aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@woody.linux-foundation.org>2007-07-13 19:46:18 -0400
committerLinus Torvalds <torvalds@woody.linux-foundation.org>2007-07-13 19:46:18 -0400
commit16cefa8c3863721fd40445a1b34dea18cd16ccfe (patch)
treec8e58ca06e2edfd667d3e6062a642b80cc58e5e7
parent4fbef206daead133085fe33905f5e842d38fb8da (diff)
parentd8558f99fbc5ef5d4ae76b893784005056450f82 (diff)
Merge git://git.linux-nfs.org/pub/linux/nfs-2.6
* git://git.linux-nfs.org/pub/linux/nfs-2.6: (122 commits) sunrpc: drop BKL around wrap and unwrap NFSv4: Make sure unlock is really an unlock when cancelling a lock NLM: fix source address of callback to client SUNRPC client: add interface for binding to a local address SUNRPC server: record the destination address of a request SUNRPC: cleanup transport creation argument passing NFSv4: Make the NFS state model work with the nosharedcache mount option NFS: Error when mounting the same filesystem with different options NFS: Add the mount option "nosharecache" NFS: Add support for mounting NFSv4 file systems with string options NFS: Add final pieces to support in-kernel mount option parsing NFS: Introduce generic mount client API NFS: Add enums and match tables for mount option parsing NFS: Improve debugging output in NFS in-kernel mount client NFS: Clean up in-kernel NFS mount NFS: Remake nfsroot_mount as a permanent part of NFS client SUNRPC: Add a convenient default for the hostname when calling rpc_create() SUNRPC: Rename rpcb_getport to be consistent with new rpcb_getport_sync name SUNRPC: Rename rpcb_getport_external routine SUNRPC: Allow rpcbind requests to be interrupted by a signal. ...
Diffstat
-rw-r--r--fs/lockd/host.c39
-rw-r--r--fs/lockd/mon.c2
-rw-r--r--fs/lockd/svc.c6
-rw-r--r--fs/nfs/Makefile4
-rw-r--r--fs/nfs/client.c28
-rw-r--r--fs/nfs/delegation.c186
-rw-r--r--fs/nfs/delegation.h26
-rw-r--r--fs/nfs/dir.c16
-rw-r--r--fs/nfs/direct.c34
-rw-r--r--fs/nfs/inode.c73
-rw-r--r--fs/nfs/internal.h4
-rw-r--r--fs/nfs/mount_clnt.c169
-rw-r--r--fs/nfs/nfs2xdr.c6
-rw-r--r--fs/nfs/nfs3proc.c4
-rw-r--r--fs/nfs/nfs3xdr.c8
-rw-r--r--fs/nfs/nfs4_fs.h40
-rw-r--r--fs/nfs/nfs4proc.c760
-rw-r--r--fs/nfs/nfs4state.c310
-rw-r--r--fs/nfs/nfs4xdr.c126
-rw-r--r--fs/nfs/nfsroot.c5
-rw-r--r--fs/nfs/pagelist.c60
-rw-r--r--fs/nfs/read.c40
-rw-r--r--fs/nfs/super.c1189
-rw-r--r--fs/nfs/write.c149
-rw-r--r--fs/nfsd/nfs4callback.c18
-rw-r--r--fs/nfsd/nfs4state.c1
-rw-r--r--include/linux/lockd/lockd.h1
-rw-r--r--include/linux/nfs4.h1
-rw-r--r--include/linux/nfs4_mount.h3
-rw-r--r--include/linux/nfs_fs.h28
-rw-r--r--include/linux/nfs_fs_sb.h8
-rw-r--r--include/linux/nfs_mount.h3
-rw-r--r--include/linux/nfs_page.h25
-rw-r--r--include/linux/nfs_xdr.h5
-rw-r--r--include/linux/sunrpc/auth.h48
-rw-r--r--include/linux/sunrpc/auth_gss.h6
-rw-r--r--include/linux/sunrpc/clnt.h33
-rw-r--r--include/linux/sunrpc/gss_api.h2
-rw-r--r--include/linux/sunrpc/rpc_pipe_fs.h2
-rw-r--r--include/linux/sunrpc/sched.h6
-rw-r--r--include/linux/sunrpc/svcsock.h1
-rw-r--r--include/linux/sunrpc/xprt.h16
-rw-r--r--net/sunrpc/auth.c370
-rw-r--r--net/sunrpc/auth_gss/auth_gss.c349
-rw-r--r--net/sunrpc/auth_gss/gss_krb5_mech.c2
-rw-r--r--net/sunrpc/auth_gss/gss_spkm3_mech.c2
-rw-r--r--net/sunrpc/auth_null.c10
-rw-r--r--net/sunrpc/auth_unix.c54
-rw-r--r--net/sunrpc/clnt.c371
-rw-r--r--net/sunrpc/rpc_pipe.c80
-rw-r--r--net/sunrpc/rpcb_clnt.c65
-rw-r--r--net/sunrpc/sched.c209
-rw-r--r--net/sunrpc/sunrpc_syms.c8
-rw-r--r--net/sunrpc/svcsock.c20
-rw-r--r--net/sunrpc/xprt.c19
-rw-r--r--net/sunrpc/xprtsock.c81
56 files changed, 3321 insertions, 1810 deletions
diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index 96070bff93fc..572601e98dcd 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -44,9 +44,8 @@ static struct nsm_handle * nsm_find(const struct sockaddr_in *sin,
44 */ 44 */
45static struct nlm_host * 45static struct nlm_host *
46nlm_lookup_host(int server, const struct sockaddr_in *sin, 46nlm_lookup_host(int server, const struct sockaddr_in *sin,
47 int proto, int version, 47 int proto, int version, const char *hostname,
48 const char *hostname, 48 int hostname_len, const struct sockaddr_in *ssin)
49 int hostname_len)
50{ 49{
51 struct hlist_head *chain; 50 struct hlist_head *chain;
52 struct hlist_node *pos; 51 struct hlist_node *pos;
@@ -54,7 +53,9 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin,
54 struct nsm_handle *nsm = NULL; 53 struct nsm_handle *nsm = NULL;
55 int hash; 54 int hash;
56 55
57 dprintk("lockd: nlm_lookup_host(%u.%u.%u.%u, p=%d, v=%d, my role=%s, name=%.*s)\n", 56 dprintk("lockd: nlm_lookup_host("NIPQUAD_FMT"->"NIPQUAD_FMT
57 ", p=%d, v=%d, my role=%s, name=%.*s)\n",
58 NIPQUAD(ssin->sin_addr.s_addr),
58 NIPQUAD(sin->sin_addr.s_addr), proto, version, 59 NIPQUAD(sin->sin_addr.s_addr), proto, version,
59 server? "server" : "client", 60 server? "server" : "client",
60 hostname_len, 61 hostname_len,
@@ -91,6 +92,8 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin,
91 continue; 92 continue;
92 if (host->h_server != server) 93 if (host->h_server != server)
93 continue; 94 continue;
95 if (!nlm_cmp_addr(&host->h_saddr, ssin))
96 continue;
94 97
95 /* Move to head of hash chain. */ 98 /* Move to head of hash chain. */
96 hlist_del(&host->h_hash); 99 hlist_del(&host->h_hash);
@@ -118,6 +121,7 @@ nlm_lookup_host(int server, const struct sockaddr_in *sin,
118 host->h_name = nsm->sm_name; 121 host->h_name = nsm->sm_name;
119 host->h_addr = *sin; 122 host->h_addr = *sin;
120 host->h_addr.sin_port = 0; /* ouch! */ 123 host->h_addr.sin_port = 0; /* ouch! */
124 host->h_saddr = *ssin;
121 host->h_version = version; 125 host->h_version = version;
122 host->h_proto = proto; 126 host->h_proto = proto;
123 host->h_rpcclnt = NULL; 127 host->h_rpcclnt = NULL;
@@ -161,15 +165,9 @@ nlm_destroy_host(struct nlm_host *host)
161 */ 165 */
162 nsm_unmonitor(host); 166 nsm_unmonitor(host);
163 167
164 if ((clnt = host->h_rpcclnt) != NULL) { 168 clnt = host->h_rpcclnt;
165 if (atomic_read(&clnt->cl_users)) { 169 if (clnt != NULL)
166 printk(KERN_WARNING 170 rpc_shutdown_client(clnt);
167 "lockd: active RPC handle\n");
168 clnt->cl_dead = 1;
169 } else {
170 rpc_destroy_client(host->h_rpcclnt);
171 }
172 }
173 kfree(host); 171 kfree(host);
174} 172}
175 173
@@ -180,8 +178,10 @@ struct nlm_host *
180nlmclnt_lookup_host(const struct sockaddr_in *sin, int proto, int version, 178nlmclnt_lookup_host(const struct sockaddr_in *sin, int proto, int version,
181 const char *hostname, int hostname_len) 179 const char *hostname, int hostname_len)
182{ 180{
181 struct sockaddr_in ssin = {0};
182
183 return nlm_lookup_host(0, sin, proto, version, 183 return nlm_lookup_host(0, sin, proto, version,
184 hostname, hostname_len); 184 hostname, hostname_len, &ssin);
185} 185}
186 186
187/* 187/*
@@ -191,9 +191,12 @@ struct nlm_host *
191nlmsvc_lookup_host(struct svc_rqst *rqstp, 191nlmsvc_lookup_host(struct svc_rqst *rqstp,
192 const char *hostname, int hostname_len) 192 const char *hostname, int hostname_len)
193{ 193{
194 struct sockaddr_in ssin = {0};
195
196 ssin.sin_addr = rqstp->rq_daddr.addr;
194 return nlm_lookup_host(1, svc_addr_in(rqstp), 197 return nlm_lookup_host(1, svc_addr_in(rqstp),
195 rqstp->rq_prot, rqstp->rq_vers, 198 rqstp->rq_prot, rqstp->rq_vers,
196 hostname, hostname_len); 199 hostname, hostname_len, &ssin);
197} 200}
198 201
199/* 202/*
@@ -204,8 +207,9 @@ nlm_bind_host(struct nlm_host *host)
204{ 207{
205 struct rpc_clnt *clnt; 208 struct rpc_clnt *clnt;
206 209
207 dprintk("lockd: nlm_bind_host(%08x)\n", 210 dprintk("lockd: nlm_bind_host("NIPQUAD_FMT"->"NIPQUAD_FMT")\n",
208 (unsigned)ntohl(host->h_addr.sin_addr.s_addr)); 211 NIPQUAD(host->h_saddr.sin_addr),
212 NIPQUAD(host->h_addr.sin_addr));
209 213
210 /* Lock host handle */ 214 /* Lock host handle */
211 mutex_lock(&host->h_mutex); 215 mutex_lock(&host->h_mutex);
@@ -232,6 +236,7 @@ nlm_bind_host(struct nlm_host *host)
232 .protocol = host->h_proto, 236 .protocol = host->h_proto,
233 .address = (struct sockaddr *)&host->h_addr, 237 .address = (struct sockaddr *)&host->h_addr,
234 .addrsize = sizeof(host->h_addr), 238 .addrsize = sizeof(host->h_addr),
239 .saddress = (struct sockaddr *)&host->h_saddr,
235 .timeout = &timeparms, 240 .timeout = &timeparms,
236 .servername = host->h_name, 241 .servername = host->h_name,
237 .program = &nlm_program, 242 .program = &nlm_program,
diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c
index 2102e2d0134d..3353ed8421a7 100644
--- a/fs/lockd/mon.c
+++ b/fs/lockd/mon.c
@@ -61,6 +61,7 @@ nsm_mon_unmon(struct nsm_handle *nsm, u32 proc, struct nsm_res *res)
61 status); 61 status);
62 else 62 else
63 status = 0; 63 status = 0;
64 rpc_shutdown_client(clnt);
64 out: 65 out:
65 return status; 66 return status;
66} 67}
@@ -138,7 +139,6 @@ nsm_create(void)
138 .program = &nsm_program, 139 .program = &nsm_program,
139 .version = SM_VERSION, 140 .version = SM_VERSION,
140 .authflavor = RPC_AUTH_NULL, 141 .authflavor = RPC_AUTH_NULL,
141 .flags = (RPC_CLNT_CREATE_ONESHOT),
142 }; 142 };
143 143
144 return rpc_create(&args); 144 return rpc_create(&args);
diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index 126b1bf02c0e..26809325469c 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -123,9 +123,6 @@ lockd(struct svc_rqst *rqstp)
123 /* Process request with signals blocked, but allow SIGKILL. */ 123 /* Process request with signals blocked, but allow SIGKILL. */
124 allow_signal(SIGKILL); 124 allow_signal(SIGKILL);
125 125
126 /* kick rpciod */
127 rpciod_up();
128
129 dprintk("NFS locking service started (ver " LOCKD_VERSION ").\n"); 126 dprintk("NFS locking service started (ver " LOCKD_VERSION ").\n");
130 127
131 if (!nlm_timeout) 128 if (!nlm_timeout)
@@ -202,9 +199,6 @@ lockd(struct svc_rqst *rqstp)
202 /* Exit the RPC thread */ 199 /* Exit the RPC thread */
203 svc_exit_thread(rqstp); 200 svc_exit_thread(rqstp);
204 201
205 /* release rpciod */
206 rpciod_down();
207
208 /* Release module */ 202 /* Release module */
209 unlock_kernel(); 203 unlock_kernel();
210 module_put_and_exit(0); 204 module_put_and_exit(0);
diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile
index f4580b44eef4..b55cb236cf74 100644
--- a/fs/nfs/Makefile
+++ b/fs/nfs/Makefile
@@ -6,8 +6,8 @@ obj-$(CONFIG_NFS_FS) += nfs.o
6 6
7nfs-y := client.o dir.o file.o getroot.o inode.o super.o nfs2xdr.o \ 7nfs-y := client.o dir.o file.o getroot.o inode.o super.o nfs2xdr.o \
8 pagelist.o proc.o read.o symlink.o unlink.o \ 8 pagelist.o proc.o read.o symlink.o unlink.o \
9 write.o namespace.o 9 write.o namespace.o mount_clnt.o
10nfs-$(CONFIG_ROOT_NFS) += nfsroot.o mount_clnt.o 10nfs-$(CONFIG_ROOT_NFS) += nfsroot.o
11nfs-$(CONFIG_NFS_V3) += nfs3proc.o nfs3xdr.o 11nfs-$(CONFIG_NFS_V3) += nfs3proc.o nfs3xdr.o
12nfs-$(CONFIG_NFS_V3_ACL) += nfs3acl.o 12nfs-$(CONFIG_NFS_V3_ACL) += nfs3acl.o
13nfs-$(CONFIG_NFS_V4) += nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \ 13nfs-$(CONFIG_NFS_V4) += nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 881fa4900923..ccb455053ee4 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -102,19 +102,10 @@ static struct nfs_client *nfs_alloc_client(const char *hostname,
102 int nfsversion) 102 int nfsversion)
103{ 103{
104 struct nfs_client *clp; 104 struct nfs_client *clp;
105 int error;
106 105
107 if ((clp = kzalloc(sizeof(*clp), GFP_KERNEL)) == NULL) 106 if ((clp = kzalloc(sizeof(*clp), GFP_KERNEL)) == NULL)
108 goto error_0; 107 goto error_0;
109 108
110 error = rpciod_up();
111 if (error < 0) {
112 dprintk("%s: couldn't start rpciod! Error = %d\n",
113 __FUNCTION__, error);
114 goto error_1;
115 }
116 __set_bit(NFS_CS_RPCIOD, &clp->cl_res_state);
117
118 if (nfsversion == 4) { 109 if (nfsversion == 4) {
119 if (nfs_callback_up() < 0) 110 if (nfs_callback_up() < 0)
120 goto error_2; 111 goto error_2;
@@ -139,8 +130,6 @@ static struct nfs_client *nfs_alloc_client(const char *hostname,
139#ifdef CONFIG_NFS_V4 130#ifdef CONFIG_NFS_V4
140 init_rwsem(&clp->cl_sem); 131 init_rwsem(&clp->cl_sem);
141 INIT_LIST_HEAD(&clp->cl_delegations); 132 INIT_LIST_HEAD(&clp->cl_delegations);
142 INIT_LIST_HEAD(&clp->cl_state_owners);
143 INIT_LIST_HEAD(&clp->cl_unused);
144 spin_lock_init(&clp->cl_lock); 133 spin_lock_init(&clp->cl_lock);
145 INIT_DELAYED_WORK(&clp->cl_renewd, nfs4_renew_state); 134 INIT_DELAYED_WORK(&clp->cl_renewd, nfs4_renew_state);
146 rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS client"); 135 rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS client");
@@ -154,9 +143,6 @@ error_3:
154 if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state)) 143 if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state))
155 nfs_callback_down(); 144 nfs_callback_down();
156error_2: 145error_2:
157 rpciod_down();
158 __clear_bit(NFS_CS_RPCIOD, &clp->cl_res_state);
159error_1:
160 kfree(clp); 146 kfree(clp);
161error_0: 147error_0:
162 return NULL; 148 return NULL;
@@ -167,16 +153,7 @@ static void nfs4_shutdown_client(struct nfs_client *clp)
167#ifdef CONFIG_NFS_V4 153#ifdef CONFIG_NFS_V4
168 if (__test_and_clear_bit(NFS_CS_RENEWD, &clp->cl_res_state)) 154 if (__test_and_clear_bit(NFS_CS_RENEWD, &clp->cl_res_state))
169 nfs4_kill_renewd(clp); 155 nfs4_kill_renewd(clp);
170 while (!list_empty(&clp->cl_unused)) { 156 BUG_ON(!RB_EMPTY_ROOT(&clp->cl_state_owners));
171 struct nfs4_state_owner *sp;
172
173 sp = list_entry(clp->cl_unused.next,
174 struct nfs4_state_owner,
175 so_list);
176 list_del(&sp->so_list);
177 kfree(sp);
178 }
179 BUG_ON(!list_empty(&clp->cl_state_owners));
180 if (__test_and_clear_bit(NFS_CS_IDMAP, &clp->cl_res_state)) 157 if (__test_and_clear_bit(NFS_CS_IDMAP, &clp->cl_res_state))
181 nfs_idmap_delete(clp); 158 nfs_idmap_delete(clp);
182#endif 159#endif
@@ -198,9 +175,6 @@ static void nfs_free_client(struct nfs_client *clp)
198 if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state)) 175 if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state))
199 nfs_callback_down(); 176 nfs_callback_down();
200 177
201 if (__test_and_clear_bit(NFS_CS_RPCIOD, &clp->cl_res_state))
202 rpciod_down();
203
204 kfree(clp->cl_hostname); 178 kfree(clp->cl_hostname);
205 kfree(clp); 179 kfree(clp);
206 180
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index 7f37d1bea83f..20ac403469a0 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -27,6 +27,13 @@ static void nfs_free_delegation(struct nfs_delegation *delegation)
27 kfree(delegation); 27 kfree(delegation);
28} 28}
29 29
30static void nfs_free_delegation_callback(struct rcu_head *head)
31{
32 struct nfs_delegation *delegation = container_of(head, struct nfs_delegation, rcu);
33
34 nfs_free_delegation(delegation);
35}
36
30static int nfs_delegation_claim_locks(struct nfs_open_context *ctx, struct nfs4_state *state) 37static int nfs_delegation_claim_locks(struct nfs_open_context *ctx, struct nfs4_state *state)
31{ 38{
32 struct inode *inode = state->inode; 39 struct inode *inode = state->inode;
@@ -57,7 +64,7 @@ out_err:
57 return status; 64 return status;
58} 65}
59 66
60static void nfs_delegation_claim_opens(struct inode *inode) 67static void nfs_delegation_claim_opens(struct inode *inode, const nfs4_stateid *stateid)
61{ 68{
