xfs
[Top] [All Lists]

Re: [PATCH 05/19] SUNRPC: track whether a request is coming from a loop-

To: Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
Subject: Re: [PATCH 05/19] SUNRPC: track whether a request is coming from a loop-back interface.
From: NeilBrown <neilb@xxxxxxx>
Date: Thu, 17 Apr 2014 09:25:04 +1000
Cc: linux-mm@xxxxxxxxx, linux-nfs@xxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx, netdev@xxxxxxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20140416104702.264cde48@xxxxxxxxxxxxxxxxxxxxxxx>
References: <20140416033623.10604.69237.stgit@xxxxxxxxxxxxxx> <20140416040336.10604.14822.stgit@xxxxxxxxxxxxxx> <20140416104702.264cde48@xxxxxxxxxxxxxxxxxxxxxxx>
On Wed, 16 Apr 2014 10:47:02 -0400 Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
wrote:

> On Wed, 16 Apr 2014 14:03:36 +1000
> NeilBrown <neilb@xxxxxxx> wrote:
> 
> > If an incoming NFS request is coming from the local host, then
> > nfsd will need to perform some special handling.  So detect that
> > possibility and make the source visible in rq_local.
> > 
> > Signed-off-by: NeilBrown <neilb@xxxxxxx>
> > ---
> >  include/linux/sunrpc/svc.h      |    1 +
> >  include/linux/sunrpc/svc_xprt.h |    1 +
> >  net/sunrpc/svcsock.c            |   10 ++++++++++
> >  3 files changed, 12 insertions(+)
> > 
> > diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
> > index 04e763221246..a0dbbd1e00e9 100644
> > --- a/include/linux/sunrpc/svc.h
> > +++ b/include/linux/sunrpc/svc.h
> > @@ -254,6 +254,7 @@ struct svc_rqst {
> >     u32                     rq_prot;        /* IP protocol */
> >     unsigned short
> >                             rq_secure  : 1; /* secure port */
> > +   unsigned short          rq_local   : 1; /* local request */
> >  
> >     void *                  rq_argp;        /* decoded arguments */
> >     void *                  rq_resp;        /* xdr'd results */
> > diff --git a/include/linux/sunrpc/svc_xprt.h 
> > b/include/linux/sunrpc/svc_xprt.h
> > index b05963f09ebf..b99bdfb0fcf9 100644
> > --- a/include/linux/sunrpc/svc_xprt.h
> > +++ b/include/linux/sunrpc/svc_xprt.h
> > @@ -63,6 +63,7 @@ struct svc_xprt {
> >  #define    XPT_DETACHED    10              /* detached from tempsocks list 
> > */
> >  #define XPT_LISTENER       11              /* listening endpoint */
> >  #define XPT_CACHE_AUTH     12              /* cache auth info */
> > +#define XPT_LOCAL  13              /* connection from loopback interface */
> >  
> >     struct svc_serv         *xpt_server;    /* service for transport */
> >     atomic_t                xpt_reserved;   /* space on outq that is rsvd */
> > diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> > index b6e59f0a9475..193115fe968c 100644
> > --- a/net/sunrpc/svcsock.c
> > +++ b/net/sunrpc/svcsock.c
> > @@ -811,6 +811,7 @@ static struct svc_xprt *svc_tcp_accept(struct svc_xprt 
> > *xprt)
> >     struct socket   *newsock;
> >     struct svc_sock *newsvsk;
> >     int             err, slen;
> > +   struct dst_entry *dst;
> >     RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
> >  
> >     dprintk("svc: tcp_accept %p sock %p\n", svsk, sock);
> > @@ -867,6 +868,14 @@ static struct svc_xprt *svc_tcp_accept(struct svc_xprt 
> > *xprt)
> >     }
> >     svc_xprt_set_local(&newsvsk->sk_xprt, sin, slen);
> >  
> > +   clear_bit(XPT_LOCAL, &newsvsk->sk_xprt.xpt_flags);
> > +   rcu_read_lock();
> > +   dst = rcu_dereference(newsock->sk->sk_dst_cache);
> > +   if (dst && dst->dev &&
> > +       (dst->dev->features & NETIF_F_LOOPBACK))
> > +           set_bit(XPT_LOCAL, &newsvsk->sk_xprt.xpt_flags);
> > +   rcu_read_unlock();
> > +
> 
> In the use-case you describe, the client is unlikely to have mounted
> "localhost", but is more likely to be mounting a floating address in
> the cluster.
> 
> Will this flag end up being set in such a situation? It looks like
> NETIF_F_LOOPBACK isn't set on all adapters -- mostly on "lo" and a few
> others that support MAC-LOOPBACK.

I was hoping someone on net-dev would have commented if it didn't work - I
probably should have explicitly asked.

My testing certainly suggests that this works if I use any local address to
perform the mount, not just 127.0.0.1.
I don't know enough about routing and neighbours and the dst cache to be
certain but my shallow understanding was always that traffic to any local
address was pseudo-routed through the lo interface and would never get
anywhere near e.g. the eth0 interface.

Can any network people confirm?

Thanks,
NeilBrown


> 
> 
> >     if (serv->sv_stats)
> >             serv->sv_stats->nettcpconn++;
> >  
> > @@ -1112,6 +1121,7 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
> >  
> >     rqstp->rq_xprt_ctxt   = NULL;
> >     rqstp->rq_prot        = IPPROTO_TCP;
> > +   rqstp->rq_local       = !!test_bit(XPT_LOCAL, &svsk->sk_xprt.xpt_flags);
> >  
> >     p = (__be32 *)rqstp->rq_arg.head[0].iov_base;
> >     calldir = p[1];
> > 
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> 

Attachment: signature.asc
Description: PGP signature

<Prev in Thread] Current Thread [Next in Thread>