Hi, Christoph!
在 2022/10/18 15:38, Christoph Hellwig 写道:
Zero out the pointers to the holder related kobjects so that
bd_unlink_disk_holder does the right thing when dm cleans up the delayed
holders after add_disk fails.
I'm planning to do the similar thing in the normal path:
1) in del_gendisk: (add a new api kobject_put_and_test)
if (kobject_put_and_test(bd_holder_dir/slave_dir))
bd_holder_dir/slave_dir = NULL;
2) in bd_link_disk_holder, get bd_holder_dir first:
if (!kobject_get_unless_zero(bd_holder_dir))
return -ENODEV;
...
bd_find_holder_disk()
Do you think this is ok?
Thanks,
Kuai
Fixes: 89f871af1b26 ("dm: delay registering the gendisk")
Reported-by: Yu Kuai <yukuai1@xxxxxxxxxxxxxxx>
Signed-off-by: Christoph Hellwig <hch@xxxxxx>
---
block/genhd.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/block/genhd.c b/block/genhd.c
index 6123005154b2a..734c2e74f42e3 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -530,8 +530,10 @@ int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
bd_unregister_all_holders(disk);
out_put_slave_dir:
kobject_put(disk->slave_dir);
+ disk->slave_dir = NULL;
out_put_holder_dir:
kobject_put(disk->part0->bd_holder_dir);
+ disk->part0->bd_holder_dir = NULL;
out_del_integrity:
blk_integrity_del(disk);
out_del_block_link:
