Commit 4abcd80f authored by Linus Torvalds's avatar Linus Torvalds
Browse files

Merge tag 'bcachefs-2024-11-13' of git://evilpiepirate.org/bcachefs 

Pull bcachefs fixes from Kent Overstreet:
 "This fixes one minor regression from the btree cache fixes (in the
  scan_for_btree_nodes repair path) - and the shutdown path fix is the
  big one here, in terms of bugs closed:

   - Assorted tiny syzbot fixes

   - Shutdown path fix: "bch2_btree_write_buffer_flush_going_ro()"

     The shutdown path wasn't flushing the btree write buffer, leading
     to shutting down while we still had operations in flight. This
     fixes a whole slew of syzbot bugs, and undoubtedly other strange
     heisenbugs.

* tag 'bcachefs-2024-11-13' of git://evilpiepirate.org/bcachefs:
  bcachefs: Fix assertion pop in bch2_ptr_swab()
  bcachefs: Fix journal_entry_dev_usage_to_text() overrun
  bcachefs: Allow for unknown key types in backpointers fsck
  bcachefs: Fix assertion pop in topology repair
  bcachefs: Fix hidden btree errors when reading roots
  bcachefs: Fix validate_bset() repair path
  bcachefs: Fix missing validation for bch_backpointer.level
  bcachefs: Fix bch_member.btree_bitmap_shift validation
  bcachefs: bch2_btree_write_buffer_flush_going_ro()
parents 0a9b9d17 840c2fbc

fs/bcachefs/backpointers.c

+12 −5
Original line number Diff line number Diff line
@@ -52,6 +52,12 @@ int bch2_backpointer_validate(struct bch_fs *c, struct bkey_s_c k, 
			      enum bch_validate_flags flags)

{

	struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(k);

	int ret = 0;



	bkey_fsck_err_on(bp.v->level > BTREE_MAX_DEPTH,

			 c, backpointer_level_bad,

			 "backpointer level bad: %u >= %u",

			 bp.v->level, BTREE_MAX_DEPTH);



	rcu_read_lock();

	struct bch_dev *ca = bch2_dev_rcu_noerror(c, bp.k->p.inode);
@@ -64,7 +70,6 @@ int bch2_backpointer_validate(struct bch_fs *c, struct bkey_s_c k, 
	struct bpos bucket = bp_pos_to_bucket(ca, bp.k->p);

	struct bpos bp_pos = bucket_pos_to_bp_noerror(ca, bucket, bp.v->bucket_offset);

	rcu_read_unlock();

	int ret = 0;



	bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||

			 !bpos_eq(bp.k->p, bp_pos),
@@ -947,9 +952,13 @@ int bch2_check_extents_to_backpointers(struct bch_fs *c) 
static int check_one_backpointer(struct btree_trans *trans,

				 struct bbpos start,

				 struct bbpos end,

				 struct bkey_s_c_backpointer bp,

				 struct bkey_s_c bp_k,

				 struct bkey_buf *last_flushed)

{

	if (bp_k.k->type != KEY_TYPE_backpointer)

		return 0;



	struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(bp_k);

	struct bch_fs *c = trans->c;

	struct btree_iter iter;

	struct bbpos pos = bp_to_bbpos(*bp.v);
@@ -1004,9 +1013,7 @@ static int bch2_check_backpointers_to_extents_pass(struct btree_trans *trans, 
				  POS_MIN, BTREE_ITER_prefetch, k,

				  NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({

			progress_update_iter(trans, &progress, &iter, "backpointers_to_extents");

			check_one_backpointer(trans, start, end,

					      bkey_s_c_to_backpointer(k),

					      &last_flushed);

			check_one_backpointer(trans, start, end, k, &last_flushed);

	}));



	bch2_bkey_buf_exit(&last_flushed, c);

fs/bcachefs/btree_gc.c

+1 −1
Original line number Diff line number Diff line
@@ -182,7 +182,7 @@ static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max) 
	bch2_btree_node_drop_keys_outside_node(b);



	mutex_lock(&c->btree_cache.lock);

	bch2_btree_node_hash_remove(&c->btree_cache, b);

