_text(struct printbuf *,
 				struct bch_fs *, struct bkey_s_c);
 int bch2_trigger_indirect_inline_data(struct btree_trans *,
@@ -47,7 +43,7 @@ int bch2_trigger_indirect_inline_data(struct btree_trans *,
 			      enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_indirect_inline_data ((struct bkey_ops) {	\
-	.key_invalid	= bch2_indirect_inline_data_invalid,	\
+	.key_validate	= bch2_indirect_inline_data_validate,	\
 	.val_to_text	= bch2_indirect_inline_data_to_text,	\
 	.trigger	= bch2_trigger_indirect_inline_data,	\
 	.min_val_size	= 8,					\

libbcachefs/sb-counters_format.h

@@ -81,7 +81,8 @@
 	x(trans_restart_write_buffer_flush,		75)	\
 	x(trans_restart_split_race,			76)	\
 	x(write_buffer_flush_slowpath,			77)	\
-	x(write_buffer_flush_sync,			78)
+	x(write_buffer_flush_sync,			78)	\
+	x(trans_restart_freeing_inode,			79)
 
 enum bch_persistent_counters {
 #define x(t, n, ...) BCH_COUNTER_##t,

libbcachefs/sb-downgrade.c

@@ -72,7 +72,10 @@
 	  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)
+	  BCH_FSCK_ERR_accounting_key_junk_at_end)		\
+	x(disk_accounting_inum,					\
+	  BIT_ULL(BCH_RECOVERY_PASS_check_allocations),		\
+	  BCH_FSCK_ERR_accounting_mismatch)
 
 #define DOWNGRADE_TABLE()					\
 	x(bucket_stripe_sectors,				\
@@ -104,6 +107,7 @@
 	  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_accounting_replicas_not_marked,		\
 	  BCH_FSCK_ERR_bkey_version_in_future)
 
 struct upgrade_downgrade_entry {

libbcachefs/snapshot.c

@@ -31,15 +31,14 @@ void bch2_snapshot_tree_to_text(struct printbuf *out, struct bch_fs *c,
 		   le32_to_cpu(t.v->root_snapshot));
 }
 
-int bch2_snapshot_tree_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_snapshot_tree_validate(struct bch_fs *c, struct bkey_s_c k,
-			       enum bch_validate_flags flags,
+			       enum bch_validate_flags flags)
-			       struct printbuf *err)
 {
 	int ret = 0;
 
 	bkey_fsck_err_on(bkey_gt(k.k->p, POS(0, U32_MAX)) ||
-			 bkey_lt(k.k->p, POS(0, 1)), c, err,
+			 bkey_lt(k.k->p, POS(0, 1)),
-			 snapshot_tree_pos_bad,
+			 c, snapshot_tree_pos_bad,
 			 "bad pos");
 fsck_err:
 	return ret;
@@ -225,55 +224,54 @@ void bch2_snapshot_to_text(struct printbuf *out, struct bch_fs *c,
 			   le32_to_cpu(s.v->skip[2]));
 }
 
-int bch2_snapshot_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_snapshot_validate(struct bch_fs *c, struct bkey_s_c k,
-			  enum bch_validate_flags flags,
+			  enum bch_validate_flags flags)
-			  struct printbuf *err)
 {
 	struct bkey_s_c_snapshot s;
 	u32 i, id;
 	int ret = 0;
 
 	bkey_fsck_err_on(bkey_gt(k.k->p, POS(0, U32_MAX)) ||
-			 bkey_lt(k.k->p, POS(0, 1)), c, err,
+			 bkey_lt(k.k->p, POS(0, 1)),
-			 snapshot_pos_bad,
+			 c, snapshot_pos_bad,
 			 "bad pos");
 
 	s = bkey_s_c_to_snapshot(k);
 
 	id = le32_to_cpu(s.v->parent);
-	bkey_fsck_err_on(id && id <= k.k->p.offset, c, err,
+	bkey_fsck_err_on(id && id <= k.k->p.offset,
-			 snapshot_parent_bad,
+			 c, snapshot_parent_bad,
 			 "bad parent node (%u <= %llu)",
 			 id, k.k->p.offset);
 
-	bkey_fsck_err_on(le32_to_cpu(s.v->children[0]) < le32_to_cpu(s.v->children[1]), c, err,
+	bkey_fsck_err_on(le32_to_cpu(s.v->children[0]) < le32_to_cpu(s.v->children[1]),
-			 snapshot_children_not_normalized,
+			 c, snapshot_children_not_normalized,
 			 "children not normalized");
 