62 struct nfs_inode *nfsi = NFS_I(inode); 69 struct nfs_inode *nfsi = NFS_I(inode);
63 struct nfs_open_context *ctx; 70 struct nfs_open_context *ctx;
@@ -72,9 +79,11 @@ again:
72 continue; 79 continue;
73 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) 80 if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
74 continue; 81 continue;
82 if (memcmp(state->stateid.data, stateid->data, sizeof(state->stateid.data)) != 0)
83 continue;
75 get_nfs_open_context(ctx); 84 get_nfs_open_context(ctx);
76 spin_unlock(&inode->i_lock); 85 spin_unlock(&inode->i_lock);
77 err = nfs4_open_delegation_recall(ctx->dentry, state); 86 err = nfs4_open_delegation_recall(ctx, state, stateid);
78 if (err >= 0) 87 if (err >= 0)
79 err = nfs_delegation_claim_locks(ctx, state); 88 err = nfs_delegation_claim_locks(ctx, state);
80 put_nfs_open_context(ctx); 89 put_nfs_open_context(ctx);
@@ -115,10 +124,6 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct
115 struct nfs_delegation *delegation; 124 struct nfs_delegation *delegation;
116 int status = 0; 125 int status = 0;
117 126
118 /* Ensure we first revalidate the attributes and page cache! */
119 if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR)))
120 __nfs_revalidate_inode(NFS_SERVER(inode), inode);
121
122 delegation = kmalloc(sizeof(*delegation), GFP_KERNEL); 127 delegation = kmalloc(sizeof(*delegation), GFP_KERNEL);
123 if (delegation == NULL) 128 if (delegation == NULL)
124 return -ENOMEM; 129 return -ENOMEM;
@@ -131,10 +136,10 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct
131 delegation->inode = inode; 136 delegation->inode = inode;
132 137
133 spin_lock(&clp->cl_lock); 138 spin_lock(&clp->cl_lock);
134 if (nfsi->delegation == NULL) { 139 if (rcu_dereference(nfsi->delegation) == NULL) {
135 list_add(&delegation->super_list, &clp->cl_delegations); 140 list_add_rcu(&delegation->super_list, &clp->cl_delegations);
136 nfsi->delegation = delegation;
137 nfsi->delegation_state = delegation->type; 141 nfsi->delegation_state = delegation->type;
142 rcu_assign_pointer(nfsi->delegation, delegation);
138 delegation = NULL; 143 delegation = NULL;
139 } else { 144 } else {
140 if (memcmp(&delegation->stateid, &nfsi->delegation->stateid, 145 if (memcmp(&delegation->stateid, &nfsi->delegation->stateid,
@@ -145,6 +150,12 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct
145 status = -EIO; 150 status = -EIO;
146 } 151 }
147 } 152 }
153
154 /* Ensure we revalidate the attributes and page cache! */
155 spin_lock(&inode->i_lock);
156 nfsi->cache_validity |= NFS_INO_REVAL_FORCED;
157 spin_unlock(&inode->i_lock);
158
148 spin_unlock(&clp->cl_lock); 159 spin_unlock(&clp->cl_lock);
149 kfree(delegation); 160 kfree(delegation);
150 return status; 161 return status;
@@ -155,7 +166,7 @@ static int nfs_do_return_delegation(struct inode *inode, struct nfs_delegation *
155 int res = 0; 166 int res = 0;
156 167
157 res = nfs4_proc_delegreturn(inode, delegation->cred, &delegation->stateid); 168 res = nfs4_proc_delegreturn(inode, delegation->cred, &delegation->stateid);
158 nfs_free_delegation(delegation); 169 call_rcu(&delegation->rcu, nfs_free_delegation_callback);
159 return res; 170 return res;
160} 171}
161 172
@@ -170,33 +181,55 @@ static void nfs_msync_inode(struct inode *inode)
170/* 181/*
171 * Basic procedure for returning a delegation to the server 182 * Basic procedure for returning a delegation to the server
172 */ 183 */
173int __nfs_inode_return_delegation(struct inode *inode) 184static int __nfs_inode_return_delegation(struct inode *inode, struct nfs_delegation *delegation)
174{ 185{
175 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client; 186 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
176 struct nfs_inode *nfsi = NFS_I(inode); 187 struct nfs_inode *nfsi = NFS_I(inode);
177 struct nfs_delegation *delegation;
178 int res = 0;
179 188
180 nfs_msync_inode(inode); 189 nfs_msync_inode(inode);
181 down_read(&clp->cl_sem); 190 down_read(&clp->cl_sem);
182 /* Guard against new delegated open calls */ 191 /* Guard against new delegated open calls */
183 down_write(&nfsi->rwsem); 192 down_write(&nfsi->rwsem);
184 spin_lock(&clp->cl_lock); 193 nfs_delegation_claim_opens(inode, &delegation->stateid);
185 delegation = nfsi->delegation;
186 if (delegation != NULL) {
187 list_del_init(&delegation->super_list);
188 nfsi->delegation = NULL;
189 nfsi->delegation_state = 0;
190 }
191 spin_unlock(&clp->cl_lock);
192 nfs_delegation_claim_opens(inode);
193 up_write(&nfsi->rwsem); 194 up_write(&nfsi->rwsem);
194 up_read(&clp->cl_sem); 195 up_read(&clp->cl_sem);
195 nfs_msync_inode(inode); 196 nfs_msync_inode(inode);
196 197
197 if (delegation != NULL) 198 return nfs_do_return_delegation(inode, delegation);
198 res = nfs_do_return_delegation(inode, delegation); 199}
199 return res; 200
201static struct nfs_delegation *nfs_detach_delegation_locked(struct nfs_inode *nfsi, const nfs4_stateid *stateid)
202{
203 struct nfs_delegation *delegation = rcu_dereference(nfsi->delegation);
204
205 if (delegation == NULL)
206 goto nomatch;
207 if (stateid != NULL && memcmp(delegation->stateid.data, stateid->data,
208 sizeof(delegation->stateid.data)) != 0)
209 goto nomatch;
210 list_del_rcu(&delegation->super_list);
211 nfsi->delegation_state = 0;
212 rcu_assign_pointer(nfsi->delegation, NULL);
213 return delegation;
214nomatch:
215 return NULL;
216}
217
218int nfs_inode_return_delegation(struct inode *inode)
219{
220 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
221 struct nfs_inode *nfsi = NFS_I(inode);
222 struct nfs_delegation *delegation;
223 int err = 0;
224
225 if (rcu_dereference(nfsi->delegation) != NULL) {
226 spin_lock(&clp->cl_lock);
227 delegation = nfs_detach_delegation_locked(nfsi, NULL);
228 spin_unlock(&clp->cl_lock);
229 if (delegation != NULL)
230 err = __nfs_inode_return_delegation(inode, delegation);
231 }
232 return err;
200} 233}
201 234
202/* 235/*
@@ -211,19 +244,23 @@ void nfs_return_all_delegations(struct super_block *sb)
211 if (clp == NULL) 244 if (clp == NULL)
212 return; 245 return;
213restart: 246restart:
214 spin_lock(&clp->cl_lock); 247 rcu_read_lock();
215 list_for_each_entry(delegation, &clp->cl_delegations, super_list) { 248 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list) {
216 if (delegation->inode->i_sb != sb) 249 if (delegation->inode->i_sb != sb)
217 continue; 250 continue;
218 inode = igrab(delegation->inode); 251 inode = igrab(delegation->inode);
219 if (inode == NULL) 252 if (inode == NULL)
220 continue; 253 continue;
254 spin_lock(&clp->cl_lock);
255 delegation = nfs_detach_delegation_locked(NFS_I(inode), NULL);
221 spin_unlock(&clp->cl_lock); 256 spin_unlock(&clp->cl_lock);
222 nfs_inode_return_delegation(inode); 257 rcu_read_unlock();
258 if (delegation != NULL)
259 __nfs_inode_return_delegation(inode, delegation);
223 iput(inode); 260 iput(inode);
224 goto restart; 261 goto restart;
225 } 262 }
226 spin_unlock(&clp->cl_lock); 263 rcu_read_unlock();
227} 264}
228 265
229static int nfs_do_expire_all_delegations(void *ptr) 266static int nfs_do_expire_all_delegations(void *ptr)
@@ -234,22 +271,26 @@ static int nfs_do_expire_all_delegations(void *ptr)
234 271
235 allow_signal(SIGKILL); 272 allow_signal(SIGKILL);
236restart: 273restart:
237 spin_lock(&clp->cl_lock);
238 if (test_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) != 0) 274 if (test_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) != 0)
239 goto out; 275 goto out;
240 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) 276 if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0)
241 goto out; 277 goto out;
242 list_for_each_entry(delegation, &clp->cl_delegations, super_list) { 278 rcu_read_lock();
279 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list) {
243 inode = igrab(delegation->inode); 280 inode = igrab(delegation->inode);
244 if (inode == NULL) 281 if (inode == NULL)
245 continue; 282 continue;
283 spin_lock(&clp->cl_lock);
284 delegation = nfs_detach_delegation_locked(NFS_I(inode), NULL);
246 spin_unlock(&clp->cl_lock); 285 spin_unlock(&clp->cl_lock);
247 nfs_inode_return_delegation(inode); 286 rcu_read_unlock();
287 if (delegation)
288 __nfs_inode_return_delegation(inode, delegation);
248 iput(inode); 289 iput(inode);
249 goto restart; 290 goto restart;
250 } 291 }
292 rcu_read_unlock();
251out: 293out:
252 spin_unlock(&clp->cl_lock);
253 nfs_put_client(clp); 294 nfs_put_client(clp);
254 module_put_and_exit(0); 295 module_put_and_exit(0);
255} 296}
@@ -280,17 +321,21 @@ void nfs_handle_cb_pathdown(struct nfs_client *clp)
280 if (clp == NULL) 321 if (clp == NULL)
281 return; 322 return;
282restart: 323restart:
283 spin_lock(&clp->cl_lock); 324 rcu_read_lock();
284 list_for_each_entry(delegation, &clp->cl_delegations, super_list) { 325 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list) {
285 inode = igrab(delegation->inode); 326 inode = igrab(delegation->inode);
286 if (inode == NULL) 327 if (inode == NULL)
287 continue; 328 continue;
329 spin_lock(&clp->cl_lock);
330 delegation = nfs_detach_delegation_locked(NFS_I(inode), NULL);
288 spin_unlock(&clp->cl_lock); 331 spin_unlock(&clp->cl_lock);
289 nfs_inode_return_delegation(inode); 332 rcu_read_unlock();
333 if (delegation != NULL)
334 __nfs_inode_return_delegation(inode, delegation);
290 iput(inode); 335 iput(inode);
291 goto restart; 336 goto restart;
292 } 337 }
293 spin_unlock(&clp->cl_lock); 338 rcu_read_unlock();
294} 339}
295 340
296struct recall_threadargs { 341struct recall_threadargs {
@@ -316,21 +361,14 @@ static int recall_thread(void *data)
316 down_read(&clp->cl_sem); 361 down_read(&clp->cl_sem);
317 down_write(&nfsi->rwsem); 362 down_write(&nfsi->rwsem);
318 spin_lock(&clp->cl_lock); 363 spin_lock(&clp->cl_lock);
319 delegation = nfsi->delegation; 364 delegation = nfs_detach_delegation_locked(nfsi, args->stateid);
320 if (delegation != NULL && memcmp(delegation->stateid.data, 365 if (delegation != NULL)
321 args->stateid->data,
322 sizeof(delegation->stateid.data)) == 0) {
323 list_del_init(&delegation->super_list);
324 nfsi->delegation = NULL;
325 nfsi->delegation_state = 0;
326 args->result = 0; 366 args->result = 0;
327 } else { 367 else
328 delegation = NULL;
329 args->result = -ENOENT; 368 args->result = -ENOENT;
330 }
331 spin_unlock(&clp->cl_lock); 369 spin_unlock(&clp->cl_lock);
332 complete(&args->started); 370 complete(&args->started);
333 nfs_delegation_claim_opens(inode); 371 nfs_delegation_claim_opens(inode, args->stateid);
334 up_write(&nfsi->rwsem); 372 up_write(&nfsi->rwsem);
335 up_read(&clp->cl_sem); 373 up_read(&clp->cl_sem);
336 nfs_msync_inode(inode); 374 nfs_msync_inode(inode);
@@ -371,14 +409,14 @@ struct inode *nfs_delegation_find_inode(struct nfs_client *clp, const struct nfs
371{ 409{
372 struct nfs_delegation *delegation; 410 struct nfs_delegation *delegation;
373 struct inode *res = NULL; 411 struct inode *res = NULL;
374 spin_lock(&clp->cl_lock); 412 rcu_read_lock();
375 list_for_each_entry(delegation, &clp->cl_delegations, super_list) { 413 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list) {
376 if (nfs_compare_fh(fhandle, &NFS_I(delegation->inode)->fh) == 0) { 414 if (nfs_compare_fh(fhandle, &NFS_I(delegation->inode)->fh) == 0) {
377 res = igrab(delegation->inode); 415 res = igrab(delegation->inode);
378 break; 416 break;
379 } 417 }
380 } 418 }
381 spin_unlock(&clp->cl_lock); 419 rcu_read_unlock();
382 return res; 420 return res;
383} 421}
384 422
@@ -388,10 +426,10 @@ struct inode *nfs_delegation_find_inode(struct nfs_client *clp, const struct nfs
388void nfs_delegation_mark_reclaim(struct nfs_client *clp) 426void nfs_delegation_mark_reclaim(struct nfs_client *clp)
389{ 427{
390 struct nfs_delegation *delegation; 428 struct nfs_delegation *delegation;
391 spin_lock(&clp->cl_lock); 429 rcu_read_lock();
392 list_for_each_entry(delegation, &clp->cl_delegations, super_list) 430 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list)
393 delegation->flags |= NFS_DELEGATION_NEED_RECLAIM; 431 delegation->flags |= NFS_DELEGATION_NEED_RECLAIM;
394 spin_unlock(&clp->cl_lock); 432 rcu_read_unlock();
395} 433}
396 434
397/* 435/*
@@ -399,39 +437,35 @@ void nfs_delegation_mark_reclaim(struct nfs_client *clp)
399 */ 437 */
400void nfs_delegation_reap_unclaimed(struct nfs_client *clp) 438void nfs_delegation_reap_unclaimed(struct nfs_client *clp)
401{ 439{
402 struct nfs_delegation *delegation, *n; 440 struct nfs_delegation *delegation;
403 LIST_HEAD(head); 441restart:
404 spin_lock(&clp->cl_lock); 442 rcu_read_lock();
405 list_for_each_entry_safe(delegation, n, &clp->cl_delegations, super_list) { 443 list_for_each_entry_rcu(delegation, &clp->cl_delegations, super_list) {
406 if ((delegation->flags & NFS_DELEGATION_NEED_RECLAIM) == 0) 444 if ((delegation->flags & NFS_DELEGATION_NEED_RECLAIM) == 0)
407 continue; 445 continue;
408 list_move(&delegation->super_list, &head); 446 spin_lock(&clp->cl_lock);
409 NFS_I(delegation->inode)->delegation = NULL; 447 delegation = nfs_detach_delegation_locked(NFS_I(delegation->inode), NULL);
410 NFS_I(delegation->inode)->delegation_state = 0; 448 spin_unlock(&clp->cl_lock);
411 } 449 rcu_read_unlock();
412 spin_unlock(&clp->cl_lock); 450 if (delegation != NULL)
413 while(!list_empty(&head)) { 451 call_rcu(&delegation->rcu, nfs_free_delegation_callback);
414 delegation = list_entry(head.next, struct nfs_delegation, super_list); 452 goto restart;
415 list_del(&delegation->super_list);
416 nfs_free_delegation(delegation);
417 } 453 }
454 rcu_read_unlock();
418} 455}
419 456
420int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode) 457int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode)
421{ 458{
422 struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
423 struct nfs_inode *nfsi = NFS_I(inode); 459 struct nfs_inode *nfsi = NFS_I(inode);
424 struct nfs_delegation *delegation; 460 struct nfs_delegation *delegation;
425 int res = 0; 461 int ret = 0;
426 462
427 if (nfsi->delegation_state == 0) 463 rcu_read_lock();
428 return 0; 464 delegation = rcu_dereference(nfsi->delegation);
429 spin_lock(&clp->cl_lock);
430 delegation = nfsi->delegation;
431 if (delegation != NULL) { 465 if (delegation != NULL) {
432 memcpy(dst->data, delegation->stateid.data, sizeof(dst->data)); 466 memcpy(dst->data, delegation->stateid.data, sizeof(dst->data));
433 res = 1; 467 ret = 1;
434 } 468 }
435 spin_unlock(&clp->cl_lock); 469 rcu_read_unlock();
436 return res; 470 return ret;
437} 471}
diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h
index 2cfd4b24c7fe..5874ce7fdbae 100644
--- a/fs/nfs/delegation.h
+++ b/fs/nfs/delegation.h
@@ -22,11 +22,12 @@ struct nfs_delegation {
22 long flags; 22 long flags;
23 loff_t maxsize; 23 loff_t maxsize;
24 __u64 change_attr; 24 __u64 change_attr;
25 struct rcu_head rcu;
25}; 26};
26 27
27int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); 28int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
28void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); 29void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res);
29int __nfs_inode_return_delegation(struct inode *inode); 30int nfs_inode_return_delegation(struct inode *inode);
