On Thu, 8 Aug 2013 15:51:30 +0000 "Myklebust, Trond" <Trond.Myklebust@xxxxxxxxxx> wrote: > On Thu, 2013-08-08 at 12:59 +1000, NeilBrown wrote: > > Hi, > > I'm trying to track down a strange problem with state ids going bad > > (possibly linked to ntp changing the system time on the non-Linux server) > > and am still learning about how the state management works. > > > > But I've come across an error where I don't think there should be one. > > > > For whatever reason the client gets a BAD_STATEID on a file that it has a > > lock on. The open gets a write delegation so that when it runs > > nfs4_reclaim_locks(), nfs4_lock_reclaim aborts early without doing anything > > (it doesn't need to because there is a delegation). > > But the code below then checks that NFS_LOCK_INITIALIZED is set on all lock > > states. But it isn't because nfs4_clear_open_state cleared it and > > nfs4_lock_reclaim didn't bother setting it. > > > > So I think the error should only be printed if there is no delegated state, > > hence this patch. > > > > Does it look right, or have I misunderstood something? > > > > Hi Neil, > > That analysis looks correct. Can you resend the patch with an > appropriate signed-off-by and changelog entry? Thanks. I've resent separately. NeilBrown
Attachment:
signature.asc
Description: PGP signature
