aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--fs/lockd/clntlock.c51
-rw-r--r--fs/lockd/mon.c8
-rw-r--r--fs/lockd/svc.c42
-rw-r--r--fs/nfs/callback.c31
-rw-r--r--fs/nfs/callback.h1
-rw-r--r--fs/nfs/client.c116
-rw-r--r--fs/nfs/dir.c9
-rw-r--r--fs/nfs/file.c32
-rw-r--r--fs/nfs/getroot.c4
-rw-r--r--fs/nfs/inode.c309
-rw-r--r--fs/nfs/internal.h4
-rw-r--r--fs/nfs/nfs2xdr.c9
-rw-r--r--fs/nfs/nfs3proc.c1
-rw-r--r--fs/nfs/nfs3xdr.c37
-rw-r--r--fs/nfs/nfs4proc.c47
-rw-r--r--fs/nfs/nfs4state.c10
-rw-r--r--fs/nfs/nfs4xdr.c213
-rw-r--r--fs/nfs/pagelist.c11
-rw-r--r--fs/nfs/proc.c1
-rw-r--r--fs/nfs/super.c4
-rw-r--r--fs/nfs/write.c53
-rw-r--r--fs/nfsd/nfsctl.c6
-rw-r--r--fs/nfsd/nfssvc.c5
-rw-r--r--include/linux/nfs_fs.h4
-rw-r--r--include/linux/nfs_fs_sb.h5
-rw-r--r--include/linux/nfs_xdr.h59
-rw-r--r--include/linux/sunrpc/svc.h9
-rw-r--r--include/linux/sunrpc/svc_xprt.h52
-rw-r--r--include/linux/sunrpc/xprt.h2
-rw-r--r--net/sunrpc/Kconfig22
-rw-r--r--net/sunrpc/clnt.c48
-rw-r--r--net/sunrpc/rpcb_clnt.c103
-rw-r--r--net/sunrpc/svc.c158
-rw-r--r--net/sunrpc/svc_xprt.c31
-rw-r--r--net/sunrpc/svcsock.c40
-rw-r--r--net/sunrpc/xprt.c89
-rw-r--r--net/sunrpc/xprtrdma/rpc_rdma.c26
-rw-r--r--net/sunrpc/xprtrdma/svc_rdma_sendto.c8
-rw-r--r--net/sunrpc/xprtsock.c363
39 files changed, 1178 insertions, 845 deletions
diff --git a/fs/lockd/clntlock.c b/fs/lockd/clntlock.c
index aedc47a264c1..1f3b0fc0d351 100644
--- a/fs/lockd/clntlock.c
+++ b/fs/lockd/clntlock.c
@@ -139,55 +139,6 @@ int nlmclnt_block(struct nlm_wait *block, struct nlm_rqst *req, long timeout)
139 return 0; 139 return 0;
140} 140}
141 141
142#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
143static const struct in6_addr *nlmclnt_map_v4addr(const struct sockaddr *sap,
144 struct in6_addr *addr_mapped)
145{
146 const struct sockaddr_in *sin = (const struct sockaddr_in *)sap;
147
148 switch (sap->sa_family) {
149 case AF_INET6:
150 return &((const struct sockaddr_in6 *)sap)->sin6_addr;
151 case AF_INET:
152 ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, addr_mapped);
153 return addr_mapped;
154 }
155
156 return NULL;
157}
158
159/*
160 * If lockd is using a PF_INET6 listener, all incoming requests appear
161 * to come from AF_INET6 remotes. The address of AF_INET remotes are
162 * mapped to AF_INET6 automatically by the network layer. In case the
163 * user passed an AF_INET server address at mount time, ensure both
164 * addresses are AF_INET6 before comparing them.
165 */
166static int nlmclnt_cmp_addr(const struct nlm_host *host,
167 const struct sockaddr *sap)
168{
169 const struct in6_addr *addr1;
170 const struct in6_addr *addr2;
171 struct in6_addr addr1_mapped;
172 struct in6_addr addr2_mapped;
173
174 addr1 = nlmclnt_map_v4addr(nlm_addr(host), &addr1_mapped);
175 if (likely(addr1 != NULL)) {
176 addr2 = nlmclnt_map_v4addr(sap, &addr2_mapped);
177 if (likely(addr2 != NULL))
178 return ipv6_addr_equal(addr1, addr2);
179 }
180
181 return 0;
182}
183#else /* !(CONFIG_IPV6 || CONFIG_IPV6_MODULE) */
184static int nlmclnt_cmp_addr(const struct nlm_host *host,
185 const struct sockaddr *sap)
186{
187 return nlm_cmp_addr(nlm_addr(host), sap);
188}
189#endif /* !(CONFIG_IPV6 || CONFIG_IPV6_MODULE) */
190
191/* 142/*
192 * The server lockd has called us back to tell us the lock was granted 143 * The server lockd has called us back to tell us the lock was granted
193 */ 144 */
@@ -215,7 +166,7 @@ __be32 nlmclnt_grant(const struct sockaddr *addr, const struct nlm_lock *lock)
215 */ 166 */
216 if (fl_blocked->fl_u.nfs_fl.owner->pid != lock->svid) 167 if (fl_blocked->fl_u.nfs_fl.owner->pid != lock->svid)
217 continue; 168 continue;
218 if (!nlmclnt_cmp_addr(block->b_host, addr)) 169 if (!nlm_cmp_addr(nlm_addr(block->b_host), addr))
219 continue; 170 continue;
220 if (nfs_compare_fh(NFS_FH(fl_blocked->fl_file->f_path.dentry->d_inode) ,fh) != 0) 171 if (nfs_compare_fh(NFS_FH(fl_blocked->fl_file->f_path.dentry->d_inode) ,fh) != 0)
221 continue; 172 continue;
diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c
index 5e2c4d5ac827..6d5d4a4169e5 100644
--- a/fs/lockd/mon.c
+++ b/fs/lockd/mon.c
@@ -16,6 +16,8 @@
16#include <linux/sunrpc/svc.h> 16#include <linux/sunrpc/svc.h>
17#include <linux/lockd/lockd.h> 17#include <linux/lockd/lockd.h>
18 18
19#include <asm/unaligned.h>
20
19#define NLMDBG_FACILITY NLMDBG_MONITOR 21#define NLMDBG_FACILITY NLMDBG_MONITOR
20#define NSM_PROGRAM 100024 22#define NSM_PROGRAM 100024
21#define NSM_VERSION 1 23#define NSM_VERSION 1
@@ -274,10 +276,12 @@ static void nsm_init_private(struct nsm_handle *nsm)
274{ 276{
275 u64 *p = (u64 *)&nsm->sm_priv.data; 277 u64 *p = (u64 *)&nsm->sm_priv.data;
276 struct timespec ts; 278 struct timespec ts;
279 s64 ns;
277 280
278 ktime_get_ts(&ts); 281 ktime_get_ts(&ts);
279 *p++ = timespec_to_ns(&ts); 282 ns = timespec_to_ns(&ts);
280 *p = (unsigned long)nsm; 283 put_unaligned(ns, p);
284 put_unaligned((unsigned long)nsm, p + 1);
281} 285}
282 286
283static struct nsm_handle *nsm_create_handle(const struct sockaddr *sap, 287static struct nsm_handle *nsm_create_handle(const struct sockaddr *sap,
diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index 64f1c31b5853..abf83881f68a 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -53,17 +53,6 @@ static struct svc_rqst *nlmsvc_rqst;
53unsigned long nlmsvc_timeout; 53unsigned long nlmsvc_timeout;
54 54
55/* 55/*
56 * If the kernel has IPv6 support available, always listen for
57 * both AF_INET and AF_INET6 requests.
58 */
59#if (defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)) && \
60 defined(CONFIG_SUNRPC_REGISTER_V4)
61static const sa_family_t nlmsvc_family = AF_INET6;
62#else /* (CONFIG_IPV6 || CONFIG_IPV6_MODULE) && CONFIG_SUNRPC_REGISTER_V4 */
63static const sa_family_t nlmsvc_family = AF_INET;
64#endif /* (CONFIG_IPV6 || CONFIG_IPV6_MODULE) && CONFIG_SUNRPC_REGISTER_V4 */
65
66/*
67 * These can be set at insmod time (useful for NFS as root filesystem), 56 * These can be set at insmod time (useful for NFS as root filesystem),
68 * and also changed through the sysctl interface. -- Jamie Lokier, Aug 2003 57 * and also changed through the sysctl interface. -- Jamie Lokier, Aug 2003
69 */ 58 */
@@ -204,19 +193,30 @@ lockd(void *vrqstp)
204 return 0; 193 return 0;
205} 194}
206 195
207static int create_lockd_listener(struct svc_serv *serv, char *name, 196static int create_lockd_listener(struct svc_serv *serv, const char *name,
208 unsigned short port) 197 const int family, const unsigned short port)
209{ 198{
210 struct svc_xprt *xprt; 199 struct svc_xprt *xprt;
211 200
212 xprt = svc_find_xprt(serv, name, 0, 0); 201 xprt = svc_find_xprt(serv, name, family, 0);
213 if (xprt == NULL) 202 if (xprt == NULL)
214 return svc_create_xprt(serv, name, port, SVC_SOCK_DEFAULTS); 203 return svc_create_xprt(serv, name, family, port,
215 204 SVC_SOCK_DEFAULTS);
216 svc_xprt_put(xprt); 205 svc_xprt_put(xprt);
217 return 0; 206 return 0;
218} 207}
219 208
209static int create_lockd_family(struct svc_serv *serv, const int family)
210{
211 int err;
212
213 err = create_lockd_listener(serv, "udp", family, nlm_udpport);
214 if (err < 0)
215 return err;
216
217 return create_lockd_listener(serv, "tcp", family, nlm_tcpport);
218}
219
220/* 220/*
221 * Ensure there are active UDP and TCP listeners for lockd. 221 * Ensure there are active UDP and TCP listeners for lockd.
222 * 222 *
@@ -232,13 +232,15 @@ static int make_socks(struct svc_serv *serv)
232 static int warned; 232 static int warned;
233 int err; 233 int err;
234 234
235 err = create_lockd_listener(serv, "udp", nlm_udpport); 235 err = create_lockd_family(serv, PF_INET);
236 if (err < 0) 236 if (err < 0)
237 goto out_err; 237 goto out_err;
238 238
239 err = create_lockd_listener(serv, "tcp", nlm_tcpport); 239#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
240 if (err < 0) 240 err = create_lockd_family(serv, PF_INET6);
241 if (err < 0 && err != -EAFNOSUPPORT)
241 goto out_err; 242 goto out_err;
243#endif /* CONFIG_IPV6 || CONFIG_IPV6_MODULE */
242 244
243 warned = 0; 245 warned = 0;
244 return 0; 246 return 0;
@@ -274,7 +276,7 @@ int lockd_up(void)
274 "lockd_up: no pid, %d users??\n", nlmsvc_users); 276 "lockd_up: no pid, %d users??\n", nlmsvc_users);
275 277
276 error = -ENOMEM; 278 error = -ENOMEM;
277 serv = svc_create(&nlmsvc_program, LOCKD_BUFSIZE, nlmsvc_family, NULL); 279 serv = svc_create(&nlmsvc_program, LOCKD_BUFSIZE, NULL);
278 if (!serv) { 280 if (!serv) {
279 printk(KERN_WARNING "lockd_up: create service failed\n"); 281 printk(KERN_WARNING "lockd_up: create service failed\n");
280 goto out; 282 goto out;
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index 3e634f2a1083..a886e692ddd0 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -38,19 +38,10 @@ static struct svc_program nfs4_callback_program;
38 38
39unsigned int nfs_callback_set_tcpport; 39unsigned int nfs_callback_set_tcpport;
40unsigned short nfs_callback_tcpport; 40unsigned short nfs_callback_tcpport;
41unsigned short nfs_callback_tcpport6;
41static const int nfs_set_port_min = 0; 42static const int nfs_set_port_min = 0;
42static const int nfs_set_port_max = 65535; 43static const int nfs_set_port_max = 65535;
43 44
44/*
45 * If the kernel has IPv6 support available, always listen for
46 * both AF_INET and AF_INET6 requests.
47 */
48#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
49static const sa_family_t nfs_callback_family = AF_INET6;
50#else
51static const sa_family_t nfs_callback_family = AF_INET;
52#endif
53
54static int param_set_port(const char *val, struct kernel_param *kp) 45static int param_set_port(const char *val, struct kernel_param *kp)
55{ 46{
56 char *endp; 47 char *endp;
@@ -116,19 +107,29 @@ int nfs_callback_up(void)
116 mutex_lock(&nfs_callback_mutex); 107 mutex_lock(&nfs_callback_mutex);
117 if (nfs_callback_info.users++ || nfs_callback_info.task != NULL) 108 if (nfs_callback_info.users++ || nfs_callback_info.task != NULL)
118 goto out; 109 goto out;
119 serv = svc_create(&nfs4_callback_program, NFS4_CALLBACK_BUFSIZE, 110 serv = svc_create(&nfs4_callback_program, NFS4_CALLBACK_BUFSIZE, NULL);
120 nfs_callback_family, NULL);
121 ret = -ENOMEM; 111 ret = -ENOMEM;
122 if (!serv) 112 if (!serv)
123 goto out_err; 113 goto out_err;
124 114
125 ret = svc_create_xprt(serv, "tcp", nfs_callback_set_tcpport, 115 ret = svc_create_xprt(serv, "tcp", PF_INET,
126 SVC_SOCK_ANONYMOUS); 116 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
127 if (ret <= 0) 117 if (ret <= 0)
128 goto out_err; 118 goto out_err;
129 nfs_callback_tcpport = ret; 119 nfs_callback_tcpport = ret;
130 dprintk("NFS: Callback listener port = %u (af %u)\n", 120 dprintk("NFS: Callback listener port = %u (af %u)\n",
131 nfs_callback_tcpport, nfs_callback_family); 121 nfs_callback_tcpport, PF_INET);
122
123#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
124 ret = svc_create_xprt(serv, "tcp", PF_INET6,
125 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
126 if (ret > 0) {
127 nfs_callback_tcpport6 = ret;
128 dprintk("NFS: Callback listener port = %u (af %u)\n",
129 nfs_callback_tcpport6, PF_INET6);
130 } else if (ret != -EAFNOSUPPORT)
131 goto out_err;
132#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
132 133
133 nfs_callback_info.rqst = svc_prepare_thread(serv, &serv->sv_pools[0]); 134 nfs_callback_info.rqst = svc_prepare_thread(serv, &serv->sv_pools[0]);
134 if (IS_ERR(nfs_callback_info.rqst)) { 135 if (IS_ERR(nfs_callback_info.rqst)) {
diff --git a/fs/nfs/callback.h b/fs/nfs/callback.h
index bb25d2135ff1..e110e286a262 100644
--- a/fs/nfs/callback.h
+++ b/fs/nfs/callback.h
@@ -72,5 +72,6 @@ extern void nfs_callback_down(void);
72 72
73extern unsigned int nfs_callback_set_tcpport; 73extern unsigned int nfs_callback_set_tcpport;
74extern unsigned short nfs_callback_tcpport; 74extern unsigned short nfs_callback_tcpport;
75extern unsigned short nfs_callback_tcpport6;
75 76
76#endif /* __LINUX_FS_NFS_CALLBACK_H */ 77#endif /* __LINUX_FS_NFS_CALLBACK_H */
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 2277421656e7..aba38017bdef 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -224,38 +224,6 @@ void nfs_put_client(struct nfs_client *clp)
224} 224}
225 225
226#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 226#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
227static const struct in6_addr *nfs_map_ipv4_addr(const struct sockaddr *sa, struct in6_addr *addr_mapped)
228{
229 switch (sa->sa_family) {
230 default:
231 return NULL;
232 case AF_INET6:
233 return &((const struct sockaddr_in6 *)sa)->sin6_addr;
234 break;
235 case AF_INET:
236 ipv6_addr_set_v4mapped(((const struct sockaddr_in *)sa)->sin_addr.s_addr,
237 addr_mapped);
238 return addr_mapped;
239 }
240}
241
242static int nfs_sockaddr_match_ipaddr(const struct sockaddr *sa1,
243 const struct sockaddr *sa2)
244{
245 const struct in6_addr *addr1;
246 const struct in6_addr *addr2;
247 struct in6_addr addr1_mapped;
248 struct in6_addr addr2_mapped;
249
250 addr1 = nfs_map_ipv4_addr(sa1, &addr1_mapped);
251 if (likely(addr1 != NULL)) {
252 addr2 = nfs_map_ipv4_addr(sa2, &addr2_mapped);
253 if (likely(addr2 != NULL))
254 return ipv6_addr_equal(addr1, addr2);
255 }
256 return 0;
257}
258
259/* 227/*
260 * Test if two ip6 socket addresses refer to the same socket by 228 * Test if two ip6 socket addresses refer to the same socket by
261 * comparing relevant fields. The padding bytes specifically, are not 229 * comparing relevant fields. The padding bytes specifically, are not
@@ -267,38 +235,21 @@ static int nfs_sockaddr_match_ipaddr(const struct sockaddr *sa1,
267 * 235 *
268 * The caller should ensure both socket addresses are AF_INET6. 236 * The caller should ensure both socket addresses are AF_INET6.
269 */ 237 */
270static int nfs_sockaddr_cmp_ip6(const struct sockaddr *sa1, 238static int nfs_sockaddr_match_ipaddr6(const struct sockaddr *sa1,
271 const struct sockaddr *sa2) 239 const struct sockaddr *sa2)
272{ 240{
273 const struct sockaddr_in6 *saddr1 = (const struct sockaddr_in6 *)sa1; 241 const struct sockaddr_in6 *sin1 = (const struct sockaddr_in6 *)sa1;
274 const struct sockaddr_in6 *saddr2 = (const struct sockaddr_in6 *)sa2; 242 const struct sockaddr_in6 *sin2 = (const struct sockaddr_in6 *)sa2;
275 243
276 if (!ipv6_addr_equal(&saddr1->sin6_addr, 244 if (ipv6_addr_scope(&sin1->sin6_addr) == IPV6_ADDR_SCOPE_LINKLOCAL &&
277 &saddr1->sin6_addr)) 245 sin1->sin6_scope_id != sin2->sin6_scope_id)
278 return 0; 246 return 0;
279 if (ipv6_addr_scope(&saddr1->sin6_addr) == IPV6_ADDR_SCOPE_LINKLOCAL &&
280 saddr1->sin6_scope_id != saddr2->sin6_scope_id)
281 return 0;
282 return saddr1->sin6_port == saddr2->sin6_port;
283}
284#else
285static int nfs_sockaddr_match_ipaddr4(const struct sockaddr_in *sa1,
286 const struct sockaddr_in *sa2)
287{
288 return sa1->sin_addr.s_addr == sa2->sin_addr.s_addr;
289}
290 247
291static int nfs_sockaddr_match_ipaddr(const struct sockaddr *sa1, 248 return ipv6_addr_equal(&sin1->sin6_addr, &sin1->sin6_addr);
292 const struct sockaddr *sa2)
293{
294 if (unlikely(sa1->sa_family != AF_INET || sa2->sa_family != AF_INET))
295 return 0;
296 return nfs_sockaddr_match_ipaddr4((const struct sockaddr_in *)sa1,
297 (const struct sockaddr_in *)sa2);
298} 249}
299 250#else /* !defined(CONFIG_IPV6) && !defined(CONFIG_IPV6_MODULE) */
300static int nfs_sockaddr_cmp_ip6(const struct sockaddr * sa1, 251static int nfs_sockaddr_match_ipaddr6(const struct sockaddr *sa1,
301 const struct sockaddr * sa2) 252 const struct sockaddr *sa2)
302{ 253{
303 return 0; 254 return 0;
304} 255}
@@ -311,20 +262,57 @@ static int nfs_sockaddr_cmp_ip6(const struct sockaddr * sa1,
311 * 262 *
312 * The caller should ensure both socket addresses are AF_INET. 263 * The caller should ensure both socket addresses are AF_INET.
313 */ 264 */
265static int nfs_sockaddr_match_ipaddr4(const struct sockaddr *sa1,
266 const struct sockaddr *sa2)
267{
268 const struct sockaddr_in *sin1 = (const struct sockaddr_in *)sa1;
269 const struct sockaddr_in *sin2 = (const struct sockaddr_in *)sa2;
270
271 return sin1->sin_addr.s_addr == sin2->sin_addr.s_addr;
272}
273
274static int nfs_sockaddr_cmp_ip6(const struct sockaddr *sa1,
275 const struct sockaddr *sa2)
276{
277 const struct sockaddr_in6 *sin1 = (const struct sockaddr_in6 *)sa1;
278 const struct sockaddr_in6 *sin2 = (const struct sockaddr_in6 *)sa2;
279
280 return nfs_sockaddr_match_ipaddr6(sa1, sa2) &&
281 (sin1->sin6_port == sin2->sin6_port);
282}
283
314static int nfs_sockaddr_cmp_ip4(const struct sockaddr *sa1, 284static int nfs_sockaddr_cmp_ip4(const struct sockaddr *sa1,
315 const struct sockaddr *sa2) 285 const struct sockaddr *sa2)
316{ 286{
317 const struct sockaddr_in *saddr1 = (const struct sockaddr_in *)sa1; 287 const struct sockaddr_in *sin1 = (const struct sockaddr_in *)sa1;
318 const struct sockaddr_in *saddr2 = (const struct sockaddr_in *)sa2; 288 const struct sockaddr_in *sin2 = (const struct sockaddr_in *)sa2;
319 289
320 if (saddr1->sin_addr.s_addr != saddr2->sin_addr.s_addr) 290 return nfs_sockaddr_match_ipaddr4(sa1, sa2) &&
291 (sin1->sin_port == sin2->sin_port);
292}
293
294/*
295 * Test if two socket addresses represent the same actual socket,
296 * by comparing (only) relevant fields, excluding the port number.
297 */
298static int nfs_sockaddr_match_ipaddr(const struct sockaddr *sa1,
299 const struct sockaddr *sa2)
300{
301 if (sa1->sa_family != sa2->sa_family)
321 return 0; 302 return 0;
322 return saddr1->sin_port == saddr2->sin_port; 303
304 switch (sa1->sa_family) {
305 case AF_INET:
306 return nfs_sockaddr_match_ipaddr4(sa1, sa2);
307 case AF_INET6:
308 return nfs_sockaddr_match_ipaddr6(sa1, sa2);
309 }
310 return 0;
323} 311}
324 312
325/* 313/*
326 * Test if two socket addresses represent the same actual socket, 314 * Test if two socket addresses represent the same actual socket,
327 * by comparing (only) relevant fields. 315 * by comparing (only) relevant fields, including the port number.