30int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid); 31int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid);
31 32
32struct inode *nfs_delegation_find_inode(struct nfs_client *clp, const struct nfs_fh *fhandle); 33struct inode *nfs_delegation_find_inode(struct nfs_client *clp, const struct nfs_fh *fhandle);
@@ -39,27 +40,24 @@ void nfs_delegation_reap_unclaimed(struct nfs_client *clp);
39 40
40/* NFSv4 delegation-related procedures */ 41/* NFSv4 delegation-related procedures */
41int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid); 42int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid);
42int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state); 43int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid);
43int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl); 44int nfs4_lock_delegation_recall(struct nfs4_state *state, struct file_lock *fl);
44int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode); 45int nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode);
45 46
46static inline int nfs_have_delegation(struct inode *inode, int flags) 47static inline int nfs_have_delegation(struct inode *inode, int flags)
47{ 48{
49 struct nfs_delegation *delegation;
50 int ret = 0;
51
48 flags &= FMODE_READ|FMODE_WRITE; 52 flags &= FMODE_READ|FMODE_WRITE;
49 smp_rmb(); 53 rcu_read_lock();
50 if ((NFS_I(inode)->delegation_state & flags) == flags) 54 delegation = rcu_dereference(NFS_I(inode)->delegation);
51 return 1; 55 if (delegation != NULL && (delegation->type & flags) == flags)
52 return 0; 56 ret = 1;
57 rcu_read_unlock();
58 return ret;
53} 59}
54 60
55static inline int nfs_inode_return_delegation(struct inode *inode)
56{
57 int err = 0;
58
59 if (NFS_I(inode)->delegation != NULL)
60 err = __nfs_inode_return_delegation(inode);
61 return err;
62}
63#else 61#else
64static inline int nfs_have_delegation(struct inode *inode, int flags) 62static inline int nfs_have_delegation(struct inode *inode, int flags)
65{ 63{
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index c27258b5d3e1..322141f4ab48 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -897,14 +897,13 @@ int nfs_is_exclusive_create(struct inode *dir, struct nameidata *nd)
897 return (nd->intent.open.flags & O_EXCL) != 0; 897 return (nd->intent.open.flags & O_EXCL) != 0;
898} 898}
899 899
900static inline int nfs_reval_fsid(struct vfsmount *mnt, struct inode *dir, 900static inline int nfs_reval_fsid(struct inode *dir, const struct nfs_fattr *fattr)
901 struct nfs_fh *fh, struct nfs_fattr *fattr)
902{ 901{
903 struct nfs_server *server = NFS_SERVER(dir); 902 struct nfs_server *server = NFS_SERVER(dir);
904 903
905 if (!nfs_fsid_equal(&server->fsid, &fattr->fsid)) 904 if (!nfs_fsid_equal(&server->fsid, &fattr->fsid))
906 /* Revalidate fsid on root dir */ 905 /* Revalidate fsid using the parent directory */
907 return __nfs_revalidate_inode(server, mnt->mnt_root->d_inode); 906 return __nfs_revalidate_inode(server, dir);
908 return 0; 907 return 0;
909} 908}
910 909
@@ -946,7 +945,7 @@ static struct dentry *nfs_lookup(struct inode *dir, struct dentry * dentry, stru
946 res = ERR_PTR(error); 945 res = ERR_PTR(error);
947 goto out_unlock; 946 goto out_unlock;
948 } 947 }
949 error = nfs_reval_fsid(nd->mnt, dir, &fhandle, &fattr); 948 error = nfs_reval_fsid(dir, &fattr);
950 if (error < 0) { 949 if (error < 0) {
951 res = ERR_PTR(error); 950 res = ERR_PTR(error);
952 goto out_unlock; 951 goto out_unlock;
@@ -1244,7 +1243,7 @@ static int nfs_create(struct inode *dir, struct dentry *dentry, int mode,
1244 attr.ia_mode = mode; 1243 attr.ia_mode = mode;
1245 attr.ia_valid = ATTR_MODE; 1244 attr.ia_valid = ATTR_MODE;
1246 1245
1247 if (nd && (nd->flags & LOOKUP_CREATE)) 1246 if ((nd->flags & LOOKUP_CREATE) != 0)
1248 open_flags = nd->intent.open.flags; 1247 open_flags = nd->intent.open.flags;
1249 1248
1250 lock_kernel(); 1249 lock_kernel();
@@ -1535,7 +1534,7 @@ static int nfs_symlink(struct inode *dir, struct dentry *dentry, const char *sym
1535 1534
1536 lock_kernel(); 1535 lock_kernel();
1537 1536
1538 page = alloc_page(GFP_KERNEL); 1537 page = alloc_page(GFP_HIGHUSER);
1539 if (!page) { 1538 if (!page) {
1540 unlock_kernel(); 1539 unlock_kernel();
1541 return -ENOMEM; 1540 return -ENOMEM;
@@ -1744,8 +1743,8 @@ int nfs_access_cache_shrinker(int nr_to_scan, gfp_t gfp_mask)
1744 struct nfs_inode *nfsi; 1743 struct nfs_inode *nfsi;
1745 struct nfs_access_entry *cache; 1744 struct nfs_access_entry *cache;
1746 1745
1747 spin_lock(&nfs_access_lru_lock);
1748restart: 1746restart:
1747 spin_lock(&nfs_access_lru_lock);
1749 list_for_each_entry(nfsi, &nfs_access_lru_list, access_cache_inode_lru) { 1748 list_for_each_entry(nfsi, &nfs_access_lru_list, access_cache_inode_lru) {
1750 struct inode *inode; 1749 struct inode *inode;
1751 1750
@@ -1770,6 +1769,7 @@ remove_lru_entry:
1770 clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags); 1769 clear_bit(NFS_INO_ACL_LRU_SET, &nfsi->flags);
1771 } 1770 }
1772 spin_unlock(&inode->i_lock); 1771 spin_unlock(&inode->i_lock);
1772 spin_unlock(&nfs_access_lru_lock);
1773 iput(inode); 1773 iput(inode);
1774 goto restart; 1774 goto restart;
1775 } 1775 }
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 00eee87510fe..a5c82b6f3b45 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -266,7 +266,7 @@ static const struct rpc_call_ops nfs_read_direct_ops = {
266static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos) 266static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos)
267{ 267{
268 struct nfs_open_context *ctx = dreq->ctx; 268 struct nfs_open_context *ctx = dreq->ctx;
269 struct inode *inode = ctx->dentry->d_inode; 269 struct inode *inode = ctx->path.dentry->d_inode;
270 size_t rsize = NFS_SERVER(inode)->rsize; 270 size_t rsize = NFS_SERVER(inode)->rsize;
271 unsigned int pgbase; 271 unsigned int pgbase;
272 int result; 272 int result;
@@ -295,9 +295,14 @@ static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned lo
295 break; 295 break;
296 } 296 }
297 if ((unsigned)result < data->npages) { 297 if ((unsigned)result < data->npages) {
298 nfs_direct_release_pages(data->pagevec, result); 298 bytes = result * PAGE_SIZE;
299 nfs_readdata_release(data); 299 if (bytes <= pgbase) {
300 break; 300 nfs_direct_release_pages(data->pagevec, result);
301 nfs_readdata_release(data);
302 break;
303 }
304 bytes -= pgbase;
305 data->npages = result;
301 } 306 }
302 307
303 get_dreq(dreq); 308 get_dreq(dreq);
@@ -601,7 +606,7 @@ static const struct rpc_call_ops nfs_write_direct_ops = {
601static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos, int sync) 606static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos, int sync)
602{ 607{
603 struct nfs_open_context *ctx = dreq->ctx; 608 struct nfs_open_context *ctx = dreq->ctx;
604 struct inode *inode = ctx->dentry->d_inode; 609 struct inode *inode = ctx->path.dentry->d_inode;
605 size_t wsize = NFS_SERVER(inode)->wsize; 610 size_t wsize = NFS_SERVER(inode)->wsize;
606 unsigned int pgbase; 611 unsigned int pgbase;
607 int result; 612 int result;
@@ -630,9 +635,14 @@ static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned l
630 break; 635 break;
631 } 636 }
632 if ((unsigned)result < data->npages) { 637 if ((unsigned)result < data->npages) {
633 nfs_direct_release_pages(data->pagevec, result); 638 bytes = result * PAGE_SIZE;
634 nfs_writedata_release(data); 639 if (bytes <= pgbase) {
635 break; 640 nfs_direct_release_pages(data->pagevec, result);
641 nfs_writedata_release(data);
642 break;
643 }
644 bytes -= pgbase;
645 data->npages = result;
636 } 646 }
637 647
638 get_dreq(dreq); 648 get_dreq(dreq);
@@ -763,10 +773,8 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
763 (unsigned long) count, (long long) pos); 773 (unsigned long) count, (long long) pos);
764 774
765 if (nr_segs != 1) 775 if (nr_segs != 1)
766 return -EINVAL;
767
768 if (count < 0)
769 goto out; 776 goto out;
777
770 retval = -EFAULT; 778 retval = -EFAULT;
771 if (!access_ok(VERIFY_WRITE, buf, count)) 779 if (!access_ok(VERIFY_WRITE, buf, count))
772 goto out; 780 goto out;
@@ -814,7 +822,7 @@ out:
814ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov, 822ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
815 unsigned long nr_segs, loff_t pos) 823 unsigned long nr_segs, loff_t pos)
816{ 824{
817 ssize_t retval; 825 ssize_t retval = -EINVAL;
818 struct file *file = iocb->ki_filp; 826 struct file *file = iocb->ki_filp;
819 struct address_space *mapping = file->f_mapping; 827 struct address_space *mapping = file->f_mapping;
820 /* XXX: temporary */ 828 /* XXX: temporary */
@@ -827,7 +835,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
827 (unsigned long) count, (long long) pos); 835 (unsigned long) count, (long long) pos);
828 836
829 if (nr_segs != 1) 837 if (nr_segs != 1)
830 return -EINVAL; 838 goto out;
831 839
832 retval = generic_write_checks(file, &pos, &count, 0); 840 retval = generic_write_checks(file, &pos, &count, 0);
833 if (retval) 841 if (retval)
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index bd9f5a836592..3d9fccf4ef93 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -461,14 +461,14 @@ static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, str
461 461
462 ctx = kmalloc(sizeof(*ctx), GFP_KERNEL); 462 ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
463 if (ctx != NULL) { 463 if (ctx != NULL) {
464 atomic_set(&ctx->count, 1); 464 ctx->path.dentry = dget(dentry);
465 ctx->dentry = dget(dentry); 465 ctx->path.mnt = mntget(mnt);
466 ctx->vfsmnt = mntget(mnt);
467 ctx->cred = get_rpccred(cred); 466 ctx->cred = get_rpccred(cred);
468 ctx->state = NULL; 467 ctx->state = NULL;
469 ctx->lockowner = current->files; 468 ctx->lockowner = current->files;
470 ctx->error = 0; 469 ctx->error = 0;
471 ctx->dir_cookie = 0; 470 ctx->dir_cookie = 0;
471 kref_init(&ctx->kref);
472 } 472 }
473 return ctx; 473 return ctx;
474} 474}
@@ -476,27 +476,33 @@ static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, str
476struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx) 476struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
477{ 477{
478 if (ctx != NULL) 478 if (ctx != NULL)
479 atomic_inc(&ctx->count); 479 kref_get(&ctx->kref);
480 return ctx; 480 return ctx;
481} 481}
482 482
483void put_nfs_open_context(struct nfs_open_context *ctx) 483static void nfs_free_open_context(struct kref *kref)
484{ 484{
485 if (atomic_dec_and_test(&ctx->count)) { 485 struct nfs_open_context *ctx = container_of(kref,
486 if (!list_empty(&ctx->list)) { 486 struct nfs_open_context, kref);
487 struct inode *inode = ctx->dentry->d_inode; 487
488 spin_lock(&inode->i_lock); 488 if (!list_empty(&ctx->list)) {
489 list_del(&ctx->list); 489 struct inode *inode = ctx->path.dentry->d_inode;
490 spin_unlock(&inode->i_lock); 490 spin_lock(&inode->i_lock);
491 } 491 list_del(&ctx->list);
492 if (ctx->state != NULL) 492 spin_unlock(&inode->i_lock);
493 nfs4_close_state(ctx->state, ctx->mode);
494 if (ctx->cred != NULL)
495 put_rpccred(ctx->cred);
496 dput(ctx->dentry);
497 mntput(ctx->vfsmnt);
498 kfree(ctx);
499 } 493 }
494 if (ctx->state != NULL)
495 nfs4_close_state(&ctx->path, ctx->state, ctx->mode);
496 if (ctx->cred != NULL)
497 put_rpccred(ctx->cred);
498 dput(ctx->path.dentry);
499 mntput(ctx->path.mnt);
500 kfree(ctx);
501}
502
503void put_nfs_open_context(struct nfs_open_context *ctx)
504{
505 kref_put(&ctx->kref, nfs_free_open_context);
500} 506}
501 507
502/* 508/*
@@ -961,8 +967,8 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
961 goto out_changed; 967 goto out_changed;
962 968
963 server = NFS_SERVER(inode); 969 server = NFS_SERVER(inode);
964 /* Update the fsid if and only if this is the root directory */ 970 /* Update the fsid? */
965 if (inode == inode->i_sb->s_root->d_inode 971 if (S_ISDIR(inode->i_mode)
966 && !nfs_fsid_equal(&server->fsid, &fattr->fsid)) 972 && !nfs_fsid_equal(&server->fsid, &fattr->fsid))
967 server->fsid = fattr->fsid; 973 server->fsid = fattr->fsid;
968 974
@@ -1066,8 +1072,10 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1066 invalid &= ~NFS_INO_INVALID_DATA; 1072 invalid &= ~NFS_INO_INVALID_DATA;
1067 if (data_stable) 1073 if (data_stable)
1068 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE); 1074 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1069 if (!nfs_have_delegation(inode, FMODE_READ)) 1075 if (!nfs_have_delegation(inode, FMODE_READ) ||
1076 (nfsi->cache_validity & NFS_INO_REVAL_FORCED))
1070 nfsi->cache_validity |= invalid; 1077 nfsi->cache_validity |= invalid;
1078 nfsi->cache_validity &= ~NFS_INO_REVAL_FORCED;
1071 1079
1072 return 0; 1080 return 0;
1073 out_changed: 1081 out_changed:
@@ -1103,27 +1111,10 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1103 */ 1111 */
1104void nfs4_clear_inode(struct inode *inode) 1112void nfs4_clear_inode(struct inode *inode)
1105{ 1113{
1106 struct nfs_inode *nfsi = NFS_I(inode);
1107
1108 /* If we are holding a delegation, return it! */ 1114 /* If we are holding a delegation, return it! */
1109 nfs_inode_return_delegation(inode); 1115 nfs_inode_return_delegation(inode);
1110 /* First call standard NFS clear_inode() code */ 1116 /* First call standard NFS clear_inode() code */
1111 nfs_clear_inode(inode); 1117 nfs_clear_inode(inode);
1112 /* Now clear out any remaining state */
1113 while (!list_empty(&nfsi->open_states)) {
1114 struct nfs4_state *state;
1115
1116 state = list_entry(nfsi->open_states.next,
1117 struct nfs4_state,
1118 inode_states);
1119 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1120 __FUNCTION__,
1121 inode->i_sb->s_id,
1122 (long long)NFS_FILEID(inode),
1123 state);
1124 BUG_ON(atomic_read(&state->count) != 1);
1125 nfs4_close_state(state, state->state);
1126 }
1127} 1118}
1128#endif 1119#endif
1129 1120
@@ -1165,15 +1156,11 @@ static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flag
1165 struct nfs_inode *nfsi = (struct nfs_inode *) foo; 1156 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1166 1157
1167 inode_init_once(&nfsi->vfs_inode); 1158 inode_init_once(&nfsi->vfs_inode);
1168 spin_lock_init(&nfsi->req_lock);
1169 INIT_LIST_HEAD(&nfsi->dirty);
1170 INIT_LIST_HEAD(&nfsi->commit);
1171 INIT_LIST_HEAD(&nfsi->open_files); 1159 INIT_LIST_HEAD(&nfsi->open_files);
1172 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru); 1160 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
1173 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru); 1161 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
1174 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC); 1162 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1175 atomic_set(&nfsi->data_updates, 0); 1163 atomic_set(&nfsi->data_updates, 0);
1176 nfsi->ndirty = 0;
1177 nfsi->ncommit = 0; 1164 nfsi->ncommit = 0;
1178 nfsi->npages = 0; 1165 nfsi->npages = 0;
1179 nfs4_init_once(nfsi); 1166 nfs4_init_once(nfsi);
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index ad2b40db1e65..76cf55d57101 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -183,9 +183,9 @@ unsigned long nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
183/* 183/*
184 * Calculate the number of 512byte blocks used. 184 * Calculate the number of 512byte blocks used.
