Re: [PATCH 0/5] locks: move most locks_release_private calls outside of i_lock

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, 12 Aug 2014 08:32:29 -0700
Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:

> Btw, I might be missing something here, but wouldn't it be better
> to reference count the file_lock structure and grab a reference to
> it where we currently call (__)locks_copy_lock?
> 

It's not really possible with the way this code works at the moment.
The problem there is that struct file_lock can represent any of:

- a lock request (copied from userland into a kernel structure)
- a lock record (that lives on the i_flock list)
- a conflicting lock record (returned by GETLK-type request)

In many cases we call (__)locks_copy_lock to copy from one "use-type" of
struct file_lock to another, and doing that with refcounts makes that a
bit difficult to manage.

If I were designing this code from scratch, I'd have probably made each
use-type a distinct structure. Maybe we should do that anyway at
some point -- I'm not sure.

Now, all that said...I think we will end up needing to do some sort of
refcounting to fix the leasing code at some point. Currently,
->setlease operations can't block, which is a significant impediment to
adding leases to clustered filesystems and the like. It would be nice
to lift that limitation and that may require making leases be
refcounted (or maybe RCU-managed).

-- 
Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Filesystem Development]     [Linux USB Development]     [Linux Media Development]     [Video for Linux]     [Linux NILFS]     [Linux Audio Users]     [Yosemite Info]     [Linux SCSI]

  Powered by Linux