Re: [PATCH 1/4] NFSv4: Fix delegation return in cases where we have to retry

[Trond Myklebust <trondmy@hammerspace.com>]

On Thu, 2021-05-20 at 14:29 -0400, J. Bruce Fields wrote:
> On Thu, May 20, 2021 at 12:38:59PM -0400, trondmy@kernel.org wrote:
> > From: Trond Myklebust <trond.myklebust@hammerspace.com>
> > 
> > If we're unable to immediately recover all locks because the server
> > is
> > unable to immediately service our reclaim calls, then we want to
> > retry
> > after we've finished servicing all the other asynchronous
> > delegation
> > returns on our queue.
> 
> So, there's a situation where the server can't service a reclaim
> until
> some other delegation is returned?  I'm not seeing how that happens.
> 

I can and I do... pNFS can be complicated...

> --b.
> 
> > 
> > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
> > ---
> >  fs/nfs/delegation.c | 71 +++++++++++++++++++++++++++++++++++------
> > ----
> >  fs/nfs/delegation.h |  1 +
> >  fs/nfs/nfs4_fs.h    |  1 +
> >  3 files changed, 58 insertions(+), 15 deletions(-)
> > 
> > diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
> > index e6ec6f09ac6e..7c45ac3c3b0b 100644
> > --- a/fs/nfs/delegation.c
> > +++ b/fs/nfs/delegation.c
> > @@ -75,6 +75,13 @@ void nfs_mark_delegation_referenced(struct
> > nfs_delegation *delegation)
> >         set_bit(NFS_DELEGATION_REFERENCED, &delegation->flags);
> >  }
> >  
> > +static void nfs_mark_return_delegation(struct nfs_server *server,
> > +                                      struct nfs_delegation
> > *delegation)
> > +{
> > +       set_bit(NFS_DELEGATION_RETURN, &delegation->flags);
> > +       set_bit(NFS4CLNT_DELEGRETURN, &server->nfs_client-
> > >cl_state);
> > +}
> > +
> >  static bool
> >  nfs4_is_valid_delegation(const struct nfs_delegation *delegation,
> >                 fmode_t flags)
> > @@ -293,6 +300,7 @@ nfs_start_delegation_return_locked(struct
> > nfs_inode *nfsi)
> >                 goto out;
> >         spin_lock(&delegation->lock);
> >         if (!test_and_set_bit(NFS_DELEGATION_RETURNING,
> > &delegation->flags)) {
> > +               clear_bit(NFS_DELEGATION_RETURN_DELAYED,
> > &delegation->flags);
> >                 /* Refcount matched in nfs_end_delegation_return()
> > */
> >                 ret = nfs_get_delegation(delegation);
> >         }
> > @@ -314,16 +322,17 @@ nfs_start_delegation_return(struct nfs_inode
> > *nfsi)
> >         return delegation;
> >  }
> >  
> > -static void
> > -nfs_abort_delegation_return(struct nfs_delegation *delegation,
> > -               struct nfs_client *clp)
> > +static void nfs_abort_delegation_return(struct nfs_delegation
> > *delegation,
> > +                                       struct nfs_client *clp, int
> > err)
> >  {
> >  
> >         spin_lock(&delegation->lock);
> >         clear_bit(NFS_DELEGATION_RETURNING, &delegation->flags);
> > -       set_bit(NFS_DELEGATION_RETURN, &delegation->flags);
> > +       if (err == -EAGAIN) {
> > +               set_bit(NFS_DELEGATION_RETURN_DELAYED, &delegation-
> > >flags);
> > +               set_bit(NFS4CLNT_DELEGRETURN_DELAYED, &clp-
> > >cl_state);
> > +       }
> >         spin_unlock(&delegation->lock);
> > -       set_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state);
> >  }
> >  
> >  static struct nfs_delegation *
> > @@ -539,7 +548,7 @@ static int nfs_end_delegation_return(struct
> > inode *inode, struct nfs_delegation
> >         } while (err == 0);
> >  
> >         if (err) {
> > -               nfs_abort_delegation_return(delegation, clp);
> > +               nfs_abort_delegation_return(delegation, clp, err);
> >                 goto out;
> >         }
> >  
> > @@ -568,6 +577,7 @@ static bool nfs_delegation_need_return(struct
> > nfs_delegation *delegation)
> >         if (ret)
> >                 clear_bit(NFS_DELEGATION_RETURN_IF_CLOSED,
> > &delegation->flags);
> >         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags)
