Re: losetup -d --force for zombie loop devices?

[Mike Frysinger <vapier@gentoo.org>]

[-- Attachment #1: Type: Text/Plain, Size: 2153 bytes --]

On Friday 11 January 2013 19:54:16 Phillip Susi wrote:
> On 01/11/2013 06:52 PM, Mike Frysinger wrote:
> > On Thursday 03 May 2012 10:29:07 Phillip Susi wrote:
> >> If you want to hide mounts from certain processes, that is what
> >> unshare is for.  Hiding a mount from all processes does not make
> >> sense.  If you know a mount is gone and beyond recovery ( like in
> >> this loop over nfs case, or removed media ), then it should be
> >> forcibly unmounted, not simply made invisible and doomed to
> >> remain a zombie mount until the system is rebooted.
> > 
> > in an ideal world, maybe unshare might work.  in the real world, it
> > doesn't. you can use it only on *new* processes, not ones that are
> > already running. nor can you do `unshare shutdown` and have it work
> > since that simply signals a long running init process to initiate a
> > shutdown.
> > 
> > an nfs server goes afk and attempts to `umount` it timeout, as well
> > as many desktop programs (like kde io daemons that like to walk
> > available mount points) or shutdown processes.  no call to
> > `unshare` will fix this, but certainly forcibly removing it with
> > `umount -l` will.
> 
> A bit of a delayed response there

the recent unshare patches reminded me of this.  and i had actually used 
unshare in the interim so i know how it works now.

> but my point was that what you are
> looking for is umount -f, not umount -l.

and my point is that `umount -f` doesn't always work which means `umount -l` 
is sometimes the only way to remove a mount point.  an unresponsive remote or 
something is holding open a reference (which doesn't show up in `lsof -n`).

> It also used to be that once
> you detached a mount with umount -l, you could not reattach it.
> Attempts to remount would fail with EBUSY and so this was a horribly
> broken state to be in.  At some point it seems this was changed and
> now you *can* reattach, so lazy unmount is no longer pure evil, but
> still it is no forced unmount.

it hasn't been that way for quite some time, and for network based mounts, 
it's pretty much never been that way ?
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]