328 */ 316 */
329static int nfs_sockaddr_cmp(const struct sockaddr *sa1, 317static int nfs_sockaddr_cmp(const struct sockaddr *sa1,
330 const struct sockaddr *sa2) 318 const struct sockaddr *sa2)
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 78bf72fc1db3..370b190a09d1 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -1624,8 +1624,7 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry,
1624 } else if (atomic_read(&new_dentry->d_count) > 1) 1624 } else if (atomic_read(&new_dentry->d_count) > 1)
1625 /* dentry still busy? */ 1625 /* dentry still busy? */
1626 goto out; 1626 goto out;
1627 } else 1627 }
1628 nfs_drop_nlink(new_inode);
1629 1628
1630go_ahead: 1629go_ahead:
1631 /* 1630 /*
@@ -1638,10 +1637,8 @@ go_ahead:
1638 } 1637 }
1639 nfs_inode_return_delegation(old_inode); 1638 nfs_inode_return_delegation(old_inode);
1640 1639
1641 if (new_inode != NULL) { 1640 if (new_inode != NULL)
1642 nfs_inode_return_delegation(new_inode); 1641 nfs_inode_return_delegation(new_inode);
1643 d_delete(new_dentry);
1644 }
1645 1642
1646 error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name, 1643 error = NFS_PROTO(old_dir)->rename(old_dir, &old_dentry->d_name,
1647 new_dir, &new_dentry->d_name); 1644 new_dir, &new_dentry->d_name);
@@ -1650,6 +1647,8 @@ out:
1650 if (rehash) 1647 if (rehash)
1651 d_rehash(rehash); 1648 d_rehash(rehash);
1652 if (!error) { 1649 if (!error) {
1650 if (new_inode != NULL)
1651 nfs_drop_nlink(new_inode);
1653 d_move(old_dentry, new_dentry); 1652 d_move(old_dentry, new_dentry);
1654 nfs_set_verifier(new_dentry, 1653 nfs_set_verifier(new_dentry,
1655 nfs_save_change_attribute(new_dir)); 1654 nfs_save_change_attribute(new_dir));
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index cec79392e4ba..0abf3f331f56 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -64,11 +64,7 @@ const struct file_operations nfs_file_operations = {
64 .write = do_sync_write, 64 .write = do_sync_write,
65 .aio_read = nfs_file_read, 65 .aio_read = nfs_file_read,
66 .aio_write = nfs_file_write, 66 .aio_write = nfs_file_write,
67#ifdef CONFIG_MMU
68 .mmap = nfs_file_mmap, 67 .mmap = nfs_file_mmap,
69#else
70 .mmap = generic_file_mmap,
71#endif
72 .open = nfs_file_open, 68 .open = nfs_file_open,
73 .flush = nfs_file_flush, 69 .flush = nfs_file_flush,
74 .release = nfs_file_release, 70 .release = nfs_file_release,
@@ -141,9 +137,6 @@ nfs_file_release(struct inode *inode, struct file *filp)
141 dentry->d_parent->d_name.name, 137 dentry->d_parent->d_name.name,
142 dentry->d_name.name); 138 dentry->d_name.name);
143 139
144 /* Ensure that dirty pages are flushed out with the right creds */
145 if (filp->f_mode & FMODE_WRITE)
146 nfs_wb_all(dentry->d_inode);
147 nfs_inc_stats(inode, NFSIOS_VFSRELEASE); 140 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
148 return nfs_release(inode, filp); 141 return nfs_release(inode, filp);
149} 142}
@@ -235,7 +228,6 @@ nfs_file_flush(struct file *file, fl_owner_t id)
235 struct nfs_open_context *ctx = nfs_file_open_context(file); 228 struct nfs_open_context *ctx = nfs_file_open_context(file);
236 struct dentry *dentry = file->f_path.dentry; 229 struct dentry *dentry = file->f_path.dentry;
237 struct inode *inode = dentry->d_inode; 230 struct inode *inode = dentry->d_inode;
238 int status;
239 231
240 dprintk("NFS: flush(%s/%s)\n", 232 dprintk("NFS: flush(%s/%s)\n",
241 dentry->d_parent->d_name.name, 233 dentry->d_parent->d_name.name,
@@ -245,11 +237,8 @@ nfs_file_flush(struct file *file, fl_owner_t id)
245 return 0; 237 return 0;
246 nfs_inc_stats(inode, NFSIOS_VFSFLUSH); 238 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
247 239
248 /* Ensure that data+attribute caches are up to date after close() */ 240 /* Flush writes to the server and return any errors */
249 status = nfs_do_fsync(ctx, inode); 241 return nfs_do_fsync(ctx, inode);
250 if (!status)
251 nfs_revalidate_inode(NFS_SERVER(inode), inode);
252 return status;
253} 242}
254 243
255static ssize_t 244static ssize_t
@@ -304,11 +293,13 @@ nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
304 dprintk("NFS: mmap(%s/%s)\n", 293 dprintk("NFS: mmap(%s/%s)\n",
305 dentry->d_parent->d_name.name, dentry->d_name.name); 294 dentry->d_parent->d_name.name, dentry->d_name.name);
306 295
307 status = nfs_revalidate_mapping(inode, file->f_mapping); 296 /* Note: generic_file_mmap() returns ENOSYS on nommu systems
297 * so we call that before revalidating the mapping
298 */
299 status = generic_file_mmap(file, vma);
308 if (!status) { 300 if (!status) {
309 vma->vm_ops = &nfs_file_vm_ops; 301 vma->vm_ops = &nfs_file_vm_ops;
310 vma->vm_flags |= VM_CAN_NONLINEAR; 302 status = nfs_revalidate_mapping(inode, file->f_mapping);
311 file_accessed(file);
312 } 303 }
313 return status; 304 return status;
314} 305}
@@ -354,6 +345,15 @@ static int nfs_write_begin(struct file *file, struct address_space *mapping,
354 file->f_path.dentry->d_name.name, 345 file->f_path.dentry->d_name.name,
355 mapping->host->i_ino, len, (long long) pos); 346 mapping->host->i_ino, len, (long long) pos);
356 347
348 /*
349 * Prevent starvation issues if someone is doing a consistency
350 * sync-to-disk
351 */
352 ret = wait_on_bit(&NFS_I(mapping->host)->flags, NFS_INO_FLUSHING,
353 nfs_wait_bit_killable, TASK_KILLABLE);
354 if (ret)
355 return ret;
356
357 page = grab_cache_page_write_begin(mapping, index, flags); 357 page = grab_cache_page_write_begin(mapping, index, flags);
358 if (!page) 358 if (!page)
359 return -ENOMEM; 359 return -ENOMEM;
diff --git a/fs/nfs/getroot.c b/fs/nfs/getroot.c
index b7c9b2df1f29..46177cb87064 100644
--- a/fs/nfs/getroot.c
+++ b/fs/nfs/getroot.c
@@ -156,7 +156,7 @@ int nfs4_path_walk(struct nfs_server *server,
156 return ret; 156 return ret;
157 } 157 }
158 158
159 if (fattr.type != NFDIR) { 159 if (!S_ISDIR(fattr.mode)) {
160 printk(KERN_ERR "nfs4_get_root:" 160 printk(KERN_ERR "nfs4_get_root:"
161 " getroot encountered non-directory\n"); 161 " getroot encountered non-directory\n");
162 return -ENOTDIR; 162 return -ENOTDIR;
@@ -213,7 +213,7 @@ eat_dot_dir:
213 return ret; 213 return ret;
214 } 214 }
215 215
216 if (fattr.type != NFDIR) { 216 if (!S_ISDIR(fattr.mode)) {
217 printk(KERN_ERR "nfs4_get_root:" 217 printk(KERN_ERR "nfs4_get_root:"
218 " lookupfh encountered non-directory\n"); 218 " lookupfh encountered non-directory\n");
219 return -ENOTDIR; 219 return -ENOTDIR;
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 0c381686171e..a834d1d850b7 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -66,6 +66,18 @@ nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
66} 66}
67 67
68/** 68/**
69 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
70 * @word: long word containing the bit lock
71 */
72int nfs_wait_bit_killable(void *word)
73{
74 if (fatal_signal_pending(current))
75 return -ERESTARTSYS;
76 schedule();
77 return 0;
78}
79
80/**
69 * nfs_compat_user_ino64 - returns the user-visible inode number 81 * nfs_compat_user_ino64 - returns the user-visible inode number
70 * @fileid: 64-bit fileid 82 * @fileid: 64-bit fileid
71 * 83 *
@@ -249,13 +261,10 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
249 struct inode *inode = ERR_PTR(-ENOENT); 261 struct inode *inode = ERR_PTR(-ENOENT);
250 unsigned long hash; 262 unsigned long hash;
251 263
252 if ((fattr->valid & NFS_ATTR_FATTR) == 0) 264 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) == 0)
253 goto out_no_inode; 265 goto out_no_inode;
254 266 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) == 0)
255 if (!fattr->nlink) {
256 printk("NFS: Buggy server - nlink == 0!\n");
257 goto out_no_inode; 267 goto out_no_inode;
258 }
259 268
260 hash = nfs_fattr_to_ino_t(fattr); 269 hash = nfs_fattr_to_ino_t(fattr);
261 270
@@ -291,7 +300,8 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
291 && fattr->size <= NFS_LIMIT_READDIRPLUS) 300 && fattr->size <= NFS_LIMIT_READDIRPLUS)
292 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags); 301 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_I(inode)->flags);
293 /* Deal with crossing mountpoints */ 302 /* Deal with crossing mountpoints */
294 if (!nfs_fsid_equal(&NFS_SB(sb)->fsid, &fattr->fsid)) { 303 if ((fattr->valid & NFS_ATTR_FATTR_FSID)
304 && !nfs_fsid_equal(&NFS_SB(sb)->fsid, &fattr->fsid)) {
295 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL) 305 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)
296 inode->i_op = &nfs_referral_inode_operations; 306 inode->i_op = &nfs_referral_inode_operations;
297 else 307 else
@@ -304,28 +314,45 @@ nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
304 else 314 else
305 init_special_inode(inode, inode->i_mode, fattr->rdev); 315 init_special_inode(inode, inode->i_mode, fattr->rdev);
306 316
317 memset(&inode->i_atime, 0, sizeof(inode->i_atime));
318 memset(&inode->i_mtime, 0, sizeof(inode->i_mtime));
319 memset(&inode->i_ctime, 0, sizeof(inode->i_ctime));
320 nfsi->change_attr = 0;
321 inode->i_size = 0;
322 inode->i_nlink = 0;
323 inode->i_uid = -2;
324 inode->i_gid = -2;
325 inode->i_blocks = 0;
326 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
327
307 nfsi->read_cache_jiffies = fattr->time_start; 328 nfsi->read_cache_jiffies = fattr->time_start;
308 nfsi->attr_gencount = fattr->gencount; 329 nfsi->attr_gencount = fattr->gencount;
309 inode->i_atime = fattr->atime; 330 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
310 inode->i_mtime = fattr->mtime; 331 inode->i_atime = fattr->atime;
311 inode->i_ctime = fattr->ctime; 332 if (fattr->valid & NFS_ATTR_FATTR_MTIME)
312 if (fattr->valid & NFS_ATTR_FATTR_V4) 333 inode->i_mtime = fattr->mtime;
334 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
335 inode->i_ctime = fattr->ctime;
336 if (fattr->valid & NFS_ATTR_FATTR_CHANGE)
313 nfsi->change_attr = fattr->change_attr; 337 nfsi->change_attr = fattr->change_attr;
314 inode->i_size = nfs_size_to_loff_t(fattr->size); 338 if (fattr->valid & NFS_ATTR_FATTR_SIZE)
315 inode->i_nlink = fattr->nlink; 339 inode->i_size = nfs_size_to_loff_t(fattr->size);
316 inode->i_uid = fattr->uid; 340 if (fattr->valid & NFS_ATTR_FATTR_NLINK)
317 inode->i_gid = fattr->gid; 341 inode->i_nlink = fattr->nlink;
318 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) { 342 if (fattr->valid & NFS_ATTR_FATTR_OWNER)
343 inode->i_uid = fattr->uid;
344 if (fattr->valid & NFS_ATTR_FATTR_GROUP)
345 inode->i_gid = fattr->gid;
346 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
347 inode->i_blocks = fattr->du.nfs2.blocks;
348 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
319 /* 349 /*
320 * report the blocks in 512byte units 350 * report the blocks in 512byte units
321 */ 351 */
322 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used); 352 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
323 } else {
324 inode->i_blocks = fattr->du.nfs2.blocks;
325 } 353 }
326 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode); 354 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
327 nfsi->attrtimeo_timestamp = now; 355 nfsi->attrtimeo_timestamp = now;
328 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
329 nfsi->access_cache = RB_ROOT; 356 nfsi->access_cache = RB_ROOT;
330 357
331 unlock_new_inode(inode); 358 unlock_new_inode(inode);
@@ -514,6 +541,32 @@ int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
514 return err; 541 return err;
515} 542}
516 543
544/**
545 * nfs_close_context - Common close_context() routine NFSv2/v3
546 * @ctx: pointer to context
547 * @is_sync: is this a synchronous close
548 *
549 * always ensure that the attributes are up to date if we're mounted
550 * with close-to-open semantics
551 */
552void nfs_close_context(struct nfs_open_context *ctx, int is_sync)
553{
554 struct inode *inode;
555 struct nfs_server *server;
556
557 if (!(ctx->mode & FMODE_WRITE))
558 return;
559 if (!is_sync)
560 return;
561 inode = ctx->path.dentry->d_inode;
562 if (!list_empty(&NFS_I(inode)->open_files))
563 return;
564 server = NFS_SERVER(inode);
565 if (server->flags & NFS_MOUNT_NOCTO)
566 return;
567 nfs_revalidate_inode(server, inode);
568}
569
517static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred) 570static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
518{ 571{
519 struct nfs_open_context *ctx; 572 struct nfs_open_context *ctx;
@@ -540,24 +593,15 @@ struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
540 return ctx; 593 return ctx;
541} 594}
542 595
543static void __put_nfs_open_context(struct nfs_open_context *ctx, int wait) 596static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
544{ 597{
545 struct inode *inode; 598 struct inode *inode = ctx->path.dentry->d_inode;
546
547 if (ctx == NULL)
548 return;
549 599
550 inode = ctx->path.dentry->d_inode;
551 if (!atomic_dec_and_lock(&ctx->count, &inode->i_lock)) 600 if (!atomic_dec_and_lock(&ctx->count, &inode->i_lock))
552 return; 601 return;
553 list_del(&ctx->list); 602 list_del(&ctx->list);
554 spin_unlock(&inode->i_lock); 603 spin_unlock(&inode->i_lock);
555 if (ctx->state != NULL) { 604 NFS_PROTO(inode)->close_context(ctx, is_sync);
556 if (wait)
557 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
558 else
559 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
560 }
561 if (ctx->cred != NULL) 605 if (ctx->cred != NULL)
562 put_rpccred(ctx->cred); 606 put_rpccred(ctx->cred);
563 path_put(&ctx->path); 607 path_put(&ctx->path);
@@ -670,9 +714,6 @@ __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
670 if (NFS_STALE(inode)) 714 if (NFS_STALE(inode))
671 goto out; 715 goto out;
672 716
673 if (NFS_STALE(inode))
674 goto out;
675
676 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE); 717 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
677 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr); 718 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
678 if (status != 0) { 719 if (status != 0) {
@@ -815,25 +856,31 @@ static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
815{ 856{
816 struct nfs_inode *nfsi = NFS_I(inode); 857 struct nfs_inode *nfsi = NFS_I(inode);
817 858
818 if ((fattr->valid & NFS_ATTR_WCC_V4) != 0 && 859 if ((fattr->valid & NFS_ATTR_FATTR_PRECHANGE)
819 nfsi->change_attr == fattr->pre_change_attr) { 860 && (fattr->valid & NFS_ATTR_FATTR_CHANGE)
861 && nfsi->change_attr == fattr->pre_change_attr) {
820 nfsi->change_attr = fattr->change_attr; 862 nfsi->change_attr = fattr->change_attr;
821 if (S_ISDIR(inode->i_mode)) 863 if (S_ISDIR(inode->i_mode))
822 nfsi->cache_validity |= NFS_INO_INVALID_DATA; 864 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
823 } 865 }
824 /* If we have atomic WCC data, we may update some attributes */ 866 /* If we have atomic WCC data, we may update some attributes */
825 if ((fattr->valid & NFS_ATTR_WCC) != 0) { 867 if ((fattr->valid & NFS_ATTR_FATTR_PRECTIME)
826 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) 868 && (fattr->valid & NFS_ATTR_FATTR_CTIME)
869 && timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
827 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime)); 870 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
828 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) { 871
872 if ((fattr->valid & NFS_ATTR_FATTR_PREMTIME)
873 && (fattr->valid & NFS_ATTR_FATTR_MTIME)
874 && timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
829 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime)); 875 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
830 if (S_ISDIR(inode->i_mode)) 876 if (S_ISDIR(inode->i_mode))
831 nfsi->cache_validity |= NFS_INO_INVALID_DATA; 877 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
832 }
833 if (i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size) &&
834 nfsi->npages == 0)
835 i_size_write(inode, nfs_size_to_loff_t(fattr->size));
836 } 878 }
879 if ((fattr->valid & NFS_ATTR_FATTR_PRESIZE)
880 && (fattr->valid & NFS_ATTR_FATTR_SIZE)
881 && i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size)
882 && nfsi->npages == 0)
883 i_size_write(inode, nfs_size_to_loff_t(fattr->size));
837} 884}
838 885
839/** 886/**
@@ -853,35 +900,39 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat
853 900
854 901
855 /* Has the inode gone and changed behind our back? */ 902 /* Has the inode gone and changed behind our back? */
856 if (nfsi->fileid != fattr->fileid 903 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid)
857 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) { 904 return -EIO;
905 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
858 return -EIO; 906 return -EIO;
859 }
860 907
861 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 && 908 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
862 nfsi->change_attr != fattr->change_attr) 909 nfsi->change_attr != fattr->change_attr)
863 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE; 910 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
864 911
865 /* Verify a few of the more important attributes */ 912 /* Verify a few of the more important attributes */
866 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) 913 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) && !timespec_equal(&inode->i_mtime, &fattr->mtime))
867 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE; 914 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
868 915
869 cur_size = i_size_read(inode); 916 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
870 new_isize = nfs_size_to_loff_t(fattr->size); 917 cur_size = i_size_read(inode);
871 if (cur_size != new_isize && nfsi->npages == 0) 918 new_isize = nfs_size_to_loff_t(fattr->size);
872 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE; 919 if (cur_size != new_isize && nfsi->npages == 0)
920 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
921 }
873 922
874 /* Have any file permissions changed? */ 923 /* Have any file permissions changed? */
875 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) 924 if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO))
876 || inode->i_uid != fattr->uid 925 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
877 || inode->i_gid != fattr->gid) 926 if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && inode->i_uid != fattr->uid)
927 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
928 if ((fattr->valid & NFS_ATTR_FATTR_GROUP) && inode->i_gid != fattr->gid)
878 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL; 929 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
879 930
880 /* Has the link count changed? */ 931 /* Has the link count changed? */
881 if (inode->i_nlink != fattr->nlink) 932 if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink)
882 invalid |= NFS_INO_INVALID_ATTR; 933 invalid |= NFS_INO_INVALID_ATTR;
883 934
884 if (!timespec_equal(&inode->i_atime, &fattr->atime)) 935 if ((fattr->valid & NFS_ATTR_FATTR_ATIME) && !timespec_equal(&inode->i_atime, &fattr->atime))
885 invalid |= NFS_INO_INVALID_ATIME; 936 invalid |= NFS_INO_INVALID_ATIME;
886 937
887 if (invalid != 0) 938 if (invalid != 0)
@@ -893,11 +944,15 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat
893 944
894static int nfs_ctime_need_update(const struct inode *inode, const struct nfs_fattr *fattr) 945static int nfs_ctime_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
895{ 946{
947 if (!(fattr->valid & NFS_ATTR_FATTR_CTIME))
948 return 0;
896 return timespec_compare(&fattr->ctime, &inode->i_ctime) > 0; 949 return timespec_compare(&fattr->ctime, &inode->i_ctime) > 0;
897} 950}
898 951
899static int nfs_size_need_update(const struct inode *inode, const struct nfs_fattr *fattr) 952static int nfs_size_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
900{ 953{
954 if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
955 return 0;
901 return nfs_size_to_loff_t(fattr->size) > i_size_read(inode); 956 return nfs_size_to_loff_t(fattr->size) > i_size_read(inode);
902} 957}
903 958
@@ -1033,20 +1088,31 @@ int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fa
1033 /* Don't do a WCC update if these attributes are already stale */ 1088 /* Don't do a WCC update if these attributes are already stale */
1034 if ((fattr->valid & NFS_ATTR_FATTR) == 0 || 1089 if ((fattr->valid & NFS_ATTR_FATTR) == 0 ||
1035 !nfs_inode_attrs_need_update(inode, fattr)) { 1090 !nfs_inode_attrs_need_update(inode, fattr)) {
1036 fattr->valid &= ~(NFS_ATTR_WCC_V4|NFS_ATTR_WCC); 1091 fattr->valid &= ~(NFS_ATTR_FATTR_PRECHANGE
1092 | NFS_ATTR_FATTR_PRESIZE
1093 | NFS_ATTR_FATTR_PREMTIME
1094 | NFS_ATTR_FATTR_PRECTIME);
1037 goto out_noforce; 1095 goto out_noforce;
1038 } 1096 }
1039 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 && 1097 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1040 (fattr->valid & NFS_ATTR_WCC_V4) == 0) { 1098 (fattr->valid & NFS_ATTR_FATTR_PRECHANGE) == 0) {
1041 fattr->pre_change_attr = NFS_I(inode)->change_attr; 1099 fattr->pre_change_attr = NFS_I(inode)->change_attr;
1042 fattr->valid |= NFS_ATTR_WCC_V4; 1100 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
1043 } 1101 }
1044 if ((fattr->valid & NFS_ATTR_FATTR) != 0 && 1102 if ((fattr->valid & NFS_ATTR_FATTR_CTIME) != 0 &&
1045 (fattr->valid & NFS_ATTR_WCC) == 0) { 1103 (fattr->valid & NFS_ATTR_FATTR_PRECTIME) == 0) {
1046 memcpy(&fattr->pre_ctime, &inode->i_ctime, sizeof(fattr->pre_ctime)); 1104 memcpy(&fattr->pre_ctime, &inode->i_ctime, sizeof(fattr->pre_ctime));
1105 fattr->valid |= NFS_ATTR_FATTR_PRECTIME;
1106 }
1107 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) != 0 &&
1108 (fattr->valid & NFS_ATTR_FATTR_PREMTIME) == 0) {
1047 memcpy(&fattr->pre_mtime, &inode->i_mtime, sizeof(fattr->pre_mtime)); 1109 memcpy(&fattr->pre_mtime, &inode->i_mtime, sizeof(fattr->pre_mtime));
1110 fattr->valid |= NFS_ATTR_FATTR_PREMTIME;
1111 }
1112 if ((fattr->valid & NFS_ATTR_FATTR_SIZE) != 0 &&
1113 (fattr->valid & NFS_ATTR_FATTR_PRESIZE) == 0) {
1048 fattr->pre_size = i_size_read(inode); 1114 fattr->pre_size = i_size_read(inode);
1049 fattr->valid |= NFS_ATTR_WCC; 1115 fattr->valid |= NFS_ATTR_FATTR_PRESIZE;
1050 } 1116 }
1051out_noforce: 1117out_noforce:
1052 status = nfs_post_op_update_inode_locked(inode, fattr); 1118 status = nfs_post_op_update_inode_locked(inode, fattr);
@@ -1078,18 +1144,18 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1078 __func__, inode->i_sb->s_id, inode->i_ino, 1144 __func__, inode->i_sb->s_id, inode->i_ino,
1079 atomic_read(&inode->i_count), fattr->valid); 1145 atomic_read(&inode->i_count), fattr->valid);
1080 1146
1081 if (nfsi->fileid != fattr->fileid) 1147 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid)
1082 goto out_fileid; 1148 goto out_fileid;
1083 1149
1084 /* 1150 /*
1085 * Make sure the inode's type hasn't changed. 1151 * Make sure the inode's type hasn't changed.
1086 */ 1152 */
1087 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) 1153 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1088 goto out_changed; 1154 goto out_changed;
1089 1155
1090 server = NFS_SERVER(inode); 1156 server = NFS_SERVER(inode);
1091 /* Update the fsid? */ 1157 /* Update the fsid? */
1092 if (S_ISDIR(inode->i_mode) && 1158 if (S_ISDIR(inode->i_mode) && (fattr->valid & NFS_ATTR_FATTR_FSID) &&
1093 !nfs_fsid_equal(&server->fsid, &fattr->fsid) && 1159 !nfs_fsid_equal(&server->fsid, &fattr->fsid) &&
1094 !test_bit(NFS_INO_MOUNTPOINT, &nfsi->flags)) 1160 !test_bit(NFS_INO_MOUNTPOINT, &nfsi->flags))
1095 server->fsid = fattr->fsid; 1161 server->fsid = fattr->fsid;
@@ -1099,14 +1165,27 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1099 */ 1165 */
1100 nfsi->read_cache_jiffies = fattr->time_start; 1166 nfsi->read_cache_jiffies = fattr->time_start;
1101 1167
1102 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ATIME 1168 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) || (fattr->valid & (NFS_ATTR_FATTR_MTIME|NFS_ATTR_FATTR_CTIME)))
1103 | NFS_INO_REVAL_PAGECACHE); 1169 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR
1170 | NFS_INO_INVALID_ATIME
1171 | NFS_INO_REVAL_PAGECACHE);
1104 1172
1105 /* Do atomic weak cache consistency updates */ 1173 /* Do atomic weak cache consistency updates */
1106 nfs_wcc_update_inode(inode, fattr); 1174 nfs_wcc_update_inode(inode, fattr);
1107 1175
1108 /* More cache consistency checks */ 1176 /* More cache consistency checks */
1109 if (!(fattr->valid & NFS_ATTR_FATTR_V4)) { 1177 if (fattr->valid & NFS_ATTR_FATTR_CHANGE) {
1178 if (nfsi->change_attr != fattr->change_attr) {
1179 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1180 inode->i_sb->s_id, inode->i_ino);
1181 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1182 if (S_ISDIR(inode->i_mode))
1183 nfs_force_lookup_revalidate(inode);
1184 nfsi->change_attr = fattr->change_attr;
1185 }
1186 }
1187
1188 if (fattr->valid & NFS_ATTR_FATTR_MTIME) {
1110 /* NFSv2/v3: Check if the mtime agrees */ 1189 /* NFSv2/v3: Check if the mtime agrees */
1111 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) { 1190 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1112 dprintk("NFS: mtime change on server for file %s/%ld\n", 1191 dprintk("NFS: mtime change on server for file %s/%ld\n",
@@ -1114,59 +1193,80 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1114 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; 1193 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1115 if (S_ISDIR(inode->i_mode)) 1194 if (S_ISDIR(inode->i_mode))
1116 nfs_force_lookup_revalidate(inode); 1195 nfs_force_lookup_revalidate(inode);
1196 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1117 } 1197 }
1198 }
1199 if (fattr->valid & NFS_ATTR_FATTR_CTIME) {
1118 /* If ctime has changed we should definitely clear access+acl caches */ 1200 /* If ctime has changed we should definitely clear access+acl caches */
1119 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) 1201 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1120 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; 1202 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1121 } else if (nfsi->change_attr != fattr->change_attr) { 1203 /* and probably clear data for a directory too as utimes can cause
1122 dprintk("NFS: change_attr change on server for file %s/%ld\n", 1204 * havoc with our cache.
