Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux

block: WARN_ON_ONCE() when we remove active partitions

The logic for disk->open_partitions is:

blkdev_get_by_*()
-> bdev_is_partition()
-> blkdev_get_part()
-> blkdev_get_whole() // bdev_whole->bd_openers++
-> if (part->bd_openers == 0)
disk->open_partitions++
part->bd_openers

In other words, when we first claim/open a partition we increment
disk->open_partitions and only when all part->bd_openers are closed will
disk->open_partitions be zero. That should mean that
disk->open_partitions is always > 0 as long as there's anyone that
has an open partition.

So the check for disk->open_partitions should mean that we can never
remove an active partition that has a holder and holder ops set. Assert
that in the code. The main disk isn't removed so that check doesn't work
for disk->part0 which is what we want. After all we only care about
partition not about the main disk.

Link: https://lore.kernel.org/r/20231017184823.1383356-3-hch@lst.de
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Christian Brauner <brauner@kernel.org>

+17 -13
+17 -13
block/partitions/core.c
··· 274 274 put_device(&part->bd_device); 275 275 } 276 276 277 - static void delete_partition(struct block_device *part) 278 - { 279 - /* 280 - * Remove the block device from the inode hash, so that it cannot be 281 - * looked up any more even when openers still hold references. 282 - */ 283 - remove_inode_hash(part->bd_inode); 284 - bdev_mark_dead(part, false); 285 - drop_partition(part); 286 - } 287 - 288 277 static ssize_t whole_disk_show(struct device *dev, 289 278 struct device_attribute *attr, char *buf) 290 279 { ··· 663 674 sync_blockdev(disk->part0); 664 675 invalidate_bdev(disk->part0); 665 676 666 - xa_for_each_start(&disk->part_tbl, idx, part, 1) 667 - delete_partition(part); 677 + xa_for_each_start(&disk->part_tbl, idx, part, 1) { 678 + /* 679 + * Remove the block device from the inode hash, so that 680 + * it cannot be looked up any more even when openers 681 + * still hold references. 682 + */ 683 + remove_inode_hash(part->bd_inode); 684 + 685 + /* 686 + * If @disk->open_partitions isn't elevated but there's 687 + * still an active holder of that block device things 688 + * are broken. 689 + */ 690 + WARN_ON_ONCE(atomic_read(&part->bd_openers)); 691 + invalidate_bdev(part); 692 + drop_partition(part); 693 + } 668 694 clear_bit(GD_NEED_PART_SCAN, &disk->state); 669 695 670 696 /*