-	bkey_fsck_err_on(s.v->children[0] && s.v->children[0] == s.v->children[1], c, err,
+	bkey_fsck_err_on(s.v->children[0] && s.v->children[0] == s.v->children[1],
-			 snapshot_child_duplicate,
+			 c, snapshot_child_duplicate,
 			 "duplicate child nodes");
 
 	for (i = 0; i < 2; i++) {
 		id = le32_to_cpu(s.v->children[i]);
 
-		bkey_fsck_err_on(id >= k.k->p.offset, c, err,
+		bkey_fsck_err_on(id >= k.k->p.offset,
-				 snapshot_child_bad,
+				 c, snapshot_child_bad,
 				 "bad child node (%u >= %llu)",
 				 id, k.k->p.offset);
 	}
 
 	if (bkey_val_bytes(k.k) > offsetof(struct bch_snapshot, skip)) {
 		bkey_fsck_err_on(le32_to_cpu(s.v->skip[0]) > le32_to_cpu(s.v->skip[1]) ||
-				 le32_to_cpu(s.v->skip[1]) > le32_to_cpu(s.v->skip[2]), c, err,
+				 le32_to_cpu(s.v->skip[1]) > le32_to_cpu(s.v->skip[2]),
-				 snapshot_skiplist_not_normalized,
+				 c, snapshot_skiplist_not_normalized,
 				 "skiplist not normalized");
 
 		for (i = 0; i < ARRAY_SIZE(s.v->skip); i++) {
 			id = le32_to_cpu(s.v->skip[i]);
 
-			bkey_fsck_err_on(id && id < le32_to_cpu(s.v->parent), c, err,
+			bkey_fsck_err_on(id && id < le32_to_cpu(s.v->parent),
-					 snapshot_skiplist_bad,
+					 c, snapshot_skiplist_bad,
 					 "bad skiplist node %u", id);
 		}
 	}

libbcachefs/snapshot.h

@@ -5,11 +5,11 @@
 enum bch_validate_flags;
 
 void bch2_snapshot_tree_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
-int bch2_snapshot_tree_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_snapshot_tree_validate(struct bch_fs *, struct bkey_s_c,
-			       enum bch_validate_flags, struct printbuf *);
+			       enum bch_validate_flags);
 
 #define bch2_bkey_ops_snapshot_tree ((struct bkey_ops) {	\
-	.key_invalid	= bch2_snapshot_tree_invalid,		\
+	.key_validate	= bch2_snapshot_tree_validate,		\
 	.val_to_text	= bch2_snapshot_tree_to_text,		\
 	.min_val_size	= 8,					\
 })
@@ -19,14 +19,13 @@ struct bkey_i_snapshot_tree *__bch2_snapshot_tree_create(struct btree_trans *);
 int bch2_snapshot_tree_lookup(struct btree_trans *, u32, struct bch_snapshot_tree *);
 
 void bch2_snapshot_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
-int bch2_snapshot_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_snapshot_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			  enum bch_validate_flags, struct printbuf *);
 int bch2_mark_snapshot(struct btree_trans *, enum btree_id, unsigned,
 		       struct bkey_s_c, struct bkey_s,
 		       enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_snapshot ((struct bkey_ops) {		\
-	.key_invalid	= bch2_snapshot_invalid,		\
+	.key_validate	= bch2_snapshot_validate,		\
 	.val_to_text	= bch2_snapshot_to_text,		\
 	.trigger	= bch2_mark_snapshot,			\
 	.min_val_size	= 24,					\

libbcachefs/subvolume.c

@@ -207,23 +207,23 @@ int bch2_check_subvol_children(struct bch_fs *c)
 
 /* Subvolumes: */
 
-int bch2_subvolume_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_subvolume_validate(struct bch_fs *c, struct bkey_s_c k,
-			   enum bch_validate_flags flags, struct printbuf *err)
+			   enum bch_validate_flags flags)
 {
 	struct bkey_s_c_subvolume subvol = bkey_s_c_to_subvolume(k);
 	int ret = 0;
 
 	bkey_fsck_err_on(bkey_lt(k.k->p, SUBVOL_POS_MIN) ||
-			 bkey_gt(k.k->p, SUBVOL_POS_MAX), c, err,
+			 bkey_gt(k.k->p, SUBVOL_POS_MAX),
-			 subvol_pos_bad,
+			 c, subvol_pos_bad,
 			 "invalid pos");
 
-	bkey_fsck_err_on(!subvol.v->snapshot, c, err,
+	bkey_fsck_err_on(!subvol.v->snapshot,
-			 subvol_snapshot_bad,
+			 c, subvol_snapshot_bad,
 			 "invalid snapshot");
 
