Merge tag 'bcachefs-2024-07-12' of https://evilpiepirate.org/git/bcachefs

Pull more bcachefs fixes from Kent Overstreet:

 - revert the SLAB_ACCOUNT patch, something crazy is going on in memcg
   and someone forgot to test

 - minor fixes: missing rcu_read_lock(), scheduling while atomic (in an
   emergency shutdown path)

 - two lockdep fixes; these could have gone earlier, but were left to
   bake awhile

* tag 'bcachefs-2024-07-12' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: bch2_gc_btree() should not use btree_root_lock
  bcachefs: Set PF_MEMALLOC_NOFS when trans->locked
  bcachefs; Use trans_unlock_long() when waiting on allocator
  Revert "bcachefs: Mark bch_inode_info as SLAB_ACCOUNT"
  bcachefs: fix scheduling while atomic in break_cycle()
  bcachefs: Fix RCU splat

11 files changed, 87 insertions, 24 deletions

diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 7c72a9e6f511..a0deb8266011 100644
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -641,16 +641,30 @@ static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree, bool in
 		target_depth = 0;
 
 	/* root */
-	mutex_lock(&c->btree_root_lock);
-	struct btree *b = bch2_btree_id_root(c, btree)->b;
-	if (!btree_node_fake(b)) {
+	do {
+retry_root:
+		bch2_trans_begin(trans);
+
+		struct btree_iter iter;
+		bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN,
+					  0, bch2_btree_id_root(c, btree)->b->c.level, 0);
+		struct btree *b = bch2_btree_iter_peek_node(&iter);
+		ret = PTR_ERR_OR_ZERO(b);
+		if (ret)
+			goto err_root;
+
+		if (b != btree_node_root(c, b)) {
+			bch2_trans_iter_exit(trans, &iter);
+			goto retry_root;
+		}
+
 		gc_pos_set(c, gc_pos_btree(btree, b->c.level + 1, SPOS_MAX));
-		ret = lockrestart_do(trans,
-			bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1,
-					 NULL, NULL, bkey_i_to_s_c(&b->key), initial));
+		struct bkey_s_c k = bkey_i_to_s_c(&b->key);
+		ret = bch2_gc_mark_key(trans, btree, b->c.level + 1, NULL, NULL, k, initial);
 		level = b->c.level;
-	}
-	mutex_unlock(&c->btree_root_lock);
+err_root:
+		bch2_trans_iter_exit(trans, &iter);
+	} while (bch2_err_matches(ret, BCH_ERR_transaction_restart));
 
 	if (ret)
 		return ret;
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index 0ed9e6574fcd..19352a08ea20 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -996,7 +996,7 @@ retry_all:
 
 	bch2_trans_unlock(trans);
 	cond_resched();
-	trans->locked = true;
+	trans_set_locked(trans);
 
 	if (unlikely(trans->memory_allocation_failure)) {
 		struct closure cl;
@@ -3089,7 +3089,8 @@ u32 bch2_trans_begin(struct btree_trans *trans)
 		bch2_trans_srcu_unlock(trans);
 
 	trans->last_begin_ip = _RET_IP_;
-	trans->locked  = true;
+
+	trans_set_locked(trans);
 
 	if (trans->restarted) {
 		bch2_btree_path_traverse_all(trans);
@@ -3159,7 +3160,6 @@ got_trans:
 	trans->last_begin_time	= local_clock();
 	trans->fn_idx		= fn_idx;
 	trans->locking_wait.task = current;
-	trans->locked		= true;
 	trans->journal_replay_not_finished =
 		unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)) &&
 		atomic_inc_not_zero(&c->journal_keys.ref);
@@ -3193,6 +3193,7 @@ got_trans:
 	trans->srcu_idx		= srcu_read_lock(&c->btree_trans_barrier);
 	trans->srcu_lock_time	= jiffies;
 	trans->srcu_held	= true;
+	trans_set_locked(trans);
 
 	closure_init_stack_release(&trans->ref);
 	return trans;
diff --git a/fs/bcachefs/btree_locking.c b/fs/bcachefs/btree_locking.c
index d66fff22109a..c51826fd557f 100644
--- a/fs/bcachefs/btree_locking.c
+++ b/fs/bcachefs/btree_locking.c
@@ -231,7 +231,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle)
 			prt_newline(&buf);
 		}
 
-		bch2_print_string_as_lines(KERN_ERR, buf.buf);
+		bch2_print_string_as_lines_nonblocking(KERN_ERR, buf.buf);
 		printbuf_exit(&buf);
 		BUG();
 	}
@@ -792,7 +792,7 @@ static inline int __bch2_trans_relock(struct btree_trans *trans, bool trace)
 			return bch2_trans_relock_fail(trans, path, &f, trace);
 	}
 
-	trans->locked = true;
+	trans_set_locked(trans);
 out:
 	bch2_trans_verify_locks(trans);
 	return 0;
@@ -812,16 +812,14 @@ void bch2_trans_unlock_noassert(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock(struct btree_trans *trans)
 {
 	__bch2_trans_unlock(trans);
 
-	trans->locked = false;
-	trans->last_unlock_ip = _RET_IP_;
+	trans_set_unlocked(trans);
 }
 
 void bch2_trans_unlock_long(struct btree_trans *trans)
diff --git a/fs/bcachefs/btree_locking.h b/fs/bcachefs/btree_locking.h
index 7f41545b9147..75a6274c7d27 100644
--- a/fs/bcachefs/btree_locking.h
+++ b/fs/bcachefs/btree_locking.h
@@ -193,6 +193,28 @@ int bch2_six_check_for_deadlock(struct six_lock *lock, void *p);
 