1123 inode->i_sb->s_id, inode->i_ino); 1205 */
1124 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL; 1206 if (S_ISDIR(inode->i_mode)) {
1125 if (S_ISDIR(inode->i_mode)) 1207 invalid |= NFS_INO_INVALID_DATA;
1126 nfs_force_lookup_revalidate(inode); 1208 nfs_force_lookup_revalidate(inode);
1209 }
1210 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1211 }
1127 } 1212 }
1128 1213
1129 /* Check if our cached file size is stale */ 1214 /* Check if our cached file size is stale */
1130 new_isize = nfs_size_to_loff_t(fattr->size); 1215 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1131 cur_isize = i_size_read(inode); 1216 new_isize = nfs_size_to_loff_t(fattr->size);
1132 if (new_isize != cur_isize) { 1217 cur_isize = i_size_read(inode);
1133 /* Do we perhaps have any outstanding writes, or has 1218 if (new_isize != cur_isize) {
1134 * the file grown beyond our last write? */ 1219 /* Do we perhaps have any outstanding writes, or has
1135 if (nfsi->npages == 0 || new_isize > cur_isize) { 1220 * the file grown beyond our last write? */
1136 i_size_write(inode, new_isize); 1221 if (nfsi->npages == 0 || new_isize > cur_isize) {
1137 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA; 1222 i_size_write(inode, new_isize);
1223 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1224 }
1225 dprintk("NFS: isize change on server for file %s/%ld\n",
1226 inode->i_sb->s_id, inode->i_ino);
1138 } 1227 }
1139 dprintk("NFS: isize change on server for file %s/%ld\n",
1140 inode->i_sb->s_id, inode->i_ino);
1141 } 1228 }
1142 1229
1143 1230
1144 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime)); 1231 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
1145 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime)); 1232 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1146 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1147 nfsi->change_attr = fattr->change_attr;
1148
1149 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1150 inode->i_uid != fattr->uid ||
1151 inode->i_gid != fattr->gid)
1152 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1153 1233
1154 if (inode->i_nlink != fattr->nlink) 1234 if (fattr->valid & NFS_ATTR_FATTR_MODE) {
1155 invalid |= NFS_INO_INVALID_ATTR; 1235 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) {
1236 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1237 inode->i_mode = fattr->mode;
1238 }
1239 }
1240 if (fattr->valid & NFS_ATTR_FATTR_OWNER) {
1241 if (inode->i_uid != fattr->uid) {
1242 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1243 inode->i_uid = fattr->uid;
1244 }
1245 }
1246 if (fattr->valid & NFS_ATTR_FATTR_GROUP) {
1247 if (inode->i_gid != fattr->gid) {
1248 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1249 inode->i_gid = fattr->gid;
1250 }
1251 }
1156 1252
1157 inode->i_mode = fattr->mode; 1253 if (fattr->valid & NFS_ATTR_FATTR_NLINK) {
1158 inode->i_nlink = fattr->nlink; 1254 if (inode->i_nlink != fattr->nlink) {
1159 inode->i_uid = fattr->uid; 1255 invalid |= NFS_INO_INVALID_ATTR;
1160 inode->i_gid = fattr->gid; 1256 if (S_ISDIR(inode->i_mode))
1257 invalid |= NFS_INO_INVALID_DATA;
1258 inode->i_nlink = fattr->nlink;
1259 }
1260 }
1161 1261
1162 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) { 1262 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
1163 /* 1263 /*
1164 * report the blocks in 512byte units 1264 * report the blocks in 512byte units
1165 */ 1265 */
1166 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used); 1266 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1167 } else {
1168 inode->i_blocks = fattr->du.nfs2.blocks;
1169 } 1267 }
1268 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
1269 inode->i_blocks = fattr->du.nfs2.blocks;
1170 1270
1171 /* Update attrtimeo value if we're out of the unstable period */ 1271 /* Update attrtimeo value if we're out of the unstable period */
1172 if (invalid & NFS_INO_INVALID_ATTR) { 1272 if (invalid & NFS_INO_INVALID_ATTR) {
@@ -1274,7 +1374,6 @@ static void init_once(void *foo)
1274 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru); 1374 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
1275 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru); 1375 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
1276 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC); 1376 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1277 nfsi->ncommit = 0;
1278 nfsi->npages = 0; 1377 nfsi->npages = 0;
1279 atomic_set(&nfsi->silly_count, 1); 1378 atomic_set(&nfsi->silly_count, 1);
1280 INIT_HLIST_HEAD(&nfsi->silly_list); 1379 INIT_HLIST_HEAD(&nfsi->silly_list);
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index 340ede8f608f..2041f68ff1cc 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -152,6 +152,9 @@ extern __be32 *nfs4_decode_dirent(__be32 *p, struct nfs_entry *entry, int plus);
152extern struct rpc_procinfo nfs4_procedures[]; 152extern struct rpc_procinfo nfs4_procedures[];
153#endif 153#endif
154 154
155/* proc.c */
156void nfs_close_context(struct nfs_open_context *ctx, int is_sync);
157
155/* dir.c */ 158/* dir.c */
156extern int nfs_access_cache_shrinker(int nr_to_scan, gfp_t gfp_mask); 159extern int nfs_access_cache_shrinker(int nr_to_scan, gfp_t gfp_mask);
157 160
@@ -165,6 +168,7 @@ extern void nfs_clear_inode(struct inode *);
165extern void nfs4_clear_inode(struct inode *); 168extern void nfs4_clear_inode(struct inode *);
166#endif 169#endif
167void nfs_zap_acl_cache(struct inode *inode); 170void nfs_zap_acl_cache(struct inode *inode);
171extern int nfs_wait_bit_killable(void *word);
168 172
169/* super.c */ 173/* super.c */
170void nfs_parse_ip_address(char *, size_t, struct sockaddr *, size_t *); 174void nfs_parse_ip_address(char *, size_t, struct sockaddr *, size_t *);
diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c
index 28bab67d1519..c862c9340f9a 100644
--- a/fs/nfs/nfs2xdr.c
+++ b/fs/nfs/nfs2xdr.c
@@ -120,8 +120,8 @@ xdr_decode_time(__be32 *p, struct timespec *timep)
120static __be32 * 120static __be32 *
121xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr) 121xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr)
122{ 122{
123 u32 rdev; 123 u32 rdev, type;
124 fattr->type = (enum nfs_ftype) ntohl(*p++); 124 type = ntohl(*p++);
125 fattr->mode = ntohl(*p++); 125 fattr->mode = ntohl(*p++);
126 fattr->nlink = ntohl(*p++); 126 fattr->nlink = ntohl(*p++);
127 fattr->uid = ntohl(*p++); 127 fattr->uid = ntohl(*p++);
@@ -136,10 +136,9 @@ xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr)
136 p = xdr_decode_time(p, &fattr->atime); 136 p = xdr_decode_time(p, &fattr->atime);
137 p = xdr_decode_time(p, &fattr->mtime); 137 p = xdr_decode_time(p, &fattr->mtime);
138 p = xdr_decode_time(p, &fattr->ctime); 138 p = xdr_decode_time(p, &fattr->ctime);
139 fattr->valid |= NFS_ATTR_FATTR; 139 fattr->valid |= NFS_ATTR_FATTR_V2;
140 fattr->rdev = new_decode_dev(rdev); 140 fattr->rdev = new_decode_dev(rdev);
141 if (fattr->type == NFCHR && rdev == NFS2_FIFO_DEV) { 141 if (type == NFCHR && rdev == NFS2_FIFO_DEV) {
142 fattr->type = NFFIFO;
143 fattr->mode = (fattr->mode & ~S_IFMT) | S_IFIFO; 142 fattr->mode = (fattr->mode & ~S_IFMT) | S_IFIFO;
144 fattr->rdev = 0; 143 fattr->rdev = 0;
145 } 144 }
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index c55be7a7679e..b82fe6847f14 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -834,4 +834,5 @@ const struct nfs_rpc_ops nfs_v3_clientops = {
834 .commit_done = nfs3_commit_done, 834 .commit_done = nfs3_commit_done,
835 .lock = nfs3_proc_lock, 835 .lock = nfs3_proc_lock,
836 .clear_acl_cache = nfs3_forget_cached_acls, 836 .clear_acl_cache = nfs3_forget_cached_acls,
837 .close_context = nfs_close_context,
837}; 838};
diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c
index 6cdeacffde46..e6a1932c7110 100644
--- a/fs/nfs/nfs3xdr.c
+++ b/fs/nfs/nfs3xdr.c
@@ -91,19 +91,15 @@
91/* 91/*
92 * Map file type to S_IFMT bits 92 * Map file type to S_IFMT bits
93 */ 93 */
94static struct { 94static const umode_t nfs_type2fmt[] = {
95 unsigned int mode; 95 [NF3BAD] = 0,
96 unsigned int nfs2type; 96 [NF3REG] = S_IFREG,
97} nfs_type2fmt[] = { 97 [NF3DIR] = S_IFDIR,
98 { 0, NFNON }, 98 [NF3BLK] = S_IFBLK,
99 { S_IFREG, NFREG }, 99 [NF3CHR] = S_IFCHR,
100 { S_IFDIR, NFDIR }, 100 [NF3LNK] = S_IFLNK,
101 { S_IFBLK, NFBLK }, 101 [NF3SOCK] = S_IFSOCK,
102 { S_IFCHR, NFCHR }, 102 [NF3FIFO] = S_IFIFO,
103 { S_IFLNK, NFLNK },
104 { S_IFSOCK, NFSOCK },
105 { S_IFIFO, NFFIFO },
106 { 0, NFBAD }
107}; 103};
108 104
109/* 105/*
@@ -148,13 +144,12 @@ static __be32 *
148xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr) 144xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr)
149{ 145{
150 unsigned int type, major, minor; 146 unsigned int type, major, minor;
151 int fmode; 147 umode_t fmode;
152 148
153 type = ntohl(*p++); 149 type = ntohl(*p++);
154 if (type >= NF3BAD) 150 if (type > NF3FIFO)
155 type = NF3BAD; 151 type = NF3NON;
156 fmode = nfs_type2fmt[type].mode; 152 fmode = nfs_type2fmt[type];
157 fattr->type = nfs_type2fmt[type].nfs2type;
158 fattr->mode = (ntohl(*p++) & ~S_IFMT) | fmode; 153 fattr->mode = (ntohl(*p++) & ~S_IFMT) | fmode;
159 fattr->nlink = ntohl(*p++); 154 fattr->nlink = ntohl(*p++);
160 fattr->uid = ntohl(*p++); 155 fattr->uid = ntohl(*p++);
@@ -177,7 +172,7 @@ xdr_decode_fattr(__be32 *p, struct nfs_fattr *fattr)
177 p = xdr_decode_time3(p, &fattr->ctime); 172 p = xdr_decode_time3(p, &fattr->ctime);
178 173
179 /* Update the mode bits */ 174 /* Update the mode bits */
180 fattr->valid |= (NFS_ATTR_FATTR | NFS_ATTR_FATTR_V3); 175 fattr->valid |= NFS_ATTR_FATTR_V3;
181 return p; 176 return p;
182} 177}
183 178
@@ -233,7 +228,9 @@ xdr_decode_wcc_attr(__be32 *p, struct nfs_fattr *fattr)
233 p = xdr_decode_hyper(p, &fattr->pre_size); 228 p = xdr_decode_hyper(p, &fattr->pre_size);
234 p = xdr_decode_time3(p, &fattr->pre_mtime); 229 p = xdr_decode_time3(p, &fattr->pre_mtime);
235 p = xdr_decode_time3(p, &fattr->pre_ctime); 230 p = xdr_decode_time3(p, &fattr->pre_ctime);
236 fattr->valid |= NFS_ATTR_WCC; 231 fattr->valid |= NFS_ATTR_FATTR_PRESIZE
232 | NFS_ATTR_FATTR_PREMTIME
233 | NFS_ATTR_FATTR_PRECTIME;
237 return p; 234 return p;
238} 235}
239 236
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 8dde84b988d9..97bacccff579 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -193,14 +193,6 @@ static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dent
193 kunmap_atomic(start, KM_USER0); 193 kunmap_atomic(start, KM_USER0);
194} 194}
195 195
196static int nfs4_wait_bit_killable(void *word)
197{
198 if (fatal_signal_pending(current))
199 return -ERESTARTSYS;
200 schedule();
201 return 0;
202}
203
204static int nfs4_wait_clnt_recover(struct nfs_client *clp) 196static int nfs4_wait_clnt_recover(struct nfs_client *clp)
205{ 197{
206 int res; 198 int res;
@@ -208,7 +200,7 @@ static int nfs4_wait_clnt_recover(struct nfs_client *clp)
208 might_sleep(); 200 might_sleep();
209 201
210 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING, 202 res = wait_on_bit(&clp->cl_state, NFS4CLNT_MANAGER_RUNNING,
211 nfs4_wait_bit_killable, TASK_KILLABLE); 203 nfs_wait_bit_killable, TASK_KILLABLE);
212 return res; 204 return res;
213} 205}
214 206
@@ -1439,7 +1431,7 @@ int nfs4_do_close(struct path *path, struct nfs4_state *state, int wait)
1439 if (calldata->arg.seqid == NULL) 1431 if (calldata->arg.seqid == NULL)
1440 goto out_free_calldata; 1432 goto out_free_calldata;
1441 calldata->arg.fmode = 0; 1433 calldata->arg.fmode = 0;
1442 calldata->arg.bitmask = server->attr_bitmask; 1434 calldata->arg.bitmask = server->cache_consistency_bitmask;
1443 calldata->res.fattr = &calldata->fattr; 1435 calldata->res.fattr = &calldata->fattr;
1444 calldata->res.seqid = calldata->arg.seqid; 1436 calldata->res.seqid = calldata->arg.seqid;
1445 calldata->res.server = server; 1437 calldata->res.server = server;
@@ -1580,6 +1572,15 @@ out_drop:
1580 return 0; 1572 return 0;
1581} 1573}
1582 1574
1575void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
1576{
1577 if (ctx->state == NULL)
1578 return;
1579 if (is_sync)
1580 nfs4_close_sync(&ctx->path, ctx->state, ctx->mode);
1581 else
1582 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
1583}
1583 1584
1584static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle) 1585static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
1585{ 1586{
@@ -1600,6 +1601,9 @@ static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *f
1600 server->caps |= NFS_CAP_HARDLINKS; 1601 server->caps |= NFS_CAP_HARDLINKS;
1601 if (res.has_symlinks != 0) 1602 if (res.has_symlinks != 0)
1602 server->caps |= NFS_CAP_SYMLINKS; 1603 server->caps |= NFS_CAP_SYMLINKS;
1604 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
1605 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
1606 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
1603 server->acl_bitmask = res.acl_bitmask; 1607 server->acl_bitmask = res.acl_bitmask;
1604 } 1608 }
1605 return status; 1609 return status;
@@ -2079,7 +2083,7 @@ static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
2079 struct nfs_removeargs *args = msg->rpc_argp; 2083 struct nfs_removeargs *args = msg->rpc_argp;
2080 struct nfs_removeres *res = msg->rpc_resp; 2084 struct nfs_removeres *res = msg->rpc_resp;
2081 2085
2082 args->bitmask = server->attr_bitmask; 2086 args->bitmask = server->cache_consistency_bitmask;
2083 res->server = server; 2087 res->server = server;
2084 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE]; 2088 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
2085} 2089}
@@ -2323,7 +2327,7 @@ static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
2323 .pages = &page, 2327 .pages = &page,
2324 .pgbase = 0, 2328 .pgbase = 0,
2325 .count = count, 2329 .count = count,
2326 .bitmask = NFS_SERVER(dentry->d_inode)->attr_bitmask, 2330 .bitmask = NFS_SERVER(dentry->d_inode)->cache_consistency_bitmask,
2327 }; 2331 };
2328 struct nfs4_readdir_res res; 2332 struct nfs4_readdir_res res;
2329 struct rpc_message msg = { 2333 struct rpc_message msg = {
@@ -2552,7 +2556,7 @@ static void nfs4_proc_write_setup(struct nfs_write_data *data, struct rpc_messag
2552{ 2556{
2553 struct nfs_server *server = NFS_SERVER(data->inode); 2557 struct nfs_server *server = NFS_SERVER(data->inode);
2554 2558
2555 data->args.bitmask = server->attr_bitmask; 2559 data->args.bitmask = server->cache_consistency_bitmask;
2556 data->res.server = server; 2560 data->res.server = server;
2557 data->timestamp = jiffies; 2561 data->timestamp = jiffies;
2558 2562
@@ -2575,7 +2579,7 @@ static void nfs4_proc_commit_setup(struct nfs_write_data *data, struct rpc_messa
2575{ 2579{
2576 struct nfs_server *server = NFS_SERVER(data->inode); 2580 struct nfs_server *server = NFS_SERVER(data->inode);
2577 2581
2578 data->args.bitmask = server->attr_bitmask; 2582 data->args.bitmask = server->cache_consistency_bitmask;
2579 data->res.server = server; 2583 data->res.server = server;
2580 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT]; 2584 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
2581} 2585}
@@ -3678,6 +3682,19 @@ ssize_t nfs4_listxattr(struct dentry *dentry, char *buf, size_t buflen)
3678 return len; 3682 return len;
3679} 3683}
3680 3684
3685static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
3686{
3687 if (!((fattr->valid & NFS_ATTR_FATTR_FILEID) &&
3688 (fattr->valid & NFS_ATTR_FATTR_FSID) &&
3689 (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)))
3690 return;
3691
3692 fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3693 NFS_ATTR_FATTR_NLINK;
3694 fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3695 fattr->nlink = 2;
3696}
3697
3681int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, 3698int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
3682 struct nfs4_fs_locations *fs_locations, struct page *page) 3699 struct nfs4_fs_locations *fs_locations, struct page *page)
3683{ 3700{
@@ -3704,6 +3721,7 @@ int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name,
3704 fs_locations->server = server; 3721 fs_locations->server = server;
3705 fs_locations->nlocations = 0; 3722 fs_locations->nlocations = 0;
3706 status = rpc_call_sync(server->client, &msg, 0); 3723 status = rpc_call_sync(server->client, &msg, 0);
3724 nfs_fixup_referral_attributes(&fs_locations->fattr);
3707 dprintk("%s: returned status = %d\n", __func__, status); 3725 dprintk("%s: returned status = %d\n", __func__, status);
3708 return status; 3726 return status;
3709} 3727}
@@ -3767,6 +3785,7 @@ const struct nfs_rpc_ops nfs_v4_clientops = {
3767 .commit_done = nfs4_commit_done, 3785 .commit_done = nfs4_commit_done,
3768 .lock = nfs4_proc_lock, 3786 .lock = nfs4_proc_lock,
3769 .clear_acl_cache = nfs4_zap_acl_attr, 3787 .clear_acl_cache = nfs4_zap_acl_attr,
3788 .close_context = nfs4_close_context,
3770}; 3789};
3771 3790
3772/* 3791/*
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index 2022fe47966f..0298e909559f 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -62,8 +62,14 @@ static LIST_HEAD(nfs4_clientid_list);
62 62
63static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred) 63static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred)
64{ 64{
65 int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, 65 unsigned short port;
66 nfs_callback_tcpport, cred); 66 int status;
67
68 port = nfs_callback_tcpport;
69 if (clp->cl_addr.ss_family == AF_INET6)
70 port = nfs_callback_tcpport6;
71
72 status = nfs4_proc_setclientid(clp, NFS4_CALLBACK, port, cred);
67 if (status == 0) 73 if (status == 0)
68 status = nfs4_proc_setclientid_confirm(clp, cred); 74 status = nfs4_proc_setclientid_confirm(clp, cred);
69 if (status == 0) 75 if (status == 0)
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index d1e4c8f8a0a9..1690f0e44b91 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -522,20 +522,17 @@ static int nfs4_stat_to_errno(int);
522 decode_lookup_maxsz + \ 522 decode_lookup_maxsz + \
523 decode_fs_locations_maxsz) 523 decode_fs_locations_maxsz)
524 524
525static struct { 525static const umode_t nfs_type2fmt[] = {
526 unsigned int mode; 526 [NF4BAD] = 0,
527 unsigned int nfs2type; 527 [NF4REG] = S_IFREG,
528} nfs_type2fmt[] = { 528 [NF4DIR] = S_IFDIR,
529 { 0, NFNON }, 529 [NF4BLK] = S_IFBLK,
530 { S_IFREG, NFREG }, 530 [NF4CHR] = S_IFCHR,
531 { S_IFDIR, NFDIR }, 531 [NF4LNK] = S_IFLNK,
532 { S_IFBLK, NFBLK }, 532 [NF4SOCK] = S_IFSOCK,
533 { S_IFCHR, NFCHR }, 533 [NF4FIFO] = S_IFIFO,
534 { S_IFLNK, NFLNK }, 534 [NF4ATTRDIR] = 0,
535 { S_IFSOCK, NFSOCK }, 535 [NF4NAMEDATTR] = 0,
536 { S_IFIFO, NFFIFO },
537 { 0, NFNON },
538 { 0, NFNON },
539}; 536};
540 537
541struct compound_hdr { 538struct compound_hdr {
@@ -2160,6 +2157,7 @@ static int decode_attr_supported(struct xdr_stream *xdr, uint32_t *bitmap, uint3
2160static int decode_attr_type(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *type) 2157static int decode_attr_type(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *type)
2161{ 2158{
2162 __be32 *p; 2159 __be32 *p;
2160 int ret = 0;
2163 2161
2164 *type = 0; 2162 *type = 0;
2165 if (unlikely(bitmap[0] & (FATTR4_WORD0_TYPE - 1U))) 2163 if (unlikely(bitmap[0] & (FATTR4_WORD0_TYPE - 1U)))
@@ -2172,14 +2170,16 @@ static int decode_attr_type(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *
2172 return -EIO; 2170 return -EIO;
2173 } 2171 }
2174 bitmap[0] &= ~FATTR4_WORD0_TYPE; 2172 bitmap[0] &= ~FATTR4_WORD0_TYPE;
2173 ret = NFS_ATTR_FATTR_TYPE;
2175 } 2174 }
2176 dprintk("%s: type=0%o\n", __func__, nfs_type2fmt[*type].nfs2type); 2175 dprintk("%s: type=0%o\n", __func__, nfs_type2fmt[*type]);
2177 return 0; 2176 return ret;
2178} 2177}
2179 2178
2180static int decode_attr_change(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *change) 2179static int decode_attr_change(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *change)
2181{ 2180{
2182 __be32 *p; 2181 __be32 *p;
2182 int ret = 0;
2183 2183
2184 *change = 0; 2184 *change = 0;
2185 if (unlikely(bitmap[0] & (FATTR4_WORD0_CHANGE - 1U))) 2185 if (unlikely(bitmap[0] & (FATTR4_WORD0_CHANGE - 1U)))
@@ -2188,15 +2188,17 @@ static int decode_attr_change(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t
2188 READ_BUF(8); 2188 READ_BUF(8);
2189 READ64(*change); 2189 READ64(*change);
2190 bitmap[0] &= ~FATTR4_WORD0_CHANGE; 2190 bitmap[0] &= ~FATTR4_WORD0_CHANGE;
2191 ret = NFS_ATTR_FATTR_CHANGE;
2191 } 2192 }
2192 dprintk("%s: change attribute=%Lu\n", __func__, 2193 dprintk("%s: change attribute=%Lu\n", __func__,
2193 (unsigned long long)*change); 2194 (unsigned long long)*change);
2194 return 0; 2195 return ret;
2195} 2196}
2196 2197
2197static int decode_attr_size(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *size) 2198static int decode_attr_size(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *size)
2198{ 2199{
2199 __be32 *p; 2200 __be32 *p;
2201 int ret = 0;
2200 2202
2201 *size = 0; 2203 *size = 0;
2202 if (unlikely(bitmap[0] & (FATTR4_WORD0_SIZE - 1U))) 2204 if (unlikely(bitmap[0] & (FATTR4_WORD0_SIZE - 1U)))