185 */ 185 */
186static inline unsigned long nfs_calc_block_size(u64 tsize) 186static inline blkcnt_t nfs_calc_block_size(u64 tsize)
187{ 187{
188 loff_t used = (tsize + 511) >> 9; 188 blkcnt_t used = (tsize + 511) >> 9;
189 return (used > ULONG_MAX) ? ULONG_MAX : used; 189 return (used > ULONG_MAX) ? ULONG_MAX : used;
190} 190}
191 191
diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c
index ca5a266a3140..8afd9f7e7a97 100644
--- a/fs/nfs/mount_clnt.c
+++ b/fs/nfs/mount_clnt.c
@@ -1,7 +1,5 @@
1/* 1/*
2 * linux/fs/nfs/mount_clnt.c 2 * In-kernel MOUNT protocol client
3 *
4 * MOUNT client to support NFSroot.
5 * 3 *
6 * Copyright (C) 1997, Olaf Kirch <okir@monad.swb.de> 4 * Copyright (C) 1997, Olaf Kirch <okir@monad.swb.de>
7 */ 5 */
@@ -18,33 +16,31 @@
18#include <linux/nfs_fs.h> 16#include <linux/nfs_fs.h>
19 17
20#ifdef RPC_DEBUG 18#ifdef RPC_DEBUG
21# define NFSDBG_FACILITY NFSDBG_ROOT 19# define NFSDBG_FACILITY NFSDBG_MOUNT
22#endif 20#endif
23 21
24/*
25#define MOUNT_PROGRAM 100005
26#define MOUNT_VERSION 1
27#define MOUNT_MNT 1
28#define MOUNT_UMNT 3
29 */
30
31static struct rpc_clnt * mnt_create(char *, struct sockaddr_in *,
32 int, int);
33static struct rpc_program mnt_program; 22static struct rpc_program mnt_program;
34 23
35struct mnt_fhstatus { 24struct mnt_fhstatus {
36 unsigned int status; 25 u32 status;
37 struct nfs_fh * fh; 26 struct nfs_fh *fh;
38}; 27};
39 28
40/* 29/**
41 * Obtain an NFS file handle for the given host and path 30 * nfs_mount - Obtain an NFS file handle for the given host and path
31 * @addr: pointer to server's address
32 * @len: size of server's address
33 * @hostname: name of server host, or NULL
34 * @path: pointer to string containing export path to mount
35 * @version: mount version to use for this request
36 * @protocol: transport protocol to use for thie request
37 * @fh: pointer to location to place returned file handle
38 *
39 * Uses default timeout parameters specified by underlying transport.
42 */ 40 */
43int 41int nfs_mount(struct sockaddr *addr, size_t len, char *hostname, char *path,
44nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh, 42 int version, int protocol, struct nfs_fh *fh)
45 int version, int protocol)
46{ 43{
47 struct rpc_clnt *mnt_clnt;
48 struct mnt_fhstatus result = { 44 struct mnt_fhstatus result = {
49 .fh = fh 45 .fh = fh
50 }; 46 };
@@ -52,16 +48,25 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh,
52 .rpc_argp = path, 48 .rpc_argp = path,
53 .rpc_resp = &result, 49 .rpc_resp = &result,
54 }; 50 };
55 char hostname[32]; 51 struct rpc_create_args args = {
52 .protocol = protocol,
53 .address = addr,
54 .addrsize = len,
55 .servername = hostname,
56 .program = &mnt_program,
57 .version = version,
58 .authflavor = RPC_AUTH_UNIX,
59 .flags = RPC_CLNT_CREATE_INTR,
60 };
61 struct rpc_clnt *mnt_clnt;
56 int status; 62 int status;
57 63
58 dprintk("NFS: nfs_mount(%08x:%s)\n", 64 dprintk("NFS: sending MNT request for %s:%s\n",
59 (unsigned)ntohl(addr->sin_addr.s_addr), path); 65 (hostname ? hostname : "server"), path);
60 66
61 sprintf(hostname, "%u.%u.%u.%u", NIPQUAD(addr->sin_addr.s_addr)); 67 mnt_clnt = rpc_create(&args);
62 mnt_clnt = mnt_create(hostname, addr, version, protocol);
63 if (IS_ERR(mnt_clnt)) 68 if (IS_ERR(mnt_clnt))
64 return PTR_ERR(mnt_clnt); 69 goto out_clnt_err;
65 70
66 if (version == NFS_MNT3_VERSION) 71 if (version == NFS_MNT3_VERSION)
67 msg.rpc_proc = &mnt_clnt->cl_procinfo[MOUNTPROC3_MNT]; 72 msg.rpc_proc = &mnt_clnt->cl_procinfo[MOUNTPROC3_MNT];
@@ -69,33 +74,39 @@ nfsroot_mount(struct sockaddr_in *addr, char *path, struct nfs_fh *fh,
69 msg.rpc_proc = &mnt_clnt->cl_procinfo[MNTPROC_MNT]; 74 msg.rpc_proc = &mnt_clnt->cl_procinfo[MNTPROC_MNT];
70 75
71 status = rpc_call_sync(mnt_clnt, &msg, 0); 76 status = rpc_call_sync(mnt_clnt, &msg, 0);
72 return status < 0? status : (result.status? -EACCES : 0); 77 rpc_shutdown_client(mnt_clnt);
73}
74 78
75static struct rpc_clnt * 79 if (status < 0)
76mnt_create(char *hostname, struct sockaddr_in *srvaddr, int version, 80 goto out_call_err;
77 int protocol) 81 if (result.status != 0)
78{ 82 goto out_mnt_err;
79 struct rpc_create_args args = { 83
80 .protocol = protocol, 84 dprintk("NFS: MNT request succeeded\n");
81 .address = (struct sockaddr *)srvaddr, 85 status = 0;
82 .addrsize = sizeof(*srvaddr), 86
83 .servername = hostname, 87out:
84 .program = &mnt_program, 88 return status;
85 .version = version, 89
86 .authflavor = RPC_AUTH_UNIX, 90out_clnt_err:
87 .flags = (RPC_CLNT_CREATE_ONESHOT | 91 status = PTR_ERR(mnt_clnt);
88 RPC_CLNT_CREATE_INTR), 92 dprintk("NFS: failed to create RPC client, status=%d\n", status);
89 }; 93 goto out;
94
95out_call_err:
96 dprintk("NFS: failed to start MNT request, status=%d\n", status);
97 goto out;
90 98
91 return rpc_create(&args); 99out_mnt_err:
100 dprintk("NFS: MNT server returned result %d\n", result.status);
101 status = -EACCES;
102 goto out;
92} 103}
93 104
94/* 105/*
95 * XDR encode/decode functions for MOUNT 106 * XDR encode/decode functions for MOUNT
96 */ 107 */
97static int 108static int xdr_encode_dirpath(struct rpc_rqst *req, __be32 *p,
98xdr_encode_dirpath(struct rpc_rqst *req, __be32 *p, const char *path) 109 const char *path)
99{ 110{
100 p = xdr_encode_string(p, path); 111 p = xdr_encode_string(p, path);
101 112
@@ -103,8 +114,8 @@ xdr_encode_dirpath(struct rpc_rqst *req, __be32 *p, const char *path)
103 return 0; 114 return 0;
104} 115}
105 116
106static int 117static int xdr_decode_fhstatus(struct rpc_rqst *req, __be32 *p,
107xdr_decode_fhstatus(struct rpc_rqst *req, __be32 *p, struct mnt_fhstatus *res) 118 struct mnt_fhstatus *res)
108{ 119{
109 struct nfs_fh *fh = res->fh; 120 struct nfs_fh *fh = res->fh;
110 121
@@ -115,8 +126,8 @@ xdr_decode_fhstatus(struct rpc_rqst *req, __be32 *p, struct mnt_fhstatus *res)
115 return 0; 126 return 0;
116} 127}
117 128
118static int 129static int xdr_decode_fhstatus3(struct rpc_rqst *req, __be32 *p,
119xdr_decode_fhstatus3(struct rpc_rqst *req, __be32 *p, struct mnt_fhstatus *res) 130 struct mnt_fhstatus *res)
120{ 131{
121 struct nfs_fh *fh = res->fh; 132 struct nfs_fh *fh = res->fh;
122 133
@@ -135,53 +146,53 @@ xdr_decode_fhstatus3(struct rpc_rqst *req, __be32 *p, struct mnt_fhstatus *res)
135#define MNT_fhstatus_sz (1 + 8) 146#define MNT_fhstatus_sz (1 + 8)
136#define MNT_fhstatus3_sz (1 + 16) 147#define MNT_fhstatus3_sz (1 + 16)
137 148
138static struct rpc_procinfo mnt_procedures[] = { 149static struct rpc_procinfo mnt_procedures[] = {
139[MNTPROC_MNT] = { 150 [MNTPROC_MNT] = {
140 .p_proc = MNTPROC_MNT, 151 .p_proc = MNTPROC_MNT,
141 .p_encode = (kxdrproc_t) xdr_encode_dirpath, 152 .p_encode = (kxdrproc_t) xdr_encode_dirpath,
142 .p_decode = (kxdrproc_t) xdr_decode_fhstatus, 153 .p_decode = (kxdrproc_t) xdr_decode_fhstatus,
143 .p_arglen = MNT_dirpath_sz, 154 .p_arglen = MNT_dirpath_sz,
144 .p_replen = MNT_fhstatus_sz, 155 .p_replen = MNT_fhstatus_sz,
145 .p_statidx = MNTPROC_MNT, 156 .p_statidx = MNTPROC_MNT,
146 .p_name = "MOUNT", 157 .p_name = "MOUNT",
147 }, 158 },
148}; 159};
149 160
150static struct rpc_procinfo mnt3_procedures[] = { 161static struct rpc_procinfo mnt3_procedures[] = {
151[MOUNTPROC3_MNT] = { 162 [MOUNTPROC3_MNT] = {
152 .p_proc = MOUNTPROC3_MNT, 163 .p_proc = MOUNTPROC3_MNT,
153 .p_encode = (kxdrproc_t) xdr_encode_dirpath, 164 .p_encode = (kxdrproc_t) xdr_encode_dirpath,
154 .p_decode = (kxdrproc_t) xdr_decode_fhstatus3, 165 .p_decode = (kxdrproc_t) xdr_decode_fhstatus3,
155 .p_arglen = MNT_dirpath_sz, 166 .p_arglen = MNT_dirpath_sz,
156 .p_replen = MNT_fhstatus3_sz, 167 .p_replen = MNT_fhstatus3_sz,
157 .p_statidx = MOUNTPROC3_MNT, 168 .p_statidx = MOUNTPROC3_MNT,
158 .p_name = "MOUNT", 169 .p_name = "MOUNT",
159 }, 170 },
160}; 171};
161 172
162 173
163static struct rpc_version mnt_version1 = { 174static struct rpc_version mnt_version1 = {
164 .number = 1, 175 .number = 1,
165 .nrprocs = 2, 176 .nrprocs = 2,
166 .procs = mnt_procedures 177 .procs = mnt_procedures,
167}; 178};
168 179
169static struct rpc_version mnt_version3 = { 180static struct rpc_version mnt_version3 = {
170 .number = 3, 181 .number = 3,
171 .nrprocs = 2, 182 .nrprocs = 2,
172 .procs = mnt3_procedures 183 .procs = mnt3_procedures,
173}; 184};
174 185
175static struct rpc_version * mnt_version[] = { 186static struct rpc_version *mnt_version[] = {
176 NULL, 187 NULL,
177 &mnt_version1, 188 &mnt_version1,
178 NULL, 189 NULL,
179 &mnt_version3, 190 &mnt_version3,
180}; 191};
181 192
182static struct rpc_stat mnt_stats; 193static struct rpc_stat mnt_stats;
183 194
184static struct rpc_program mnt_program = { 195static struct rpc_program mnt_program = {
185 .name = "mount", 196 .name = "mount",
186 .number = NFS_MNT_PROGRAM, 197 .number = NFS_MNT_PROGRAM,
187 .nrvers = ARRAY_SIZE(mnt_version), 198 .nrvers = ARRAY_SIZE(mnt_version),
diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c
index cd3ca7b5d3db..7fcc78f2aa71 100644
--- a/fs/nfs/nfs2xdr.c
+++ b/fs/nfs/nfs2xdr.c
@@ -223,7 +223,7 @@ nfs_xdr_diropargs(struct rpc_rqst *req, __be32 *p, struct nfs_diropargs *args)
223static int 223static int
224nfs_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) 224nfs_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args)
225{ 225{
226 struct rpc_auth *auth = req->rq_task->tk_auth; 226 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
227 unsigned int replen; 227 unsigned int replen;
228 u32 offset = (u32)args->offset; 228 u32 offset = (u32)args->offset;
229 u32 count = args->count; 229 u32 count = args->count;
@@ -380,7 +380,7 @@ static int
380nfs_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs_readdirargs *args) 380nfs_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs_readdirargs *args)
381{ 381{
382 struct rpc_task *task = req->rq_task; 382 struct rpc_task *task = req->rq_task;
383 struct rpc_auth *auth = task->tk_auth; 383 struct rpc_auth *auth = task->tk_msg.rpc_cred->cr_auth;
384 unsigned int replen; 384 unsigned int replen;
385 u32 count = args->count; 385 u32 count = args->count;
386 386
@@ -541,7 +541,7 @@ nfs_xdr_diropres(struct rpc_rqst *req, __be32 *p, struct nfs_diropok *res)
541static int 541static int
542nfs_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs_readlinkargs *args) 542nfs_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs_readlinkargs *args)
543{ 543{
544 struct rpc_auth *auth = req->rq_task->tk_auth; 544 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
545 unsigned int replen; 545 unsigned int replen;
546 546
547 p = xdr_encode_fhandle(p, args->fh); 547 p = xdr_encode_fhandle(p, args->fh);
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index 45268d6def2e..814d886b6aa4 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -335,9 +335,7 @@ again:
335 * not sure this buys us anything (and I'd have 335 * not sure this buys us anything (and I'd have
336 * to revamp the NFSv3 XDR code) */ 336 * to revamp the NFSv3 XDR code) */
337 status = nfs3_proc_setattr(dentry, &fattr, sattr); 337 status = nfs3_proc_setattr(dentry, &fattr, sattr);
338 if (status == 0) 338 nfs_post_op_update_inode(dentry->d_inode, &fattr);
339 nfs_setattr_update_inode(dentry->d_inode, sattr);
340 nfs_refresh_inode(dentry->d_inode, &fattr);
341 dprintk("NFS reply setattr (post-create): %d\n", status); 339 dprintk("NFS reply setattr (post-create): %d\n", status);
342 } 340 }
343 if (status != 0) 341 if (status != 0)
diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c
index b51df8eb9f01..b4647a22f349 100644
--- a/fs/nfs/nfs3xdr.c
+++ b/fs/nfs/nfs3xdr.c
@@ -319,7 +319,7 @@ nfs3_xdr_accessargs(struct rpc_rqst *req, __be32 *p, struct nfs3_accessargs *arg
319static int 319static int
320nfs3_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) 320nfs3_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args)
321{ 321{
322 struct rpc_auth *auth = req->rq_task->tk_auth; 322 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
323 unsigned int replen; 323 unsigned int replen;
324 u32 count = args->count; 324 u32 count = args->count;
325 325
@@ -458,7 +458,7 @@ nfs3_xdr_linkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_linkargs *args)
458static int 458static int
459nfs3_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readdirargs *args) 459nfs3_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readdirargs *args)
460{ 460{
461 struct rpc_auth *auth = req->rq_task->tk_auth; 461 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
462 unsigned int replen; 462 unsigned int replen;
463 u32 count = args->count; 463 u32 count = args->count;
464 464
@@ -643,7 +643,7 @@ static int
643nfs3_xdr_getaclargs(struct rpc_rqst *req, __be32 *p, 643nfs3_xdr_getaclargs(struct rpc_rqst *req, __be32 *p,
644 struct nfs3_getaclargs *args) 644 struct nfs3_getaclargs *args)
645{ 645{
646 struct rpc_auth *auth = req->rq_task->tk_auth; 646 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
647 unsigned int replen; 647 unsigned int replen;
648 648
649 p = xdr_encode_fhandle(p, args->fh); 649 p = xdr_encode_fhandle(p, args->fh);
@@ -773,7 +773,7 @@ nfs3_xdr_accessres(struct rpc_rqst *req, __be32 *p, struct nfs3_accessres *res)
773static int 773static int
774nfs3_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readlinkargs *args) 774nfs3_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readlinkargs *args)
775{ 775{
776 struct rpc_auth *auth = req->rq_task->tk_auth; 776 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
777 unsigned int replen; 777 unsigned int replen;
778 778
779 p = xdr_encode_fhandle(p, args->fh); 779 p = xdr_encode_fhandle(p, args->fh);
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index cf3a17eb5c09..6c028e734fe6 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -70,19 +70,26 @@ static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status
70 seqid->flags |= NFS_SEQID_CONFIRMED; 70 seqid->flags |= NFS_SEQID_CONFIRMED;
71} 71}
72 72
73struct nfs_unique_id {
74 struct rb_node rb_node;
75 __u64 id;
76};
77
73/* 78/*
74 * NFS4 state_owners and lock_owners are simply labels for ordered 79 * NFS4 state_owners and lock_owners are simply labels for ordered
75 * sequences of RPC calls. Their sole purpose is to provide once-only 80 * sequences of RPC calls. Their sole purpose is to provide once-only
76 * semantics by allowing the server to identify replayed requests. 81 * semantics by allowing the server to identify replayed requests.
