From: Chuck Lever <chuck.lever@oracle.com>
To: trond.myklebust@netapp.com
Cc: linux-nfs@vger.kernel.org
Subject: [PATCH 02/12] SUNRPC: Add API to acquire source address
Date: Mon, 14 Mar 2011 08:56:42 -0400 [thread overview]
Message-ID: <20110314125642.2413.4936.stgit@matisse.1015granger.net> (raw)
In-Reply-To: <20110314125204.2413.8870.stgit@matisse.1015granger.net>
NFSv4.0 clients have to send endpoint information for their callback
service to NFSv4.0 servers during SETCLIENTID negotiation.
Traditionally, user space provides the callback endpoint IP address
via the "clientaddr=" mount option.
During an NFSv4 migration event, it is possible that an FSID may be
migrated to a destination server that is accessible by a different NIC
than the source server. The client must update the callback endpoint
information on the destination server so that it can maintain leases
and allow delegation. Without a new "clientaddr=" option, though, the
kernel itself must construct an appropriate IP address for the
callback service.
Provide an API in the RPC client for upper layers to acquire a source
address for a remote. We copy the mechanism used by the mount
command: set up a connected UDP socket to the remote, then scrape the
source address off the socket.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
include/linux/sunrpc/clnt.h | 1
net/sunrpc/clnt.c | 132 +++++++++++++++++++++++++++++++++++++++++++
2 files changed, 133 insertions(+), 0 deletions(-)
diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index 70acded..ca48893 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -169,6 +169,7 @@ size_t rpc_max_payload(struct rpc_clnt *);
void rpc_force_rebind(struct rpc_clnt *);
size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t);
const char *rpc_peeraddr2str(struct rpc_clnt *, enum rpc_display_format_t);
+int rpc_localaddr(struct rpc_clnt *, struct sockaddr *, size_t *);
size_t rpc_ntop(const struct sockaddr *, char *, const size_t);
size_t rpc_pton(const char *, const size_t,
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 042e7da..49d8460 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -930,6 +930,138 @@ const char *rpc_peeraddr2str(struct rpc_clnt *clnt,
}
EXPORT_SYMBOL_GPL(rpc_peeraddr2str);
+/*
+ * Try a getsockname() on a connected datagram socket. Using a
+ * connected datagram socket prevents leaving a socket in TIME_WAIT.
+ * This conserves the ephemeral port number space.
+ *
+ * Returns zero and fills in "buf" and "bufsize" if successful;
+ * otherwise, a negative errno is returned.
+ */
+static int rpc_sockname(struct net *net, struct sockaddr *sap, size_t salen,
+ struct sockaddr *buf, size_t *bufsize)
+{
+ struct socket *sock;
+ int err, buflen;
+
+ err = __sock_create(net, sap->sa_family,
+ SOCK_DGRAM, IPPROTO_UDP, &sock, 1);
+ if (err < 0) {
+ dprintk("RPC: can't create UDP socket (%d)\n", -err);
+ goto out;
+ }
+
+ err = kernel_bind(sock, sap, salen);
+ if (err < 0) {
+ dprintk("RPC: can't bind UDP socket (%d)\n", -err);
+ goto out_release;
+ }
+
+ err = kernel_connect(sock, sap, salen, 0);
+ if (err < 0) {
+ dprintk("RPC: can't connect UDP socket (%d)\n", -err);
+ goto out_release;
+ }
+
+ err = kernel_getsockname(sock, buf, &buflen);
+ if (err < 0) {
+ dprintk("RPC: getsockname failed (%d)\n", -err);
+ goto out_release;
+ }
+
+ err = 0;
+ *bufsize = buflen;
+ if (buf->sa_family == AF_INET6) {
+ struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)buf;
+ sin6->sin6_scope_id = 0;
+ }
+
+out_release:
+ sock_release(sock);
+out:
+ return err;
+}
+
+/*
+ * Scraping a connected socket failed, so we don't have a useable
+ * local address. Fallback: generate an address that will prevent
+ * the server from calling us back.
+ *
+ * Returns zero and fills in "buf" and "bufsize" if successful;
+ * otherwise, a negative errno is returned.