@@ -2205,9 +2207,10 @@ static int decode_attr_size(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *
2205 READ_BUF(8); 2207 READ_BUF(8);
2206 READ64(*size); 2208 READ64(*size);
2207 bitmap[0] &= ~FATTR4_WORD0_SIZE; 2209 bitmap[0] &= ~FATTR4_WORD0_SIZE;
2210 ret = NFS_ATTR_FATTR_SIZE;
2208 } 2211 }
2209 dprintk("%s: file size=%Lu\n", __func__, (unsigned long long)*size); 2212 dprintk("%s: file size=%Lu\n", __func__, (unsigned long long)*size);
2210 return 0; 2213 return ret;
2211} 2214}
2212 2215
2213static int decode_attr_link_support(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *res) 2216static int decode_attr_link_support(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *res)
@@ -2245,6 +2248,7 @@ static int decode_attr_symlink_support(struct xdr_stream *xdr, uint32_t *bitmap,
2245static int decode_attr_fsid(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_fsid *fsid) 2248static int decode_attr_fsid(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_fsid *fsid)
2246{ 2249{
2247 __be32 *p; 2250 __be32 *p;
2251 int ret = 0;
2248 2252
2249 fsid->major = 0; 2253 fsid->major = 0;
2250 fsid->minor = 0; 2254 fsid->minor = 0;
@@ -2255,11 +2259,12 @@ static int decode_attr_fsid(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs
2255 READ64(fsid->major); 2259 READ64(fsid->major);
2256 READ64(fsid->minor); 2260 READ64(fsid->minor);
2257 bitmap[0] &= ~FATTR4_WORD0_FSID; 2261 bitmap[0] &= ~FATTR4_WORD0_FSID;
2262 ret = NFS_ATTR_FATTR_FSID;
2258 } 2263 }
2259 dprintk("%s: fsid=(0x%Lx/0x%Lx)\n", __func__, 2264 dprintk("%s: fsid=(0x%Lx/0x%Lx)\n", __func__,
2260 (unsigned long long)fsid->major, 2265 (unsigned long long)fsid->major,
2261 (unsigned long long)fsid->minor); 2266 (unsigned long long)fsid->minor);
2262 return 0; 2267 return ret;
2263} 2268}
2264 2269
2265static int decode_attr_lease_time(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *res) 2270static int decode_attr_lease_time(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *res)
@@ -2297,6 +2302,7 @@ static int decode_attr_aclsupport(struct xdr_stream *xdr, uint32_t *bitmap, uint
2297static int decode_attr_fileid(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *fileid) 2302static int decode_attr_fileid(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *fileid)
2298{ 2303{
2299 __be32 *p; 2304 __be32 *p;
2305 int ret = 0;
2300 2306
2301 *fileid = 0; 2307 *fileid = 0;
2302 if (unlikely(bitmap[0] & (FATTR4_WORD0_FILEID - 1U))) 2308 if (unlikely(bitmap[0] & (FATTR4_WORD0_FILEID - 1U)))
@@ -2305,14 +2311,16 @@ static int decode_attr_fileid(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t
2305 READ_BUF(8); 2311 READ_BUF(8);
2306 READ64(*fileid); 2312 READ64(*fileid);
2307 bitmap[0] &= ~FATTR4_WORD0_FILEID; 2313 bitmap[0] &= ~FATTR4_WORD0_FILEID;
2314 ret = NFS_ATTR_FATTR_FILEID;
2308 } 2315 }
2309 dprintk("%s: fileid=%Lu\n", __func__, (unsigned long long)*fileid); 2316 dprintk("%s: fileid=%Lu\n", __func__, (unsigned long long)*fileid);
2310 return 0; 2317 return ret;
2311} 2318}
2312 2319
2313static int decode_attr_mounted_on_fileid(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *fileid) 2320static int decode_attr_mounted_on_fileid(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *fileid)
2314{ 2321{
2315 __be32 *p; 2322 __be32 *p;
2323 int ret = 0;
2316 2324
2317 *fileid = 0; 2325 *fileid = 0;
2318 if (unlikely(bitmap[1] & (FATTR4_WORD1_MOUNTED_ON_FILEID - 1U))) 2326 if (unlikely(bitmap[1] & (FATTR4_WORD1_MOUNTED_ON_FILEID - 1U)))
@@ -2321,9 +2329,10 @@ static int decode_attr_mounted_on_fileid(struct xdr_stream *xdr, uint32_t *bitma
2321 READ_BUF(8); 2329 READ_BUF(8);
2322 READ64(*fileid); 2330 READ64(*fileid);
2323 bitmap[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID; 2331 bitmap[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
2332 ret = NFS_ATTR_FATTR_FILEID;
2324 } 2333 }
2325 dprintk("%s: fileid=%Lu\n", __func__, (unsigned long long)*fileid); 2334 dprintk("%s: fileid=%Lu\n", __func__, (unsigned long long)*fileid);
2326 return 0; 2335 return ret;
2327} 2336}
2328 2337
2329static int decode_attr_files_avail(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *res) 2338static int decode_attr_files_avail(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *res)
@@ -2479,6 +2488,8 @@ static int decode_attr_fs_locations(struct xdr_stream *xdr, uint32_t *bitmap, st
2479 if (res->nlocations < NFS4_FS_LOCATIONS_MAXENTRIES) 2488 if (res->nlocations < NFS4_FS_LOCATIONS_MAXENTRIES)
2480 res->nlocations++; 2489 res->nlocations++;
2481 } 2490 }
2491 if (res->nlocations != 0)
2492 status = NFS_ATTR_FATTR_V4_REFERRAL;
2482out: 2493out:
2483 dprintk("%s: fs_locations done, error = %d\n", __func__, status); 2494 dprintk("%s: fs_locations done, error = %d\n", __func__, status);
2484 return status; 2495 return status;
@@ -2580,26 +2591,30 @@ static int decode_attr_maxwrite(struct xdr_stream *xdr, uint32_t *bitmap, uint32
2580 return status; 2591 return status;
2581} 2592}
2582 2593
2583static int decode_attr_mode(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *mode) 2594static int decode_attr_mode(struct xdr_stream *xdr, uint32_t *bitmap, umode_t *mode)
2584{ 2595{
2596 uint32_t tmp;
2585 __be32 *p; 2597 __be32 *p;
2598 int ret = 0;
2586 2599
2587 *mode = 0; 2600 *mode = 0;
2588 if (unlikely(bitmap[1] & (FATTR4_WORD1_MODE - 1U))) 2601 if (unlikely(bitmap[1] & (FATTR4_WORD1_MODE - 1U)))
2589 return -EIO; 2602 return -EIO;
2590 if (likely(bitmap[1] & FATTR4_WORD1_MODE)) { 2603 if (likely(bitmap[1] & FATTR4_WORD1_MODE)) {
2591 READ_BUF(4); 2604 READ_BUF(4);
2592 READ32(*mode); 2605 READ32(tmp);
2593 *mode &= ~S_IFMT; 2606 *mode = tmp & ~S_IFMT;
2594 bitmap[1] &= ~FATTR4_WORD1_MODE; 2607 bitmap[1] &= ~FATTR4_WORD1_MODE;
2608 ret = NFS_ATTR_FATTR_MODE;
2595 } 2609 }
2596 dprintk("%s: file mode=0%o\n", __func__, (unsigned int)*mode); 2610 dprintk("%s: file mode=0%o\n", __func__, (unsigned int)*mode);
2597 return 0; 2611 return ret;
2598} 2612}
2599 2613
2600static int decode_attr_nlink(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *nlink) 2614static int decode_attr_nlink(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t *nlink)
2601{ 2615{
2602 __be32 *p; 2616 __be32 *p;
2617 int ret = 0;
2603 2618
2604 *nlink = 1; 2619 *nlink = 1;
2605 if (unlikely(bitmap[1] & (FATTR4_WORD1_NUMLINKS - 1U))) 2620 if (unlikely(bitmap[1] & (FATTR4_WORD1_NUMLINKS - 1U)))
@@ -2608,15 +2623,17 @@ static int decode_attr_nlink(struct xdr_stream *xdr, uint32_t *bitmap, uint32_t
2608 READ_BUF(4); 2623 READ_BUF(4);
2609 READ32(*nlink); 2624 READ32(*nlink);
2610 bitmap[1] &= ~FATTR4_WORD1_NUMLINKS; 2625 bitmap[1] &= ~FATTR4_WORD1_NUMLINKS;
2626 ret = NFS_ATTR_FATTR_NLINK;
2611 } 2627 }
2612 dprintk("%s: nlink=%u\n", __func__, (unsigned int)*nlink); 2628 dprintk("%s: nlink=%u\n", __func__, (unsigned int)*nlink);
2613 return 0; 2629 return ret;
2614} 2630}
2615 2631
2616static int decode_attr_owner(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_client *clp, uint32_t *uid) 2632static int decode_attr_owner(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_client *clp, uint32_t *uid)
2617{ 2633{
2618 uint32_t len; 2634 uint32_t len;
2619 __be32 *p; 2635 __be32 *p;
2636 int ret = 0;
2620 2637
2621 *uid = -2; 2638 *uid = -2;
2622 if (unlikely(bitmap[1] & (FATTR4_WORD1_OWNER - 1U))) 2639 if (unlikely(bitmap[1] & (FATTR4_WORD1_OWNER - 1U)))
@@ -2626,7 +2643,9 @@ static int decode_attr_owner(struct xdr_stream *xdr, uint32_t *bitmap, struct nf
2626 READ32(len); 2643 READ32(len);
2627 READ_BUF(len); 2644 READ_BUF(len);
2628 if (len < XDR_MAX_NETOBJ) { 2645 if (len < XDR_MAX_NETOBJ) {
2629 if (nfs_map_name_to_uid(clp, (char *)p, len, uid) != 0) 2646 if (nfs_map_name_to_uid(clp, (char *)p, len, uid) == 0)
2647 ret = NFS_ATTR_FATTR_OWNER;
2648 else
2630 dprintk("%s: nfs_map_name_to_uid failed!\n", 2649 dprintk("%s: nfs_map_name_to_uid failed!\n",
2631 __func__); 2650 __func__);
2632 } else 2651 } else
@@ -2635,13 +2654,14 @@ static int decode_attr_owner(struct xdr_stream *xdr, uint32_t *bitmap, struct nf
2635 bitmap[1] &= ~FATTR4_WORD1_OWNER; 2654 bitmap[1] &= ~FATTR4_WORD1_OWNER;
2636 } 2655 }
2637 dprintk("%s: uid=%d\n", __func__, (int)*uid); 2656 dprintk("%s: uid=%d\n", __func__, (int)*uid);
2638 return 0; 2657 return ret;
2639} 2658}
2640 2659
2641static int decode_attr_group(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_client *clp, uint32_t *gid) 2660static int decode_attr_group(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_client *clp, uint32_t *gid)
2642{ 2661{
2643 uint32_t len; 2662 uint32_t len;
2644 __be32 *p; 2663 __be32 *p;
2664 int ret = 0;
2645 2665
2646 *gid = -2; 2666 *gid = -2;
2647 if (unlikely(bitmap[1] & (FATTR4_WORD1_OWNER_GROUP - 1U))) 2667 if (unlikely(bitmap[1] & (FATTR4_WORD1_OWNER_GROUP - 1U)))
@@ -2651,7 +2671,9 @@ static int decode_attr_group(struct xdr_stream *xdr, uint32_t *bitmap, struct nf
2651 READ32(len); 2671 READ32(len);
2652 READ_BUF(len); 2672 READ_BUF(len);
2653 if (len < XDR_MAX_NETOBJ) { 2673 if (len < XDR_MAX_NETOBJ) {
2654 if (nfs_map_group_to_gid(clp, (char *)p, len, gid) != 0) 2674 if (nfs_map_group_to_gid(clp, (char *)p, len, gid) == 0)
2675 ret = NFS_ATTR_FATTR_GROUP;
2676 else
2655 dprintk("%s: nfs_map_group_to_gid failed!\n", 2677 dprintk("%s: nfs_map_group_to_gid failed!\n",
2656 __func__); 2678 __func__);
2657 } else 2679 } else
@@ -2660,13 +2682,14 @@ static int decode_attr_group(struct xdr_stream *xdr, uint32_t *bitmap, struct nf
2660 bitmap[1] &= ~FATTR4_WORD1_OWNER_GROUP; 2682 bitmap[1] &= ~FATTR4_WORD1_OWNER_GROUP;
2661 } 2683 }
2662 dprintk("%s: gid=%d\n", __func__, (int)*gid); 2684 dprintk("%s: gid=%d\n", __func__, (int)*gid);
2663 return 0; 2685 return ret;
2664} 2686}
2665 2687
2666static int decode_attr_rdev(struct xdr_stream *xdr, uint32_t *bitmap, dev_t *rdev) 2688static int decode_attr_rdev(struct xdr_stream *xdr, uint32_t *bitmap, dev_t *rdev)
2667{ 2689{
2668 uint32_t major = 0, minor = 0; 2690 uint32_t major = 0, minor = 0;
2669 __be32 *p; 2691 __be32 *p;
2692 int ret = 0;
2670 2693
2671 *rdev = MKDEV(0,0); 2694 *rdev = MKDEV(0,0);
2672 if (unlikely(bitmap[1] & (FATTR4_WORD1_RAWDEV - 1U))) 2695 if (unlikely(bitmap[1] & (FATTR4_WORD1_RAWDEV - 1U)))
@@ -2681,9 +2704,10 @@ static int decode_attr_rdev(struct xdr_stream *xdr, uint32_t *bitmap, dev_t *rde
2681 if (MAJOR(tmp) == major && MINOR(tmp) == minor) 2704 if (MAJOR(tmp) == major && MINOR(tmp) == minor)
2682 *rdev = tmp; 2705 *rdev = tmp;
2683 bitmap[1] &= ~ FATTR4_WORD1_RAWDEV; 2706 bitmap[1] &= ~ FATTR4_WORD1_RAWDEV;
2707 ret = NFS_ATTR_FATTR_RDEV;
2684 } 2708 }
2685 dprintk("%s: rdev=(0x%x:0x%x)\n", __func__, major, minor); 2709 dprintk("%s: rdev=(0x%x:0x%x)\n", __func__, major, minor);
2686 return 0; 2710 return ret;
2687} 2711}
2688 2712
2689static int decode_attr_space_avail(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *res) 2713static int decode_attr_space_avail(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *res)
@@ -2740,6 +2764,7 @@ static int decode_attr_space_total(struct xdr_stream *xdr, uint32_t *bitmap, uin
2740static int decode_attr_space_used(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *used) 2764static int decode_attr_space_used(struct xdr_stream *xdr, uint32_t *bitmap, uint64_t *used)
2741{ 2765{
2742 __be32 *p; 2766 __be32 *p;
2767 int ret = 0;
2743 2768
2744 *used = 0; 2769 *used = 0;
2745 if (unlikely(bitmap[1] & (FATTR4_WORD1_SPACE_USED - 1U))) 2770 if (unlikely(bitmap[1] & (FATTR4_WORD1_SPACE_USED - 1U)))
@@ -2748,10 +2773,11 @@ static int decode_attr_space_used(struct xdr_stream *xdr, uint32_t *bitmap, uint
2748 READ_BUF(8); 2773 READ_BUF(8);
2749 READ64(*used); 2774 READ64(*used);
2750 bitmap[1] &= ~FATTR4_WORD1_SPACE_USED; 2775 bitmap[1] &= ~FATTR4_WORD1_SPACE_USED;
2776 ret = NFS_ATTR_FATTR_SPACE_USED;
2751 } 2777 }
2752 dprintk("%s: space used=%Lu\n", __func__, 2778 dprintk("%s: space used=%Lu\n", __func__,
2753 (unsigned long long)*used); 2779 (unsigned long long)*used);
2754 return 0; 2780 return ret;
2755} 2781}
2756 2782
2757static int decode_attr_time(struct xdr_stream *xdr, struct timespec *time) 2783static int decode_attr_time(struct xdr_stream *xdr, struct timespec *time)
@@ -2778,6 +2804,8 @@ static int decode_attr_time_access(struct xdr_stream *xdr, uint32_t *bitmap, str
2778 return -EIO; 2804 return -EIO;
2779 if (likely(bitmap[1] & FATTR4_WORD1_TIME_ACCESS)) { 2805 if (likely(bitmap[1] & FATTR4_WORD1_TIME_ACCESS)) {
2780 status = decode_attr_time(xdr, time); 2806 status = decode_attr_time(xdr, time);
2807 if (status == 0)
2808 status = NFS_ATTR_FATTR_ATIME;
2781 bitmap[1] &= ~FATTR4_WORD1_TIME_ACCESS; 2809 bitmap[1] &= ~FATTR4_WORD1_TIME_ACCESS;
2782 } 2810 }
2783 dprintk("%s: atime=%ld\n", __func__, (long)time->tv_sec); 2811 dprintk("%s: atime=%ld\n", __func__, (long)time->tv_sec);
@@ -2794,6 +2822,8 @@ static int decode_attr_time_metadata(struct xdr_stream *xdr, uint32_t *bitmap, s
2794 return -EIO; 2822 return -EIO;
2795 if (likely(bitmap[1] & FATTR4_WORD1_TIME_METADATA)) { 2823 if (likely(bitmap[1] & FATTR4_WORD1_TIME_METADATA)) {
2796 status = decode_attr_time(xdr, time); 2824 status = decode_attr_time(xdr, time);
2825 if (status == 0)
2826 status = NFS_ATTR_FATTR_CTIME;
2797 bitmap[1] &= ~FATTR4_WORD1_TIME_METADATA; 2827 bitmap[1] &= ~FATTR4_WORD1_TIME_METADATA;
2798 } 2828 }
2799 dprintk("%s: ctime=%ld\n", __func__, (long)time->tv_sec); 2829 dprintk("%s: ctime=%ld\n", __func__, (long)time->tv_sec);
@@ -2810,6 +2840,8 @@ static int decode_attr_time_modify(struct xdr_stream *xdr, uint32_t *bitmap, str
2810 return -EIO; 2840 return -EIO;
2811 if (likely(bitmap[1] & FATTR4_WORD1_TIME_MODIFY)) { 2841 if (likely(bitmap[1] & FATTR4_WORD1_TIME_MODIFY)) {
2812 status = decode_attr_time(xdr, time); 2842 status = decode_attr_time(xdr, time);
2843 if (status == 0)
2844 status = NFS_ATTR_FATTR_MTIME;
2813 bitmap[1] &= ~FATTR4_WORD1_TIME_MODIFY; 2845 bitmap[1] &= ~FATTR4_WORD1_TIME_MODIFY;
2814 } 2846 }
2815 dprintk("%s: mtime=%ld\n", __func__, (long)time->tv_sec); 2847 dprintk("%s: mtime=%ld\n", __func__, (long)time->tv_sec);
@@ -2994,63 +3026,116 @@ static int decode_getfattr(struct xdr_stream *xdr, struct nfs_fattr *fattr, cons
2994 uint32_t attrlen, 3026 uint32_t attrlen,
2995 bitmap[2] = {0}, 3027 bitmap[2] = {0},
2996 type; 3028 type;
2997 int status, fmode = 0; 3029 int status;
3030 umode_t fmode = 0;
2998 uint64_t fileid; 3031 uint64_t fileid;
2999 3032
3000 if ((status = decode_op_hdr(xdr, OP_GETATTR)) != 0) 3033 status = decode_op_hdr(xdr, OP_GETATTR);
3001 goto xdr_error; 3034 if (status < 0)
3002 if ((status = decode_attr_bitmap(xdr, bitmap)) != 0)
3003 goto xdr_error; 3035 goto xdr_error;
3004 3036
3005 fattr->bitmap[0] = bitmap[0]; 3037 status = decode_attr_bitmap(xdr, bitmap);
3006 fattr->bitmap[1] = bitmap[1]; 3038 if (status < 0)
3039 goto xdr_error;
3007 3040
3008 if ((status = decode_attr_length(xdr, &attrlen, &savep)) != 0) 3041 status = decode_attr_length(xdr, &attrlen, &savep);
3042 if (status < 0)
3009 goto xdr_error; 3043 goto xdr_error;
3010 3044
3011 3045
3012 if ((status = decode_attr_type(xdr, bitmap, &type)) != 0) 3046 status = decode_attr_type(xdr, bitmap, &type);
3047 if (status < 0)
3013 goto xdr_error; 3048 goto xdr_error;
3014 fattr->type = nfs_type2fmt[type].nfs2type; 3049 fattr->mode = 0;
3015 fmode = nfs_type2fmt[type].mode; 3050 if (status != 0) {
3051 fattr->mode |= nfs_type2fmt[type];
3052 fattr->valid |= status;
3053 }
3016 3054
3017 if ((status = decode_attr_change(xdr, bitmap, &fattr->change_attr)) != 0) 3055 status = decode_attr_change(xdr, bitmap, &fattr->change_attr);
3056 if (status < 0)
3018 goto xdr_error; 3057 goto xdr_error;
3019 if ((status = decode_attr_size(xdr, bitmap, &fattr->size)) != 0) 3058 fattr->valid |= status;
3059
3060 status = decode_attr_size(xdr, bitmap, &fattr->size);
3061 if (status < 0)
3020 goto xdr_error; 3062 goto xdr_error;
3021 if ((status = decode_attr_fsid(xdr, bitmap, &fattr->fsid)) != 0) 3063 fattr->valid |= status;
3064
3065 status = decode_attr_fsid(xdr, bitmap, &fattr->fsid);
3066 if (status < 0)
3022 goto xdr_error; 3067 goto xdr_error;
3023 if ((status = decode_attr_fileid(xdr, bitmap, &fattr->fileid)) != 0) 3068 fattr->valid |= status;
3069
3070 status = decode_attr_fileid(xdr, bitmap, &fattr->fileid);
3071 if (status < 0)
3024 goto xdr_error; 3072 goto xdr_error;
3025 if ((status = decode_attr_fs_locations(xdr, bitmap, container_of(fattr, 3073 fattr->valid |= status;
3074
3075 status = decode_attr_fs_locations(xdr, bitmap, container_of(fattr,
3026 struct nfs4_fs_locations, 3076 struct nfs4_fs_locations,
3027 fattr))) != 0) 3077 fattr));
3078 if (status < 0)
3028 goto xdr_error; 3079 goto xdr_error;
3029 if ((status = decode_attr_mode(xdr, bitmap, &fattr->mode)) != 0) 3080 fattr->valid |= status;
3081
3082 status = decode_attr_mode(xdr, bitmap, &fmode);
3083 if (status < 0)
3030 goto xdr_error; 3084 goto xdr_error;
3031 fattr->mode |= fmode; 3085 if (status != 0) {
3032 if ((status = decode_attr_nlink(xdr, bitmap, &fattr->nlink)) != 0) 3086 fattr->mode |= fmode;
3087 fattr->valid |= status;
3088 }
3089
3090 status = decode_attr_nlink(xdr, bitmap, &fattr->nlink);
3091 if (status < 0)
3033 goto xdr_error; 3092 goto xdr_error;
3034 if ((status = decode_attr_owner(xdr, bitmap, server->nfs_client, &fattr->uid)) != 0) 3093 fattr->valid |= status;
3094
3095 status = decode_attr_owner(xdr, bitmap, server->nfs_client, &fattr->uid);
3096 if (status < 0)
3035 goto xdr_error; 3097 goto xdr_error;
3036 if ((status = decode_attr_group(xdr, bitmap, server->nfs_client, &fattr->gid)) != 0) 3098 fattr->valid |= status;
3099
3100 status = decode_attr_group(xdr, bitmap, server->nfs_client, &fattr->gid);
3101 if (status < 0)
3037 goto xdr_error; 3102 goto xdr_error;
3038 if ((status = decode_attr_rdev(xdr, bitmap, &fattr->rdev)) != 0) 3103 fattr->valid |= status;
3104
3105 status = decode_attr_rdev(xdr, bitmap, &fattr->rdev);
3106 if (status < 0)
3039 goto xdr_error; 3107 goto xdr_error;
3040 if ((status = decode_attr_space_used(xdr, bitmap, &fattr->du.nfs3.used)) != 0) 3108 fattr->valid |= status;
3109
3110 status = decode_attr_space_used(xdr, bitmap, &fattr->du.nfs3.used);
3111 if (status < 0)
3041 goto xdr_error; 3112 goto xdr_error;
3042 if ((status = decode_attr_time_access(xdr, bitmap, &fattr->atime)) != 0) 3113 fattr->valid |= status;
3114
3115 status = decode_attr_time_access(xdr, bitmap, &fattr->atime);
3116 if (status < 0)
3043 goto xdr_error; 3117 goto xdr_error;
3044 if ((status = decode_attr_time_metadata(xdr, bitmap, &fattr->ctime)) != 0) 3118 fattr->valid |= status;
3119
3120 status = decode_attr_time_metadata(xdr, bitmap, &fattr->ctime);
3121 if (status < 0)
3045 goto xdr_error; 3122 goto xdr_error;
3046 if ((status = decode_attr_time_modify(xdr, bitmap, &fattr->mtime)) != 0) 3123 fattr->valid |= status;
3124
3125 status = decode_attr_time_modify(xdr, bitmap, &fattr->mtime);
3126 if (status < 0)
3047 goto xdr_error; 3127 goto xdr_error;
3048 if ((status = decode_attr_mounted_on_fileid(xdr, bitmap, &fileid)) != 0) 3128 fattr->valid |= status;
3129
3130 status = decode_attr_mounted_on_fileid(xdr, bitmap, &fileid);
3131 if (status < 0)
3049 goto xdr_error; 3132 goto xdr_error;
3050 if (fattr->fileid == 0 && fileid != 0) 3133 if (status != 0 && !(fattr->valid & status)) {
3051 fattr->fileid = fileid; 3134 fattr->fileid = fileid;
3052 if ((status = verify_attr_len(xdr, savep, attrlen)) == 0) 3135 fattr->valid |= status;
3053 fattr->valid = NFS_ATTR_FATTR | NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4; 3136 }
3137
3138 status = verify_attr_len(xdr, savep, attrlen);
3054xdr_error: 3139xdr_error:
3055 dprintk("%s: xdr returned %d\n", __func__, -status); 3140 dprintk("%s: xdr returned %d\n", __func__, -status);
3056 return status; 3141 return status;
@@ -4078,9 +4163,7 @@ static int nfs4_xdr_dec_setattr(struct rpc_rqst *rqstp, __be32 *p, struct nfs_se
4078 status = decode_setattr(&xdr, res); 4163 status = decode_setattr(&xdr, res);
4079 if (status) 4164 if (status)
4080 goto out; 4165 goto out;
4081 status = decode_getfattr(&xdr, res->fattr, res->server); 4166 decode_getfattr(&xdr, res->fattr, res->server);
4082 if (status == NFS4ERR_DELAY)
4083 status = 0;
4084out: 4167out:
4085 return status; 4168 return status;
4086} 4169}
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index 7f079209d70a..e2975939126a 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -176,17 +176,6 @@ void nfs_release_request(struct nfs_page *req)
176 kref_put(&req->wb_kref, nfs_free_request); 176 kref_put(&req->wb_kref, nfs_free_request);
177} 177}
178 178
179static int nfs_wait_bit_killable(void *word)
180{
181 int ret = 0;
182
183 if (fatal_signal_pending(current))
184 ret = -ERESTARTSYS;
185 else
186 schedule();
187 return ret;
188}
189
190/** 179/**
191 * nfs_wait_on_request - Wait for a request to complete. 180 * nfs_wait_on_request - Wait for a request to complete.
192 * @req: request to wait upon. 181 * @req: request to wait upon.
diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
index 193465210d7c..7be72d90d49d 100644
--- a/fs/nfs/proc.c
+++ b/fs/nfs/proc.c
@@ -663,4 +663,5 @@ const struct nfs_rpc_ops nfs_v2_clientops = {
663 .commit_setup = nfs_proc_commit_setup, 663 .commit_setup = nfs_proc_commit_setup,
664 .lock = nfs_proc_lock, 664 .lock = nfs_proc_lock,
665 .lock_check_bounds = nfs_lock_check_bounds, 665 .lock_check_bounds = nfs_lock_check_bounds,
666 .close_context = nfs_close_context,
666}; 667};
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index d6686f4786dc..0942fcbbad3c 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -1018,6 +1018,7 @@ static int nfs_parse_mount_options(char *raw,
1018 case Opt_rdma: 1018 case Opt_rdma:
1019 mnt->flags |= NFS_MOUNT_TCP; /* for side protocols */ 1019 mnt->flags |= NFS_MOUNT_TCP; /* for side protocols */
1020 mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA; 1020 mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA;
1021 xprt_load_transport(p);
1021 break; 1022 break;
1022 case Opt_acl: 1023 case Opt_acl:
1023 mnt->flags &= ~NFS_MOUNT_NOACL; 1024 mnt->flags &= ~NFS_MOUNT_NOACL;
@@ -1205,12 +1206,14 @@ static int nfs_parse_mount_options(char *raw,
1205 /* vector side protocols to TCP */ 1206 /* vector side protocols to TCP */
1206 mnt->flags |= NFS_MOUNT_TCP; 1207 mnt->flags |= NFS_MOUNT_TCP;
1207 mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA; 1208 mnt->nfs_server.protocol = XPRT_TRANSPORT_RDMA;
1209 xprt_load_transport(string);
1208 break; 1210 break;
1209 default: 1211 default:
1210 errors++; 1212 errors++;
1211 dfprintk(MOUNT, "NFS: unrecognized " 1213 dfprintk(MOUNT, "NFS: unrecognized "
1212 "transport protocol\n"); 1214 "transport protocol\n");
1213 } 1215 }
1216 kfree(string);
1214 break; 1217 break;
1215 case Opt_mountproto: 1218 case Opt_mountproto:
1216 string = match_strdup(args); 1219 string = match_strdup(args);
@@ -1218,7 +1221,6 @@ static int nfs_parse_mount_options(char *raw,
1218 goto out_nomem; 1221 goto out_nomem;
1219 token = match_token(string, 1222 token = match_token(string,
1220 nfs_xprt_protocol_tokens, args); 1223 nfs_xprt_protocol_tokens, args);
1221 kfree(string);
1222 1224
1223 switch (token) { 1225 switch (token) {
1224 case Opt_xprt_udp: 1226 case Opt_xprt_udp:
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 9f9845859fc1..e560a78995a3 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -313,19 +313,34 @@ static int nfs_writepages_callback(struct page *page, struct writeback_control *
313int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc) 313int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
314{ 314{
315 struct inode *inode = mapping->host; 315 struct inode *inode = mapping->host;
316 unsigned long *bitlock = &NFS_I(inode)->flags;
316 struct nfs_pageio_descriptor pgio; 317 struct nfs_pageio_descriptor pgio;
317 int err; 318 int err;
318 319
320 /* Stop dirtying of new pages while we sync */
321 err = wait_on_bit_lock(bitlock, NFS_INO_FLUSHING,
322 nfs_wait_bit_killable, TASK_KILLABLE);
323 if (err)
324 goto out_err;
325
319 nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES); 326 nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);
320 327
321 nfs_pageio_init_write(&pgio, inode, wb_priority(wbc)); 328 nfs_pageio_init_write(&pgio, inode, wb_priority(wbc));
322 err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio); 329 err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio);
323 nfs_pageio_complete(&pgio); 330 nfs_pageio_complete(&pgio);
331
332 clear_bit_unlock(NFS_INO_FLUSHING, bitlock);
333 smp_mb__after_clear_bit();
334 wake_up_bit(bitlock, NFS_INO_FLUSHING);
335
324 if (err < 0) 336 if (err < 0)
325 return err; 337 goto out_err;
326 if (pgio.pg_error < 0) 338 err = pgio.pg_error;
327 return pgio.pg_error; 339 if (err < 0)
340 goto out_err;
328 return 0; 341 return 0;
342out_err:
343 return err;
329} 344}
330 345
331/* 346/*
@@ -404,7 +419,6 @@ nfs_mark_request_commit(struct nfs_page *req)
404 struct nfs_inode *nfsi = NFS_I(inode); 419 struct nfs_inode *nfsi = NFS_I(inode);
405 420
406 spin_lock(&inode->i_lock); 421 spin_lock(&inode->i_lock);
407 nfsi->ncommit++;
408 set_bit(PG_CLEAN, &(req)->wb_flags); 422 set_bit(PG_CLEAN, &(req)->wb_flags);
409 radix_tree_tag_set(&nfsi->nfs_page_tree, 423 radix_tree_tag_set(&nfsi->nfs_page_tree,
410 req->wb_index, 424 req->wb_index,
@@ -524,6 +538,12 @@ static void nfs_cancel_commit_list(struct list_head *head)
524} 538}
525 539
526#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4) 540#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
541static int
542nfs_need_commit(struct nfs_inode *nfsi)
543{
544 return radix_tree_tagged(&nfsi->nfs_page_tree, NFS_PAGE_TAG_COMMIT);
545}
546
527/* 547/*
528 * nfs_scan_commit - Scan an inode for commit requests 548 * nfs_scan_commit - Scan an inode for commit requests
529 * @inode: NFS inode to scan 549 * @inode: NFS inode to scan
@@ -538,16 +558,18 @@ static int
538nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages) 558nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages)
539{ 559{
540 struct nfs_inode *nfsi = NFS_I(inode); 560 struct nfs_inode *nfsi = NFS_I(inode);
541 int res = 0;
542 561
543 if (nfsi->ncommit != 0) { 562 if (!nfs_need_commit(nfsi))
544 res = nfs_scan_list(nfsi, dst, idx_start, npages, 563 return 0;
545 NFS_PAGE_TAG_COMMIT); 564
546 nfsi->ncommit -= res; 565 return nfs_scan_list(nfsi, dst, idx_start, npages, NFS_PAGE_TAG_COMMIT);
547 }
548 return res;
549} 566}
550#else 567#else
568static inline int nfs_need_commit(struct nfs_inode *nfsi)
569{
570 return 0;
571}
572
551static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages) 573static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, unsigned int npages)
552{ 574{
553 return 0; 575 return 0;
@@ -820,7 +842,7 @@ static int nfs_write_rpcsetup(struct nfs_page *req,
820 data->args.stable = NFS_UNSTABLE; 842 data->args.stable = NFS_UNSTABLE;
821 if (how & FLUSH_STABLE) { 843 if (how & FLUSH_STABLE) {
822 data->args.stable = NFS_DATA_SYNC; 844 data->args.stable = NFS_DATA_SYNC;
823 if (!NFS_I(inode)->ncommit) 845 if (!nfs_need_commit(NFS_I(inode)))
824 data->args.stable = NFS_FILE_SYNC; 846 data->args.stable = NFS_FILE_SYNC;
825 } 847 }
826 848
@@ -1425,18 +1447,13 @@ static int nfs_write_mapping(struct address_space *mapping, int how)
1425{ 1447{
1426 struct writeback_control wbc = { 1448 struct writeback_control wbc = {
1427 .bdi = mapping->backing_dev_info, 1449 .bdi = mapping->backing_dev_info,
1428 .sync_mode = WB_SYNC_NONE, 1450 .sync_mode = WB_SYNC_ALL,
1429 .nr_to_write = LONG_MAX, 1451 .nr_to_write = LONG_MAX,
1430 .range_start = 0, 1452 .range_start = 0,
1431 .range_end = LLONG_MAX, 1453 .range_end = LLONG_MAX,
1432 .for_writepages = 1, 1454 .for_writepages = 1,
1433 }; 1455 };
1434 int ret;
1435 1456
1436 ret = __nfs_write_mapping(mapping, &wbc, how);
1437 if (ret < 0)
1438 return ret;
1439 wbc.sync_mode = WB_SYNC_ALL;
1440 return __nfs_write_mapping(mapping, &wbc, how); 1457 return __nfs_write_mapping(mapping, &wbc, how);
1441} 1458}
1442 1459
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index 3d93b2064ce5..a4ed8644d69c 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -938,10 +938,12 @@ static ssize_t __write_ports(struct file *file, char *buf, size_t size)
938 char transport[16]; 938 char transport[16];
939 int port; 939 int port;
940 if (sscanf(buf, "%15s %4d", transport, &port) == 2) { 940 if (sscanf(buf, "%15s %4d", transport, &port) == 2) {
941 if (port < 1 || port > 65535)
942 return -EINVAL;
941 err = nfsd_create_serv(); 943 err = nfsd_create_serv();
942 if (!err) { 944 if (!err) {
943 err = svc_create_xprt(nfsd_serv, 945 err = svc_create_xprt(nfsd_serv,
944 transport, port, 946 transport, PF_INET, port,
945 SVC_SOCK_ANONYMOUS); 947 SVC_SOCK_ANONYMOUS);
946 if (err == -ENOENT) 948 if (err == -ENOENT)
947 /* Give a reasonable perror msg for 949 /* Give a reasonable perror msg for
@@ -960,7 +962,7 @@ static ssize_t __write_ports(struct file *file, char *buf, size_t size)
960 char transport[16]; 962 char transport[16];
961 int port; 963 int port;
962 if (sscanf(&buf[1], "%15s %4d", transport, &port) == 2) { 964 if (sscanf(&buf[1], "%15s %4d", transport, &port) == 2) {
963 if (port == 0) 965 if (port < 1 || port > 65535)
964 return -EINVAL; 966 return -EINVAL;
965 if (nfsd_serv) { 967 if (nfsd_serv) {
966 xprt = svc_find_xprt(nfsd_serv, transport, 968 xprt = svc_find_xprt(nfsd_serv, transport,
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index 07e4f5d7baa8..bc3567bab8c4 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -229,7 +229,6 @@ int nfsd_create_serv(void)
229 229
230 atomic_set(&nfsd_busy, 0); 230 atomic_set(&nfsd_busy, 0);
231 nfsd_serv = svc_create_pooled(&nfsd_program, nfsd_max_blksize, 231 nfsd_serv = svc_create_pooled(&nfsd_program, nfsd_max_blksize,
232 AF_INET,
233 nfsd_last_thread, nfsd, THIS_MODULE); 232 nfsd_last_thread, nfsd, THIS_MODULE);
234 if (nfsd_serv == NULL) 233 if (nfsd_serv == NULL)
235 err = -ENOMEM; 234 err = -ENOMEM;
@@ -244,7 +243,7 @@ static int nfsd_init_socks(int port)
244 if (!list_empty(&nfsd_serv->sv_permsocks)) 243 if (!list_empty(&nfsd_serv->sv_permsocks))
245 return 0; 244 return 0;
246 245
247 error = svc_create_xprt(nfsd_serv, "udp", port, 246 error = svc_create_xprt(nfsd_serv, "udp", PF_INET, port,
248 SVC_SOCK_DEFAULTS); 247 SVC_SOCK_DEFAULTS);
249 if (error < 0) 248 if (error < 0)
250 return error; 249 return error;
@@ -253,7 +252,7 @@ static int nfsd_init_socks(int port)
253 if (error < 0) 252 if (error < 0)
254 return error; 253 return error;
255 254
256 error = svc_create_xprt(nfsd_serv, "tcp", port, 255 error = svc_create_xprt(nfsd_serv, "tcp", PF_INET, port,
257 SVC_SOCK_DEFAULTS); 256 SVC_SOCK_DEFAULTS);
258 if (error < 0) 257 if (error < 0)
259 return error; 258 return error;
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 8cc8807f77d6..bde2557c2a9c 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -166,8 +166,7 @@ struct nfs_inode {
166 */ 166 */
167 struct radix_tree_root nfs_page_tree; 167 struct radix_tree_root nfs_page_tree;
168 168
169 unsigned long ncommit, 169 unsigned long npages;
170 npages;
171 170
172 /* Open contexts for shared mmap writes */ 171 /* Open contexts for shared mmap writes */
173 struct list_head open_files; 172 struct list_head open_files;
@@ -207,6 +206,7 @@ struct nfs_inode {
207#define NFS_INO_STALE (1) /* possible stale inode */ 206#define NFS_INO_STALE (1) /* possible stale inode */
208#define NFS_INO_ACL_LRU_SET (2) /* Inode is on the LRU list */ 207#define NFS_INO_ACL_LRU_SET (2) /* Inode is on the LRU list */
209#define NFS_INO_MOUNTPOINT (3) /* inode is remote mountpoint */ 208#define NFS_INO_MOUNTPOINT (3) /* inode is remote mountpoint */
209#define NFS_INO_FLUSHING (4) /* inode is flushing out data */
210 210
211static inline struct nfs_inode *NFS_I(const struct inode *inode) 211static inline struct nfs_inode *NFS_I(const struct inode *inode)
212{ 212{
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 9bb81aec91cf..29b1e40dce99 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -106,6 +106,11 @@ struct nfs_server {
106 u32 attr_bitmask[2];/* V4 bitmask representing the set 106 u32 attr_bitmask[2];/* V4 bitmask representing the set
107 of attributes supported on this 107 of attributes supported on this
108 filesystem */ 108 filesystem */
109 u32 cache_consistency_bitmask[2];
110 /* V4 bitmask representing the subset
111 of change attribute, size, ctime
112 and mtime attributes supported by
113 the server */
109 u32 acl_bitmask; /* V4 bitmask representing the ACEs 114 u32 acl_bitmask; /* V4 bitmask representing the ACEs
110 that are supported on this 115 that are supported on this
111 filesystem */ 116 filesystem */
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 43a713fce11c..b89c34e40bc2 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -27,12 +27,8 @@ static inline int nfs_fsid_equal(const struct nfs_fsid *a, const struct nfs_fsid
27} 27}
28 28
29struct nfs_fattr { 29struct nfs_fattr {
30 unsigned short valid; /* which fields are valid */ 30 unsigned int valid; /* which fields are valid */
31 __u64 pre_size; /* pre_op_attr.size */ 31 umode_t mode;
32 struct timespec pre_mtime; /* pre_op_attr.mtime */
33 struct timespec pre_ctime; /* pre_op_attr.ctime */
34 enum nfs_ftype type; /* always use NFSv2 types */
35 __u32 mode;
36 __u32 nlink; 32 __u32 nlink;
37 __u32 uid; 33 __u32 uid;
38 __u32 gid; 34 __u32 gid;
@@ -52,19 +48,55 @@ struct nfs_fattr {
52 struct timespec atime; 48 struct timespec atime;
53 struct timespec mtime; 49 struct timespec mtime;
54 struct timespec ctime; 50 struct timespec ctime;
55 __u32 bitmap[2]; /* NFSv4 returned attribute bitmap */
56 __u64 change_attr; /* NFSv4 change attribute */ 51 __u64 change_attr; /* NFSv4 change attribute */
57 __u64 pre_change_attr;/* pre-op NFSv4 change attribute */ 52 __u64 pre_change_attr;/* pre-op NFSv4 change attribute */
53 __u64 pre_size; /* pre_op_attr.size */
54 struct timespec pre_mtime; /* pre_op_attr.mtime */
55 struct timespec pre_ctime; /* pre_op_attr.ctime */
58 unsigned long time_start; 56 unsigned long time_start;
59 unsigned long gencount; 57 unsigned long gencount;
60}; 58};
61 59
62#define NFS_ATTR_WCC 0x0001 /* pre-op WCC data */ 60#define NFS_ATTR_FATTR_TYPE (1U << 0)
63#define NFS_ATTR_FATTR 0x0002 /* post-op attributes */ 61#define NFS_ATTR_FATTR_MODE (1U << 1)
64#define NFS_ATTR_FATTR_V3 0x0004 /* NFSv3 attributes */ 62#define NFS_ATTR_FATTR_NLINK (1U << 2)
65#define NFS_ATTR_FATTR_V4 0x0008 /* NFSv4 change attribute */ 63#define NFS_ATTR_FATTR_OWNER (1U << 3)
66#define NFS_ATTR_WCC_V4 0x0010 /* pre-op change attribute */ 64#define NFS_ATTR_FATTR_GROUP (1U << 4)
67#define NFS_ATTR_FATTR_V4_REFERRAL 0x0020 /* NFSv4 referral */ 65#define NFS_ATTR_FATTR_RDEV (1U << 5)
66#define NFS_ATTR_FATTR_SIZE (1U << 6)
67#define NFS_ATTR_FATTR_PRESIZE (1U << 7)
68#define NFS_ATTR_FATTR_BLOCKS_USED (1U << 8)
69#define NFS_ATTR_FATTR_SPACE_USED (1U << 9)
70#define NFS_ATTR_FATTR_FSID (1U << 10)
71#define NFS_ATTR_FATTR_FILEID (1U << 11)
72#define NFS_ATTR_FATTR_ATIME (1U << 12)
73#define NFS_ATTR_FATTR_MTIME (1U << 13)
74#define NFS_ATTR_FATTR_CTIME (1U << 14)
75#define NFS_ATTR_FATTR_PREMTIME (1U << 15)
76#define NFS_ATTR_FATTR_PRECTIME (1U << 16)
77#define NFS_ATTR_FATTR_CHANGE (1U << 17)
78#define NFS_ATTR_FATTR_PRECHANGE (1U << 18)
79#define NFS_ATTR_FATTR_V4_REFERRAL (1U << 19) /* NFSv4 referral */
80
81#define NFS_ATTR_FATTR (NFS_ATTR_FATTR_TYPE \
82 | NFS_ATTR_FATTR_MODE \
83 | NFS_ATTR_FATTR_NLINK \
84 | NFS_ATTR_FATTR_OWNER \
85 | NFS_ATTR_FATTR_GROUP \
86 | NFS_ATTR_FATTR_RDEV \
87 | NFS_ATTR_FATTR_SIZE \
88 | NFS_ATTR_FATTR_FSID \
89 | NFS_ATTR_FATTR_FILEID \
90 | NFS_ATTR_FATTR_ATIME \
91 | NFS_ATTR_FATTR_MTIME \
92 | NFS_ATTR_FATTR_CTIME)
93#define NFS_ATTR_FATTR_V2 (NFS_ATTR_FATTR \
94 | NFS_ATTR_FATTR_BLOCKS_USED)
95#define NFS_ATTR_FATTR_V3 (NFS_ATTR_FATTR \
96 | NFS_ATTR_FATTR_SPACE_USED)
97#define NFS_ATTR_FATTR_V4 (NFS_ATTR_FATTR \
98 | NFS_ATTR_FATTR_SPACE_USED \
99 | NFS_ATTR_FATTR_CHANGE)
68 100
69/* 101/*
70 * Info on the file system 102 * Info on the file system
@@ -836,6 +868,7 @@ struct nfs_rpc_ops {
836 int (*lock)(struct file *, int, struct file_lock *); 868 int (*lock)(struct file *, int, struct file_lock *);
837 int (*lock_check_bounds)(const struct file_lock *); 869 int (*lock_check_bounds)(const struct file_lock *);
838 void (*clear_acl_cache)(struct inode *); 870 void (*clear_acl_cache)(struct inode *);
871 void (*close_context)(struct nfs_open_context *ctx, int);
839}; 872};
840 873
841/* 874/*
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 3435d24bfe55..d3a4c0231933 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -69,7 +69,6 @@ struct svc_serv {
69 struct list_head sv_tempsocks; /* all temporary sockets */ 69 struct list_head sv_tempsocks; /* all temporary sockets */
70 int sv_tmpcnt; /* count of temporary sockets */ 70 int sv_tmpcnt; /* count of temporary sockets */
71 struct timer_list sv_temptimer; /* timer for aging temporary sockets */ 71 struct timer_list sv_temptimer; /* timer for aging temporary sockets */
72 sa_family_t sv_family; /* listener's address family */
73 72
74 char * sv_name; /* service name */ 73 char * sv_name; /* service name */
75 74
@@ -385,19 +384,19 @@ struct svc_procedure {
385/* 384/*
386 * Function prototypes. 385 * Function prototypes.
387 */ 386 */
388struct svc_serv *svc_create(struct svc_program *, unsigned int, sa_family_t, 387struct svc_serv *svc_create(struct svc_program *, unsigned int,
389 void (*shutdown)(struct svc_serv *)); 388 void (*shutdown)(struct svc_serv *));
390struct svc_rqst *svc_prepare_thread(struct svc_serv *serv, 389struct svc_rqst *svc_prepare_thread(struct svc_serv *serv,
391 struct svc_pool *pool); 390 struct svc_pool *pool);
392void svc_exit_thread(struct svc_rqst *); 391void svc_exit_thread(struct svc_rqst *);
393struct svc_serv * svc_create_pooled(struct svc_program *, unsigned int, 392struct svc_serv * svc_create_pooled(struct svc_program *, unsigned int,
394 sa_family_t, void (*shutdown)(struct svc_serv *), 393 void (*shutdown)(struct svc_serv *),
395 svc_thread_fn, struct module *); 394 svc_thread_fn, struct module *);
396int svc_set_num_threads(struct svc_serv *, struct svc_pool *, int); 395int svc_set_num_threads(struct svc_serv *, struct svc_pool *, int);
397void svc_destroy(struct svc_serv *); 396void svc_destroy(struct svc_serv *);
398int svc_process(struct svc_rqst *); 397int svc_process(struct svc_rqst *);
399int svc_register(const struct svc_serv *, const unsigned short, 398int svc_register(const struct svc_serv *, const int,
400 const unsigned short); 399 const unsigned short, const unsigned short);
401 400
402void svc_wake_up(struct svc_serv *); 401void svc_wake_up(struct svc_serv *);
403void svc_reserve(struct svc_rqst *rqstp, int space); 402void svc_reserve(struct svc_rqst *rqstp, int space);
diff --git a/include/linux/sunrpc/svc_xprt.h b/include/linux/sunrpc/svc_xprt.h
index 0127daca4354..0d9cb6ef28b0 100644
--- a/include/linux/sunrpc/svc_xprt.h
+++ b/include/linux/sunrpc/svc_xprt.h
@@ -71,7 +71,8 @@ int svc_reg_xprt_class(struct svc_xprt_class *);
71void svc_unreg_xprt_class(struct svc_xprt_class *); 71void svc_unreg_xprt_class(struct svc_xprt_class *);
72void svc_xprt_init(struct svc_xprt_class *, struct svc_xprt *, 72void svc_xprt_init(struct svc_xprt_class *, struct svc_xprt *,
73 struct svc_serv *); 73 struct svc_serv *);
74int svc_create_xprt(struct svc_serv *, char *, unsigned short, int); 74int svc_create_xprt(struct svc_serv *, const char *, const int,
75 const unsigned short, int);
75void svc_xprt_enqueue(struct svc_xprt *xprt); 76void svc_xprt_enqueue(struct svc_xprt *xprt);
76void svc_xprt_received(struct svc_xprt *); 77void svc_xprt_received(struct svc_xprt *);
77void svc_xprt_put(struct svc_xprt *xprt); 78void svc_xprt_put(struct svc_xprt *xprt);
@@ -80,7 +81,8 @@ void svc_close_xprt(struct svc_xprt *xprt);
80void svc_delete_xprt(struct svc_xprt *xprt); 81void svc_delete_xprt(struct svc_xprt *xprt);
81int svc_port_is_privileged(struct sockaddr *sin); 82int svc_port_is_privileged(struct sockaddr *sin);
82int svc_print_xprts(char *buf, int maxlen); 83int svc_print_xprts(char *buf, int maxlen);
83struct svc_xprt *svc_find_xprt(struct svc_serv *, char *, int, int); 84struct svc_xprt *svc_find_xprt(struct svc_serv *serv, const char *xcl_name,
85 const sa_family_t af, const unsigned short port);
84int svc_xprt_names(struct svc_serv *serv, char *buf, int buflen); 86int svc_xprt_names(struct svc_serv *serv, char *buf, int buflen);
85 87
86static inline void svc_xprt_get(struct svc_xprt *xprt) 88static inline void svc_xprt_get(struct svc_xprt *xprt)
@@ -88,29 +90,32 @@ static inline void svc_xprt_get(struct svc_xprt *xprt)
88 kref_get(&xprt->xpt_ref); 90 kref_get(&xprt->xpt_ref);
89} 91}
90static inline void svc_xprt_set_local(struct svc_xprt *xprt, 92static inline void svc_xprt_set_local(struct svc_xprt *xprt,
91 struct sockaddr *sa, int salen) 93 const struct sockaddr *sa,
94 const size_t salen)
92{ 95{
93 memcpy(&xprt->xpt_local, sa, salen); 96 memcpy(&xprt->xpt_local, sa, salen);
94 xprt->xpt_locallen = salen; 97 xprt->xpt_locallen = salen;
95} 98}
96static inline void svc_xprt_set_remote(struct svc_xprt *xprt, 99static inline void svc_xprt_set_remote(struct svc_xprt *xprt,
97 struct sockaddr *sa, int salen) 100 const struct sockaddr *sa,
101 const size_t salen)
98{ 102{
99 memcpy(&xprt->xpt_remote, sa, salen); 103 memcpy(&xprt->xpt_remote, sa, salen);
100 xprt->xpt_remotelen = salen; 104 xprt->xpt_remotelen = salen;
101} 105}
102static inline unsigned short svc_addr_port(struct sockaddr *sa) 106static inline unsigned short svc_addr_port(const struct sockaddr *sa)
103{ 107{
104 unsigned short ret = 0; 108 const struct sockaddr_in *sin = (const struct sockaddr_in *)sa;
109 const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)sa;
110
105 switch (sa->sa_family) { 111 switch (sa->sa_family) {
106 case AF_INET: 112 case AF_INET:
107 ret = ntohs(((struct sockaddr_in *)sa)->sin_port); 113 return ntohs(sin->sin_port);
108 break;
109 case AF_INET6: 114 case AF_INET6:
110 ret = ntohs(((struct sockaddr_in6 *)sa)->sin6_port); 115 return ntohs(sin6->sin6_port);
111 break;
112 } 116 }
113 return ret; 117
118 return 0;
114} 119}
115 120
116static inline size_t svc_addr_len(struct sockaddr *sa) 121static inline size_t svc_addr_len(struct sockaddr *sa)
@@ -124,36 +129,39 @@ static inline size_t svc_addr_len(struct sockaddr *sa)
124 return -EAFNOSUPPORT; 129 return -EAFNOSUPPORT;
125} 130}
126 131
127static inline unsigned short svc_xprt_local_port(struct svc_xprt *xprt) 132static inline unsigned short svc_xprt_local_port(const struct svc_xprt *xprt)
128{ 133{
129 return svc_addr_port((struct sockaddr *)&xprt->xpt_local); 134 return svc_addr_port((const struct sockaddr *)&xprt->xpt_local);
130} 135}
131 136
132static inline unsigned short svc_xprt_remote_port(struct svc_xprt *xprt) 137static inline unsigned short svc_xprt_remote_port(const struct svc_xprt *xprt)
133{ 138{
134 return svc_addr_port((struct sockaddr *)&xprt->xpt_remote); 139 return svc_addr_port((const struct sockaddr *)&xprt->xpt_remote);
135} 140}
136 141
137static inline char *__svc_print_addr(struct sockaddr *addr, 142static inline char *__svc_print_addr(const struct sockaddr *addr,
138 char *buf, size_t len) 143 char *buf, const size_t len)
139{ 144{
145 const struct sockaddr_in *sin = (const struct sockaddr_in *)addr;
146 const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)addr;
147
140 switch (addr->sa_family) { 148 switch (addr->sa_family) {
141 case AF_INET: 149 case AF_INET:
142 snprintf(buf, len, "%pI4, port=%u", 150 snprintf(buf, len, "%pI4, port=%u", &sin->sin_addr,
143 &((struct sockaddr_in *)addr)->sin_addr, 151 ntohs(sin->sin_port));
144 ntohs(((struct sockaddr_in *) addr)->sin_port));
145 break; 152 break;
146 153
147 case AF_INET6: 154 case AF_INET6:
148 snprintf(buf, len, "%pI6, port=%u", 155 snprintf(buf, len, "%pI6, port=%u",
149 &((struct sockaddr_in6 *)addr)->sin6_addr, 156 &sin6->sin6_addr,
150 ntohs(((struct sockaddr_in6 *) addr)->sin6_port)); 157 ntohs(sin6->sin6_port));
151 break; 158 break;
152 159
153 default: 160 default:
154 snprintf(buf, len, "unknown address type: %d", addr->sa_family); 161 snprintf(buf, len, "unknown address type: %d", addr->sa_family);
155 break; 162 break;
156 } 163 }
164
157 return buf; 165 return buf;
158} 166}
159#endif /* SUNRPC_SVC_XPRT_H */ 167#endif /* SUNRPC_SVC_XPRT_H */
diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index 11fc71d50c1e..1758d9f5b5c3 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -235,6 +235,7 @@ static inline __be32 *xprt_skip_transport_header(struct rpc_xprt *xprt, __be32 *
235 */ 235 */
236int xprt_register_transport(struct xprt_class *type); 236int xprt_register_transport(struct xprt_class *type);
237int xprt_unregister_transport(struct xprt_class *type); 237int xprt_unregister_transport(struct xprt_class *type);
238int xprt_load_transport(const char *);
238void xprt_set_retrans_timeout_def(struct rpc_task *task); 239void xprt_set_retrans_timeout_def(struct rpc_task *task);
239void xprt_set_retrans_timeout_rtt(struct rpc_task *task); 240void xprt_set_retrans_timeout_rtt(struct rpc_task *task);
240void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status); 241void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status);
@@ -259,6 +260,7 @@ void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie);
259#define XPRT_BOUND (4) 260#define XPRT_BOUND (4)
260#define XPRT_BINDING (5) 261#define XPRT_BINDING (5)
261#define XPRT_CLOSING (6) 262#define XPRT_CLOSING (6)
263#define XPRT_CONNECTION_ABORT (7)
262 264
263static inline void xprt_set_connected(struct rpc_xprt *xprt) 265static inline void xprt_set_connected(struct rpc_xprt *xprt)
264{ 266{
diff --git a/net/sunrpc/Kconfig b/net/sunrpc/Kconfig
index 5592883e1e4a..afd91c78ce8e 100644
--- a/net/sunrpc/Kconfig
+++ b/net/sunrpc/Kconfig
@@ -17,28 +17,6 @@ config SUNRPC_XPRT_RDMA
17 17
18 If unsure, say N. 18 If unsure, say N.
19 19
20config SUNRPC_REGISTER_V4
21 bool "Register local RPC services via rpcbind v4 (EXPERIMENTAL)"
22 depends on SUNRPC && EXPERIMENTAL
23 default n
24 help
25 Sun added support for registering RPC services at an IPv6
26 address by creating two new versions of the rpcbind protocol
27 (RFC 1833).
28
29 This option enables support in the kernel RPC server for
30 registering kernel RPC services via version 4 of the rpcbind
31 protocol. If you enable this option, you must run a portmapper
32 daemon that supports rpcbind protocol version 4.
33
34 Serving NFS over IPv6 from knfsd (the kernel's NFS server)
35 requires that you enable this option and use a portmapper that
36 supports rpcbind version 4.
37
38 If unsure, say N to get traditional behavior (register kernel
39 RPC services using only rpcbind version 2). Distributions
40 using the legacy Linux portmapper daemon must say N here.
