Re: [PATCH v4 2/5] fuse: Support fuse filesystems outside of init_user_ns

[Andy Lutomirski]

On Wed, Oct 15, 2014 at 3:59 PM, Seth Forshee
<seth.forshee@xxxxxxxxxxxxx> wrote:
> On Wed, Oct 15, 2014 at 10:05:46AM -0700, Andy Lutomirski wrote:
>> On Wed, Oct 15, 2014 at 8:05 AM, Seth Forshee
>> <seth.forshee@xxxxxxxxxxxxx> wrote:
>> > On Wed, Oct 15, 2014 at 07:49:39AM -0700, Andy Lutomirski wrote:
>> >> On 10/14/2014 07:25 AM, Seth Forshee wrote:
>> >> > Update fuse to translate uids and gids to/from the user namspace
>> >> > of the process servicing requests on /dev/fuse. Any ids which do
>> >> > not map into the namespace will result in errors. inodes will
>> >> > also be marked bad when unmappable ids are received from
>> >> > userspace.
>> >> >
>> >> > Due to security concerns the namespace used should be fixed,
>> >> > otherwise a user might be able to gain elevated privileges or
>> >> > influence processes that the user would otherwise be unable to
>> >> > manipulate. Thus the namespace of the mounting process is used
>> >> > for all translations, and this namespace is required to be the
>> >> > same as the one in use when /dev/fuse was opened.
>> >> >
>> >>
>> >> I'm not sure that this is necessary if my nosuid patch goes in, but I
>> >> also don't think it makes any sense to hold this up while we find a
>> >> perfect solution.
>> >>
>> >> Is there a decent way to extend this to different translation schemes in
>> >> the future (e.g. a flag at fs setup that could be used)?
>> >
>> > I think it would be possible to relax the translation scheme
>> > restrictions in the future, certainly that's easier than tightening down
>> > a looser restriction. I still favor picking one namespace to use for
>> > translation (surely that's how it would work with other filesystems
>> > anyway) rather than using the current namespace during /dev/fuse I/O. I
>> > did an implementation using the latter technique, and it's far more
>> > complex with no benefits that I can see.
>>
>> Long term, I think we'll want more flexible translations for
>> filesystems on removable media, even when both the mounter and the
>> accessing process are in the init user namespace.  But this can wait.
>
> You've piqued my interest. What are you thinking of which would require
> this flexibility?
>

For an easy example, if I stick a USB stick into my computer and copy
a file to it, I probably want the file to be owned by uid 0 in the FS
metadata (but still owned by me as reported by stat(2) and friends).

For a more complex example, tools like Sandstorm (http://sandstorm.io)
probably want to use FUSE mounted by an outer (non-root) userns and
accessed from an inner userns.  With your patches, this *might* work,
but it might also be a little tricky.

I can also see this ability being extremely useful for NFS and other
network filesystems, where keeping all the uids in sync is currently a
royal PITA.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/