[Devel] [PATCH] nfs: abort delegation in dying VE

[Andrei Vagin]

On Wed, Nov 15, 2017 at 11:41:00AM -0800, Andrei Vagin wrote:
> On Wed, Nov 15, 2017 at 07:55:02PM +0300, Kirill Tkhai wrote:
> > On 15.11.2017 19:50, Stanislav Kinsburskiy wrote:
> > > Don't queue delegation request, if ve init is exiting.
> > > 
> > > https://jira.sw.ru/browse/PSBM-77061
> > > 
> > > Inspired-by: Kirill Tkhai <ktkhai at virtuozzo.com>
> > > Signed-off-by: Stanislav Kinsburskiy <skinsbursky at virtuozzo.com>
> > > ---
> > >  fs/nfs/delegation.c |   16 +++++++++++++++-
> > >  1 file changed, 15 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
> > > index 66af497..2422754 100644
> > > --- a/fs/nfs/delegation.c
> > > +++ b/fs/nfs/delegation.c
> > > @@ -189,15 +189,29 @@ void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred,
> > >  	nfs_inode_set_delegation(inode, cred, res);
> > >  }
> > >  
> > > +static bool ve_abort_delegation(struct inode *inode)
> > > +{
> > > +	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
> > > +	struct rpc_xprt *xprt;
> > > +
> > > +	rcu_read_lock();
> > > +	xprt = rcu_dereference(clp->cl_rpcclient->cl_xprt);
> > > +	rcu_read_unlock();
> > > +
> > > +	return xprt->xprt_net->owner_ve->ve_netns == NULL;
> > 
> > Usually, memory pointed by a pointer, which was obtained via rcu, has to be used
> > in rcu_read_* brackets:
> > 
> > rcu_read_lock();
> > xprt = rcu_dereference(clp->cl_rpcclient->cl_xprt);
> > ret = (xprt->xprt_net->owner_ve->ve_netns == NULL);
> > rcu_read_unlock();
> > 
> > return ret;
> > 
> > If there is no an exception, we have to do something like above.
> 
> It is true when you want to dereference this pointer, otherwise I don't
> see any reason to take rcu_read_lock().

Oops, here is exectly this case. I didn't read the proposed code. Sorry.
> 
> > 
> > > +}
> > > +
> > >  static int nfs_do_return_delegation(struct inode *inode, struct nfs_delegation *delegation, int issync)
> > >  {
> > >  	int res = 0;
> > >  
> > > -	if (!test_bit(NFS_DELEGATION_REVOKED, &delegation->flags))
> > > +	if (!test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) &&
> > > +	    !ve_abort_delegation(inode)) {
> > >  		res = nfs4_proc_delegreturn(inode,
> > >  				delegation->cred,
> > >  				&delegation->stateid,
> > >  				issync);
> > > +	}
> > >  	nfs_free_delegation(delegation);
> > >  	return res;
> > >  }
> > > 
> > _______________________________________________
> > Devel mailing list
> > Devel at openvz.org
> > https://lists.openvz.org/mailman/listinfo/devel
> _______________________________________________
> Devel mailing list
> Devel at openvz.org
> https://lists.openvz.org/mailman/listinfo/devel