Merge tag 'bcachefs-2024-06-22' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:
 "Lots of (mostly boring) fixes for syzbot bugs and rare(r) CI bugs.

  The LRU_TIME_BITS fix was slightly more involved; we only have 48 bits
  for the LRU position (we would prefer 64), so wraparound is possible
  for the cached data LRUs on a filesystem that has done sufficient
  (petabytes) reads; this is now handled.

  One notable user reported bugfix, where we were forgetting to
  correctly set the bucket data type, which should have been
  BCH_DATA_need_gc_gens instead of BCH_DATA_free; this was causing us to
  go emergency read-only on a filesystem that had seen heavy enough use
  to see bucket gen wraparoud.

  We're now starting to fix simple (safe) errors without requiring user
  intervention - i.e. a small incremental step towards full self
  healing.

  This is currently limited to just certain allocation information
  counters, and the error is still logged in the superblock; see that
  patch for more information. ("bcachefs: Fix safe errors by default")"

* tag 'bcachefs-2024-06-22' of https://evilpiepirate.org/git/bcachefs: (22 commits)
  bcachefs: Move the ei_flags setting to after initialization
  bcachefs: Fix a UAF after write_super()
  bcachefs: Use bch2_print_string_as_lines for long err
  bcachefs: Fix I_NEW warning in race path in bch2_inode_insert()
  bcachefs: Replace bare EEXIST with private error codes
  bcachefs: Fix missing alloc_data_type_set()
  closures: Change BUG_ON() to WARN_ON()
  bcachefs: fix alignment of VMA for memory mapped files on THP
  bcachefs: Fix safe errors by default
  bcachefs: Fix bch2_trans_put()
  bcachefs: set_worker_desc() for delete_dead_snapshots
  bcachefs: Fix bch2_sb_downgrade_update()
  bcachefs: Handle cached data LRU wraparound
  bcachefs: Guard against overflowing LRU_TIME_BITS
  bcachefs: delete_dead_snapshots() doesn't need to go RW
  bcachefs: Fix early init error path in journal code
  bcachefs: Check for invalid btree IDs
  bcachefs: Fix btree ID bitmasks
  bcachefs: Fix shift overflow in read_one_super()
  bcachefs: Fix a locking bug in the do_discard_fast() path
  ...

Author: torvalds

fs/bcachefs/alloc_background.c

@@ -259,6 +259,14 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
 			 "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,
+				 alloc_key_io_time_bad,
+				 "invalid io_time[%s]: %llu, max %llu",
+				 i == READ ? "read" : "write",
+				 a.v->io_time[i], LRU_TIME_MAX);
+
 	switch (a.v->data_type) {
 	case BCH_DATA_free:
 	case BCH_DATA_need_gc_gens:
@@ -757,8 +765,8 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 		alloc_data_type_set(new_a, new_a->data_type);
 
 		if (bch2_bucket_sectors_total(*new_a) > bch2_bucket_sectors_total(*old_a)) {
-			new_a->io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
-			new_a->io_time[WRITE]= max_t(u64, 1, atomic64_read(&c->io_clock[WRITE].now));
+			new_a->io_time[READ] = bch2_current_io_time(c, READ);
+			new_a->io_time[WRITE]= bch2_current_io_time(c, WRITE);
 			SET_BCH_ALLOC_V4_NEED_INC_GEN(new_a, true);
 			SET_BCH_ALLOC_V4_NEED_DISCARD(new_a, true);
 		}
@@ -768,6 +776,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 		    !bch2_bucket_is_open_safe(c, new.k->p.inode, new.k->p.offset)) {
 			new_a->gen++;
 			SET_BCH_ALLOC_V4_NEED_INC_GEN(new_a, false);
+			alloc_data_type_set(new_a, new_a->data_type);
 		}
 
 		if (old_a->data_type != new_a->data_type ||
@@ -781,7 +790,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 
 		if (new_a->data_type == BCH_DATA_cached &&
 		    !new_a->io_time[READ])
-			new_a->io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
+			new_a->io_time[READ] = bch2_current_io_time(c, READ);
 
 		u64 old_lru = alloc_lru_idx_read(*old_a);
 		u64 new_lru = alloc_lru_idx_read(*new_a);
@@ -882,7 +891,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
 			closure_wake_up(&c->freelist_wait);
 
 		if (statechange(a->data_type == BCH_DATA_need_discard) &&
-		    !bch2_bucket_is_open(c, new.k->p.inode, new.k->p.offset) &&
+		    !bch2_bucket_is_open_safe(c, new.k->p.inode, new.k->p.offset) &&
 		    bucket_flushed(new_a))
 			bch2_discard_one_bucket_fast(c, new.k->p);
 
