Hi Gunnar, * Gunnar Wolf <gw...@gwolf.org> [220416 19:54]: > If I request losetup to detach from a loop device and it cannot be > detached, I would expect it to notify me accordingly. However: > > # losetup --list > NAME SIZELIMIT OFFSET AUTOCLEAR RO BACK-FILE > DIO LOG-SEC > /dev/loop1 0 0 1 0 > /home/gwolf/vcs/raspi3-image-spec/raspi3.img (deleted) 0 512 > # if losetup --detach /dev/loop1 ; then echo SUCCESS; else echo FAIL; fi > SUCCESS > # losetup --list > NAME SIZELIMIT OFFSET AUTOCLEAR RO BACK-FILE > DIO LOG-SEC > /dev/loop1 0 0 1 0 > /home/gwolf/vcs/raspi3-image-spec/raspi3.img (deleted) 0 512 > > When requesting debug information, it seems losetup actually believes > the loopback was dropped - but it is false: > > # LOOPDEV_DEBUG=all losetup --detach /dev/loop1 > 9932: loopdev: CXT: [0x7ffeaa764d60]: initialize context > 9932: loopdev: CXT: [0x7ffeaa764d60]: init: ignore ioctls > 9932: loopdev: CXT: [0x7ffeaa764d60]: init: loop-control detected > 9932: loopdev: CXT: [0x7ffeaa764d60]: /dev/loop1 name assigned > 9932: loopdev: CXT: [0x7ffeaa764d60]: open /dev/loop1 [ro]: Success > 9932: loopdev: CXT: [0x7ffeaa764d60]: device removed > 9932: loopdev: CXT: [0x7ffeaa764d60]: de-initialize > 9932: loopdev: CXT: [0x7ffeaa764d60]: closing old open fd > 9932: loopdev: ITER: [0x7ffeaa764f18]: de-initialize > # losetup --list > NAME SIZELIMIT OFFSET AUTOCLEAR RO BACK-FILE > DIO LOG-SEC > /dev/loop1 0 0 1 0 > /home/gwolf/vcs/raspi3-image-spec/raspi3.img (deleted) 0 512
Sorry for the long time with no reply to your bug. I took a closer look today; I can reproduce this -if- I have a filesystem mounted using that loop device. This comment from upstream explains whats going on: https://github.com/util-linux/util-linux/issues/484#issuecomment-321506607 Apparently losetup cannot (really) know that the kernel ignores the detach ioctl. I would say there is no bug in util-linux. Maybe you want to talk to the kernel folks? Best, Chris
