On Fri, 2020-01-31 at 19:35 +0800, Ian Kent wrote: > On Fri, 2020-01-31 at 19:29 +0800, Ian Kent wrote: > > On Fri, 2020-01-31 at 19:16 +0800, Ian Kent wrote: > > > On Fri, 2020-01-31 at 07:55 +0100, Marc Lehmann wrote: > > > > > > Sounds like I have enough information to duplicate the > > > > > > problem > > > > > > so > > > > > > I'll have a look see. > > > > > > > > > > Oh, as usual, what's the autofs source version of your > > > > > package > > > > > please, > > > > > since I'll be using the current development source to work on > > > > > it. > > > > > > > > It's the one mentioned in my original report (that you probably > > > > didn't > > > > get): 5.1.2-4 > > > > > > I grabbed the Debian source package and unpacked it on an old > > > Ubuntu > > > VM I have. > > > > > > It looks like this package has the change that adds handling of > > > symlinks to umount_multi() so this should not be calling umount() > > > at all. > > > > Ahh, Haa, I see the problem in the Debian package source. > > > > As you say, there is a call to is_mounted() (which calls that > > ioctl) > > at the top of umount_multi(): > > > > if (is_mounted(_PATH_MOUNTED, path, MNTS_ALL)) > > goto real_mount; > > > > and that branches to the real mount umount code, this is now gone > > in the current source. > > > > I'll have a look at the change history and see if I can locate > > the patch (or patches) that removed it. > > There it is, and it looks independent of other changes ... > > autofs-5.1.3 - fix symlink false negative in umount_multi()
And you can find this at: https://mirrors.edge.kernel.org/pub/linux/daemons/autofs/v5/patches-5.1.4/autofs-5.1.3-fix-symlink-false-negative-in-umount_multi.patch Ian