-	bkey_fsck_err_on(!subvol.v->inode, c, err,
+	bkey_fsck_err_on(!subvol.v->inode,
-			 subvol_inode_bad,
+			 c, subvol_inode_bad,
 			 "invalid inode");
 fsck_err:
 	return ret;

libbcachefs/subvolume.h

@@ -10,15 +10,14 @@ enum bch_validate_flags;
 int bch2_check_subvols(struct bch_fs *);
 int bch2_check_subvol_children(struct bch_fs *);
 
-int bch2_subvolume_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_subvolume_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-			   enum bch_validate_flags, struct printbuf *);
 void bch2_subvolume_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 int bch2_subvolume_trigger(struct btree_trans *, enum btree_id, unsigned,
 			   struct bkey_s_c, struct bkey_s,
 			   enum btree_iter_update_trigger_flags);
 
 #define bch2_bkey_ops_subvolume ((struct bkey_ops) {		\
-	.key_invalid	= bch2_subvolume_invalid,		\
+	.key_validate	= bch2_subvolume_validate,		\
 	.val_to_text	= bch2_subvolume_to_text,		\
 	.trigger	= bch2_subvolume_trigger,		\
 	.min_val_size	= 16,					\

libbcachefs/subvolume_types.h

@@ -30,7 +30,8 @@ struct snapshot_table {
 };
 
 typedef struct {
-	u32		subvol;
+	/* we can't have padding in this struct: */
+	u64		subvol;
 	u64		inum;
 } subvol_inum;
 

libbcachefs/trace.c

@@ -4,6 +4,7 @@
 #include "buckets.h"
 #include "btree_cache.h"
 #include "btree_iter.h"
+#include "btree_key_cache.h"
 #include "btree_locking.h"
 #include "btree_update_interior.h"
 #include "keylist.h"

libbcachefs/trace.h

@@ -993,10 +993,33 @@ TRACE_EVENT(trans_restart_split_race,
 		  __entry->u64s_remaining)
 );
 
-DEFINE_EVENT(transaction_event,	trans_blocked_journal_reclaim,
+TRACE_EVENT(trans_blocked_journal_reclaim,
 	TP_PROTO(struct btree_trans *trans,
 		 unsigned long caller_ip),
-	TP_ARGS(trans, caller_ip)
+	TP_ARGS(trans, caller_ip),
+
+	TP_STRUCT__entry(
+		__array(char,			trans_fn, 32	)
+		__field(unsigned long,		caller_ip	)
+
+		__field(unsigned long,		key_cache_nr_keys	)
+		__field(unsigned long,		key_cache_nr_dirty	)
+		__field(long,			must_wait		)
+	),
+
+	TP_fast_assign(
+		strscpy(__entry->trans_fn, trans->fn, sizeof(__entry->trans_fn));
+		__entry->caller_ip		= caller_ip;
+		__entry->key_cache_nr_keys	= atomic_long_read(&trans->c->btree_key_cache.nr_keys);
+		__entry->key_cache_nr_dirty	= atomic_long_read(&trans->c->btree_key_cache.nr_dirty);
+		__entry->must_wait		= __bch2_btree_key_cache_must_wait(trans->c);
+	),
+
+	TP_printk("%s %pS key cache keys %lu dirty %lu must_wait %li",
+		  __entry->trans_fn, (void *) __entry->caller_ip,
+		  __entry->key_cache_nr_keys,
+		  __entry->key_cache_nr_dirty,
+		  __entry->must_wait)
 );
 
 TRACE_EVENT(trans_restart_journal_preres_get,
@@ -1293,6 +1316,12 @@ TRACE_EVENT(trans_restart_key_cache_key_realloced,
 		  __entry->new_u64s)
 );
 
+DEFINE_EVENT(transaction_event,	trans_restart_freeing_inode,
+	TP_PROTO(struct btree_trans *trans,
+		 unsigned long caller_ip),
+	TP_ARGS(trans, caller_ip)
+);
+
 TRACE_EVENT(path_downgrade,
 	TP_PROTO(struct btree_trans *trans,
 		 unsigned long caller_ip,

libbcachefs/xattr.c

@@ -70,17 +70,16 @@ const struct bch_hash_desc bch2_xattr_hash_desc = {
 	.cmp_bkey	= xattr_cmp_bkey,
 };
 
-int bch2_xattr_invalid(struct bch_fs *c, struct bkey_s_c k,
+int bch2_xattr_validate(struct bch_fs *c, struct bkey_s_c k,
-		       enum bch_validate_flags flags,
+		       enum bch_validate_flags flags)
-		       struct printbuf *err)
 {
 	struct bkey_s_c_xattr xattr = bkey_s_c_to_xattr(k);
 	unsigned val_u64s = xattr_val_u64s(xattr.v->x_name_len,
 					   le16_to_cpu(xattr.v->x_val_len));
 	int ret = 0;
 