@@ -1579,7 +1588,7 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
 		if (ret)
 			goto err;
 
-		a_mut->v.io_time[READ] = atomic64_read(&c->io_clock[READ].now);
+		a_mut->v.io_time[READ] = bch2_current_io_time(c, READ);
 		ret = bch2_trans_update(trans, alloc_iter,
 					&a_mut->k_i, BTREE_TRIGGER_norun);
 		if (ret)
@@ -1634,7 +1643,7 @@ static int discard_in_flight_add(struct bch_fs *c, struct bpos bucket)
 	mutex_lock(&c->discard_buckets_in_flight_lock);
 	darray_for_each(c->discard_buckets_in_flight, i)
 		if (bkey_eq(*i, bucket)) {
-			ret = -EEXIST;
+			ret = -BCH_ERR_EEXIST_discard_in_flight_add;
 			goto out;
 		}
 
@@ -1788,8 +1797,9 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
 	}
 
 	SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
-	alloc_data_type_set(&a->v, a->v.data_type);
 write:
+	alloc_data_type_set(&a->v, a->v.data_type);
+
 	ret =   bch2_trans_update(trans, &iter, &a->k_i, 0) ?:
 		bch2_trans_commit(trans, NULL, NULL,
 				  BCH_WATERMARK_btree|
@@ -1975,8 +1985,8 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 	a->v.data_type		= 0;
 	a->v.dirty_sectors	= 0;
 	a->v.cached_sectors	= 0;
-	a->v.io_time[READ]	= atomic64_read(&c->io_clock[READ].now);
-	a->v.io_time[WRITE]	= atomic64_read(&c->io_clock[WRITE].now);
+	a->v.io_time[READ]	= bch2_current_io_time(c, READ);
+	a->v.io_time[WRITE]	= bch2_current_io_time(c, WRITE);
 
 	ret = bch2_trans_commit(trans, NULL, NULL,
 				BCH_WATERMARK_btree|
@@ -2011,6 +2021,21 @@ static int invalidate_one_bucket(struct btree_trans *trans,
 	goto out;
 }
 
+static struct bkey_s_c next_lru_key(struct btree_trans *trans, struct btree_iter *iter,
+				    struct bch_dev *ca, bool *wrapped)
+{
+	struct bkey_s_c k;
+again:
+	k = bch2_btree_iter_peek_upto(iter, lru_pos(ca->dev_idx, U64_MAX, LRU_TIME_MAX));
+	if (!k.k && !*wrapped) {
+		bch2_btree_iter_set_pos(iter, lru_pos(ca->dev_idx, 0, 0));
+		*wrapped = true;
+		goto again;
+	}
+
+	return k;
+}
+
 static void bch2_do_invalidates_work(struct work_struct *work)
 {
 	struct bch_fs *c = container_of(work, struct bch_fs, invalidate_work);
@@ -2024,12 +2049,33 @@ static void bch2_do_invalidates_work(struct work_struct *work)
 	for_each_member_device(c, ca) {
 		s64 nr_to_invalidate =
 			should_invalidate_buckets(ca, bch2_dev_usage_read(ca));
+		struct btree_iter iter;
+		bool wrapped = false;
+
+		bch2_trans_iter_init(trans, &iter, BTREE_ID_lru,
+				     lru_pos(ca->dev_idx, 0,
+					     ((bch2_current_io_time(c, READ) + U32_MAX) &
+					      LRU_TIME_MAX)), 0);
 
-		ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
-				lru_pos(ca->dev_idx, 0, 0),
-				lru_pos(ca->dev_idx, U64_MAX, LRU_TIME_MAX),
-				BTREE_ITER_intent, k,
-			invalidate_one_bucket(trans, &iter, k, &nr_to_invalidate));
+		while (true) {
+			bch2_trans_begin(trans);
+
+			struct bkey_s_c k = next_lru_key(trans, &iter, ca, &wrapped);
+			ret = bkey_err(k);
+			if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+				continue;
+			if (ret)
+				break;
+			if (!k.k)
+				break;
+
+			ret = invalidate_one_bucket(trans, &iter, k, &nr_to_invalidate);
+			if (ret)
+				break;
+
+			bch2_btree_iter_advance(&iter);
+		}
+		bch2_trans_iter_exit(trans, &iter);
 
 		if (ret < 0) {
 			bch2_dev_put(ca);
@@ -2204,7 +2250,7 @@ int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
 	if (ret)
 		return ret;
 
