fvogt added a comment.
In D15739#486043 <https://phabricator.kde.org/D15739#486043>, @bruns wrote: > In D15739#486020 <https://phabricator.kde.org/D15739#486020>, @fvogt wrote: > > > In D15739#486009 <https://phabricator.kde.org/D15739#486009>, @bruns wrote: > > > > > It actually already does, via the fstab backend (currently network (SMB/NFS) and various fuse mounts). Fstab in this case also includes everything in MTAB, should be quite easy to extend this. > > > > > > What is the output of `cat /proc/self/mounts` (fell free to remove anything irrelevant, like cgroups)? > > > > > > The entries involved in / are these: > > > > /dev/sr0 /run/initramfs/live iso9660 ro,relatime 0 0 > > /dev/loop0 /run/initramfs/squashfs_container squashfs ro,relatime 0 0 > > /dev/loop1 /run/rootfsbase ext4 ro,relatime,data=ordered 0 0 > > LiveOS_rootfs / overlay rw,relatime,lowerdir=/run/rootfsbase,upperdir=/run/overlayfs/rw,workdir=/run/overlayfs/work 0 0 > > > > > Hm, loop1 is the `lowerdir` of the overlay - how are loop0 and sr0 involved, are these the backing files? /dev/sr0 contains a squashfs, which is visible as /dev/loop0. /dev/loop0 contains an ext4 image, which is visible as /dev/loop1. So it's doubly indirect. > Though, the relevant part is `mntent.mnt.type == "overlay"` and `mntent.mnt_dir == "/"`. Matching for "overlay" is probably sufficient. If adding such a special case makes sense, yes. I'd even argue about 'mntent.mnt_dir == "/" && isKnownFilesystem(mntent)' or something like that to ensure that an entry for `/` is always provided. REPOSITORY R241 KIO REVISION DETAIL https://phabricator.kde.org/D15739 To: meven, #dolphin, #vdg, tcanabrava, ngraham, elvisangelaccio Cc: andriusr, fvogt, meven, elvisangelaccio, Codezela, davidc, tcanabrava, ndavis, romangg, bruns, davidedmundson, abetts, svenmauch, broulik, acrouthamel, kde-frameworks-devel, LeGast00n, michaelh, ngraham