77 */ 82 */
78struct nfs4_state_owner { 83struct nfs4_state_owner {
79 spinlock_t so_lock; 84 struct nfs_unique_id so_owner_id;
80 struct list_head so_list; /* per-clientid list of state_owners */
81 struct nfs_client *so_client; 85 struct nfs_client *so_client;
82 u32 so_id; /* 32-bit identifier, unique */ 86 struct nfs_server *so_server;
83 atomic_t so_count; 87 struct rb_node so_client_node;
84 88
85 struct rpc_cred *so_cred; /* Associated cred */ 89 struct rpc_cred *so_cred; /* Associated cred */
90
91 spinlock_t so_lock;
92 atomic_t so_count;
86 struct list_head so_states; 93 struct list_head so_states;
87 struct list_head so_delegations; 94 struct list_head so_delegations;
88 struct nfs_seqid_counter so_seqid; 95 struct nfs_seqid_counter so_seqid;
@@ -108,7 +115,7 @@ struct nfs4_lock_state {
108#define NFS_LOCK_INITIALIZED 1 115#define NFS_LOCK_INITIALIZED 1
109 int ls_flags; 116 int ls_flags;
110 struct nfs_seqid_counter ls_seqid; 117 struct nfs_seqid_counter ls_seqid;
111 u32 ls_id; 118 struct nfs_unique_id ls_id;
112 nfs4_stateid ls_stateid; 119 nfs4_stateid ls_stateid;
113 atomic_t ls_count; 120 atomic_t ls_count;
114}; 121};
@@ -116,7 +123,10 @@ struct nfs4_lock_state {
116/* bits for nfs4_state->flags */ 123/* bits for nfs4_state->flags */
117enum { 124enum {
118 LK_STATE_IN_USE, 125 LK_STATE_IN_USE,
119 NFS_DELEGATED_STATE, 126 NFS_DELEGATED_STATE, /* Current stateid is delegation */
127 NFS_O_RDONLY_STATE, /* OPEN stateid has read-only state */
128 NFS_O_WRONLY_STATE, /* OPEN stateid has write-only state */
129 NFS_O_RDWR_STATE, /* OPEN stateid has read/write state */
120}; 130};
121 131
122struct nfs4_state { 132struct nfs4_state {
@@ -130,11 +140,14 @@ struct nfs4_state {
130 unsigned long flags; /* Do we hold any locks? */ 140 unsigned long flags; /* Do we hold any locks? */
131 spinlock_t state_lock; /* Protects the lock_states list */ 141 spinlock_t state_lock; /* Protects the lock_states list */
132 142
133 nfs4_stateid stateid; 143 seqlock_t seqlock; /* Protects the stateid/open_stateid */
144 nfs4_stateid stateid; /* Current stateid: may be delegation */
145 nfs4_stateid open_stateid; /* OPEN stateid */
134 146
135 unsigned int n_rdonly; 147 /* The following 3 fields are protected by owner->so_lock */
136 unsigned int n_wronly; 148 unsigned int n_rdonly; /* Number of read-only references */
137 unsigned int n_rdwr; 149 unsigned int n_wronly; /* Number of write-only references */
150 unsigned int n_rdwr; /* Number of read/write references */
138 int state; /* State on the server (R,W, or RW) */ 151 int state; /* State on the server (R,W, or RW) */
139 atomic_t count; 152 atomic_t count;
140}; 153};
@@ -165,7 +178,7 @@ extern int nfs4_proc_setclientid(struct nfs_client *, u32, unsigned short, struc
165extern int nfs4_proc_setclientid_confirm(struct nfs_client *, struct rpc_cred *); 178extern int nfs4_proc_setclientid_confirm(struct nfs_client *, struct rpc_cred *);
166extern int nfs4_proc_async_renew(struct nfs_client *, struct rpc_cred *); 179extern int nfs4_proc_async_renew(struct nfs_client *, struct rpc_cred *);
167extern int nfs4_proc_renew(struct nfs_client *, struct rpc_cred *); 180extern int nfs4_proc_renew(struct nfs_client *, struct rpc_cred *);
168extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state); 181extern int nfs4_do_close(struct path *path, struct nfs4_state *state);
169extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); 182extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *);
170extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *); 183extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *);
171extern int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle); 184extern int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle);
@@ -189,14 +202,13 @@ extern void nfs4_renew_state(struct work_struct *);
189 202
190/* nfs4state.c */ 203/* nfs4state.c */
191struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp); 204struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp);
192extern u32 nfs4_alloc_lockowner_id(struct nfs_client *);
193 205
194extern struct nfs4_state_owner * nfs4_get_state_owner(struct nfs_server *, struct rpc_cred *); 206extern struct nfs4_state_owner * nfs4_get_state_owner(struct nfs_server *, struct rpc_cred *);
195extern void nfs4_put_state_owner(struct nfs4_state_owner *); 207extern void nfs4_put_state_owner(struct nfs4_state_owner *);
196extern void nfs4_drop_state_owner(struct nfs4_state_owner *); 208extern void nfs4_drop_state_owner(struct nfs4_state_owner *);
197extern struct nfs4_state * nfs4_get_open_state(struct inode *, struct nfs4_state_owner *); 209extern struct nfs4_state * nfs4_get_open_state(struct inode *, struct nfs4_state_owner *);
198extern void nfs4_put_open_state(struct nfs4_state *); 210extern void nfs4_put_open_state(struct nfs4_state *);
199extern void nfs4_close_state(struct nfs4_state *, mode_t); 211extern void nfs4_close_state(struct path *, struct nfs4_state *, mode_t);
200extern void nfs4_state_set_mode_locked(struct nfs4_state *, mode_t); 212extern void nfs4_state_set_mode_locked(struct nfs4_state *, mode_t);
201extern void nfs4_schedule_state_recovery(struct nfs_client *); 213extern void nfs4_schedule_state_recovery(struct nfs_client *);
202extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp); 214extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp);
@@ -222,7 +234,7 @@ extern struct svc_version nfs4_callback_version1;
222 234
223#else 235#else
224 236
225#define nfs4_close_state(a, b) do { } while (0) 237#define nfs4_close_state(a, b, c) do { } while (0)
226 238
227#endif /* CONFIG_NFS_V4 */ 239#endif /* CONFIG_NFS_V4 */
228#endif /* __LINUX_FS_NFS_NFS4_FS.H */ 240#endif /* __LINUX_FS_NFS_NFS4_FS.H */
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 648e0ac0f90e..fee2da856c95 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -65,6 +65,7 @@ static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *)
65static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); 65static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
66static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception); 66static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
67static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs_client *clp); 67static int nfs4_wait_clnt_recover(struct rpc_clnt *clnt, struct nfs_client *clp);
68static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags);
68 69
69/* Prevent leaks of NFSv4 errors into userland */ 70/* Prevent leaks of NFSv4 errors into userland */
70int nfs4_map_errors(int err) 71int nfs4_map_errors(int err)
@@ -214,27 +215,39 @@ static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
214} 215}
215 216
216struct nfs4_opendata { 217struct nfs4_opendata {
217 atomic_t count; 218 struct kref kref;
218 struct nfs_openargs o_arg; 219 struct nfs_openargs o_arg;
219 struct nfs_openres o_res; 220 struct nfs_openres o_res;
220 struct nfs_open_confirmargs c_arg; 221 struct nfs_open_confirmargs c_arg;
221 struct nfs_open_confirmres c_res; 222 struct nfs_open_confirmres c_res;
222 struct nfs_fattr f_attr; 223 struct nfs_fattr f_attr;
223 struct nfs_fattr dir_attr; 224 struct nfs_fattr dir_attr;
224 struct dentry *dentry; 225 struct path path;
225 struct dentry *dir; 226 struct dentry *dir;
226 struct nfs4_state_owner *owner; 227 struct nfs4_state_owner *owner;
228 struct nfs4_state *state;
227 struct iattr attrs; 229 struct iattr attrs;
228 unsigned long timestamp; 230 unsigned long timestamp;
231 unsigned int rpc_done : 1;
229 int rpc_status; 232 int rpc_status;
230 int cancelled; 233 int cancelled;
231}; 234};
232 235
233static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry, 236
237static void nfs4_init_opendata_res(struct nfs4_opendata *p)
238{
239 p->o_res.f_attr = &p->f_attr;
240 p->o_res.dir_attr = &p->dir_attr;
241 p->o_res.server = p->o_arg.server;
242 nfs_fattr_init(&p->f_attr);
243 nfs_fattr_init(&p->dir_attr);
244}
245
246static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path,
234 struct nfs4_state_owner *sp, int flags, 247 struct nfs4_state_owner *sp, int flags,
235 const struct iattr *attrs) 248 const struct iattr *attrs)
236{ 249{
237 struct dentry *parent = dget_parent(dentry); 250 struct dentry *parent = dget_parent(path->dentry);
238 struct inode *dir = parent->d_inode; 251 struct inode *dir = parent->d_inode;
239 struct nfs_server *server = NFS_SERVER(dir); 252 struct nfs_server *server = NFS_SERVER(dir);
240 struct nfs4_opendata *p; 253 struct nfs4_opendata *p;
@@ -245,24 +258,19 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
245 p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); 258 p->o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
246 if (p->o_arg.seqid == NULL) 259 if (p->o_arg.seqid == NULL)
247 goto err_free; 260 goto err_free;
248 atomic_set(&p->count, 1); 261 p->path.mnt = mntget(path->mnt);
249 p->dentry = dget(dentry); 262 p->path.dentry = dget(path->dentry);
250 p->dir = parent; 263 p->dir = parent;
251 p->owner = sp; 264 p->owner = sp;
252 atomic_inc(&sp->so_count); 265 atomic_inc(&sp->so_count);
253 p->o_arg.fh = NFS_FH(dir); 266 p->o_arg.fh = NFS_FH(dir);
254 p->o_arg.open_flags = flags, 267 p->o_arg.open_flags = flags,
255 p->o_arg.clientid = server->nfs_client->cl_clientid; 268 p->o_arg.clientid = server->nfs_client->cl_clientid;
256 p->o_arg.id = sp->so_id; 269 p->o_arg.id = sp->so_owner_id.id;
257 p->o_arg.name = &dentry->d_name; 270 p->o_arg.name = &p->path.dentry->d_name;
258 p->o_arg.server = server; 271 p->o_arg.server = server;
259 p->o_arg.bitmask = server->attr_bitmask; 272 p->o_arg.bitmask = server->attr_bitmask;
260 p->o_arg.claim = NFS4_OPEN_CLAIM_NULL; 273 p->o_arg.claim = NFS4_OPEN_CLAIM_NULL;
261 p->o_res.f_attr = &p->f_attr;
262 p->o_res.dir_attr = &p->dir_attr;
263 p->o_res.server = server;
264 nfs_fattr_init(&p->f_attr);
265 nfs_fattr_init(&p->dir_attr);
266 if (flags & O_EXCL) { 274 if (flags & O_EXCL) {
267 u32 *s = (u32 *) p->o_arg.u.verifier.data; 275 u32 *s = (u32 *) p->o_arg.u.verifier.data;
268 s[0] = jiffies; 276 s[0] = jiffies;
@@ -274,6 +282,8 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
274 p->c_arg.fh = &p->o_res.fh; 282 p->c_arg.fh = &p->o_res.fh;
275 p->c_arg.stateid = &p->o_res.stateid; 283 p->c_arg.stateid = &p->o_res.stateid;
276 p->c_arg.seqid = p->o_arg.seqid; 284 p->c_arg.seqid = p->o_arg.seqid;
285 nfs4_init_opendata_res(p);
286 kref_init(&p->kref);
277 return p; 287 return p;
278err_free: 288err_free:
279 kfree(p); 289 kfree(p);
@@ -282,27 +292,25 @@ err:
282 return NULL; 292 return NULL;
283} 293}
284 294
285static void nfs4_opendata_free(struct nfs4_opendata *p) 295static void nfs4_opendata_free(struct kref *kref)
286{ 296{
287 if (p != NULL && atomic_dec_and_test(&p->count)) { 297 struct nfs4_opendata *p = container_of(kref,
288 nfs_free_seqid(p->o_arg.seqid); 298 struct nfs4_opendata, kref);
289 nfs4_put_state_owner(p->owner); 299
290 dput(p->dir); 300 nfs_free_seqid(p->o_arg.seqid);
291 dput(p->dentry); 301 if (p->state != NULL)
292 kfree(p); 302 nfs4_put_open_state(p->state);
293 } 303 nfs4_put_state_owner(p->owner);
304 dput(p->dir);
305 dput(p->path.dentry);
306 mntput(p->path.mnt);
307 kfree(p);
294} 308}
295 309
296/* Helper for asynchronous RPC calls */ 310static void nfs4_opendata_put(struct nfs4_opendata *p)
297static int nfs4_call_async(struct rpc_clnt *clnt,
298 const struct rpc_call_ops *tk_ops, void *calldata)
299{ 311{
300 struct rpc_task *task; 312 if (p != NULL)
301 313 kref_put(&p->kref, nfs4_opendata_free);
302 if (!(task = rpc_new_task(clnt, RPC_TASK_ASYNC, tk_ops, calldata)))
303 return -ENOMEM;
304 rpc_execute(task);
305 return 0;
306} 314}
307 315
308static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task) 316static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
@@ -316,7 +324,34 @@ static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
316 return ret; 324 return ret;
317} 325}
318 326
319static inline void update_open_stateflags(struct nfs4_state *state, mode_t open_flags) 327static int can_open_cached(struct nfs4_state *state, int mode)
328{
329 int ret = 0;
330 switch (mode & (FMODE_READ|FMODE_WRITE|O_EXCL)) {
331 case FMODE_READ:
332 ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0;
333 ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0;
334 break;
335 case FMODE_WRITE:
336 ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0;
337 ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0;
338 break;
339 case FMODE_READ|FMODE_WRITE:
340 ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0;
341 }
342 return ret;
343}
344
345static int can_open_delegated(struct nfs_delegation *delegation, mode_t open_flags)
346{
347 if ((delegation->type & open_flags) != open_flags)
348 return 0;
349 if (delegation->flags & NFS_DELEGATION_NEED_RECLAIM)
350 return 0;
351 return 1;
352}
353
354static void update_open_stateflags(struct nfs4_state *state, mode_t open_flags)
320{ 355{
321 switch (open_flags) { 356 switch (open_flags) {
322 case FMODE_WRITE: 357 case FMODE_WRITE:
@@ -328,41 +363,176 @@ static inline void update_open_stateflags(struct nfs4_state *state, mode_t open_
328 case FMODE_READ|FMODE_WRITE: 363 case FMODE_READ|FMODE_WRITE:
329 state->n_rdwr++; 364 state->n_rdwr++;
330 } 365 }
366 nfs4_state_set_mode_locked(state, state->state | open_flags);
331} 367}
332 368
333static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) 369static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
334{ 370{
335 struct inode *inode = state->inode; 371 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
372 memcpy(state->stateid.data, stateid->data, sizeof(state->stateid.data));
373 memcpy(state->open_stateid.data, stateid->data, sizeof(state->open_stateid.data));
374 switch (open_flags) {
375 case FMODE_READ:
376 set_bit(NFS_O_RDONLY_STATE, &state->flags);
377 break;
378 case FMODE_WRITE:
379 set_bit(NFS_O_WRONLY_STATE, &state->flags);
380 break;
381 case FMODE_READ|FMODE_WRITE:
382 set_bit(NFS_O_RDWR_STATE, &state->flags);
383 }
384}
385
386static void nfs_set_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
387{
388 write_seqlock(&state->seqlock);
389 nfs_set_open_stateid_locked(state, stateid, open_flags);
390 write_sequnlock(&state->seqlock);
391}
336 392
393static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *deleg_stateid, int open_flags)
394{
337 open_flags &= (FMODE_READ|FMODE_WRITE); 395 open_flags &= (FMODE_READ|FMODE_WRITE);
338 /* Protect against nfs4_find_state_byowner() */ 396 /*
397 * Protect the call to nfs4_state_set_mode_locked and
398 * serialise the stateid update
399 */
400 write_seqlock(&state->seqlock);
401 if (deleg_stateid != NULL) {
402 memcpy(state->stateid.data, deleg_stateid->data, sizeof(state->stateid.data));
403 set_bit(NFS_DELEGATED_STATE, &state->flags);
404 }
405 if (open_stateid != NULL)
406 nfs_set_open_stateid_locked(state, open_stateid, open_flags);
407 write_sequnlock(&state->seqlock);
339 spin_lock(&state->owner->so_lock); 408 spin_lock(&state->owner->so_lock);
340 spin_lock(&inode->i_lock);
341 memcpy(&state->stateid, stateid, sizeof(state->stateid));
342 update_open_stateflags(state, open_flags); 409 update_open_stateflags(state, open_flags);
343 nfs4_state_set_mode_locked(state, state->state | open_flags);
344 spin_unlock(&inode->i_lock);
345 spin_unlock(&state->owner->so_lock); 410 spin_unlock(&state->owner->so_lock);
346} 411}
347 412
413static void nfs4_return_incompatible_delegation(struct inode *inode, mode_t open_flags)
414{
415 struct nfs_delegation *delegation;
416
417 rcu_read_lock();
418 delegation = rcu_dereference(NFS_I(inode)->delegation);
419 if (delegation == NULL || (delegation->type & open_flags) == open_flags) {
420 rcu_read_unlock();
421 return;
422 }
423 rcu_read_unlock();
424 nfs_inode_return_delegation(inode);
425}
426
427static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
428{
429 struct nfs4_state *state = opendata->state;
430 struct nfs_inode *nfsi = NFS_I(state->inode);