-	now = atomic64_read(&c->io_clock[rw].now);
+	now = bch2_current_io_time(c, rw);
 	if (a->v.io_time[rw] == now)
 		goto out;
 

fs/bcachefs/alloc_background.h

@@ -141,7 +141,13 @@ static inline u64 alloc_lru_idx_fragmentation(struct bch_alloc_v4 a,
 	    !bch2_bucket_sectors_fragmented(ca, a))
 		return 0;
 
-	u64 d = bch2_bucket_sectors_dirty(a);
+	/*
+	 * avoid overflowing LRU_TIME_BITS on a corrupted fs, when
+	 * bucket_sectors_dirty is (much) bigger than bucket_size
+	 */
+	u64 d = min(bch2_bucket_sectors_dirty(a),
+		    ca->mi.bucket_size);
+
 	return div_u64(d * (1ULL << 31), ca->mi.bucket_size);
 }
 

fs/bcachefs/bcachefs.h

@@ -1214,6 +1214,11 @@ static inline s64 bch2_current_time(const struct bch_fs *c)
 	return timespec_to_bch2_time(c, now);
 }
 
+static inline u64 bch2_current_io_time(const struct bch_fs *c, int rw)
+{
+	return max(1ULL, (u64) atomic64_read(&c->io_clock[rw].now) & LRU_TIME_MAX);
+}
+
 static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
 {
 	struct stdio_redirect *stdio = c->stdio;

fs/bcachefs/bcachefs_format.h

@@ -476,6 +476,9 @@ struct bch_lru {
 
 #define LRU_ID_STRIPES		(1U << 16)
 
+#define LRU_TIME_BITS	48
+#define LRU_TIME_MAX	((1ULL << LRU_TIME_BITS) - 1)
+
 /* Optional/variable size superblock sections: */
 
 struct bch_sb_field {
@@ -987,8 +990,9 @@ enum bch_version_upgrade_opts {
 
 #define BCH_ERROR_ACTIONS()		\
 	x(continue,		0)	\
-	x(ro,			1)	\
-	x(panic,		2)
+	x(fix_safe,		1)	\
+	x(panic,		2)	\
+	x(ro,			3)
 
 enum bch_error_actions {
 #define x(t, n) BCH_ON_ERROR_##t = n,
@@ -1382,9 +1386,10 @@ enum btree_id {
 
 /*
  * Maximum number of btrees that we will _ever_ have under the current scheme,
- * where we refer to them with bitfields
+ * where we refer to them with 64 bit bitfields - and we also need a bit for
+ * the interior btree node type:
  */
-#define BTREE_ID_NR_MAX		64
+#define BTREE_ID_NR_MAX		63
 
 static inline bool btree_id_is_alloc(enum btree_id id)
 {

fs/bcachefs/bkey.c

@@ -1064,7 +1064,7 @@ void bch2_bkey_swab_key(const struct bkey_format *_f, struct bkey_packed *k)
 {
 	const struct bkey_format *f = bkey_packed(k) ? _f : &bch2_bkey_format_current;
 	u8 *l = k->key_start;
-	u8 *h = (u8 *) (k->_data + f->key_u64s) - 1;
+	u8 *h = (u8 *) ((u64 *) k->_data + f->key_u64s) - 1;
 
 	while (l < h) {
 		swap(*l, *h);

fs/bcachefs/bkey_methods.c

@@ -398,8 +398,12 @@ void __bch2_bkey_compat(unsigned level, enum btree_id btree_id,
 	for (i = 0; i < nr_compat; i++)
 	switch (!write ? i : nr_compat - 1 - i) {
 	case 0:
-		if (big_endian != CPU_BIG_ENDIAN)
+		if (big_endian != CPU_BIG_ENDIAN) {
+			bch2_bkey_swab_key(f, k);
+		} else if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG)) {
 			bch2_bkey_swab_key(f, k);
+			bch2_bkey_swab_key(f, k);
+		}
 		break;
 	case 1:
 		if (version < bcachefs_metadata_version_bkey_renumber)

fs/bcachefs/bkey_methods.h

@@ -129,7 +129,8 @@ static inline void bch2_bkey_compat(unsigned level, enum btree_id btree_id,
 			       struct bkey_packed *k)
 {
 	if (version < bcachefs_metadata_version_current ||
-	    big_endian != CPU_BIG_ENDIAN)
+	    big_endian != CPU_BIG_ENDIAN ||
+	    IS_ENABLED(CONFIG_BCACHEFS_DEBUG))
 		__bch2_bkey_compat(level, btree_id, version,
 				   big_endian, write, f, k);
 

