Hello,
these patches fix races happening when devices are frequently destroyed and
recreated in association of block device inode with corresponding gendisk.
Generally when such race happen it results in use-after-free issues, block
device page cache inconsistencies, or other problems. I have verified these
patches fix use-after-free issues that could be reproduced by frequent creation
and destruction of loop device. I was not able to reproduce races reported by
Hou Tao [1] related to gendisk-blkdev association (at least I was not able to
hit any issues with stressing using scsi_debug in various ways with or without
patches). My patches should fix those but it would be great if Tao could verify
that. Any comments and review welcome!
Honza
[1] https://www.spinics.net/lists/linux-block/msg20015.html