	__bch2_btree_node_hash_remove(&c->btree_cache, b);



	bkey_copy(&b->key, &new->k_i);

	ret = __bch2_btree_node_hash_insert(&c->btree_cache, b);

fs/bcachefs/btree_io.c

+1 −5
Original line number Diff line number Diff line
@@ -733,11 +733,8 @@ static int validate_bset(struct bch_fs *c, struct bch_dev *ca, 
			 c, ca, b, i, NULL,

			 bset_past_end_of_btree_node,

			 "bset past end of btree node (offset %u len %u but written %zu)",

			 offset, sectors, ptr_written ?: btree_sectors(c))) {

			 offset, sectors, ptr_written ?: btree_sectors(c)))

		i->u64s = 0;

		ret = 0;

		goto out;

	}



	btree_err_on(offset && !i->u64s,

		     -BCH_ERR_btree_node_read_err_fixable,
@@ -829,7 +826,6 @@ static int validate_bset(struct bch_fs *c, struct bch_dev *ca, 
			       BSET_BIG_ENDIAN(i), write,

			       &bn->format);

	}

out:

fsck_err:

	printbuf_exit(&buf2);

	printbuf_exit(&buf1);

fs/bcachefs/btree_update_interior.c

+2 −1
Original line number Diff line number Diff line
@@ -2398,7 +2398,8 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans, 
	if (new_hash) {

		mutex_lock(&c->btree_cache.lock);

		bch2_btree_node_hash_remove(&c->btree_cache, new_hash);

		bch2_btree_node_hash_remove(&c->btree_cache, b);



		__bch2_btree_node_hash_remove(&c->btree_cache, b);



		bkey_copy(&b->key, new_key);

		ret = __bch2_btree_node_hash_insert(&c->btree_cache, b);

fs/bcachefs/btree_write_buffer.c

+27 −3
Original line number Diff line number Diff line
@@ -277,6 +277,10 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans) 
	bool accounting_replay_done = test_bit(BCH_FS_accounting_replay_done, &c->flags);

	int ret = 0;



	ret = bch2_journal_error(&c->journal);

	if (ret)

		return ret;



	bch2_trans_unlock(trans);

	bch2_trans_begin(trans);
@@ -491,7 +495,8 @@ static int fetch_wb_keys_from_journal(struct bch_fs *c, u64 seq) 
	return ret;

}



static int btree_write_buffer_flush_seq(struct btree_trans *trans, u64 seq)

static int btree_write_buffer_flush_seq(struct btree_trans *trans, u64 seq,

					bool *did_work)

{

	struct bch_fs *c = trans->c;

	struct btree_write_buffer *wb = &c->btree_write_buffer;
@@ -502,6 +507,8 @@ static int btree_write_buffer_flush_seq(struct btree_trans *trans, u64 seq) 


		fetch_from_journal_err = fetch_wb_keys_from_journal(c, seq);



		*did_work |= wb->inc.keys.nr || wb->flushing.keys.nr;



		/*

		 * On memory allocation failure, bch2_btree_write_buffer_flush_locked()

		 * is not guaranteed to empty wb->inc:
@@ -521,17 +528,34 @@ static int bch2_btree_write_buffer_journal_flush(struct journal *j, 
				struct journal_entry_pin *_pin, u64 seq)

{

	struct bch_fs *c = container_of(j, struct bch_fs, journal);

	bool did_work = false;



	return bch2_trans_run(c, btree_write_buffer_flush_seq(trans, seq));

	return bch2_trans_run(c, btree_write_buffer_flush_seq(trans, seq, &did_work));

}



int bch2_btree_write_buffer_flush_sync(struct btree_trans *trans)

{

	struct bch_fs *c = trans->c;

	bool did_work = false;



	trace_and_count(c, write_buffer_flush_sync, trans, _RET_IP_);



	return btree_write_buffer_flush_seq(trans, journal_cur_seq(&c->journal));

	return btree_write_buffer_flush_seq(trans, journal_cur_seq(&c->journal), &did_work);

}



/*

 * The write buffer requires flushing when going RO: keys in the journal for the

 * write buffer don't have a journal pin yet

 */

bool bch2_btree_write_buffer_flush_going_ro(struct bch_fs *c)

{

	if (bch2_journal_error(&c->journal))

		return false;



	bool did_work = false;

	bch2_trans_run(c, btree_write_buffer_flush_seq(trans,

				journal_cur_seq(&c->journal), &did_work));

	return did_work;

}



int bch2_btree_write_buffer_flush_nocheck_rw(struct btree_trans *trans)