431 struct nfs_delegation *delegation;
432 int open_mode = opendata->o_arg.open_flags & (FMODE_READ|FMODE_WRITE|O_EXCL);
433 nfs4_stateid stateid;
434 int ret = -EAGAIN;
435
436 rcu_read_lock();
437 delegation = rcu_dereference(nfsi->delegation);
438 for (;;) {
439 if (can_open_cached(state, open_mode)) {
440 spin_lock(&state->owner->so_lock);
441 if (can_open_cached(state, open_mode)) {
442 update_open_stateflags(state, open_mode);
443 spin_unlock(&state->owner->so_lock);
444 rcu_read_unlock();
445 goto out_return_state;
446 }
447 spin_unlock(&state->owner->so_lock);
448 }
449 if (delegation == NULL)
450 break;
451 if (!can_open_delegated(delegation, open_mode))
452 break;
453 /* Save the delegation */
454 memcpy(stateid.data, delegation->stateid.data, sizeof(stateid.data));
455 rcu_read_unlock();
456 lock_kernel();
457 ret = _nfs4_do_access(state->inode, state->owner->so_cred, open_mode);
458 unlock_kernel();
459 if (ret != 0)
460 goto out;
461 ret = -EAGAIN;
462 rcu_read_lock();
463 delegation = rcu_dereference(nfsi->delegation);
464 /* If no delegation, try a cached open */
465 if (delegation == NULL)
466 continue;
467 /* Is the delegation still valid? */
468 if (memcmp(stateid.data, delegation->stateid.data, sizeof(stateid.data)) != 0)
469 continue;
470 rcu_read_unlock();
471 update_open_stateid(state, NULL, &stateid, open_mode);
472 goto out_return_state;
473 }
474 rcu_read_unlock();
475out:
476 return ERR_PTR(ret);
477out_return_state:
478 atomic_inc(&state->count);
479 return state;
480}
481
348static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data) 482static struct nfs4_state *nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
349{ 483{
350 struct inode *inode; 484 struct inode *inode;
351 struct nfs4_state *state = NULL; 485 struct nfs4_state *state = NULL;
486 struct nfs_delegation *delegation;
487 nfs4_stateid *deleg_stateid = NULL;
488 int ret;
352 489
353 if (!(data->f_attr.valid & NFS_ATTR_FATTR)) 490 if (!data->rpc_done) {
491 state = nfs4_try_open_cached(data);
354 goto out; 492 goto out;
493 }
494
495 ret = -EAGAIN;
496 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
497 goto err;
355 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr); 498 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr);
499 ret = PTR_ERR(inode);
356 if (IS_ERR(inode)) 500 if (IS_ERR(inode))
357 goto out; 501 goto err;
502 ret = -ENOMEM;
358 state = nfs4_get_open_state(inode, data->owner); 503 state = nfs4_get_open_state(inode, data->owner);
359 if (state == NULL) 504 if (state == NULL)
360 goto put_inode; 505 goto err_put_inode;
361 update_open_stateid(state, &data->o_res.stateid, data->o_arg.open_flags); 506 if (data->o_res.delegation_type != 0) {
362put_inode: 507 int delegation_flags = 0;
508
509 rcu_read_lock();
510 delegation = rcu_dereference(NFS_I(inode)->delegation);
511 if (delegation)
512 delegation_flags = delegation->flags;
513 rcu_read_unlock();
514 if (!(delegation_flags & NFS_DELEGATION_NEED_RECLAIM))
515 nfs_inode_set_delegation(state->inode,
516 data->owner->so_cred,
517 &data->o_res);
518 else
519 nfs_inode_reclaim_delegation(state->inode,
520 data->owner->so_cred,
521 &data->o_res);
522 }
523 rcu_read_lock();
524 delegation = rcu_dereference(NFS_I(inode)->delegation);
525 if (delegation != NULL)
526 deleg_stateid = &delegation->stateid;
527 update_open_stateid(state, &data->o_res.stateid, deleg_stateid, data->o_arg.open_flags);
528 rcu_read_unlock();
363 iput(inode); 529 iput(inode);
364out: 530out:
365 return state; 531 return state;
532err_put_inode:
533 iput(inode);
534err:
535 return ERR_PTR(ret);
366} 536}
367 537
368static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state) 538static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
@@ -382,79 +552,66 @@ static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *
382 return ERR_PTR(-ENOENT); 552 return ERR_PTR(-ENOENT);
383} 553}
384 554
385static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, mode_t openflags, nfs4_stateid *stateid) 555static int nfs4_open_recover_helper(struct nfs4_opendata *opendata, mode_t openflags, struct nfs4_state **res)
386{ 556{
557 struct nfs4_state *newstate;
387 int ret; 558 int ret;
388 559
389 opendata->o_arg.open_flags = openflags; 560 opendata->o_arg.open_flags = openflags;
561 memset(&opendata->o_res, 0, sizeof(opendata->o_res));
562 memset(&opendata->c_res, 0, sizeof(opendata->c_res));
563 nfs4_init_opendata_res(opendata);
390 ret = _nfs4_proc_open(opendata); 564 ret = _nfs4_proc_open(opendata);
391 if (ret != 0) 565 if (ret != 0)
392 return ret; 566 return ret;
393 memcpy(stateid->data, opendata->o_res.stateid.data, 567 newstate = nfs4_opendata_to_nfs4_state(opendata);
394 sizeof(stateid->data)); 568 if (IS_ERR(newstate))
569 return PTR_ERR(newstate);
570 nfs4_close_state(&opendata->path, newstate, openflags);
571 *res = newstate;
395 return 0; 572 return 0;
396} 573}
397 574
398static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state) 575static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
399{ 576{
400 nfs4_stateid stateid;
401 struct nfs4_state *newstate; 577 struct nfs4_state *newstate;
402 int mode = 0;
403 int delegation = 0;
404 int ret; 578 int ret;
405 579
406 /* memory barrier prior to reading state->n_* */ 580 /* memory barrier prior to reading state->n_* */
581 clear_bit(NFS_DELEGATED_STATE, &state->flags);
407 smp_rmb(); 582 smp_rmb();
408 if (state->n_rdwr != 0) { 583 if (state->n_rdwr != 0) {
409 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &stateid); 584 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE, &newstate);
410 if (ret != 0) 585 if (ret != 0)
411 return ret; 586 return ret;
412 mode |= FMODE_READ|FMODE_WRITE; 587 if (newstate != state)
413 if (opendata->o_res.delegation_type != 0) 588 return -ESTALE;
414 delegation = opendata->o_res.delegation_type;
415 smp_rmb();
416 } 589 }
417 if (state->n_wronly != 0) { 590 if (state->n_wronly != 0) {
418 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &stateid); 591 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE, &newstate);
419 if (ret != 0) 592 if (ret != 0)
420 return ret; 593 return ret;
421 mode |= FMODE_WRITE; 594 if (newstate != state)
422 if (opendata->o_res.delegation_type != 0) 595 return -ESTALE;
423 delegation = opendata->o_res.delegation_type;
424 smp_rmb();
425 } 596 }
426 if (state->n_rdonly != 0) { 597 if (state->n_rdonly != 0) {
427 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &stateid); 598 ret = nfs4_open_recover_helper(opendata, FMODE_READ, &newstate);
428 if (ret != 0) 599 if (ret != 0)
429 return ret; 600 return ret;
430 mode |= FMODE_READ; 601 if (newstate != state)
602 return -ESTALE;
431 } 603 }
432 clear_bit(NFS_DELEGATED_STATE, &state->flags); 604 /*
433 if (mode == 0) 605 * We may have performed cached opens for all three recoveries.
434 return 0; 606 * Check if we need to update the current stateid.
435 if (opendata->o_res.delegation_type == 0) 607 */
436 opendata->o_res.delegation_type = delegation; 608 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
437 opendata->o_arg.open_flags |= mode; 609 memcmp(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data)) != 0) {
438 newstate = nfs4_opendata_to_nfs4_state(opendata); 610 write_seqlock(&state->seqlock);
439 if (newstate != NULL) { 611 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
440 if (opendata->o_res.delegation_type != 0) { 612 memcpy(state->stateid.data, state->open_stateid.data, sizeof(state->stateid.data));
441 struct nfs_inode *nfsi = NFS_I(newstate->inode); 613 write_sequnlock(&state->seqlock);
442 int delegation_flags = 0;
443 if (nfsi->delegation)
444 delegation_flags = nfsi->delegation->flags;
445 if (!(delegation_flags & NFS_DELEGATION_NEED_RECLAIM))
446 nfs_inode_set_delegation(newstate->inode,
447 opendata->owner->so_cred,
448 &opendata->o_res);
449 else
450 nfs_inode_reclaim_delegation(newstate->inode,
451 opendata->owner->so_cred,
452 &opendata->o_res);
453 }
454 nfs4_close_state(newstate, opendata->o_arg.open_flags);
455 } 614 }
456 if (newstate != state)
457 return -ESTALE;
458 return 0; 615 return 0;
459} 616}
460 617
@@ -462,41 +619,37 @@ static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *
462 * OPEN_RECLAIM: 619 * OPEN_RECLAIM:
463 * reclaim state on the server after a reboot. 620 * reclaim state on the server after a reboot.
464 */ 621 */
465static int _nfs4_do_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) 622static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
466{ 623{
467 struct nfs_delegation *delegation = NFS_I(state->inode)->delegation; 624 struct nfs_delegation *delegation;
468 struct nfs4_opendata *opendata; 625 struct nfs4_opendata *opendata;
469 int delegation_type = 0; 626 int delegation_type = 0;
470 int status; 627 int status;
471 628
472 if (delegation != NULL) { 629 opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, NULL);
473 if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
474 memcpy(&state->stateid, &delegation->stateid,
475 sizeof(state->stateid));
476 set_bit(NFS_DELEGATED_STATE, &state->flags);
477 return 0;
478 }
479 delegation_type = delegation->type;
480 }
481 opendata = nfs4_opendata_alloc(dentry, sp, 0, NULL);
482 if (opendata == NULL) 630 if (opendata == NULL)
483 return -ENOMEM; 631 return -ENOMEM;
484 opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS; 632 opendata->o_arg.claim = NFS4_OPEN_CLAIM_PREVIOUS;
485 opendata->o_arg.fh = NFS_FH(state->inode); 633 opendata->o_arg.fh = NFS_FH(state->inode);
486 nfs_copy_fh(&opendata->o_res.fh, opendata->o_arg.fh); 634 nfs_copy_fh(&opendata->o_res.fh, opendata->o_arg.fh);
635 rcu_read_lock();
636 delegation = rcu_dereference(NFS_I(state->inode)->delegation);
637 if (delegation != NULL && (delegation->flags & NFS_DELEGATION_NEED_RECLAIM) != 0)
638 delegation_type = delegation->flags;
639 rcu_read_unlock();
487 opendata->o_arg.u.delegation_type = delegation_type; 640 opendata->o_arg.u.delegation_type = delegation_type;
488 status = nfs4_open_recover(opendata, state); 641 status = nfs4_open_recover(opendata, state);
489 nfs4_opendata_free(opendata); 642 nfs4_opendata_put(opendata);
490 return status; 643 return status;
491} 644}
492 645
493static int nfs4_do_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) 646static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
494{ 647{
495 struct nfs_server *server = NFS_SERVER(state->inode); 648 struct nfs_server *server = NFS_SERVER(state->inode);
496 struct nfs4_exception exception = { }; 649 struct nfs4_exception exception = { };
497 int err; 650 int err;
498 do { 651 do {
499 err = _nfs4_do_open_reclaim(sp, state, dentry); 652 err = _nfs4_do_open_reclaim(ctx, state);
500 if (err != -NFS4ERR_DELAY) 653 if (err != -NFS4ERR_DELAY)
501 break; 654 break;
502 nfs4_handle_exception(server, err, &exception); 655 nfs4_handle_exception(server, err, &exception);
@@ -512,37 +665,35 @@ static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *sta
512 ctx = nfs4_state_find_open_context(state); 665 ctx = nfs4_state_find_open_context(state);
513 if (IS_ERR(ctx)) 666 if (IS_ERR(ctx))
514 return PTR_ERR(ctx); 667 return PTR_ERR(ctx);
515 ret = nfs4_do_open_reclaim(sp, state, ctx->dentry); 668 ret = nfs4_do_open_reclaim(ctx, state);
516 put_nfs_open_context(ctx); 669 put_nfs_open_context(ctx);
517 return ret; 670 return ret;
518} 671}
519 672
520static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) 673static int _nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
521{ 674{
522 struct nfs4_state_owner *sp = state->owner; 675 struct nfs4_state_owner *sp = state->owner;
523 struct nfs4_opendata *opendata; 676 struct nfs4_opendata *opendata;
524 int ret; 677 int ret;
525 678
526 if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) 679 opendata = nfs4_opendata_alloc(&ctx->path, sp, 0, NULL);
527 return 0;
528 opendata = nfs4_opendata_alloc(dentry, sp, 0, NULL);
529 if (opendata == NULL) 680 if (opendata == NULL)
530 return -ENOMEM; 681 return -ENOMEM;
531 opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR; 682 opendata->o_arg.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR;
532 memcpy(opendata->o_arg.u.delegation.data, state->stateid.data, 683 memcpy(opendata->o_arg.u.delegation.data, stateid->data,
533 sizeof(opendata->o_arg.u.delegation.data)); 684 sizeof(opendata->o_arg.u.delegation.data));
534 ret = nfs4_open_recover(opendata, state); 685 ret = nfs4_open_recover(opendata, state);
535 nfs4_opendata_free(opendata); 686 nfs4_opendata_put(opendata);
536 return ret; 687 return ret;
537} 688}
538 689
539int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state) 690int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid)
540{ 691{
541 struct nfs4_exception exception = { }; 692 struct nfs4_exception exception = { };
542 struct nfs_server *server = NFS_SERVER(dentry->d_inode); 693 struct nfs_server *server = NFS_SERVER(state->inode);
543 int err; 694 int err;
544 do { 695 do {
545 err = _nfs4_open_delegation_recall(dentry, state); 696 err = _nfs4_open_delegation_recall(ctx, state, stateid);
546 switch (err) { 697 switch (err) {
547 case 0: 698 case 0:
548 return err; 699 return err;
@@ -582,9 +733,10 @@ static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
582 memcpy(data->o_res.stateid.data, data->c_res.stateid.data, 733 memcpy(data->o_res.stateid.data, data->c_res.stateid.data,
583 sizeof(data->o_res.stateid.data)); 734 sizeof(data->o_res.stateid.data));
584 renew_lease(data->o_res.server, data->timestamp); 735 renew_lease(data->o_res.server, data->timestamp);
736 data->rpc_done = 1;
585 } 737 }
586 nfs_increment_open_seqid(data->rpc_status, data->c_arg.seqid);
587 nfs_confirm_seqid(&data->owner->so_seqid, data->rpc_status); 738 nfs_confirm_seqid(&data->owner->so_seqid, data->rpc_status);
739 nfs_increment_open_seqid(data->rpc_status, data->c_arg.seqid);
588} 740}
589 741
590static void nfs4_open_confirm_release(void *calldata) 742static void nfs4_open_confirm_release(void *calldata)
@@ -596,14 +748,14 @@ static void nfs4_open_confirm_release(void *calldata)
596 if (data->cancelled == 0) 748 if (data->cancelled == 0)
597 goto out_free; 749 goto out_free;
598 /* In case of error, no cleanup! */ 750 /* In case of error, no cleanup! */
599 if (data->rpc_status != 0) 751 if (!data->rpc_done)
600 goto out_free; 752 goto out_free;
601 nfs_confirm_seqid(&data->owner->so_seqid, 0); 753 nfs_confirm_seqid(&data->owner->so_seqid, 0);
602 state = nfs4_opendata_to_nfs4_state(data); 754 state = nfs4_opendata_to_nfs4_state(data);
603 if (state != NULL) 755 if (!IS_ERR(state))
604 nfs4_close_state(state, data->o_arg.open_flags); 756 nfs4_close_state(&data->path, state, data->o_arg.open_flags);
605out_free: 757out_free:
606 nfs4_opendata_free(data); 758 nfs4_opendata_put(data);
607} 759}
608 760
609static const struct rpc_call_ops nfs4_open_confirm_ops = { 761static const struct rpc_call_ops nfs4_open_confirm_ops = {
@@ -621,12 +773,9 @@ static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
621 struct rpc_task *task; 773 struct rpc_task *task;
622 int status; 774 int status;
623 775
624 atomic_inc(&data->count); 776 kref_get(&data->kref);
625 /* 777 data->rpc_done = 0;
626 * If rpc_run_task() ends up calling ->rpc_release(), we 778 data->rpc_status = 0;
627 * want to ensure that it takes the 'error' code path.
628 */
629 data->rpc_status = -ENOMEM;
630 task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_confirm_ops, data); 779 task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_confirm_ops, data);
631 if (IS_ERR(task)) 780 if (IS_ERR(task))
632 return PTR_ERR(task); 781 return PTR_ERR(task);
@@ -653,13 +802,35 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
653 802
654 if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0) 803 if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
655 return; 804 return;
805 /*
806 * Check if we still need to send an OPEN call, or if we can use
807 * a delegation instead.
