On Sat, 30 May 2015 13:55:44 -0400
Chuck Lever <chuck.lever@xxxxxxxxxx> wrote:
> Hi Jeff-
>
> On May 30, 2015, at 8:03 AM, Jeff Layton <jlayton@xxxxxxxxxxxxxxx> wrote:
>
> > Split xs_swapper into enable/disable functions and eliminate the
> > "enable" flag.
> >
> > Currently, it's racy if you have multiple swapon/swapoff operations
> > running in parallel over the same xprt. Also fix it so that we only
> > set it to a memalloc socket on a 0->1 transition and only clear it
> > on a 1->0 transition.
> >
> > Cc: Mel Gorman <mgorman@xxxxxxx>
> > Signed-off-by: Jeff Layton <jeff.layton@xxxxxxxxxxxxxxx>
> > ---
> > include/linux/sunrpc/xprt.h | 5 +++--
> > net/sunrpc/clnt.c | 4 ++--
> > net/sunrpc/xprtsock.c | 38 +++++++++++++++++++++++++-------------
> > 3 files changed, 30 insertions(+), 17 deletions(-)
> >
> > diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
> > index 8b93ef53df3c..26b1624128ec 100644
> > --- a/include/linux/sunrpc/xprt.h
> > +++ b/include/linux/sunrpc/xprt.h
> > @@ -180,7 +180,7 @@ struct rpc_xprt {
> > atomic_t num_reqs; /* total slots */
> > unsigned long state; /* transport state */
> > unsigned char resvport : 1; /* use a reserved port */
> > - unsigned int swapper; /* we're swapping over this
> > + atomic_t swapper; /* we're swapping over this
> > transport */
> > unsigned int bind_index; /* bind function index */
> >
> > @@ -345,7 +345,8 @@ void xprt_release_rqst_cong(struct rpc_task *task);
> > void xprt_disconnect_done(struct rpc_xprt *xprt);
> > void xprt_force_disconnect(struct rpc_xprt *xprt);
> > void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie);
> > -int xs_swapper(struct rpc_xprt *xprt, int enable);
> > +int xs_swapper_enable(struct rpc_xprt *xprt);
> > +void xs_swapper_disable(struct rpc_xprt *xprt);
> >
> > bool xprt_lock_connect(struct rpc_xprt *, struct rpc_task *, void *);
> > void xprt_unlock_connect(struct rpc_xprt *, void *);
> > diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
> > index 383cb778179f..804a75e71e84 100644
> > --- a/net/sunrpc/clnt.c
> > +++ b/net/sunrpc/clnt.c
> > @@ -2492,7 +2492,7 @@ retry:
> > goto retry;
> > }
> >
> > - ret = xs_swapper(xprt, 1);
> > + ret = xs_swapper_enable(xprt);
> > xprt_put(xprt);
> > }
> > return ret;
> > @@ -2519,7 +2519,7 @@ retry:
> > goto retry;
> > }
> >
> > - xs_swapper(xprt, 0);
> > + xs_swapper_disable(xprt);
> > xprt_put(xprt);
> > }
> > }
>
> Seems like xs_swapper() is specific to socket-based transports.
>
> There’s no struct sock * to use as an argument with RDMA, so xs_swapper()
> would probably dereference garbage if “swapon” was invoked on a
> proto=rdma mount point.
>
> Should these new functions be made members of the rpc_xprt_ops? The
> RDMA version of the methods can be made no-ops for now.
>
Oh, right -- rdma uses rpcrdma_xprt instead. Yeah, adding a new ops to
handle this sounds like the right thing to do. I'll roll that into the
v2 set.
Thanks!
Jeff
>
> > diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
> > index b29703996028..a2861bbfd319 100644
> > --- a/net/sunrpc/xprtsock.c
> > +++ b/net/sunrpc/xprtsock.c
> > @@ -1966,31 +1966,43 @@ static void xs_set_memalloc(struct rpc_xprt
> > *xprt) struct sock_xprt *transport = container_of(xprt, struct
> > sock_xprt, xprt);
> >
> > - if (xprt->swapper)
> > + if (atomic_read(&xprt->swapper))
> > sk_set_memalloc(transport->inet);
> > }
> >
> > /**
> > - * xs_swapper - Tag this transport as being used for swap.
> > + * xs_swapper_enable - Tag this transport as being used for swap.
> > * @xprt: transport to tag
> > - * @enable: enable/disable
> > *
> > + * Take a reference to this transport on behalf of the rpc_clnt,
> > and
> > + * optionally mark it for swapping if it wasn't already.
> > */
> > -int xs_swapper(struct rpc_xprt *xprt, int enable)
> > +int
> > +xs_swapper_enable(struct rpc_xprt *xprt)
> > {
> > struct sock_xprt *transport = container_of(xprt, struct
> > sock_xprt, xprt);
> > - int err = 0;
> >
> > - if (enable) {
> > - xprt->swapper++;
> > - xs_set_memalloc(xprt);
> > - } else if (xprt->swapper) {
> > - xprt->swapper--;
> > - sk_clear_memalloc(transport->inet);
> > - }
> > + if (atomic_inc_return(&xprt->swapper) == 1)
> > + sk_set_memalloc(transport->inet);
> > + return 0;
> > +}
> >
> > - return err;
> > +/**
> > + * xs_swapper_disable - Untag this transport as being used for
> > swap.
> > + * @xprt: transport to tag
> > + *
> > + * Drop a "swapper" reference to this xprt on behalf of the
> > rpc_clnt. If the
> > + * swapper refcount goes to 0, untag the socket as a memalloc
> > socket.
> > + */
> > +void
> > +xs_swapper_disable(struct rpc_xprt *xprt)
> > +{
> > + struct sock_xprt *transport = container_of(xprt, struct
> > sock_xprt,
> > + xprt);
> > +
> > + if (atomic_dec_and_test(&xprt->swapper))
> > + sk_clear_memalloc(transport->inet);
> > }
> > #else
> > static void xs_set_memalloc(struct rpc_xprt *xprt)
> > --
> > 2.4.1
> >
> > --
> > 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
>
> --
> Chuck Lever
> chuck[dot]lever[at]oracle[dot]com
>
>
>
--
Jeff Layton <jlayton@xxxxxxxxxxxxxxx>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href