On Mon, 15 Jul 2013 09:10:10 +0200 Irek Szczesniak wrote:
> On Thu, Jul 4, 2013 at 3:17 PM, Roland Mainz <[email protected]> wrote:
> > Hi!
> >
> > ----
> >
> > Attached (as "astksh20130628_sync_fsync_syncfs001.diff.txt") is a
> > patch which adds support for |fsync()| and |syncfs()| to the sync(1)
> > builtin.
> >
> > The main purpose of the patch is to help in three cases:
> > 1. Networked/distributed filesystems where |fsync()| or |syncfs()| can
> > used to force syncronisation with the server (for example usefull to
> > weed-out NFSv4 delegation bugs or for syncronisation primitives)
> > 2. Operating system install software (which may not use a graceful
> > shutdown before reboot)
> > 3. Logout scripts (e.g. call |syncfs()| for a user's home directory at
> > logout time to flush his/her data to storage and make (memory) buffer
> > space available for something else. This (much more fine-grained
> > approach) may be preferred over a "raw" sync(2) call on multiuser
> > systems with hundreds of users)
> >
> > Question (mainly for Glenn):
> > Do you see any need that we have to call SFIO's flush facilities
> > before calling |fsync()| or |syncfs()| ?

> IMO you have to do that to ensure that nothing is left in sfio's
> buffers. Remember the rule: First flush to get a safe state, and then
> sync to disk.

sfsync(NiL) should be called

_______________________________________________
ast-developers mailing list
[email protected]
http://lists.research.att.com/mailman/listinfo/ast-developers

Reply via email to