41
42config RPCSEC_GSS_KRB5 20config RPCSEC_GSS_KRB5
43 tristate "Secure RPC: Kerberos V mechanism (EXPERIMENTAL)" 21 tristate "Secure RPC: Kerberos V mechanism (EXPERIMENTAL)"
44 depends on SUNRPC && EXPERIMENTAL 22 depends on SUNRPC && EXPERIMENTAL
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 836f15c0c4a3..5abab094441f 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1032,27 +1032,20 @@ call_connect_status(struct rpc_task *task)
1032 dprint_status(task); 1032 dprint_status(task);
1033 1033
1034 task->tk_status = 0; 1034 task->tk_status = 0;
1035 if (status >= 0) { 1035 if (status >= 0 || status == -EAGAIN) {
1036 clnt->cl_stats->netreconn++; 1036 clnt->cl_stats->netreconn++;
1037 task->tk_action = call_transmit; 1037 task->tk_action = call_transmit;
1038 return; 1038 return;
1039 } 1039 }
1040 1040
1041 /* Something failed: remote service port may have changed */
1042 rpc_force_rebind(clnt);
1043
1044 switch (status) { 1041 switch (status) {
1045 case -ENOTCONN:
1046 case -EAGAIN:
1047 task->tk_action = call_bind;
1048 if (!RPC_IS_SOFT(task))
1049 return;
1050 /* if soft mounted, test if we've timed out */ 1042 /* if soft mounted, test if we've timed out */
1051 case -ETIMEDOUT: 1043 case -ETIMEDOUT:
1052 task->tk_action = call_timeout; 1044 task->tk_action = call_timeout;
1053 return; 1045 break;
1046 default:
1047 rpc_exit(task, -EIO);
1054 } 1048 }
1055 rpc_exit(task, -EIO);
1056} 1049}
1057 1050
1058/* 1051/*
@@ -1105,14 +1098,26 @@ static void
1105call_transmit_status(struct rpc_task *task) 1098call_transmit_status(struct rpc_task *task)
1106{ 1099{
1107 task->tk_action = call_status; 1100 task->tk_action = call_status;
1108 /* 1101 switch (task->tk_status) {
1109 * Special case: if we've been waiting on the socket's write_space() 1102 case -EAGAIN:
1110 * callback, then don't call xprt_end_transmit(). 1103 break;
1111 */ 1104 default:
1112 if (task->tk_status == -EAGAIN) 1105 xprt_end_transmit(task);
1113 return; 1106 /*
1114 xprt_end_transmit(task); 1107 * Special cases: if we've been waiting on the
1115 rpc_task_force_reencode(task); 1108 * socket's write_space() callback, or if the
1109 * socket just returned a connection error,
1110 * then hold onto the transport lock.
1111 */
1112 case -ECONNREFUSED:
1113 case -ECONNRESET:
1114 case -ENOTCONN:
1115 case -EHOSTDOWN:
1116 case -EHOSTUNREACH:
1117 case -ENETUNREACH:
1118 case -EPIPE:
1119 rpc_task_force_reencode(task);
1120 }
1116} 1121}
1117 1122
1118/* 1123/*
@@ -1152,9 +1157,12 @@ call_status(struct rpc_task *task)
1152 xprt_conditional_disconnect(task->tk_xprt, 1157 xprt_conditional_disconnect(task->tk_xprt,
1153 req->rq_connect_cookie); 1158 req->rq_connect_cookie);
1154 break; 1159 break;
1160 case -ECONNRESET:
1155 case -ECONNREFUSED: 1161 case -ECONNREFUSED:
1156 case -ENOTCONN:
1157 rpc_force_rebind(clnt); 1162 rpc_force_rebind(clnt);
1163 rpc_delay(task, 3*HZ);
1164 case -EPIPE:
1165 case -ENOTCONN:
1158 task->tk_action = call_bind; 1166 task->tk_action = call_bind;
1159 break; 1167 break;
1160 case -EAGAIN: 1168 case -EAGAIN:
diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c
index 03ae007641e4..beee6da33035 100644
--- a/net/sunrpc/rpcb_clnt.c
+++ b/net/sunrpc/rpcb_clnt.c
@@ -63,9 +63,16 @@ enum {
63 * r_owner 63 * r_owner
64 * 64 *
65 * The "owner" is allowed to unset a service in the rpcbind database. 65 * The "owner" is allowed to unset a service in the rpcbind database.
66 * We always use the following (arbitrary) fixed string. 66 *
67 * For AF_LOCAL SET/UNSET requests, rpcbind treats this string as a
68 * UID which it maps to a local user name via a password lookup.
69 * In all other cases it is ignored.
70 *
71 * For SET/UNSET requests, user space provides a value, even for
72 * network requests, and GETADDR uses an empty string. We follow
73 * those precedents here.
67 */ 74 */
68#define RPCB_OWNER_STRING "rpcb" 75#define RPCB_OWNER_STRING "0"
69#define RPCB_MAXOWNERLEN sizeof(RPCB_OWNER_STRING) 76#define RPCB_MAXOWNERLEN sizeof(RPCB_OWNER_STRING)
70 77
71static void rpcb_getport_done(struct rpc_task *, void *); 78static void rpcb_getport_done(struct rpc_task *, void *);
@@ -124,12 +131,6 @@ static const struct sockaddr_in rpcb_inaddr_loopback = {
124 .sin_port = htons(RPCBIND_PORT), 131 .sin_port = htons(RPCBIND_PORT),
125}; 132};
126 133
127static const struct sockaddr_in6 rpcb_in6addr_loopback = {
128 .sin6_family = AF_INET6,
129 .sin6_addr = IN6ADDR_LOOPBACK_INIT,
130 .sin6_port = htons(RPCBIND_PORT),
131};
132
133static struct rpc_clnt *rpcb_create_local(struct sockaddr *addr, 134static struct rpc_clnt *rpcb_create_local(struct sockaddr *addr,
134 size_t addrlen, u32 version) 135 size_t addrlen, u32 version)
135{ 136{
@@ -176,9 +177,10 @@ static struct rpc_clnt *rpcb_create(char *hostname, struct sockaddr *srvaddr,
176 return rpc_create(&args); 177 return rpc_create(&args);
177} 178}
178 179
179static int rpcb_register_call(struct sockaddr *addr, size_t addrlen, 180static int rpcb_register_call(const u32 version, struct rpc_message *msg)
180 u32 version, struct rpc_message *msg)
181{ 181{
182 struct sockaddr *addr = (struct sockaddr *)&rpcb_inaddr_loopback;
183 size_t addrlen = sizeof(rpcb_inaddr_loopback);
182 struct rpc_clnt *rpcb_clnt; 184 struct rpc_clnt *rpcb_clnt;
183 int result, error = 0; 185 int result, error = 0;
184 186
@@ -192,7 +194,7 @@ static int rpcb_register_call(struct sockaddr *addr, size_t addrlen,
192 error = PTR_ERR(rpcb_clnt); 194 error = PTR_ERR(rpcb_clnt);
193 195
194 if (error < 0) { 196 if (error < 0) {
195 printk(KERN_WARNING "RPC: failed to contact local rpcbind " 197 dprintk("RPC: failed to contact local rpcbind "
196 "server (errno %d).\n", -error); 198 "server (errno %d).\n", -error);
197 return error; 199 return error;
198 } 200 }
@@ -254,25 +256,23 @@ int rpcb_register(u32 prog, u32 vers, int prot, unsigned short port)
254 if (port) 256 if (port)
255 msg.rpc_proc = &rpcb_procedures2[RPCBPROC_SET]; 257 msg.rpc_proc = &rpcb_procedures2[RPCBPROC_SET];
256 258
257 return rpcb_register_call((struct sockaddr *)&rpcb_inaddr_loopback, 259 return rpcb_register_call(RPCBVERS_2, &msg);
258 sizeof(rpcb_inaddr_loopback),
259 RPCBVERS_2, &msg);
260} 260}
261 261
262/* 262/*
263 * Fill in AF_INET family-specific arguments to register 263 * Fill in AF_INET family-specific arguments to register
264 */ 264 */
265static int rpcb_register_netid4(struct sockaddr_in *address_to_register, 265static int rpcb_register_inet4(const struct sockaddr *sap,
266 struct rpc_message *msg) 266 struct rpc_message *msg)
267{ 267{
268 const struct sockaddr_in *sin = (const struct sockaddr_in *)sap;
268 struct rpcbind_args *map = msg->rpc_argp; 269 struct rpcbind_args *map = msg->rpc_argp;
269 unsigned short port = ntohs(address_to_register->sin_port); 270 unsigned short port = ntohs(sin->sin_port);
270 char buf[32]; 271 char buf[32];
271 272
272 /* Construct AF_INET universal address */ 273 /* Construct AF_INET universal address */
273 snprintf(buf, sizeof(buf), "%pI4.%u.%u", 274 snprintf(buf, sizeof(buf), "%pI4.%u.%u",
274 &address_to_register->sin_addr.s_addr, 275 &sin->sin_addr.s_addr, port >> 8, port & 0xff);
275 port >> 8, port & 0xff);
276 map->r_addr = buf; 276 map->r_addr = buf;
277 277
278 dprintk("RPC: %sregistering [%u, %u, %s, '%s'] with " 278 dprintk("RPC: %sregistering [%u, %u, %s, '%s'] with "
@@ -284,29 +284,27 @@ static int rpcb_register_netid4(struct sockaddr_in *address_to_register,
284 if (port) 284 if (port)
285 msg->rpc_proc = &rpcb_procedures4[RPCBPROC_SET]; 285 msg->rpc_proc = &rpcb_procedures4[RPCBPROC_SET];
286 286
287 return rpcb_register_call((struct sockaddr *)&rpcb_inaddr_loopback, 287 return rpcb_register_call(RPCBVERS_4, msg);
288 sizeof(rpcb_inaddr_loopback),
289 RPCBVERS_4, msg);
290} 288}
291 289
292/* 290/*
293 * Fill in AF_INET6 family-specific arguments to register 291 * Fill in AF_INET6 family-specific arguments to register
294 */ 292 */
295static int rpcb_register_netid6(struct sockaddr_in6 *address_to_register, 293static int rpcb_register_inet6(const struct sockaddr *sap,
296 struct rpc_message *msg) 294 struct rpc_message *msg)
297{ 295{
296 const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)sap;
298 struct rpcbind_args *map = msg->rpc_argp; 297 struct rpcbind_args *map = msg->rpc_argp;
299 unsigned short port = ntohs(address_to_register->sin6_port); 298 unsigned short port = ntohs(sin6->sin6_port);
300 char buf[64]; 299 char buf[64];
301 300
302 /* Construct AF_INET6 universal address */ 301 /* Construct AF_INET6 universal address */
303 if (ipv6_addr_any(&address_to_register->sin6_addr)) 302 if (ipv6_addr_any(&sin6->sin6_addr))
304 snprintf(buf, sizeof(buf), "::.%u.%u", 303 snprintf(buf, sizeof(buf), "::.%u.%u",
305 port >> 8, port & 0xff); 304 port >> 8, port & 0xff);
306 else 305 else
307 snprintf(buf, sizeof(buf), "%pI6.%u.%u", 306 snprintf(buf, sizeof(buf), "%pI6.%u.%u",
308 &address_to_register->sin6_addr, 307 &sin6->sin6_addr, port >> 8, port & 0xff);
309 port >> 8, port & 0xff);
310 map->r_addr = buf; 308 map->r_addr = buf;
311 309
312 dprintk("RPC: %sregistering [%u, %u, %s, '%s'] with " 310 dprintk("RPC: %sregistering [%u, %u, %s, '%s'] with "
@@ -318,9 +316,21 @@ static int rpcb_register_netid6(struct sockaddr_in6 *address_to_register,
318 if (port) 316 if (port)
319 msg->rpc_proc = &rpcb_procedures4[RPCBPROC_SET]; 317 msg->rpc_proc = &rpcb_procedures4[RPCBPROC_SET];
320 318
321 return rpcb_register_call((struct sockaddr *)&rpcb_in6addr_loopback, 319 return rpcb_register_call(RPCBVERS_4, msg);
322 sizeof(rpcb_in6addr_loopback), 320}
323 RPCBVERS_4, msg); 321
322static int rpcb_unregister_all_protofamilies(struct rpc_message *msg)
323{
324 struct rpcbind_args *map = msg->rpc_argp;
325
326 dprintk("RPC: unregistering [%u, %u, '%s'] with "
327 "local rpcbind\n",
328 map->r_prog, map->r_vers, map->r_netid);
329
330 map->r_addr = "";
331 msg->rpc_proc = &rpcb_procedures4[RPCBPROC_UNSET];
332
333 return rpcb_register_call(RPCBVERS_4, msg);
324} 334}
325 335
326/** 336/**
@@ -340,10 +350,11 @@ static int rpcb_register_netid6(struct sockaddr_in6 *address_to_register,
340 * invoke this function once for each [program, version, address, 350 * invoke this function once for each [program, version, address,
341 * netid] tuple they wish to advertise. 351 * netid] tuple they wish to advertise.
342 * 352 *
343 * Callers may also unregister RPC services that are no longer 353 * Callers may also unregister RPC services that are registered at a
344 * available by setting the port number in the passed-in address 354 * specific address by setting the port number in @address to zero.
345 * to zero. Callers pass a netid of "" to unregister all 355 * They may unregister all registered protocol families at once for
346 * transport netids associated with [program, version, address]. 356 * a service by passing a NULL @address argument. If @netid is ""
357 * then all netids for [program, version, address] are unregistered.
347 * 358 *
348 * This function uses rpcbind protocol version 4 to contact the 359 * This function uses rpcbind protocol version 4 to contact the
349 * local rpcbind daemon. The local rpcbind daemon must support 360 * local rpcbind daemon. The local rpcbind daemon must support
@@ -378,13 +389,14 @@ int rpcb_v4_register(const u32 program, const u32 version,
378 .rpc_argp = &map, 389 .rpc_argp = &map,
379 }; 390 };
380 391
392 if (address == NULL)
393 return rpcb_unregister_all_protofamilies(&msg);
394
381 switch (address->sa_family) { 395 switch (address->sa_family) {
382 case AF_INET: 396 case AF_INET:
383 return rpcb_register_netid4((struct sockaddr_in *)address, 397 return rpcb_register_inet4(address, &msg);
384 &msg);
385 case AF_INET6: 398 case AF_INET6:
386 return rpcb_register_netid6((struct sockaddr_in6 *)address, 399 return rpcb_register_inet6(address, &msg);
387 &msg);
388 } 400 }
389 401
390 return -EAFNOSUPPORT; 402 return -EAFNOSUPPORT;
@@ -579,7 +591,7 @@ void rpcb_getport_async(struct rpc_task *task)
579 map->r_xprt = xprt_get(xprt); 591 map->r_xprt = xprt_get(xprt);
580 map->r_netid = rpc_peeraddr2str(clnt, RPC_DISPLAY_NETID); 592 map->r_netid = rpc_peeraddr2str(clnt, RPC_DISPLAY_NETID);
581 map->r_addr = rpc_peeraddr2str(rpcb_clnt, RPC_DISPLAY_UNIVERSAL_ADDR); 593 map->r_addr = rpc_peeraddr2str(rpcb_clnt, RPC_DISPLAY_UNIVERSAL_ADDR);
582 map->r_owner = RPCB_OWNER_STRING; /* ignored for GETADDR */ 594 map->r_owner = "";
583 map->r_status = -EIO; 595 map->r_status = -EIO;
584 596
585 child = rpcb_call_async(rpcb_clnt, map, proc); 597 child = rpcb_call_async(rpcb_clnt, map, proc);
@@ -703,11 +715,16 @@ static int rpcb_decode_getaddr(struct rpc_rqst *req, __be32 *p,
703 *portp = 0; 715 *portp = 0;
704 addr_len = ntohl(*p++); 716 addr_len = ntohl(*p++);
705 717
718 if (addr_len == 0) {
719 dprintk("RPC: rpcb_decode_getaddr: "
720 "service is not registered\n");
721 return 0;
722 }
723
706 /* 724 /*
707 * Simple sanity check. The smallest possible universal 725 * Simple sanity check.
708 * address is an IPv4 address string containing 11 bytes.
709 */ 726 */
710 if (addr_len < 11 || addr_len > RPCBIND_MAXUADDRLEN) 727 if (addr_len > RPCBIND_MAXUADDRLEN)
711 goto out_err; 728 goto out_err;
712 729
713 /* 730 /*
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index bb507e2bb94d..9f2f2412a2f3 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -359,7 +359,7 @@ svc_pool_for_cpu(struct svc_serv *serv, int cpu)
359 */ 359 */
360static struct svc_serv * 360static struct svc_serv *
361__svc_create(struct svc_program *prog, unsigned int bufsize, int npools, 361__svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
362 sa_family_t family, void (*shutdown)(struct svc_serv *serv)) 362 void (*shutdown)(struct svc_serv *serv))
363{ 363{
364 struct svc_serv *serv; 364 struct svc_serv *serv;
365 unsigned int vers; 365 unsigned int vers;
@@ -368,7 +368,6 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
368 368
369 if (!(serv = kzalloc(sizeof(*serv), GFP_KERNEL))) 369 if (!(serv = kzalloc(sizeof(*serv), GFP_KERNEL)))
370 return NULL; 370 return NULL;
371 serv->sv_family = family;
372 serv->sv_name = prog->pg_name; 371 serv->sv_name = prog->pg_name;
373 serv->sv_program = prog; 372 serv->sv_program = prog;
374 serv->sv_nrthreads = 1; 373 serv->sv_nrthreads = 1;
@@ -427,21 +426,21 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
427 426
428struct svc_serv * 427struct svc_serv *
429svc_create(struct svc_program *prog, unsigned int bufsize, 428svc_create(struct svc_program *prog, unsigned int bufsize,
430 sa_family_t family, void (*shutdown)(struct svc_serv *serv)) 429 void (*shutdown)(struct svc_serv *serv))
431{ 430{
432 return __svc_create(prog, bufsize, /*npools*/1, family, shutdown); 431 return __svc_create(prog, bufsize, /*npools*/1, shutdown);
433} 432}
434EXPORT_SYMBOL_GPL(svc_create); 433EXPORT_SYMBOL_GPL(svc_create);
435 434
436struct svc_serv * 435struct svc_serv *
437svc_create_pooled(struct svc_program *prog, unsigned int bufsize, 436svc_create_pooled(struct svc_program *prog, unsigned int bufsize,
438 sa_family_t family, void (*shutdown)(struct svc_serv *serv), 437 void (*shutdown)(struct svc_serv *serv),
439 svc_thread_fn func, struct module *mod) 438 svc_thread_fn func, struct module *mod)
440{ 439{
441 struct svc_serv *serv; 440 struct svc_serv *serv;
442 unsigned int npools = svc_pool_map_get(); 441 unsigned int npools = svc_pool_map_get();
443 442
444 serv = __svc_create(prog, bufsize, npools, family, shutdown); 443 serv = __svc_create(prog, bufsize, npools, shutdown);
445 444
446 if (serv != NULL) { 445 if (serv != NULL) {
447 serv->sv_function = func; 446 serv->sv_function = func;
@@ -719,8 +718,6 @@ svc_exit_thread(struct svc_rqst *rqstp)
719} 718}
720EXPORT_SYMBOL_GPL(svc_exit_thread); 719EXPORT_SYMBOL_GPL(svc_exit_thread);
721 720
722#ifdef CONFIG_SUNRPC_REGISTER_V4
723
724/* 721/*
725 * Register an "inet" protocol family netid with the local 722 * Register an "inet" protocol family netid with the local
726 * rpcbind daemon via an rpcbind v4 SET request. 723 * rpcbind daemon via an rpcbind v4 SET request.
@@ -735,12 +732,13 @@ static int __svc_rpcb_register4(const u32 program, const u32 version,
735 const unsigned short protocol, 732 const unsigned short protocol,
736 const unsigned short port) 733 const unsigned short port)
737{ 734{
738 struct sockaddr_in sin = { 735 const struct sockaddr_in sin = {
739 .sin_family = AF_INET, 736 .sin_family = AF_INET,
740 .sin_addr.s_addr = htonl(INADDR_ANY), 737 .sin_addr.s_addr = htonl(INADDR_ANY),
741 .sin_port = htons(port), 738 .sin_port = htons(port),
742 }; 739 };
743 char *netid; 740 const char *netid;
741 int error;
744 742
745 switch (protocol) { 743 switch (protocol) {
746 case IPPROTO_UDP: 744 case IPPROTO_UDP:
@@ -750,13 +748,23 @@ static int __svc_rpcb_register4(const u32 program, const u32 version,
750 netid = RPCBIND_NETID_TCP; 748 netid = RPCBIND_NETID_TCP;
751 break; 749 break;
752 default: 750 default:
753 return -EPROTONOSUPPORT; 751 return -ENOPROTOOPT;
754 } 752 }
755 753
756 return rpcb_v4_register(program, version, 754 error = rpcb_v4_register(program, version,
757 (struct sockaddr *)&sin, netid); 755 (const struct sockaddr *)&sin, netid);
756
757 /*
758 * User space didn't support rpcbind v4, so retry this
759 * registration request with the legacy rpcbind v2 protocol.
760 */
761 if (error == -EPROTONOSUPPORT)
762 error = rpcb_register(program, version, protocol, port);
763
764 return error;
758} 765}
759 766
767#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
760/* 768/*
761 * Register an "inet6" protocol family netid with the local 769 * Register an "inet6" protocol family netid with the local
762 * rpcbind daemon via an rpcbind v4 SET request. 770 * rpcbind daemon via an rpcbind v4 SET request.
@@ -771,12 +779,13 @@ static int __svc_rpcb_register6(const u32 program, const u32 version,
771 const unsigned short protocol, 779 const unsigned short protocol,
772 const unsigned short port) 780 const unsigned short port)
773{ 781{
774 struct sockaddr_in6 sin6 = { 782 const struct sockaddr_in6 sin6 = {
775 .sin6_family = AF_INET6, 783 .sin6_family = AF_INET6,
776 .sin6_addr = IN6ADDR_ANY_INIT, 784 .sin6_addr = IN6ADDR_ANY_INIT,
777 .sin6_port = htons(port), 785 .sin6_port = htons(port),
778 }; 786 };
779 char *netid; 787 const char *netid;
788 int error;
780 789
781 switch (protocol) { 790 switch (protocol) {
782 case IPPROTO_UDP: 791 case IPPROTO_UDP:
@@ -786,12 +795,22 @@ static int __svc_rpcb_register6(const u32 program, const u32 version,
786 netid = RPCBIND_NETID_TCP6; 795 netid = RPCBIND_NETID_TCP6;
787 break; 796 break;
788 default: 797 default:
789 return -EPROTONOSUPPORT; 798 return -ENOPROTOOPT;
790 } 799 }
791 800
792 return rpcb_v4_register(program, version, 801 error = rpcb_v4_register(program, version,
793 (struct sockaddr *)&sin6, netid); 802 (const struct sockaddr *)&sin6, netid);
803
804 /*
805 * User space didn't support rpcbind version 4, so we won't
806 * use a PF_INET6 listener.
807 */
808 if (error == -EPROTONOSUPPORT)
809 error = -EAFNOSUPPORT;
810
811 return error;
794} 812}
813#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
795 814
796/* 815/*
797 * Register a kernel RPC service via rpcbind version 4. 816 * Register a kernel RPC service via rpcbind version 4.
@@ -799,69 +818,43 @@ static int __svc_rpcb_register6(const u32 program, const u32 version,
799 * Returns zero on success; a negative errno value is returned 818 * Returns zero on success; a negative errno value is returned
800 * if any error occurs. 819 * if any error occurs.
801 */ 820 */
802static int __svc_register(const u32 program, const u32 version, 821static int __svc_register(const char *progname,
803 const sa_family_t family, 822 const u32 program, const u32 version,
823 const int family,
804 const unsigned short protocol, 824 const unsigned short protocol,
805 const unsigned short port) 825 const unsigned short port)
806{ 826{
807 int error; 827 int error = -EAFNOSUPPORT;
808 828
809 switch (family) { 829 switch (family) {
810 case AF_INET: 830 case PF_INET:
811 return __svc_rpcb_register4(program, version, 831 error = __svc_rpcb_register4(program, version,
812 protocol, port); 832 protocol, port);
813 case AF_INET6: 833 break;
834#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
835 case PF_INET6:
814 error = __svc_rpcb_register6(program, version, 836 error = __svc_rpcb_register6(program, version,
815 protocol, port); 837 protocol, port);
816 if (error < 0) 838#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
817 return error;
818
819 /*
820 * Work around bug in some versions of Linux rpcbind
821 * which don't allow registration of both inet and
822 * inet6 netids.
823 *
824 * Error return ignored for now.
825 */
826 __svc_rpcb_register4(program, version,
827 protocol, port);
828 return 0;
829 } 839 }
830 840
831 return -EAFNOSUPPORT; 841 if (error < 0)
832} 842 printk(KERN_WARNING "svc: failed to register %sv%u RPC "
833 843 "service (errno %d).\n", progname, version, -error);
834#else /* CONFIG_SUNRPC_REGISTER_V4 */ 844 return error;
835
836/*
837 * Register a kernel RPC service via rpcbind version 2.
838 *
839 * Returns zero on success; a negative errno value is returned
840 * if any error occurs.
841 */
842static int __svc_register(const u32 program, const u32 version,
843 sa_family_t family,
844 const unsigned short protocol,
845 const unsigned short port)
846{
847 if (family != AF_INET)
848 return -EAFNOSUPPORT;
849
850 return rpcb_register(program, version, protocol, port);
851} 845}
852 846
853#endif /* CONFIG_SUNRPC_REGISTER_V4 */
854
855/** 847/**
856 * svc_register - register an RPC service with the local portmapper 848 * svc_register - register an RPC service with the local portmapper
857 * @serv: svc_serv struct for the service to register 849 * @serv: svc_serv struct for the service to register
850 * @family: protocol family of service's listener socket
858 * @proto: transport protocol number to advertise 851 * @proto: transport protocol number to advertise
859 * @port: port to advertise 852 * @port: port to advertise
860 * 853 *
861 * Service is registered for any address in serv's address family 854 * Service is registered for any address in the passed-in protocol family
862 */ 855 */
863int svc_register(const struct svc_serv *serv, const unsigned short proto, 856int svc_register(const struct svc_serv *serv, const int family,
864 const unsigned short port) 857 const unsigned short proto, const unsigned short port)
865{ 858{
866 struct svc_program *progp; 859 struct svc_program *progp;
867 unsigned int i; 860 unsigned int i;
@@ -879,15 +872,15 @@ int svc_register(const struct svc_serv *serv, const unsigned short proto,
879 i, 872 i,
880 proto == IPPROTO_UDP? "udp" : "tcp", 873 proto == IPPROTO_UDP? "udp" : "tcp",
881 port, 874 port,
882 serv->sv_family, 875 family,
883 progp->pg_vers[i]->vs_hidden? 876 progp->pg_vers[i]->vs_hidden?
