On Sat, Jan 12, 2019 at 10:20:26AM +0800, yangerkun wrote: > > > Dave Chinner wrote on 2019/1/10 5:02: > >On Wed, Jan 09, 2019 at 01:48:45PM +0800, yangerkun wrote: > >>Case generic/108 sometimes will fail while testing ext2, and the reson > >>is that rmmod scsi_debug in _put_scsi_debug_dev may randomly fail. Now > >>add the loop to do rmmod to make sure scsi_debug can be removed > >>correctly. > > > >Why does 'rmmod scsi_debug' randomly fail? > > > >What bug does ext2 have that prevents the scsi debug module from > >being released and hence removed? > > It's not a bug with ext2, ever been existing in ext4 too. This patch > is a reference to 'https://gitlab.com/cryptsetup/cryptsetup.git' > commit d7b9ed05f0931b416c33c8eb2ff1e6efa39270ff, and the reason why > need this is that the behavior of udev cannot be speculated, so > scsi_debug may rmmod failed since udev scan open the device and take > the reference of module scsi_debug. IOWs, you copied a hack from cryptsetup tests because you didn't know about $UDEV_SETTLE_PROG and didn't think to ask if anyone knew a solution to this problem? Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