808 */
809 if (data->state != NULL) {
810 struct nfs_delegation *delegation;
811
812 if (can_open_cached(data->state, data->o_arg.open_flags & (FMODE_READ|FMODE_WRITE|O_EXCL)))
813 goto out_no_action;
814 rcu_read_lock();
815 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
816 if (delegation != NULL &&
817 (delegation->flags & NFS_DELEGATION_NEED_RECLAIM) == 0) {
818 rcu_read_unlock();
819 goto out_no_action;
820 }
821 rcu_read_unlock();
822 }
656 /* Update sequence id. */ 823 /* Update sequence id. */
657 data->o_arg.id = sp->so_id; 824 data->o_arg.id = sp->so_owner_id.id;
658 data->o_arg.clientid = sp->so_client->cl_clientid; 825 data->o_arg.clientid = sp->so_client->cl_clientid;
659 if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) 826 if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
660 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR]; 827 msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
661 data->timestamp = jiffies; 828 data->timestamp = jiffies;
662 rpc_call_setup(task, &msg, 0); 829 rpc_call_setup(task, &msg, 0);
830 return;
831out_no_action:
832 task->tk_action = NULL;
833
663} 834}
664 835
665static void nfs4_open_done(struct rpc_task *task, void *calldata) 836static void nfs4_open_done(struct rpc_task *task, void *calldata)
@@ -683,8 +854,11 @@ static void nfs4_open_done(struct rpc_task *task, void *calldata)
683 data->rpc_status = -ENOTDIR; 854 data->rpc_status = -ENOTDIR;
684 } 855 }
685 renew_lease(data->o_res.server, data->timestamp); 856 renew_lease(data->o_res.server, data->timestamp);
857 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
858 nfs_confirm_seqid(&data->owner->so_seqid, 0);
686 } 859 }
687 nfs_increment_open_seqid(data->rpc_status, data->o_arg.seqid); 860 nfs_increment_open_seqid(data->rpc_status, data->o_arg.seqid);
861 data->rpc_done = 1;
688} 862}
689 863
690static void nfs4_open_release(void *calldata) 864static void nfs4_open_release(void *calldata)
@@ -696,17 +870,17 @@ static void nfs4_open_release(void *calldata)
696 if (data->cancelled == 0) 870 if (data->cancelled == 0)
697 goto out_free; 871 goto out_free;
698 /* In case of error, no cleanup! */ 872 /* In case of error, no cleanup! */
699 if (data->rpc_status != 0) 873 if (data->rpc_status != 0 || !data->rpc_done)
700 goto out_free; 874 goto out_free;
701 /* In case we need an open_confirm, no cleanup! */ 875 /* In case we need an open_confirm, no cleanup! */
702 if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM) 876 if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
703 goto out_free; 877 goto out_free;
704 nfs_confirm_seqid(&data->owner->so_seqid, 0); 878 nfs_confirm_seqid(&data->owner->so_seqid, 0);
705 state = nfs4_opendata_to_nfs4_state(data); 879 state = nfs4_opendata_to_nfs4_state(data);
706 if (state != NULL) 880 if (!IS_ERR(state))
707 nfs4_close_state(state, data->o_arg.open_flags); 881 nfs4_close_state(&data->path, state, data->o_arg.open_flags);
708out_free: 882out_free:
709 nfs4_opendata_free(data); 883 nfs4_opendata_put(data);
710} 884}
711 885
712static const struct rpc_call_ops nfs4_open_ops = { 886static const struct rpc_call_ops nfs4_open_ops = {
@@ -727,12 +901,10 @@ static int _nfs4_proc_open(struct nfs4_opendata *data)
727 struct rpc_task *task; 901 struct rpc_task *task;
728 int status; 902 int status;
729 903
730 atomic_inc(&data->count); 904 kref_get(&data->kref);
731 /* 905 data->rpc_done = 0;
732 * If rpc_run_task() ends up calling ->rpc_release(), we 906 data->rpc_status = 0;
733 * want to ensure that it takes the 'error' code path. 907 data->cancelled = 0;
734 */
735 data->rpc_status = -ENOMEM;
736 task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_ops, data); 908 task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_open_ops, data);
737 if (IS_ERR(task)) 909 if (IS_ERR(task))
738 return PTR_ERR(task); 910 return PTR_ERR(task);
@@ -743,7 +915,7 @@ static int _nfs4_proc_open(struct nfs4_opendata *data)
743 } else 915 } else
744 status = data->rpc_status; 916 status = data->rpc_status;
745 rpc_put_task(task); 917 rpc_put_task(task);
746 if (status != 0) 918 if (status != 0 || !data->rpc_done)
747 return status; 919 return status;
748 920
749 if (o_arg->open_flags & O_CREAT) { 921 if (o_arg->open_flags & O_CREAT) {
@@ -756,7 +928,6 @@ static int _nfs4_proc_open(struct nfs4_opendata *data)
756 if (status != 0) 928 if (status != 0)
757 return status; 929 return status;
758 } 930 }
759 nfs_confirm_seqid(&data->owner->so_seqid, 0);
760 if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) 931 if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
761 return server->nfs_client->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); 932 return server->nfs_client->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
762 return 0; 933 return 0;
@@ -772,6 +943,8 @@ static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openf
772 mask |= MAY_READ; 943 mask |= MAY_READ;
773 if (openflags & FMODE_WRITE) 944 if (openflags & FMODE_WRITE)
774 mask |= MAY_WRITE; 945 mask |= MAY_WRITE;
946 if (openflags & FMODE_EXEC)
947 mask |= MAY_EXEC;
775 status = nfs_access_get_cached(inode, cred, &cache); 948 status = nfs_access_get_cached(inode, cred, &cache);
776 if (status == 0) 949 if (status == 0)
777 goto out; 950 goto out;
@@ -811,43 +984,32 @@ static int nfs4_recover_expired_lease(struct nfs_server *server)
811 * reclaim state on the server after a network partition. 984 * reclaim state on the server after a network partition.
812 * Assumes caller holds the appropriate lock 985 * Assumes caller holds the appropriate lock
813 */ 986 */
814static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) 987static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
815{ 988{
816 struct inode *inode = state->inode;
817 struct nfs_delegation *delegation = NFS_I(inode)->delegation;
818 struct nfs4_opendata *opendata; 989 struct nfs4_opendata *opendata;
819 int openflags = state->state & (FMODE_READ|FMODE_WRITE);
820 int ret; 990 int ret;
821 991
822 if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) { 992 opendata = nfs4_opendata_alloc(&ctx->path, state->owner, 0, NULL);
823 ret = _nfs4_do_access(inode, sp->so_cred, openflags);
824 if (ret < 0)
825 return ret;
826 memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid));
827 set_bit(NFS_DELEGATED_STATE, &state->flags);
828 return 0;
829 }
830 opendata = nfs4_opendata_alloc(dentry, sp, openflags, NULL);
831 if (opendata == NULL) 993 if (opendata == NULL)
832 return -ENOMEM; 994 return -ENOMEM;
833 ret = nfs4_open_recover(opendata, state); 995 ret = nfs4_open_recover(opendata, state);
834 if (ret == -ESTALE) { 996 if (ret == -ESTALE) {
835 /* Invalidate the state owner so we don't ever use it again */ 997 /* Invalidate the state owner so we don't ever use it again */
836 nfs4_drop_state_owner(sp); 998 nfs4_drop_state_owner(state->owner);
837 d_drop(dentry); 999 d_drop(ctx->path.dentry);
838 } 1000 }
839 nfs4_opendata_free(opendata); 1001 nfs4_opendata_put(opendata);
840 return ret; 1002 return ret;
841} 1003}
842 1004
843static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry) 1005static inline int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
844{ 1006{
845 struct nfs_server *server = NFS_SERVER(dentry->d_inode); 1007 struct nfs_server *server = NFS_SERVER(state->inode);
846 struct nfs4_exception exception = { }; 1008 struct nfs4_exception exception = { };
847 int err; 1009 int err;
848 1010
849 do { 1011 do {
850 err = _nfs4_open_expired(sp, state, dentry); 1012 err = _nfs4_open_expired(ctx, state);
851 if (err == -NFS4ERR_DELAY) 1013 if (err == -NFS4ERR_DELAY)
852 nfs4_handle_exception(server, err, &exception); 1014 nfs4_handle_exception(server, err, &exception);
853 } while (exception.retry); 1015 } while (exception.retry);
@@ -862,107 +1024,38 @@ static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *sta
862 ctx = nfs4_state_find_open_context(state); 1024 ctx = nfs4_state_find_open_context(state);
863 if (IS_ERR(ctx)) 1025 if (IS_ERR(ctx))
864 return PTR_ERR(ctx); 1026 return PTR_ERR(ctx);
865 ret = nfs4_do_open_expired(sp, state, ctx->dentry); 1027 ret = nfs4_do_open_expired(ctx, state);
866 put_nfs_open_context(ctx); 1028 put_nfs_open_context(ctx);
867 return ret; 1029 return ret;
868} 1030}
869 1031
870/* 1032/*
871 * Returns a referenced nfs4_state if there is an open delegation on the file 1033 * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
1034 * fields corresponding to attributes that were used to store the verifier.
1035 * Make sure we clobber those fields in the later setattr call
872 */ 1036 */
873static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res) 1037static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr)
874{
875 struct nfs_delegation *delegation;
876 struct nfs_server *server = NFS_SERVER(inode);
877 struct nfs_client *clp = server->nfs_client;
878 struct nfs_inode *nfsi = NFS_I(inode);
879 struct nfs4_state_owner *sp = NULL;
880 struct nfs4_state *state = NULL;
881 int open_flags = flags & (FMODE_READ|FMODE_WRITE);
882 int err;
883
884 err = -ENOMEM;
885 if (!(sp = nfs4_get_state_owner(server, cred))) {
886 dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
887 return err;
888 }
889 err = nfs4_recover_expired_lease(server);
890 if (err != 0)
891 goto out_put_state_owner;
892 /* Protect against reboot recovery - NOTE ORDER! */
893 down_read(&clp->cl_sem);
894 /* Protect against delegation recall */
895 down_read(&nfsi->rwsem);
896 delegation = NFS_I(inode)->delegation;
897 err = -ENOENT;
898 if (delegation == NULL || (delegation->type & open_flags) != open_flags)
899 goto out_err;
900 err = -ENOMEM;
901 state = nfs4_get_open_state(inode, sp);
902 if (state == NULL)
903 goto out_err;
904
905 err = -ENOENT;
906 if ((state->state & open_flags) == open_flags) {
907 spin_lock(&inode->i_lock);
908 update_open_stateflags(state, open_flags);
909 spin_unlock(&inode->i_lock);
910 goto out_ok;
911 } else if (state->state != 0)
912 goto out_put_open_state;
913
914 lock_kernel();
915 err = _nfs4_do_access(inode, cred, open_flags);
916 unlock_kernel();
917 if (err != 0)
918 goto out_put_open_state;
919 set_bit(NFS_DELEGATED_STATE, &state->flags);
920 update_open_stateid(state, &delegation->stateid, open_flags);
921out_ok:
922 nfs4_put_state_owner(sp);
923 up_read(&nfsi->rwsem);
924 up_read(&clp->cl_sem);
925 *res = state;
926 return 0;
927out_put_open_state:
928 nfs4_put_open_state(state);
929out_err:
930 up_read(&nfsi->rwsem);
931 up_read(&clp->cl_sem);
932 if (err != -EACCES)
933 nfs_inode_return_delegation(inode);
934out_put_state_owner:
935 nfs4_put_state_owner(sp);
936 return err;
937}
938
939static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
940{ 1038{
941 struct nfs4_exception exception = { }; 1039 if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
942 struct nfs4_state *res = ERR_PTR(-EIO); 1040 !(sattr->ia_valid & ATTR_ATIME_SET))
943 int err; 1041 sattr->ia_valid |= ATTR_ATIME;
944 1042
945 do { 1043 if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
946 err = _nfs4_open_delegated(inode, flags, cred, &res); 1044 !(sattr->ia_valid & ATTR_MTIME_SET))
947 if (err == 0) 1045 sattr->ia_valid |= ATTR_MTIME;
948 break;
949 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(inode),
950 err, &exception));
951 } while (exception.retry);
952 return res;
953} 1046}
954 1047
955/* 1048/*
956 * Returns a referenced nfs4_state 1049 * Returns a referenced nfs4_state
957 */ 1050 */
958static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res) 1051static int _nfs4_do_open(struct inode *dir, struct path *path, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
959{ 1052{
960 struct nfs4_state_owner *sp; 1053 struct nfs4_state_owner *sp;
961 struct nfs4_state *state = NULL; 1054 struct nfs4_state *state = NULL;
962 struct nfs_server *server = NFS_SERVER(dir); 1055 struct nfs_server *server = NFS_SERVER(dir);
963 struct nfs_client *clp = server->nfs_client; 1056 struct nfs_client *clp = server->nfs_client;
964 struct nfs4_opendata *opendata; 1057 struct nfs4_opendata *opendata;
965 int status; 1058 int status;
966 1059
967 /* Protect against reboot recovery conflicts */ 1060 /* Protect against reboot recovery conflicts */
968 status = -ENOMEM; 1061 status = -ENOMEM;
@@ -973,29 +1066,35 @@ static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, st
973 status = nfs4_recover_expired_lease(server); 1066 status = nfs4_recover_expired_lease(server);
974 if (status != 0) 1067 if (status != 0)
975 goto err_put_state_owner; 1068 goto err_put_state_owner;
1069 if (path->dentry->d_inode != NULL)
1070 nfs4_return_incompatible_delegation(path->dentry->d_inode, flags & (FMODE_READ|FMODE_WRITE));
976 down_read(&clp->cl_sem); 1071 down_read(&clp->cl_sem);
977 status = -ENOMEM; 1072 status = -ENOMEM;
978 opendata = nfs4_opendata_alloc(dentry, sp, flags, sattr); 1073 opendata = nfs4_opendata_alloc(path, sp, flags, sattr);
979 if (opendata == NULL) 1074 if (opendata == NULL)
980 goto err_release_rwsem; 1075 goto err_release_rwsem;
981 1076
1077 if (path->dentry->d_inode != NULL)
1078 opendata->state = nfs4_get_open_state(path->dentry->d_inode, sp);
1079
982 status = _nfs4_proc_open(opendata); 1080 status = _nfs4_proc_open(opendata);
983 if (status != 0) 1081 if (status != 0)
984 goto err_opendata_free; 1082 goto err_opendata_put;
1083
1084 if (opendata->o_arg.open_flags & O_EXCL)
1085 nfs4_exclusive_attrset(opendata, sattr);
985 1086
986 status = -ENOMEM;
987 state = nfs4_opendata_to_nfs4_state(opendata); 1087 state = nfs4_opendata_to_nfs4_state(opendata);
988 if (state == NULL) 1088 status = PTR_ERR(state);
989 goto err_opendata_free; 1089 if (IS_ERR(state))
990 if (opendata->o_res.delegation_type != 0) 1090 goto err_opendata_put;
991 nfs_inode_set_delegation(state->inode, cred, &opendata->o_res); 1091 nfs4_opendata_put(opendata);
992 nfs4_opendata_free(opendata);
993 nfs4_put_state_owner(sp); 1092 nfs4_put_state_owner(sp);
994 up_read(&clp->cl_sem); 1093 up_read(&clp->cl_sem);
995 *res = state; 1094 *res = state;
996 return 0; 1095 return 0;
997err_opendata_free: 1096err_opendata_put:
998 nfs4_opendata_free(opendata); 1097 nfs4_opendata_put(opendata);
999err_release_rwsem: 1098err_release_rwsem:
1000 up_read(&clp->cl_sem); 1099 up_read(&clp->cl_sem);
1001err_put_state_owner: 1100err_put_state_owner:
@@ -1006,14 +1105,14 @@ out_err:
1006} 1105}
1007 1106
1008 1107
1009static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred) 1108static struct nfs4_state *nfs4_do_open(struct inode *dir, struct path *path, int flags, struct iattr *sattr, struct rpc_cred *cred)
1010{ 1109{
1011 struct nfs4_exception exception = { }; 1110 struct nfs4_exception exception = { };
1012 struct nfs4_state *res; 1111 struct nfs4_state *res;
1013 int status; 1112 int status;
1014 1113
1015 do { 1114 do {
1016 status = _nfs4_do_open(dir, dentry, flags, sattr, cred, &res); 1115 status = _nfs4_do_open(dir, path, flags, sattr, cred, &res);
1017 if (status == 0) 1116 if (status == 0)
1018 break; 1117 break;
1019 /* NOTE: BAD_SEQID means the server and client disagree about the 1118 /* NOTE: BAD_SEQID means the server and client disagree about the
@@ -1028,7 +1127,9 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry,
1028 * the user though... 1127 * the user though...