884 " (but not telling portmap)" : ""); 877 " (but not telling portmap)" : "");
885 878
886 if (progp->pg_vers[i]->vs_hidden) 879 if (progp->pg_vers[i]->vs_hidden)
887 continue; 880 continue;
888 881
889 error = __svc_register(progp->pg_prog, i, 882 error = __svc_register(progp->pg_name, progp->pg_prog,
890 serv->sv_family, proto, port); 883 i, family, proto, port);
891 if (error < 0) 884 if (error < 0)
892 break; 885 break;
893 } 886 }
@@ -896,38 +889,31 @@ int svc_register(const struct svc_serv *serv, const unsigned short proto,
896 return error; 889 return error;
897} 890}
898 891
899#ifdef CONFIG_SUNRPC_REGISTER_V4 892/*
900 893 * If user space is running rpcbind, it should take the v4 UNSET
894 * and clear everything for this [program, version]. If user space
895 * is running portmap, it will reject the v4 UNSET, but won't have
896 * any "inet6" entries anyway. So a PMAP_UNSET should be sufficient
897 * in this case to clear all existing entries for [program, version].
898 */
901static void __svc_unregister(const u32 program, const u32 version, 899static void __svc_unregister(const u32 program, const u32 version,
902 const char *progname) 900 const char *progname)
903{ 901{
904 struct sockaddr_in6 sin6 = {
905 .sin6_family = AF_INET6,
906 .sin6_addr = IN6ADDR_ANY_INIT,
907 .sin6_port = 0,
908 };
909 int error; 902 int error;
910 903
911 error = rpcb_v4_register(program, version, 904 error = rpcb_v4_register(program, version, NULL, "");
912 (struct sockaddr *)&sin6, "");
913 dprintk("svc: %s(%sv%u), error %d\n",
914 __func__, progname, version, error);
915}
916
917#else /* CONFIG_SUNRPC_REGISTER_V4 */
918 905
919static void __svc_unregister(const u32 program, const u32 version, 906 /*
920 const char *progname) 907 * User space didn't support rpcbind v4, so retry this
921{ 908 * request with the legacy rpcbind v2 protocol.
922 int error; 909 */
910 if (error == -EPROTONOSUPPORT)
911 error = rpcb_register(program, version, 0, 0);
923 912
924 error = rpcb_register(program, version, 0, 0);
925 dprintk("svc: %s(%sv%u), error %d\n", 913 dprintk("svc: %s(%sv%u), error %d\n",
926 __func__, progname, version, error); 914 __func__, progname, version, error);
927} 915}
928 916
929#endif /* CONFIG_SUNRPC_REGISTER_V4 */
930
931/* 917/*
932 * All netids, bind addresses and ports registered for [program, version] 918 * All netids, bind addresses and ports registered for [program, version]
933 * are removed from the local rpcbind database (if the service is not 919 * are removed from the local rpcbind database (if the service is not
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index e588df5d6b34..2819ee093f36 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -161,7 +161,9 @@ EXPORT_SYMBOL_GPL(svc_xprt_init);
161 161
162static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl, 162static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
163 struct svc_serv *serv, 163 struct svc_serv *serv,
164 unsigned short port, int flags) 164 const int family,
165 const unsigned short port,
166 int flags)
165{ 167{
166 struct sockaddr_in sin = { 168 struct sockaddr_in sin = {
167 .sin_family = AF_INET, 169 .sin_family = AF_INET,
@@ -176,12 +178,12 @@ static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
176 struct sockaddr *sap; 178 struct sockaddr *sap;
177 size_t len; 179 size_t len;
178 180
179 switch (serv->sv_family) { 181 switch (family) {
180 case AF_INET: 182 case PF_INET:
181 sap = (struct sockaddr *)&sin; 183 sap = (struct sockaddr *)&sin;
182 len = sizeof(sin); 184 len = sizeof(sin);
183 break; 185 break;
184 case AF_INET6: 186 case PF_INET6:
185 sap = (struct sockaddr *)&sin6; 187 sap = (struct sockaddr *)&sin6;
186 len = sizeof(sin6); 188 len = sizeof(sin6);
187 break; 189 break;
@@ -192,7 +194,8 @@ static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
192 return xcl->xcl_ops->xpo_create(serv, sap, len, flags); 194 return xcl->xcl_ops->xpo_create(serv, sap, len, flags);
193} 195}
194 196
195int svc_create_xprt(struct svc_serv *serv, char *xprt_name, unsigned short port, 197int svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
198 const int family, const unsigned short port,
196 int flags) 199 int flags)
197{ 200{
198 struct svc_xprt_class *xcl; 201 struct svc_xprt_class *xcl;
@@ -209,7 +212,7 @@ int svc_create_xprt(struct svc_serv *serv, char *xprt_name, unsigned short port,
209 goto err; 212 goto err;
210 213
211 spin_unlock(&svc_xprt_class_lock); 214 spin_unlock(&svc_xprt_class_lock);
212 newxprt = __svc_xpo_create(xcl, serv, port, flags); 215 newxprt = __svc_xpo_create(xcl, serv, family, port, flags);
213 if (IS_ERR(newxprt)) { 216 if (IS_ERR(newxprt)) {
214 module_put(xcl->xcl_owner); 217 module_put(xcl->xcl_owner);
215 return PTR_ERR(newxprt); 218 return PTR_ERR(newxprt);
@@ -1033,7 +1036,13 @@ static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
1033 return dr; 1036 return dr;
1034} 1037}
1035 1038
1036/* 1039/**
1040 * svc_find_xprt - find an RPC transport instance
1041 * @serv: pointer to svc_serv to search
1042 * @xcl_name: C string containing transport's class name
1043 * @af: Address family of transport's local address
1044 * @port: transport's IP port number
1045 *
1037 * Return the transport instance pointer for the endpoint accepting 1046 * Return the transport instance pointer for the endpoint accepting
1038 * connections/peer traffic from the specified transport class, 1047 * connections/peer traffic from the specified transport class,
1039 * address family and port. 1048 * address family and port.
@@ -1042,14 +1051,14 @@ static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
1042 * wild-card, and will result in matching the first transport in the 1051 * wild-card, and will result in matching the first transport in the
1043 * service's list that has a matching class name. 1052 * service's list that has a matching class name.
1044 */ 1053 */
1045struct svc_xprt *svc_find_xprt(struct svc_serv *serv, char *xcl_name, 1054struct svc_xprt *svc_find_xprt(struct svc_serv *serv, const char *xcl_name,
1046 int af, int port) 1055 const sa_family_t af, const unsigned short port)
1047{ 1056{
1048 struct svc_xprt *xprt; 1057 struct svc_xprt *xprt;
1049 struct svc_xprt *found = NULL; 1058 struct svc_xprt *found = NULL;
1050 1059
1051 /* Sanity check the args */ 1060 /* Sanity check the args */
1052 if (!serv || !xcl_name) 1061 if (serv == NULL || xcl_name == NULL)
1053 return found; 1062 return found;
1054 1063
1055 spin_lock_bh(&serv->sv_lock); 1064 spin_lock_bh(&serv->sv_lock);
@@ -1058,7 +1067,7 @@ struct svc_xprt *svc_find_xprt(struct svc_serv *serv, char *xcl_name,
1058 continue; 1067 continue;
1059 if (af != AF_UNSPEC && af != xprt->xpt_local.ss_family) 1068 if (af != AF_UNSPEC && af != xprt->xpt_local.ss_family)
1060 continue; 1069 continue;
1061 if (port && port != svc_xprt_local_port(xprt)) 1070 if (port != 0 && port != svc_xprt_local_port(xprt))
1062 continue; 1071 continue;
1063 found = xprt; 1072 found = xprt;
1064 svc_xprt_get(xprt); 1073 svc_xprt_get(xprt);
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 5763e6460fea..9d504234af4a 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1110,7 +1110,6 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,
1110 struct svc_sock *svsk; 1110 struct svc_sock *svsk;
1111 struct sock *inet; 1111 struct sock *inet;
1112 int pmap_register = !(flags & SVC_SOCK_ANONYMOUS); 1112 int pmap_register = !(flags & SVC_SOCK_ANONYMOUS);
1113 int val;
1114 1113
1115 dprintk("svc: svc_setup_socket %p\n", sock); 1114 dprintk("svc: svc_setup_socket %p\n", sock);
1116 if (!(svsk = kzalloc(sizeof(*svsk), GFP_KERNEL))) { 1115 if (!(svsk = kzalloc(sizeof(*svsk), GFP_KERNEL))) {
@@ -1122,7 +1121,7 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,
1122 1121
1123 /* Register socket with portmapper */ 1122 /* Register socket with portmapper */
1124 if (*errp >= 0 && pmap_register) 1123 if (*errp >= 0 && pmap_register)
1125 *errp = svc_register(serv, inet->sk_protocol, 1124 *errp = svc_register(serv, inet->sk_family, inet->sk_protocol,
1126 ntohs(inet_sk(inet)->sport)); 1125 ntohs(inet_sk(inet)->sport));
1127 1126
1128 if (*errp < 0) { 1127 if (*errp < 0) {
@@ -1143,18 +1142,6 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,
1143 else 1142 else
1144 svc_tcp_init(svsk, serv); 1143 svc_tcp_init(svsk, serv);
1145 1144
1146 /*
1147 * We start one listener per sv_serv. We want AF_INET
1148 * requests to be automatically shunted to our AF_INET6
1149 * listener using a mapped IPv4 address. Make sure
1150 * no-one starts an equivalent IPv4 listener, which
1151 * would steal our incoming connections.
1152 */
1153 val = 0;
1154 if (serv->sv_family == AF_INET6)
1155 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
1156 (char *)&val, sizeof(val));
1157
1158 dprintk("svc: svc_setup_socket created %p (inet %p)\n", 1145 dprintk("svc: svc_setup_socket created %p (inet %p)\n",
1159 svsk, svsk->sk_sk); 1146 svsk, svsk->sk_sk);
1160 1147
@@ -1222,6 +1209,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
1222 struct sockaddr_storage addr; 1209 struct sockaddr_storage addr;
1223 struct sockaddr *newsin = (struct sockaddr *)&addr; 1210 struct sockaddr *newsin = (struct sockaddr *)&addr;
1224 int newlen; 1211 int newlen;
1212 int family;
1213 int val;
1225 RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]); 1214 RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
1226 1215
1227 dprintk("svc: svc_create_socket(%s, %d, %s)\n", 1216 dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1233,14 +1222,35 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
1233 "sockets supported\n"); 1222 "sockets supported\n");
1234 return ERR_PTR(-EINVAL); 1223 return ERR_PTR(-EINVAL);
1235 } 1224 }
1225
1236 type = (protocol == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM; 1226 type = (protocol == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
1227 switch (sin->sa_family) {
1228 case AF_INET6:
1229 family = PF_INET6;
1230 break;
1231 case AF_INET:
1232 family = PF_INET;
1233 break;
1234 default:
1235 return ERR_PTR(-EINVAL);
1236 }
1237 1237
1238 error = sock_create_kern(sin->sa_family, type, protocol, &sock); 1238 error = sock_create_kern(family, type, protocol, &sock);
1239 if (error < 0) 1239 if (error < 0)
1240 return ERR_PTR(error); 1240 return ERR_PTR(error);
1241 1241
1242 svc_reclassify_socket(sock); 1242 svc_reclassify_socket(sock);
1243 1243
1244 /*
1245 * If this is an PF_INET6 listener, we want to avoid
1246 * getting requests from IPv4 remotes. Those should
1247 * be shunted to a PF_INET listener via rpcbind.
1248 */
1249 val = 1;
1250 if (family == PF_INET6)
1251 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
1252 (char *)&val, sizeof(val));
1253
1244 if (type == SOCK_STREAM) 1254 if (type == SOCK_STREAM)
1245 sock->sk->sk_reuse = 1; /* allow address reuse */ 1255 sock->sk->sk_reuse = 1; /* allow address reuse */
1246 error = kernel_bind(sock, sin, len); 1256 error = kernel_bind(sock, sin, len);
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 62098d101a1f..a0bfe53f1621 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -152,6 +152,37 @@ out:
152EXPORT_SYMBOL_GPL(xprt_unregister_transport); 152EXPORT_SYMBOL_GPL(xprt_unregister_transport);
153 153
154/** 154/**
155 * xprt_load_transport - load a transport implementation
156 * @transport_name: transport to load
157 *
158 * Returns:
159 * 0: transport successfully loaded
160 * -ENOENT: transport module not available
161 */
162int xprt_load_transport(const char *transport_name)
163{
164 struct xprt_class *t;
165 char module_name[sizeof t->name + 5];
166 int result;
167
168 result = 0;
169 spin_lock(&xprt_list_lock);
170 list_for_each_entry(t, &xprt_list, list) {
171 if (strcmp(t->name, transport_name) == 0) {
172 spin_unlock(&xprt_list_lock);
173 goto out;
174 }
175 }
176 spin_unlock(&xprt_list_lock);
177 strcpy(module_name, "xprt");
178 strncat(module_name, transport_name, sizeof t->name);
179 result = request_module(module_name);
180out:
181 return result;
182}
183EXPORT_SYMBOL_GPL(xprt_load_transport);
184
185/**
155 * xprt_reserve_xprt - serialize write access to transports 186 * xprt_reserve_xprt - serialize write access to transports
156 * @task: task that is requesting access to the transport 187 * @task: task that is requesting access to the transport
157 * 188 *
@@ -580,7 +611,7 @@ void xprt_disconnect_done(struct rpc_xprt *xprt)
580 dprintk("RPC: disconnected transport %p\n", xprt); 611 dprintk("RPC: disconnected transport %p\n", xprt);
581 spin_lock_bh(&xprt->transport_lock); 612 spin_lock_bh(&xprt->transport_lock);
582 xprt_clear_connected(xprt); 613 xprt_clear_connected(xprt);
583 xprt_wake_pending_tasks(xprt, -ENOTCONN); 614 xprt_wake_pending_tasks(xprt, -EAGAIN);
584 spin_unlock_bh(&xprt->transport_lock); 615 spin_unlock_bh(&xprt->transport_lock);
585} 616}
586EXPORT_SYMBOL_GPL(xprt_disconnect_done); 617EXPORT_SYMBOL_GPL(xprt_disconnect_done);
@@ -598,7 +629,7 @@ void xprt_force_disconnect(struct rpc_xprt *xprt)
598 /* Try to schedule an autoclose RPC call */ 629 /* Try to schedule an autoclose RPC call */
599 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0) 630 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
600 queue_work(rpciod_workqueue, &xprt->task_cleanup); 631 queue_work(rpciod_workqueue, &xprt->task_cleanup);
601 xprt_wake_pending_tasks(xprt, -ENOTCONN); 632 xprt_wake_pending_tasks(xprt, -EAGAIN);
602 spin_unlock_bh(&xprt->transport_lock); 633 spin_unlock_bh(&xprt->transport_lock);
603} 634}
604 635
@@ -625,7 +656,7 @@ void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
625 /* Try to schedule an autoclose RPC call */ 656 /* Try to schedule an autoclose RPC call */
626 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0) 657 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
627 queue_work(rpciod_workqueue, &xprt->task_cleanup); 658 queue_work(rpciod_workqueue, &xprt->task_cleanup);
628 xprt_wake_pending_tasks(xprt, -ENOTCONN); 659 xprt_wake_pending_tasks(xprt, -EAGAIN);
629out: 660out:
630 spin_unlock_bh(&xprt->transport_lock); 661 spin_unlock_bh(&xprt->transport_lock);
631} 662}
@@ -695,9 +726,8 @@ static void xprt_connect_status(struct rpc_task *task)
695 } 726 }
696 727
697 switch (task->tk_status) { 728 switch (task->tk_status) {
698 case -ENOTCONN: 729 case -EAGAIN:
699 dprintk("RPC: %5u xprt_connect_status: connection broken\n", 730 dprintk("RPC: %5u xprt_connect_status: retrying\n", task->tk_pid);
700 task->tk_pid);
701 break; 731 break;
702 case -ETIMEDOUT: 732 case -ETIMEDOUT:
703 dprintk("RPC: %5u xprt_connect_status: connect attempt timed " 733 dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
@@ -818,15 +848,8 @@ int xprt_prepare_transmit(struct rpc_task *task)
818 err = req->rq_received; 848 err = req->rq_received;
819 goto out_unlock; 849 goto out_unlock;
820 } 850 }
821 if (!xprt->ops->reserve_xprt(task)) { 851 if (!xprt->ops->reserve_xprt(task))
822 err = -EAGAIN; 852 err = -EAGAIN;
823 goto out_unlock;
824 }
825
826 if (!xprt_connected(xprt)) {
827 err = -ENOTCONN;
828 goto out_unlock;
829 }
830out_unlock: 853out_unlock:
831 spin_unlock_bh(&xprt->transport_lock); 854 spin_unlock_bh(&xprt->transport_lock);
832 return err; 855 return err;
@@ -870,32 +893,26 @@ void xprt_transmit(struct rpc_task *task)
870 req->rq_connect_cookie = xprt->connect_cookie; 893 req->rq_connect_cookie = xprt->connect_cookie;
871 req->rq_xtime = jiffies; 894 req->rq_xtime = jiffies;
872 status = xprt->ops->send_request(task); 895 status = xprt->ops->send_request(task);
873 if (status == 0) { 896 if (status != 0) {
874 dprintk("RPC: %5u xmit complete\n", task->tk_pid); 897 task->tk_status = status;
875 spin_lock_bh(&xprt->transport_lock); 898 return;
899 }
876 900
877 xprt->ops->set_retrans_timeout(task); 901 dprintk("RPC: %5u xmit complete\n", task->tk_pid);
902 spin_lock_bh(&xprt->transport_lock);
878 903
879 xprt->stat.sends++; 904 xprt->ops->set_retrans_timeout(task);
880 xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
881 xprt->stat.bklog_u += xprt->backlog.qlen;
882 905
883 /* Don't race with disconnect */ 906 xprt->stat.sends++;
884 if (!xprt_connected(xprt)) 907 xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
885 task->tk_status = -ENOTCONN; 908 xprt->stat.bklog_u += xprt->backlog.qlen;
886 else if (!req->rq_received)
887 rpc_sleep_on(&xprt->pending, task, xprt_timer);
888 spin_unlock_bh(&xprt->transport_lock);
889 return;
890 }
891 909
892 /* Note: at this point, task->tk_sleeping has not yet been set, 910 /* Don't race with disconnect */
893 * hence there is no danger of the waking up task being put on 911 if (!xprt_connected(xprt))
894 * schedq, and being picked up by a parallel run of rpciod(). 912 task->tk_status = -ENOTCONN;
895 */ 913 else if (!req->rq_received)
896 task->tk_status = status; 914 rpc_sleep_on(&xprt->pending, task, xprt_timer);
897 if (status == -ECONNREFUSED) 915 spin_unlock_bh(&xprt->transport_lock);
898 rpc_sleep_on(&xprt->sending, task, NULL);
899} 916}
900 917
901static inline void do_xprt_reserve(struct rpc_task *task) 918static inline void do_xprt_reserve(struct rpc_task *task)
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 14106d26bb95..e5e28d1946a4 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -310,6 +310,19 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad)
310 __func__, pad, destp, rqst->rq_slen, curlen); 310 __func__, pad, destp, rqst->rq_slen, curlen);
311 311
312 copy_len = rqst->rq_snd_buf.page_len; 312 copy_len = rqst->rq_snd_buf.page_len;
313
314 if (rqst->rq_snd_buf.tail[0].iov_len) {
315 curlen = rqst->rq_snd_buf.tail[0].iov_len;
316 if (destp + copy_len != rqst->rq_snd_buf.tail[0].iov_base) {
317 memmove(destp + copy_len,
318 rqst->rq_snd_buf.tail[0].iov_base, curlen);
319 r_xprt->rx_stats.pullup_copy_count += curlen;
320 }
321 dprintk("RPC: %s: tail destp 0x%p len %d\n",
322 __func__, destp + copy_len, curlen);
323 rqst->rq_svec[0].iov_len += curlen;
324 }
325
313 r_xprt->rx_stats.pullup_copy_count += copy_len; 326 r_xprt->rx_stats.pullup_copy_count += copy_len;
314 npages = PAGE_ALIGN(rqst->rq_snd_buf.page_base+copy_len) >> PAGE_SHIFT; 327 npages = PAGE_ALIGN(rqst->rq_snd_buf.page_base+copy_len) >> PAGE_SHIFT;
315 for (i = 0; copy_len && i < npages; i++) { 328 for (i = 0; copy_len && i < npages; i++) {
@@ -332,17 +345,6 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad)
332 destp += curlen; 345 destp += curlen;
333 copy_len -= curlen; 346 copy_len -= curlen;
334 } 347 }
335 if (rqst->rq_snd_buf.tail[0].iov_len) {
336 curlen = rqst->rq_snd_buf.tail[0].iov_len;
337 if (destp != rqst->rq_snd_buf.tail[0].iov_base) {
338 memcpy(destp,
339 rqst->rq_snd_buf.tail[0].iov_base, curlen);
340 r_xprt->rx_stats.pullup_copy_count += curlen;
341 }
342 dprintk("RPC: %s: tail destp 0x%p len %d curlen %d\n",
343 __func__, destp, copy_len, curlen);
344 rqst->rq_svec[0].iov_len += curlen;
345 }
346 /* header now contains entire send message */ 348 /* header now contains entire send message */
347 return pad; 349 return pad;
348} 350}
@@ -656,7 +658,7 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
656 if (curlen > rqst->rq_rcv_buf.tail[0].iov_len) 658 if (curlen > rqst->rq_rcv_buf.tail[0].iov_len)
657 curlen = rqst->rq_rcv_buf.tail[0].iov_len; 659 curlen = rqst->rq_rcv_buf.tail[0].iov_len;
658 if (rqst->rq_rcv_buf.tail[0].iov_base != srcp) 660 if (rqst->rq_rcv_buf.tail[0].iov_base != srcp)
659 memcpy(rqst->rq_rcv_buf.tail[0].iov_base, srcp, curlen); 661 memmove(rqst->rq_rcv_buf.tail[0].iov_base, srcp, curlen);
660 dprintk("RPC: %s: tail srcp 0x%p len %d curlen %d\n", 662 dprintk("RPC: %s: tail srcp 0x%p len %d curlen %d\n",
661 __func__, srcp, copy_len, curlen); 663 __func__, srcp, copy_len, curlen);
662 rqst->rq_rcv_buf.tail[0].iov_len = curlen; 664 rqst->rq_rcv_buf.tail[0].iov_len = curlen;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index a3334e3b73cc..6c26a675435a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -191,7 +191,6 @@ static int map_xdr(struct svcxprt_rdma *xprt,
191 struct xdr_buf *xdr, 191 struct xdr_buf *xdr,
192 struct svc_rdma_req_map *vec) 192 struct svc_rdma_req_map *vec)
193{ 193{
194 int sge_max = (xdr->len+PAGE_SIZE-1) / PAGE_SIZE + 3;
195 int sge_no; 194 int sge_no;
196 u32 sge_bytes; 195 u32 sge_bytes;
197 u32 page_bytes; 196 u32 page_bytes;
@@ -235,7 +234,11 @@ static int map_xdr(struct svcxprt_rdma *xprt,
235 sge_no++; 234 sge_no++;
236 } 235 }
237 236
238 BUG_ON(sge_no > sge_max); 237 dprintk("svcrdma: map_xdr: sge_no %d page_no %d "
238 "page_base %u page_len %u head_len %zu tail_len %zu\n",
239 sge_no, page_no, xdr->page_base, xdr->page_len,
240 xdr->head[0].iov_len, xdr->tail[0].iov_len);
241
239 vec->count = sge_no; 242 vec->count = sge_no;
240 return 0; 243 return 0;
241} 244}
@@ -579,7 +582,6 @@ static int send_reply(struct svcxprt_rdma *rdma,
579 ctxt->sge[page_no+1].length = 0; 582 ctxt->sge[page_no+1].length = 0;
580 } 583 }
581 BUG_ON(sge_no > rdma->sc_max_sge); 584 BUG_ON(sge_no > rdma->sc_max_sge);
582 BUG_ON(sge_no > ctxt->count);
583 memset(&send_wr, 0, sizeof send_wr); 585 memset(&send_wr, 0, sizeof send_wr);
584 ctxt->wr_op = IB_WR_SEND; 586 ctxt->wr_op = IB_WR_SEND;
585 send_wr.wr_id = (unsigned long)ctxt; 587 send_wr.wr_id = (unsigned long)ctxt;
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 568330eebbfe..d40ff50887aa 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -49,6 +49,9 @@ unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
49unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; 49unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
50unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; 50unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
51 51
52#define XS_TCP_LINGER_TO (15U * HZ)
53static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
54
52/* 55/*
53 * We can register our own files under /proc/sys/sunrpc by 56 * We can register our own files under /proc/sys/sunrpc by
54 * calling register_sysctl_table() again. The files in that 57 * calling register_sysctl_table() again. The files in that
@@ -117,6 +120,14 @@ static ctl_table xs_tunables_table[] = {
117 .extra2 = &xprt_max_resvport_limit 120 .extra2 = &xprt_max_resvport_limit
118 }, 121 },
119 { 122 {
123 .procname = "tcp_fin_timeout",
124 .data = &xs_tcp_fin_timeout,
125 .maxlen = sizeof(xs_tcp_fin_timeout),
126 .mode = 0644,
127 .proc_handler = &proc_dointvec_jiffies,
128 .strategy = sysctl_jiffies
129 },
130 {
120 .ctl_name = 0, 131 .ctl_name = 0,
121 }, 132 },
122}; 133};
@@ -521,11 +532,12 @@ static void xs_nospace_callback(struct rpc_task *task)
521 * @task: task to put to sleep 532 * @task: task to put to sleep
522 * 533 *
523 */ 534 */
524static void xs_nospace(struct rpc_task *task) 535static int xs_nospace(struct rpc_task *task)
525{ 536{
526 struct rpc_rqst *req = task->tk_rqstp; 537 struct rpc_rqst *req = task->tk_rqstp;
527 struct rpc_xprt *xprt = req->rq_xprt; 538 struct rpc_xprt *xprt = req->rq_xprt;
528 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 539 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
540 int ret = 0;
529 541
530 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n", 542 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
531 task->tk_pid, req->rq_slen - req->rq_bytes_sent, 543 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
@@ -537,6 +549,7 @@ static void xs_nospace(struct rpc_task *task)
537 /* Don't race with disconnect */ 549 /* Don't race with disconnect */
538 if (xprt_connected(xprt)) { 550 if (xprt_connected(xprt)) {
539 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { 551 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) {
552 ret = -EAGAIN;
540 /* 553 /*
541 * Notify TCP that we're limited by the application 554 * Notify TCP that we're limited by the application
542 * window size 555 * window size
@@ -548,10 +561,11 @@ static void xs_nospace(struct rpc_task *task)
548 } 561 }
549 } else { 562 } else {
550 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 563 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
551 task->tk_status = -ENOTCONN; 564 ret = -ENOTCONN;
552 } 565 }
553 566
554 spin_unlock_bh(&xprt->transport_lock); 567 spin_unlock_bh(&xprt->transport_lock);
568 return ret;
555} 569}
556 570
557/** 571/**
@@ -594,6 +608,8 @@ static int xs_udp_send_request(struct rpc_task *task)
594 /* Still some bytes left; set up for a retry later. */ 608 /* Still some bytes left; set up for a retry later. */
595 status = -EAGAIN; 609 status = -EAGAIN;
596 } 610 }
611 if (!transport->sock)
612 goto out;
597 613
598 switch (status) { 614 switch (status) {
599 case -ENOTSOCK: 615 case -ENOTSOCK:
@@ -601,21 +617,19 @@ static int xs_udp_send_request(struct rpc_task *task)
601 /* Should we call xs_close() here? */ 617 /* Should we call xs_close() here? */
602 break; 618 break;
603 case -EAGAIN: 619 case -EAGAIN:
604 xs_nospace(task); 620 status = xs_nospace(task);
605 break; 621 break;
622 default:
623 dprintk("RPC: sendmsg returned unrecognized error %d\n",
624 -status);
606 case -ENETUNREACH: 625 case -ENETUNREACH:
607 case -EPIPE: 626 case -EPIPE:
608 case -ECONNREFUSED: 627 case -ECONNREFUSED:
609 /* When the server has died, an ICMP port unreachable message 628 /* When the server has died, an ICMP port unreachable message
610 * prompts ECONNREFUSED. */ 629 * prompts ECONNREFUSED. */
611 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 630 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
612 break;
613 default:
614 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
615 dprintk("RPC: sendmsg returned unrecognized error %d\n",
616 -status);
617 } 631 }
618 632out:
619 return status; 633 return status;
620} 634}
621 635
@@ -697,6 +711,8 @@ static int xs_tcp_send_request(struct rpc_task *task)
697 status = -EAGAIN; 711 status = -EAGAIN;
698 break; 712 break;
699 } 713 }
714 if (!transport->sock)
715 goto out;
700 716
701 switch (status) { 717 switch (status) {
702 case -ENOTSOCK: 718 case -ENOTSOCK:
@@ -704,23 +720,19 @@ static int xs_tcp_send_request(struct rpc_task *task)
704 /* Should we call xs_close() here? */ 720 /* Should we call xs_close() here? */
705 break; 721 break;
706 case -EAGAIN: 722 case -EAGAIN:
707 xs_nospace(task); 723 status = xs_nospace(task);
708 break; 724 break;
725 default:
726 dprintk("RPC: sendmsg returned unrecognized error %d\n",
727 -status);
709 case -ECONNRESET: 728 case -ECONNRESET:
729 case -EPIPE:
710 xs_tcp_shutdown(xprt); 730 xs_tcp_shutdown(xprt);
711 case -ECONNREFUSED: 731 case -ECONNREFUSED:
712 case -ENOTCONN: 732 case -ENOTCONN:
713 case -EPIPE:
714 status = -ENOTCONN;
715 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
716 break;
717 default:
718 dprintk("RPC: sendmsg returned unrecognized error %d\n",
719 -status);
720 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); 733 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
721 xs_tcp_shutdown(xprt);
722 } 734 }
723 735out:
724 return status; 736 return status;
725} 737}
726 738
@@ -767,23 +779,13 @@ static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *s
767 sk->sk_error_report = transport->old_error_report; 779 sk->sk_error_report = transport->old_error_report;
768} 780}
769 781
770/** 782static void xs_reset_transport(struct sock_xprt *transport)
771 * xs_close - close a socket
772 * @xprt: transport
773 *
774 * This is used when all requests are complete; ie, no DRC state remains
775 * on the server we want to save.