-	bkey_fsck_err_on(bkey_val_u64s(k.k) < val_u64s, c, err,
+	bkey_fsck_err_on(bkey_val_u64s(k.k) < val_u64s,
-			 xattr_val_size_too_small,
+			 c, xattr_val_size_too_small,
 			 "value too small (%zu < %u)",
 			 bkey_val_u64s(k.k), val_u64s);
 
@@ -88,17 +87,17 @@ int bch2_xattr_invalid(struct bch_fs *c, struct bkey_s_c k,
 	val_u64s = xattr_val_u64s(xattr.v->x_name_len,
 				  le16_to_cpu(xattr.v->x_val_len) + 4);
 
-	bkey_fsck_err_on(bkey_val_u64s(k.k) > val_u64s, c, err,
+	bkey_fsck_err_on(bkey_val_u64s(k.k) > val_u64s,
-			 xattr_val_size_too_big,
+			 c, xattr_val_size_too_big,
 			 "value too big (%zu > %u)",
 			 bkey_val_u64s(k.k), val_u64s);
 
-	bkey_fsck_err_on(!bch2_xattr_type_to_handler(xattr.v->x_type), c, err,
+	bkey_fsck_err_on(!bch2_xattr_type_to_handler(xattr.v->x_type),
-			 xattr_invalid_type,
+			 c, xattr_invalid_type,
 			 "invalid type (%u)", xattr.v->x_type);
 
-	bkey_fsck_err_on(memchr(xattr.v->x_name, '\0', xattr.v->x_name_len), c, err,
+	bkey_fsck_err_on(memchr(xattr.v->x_name, '\0', xattr.v->x_name_len),
-			 xattr_name_invalid_chars,
+			 c, xattr_name_invalid_chars,
 			 "xattr name has invalid characters");
 fsck_err:
 	return ret;
@@ -251,17 +250,27 @@ static int __bch2_xattr_emit(const char *prefix,
 	return 0;
 }
 
+static inline const char *bch2_xattr_prefix(unsigned type, struct dentry *dentry)
+{
+	const struct xattr_handler *handler = bch2_xattr_type_to_handler(type);
+
+	if (!xattr_handler_can_list(handler, dentry))
+		return NULL;
+
+	return xattr_prefix(handler);
+}
+
 static int bch2_xattr_emit(struct dentry *dentry,
 			    const struct bch_xattr *xattr,
 			    struct xattr_buf *buf)
 {
-	const struct xattr_handler *handler =
+	const char *prefix;
-		bch2_xattr_type_to_handler(xattr->x_type);
 
-	return handler && (!handler->list || handler->list(dentry))
+	prefix = bch2_xattr_prefix(xattr->x_type, dentry);
-		? __bch2_xattr_emit(handler->prefix ?: handler->name,
+	if (!prefix)
-				    xattr->x_name, xattr->x_name_len, buf)
+		return 0;
-		: 0;
+
+	return __bch2_xattr_emit(prefix, xattr->x_name, xattr->x_name_len, buf);
 }
 
 static int bch2_xattr_list_bcachefs(struct bch_fs *c,
@@ -592,10 +601,6 @@ static const struct xattr_handler bch_xattr_bcachefs_effective_handler = {
 
 const struct xattr_handler *bch2_xattr_handlers[] = {
 	&bch_xattr_user_handler,
-#ifdef CONFIG_BCACHEFS_POSIX_ACL
-	&nop_posix_acl_access,
-	&nop_posix_acl_default,
-#endif
 	&bch_xattr_trusted_handler,
 	&bch_xattr_security_handler,
 #ifndef NO_BCACHEFS_FS

libbcachefs/xattr.h

@@ -6,12 +6,11 @@
 
 extern const struct bch_hash_desc bch2_xattr_hash_desc;
 
-int bch2_xattr_invalid(struct bch_fs *, struct bkey_s_c,
+int bch2_xattr_validate(struct bch_fs *, struct bkey_s_c, enum bch_validate_flags);
-		       enum bch_validate_flags, struct printbuf *);
 void bch2_xattr_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
 
 #define bch2_bkey_ops_xattr ((struct bkey_ops) {	\
-	.key_invalid	= bch2_xattr_invalid,		\
+	.key_validate	= bch2_xattr_validate,		\
 	.val_to_text	= bch2_xattr_to_text,		\
 	.min_val_size	= 8,				\
 })

linux/generic-radix-tree.c

@@ -121,6 +121,8 @@ void *__genradix_ptr_alloc(struct __genradix *radix, size_t offset,
 		if ((v = cmpxchg_release(&radix->root, r, new_root)) == r) {
 			v = new_root;
 			new_node = NULL;
+		} else {
+			new_node->children[0] = NULL;
 		}
 	}