+ */
+static int rpc_anyaddr(int family, struct sockaddr *buf, size_t *bufsize)
+{
+ static const struct sockaddr_in sin = {
+ .sin_family = AF_INET,
+ .sin_addr.s_addr = htonl(INADDR_ANY),
+ };
+ static const struct sockaddr_in6 sin6 = {
+ .sin6_family = AF_INET6,
+ .sin6_addr = IN6ADDR_ANY_INIT,
+ };
+ size_t buflen = *bufsize;
+
+ switch (family) {
+ case AF_INET:
+ if (buflen < sizeof(sin))
+ return -EINVAL;
+ memcpy(buf, &sin, sizeof(sin));
+ *bufsize = sizeof(sin);
+ break;
+ case AF_INET6:
+ if (buflen < sizeof(sin6))
+ return -EINVAL;
+ memcpy(buf, &sin6, sizeof(sin6));
+ *bufsize = sizeof(sin);
+ default:
+ return -EAFNOSUPPORT;
+ }
+ return 0;
+}
+
+/**
+ * rpc_localaddr - discover local endpoint address for an RPC client
+ * @clnt: RPC client structure
+ * @buf: target buffer
+ * @bufsize: IN: length of target buffer; OUT: length of local address
+ *
+ * Returns zero and fills in "buf" and "bufsize" if successful;
+ * otherwise, a negative errno is returned.
+ *
+ * This works even if the underlying transport is not currently connected,
+ * or if the upper layer never previously provided a source address.
+ *
+ * The results of this function call are transient: multiple calls in
+ * succession may give different results, depending on how local
+ * networking configuration changes over time.
+ */
+int rpc_localaddr(struct rpc_clnt *clnt, struct sockaddr *buf, size_t *bufsize)
+{
+ struct sockaddr_storage address;
+ struct sockaddr *sap = (struct sockaddr *)&address;
+ struct rpc_xprt *xprt;
+ struct net *net;
+ size_t salen;
+ int err;
+
+ rcu_read_lock();
+ xprt = rcu_dereference(clnt->cl_xprt);
+ salen = xprt->addrlen;
+ memcpy(sap, &xprt->addr, salen);
+ net = get_net(xprt->xprt_net);
+ rcu_read_unlock();
+
+ err = rpc_sockname(net, sap, salen, buf, bufsize);
+ put_net(net);
+ if (err < 0) {
+ err = rpc_anyaddr(sap->sa_family, buf, bufsize);
+ if (err < 0)
+ return err;
+ }
+ return 0;
+}
+
void
rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize)
{
next prev parent reply other threads:[~2011-03-14 12:56 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-03-14 12:56 [PATCH 00/12] Snapshot of client-side NFSv4 migration support (v2) Chuck Lever
2011-03-14 12:56 ` [PATCH 01/12] SUNRPC: Remove obsolete comment Chuck Lever
2011-03-14 12:56 ` Chuck Lever [this message]
2011-03-14 12:56 ` [PATCH 03/12] NFS: Add NFS4CLNT_UPDATE_CLIENTID Chuck Lever
2011-03-14 12:57 ` [PATCH 04/12] NFS: Add a client-side function to display file handles Chuck Lever
2011-03-14 12:57 ` [PATCH 05/12] NFS: Save root file handle in nfs_server Chuck Lever
2011-03-14 12:57 ` [PATCH 06/12] NFS: Introduce NFS_ATTR_FATTR_V4_LOCATIONS Chuck Lever
2011-03-14 12:57 ` [PATCH 07/12] NFS: Introduce nfs4_proc_get_mig_status() Chuck Lever
2011-03-14 12:57 ` [PATCH 08/12] NFS: Add functions to swap transports during migration recovery Chuck Lever
2011-03-14 12:57 ` [PATCH 09/12] NFS: Add basic migration support to state manager thread Chuck Lever
2011-03-14 12:57 ` [PATCH 10/12] NFS: Remove "const" from "struct nfs_server *" fields Chuck Lever
2011-03-14 12:58 ` [PATCH 11/12] NFS: Add migration recovery callouts in nfs4proc.c Chuck Lever
2011-03-14 12:58 ` [PATCH 12/12] NFS: Implement support for NFS4ERR_LEASE_MOVED Chuck Lever
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20110314125642.2413.4936.stgit@matisse.1015granger.net \
--to=chuck.lever@oracle.com \
--cc=linux-nfs@vger.kernel.org \
--cc=trond.myklebust@netapp.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).