Re: Lease semantic proposal

From: Ira Weiny
Date: Thu Oct 03 2019 - 13:05:30 EST


On Thu, Oct 03, 2019 at 11:01:10AM +0200, Jan Kara wrote:
> On Tue 01-10-19 11:17:00, Ira Weiny wrote:
> > On Mon, Sep 23, 2019 at 04:17:59PM -0400, Jeff Layton wrote:
> > > On Mon, 2019-09-23 at 12:08 -0700, Ira Weiny wrote:
> > >
> > > Will userland require any special privileges in order to set an
> > > F_UNBREAK lease? This seems like something that could be used for DoS. I
> > > assume that these will never time out.
> >
> > Dan and I discussed this some more and yes I think the uid of the process needs
> > to be the owner of the file. I think that is a reasonable mechanism.
>
> Honestly, I'm not convinced anything more than open-for-write should be
> required. Sure unbreakable lease may result in failing truncate and other
> ops but as we discussed at LFS/MM, this is not hugely different from
> executing a file resulting in ETXTBUSY for any truncate attempt (even from
> root). So sufficiently priviledged user has to be able to easily find which
> process(es) owns the lease so that he can kill it / take other
> administrative action to release the lease. But that's about it.

Well that was kind of what I was thinking. However I wanted to be careful
about requiring write permission when doing a F_RDLCK. I think that it has to
be clearly documented _why_ write permission is required.

>
> > > How will we deal with the case where something is is squatting on an
> > > F_UNBREAK lease and isn't letting it go?
> >
> > That is a good question. I had not considered someone taking the UNBREAK
> > without pinning the file.
>
> IMHO the same answer as above - sufficiently priviledged user should be
> able to easily find the process holding the lease and kill it. Given the
> lease owner has to have write access to the file, he better should be from
> the same "security domain"...
>
> > > Leases are technically "owned" by the file description -- we can't
> > > necessarily trace it back to a single task in a threaded program. The
> > > kernel task that set the lease may have exited by the time we go
> > > looking.
> > >
> > > Will we be content trying to determine this using /proc/locks+lsof, etc,
> > > or will we need something better?
> >
> > I think using /proc/locks is our best bet. Similar to my intention to report
> > files being pinned.[1]
> >
> > In fact should we consider files with F_UNBREAK leases "pinned" and just report
> > them there?
>
> As Jeff wrote later, /proc/locks is not enough. You need PID(s) which have
> access to the lease and hold it alive. Your /proc/<pid>/ files you had in your
> patches should do that, shouldn't they? Maybe they were not tied to the
> right structure... They really need to be tied to the existence of a lease.

Yes, sorry. I misspoke above.

Right now /proc/<pid>/file_pins indicates that the file is pinned by GUP. I
think it may be reasonable to extend that to any file which has F_UNBREAK
specified. 'file_pins' may be the wrong name when we include F_UNBREAK'ed
leased files, so I will think on the name. But I think this is possible and
desired.

Ira