[lkml]   [2019]   [Jun]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH RFC 00/10] RDMA/FS DAX truncate proposal
On Thu, Jun 06, 2019 at 04:51:15PM -0300, Jason Gunthorpe wrote:
> On Thu, Jun 06, 2019 at 12:42:03PM +0200, Jan Kara wrote:
> > So I'd like to actually mandate that you *must* hold the file lease until
> > you unpin all pages in the given range (not just that you have an option to
> > hold a lease). And I believe the kernel should actually enforce this. That
> > way we maintain a sane state that if someone uses a physical location of
> > logical file offset on disk, he has a layout lease. Also once this is done,
> > sysadmin has a reasonably easy way to discover run-away RDMA application
> > and kill it if he wishes so.
> >
> > The question is on how to exactly enforce that lease is taken until all
> > pages are unpinned. I belive it could be done by tracking number of
> > long-term pinned pages within a lease. Gup_longterm could easily increment
> > the count when verifying the lease exists, gup_longterm users will somehow
> > need to propagate corresponding 'filp' (struct file pointer) to
> > put_user_pages_longterm() callsites so that they can look up appropriate
> > lease to drop reference - probably I'd just transition all gup_longterm()
> > users to a saner API similar to the one we have in mm/frame_vector.c where
> > we don't hand out page pointers but an encapsulating structure that does
> > all the necessary tracking. Removing a lease would need to block until all
> > pins are released - this is probably the most hairy part since we need to
> > handle a case if application just closes the file descriptor which
> > would
> I think if you are going to do this then the 'struct filp' that
> represents the lease should be held in the kernel (ie inside the RDMA
> umem) until the kernel is done with it.

Yea there seems merit to this. I'm still not resolving how this helps track
who has the pin across a fork.

> Actually does someone have a pointer to this userspace lease API, I'm
> not at all familiar with it, thanks

man fcntl
search for SETLEASE

But I had to add the F_LAYOUT lease type. (Personally I'm for calling it
F_LONGTERM at this point. I don't think LAYOUT is compatible with what we are
proposing here.)

Anyway, yea would be a libc change at lease for man page etc... But again I
want to get some buy in before going through all that.

> And yes, a better output format from GUP would be great..
> > Maybe we could block only on explicit lease unlock and just drop the layout
> > lease on file close and if there are still pinned pages, send SIGKILL to an
> > application as a reminder it did something stupid...
> Which process would you SIGKILL? At least for the rdma case a FD is
> holding the GUP, so to do the put_user_pages() the kernel needs to
> close the FD. I guess it would have to kill every process that has the
> FD open? Seems complicated...

Tending to agree... But I'm still not opposed to killing bad actors... ;-)

NOTE: Jason I think you need to be more clear about the FD you are speaking of.
I believe you mean the FD which refers to the RMDA context. That is what I
called it in my other email.


> Regards,
> Jason

 \ /
  Last update: 2019-06-07 00:22    [W:0.164 / U:2.284 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site