fs/bcachefs/btree_iter.c

@@ -3161,6 +3161,7 @@ struct btree_trans *__bch2_trans_get(struct bch_fs *c, unsigned fn_idx)
 list_add_done:
 	seqmutex_unlock(&c->btree_trans_lock);
 got_trans:
+	trans->ref.closure_get_happened = false;
 	trans->c		= c;
 	trans->last_begin_time	= local_clock();
 	trans->fn_idx		= fn_idx;
@@ -3235,7 +3236,6 @@ void bch2_trans_put(struct btree_trans *trans)
 	trans_for_each_update(trans, i)
 		__btree_path_put(trans->paths + i->path, true);
 	trans->nr_updates	= 0;
-	trans->locking_wait.task = NULL;
 
 	check_btree_paths_leaked(trans);
 
@@ -3256,6 +3256,13 @@ void bch2_trans_put(struct btree_trans *trans)
 	if (unlikely(trans->journal_replay_not_finished))
 		bch2_journal_keys_put(c);
 
+	/*
+	 * trans->ref protects trans->locking_wait.task, btree_paths arary; used
+	 * by cycle detector
+	 */
+	closure_sync(&trans->ref);
+	trans->locking_wait.task = NULL;
+
 	unsigned long *paths_allocated = trans->paths_allocated;
 	trans->paths_allocated	= NULL;
 	trans->paths		= NULL;
@@ -3273,8 +3280,6 @@ void bch2_trans_put(struct btree_trans *trans)
 		trans = this_cpu_xchg(c->btree_trans_bufs->trans, trans);
 
 	if (trans) {
-		closure_sync(&trans->ref);
-
 		seqmutex_lock(&c->btree_trans_lock);
 		list_del(&trans->list);
 		seqmutex_unlock(&c->btree_trans_lock);

fs/bcachefs/btree_types.h

@@ -761,13 +761,13 @@ static inline bool btree_node_type_needs_gc(enum btree_node_type type)
 
 static inline bool btree_node_type_is_extents(enum btree_node_type type)
 {
-	const unsigned mask = 0
+	const u64 mask = 0
 #define x(name, nr, flags, ...)	|((!!((flags) & BTREE_ID_EXTENTS)) << (nr + 1))
 	BCH_BTREE_IDS()
 #undef x
 	;
 
-	return (1U << type) & mask;
+	return BIT_ULL(type) & mask;
 }
 
 static inline bool btree_id_is_extents(enum btree_id btree)
@@ -777,35 +777,35 @@ static inline bool btree_id_is_extents(enum btree_id btree)
 
 static inline bool btree_type_has_snapshots(enum btree_id id)
 {
-	const unsigned mask = 0
+	const u64 mask = 0
 #define x(name, nr, flags, ...)	|((!!((flags) & BTREE_ID_SNAPSHOTS)) << nr)
 	BCH_BTREE_IDS()
 #undef x
 	;
 
-	return (1U << id) & mask;
+	return BIT_ULL(id) & mask;
 }
 
 static inline bool btree_type_has_snapshot_field(enum btree_id id)
 {
-	const unsigned mask = 0
+	const u64 mask = 0
 #define x(name, nr, flags, ...)	|((!!((flags) & (BTREE_ID_SNAPSHOT_FIELD|BTREE_ID_SNAPSHOTS))) << nr)
 	BCH_BTREE_IDS()
 #undef x
 	;
 
-	return (1U << id) & mask;
+	return BIT_ULL(id) & mask;
 }
 
 static inline bool btree_type_has_ptrs(enum btree_id id)
 {
-	const unsigned mask = 0
+	const u64 mask = 0
 #define x(name, nr, flags, ...)	|((!!((flags) & BTREE_ID_DATA)) << nr)
 	BCH_BTREE_IDS()
 #undef x
 	;
 
-	return (1U << id) & mask;
+	return BIT_ULL(id) & mask;
 }
 
 struct btree_root {

fs/bcachefs/errcode.h

@@ -116,6 +116,9 @@
 	x(ENOENT,			ENOENT_dev_idx_not_found)		\
 	x(ENOTEMPTY,			ENOTEMPTY_dir_not_empty)		\
 	x(ENOTEMPTY,			ENOTEMPTY_subvol_not_empty)		\
+	x(EEXIST,			EEXIST_str_hash_set)			\
+	x(EEXIST,			EEXIST_discard_in_flight_add)		\
+	x(EEXIST,			EEXIST_subvolume_create)		\
 	x(0,				open_buckets_empty)			\
 	x(0,				freelist_empty)				\
 	x(BCH_ERR_freelist_empty,	no_buckets_found)			\