 /* lock: */
 
+static inline void trans_set_locked(struct btree_trans *trans)
+{
+	if (!trans->locked) {
+		trans->locked = true;
+		trans->last_unlock_ip = 0;
+
+		trans->pf_memalloc_nofs = (current->flags & PF_MEMALLOC_NOFS) != 0;
+		current->flags |= PF_MEMALLOC_NOFS;
+	}
+}
+
+static inline void trans_set_unlocked(struct btree_trans *trans)
+{
+	if (trans->locked) {
+		trans->locked = false;
+		trans->last_unlock_ip = _RET_IP_;
+
+		if (!trans->pf_memalloc_nofs)
+			current->flags &= ~PF_MEMALLOC_NOFS;
+	}
+}
+
 static inline int __btree_node_lock_nopath(struct btree_trans *trans,
 					 struct btree_bkey_cached_common *b,
 					 enum six_lock_type type,
diff --git a/fs/bcachefs/btree_types.h b/fs/bcachefs/btree_types.h
index 87f485e9c552..48cb1a7d31c5 100644
--- a/fs/bcachefs/btree_types.h
+++ b/fs/bcachefs/btree_types.h
@@ -484,6 +484,7 @@ struct btree_trans {
 	bool			lock_may_not_fail:1;
 	bool			srcu_held:1;
 	bool			locked:1;
+	bool			pf_memalloc_nofs:1;
 	bool			write_locked:1;
 	bool			used_mempool:1;
 	bool			in_traverse_all:1;
diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index 743d57eba760..314ee3e0187f 100644
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -805,7 +805,7 @@ int bch2_bucket_ref_update(struct btree_trans *trans, struct bch_dev *ca,
 			"bucket %u:%zu gen %u (mem gen %u) data type %s: stale dirty ptr (gen %u)\n"
 			"while marking %s",
 			ptr->dev, bucket_nr, b_gen,
-			*bucket_gen(ca, bucket_nr),
+			bucket_gen_get(ca, bucket_nr),
 			bch2_data_type_str(bucket_data_type ?: ptr_data_type),
 			ptr->gen,
 			(printbuf_reset(&buf),
diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index 80ee0be9793e..8ad4be73860c 100644
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -116,6 +116,14 @@ static inline u8 *bucket_gen(struct bch_dev *ca, size_t b)
 	return gens->b + b;
 }
 
+static inline u8 bucket_gen_get(struct bch_dev *ca, size_t b)
+{
+	rcu_read_lock();
+	u8 gen = *bucket_gen(ca, b);
+	rcu_read_unlock();
+	return gen;
+}
+
 static inline size_t PTR_BUCKET_NR(const struct bch_dev *ca,
 				   const struct bch_extent_ptr *ptr)
 {
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index 74a1e12a7a14..fa1fee05cf8f 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -2073,8 +2073,7 @@ int __init bch2_vfs_init(void)
 {
 	int ret = -ENOMEM;
 
-	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT |
-				      SLAB_ACCOUNT);
+	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
 	if (!bch2_inode_cache)
 		goto err;
 
diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c
index 4ec979b4b23e..4583c9386e8c 100644
--- a/fs/bcachefs/io_misc.c
+++ b/fs/bcachefs/io_misc.c
@@ -125,7 +125,7 @@ err_noprint:
 	bch2_bkey_buf_exit(&old, c);
 
 	if (closure_nr_remaining(&cl) != 1) {
-		bch2_trans_unlock(trans);
+		bch2_trans_unlock_long(trans);
 		closure_sync(&cl);
 	}
 
diff --git a/fs/bcachefs/util.c b/fs/bcachefs/util.c
index de331dec2a99..4ec7e44d6e36 100644
--- a/fs/bcachefs/util.c
+++ b/fs/bcachefs/util.c
@@ -252,8 +252,10 @@ void bch2_prt_u64_base2(struct printbuf *out, u64 v)
 	bch2_prt_u64_base2_nbits(out, v, fls64(v) ?: 1);
 }
 
-void bch2_print_string_as_lines(const char *prefix, const char *lines)
+static void __bch2_print_string_as_lines(const char *prefix, const char *lines,
+					 bool nonblocking)
 {
+	bool locked = false;
 	const char *p;
 
 	if (!lines) {
@@ -261,7 +263,13 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 		return;
 	}
 
-	console_lock();
+	if (!nonblocking) {
+		console_lock();
+		locked = true;
+	} else {
+		locked = console_trylock();
+	}
+
 	while (1) {
 		p = strchrnul(lines, '\n');
 		printk("%s%.*s\n", prefix, (int) (p - lines), lines);
@@ -269,7 +277,18 @@ void bch2_print_string_as_lines(const char *prefix, const char *lines)
 			break;
 		lines = p + 1;
 	}
-	console_unlock();
+	if (locked)
+		console_unlock();
+}
+
+void bch2_print_string_as_lines(const char *prefix, const char *lines)
+{
+	return __bch2_print_string_as_lines(prefix, lines, false);
+}
+
+void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines)
+{
+	return __bch2_print_string_as_lines(prefix, lines, true);
 }
 
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *task, unsigned skipnr,
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index 5d2c470a49ac..5b0533ec4c7e 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -315,6 +315,7 @@ void bch2_prt_u64_base2_nbits(struct printbuf *, u64, unsigned);
 void bch2_prt_u64_base2(struct printbuf *, u64);
 
 void bch2_print_string_as_lines(const char *prefix, const char *lines);
+void bch2_print_string_as_lines_nonblocking(const char *prefix, const char *lines);
 
 typedef DARRAY(unsigned long) bch_stacktrace;
 int bch2_save_backtrace(bch_stacktrace *stack, struct task_struct *, unsigned, gfp_t);