The new year isn't very old yet, and already there
are questions over questions regarding the future
direction linux-vserver will take ...
anyway, interesting times and so I once again try
to get your opinion on a specific patch, the so
called Bind Mount Extensions (short BME)
(the hasty reader can skip to OPINION POLL)
what is it?
simple, the Linux kernel supports a few 'magic'
tricks in the vfs (virtual filesystem) layer, which
can be used to 'create' different views of the
underlying filesystems ... those are better known
as --bind and --rbind mounts, but also --move
mounts belong to this category.
what can I do with it?
they basically allow you to 'copy' or 'move' the
'view' of an entire directory to a different point
in the vfs hierarchy, let's take an example here:
mount --bind /var/tmp /tmp
the result is that both /tmp and /var/tmp point to
the same directory (/var/tmp) in the filesystem,
even if they originally (e.g. var) are on different
why would I need BME?
well, unfortunately the kernel design is flawed,
or better incomplete, in that way, that it does
not honor certain mount attributes, like read only
or noatime, more than that, it silently ignores
mount --bind -o ro /tmp /mnt/
please check (X) all points which apply ...
(according to your opinion)
[ ] I had no idea about this issue/deficiency
[ ] I did already know about it
[ ] I'm already using BME (or similar) patches
[ ] I will start using BME patches now
[ ] I do not need/use this 'feature' at all
[ ] I think this SHOULD get into the next
stable Linux-VServer release because ...
[ ] I am using it/want to use it
[ ] It's more a bugfix than a feature
[ ] I like the idea of ro --bind mounts
[ ] I think this should NOT be included into
the next stable release because ...
[ ] it is not thoroughly tested
[ ] it is not mature enough
[ ] it adds unnecessary code
[ ] it should better be fixed in
the mainline Linux Kernel
Vserver mailing list
Received on Fri Jan 6 22:15:29 2006