776 */
777static void xs_close(struct rpc_xprt *xprt)
778{ 783{
779 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
780 struct socket *sock = transport->sock; 784 struct socket *sock = transport->sock;
781 struct sock *sk = transport->inet; 785 struct sock *sk = transport->inet;
782 786
783 if (!sk) 787 if (sk == NULL)
784 goto clear_close_wait; 788 return;
785
786 dprintk("RPC: xs_close xprt %p\n", xprt);
787 789
788 write_lock_bh(&sk->sk_callback_lock); 790 write_lock_bh(&sk->sk_callback_lock);
789 transport->inet = NULL; 791 transport->inet = NULL;
@@ -797,8 +799,25 @@ static void xs_close(struct rpc_xprt *xprt)
797 sk->sk_no_check = 0; 799 sk->sk_no_check = 0;
798 800
799 sock_release(sock); 801 sock_release(sock);
800clear_close_wait: 802}
803
804/**
805 * xs_close - close a socket
806 * @xprt: transport
807 *
808 * This is used when all requests are complete; ie, no DRC state remains
809 * on the server we want to save.
810 */
811static void xs_close(struct rpc_xprt *xprt)
812{
813 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
814
815 dprintk("RPC: xs_close xprt %p\n", xprt);
816
817 xs_reset_transport(transport);
818
801 smp_mb__before_clear_bit(); 819 smp_mb__before_clear_bit();
820 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
802 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 821 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
803 clear_bit(XPRT_CLOSING, &xprt->state); 822 clear_bit(XPRT_CLOSING, &xprt->state);
804 smp_mb__after_clear_bit(); 823 smp_mb__after_clear_bit();
@@ -1126,6 +1145,47 @@ out:
1126 read_unlock(&sk->sk_callback_lock); 1145 read_unlock(&sk->sk_callback_lock);
1127} 1146}
1128 1147
1148/*
1149 * Do the equivalent of linger/linger2 handling for dealing with
1150 * broken servers that don't close the socket in a timely
1151 * fashion
1152 */
1153static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1154 unsigned long timeout)
1155{
1156 struct sock_xprt *transport;
1157
1158 if (xprt_test_and_set_connecting(xprt))
1159 return;
1160 set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1161 transport = container_of(xprt, struct sock_xprt, xprt);
1162 queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1163 timeout);
1164}
1165
1166static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1167{
1168 struct sock_xprt *transport;
1169
1170 transport = container_of(xprt, struct sock_xprt, xprt);
1171
1172 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1173 !cancel_delayed_work(&transport->connect_worker))
1174 return;
1175 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1176 xprt_clear_connecting(xprt);
1177}
1178
1179static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1180{
1181 smp_mb__before_clear_bit();
1182 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1183 clear_bit(XPRT_CLOSING, &xprt->state);
1184 smp_mb__after_clear_bit();
1185 /* Mark transport as closed and wake up all pending tasks */
1186 xprt_disconnect_done(xprt);
1187}
1188
1129/** 1189/**
1130 * xs_tcp_state_change - callback to handle TCP socket state changes 1190 * xs_tcp_state_change - callback to handle TCP socket state changes
1131 * @sk: socket whose state has changed 1191 * @sk: socket whose state has changed
@@ -1158,7 +1218,7 @@ static void xs_tcp_state_change(struct sock *sk)
1158 transport->tcp_flags = 1218 transport->tcp_flags =
1159 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID; 1219 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
1160 1220
1161 xprt_wake_pending_tasks(xprt, 0); 1221 xprt_wake_pending_tasks(xprt, -EAGAIN);
1162 } 1222 }
1163 spin_unlock_bh(&xprt->transport_lock); 1223 spin_unlock_bh(&xprt->transport_lock);
1164 break; 1224 break;
@@ -1171,10 +1231,10 @@ static void xs_tcp_state_change(struct sock *sk)
1171 clear_bit(XPRT_CONNECTED, &xprt->state); 1231 clear_bit(XPRT_CONNECTED, &xprt->state);
1172 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1232 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1173 smp_mb__after_clear_bit(); 1233 smp_mb__after_clear_bit();
1234 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1174 break; 1235 break;
1175 case TCP_CLOSE_WAIT: 1236 case TCP_CLOSE_WAIT:
1176 /* The server initiated a shutdown of the socket */ 1237 /* The server initiated a shutdown of the socket */
1177 set_bit(XPRT_CLOSING, &xprt->state);
1178 xprt_force_disconnect(xprt); 1238 xprt_force_disconnect(xprt);
1179 case TCP_SYN_SENT: 1239 case TCP_SYN_SENT:
1180 xprt->connect_cookie++; 1240 xprt->connect_cookie++;
@@ -1187,40 +1247,35 @@ static void xs_tcp_state_change(struct sock *sk)
1187 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; 1247 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
1188 break; 1248 break;
1189 case TCP_LAST_ACK: 1249 case TCP_LAST_ACK:
1250 set_bit(XPRT_CLOSING, &xprt->state);
1251 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
1190 smp_mb__before_clear_bit(); 1252 smp_mb__before_clear_bit();
1191 clear_bit(XPRT_CONNECTED, &xprt->state); 1253 clear_bit(XPRT_CONNECTED, &xprt->state);
1192 smp_mb__after_clear_bit(); 1254 smp_mb__after_clear_bit();
1193 break; 1255 break;
1194 case TCP_CLOSE: 1256 case TCP_CLOSE:
1195 smp_mb__before_clear_bit(); 1257 xs_tcp_cancel_linger_timeout(xprt);
1196 clear_bit(XPRT_CLOSE_WAIT, &xprt->state); 1258 xs_sock_mark_closed(xprt);
1197 clear_bit(XPRT_CLOSING, &xprt->state);
1198 smp_mb__after_clear_bit();
1199 /* Mark transport as closed and wake up all pending tasks */
1200 xprt_disconnect_done(xprt);
1201 } 1259 }
1202 out: 1260 out:
1203 read_unlock(&sk->sk_callback_lock); 1261 read_unlock(&sk->sk_callback_lock);
1204} 1262}
1205 1263
1206/** 1264/**
1207 * xs_tcp_error_report - callback mainly for catching RST events 1265 * xs_error_report - callback mainly for catching socket errors
1208 * @sk: socket 1266 * @sk: socket
1209 */ 1267 */
1210static void xs_tcp_error_report(struct sock *sk) 1268static void xs_error_report(struct sock *sk)
1211{ 1269{
1212 struct rpc_xprt *xprt; 1270 struct rpc_xprt *xprt;
1213 1271
1214 read_lock(&sk->sk_callback_lock); 1272 read_lock(&sk->sk_callback_lock);
1215 if (sk->sk_err != ECONNRESET || sk->sk_state != TCP_ESTABLISHED)
1216 goto out;
1217 if (!(xprt = xprt_from_sock(sk))) 1273 if (!(xprt = xprt_from_sock(sk)))
1218 goto out; 1274 goto out;
1219 dprintk("RPC: %s client %p...\n" 1275 dprintk("RPC: %s client %p...\n"
1220 "RPC: error %d\n", 1276 "RPC: error %d\n",
1221 __func__, xprt, sk->sk_err); 1277 __func__, xprt, sk->sk_err);
1222 1278 xprt_wake_pending_tasks(xprt, -EAGAIN);
1223 xprt_force_disconnect(xprt);
1224out: 1279out:
1225 read_unlock(&sk->sk_callback_lock); 1280 read_unlock(&sk->sk_callback_lock);
1226} 1281}
@@ -1494,6 +1549,7 @@ static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1494 sk->sk_user_data = xprt; 1549 sk->sk_user_data = xprt;
1495 sk->sk_data_ready = xs_udp_data_ready; 1550 sk->sk_data_ready = xs_udp_data_ready;
1496 sk->sk_write_space = xs_udp_write_space; 1551 sk->sk_write_space = xs_udp_write_space;
1552 sk->sk_error_report = xs_error_report;
1497 sk->sk_no_check = UDP_CSUM_NORCV; 1553 sk->sk_no_check = UDP_CSUM_NORCV;
1498 sk->sk_allocation = GFP_ATOMIC; 1554 sk->sk_allocation = GFP_ATOMIC;
1499 1555
@@ -1526,9 +1582,10 @@ static void xs_udp_connect_worker4(struct work_struct *work)
1526 goto out; 1582 goto out;
1527 1583
1528 /* Start by resetting any existing state */ 1584 /* Start by resetting any existing state */
1529 xs_close(xprt); 1585 xs_reset_transport(transport);
1530 1586
1531 if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { 1587 err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
1588 if (err < 0) {
1532 dprintk("RPC: can't create UDP transport socket (%d).\n", -err); 1589 dprintk("RPC: can't create UDP transport socket (%d).\n", -err);
1533 goto out; 1590 goto out;
1534 } 1591 }
@@ -1545,8 +1602,8 @@ static void xs_udp_connect_worker4(struct work_struct *work)
1545 xs_udp_finish_connecting(xprt, sock); 1602 xs_udp_finish_connecting(xprt, sock);
1546 status = 0; 1603 status = 0;
1547out: 1604out:
1548 xprt_wake_pending_tasks(xprt, status);
1549 xprt_clear_connecting(xprt); 1605 xprt_clear_connecting(xprt);
1606 xprt_wake_pending_tasks(xprt, status);
1550} 1607}
1551 1608
1552/** 1609/**
@@ -1567,9 +1624,10 @@ static void xs_udp_connect_worker6(struct work_struct *work)
1567 goto out; 1624 goto out;
1568 1625
1569 /* Start by resetting any existing state */ 1626 /* Start by resetting any existing state */
1570 xs_close(xprt); 1627 xs_reset_transport(transport);
1571 1628
1572 if ((err = sock_create_kern(PF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { 1629 err = sock_create_kern(PF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
1630 if (err < 0) {
1573 dprintk("RPC: can't create UDP transport socket (%d).\n", -err); 1631 dprintk("RPC: can't create UDP transport socket (%d).\n", -err);
1574 goto out; 1632 goto out;
1575 } 1633 }
@@ -1586,18 +1644,17 @@ static void xs_udp_connect_worker6(struct work_struct *work)
1586 xs_udp_finish_connecting(xprt, sock); 1644 xs_udp_finish_connecting(xprt, sock);
1587 status = 0; 1645 status = 0;
1588out: 1646out:
1589 xprt_wake_pending_tasks(xprt, status);
1590 xprt_clear_connecting(xprt); 1647 xprt_clear_connecting(xprt);
1648 xprt_wake_pending_tasks(xprt, status);
1591} 1649}
1592 1650
1593/* 1651/*
1594 * We need to preserve the port number so the reply cache on the server can 1652 * We need to preserve the port number so the reply cache on the server can
1595 * find our cached RPC replies when we get around to reconnecting. 1653 * find our cached RPC replies when we get around to reconnecting.
1596 */ 1654 */
1597static void xs_tcp_reuse_connection(struct rpc_xprt *xprt) 1655static void xs_abort_connection(struct rpc_xprt *xprt, struct sock_xprt *transport)
1598{ 1656{
1599 int result; 1657 int result;
1600 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1601 struct sockaddr any; 1658 struct sockaddr any;
1602 1659
1603 dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); 1660 dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt);
@@ -1609,11 +1666,24 @@ static void xs_tcp_reuse_connection(struct rpc_xprt *xprt)
1609 memset(&any, 0, sizeof(any)); 1666 memset(&any, 0, sizeof(any));
1610 any.sa_family = AF_UNSPEC; 1667 any.sa_family = AF_UNSPEC;
1611 result = kernel_connect(transport->sock, &any, sizeof(any), 0); 1668 result = kernel_connect(transport->sock, &any, sizeof(any), 0);
1612 if (result) 1669 if (!result)
1670 xs_sock_mark_closed(xprt);
1671 else
1613 dprintk("RPC: AF_UNSPEC connect return code %d\n", 1672 dprintk("RPC: AF_UNSPEC connect return code %d\n",
1614 result); 1673 result);
1615} 1674}
1616 1675
1676static void xs_tcp_reuse_connection(struct rpc_xprt *xprt, struct sock_xprt *transport)
1677{
1678 unsigned int state = transport->inet->sk_state;
1679
1680 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED)
1681 return;
1682 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT))
1683 return;
1684 xs_abort_connection(xprt, transport);
1685}
1686
1617static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1687static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1618{ 1688{
1619 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1689 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
@@ -1629,7 +1699,7 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1629 sk->sk_data_ready = xs_tcp_data_ready; 1699 sk->sk_data_ready = xs_tcp_data_ready;
1630 sk->sk_state_change = xs_tcp_state_change; 1700 sk->sk_state_change = xs_tcp_state_change;
1631 sk->sk_write_space = xs_tcp_write_space; 1701 sk->sk_write_space = xs_tcp_write_space;
1632 sk->sk_error_report = xs_tcp_error_report; 1702 sk->sk_error_report = xs_error_report;
1633 sk->sk_allocation = GFP_ATOMIC; 1703 sk->sk_allocation = GFP_ATOMIC;
1634 1704
1635 /* socket options */ 1705 /* socket options */
@@ -1657,37 +1727,42 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1657} 1727}
1658 1728
1659/** 1729/**
1660 * xs_tcp_connect_worker4 - connect a TCP socket to a remote endpoint 1730 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
1661 * @work: RPC transport to connect 1731 * @xprt: RPC transport to connect
1732 * @transport: socket transport to connect
1733 * @create_sock: function to create a socket of the correct type
1662 * 1734 *
1663 * Invoked by a work queue tasklet. 1735 * Invoked by a work queue tasklet.
1664 */ 1736 */
1665static void xs_tcp_connect_worker4(struct work_struct *work) 1737static void xs_tcp_setup_socket(struct rpc_xprt *xprt,
1738 struct sock_xprt *transport,
1739 struct socket *(*create_sock)(struct rpc_xprt *,
1740 struct sock_xprt *))
1666{ 1741{
1667 struct sock_xprt *transport =
1668 container_of(work, struct sock_xprt, connect_worker.work);
1669 struct rpc_xprt *xprt = &transport->xprt;
1670 struct socket *sock = transport->sock; 1742 struct socket *sock = transport->sock;
1671 int err, status = -EIO; 1743 int status = -EIO;
1672 1744
1673 if (xprt->shutdown) 1745 if (xprt->shutdown)
1674 goto out; 1746 goto out;
1675 1747
1676 if (!sock) { 1748 if (!sock) {
1677 /* start from scratch */ 1749 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1678 if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { 1750 sock = create_sock(xprt, transport);
1679 dprintk("RPC: can't create TCP transport socket (%d).\n", -err); 1751 if (IS_ERR(sock)) {
1752 status = PTR_ERR(sock);
1680 goto out; 1753 goto out;
1681 } 1754 }
1682 xs_reclassify_socket4(sock); 1755 } else {
1756 int abort_and_exit;
1683 1757
1684 if (xs_bind4(transport, sock) < 0) { 1758 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
1685 sock_release(sock); 1759 &xprt->state);
1686 goto out;
1687 }
1688 } else
1689 /* "close" the socket, preserving the local port */ 1760 /* "close" the socket, preserving the local port */
1690 xs_tcp_reuse_connection(xprt); 1761 xs_tcp_reuse_connection(xprt, transport);
1762
1763 if (abort_and_exit)
1764 goto out_eagain;
1765 }
1691 1766
1692 dprintk("RPC: worker connecting xprt %p to address: %s\n", 1767 dprintk("RPC: worker connecting xprt %p to address: %s\n",
1693 xprt, xprt->address_strings[RPC_DISPLAY_ALL]); 1768 xprt, xprt->address_strings[RPC_DISPLAY_ALL]);
@@ -1696,83 +1771,104 @@ static void xs_tcp_connect_worker4(struct work_struct *work)
1696 dprintk("RPC: %p connect status %d connected %d sock state %d\n", 1771 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
1697 xprt, -status, xprt_connected(xprt), 1772 xprt, -status, xprt_connected(xprt),
1698 sock->sk->sk_state); 1773 sock->sk->sk_state);
1699 if (status < 0) { 1774 switch (status) {
1700 switch (status) { 1775 case -ECONNREFUSED:
1701 case -EINPROGRESS: 1776 case -ECONNRESET:
1702 case -EALREADY: 1777 case -ENETUNREACH:
1703 goto out_clear; 1778 /* retry with existing socket, after a delay */
1704 case -ECONNREFUSED: 1779 case 0:
1705 case -ECONNRESET: 1780 case -EINPROGRESS:
1706 /* retry with existing socket, after a delay */ 1781 case -EALREADY:
1707 break; 1782 xprt_clear_connecting(xprt);
1708 default: 1783 return;
1709 /* get rid of existing socket, and retry */
1710 xs_tcp_shutdown(xprt);
1711 }
1712 } 1784 }
1785 /* get rid of existing socket, and retry */
1786 xs_tcp_shutdown(xprt);
1787 printk("%s: connect returned unhandled error %d\n",
1788 __func__, status);
1789out_eagain:
1790 status = -EAGAIN;
1713out: 1791out:
1714 xprt_wake_pending_tasks(xprt, status);
1715out_clear:
1716 xprt_clear_connecting(xprt); 1792 xprt_clear_connecting(xprt);
1793 xprt_wake_pending_tasks(xprt, status);
1794}
1795
1796static struct socket *xs_create_tcp_sock4(struct rpc_xprt *xprt,
1797 struct sock_xprt *transport)
1798{
1799 struct socket *sock;
1800 int err;
1801
1802 /* start from scratch */
1803 err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
1804 if (err < 0) {
1805 dprintk("RPC: can't create TCP transport socket (%d).\n",
1806 -err);
1807 goto out_err;
1808 }
1809 xs_reclassify_socket4(sock);
1810
1811 if (xs_bind4(transport, sock) < 0) {
1812 sock_release(sock);
1813 goto out_err;
1814 }
1815 return sock;
1816out_err:
1817 return ERR_PTR(-EIO);
1717} 1818}
1718 1819
1719/** 1820/**
1720 * xs_tcp_connect_worker6 - connect a TCP socket to a remote endpoint 1821 * xs_tcp_connect_worker4 - connect a TCP socket to a remote endpoint
1721 * @work: RPC transport to connect 1822 * @work: RPC transport to connect
1722 * 1823 *
1723 * Invoked by a work queue tasklet. 1824 * Invoked by a work queue tasklet.
1724 */ 1825 */
1725static void xs_tcp_connect_worker6(struct work_struct *work) 1826static void xs_tcp_connect_worker4(struct work_struct *work)
1726{ 1827{
1727 struct sock_xprt *transport = 1828 struct sock_xprt *transport =
1728 container_of(work, struct sock_xprt, connect_worker.work); 1829 container_of(work, struct sock_xprt, connect_worker.work);
1729 struct rpc_xprt *xprt = &transport->xprt; 1830 struct rpc_xprt *xprt = &transport->xprt;
1730 struct socket *sock = transport->sock;
1731 int err, status = -EIO;
1732 1831
1733 if (xprt->shutdown) 1832 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock4);
1734 goto out; 1833}
1735 1834
1736 if (!sock) { 1835static struct socket *xs_create_tcp_sock6(struct rpc_xprt *xprt,
1737 /* start from scratch */ 1836 struct sock_xprt *transport)
1738 if ((err = sock_create_kern(PF_INET6, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { 1837{
1739 dprintk("RPC: can't create TCP transport socket (%d).\n", -err); 1838 struct socket *sock;
1740 goto out; 1839 int err;
1741 } 1840
1742 xs_reclassify_socket6(sock); 1841 /* start from scratch */
1842 err = sock_create_kern(PF_INET6, SOCK_STREAM, IPPROTO_TCP, &sock);
1843 if (err < 0) {
1844 dprintk("RPC: can't create TCP transport socket (%d).\n",
1845 -err);
1846 goto out_err;
1847 }
1848 xs_reclassify_socket6(sock);
1743 1849
1744 if (xs_bind6(transport, sock) < 0) { 1850 if (xs_bind6(transport, sock) < 0) {
1745 sock_release(sock); 1851 sock_release(sock);
1746 goto out; 1852 goto out_err;
1747 } 1853 }
1748 } else 1854 return sock;
1749 /* "close" the socket, preserving the local port */ 1855out_err:
1750 xs_tcp_reuse_connection(xprt); 1856 return ERR_PTR(-EIO);
1857}
1751 1858
1752 dprintk("RPC: worker connecting xprt %p to address: %s\n", 1859/**
1753 xprt, xprt->address_strings[RPC_DISPLAY_ALL]); 1860 * xs_tcp_connect_worker6 - connect a TCP socket to a remote endpoint
1861 * @work: RPC transport to connect
1862 *
1863 * Invoked by a work queue tasklet.
1864 */
1865static void xs_tcp_connect_worker6(struct work_struct *work)
1866{
1867 struct sock_xprt *transport =
1868 container_of(work, struct sock_xprt, connect_worker.work);
1869 struct rpc_xprt *xprt = &transport->xprt;
1754 1870
1755 status = xs_tcp_finish_connecting(xprt, sock); 1871 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock6);
1756 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
1757 xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
1758 if (status < 0) {
1759 switch (status) {
1760 case -EINPROGRESS:
1761 case -EALREADY:
1762 goto out_clear;
1763 case -ECONNREFUSED:
1764 case -ECONNRESET:
1765 /* retry with existing socket, after a delay */
1766 break;
1767 default:
1768 /* get rid of existing socket, and retry */
1769 xs_tcp_shutdown(xprt);
1770 }
1771 }
1772out:
1773 xprt_wake_pending_tasks(xprt, status);
1774out_clear:
1775 xprt_clear_connecting(xprt);
1776} 1872}
1777 1873
1778/** 1874/**
@@ -1817,9 +1913,6 @@ static void xs_tcp_connect(struct rpc_task *task)
1817{ 1913{
1818 struct rpc_xprt *xprt = task->tk_xprt; 1914 struct rpc_xprt *xprt = task->tk_xprt;
1819 1915
1820 /* Initiate graceful shutdown of the socket if not already done */
1821 if (test_bit(XPRT_CONNECTED, &xprt->state))
1822 xs_tcp_shutdown(xprt);
1823 /* Exit if we need to wait for socket shutdown to complete */ 1916 /* Exit if we need to wait for socket shutdown to complete */
1824 if (test_bit(XPRT_CLOSING, &xprt->state)) 1917 if (test_bit(XPRT_CLOSING, &xprt->state))
1825 return; 1918 return;
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-15 03:37:40 -0500