Re: [PATCH 0/3] nfsd state handling fixes

[J . Bruce Fields]

On Wed, Jun 15, 2016 at 09:54:28PM -0400, Oleg Drokin wrote:
> 
> On Jun 14, 2016, at 11:28 PM, Oleg Drokin wrote:
> 
> > These three patches do the much discussed job of making nfsd state handling
> > more robust in face of races where several opens arrive for the same file
> > at the same time from the same client.
> > 
> > This does not yet handle a case when one of those opens gets an error
> > and others don't.
> > 
> > Also this is undergoing testing ATM, so please only use it as a
> > discussion/review piece for now.
> 
> With 24 hours in testing an no problems encountered, I guess
> it's safe to declare this patchset as good to go if nobody has any
> objections against it.

Great, thanks for the testing, and all the work tracking this down;
committing for 4.7 (and the first two for stable).

--b.

> 
> > 
> > Oleg Drokin (3):
> >  nfsd: Always lock state exclusively.
> >  nfsd: Extend the mutex holding region around in nfsd4_process_open2()
> >  nfsd: Make init_open_stateid() a bit more whole
> > 
> > fs/nfsd/nfs4state.c | 67 +++++++++++++++++++++++++++++------------------------
> > fs/nfsd/state.h     |  2 +-
> > 2 files changed, 38 insertions(+), 31 deletions(-)
> > 
> > -- 
> > 2.7.4