> > ||
> > +           test_bit(NFS_DELEGATION_RETURN_DELAYED, &delegation-
> > >flags) ||
> >             test_bit(NFS_DELEGATION_REVOKED, &delegation->flags))
> >                 ret = false;
> >  
> > @@ -647,6 +657,38 @@ static int
> > nfs_server_return_marked_delegations(struct nfs_server *server,
> >         return err;
> >  }
> >  
> > +static bool nfs_server_clear_delayed_delegations(struct nfs_server
> > *server)
> > +{
> > +       struct nfs_delegation *d;
> > +       bool ret = false;
> > +
> > +       list_for_each_entry_rcu (d, &server->delegations,
> > super_list) {
> > +               if (!test_bit(NFS_DELEGATION_RETURN_DELAYED, &d-
> > >flags))
> > +                       continue;
> > +               nfs_mark_return_delegation(server, d);
> > +               clear_bit(NFS_DELEGATION_RETURN_DELAYED, &d-
> > >flags);
> > +               ret = true;
> > +       }
> > +       return ret;
> > +}
> > +
> > +static bool nfs_client_clear_delayed_delegations(struct nfs_client
> > *clp)
> > +{
> > +       struct nfs_server *server;
> > +       bool ret = false;
> > +
> > +       if (!test_and_clear_bit(NFS4CLNT_DELEGRETURN_DELAYED, &clp-
> > >cl_state))
> > +               goto out;
> > +       rcu_read_lock();
> > +       list_for_each_entry_rcu (server, &clp->cl_superblocks,
> > client_link) {
> > +               if (nfs_server_clear_delayed_delegations(server))
> > +                       ret = true;
> > +       }
> > +       rcu_read_unlock();
> > +out:
> > +       return ret;
> > +}
> > +
> >  /**
> >   * nfs_client_return_marked_delegations - return previously marked
> > delegations
> >   * @clp: nfs_client to process
> > @@ -659,8 +701,14 @@ static int
> > nfs_server_return_marked_delegations(struct nfs_server *server,
> >   */
> >  int nfs_client_return_marked_delegations(struct nfs_client *clp)
> >  {
> > -       return nfs_client_for_each_server(clp,
> > -                       nfs_server_return_marked_delegations,
> > NULL);
> > +       int err = nfs_client_for_each_server(
> > +               clp, nfs_server_return_marked_delegations, NULL);
> > +       if (err)
> > +               return err;
> > +       /* If a return was delayed, sleep to prevent hard looping
> > */
> > +       if (nfs_client_clear_delayed_delegations(clp))
> > +               ssleep(1);
> > +       return 0;
> >  }
> >  
> >  /**
> > @@ -775,13 +823,6 @@ static void
> > nfs_mark_return_if_closed_delegation(struct nfs_server *server,
> >         set_bit(NFS4CLNT_DELEGRETURN, &server->nfs_client-
> > >cl_state);
> >  }
> >  
> > -static void nfs_mark_return_delegation(struct nfs_server *server,
> > -               struct nfs_delegation *delegation)
> > -{
> > -       set_bit(NFS_DELEGATION_RETURN, &delegation->flags);
> > -       set_bit(NFS4CLNT_DELEGRETURN, &server->nfs_client-
> > >cl_state);
> > -}
> > -
> >  static bool nfs_server_mark_return_all_delegations(struct
> > nfs_server *server)
> >  {
> >         struct nfs_delegation *delegation;
> > diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h
> > index c19b4fd20781..1c378992b7c0 100644
> > --- a/fs/nfs/delegation.h
> > +++ b/fs/nfs/delegation.h
> > @@ -36,6 +36,7 @@ enum {
> >         NFS_DELEGATION_REVOKED,
> >         NFS_DELEGATION_TEST_EXPIRED,
> >         NFS_DELEGATION_INODE_FREEING,
> > +       NFS_DELEGATION_RETURN_DELAYED,
> >  };
> >  
> >  int nfs_inode_set_delegation(struct inode *inode, const struct
> > cred *cred,
> > diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
> > index 065cb04222a1..4c44322c2643 100644
> > --- a/fs/nfs/nfs4_fs.h
> > +++ b/fs/nfs/nfs4_fs.h
> > @@ -45,6 +45,7 @@ enum nfs4_client_state {
> >         NFS4CLNT_RECALL_RUNNING,
> >         NFS4CLNT_RECALL_ANY_LAYOUT_READ,
> >         NFS4CLNT_RECALL_ANY_LAYOUT_RW,
> > +       NFS4CLNT_DELEGRETURN_DELAYED,
> >  };
> >  
> >  #define NFS4_RENEW_TIMEOUT             0x01
> > -- 
> > 2.31.1

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com