> So AFAICT this fixes the UAF issues. It does reintroduce the EBUSY problems > with mount racing with umount e.g. for EROFS which calls bdev_release() It shouldn't as erofs doesn't use block devices when it is in fscache mode which is when kill_anon_super() is called. For regular erofs kill_block_super() is used and notification happens as before. > after calling kill_anon_super() but I think we can live with that until we > come up with a neater solution for this problem. So feel free to add: I think ultimately we might just provide callback to kill_*_super() like we do for sget_fc() or something. But let's keep thinking of course.