1029 */ 1128 */
1030 if (status == -NFS4ERR_BAD_SEQID) { 1129 if (status == -NFS4ERR_BAD_SEQID) {
1031 printk(KERN_WARNING "NFS: v4 server returned a bad sequence-id error!\n"); 1130 printk(KERN_WARNING "NFS: v4 server %s "
1131 " returned a bad sequence-id error!\n",
1132 NFS_SERVER(dir)->nfs_client->cl_hostname);
1032 exception.retry = 1; 1133 exception.retry = 1;
1033 continue; 1134 continue;
1034 } 1135 }
@@ -1042,6 +1143,11 @@ static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry,
1042 exception.retry = 1; 1143 exception.retry = 1;
1043 continue; 1144 continue;
1044 } 1145 }
1146 if (status == -EAGAIN) {
1147 /* We must have found a delegation */
1148 exception.retry = 1;
1149 continue;
1150 }
1045 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir), 1151 res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
1046 status, &exception)); 1152 status, &exception));
1047 } while (exception.retry); 1153 } while (exception.retry);
@@ -1101,6 +1207,7 @@ static int nfs4_do_setattr(struct inode *inode, struct nfs_fattr *fattr,
1101} 1207}
1102 1208
1103struct nfs4_closedata { 1209struct nfs4_closedata {
1210 struct path path;
1104 struct inode *inode; 1211 struct inode *inode;
1105 struct nfs4_state *state; 1212 struct nfs4_state *state;
1106 struct nfs_closeargs arg; 1213 struct nfs_closeargs arg;
@@ -1117,6 +1224,8 @@ static void nfs4_free_closedata(void *data)
1117 nfs4_put_open_state(calldata->state); 1224 nfs4_put_open_state(calldata->state);
1118 nfs_free_seqid(calldata->arg.seqid); 1225 nfs_free_seqid(calldata->arg.seqid);
1119 nfs4_put_state_owner(sp); 1226 nfs4_put_state_owner(sp);
1227 dput(calldata->path.dentry);
1228 mntput(calldata->path.mnt);
1120 kfree(calldata); 1229 kfree(calldata);
1121} 1230}
1122 1231
@@ -1134,8 +1243,7 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
1134 nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid); 1243 nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid);
1135 switch (task->tk_status) { 1244 switch (task->tk_status) {
1136 case 0: 1245 case 0:
1137 memcpy(&state->stateid, &calldata->res.stateid, 1246 nfs_set_open_stateid(state, &calldata->res.stateid, calldata->arg.open_flags);
1138 sizeof(state->stateid));
1139 renew_lease(server, calldata->timestamp); 1247 renew_lease(server, calldata->timestamp);
1140 break; 1248 break;
1141 case -NFS4ERR_STALE_STATEID: 1249 case -NFS4ERR_STALE_STATEID:
@@ -1160,26 +1268,30 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data)
1160 .rpc_resp = &calldata->res, 1268 .rpc_resp = &calldata->res,
1161 .rpc_cred = state->owner->so_cred, 1269 .rpc_cred = state->owner->so_cred,
1162 }; 1270 };
1163 int mode = 0, old_mode; 1271 int clear_rd, clear_wr, clear_rdwr;
1272 int mode;
1164 1273
1165 if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0) 1274 if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
1166 return; 1275 return;
1167 /* Recalculate the new open mode in case someone reopened the file 1276
1168 * while we were waiting in line to be scheduled. 1277 mode = FMODE_READ|FMODE_WRITE;
1169 */ 1278 clear_rd = clear_wr = clear_rdwr = 0;
1170 spin_lock(&state->owner->so_lock); 1279 spin_lock(&state->owner->so_lock);
1171 spin_lock(&calldata->inode->i_lock); 1280 /* Calculate the change in open mode */
1172 mode = old_mode = state->state;
1173 if (state->n_rdwr == 0) { 1281 if (state->n_rdwr == 0) {
1174 if (state->n_rdonly == 0) 1282 if (state->n_rdonly == 0) {
1175 mode &= ~FMODE_READ; 1283 mode &= ~FMODE_READ;
1176 if (state->n_wronly == 0) 1284 clear_rd |= test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1285 clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1286 }
1287 if (state->n_wronly == 0) {
1177 mode &= ~FMODE_WRITE; 1288 mode &= ~FMODE_WRITE;
1289 clear_wr |= test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1290 clear_rdwr |= test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags);
1291 }
1178 } 1292 }
1179 nfs4_state_set_mode_locked(state, mode);
1180 spin_unlock(&calldata->inode->i_lock);
1181 spin_unlock(&state->owner->so_lock); 1293 spin_unlock(&state->owner->so_lock);
1182 if (mode == old_mode || test_bit(NFS_DELEGATED_STATE, &state->flags)) { 1294 if (!clear_rd && !clear_wr && !clear_rdwr) {
1183 /* Note: exit _without_ calling nfs4_close_done */ 1295 /* Note: exit _without_ calling nfs4_close_done */
1184 task->tk_action = NULL; 1296 task->tk_action = NULL;
1185 return; 1297 return;
@@ -1209,19 +1321,21 @@ static const struct rpc_call_ops nfs4_close_ops = {
1209 * 1321 *
1210 * NOTE: Caller must be holding the sp->so_owner semaphore! 1322 * NOTE: Caller must be holding the sp->so_owner semaphore!
1211 */ 1323 */
1212int nfs4_do_close(struct inode *inode, struct nfs4_state *state) 1324int nfs4_do_close(struct path *path, struct nfs4_state *state)
1213{ 1325{
1214 struct nfs_server *server = NFS_SERVER(inode); 1326 struct nfs_server *server = NFS_SERVER(state->inode);
1215 struct nfs4_closedata *calldata; 1327 struct nfs4_closedata *calldata;
1328 struct nfs4_state_owner *sp = state->owner;
1329 struct rpc_task *task;
1216 int status = -ENOMEM; 1330 int status = -ENOMEM;
1217 1331
1218 calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); 1332 calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
1219 if (calldata == NULL) 1333 if (calldata == NULL)
1220 goto out; 1334 goto out;
1221 calldata->inode = inode; 1335 calldata->inode = state->inode;
1222 calldata->state = state; 1336 calldata->state = state;
1223 calldata->arg.fh = NFS_FH(inode); 1337 calldata->arg.fh = NFS_FH(state->inode);
1224 calldata->arg.stateid = &state->stateid; 1338 calldata->arg.stateid = &state->open_stateid;
1225 /* Serialization for the sequence id */ 1339 /* Serialization for the sequence id */
1226 calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid); 1340 calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
1227 if (calldata->arg.seqid == NULL) 1341 if (calldata->arg.seqid == NULL)
@@ -1229,36 +1343,55 @@ int nfs4_do_close(struct inode *inode, struct nfs4_state *state)
1229 calldata->arg.bitmask = server->attr_bitmask; 1343 calldata->arg.bitmask = server->attr_bitmask;
1230 calldata->res.fattr = &calldata->fattr; 1344 calldata->res.fattr = &calldata->fattr;
1231 calldata->res.server = server; 1345 calldata->res.server = server;
1346 calldata->path.mnt = mntget(path->mnt);
1347 calldata->path.dentry = dget(path->dentry);
1232 1348
1233 status = nfs4_call_async(server->client, &nfs4_close_ops, calldata); 1349 task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_close_ops, calldata);
1234 if (status == 0) 1350 if (IS_ERR(task))
1235 goto out; 1351 return PTR_ERR(task);
1236 1352 rpc_put_task(task);
1237 nfs_free_seqid(calldata->arg.seqid); 1353 return 0;
1238out_free_calldata: 1354out_free_calldata:
1239 kfree(calldata); 1355 kfree(calldata);
1240out: 1356out:
1357 nfs4_put_open_state(state);
1358 nfs4_put_state_owner(sp);
1241 return status; 1359 return status;
1242} 1360}
1243 1361
1244static int nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state) 1362static int nfs4_intent_set_file(struct nameidata *nd, struct path *path, struct nfs4_state *state)
1245{ 1363{
1246 struct file *filp; 1364 struct file *filp;
1365 int ret;
1247 1366
1248 filp = lookup_instantiate_filp(nd, dentry, NULL); 1367 /* If the open_intent is for execute, we have an extra check to make */
1368 if (nd->intent.open.flags & FMODE_EXEC) {
1369 ret = _nfs4_do_access(state->inode,
1370 state->owner->so_cred,
1371 nd->intent.open.flags);
1372 if (ret < 0)
1373 goto out_close;
1374 }
1375 filp = lookup_instantiate_filp(nd, path->dentry, NULL);
1249 if (!IS_ERR(filp)) { 1376 if (!IS_ERR(filp)) {
1250 struct nfs_open_context *ctx; 1377 struct nfs_open_context *ctx;
1251 ctx = (struct nfs_open_context *)filp->private_data; 1378 ctx = (struct nfs_open_context *)filp->private_data;
1252 ctx->state = state; 1379 ctx->state = state;
1253 return 0; 1380 return 0;
1254 } 1381 }
1255 nfs4_close_state(state, nd->intent.open.flags); 1382 ret = PTR_ERR(filp);
1256 return PTR_ERR(filp); 1383out_close:
1384 nfs4_close_state(path, state, nd->intent.open.flags);
1385 return ret;
1257} 1386}
1258 1387
1259struct dentry * 1388struct dentry *
1260nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) 1389nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1261{ 1390{
1391 struct path path = {
1392 .mnt = nd->mnt,
1393 .dentry = dentry,
1394 };
1262 struct iattr attr; 1395 struct iattr attr;
1263 struct rpc_cred *cred; 1396 struct rpc_cred *cred;
1264 struct nfs4_state *state; 1397 struct nfs4_state *state;
@@ -1277,7 +1410,7 @@ nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1277 cred = rpcauth_lookupcred(NFS_CLIENT(dir)->cl_auth, 0); 1410 cred = rpcauth_lookupcred(NFS_CLIENT(dir)->cl_auth, 0);
1278 if (IS_ERR(cred)) 1411 if (IS_ERR(cred))
1279 return (struct dentry *)cred; 1412 return (struct dentry *)cred;
1280 state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred); 1413 state = nfs4_do_open(dir, &path, nd->intent.open.flags, &attr, cred);
1281 put_rpccred(cred); 1414 put_rpccred(cred);
1282 if (IS_ERR(state)) { 1415 if (IS_ERR(state)) {
1283 if (PTR_ERR(state) == -ENOENT) 1416 if (PTR_ERR(state) == -ENOENT)
@@ -1287,22 +1420,24 @@ nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd)
1287 res = d_add_unique(dentry, igrab(state->inode)); 1420 res = d_add_unique(dentry, igrab(state->inode));
1288 if (res != NULL) 1421 if (res != NULL)
1289 dentry = res; 1422 dentry = res;
1290 nfs4_intent_set_file(nd, dentry, state); 1423 nfs4_intent_set_file(nd, &path, state);
1291 return res; 1424 return res;
1292} 1425}
1293 1426
1294int 1427int
1295nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd) 1428nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd)
1296{ 1429{
1430 struct path path = {
1431 .mnt = nd->mnt,
1432 .dentry = dentry,
1433 };
1297 struct rpc_cred *cred; 1434 struct rpc_cred *cred;
1298 struct nfs4_state *state; 1435 struct nfs4_state *state;
1299 1436
1300 cred = rpcauth_lookupcred(NFS_CLIENT(dir)->cl_auth, 0); 1437 cred = rpcauth_lookupcred(NFS_CLIENT(dir)->cl_auth, 0);
1301 if (IS_ERR(cred)) 1438 if (IS_ERR(cred))
1302 return PTR_ERR(cred); 1439 return PTR_ERR(cred);
1303 state = nfs4_open_delegated(dentry->d_inode, openflags, cred); 1440 state = nfs4_do_open(dir, &path, openflags, NULL, cred);
1304 if (IS_ERR(state))
1305 state = nfs4_do_open(dir, dentry, openflags, NULL, cred);
1306 put_rpccred(cred); 1441 put_rpccred(cred);
1307 if (IS_ERR(state)) { 1442 if (IS_ERR(state)) {
1308 switch (PTR_ERR(state)) { 1443 switch (PTR_ERR(state)) {
@@ -1318,10 +1453,10 @@ nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, st
1318 } 1453 }
1319 } 1454 }
1320 if (state->inode == dentry->d_inode) { 1455 if (state->inode == dentry->d_inode) {
1321 nfs4_intent_set_file(nd, dentry, state); 1456 nfs4_intent_set_file(nd, &path, state);
1322 return 1; 1457 return 1;
1323 } 1458 }
1324 nfs4_close_state(state, openflags); 1459 nfs4_close_state(&path, state, openflags);
1325out_drop: 1460out_drop:
1326 d_drop(dentry); 1461 d_drop(dentry);
1327 return 0; 1462 return 0;
@@ -1559,8 +1694,6 @@ static int _nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
1559 dprintk("NFS call lookupfh %s\n", name->name); 1694 dprintk("NFS call lookupfh %s\n", name->name);
1560 status = rpc_call_sync(server->client, &msg, 0); 1695 status = rpc_call_sync(server->client, &msg, 0);
1561 dprintk("NFS reply lookupfh: %d\n", status); 1696 dprintk("NFS reply lookupfh: %d\n", status);
1562 if (status == -NFS4ERR_MOVED)
1563 status = -EREMOTE;
1564 return status; 1697 return status;
1565} 1698}
1566 1699
@@ -1571,10 +1704,13 @@ static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
1571 struct nfs4_exception exception = { }; 1704 struct nfs4_exception exception = { };
1572 int err; 1705 int err;
1573 do { 1706 do {
1574 err = nfs4_handle_exception(server, 1707 err = _nfs4_proc_lookupfh(server, dirfh, name, fhandle, fattr);
1575 _nfs4_proc_lookupfh(server, dirfh, name, 1708 /* FIXME: !!!! */
1576 fhandle, fattr), 1709 if (err == -NFS4ERR_MOVED) {
1577 &exception); 1710 err = -EREMOTE;
1711 break;
1712 }
1713 err = nfs4_handle_exception(server, err, &exception);
1578 } while (exception.retry); 1714 } while (exception.retry);
1579 return err; 1715 return err;
1580} 1716}
@@ -1582,28 +1718,10 @@ static int nfs4_proc_lookupfh(struct nfs_server *server, struct nfs_fh *dirfh,
1582static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name, 1718static int _nfs4_proc_lookup(struct inode *dir, struct qstr *name,
1583 struct nfs_fh *fhandle, struct nfs_fattr *fattr) 1719 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
1584{ 1720{
1585 int status; 1721 int status;
1586 struct nfs_server *server = NFS_SERVER(dir);
1587 struct nfs4_lookup_arg args = {
1588 .bitmask = server->attr_bitmask,
1589 .dir_fh = NFS_FH(dir),
1590 .name = name,
1591 };
1592 struct nfs4_lookup_res res = {
1593 .server = server,
1594 .fattr = fattr,
1595 .fh = fhandle,
1596 };
1597 struct rpc_message msg = {
1598 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
1599 .rpc_argp = &args,
1600 .rpc_resp = &res,
1601 };
1602
1603 nfs_fattr_init(fattr);
1604 1722
1605 dprintk("NFS call lookup %s\n", name->name); 1723 dprintk("NFS call lookup %s\n", name->name);
1606 status = rpc_call_sync(NFS_CLIENT(dir), &msg, 0); 1724 status = _nfs4_proc_lookupfh(NFS_SERVER(dir), NFS_FH(dir), name, fhandle, fattr);
1607 if (status == -NFS4ERR_MOVED) 1725 if (status == -NFS4ERR_MOVED)
1608 status = nfs4_get_referral(dir, name, fattr, fhandle); 1726 status = nfs4_get_referral(dir, name, fattr, fhandle);
1609 dprintk("NFS reply lookup: %d\n", status); 1727 dprintk("NFS reply lookup: %d\n", status);
@@ -1752,6 +1870,10 @@ static int
1752nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, 1870nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1753 int flags, struct nameidata *nd) 1871 int flags, struct nameidata *nd)
1754{ 1872{
1873 struct path path = {
1874 .mnt = nd->mnt,
1875 .dentry = dentry,
1876 };
1755 struct nfs4_state *state; 1877 struct nfs4_state *state;
1756 struct rpc_cred *cred; 1878 struct rpc_cred *cred;
1757 int status = 0; 1879 int status = 0;
@@ -1761,7 +1883,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1761 status = PTR_ERR(cred); 1883 status = PTR_ERR(cred);
1762 goto out; 1884 goto out;
1763 } 1885 }
1764 state = nfs4_do_open(dir, dentry, flags, sattr, cred); 1886 state = nfs4_do_open(dir, &path, flags, sattr, cred);
1765 put_rpccred(cred); 1887 put_rpccred(cred);
1766 if (IS_ERR(state)) { 1888 if (IS_ERR(state)) {
1767 status = PTR_ERR(state); 1889 status = PTR_ERR(state);
@@ -1773,11 +1895,12 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
1773 status = nfs4_do_setattr(state->inode, &fattr, sattr, state); 1895 status = nfs4_do_setattr(state->inode, &fattr, sattr, state);
1774 if (status == 0) 1896 if (status == 0)
1775 nfs_setattr_update_inode(state->inode, sattr); 1897 nfs_setattr_update_inode(state->inode, sattr);
1898 nfs_post_op_update_inode(state->inode, &fattr);
1776 } 1899 }
1777 if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN)) 1900 if (status == 0 && (nd->flags & LOOKUP_OPEN) != 0)
1778 status = nfs4_intent_set_file(nd, dentry, state); 1901 status = nfs4_intent_set_file(nd, &path, state);
1779 else 1902 else
1780 nfs4_close_state(state, flags); 1903 nfs4_close_state(&path, state, flags);
1781out: 1904out:
1782 return status; 1905 return status;
1783} 1906}
@@ -3008,7 +3131,7 @@ static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock
3008 if (status != 0) 3131 if (status != 0)
3009 goto out; 3132 goto out;
3010 lsp = request->fl_u.nfs4_fl.owner; 3133 lsp = request->fl_u.nfs4_fl.owner;
3011 arg.lock_owner.id = lsp->ls_id; 3134 arg.lock_owner.id = lsp->ls_id.id;
3012 status = rpc_call_sync(server->client, &msg, 0); 3135 status = rpc_call_sync(server->client, &msg, 0);
3013 switch (status) { 3136 switch (status) {
3014 case 0: 3137 case 0:
@@ -3152,6 +3275,11 @@ static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
3152{ 3275{
3153 struct nfs4_unlockdata *data; 3276 struct nfs4_unlockdata *data;
3154 3277
3278 /* Ensure this is an unlock - when canceling a lock, the
3279 * canceled lock is passed in, and it won't be an unlock.
3280 */
3281 fl->fl_type = F_UNLCK;
3282
3155 data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid); 3283 data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
3156 if (data == NULL) { 3284 if (data == NULL) {
3157 nfs_free_seqid(seqid); 3285 nfs_free_seqid(seqid);
@@ -3222,7 +3350,7 @@ static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
3222 goto out_free; 3350 goto out_free;
3223 p->arg.lock_stateid = &lsp->ls_stateid; 3351 p->arg.lock_stateid = &lsp->ls_stateid;
3224 p->arg.lock_owner.clientid = server->nfs_client->cl_clientid; 3352 p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
3225 p->arg.lock_owner.id = lsp->ls_id; 3353 p->arg.lock_owner.id = lsp->ls_id.id;
3226 p->lsp = lsp; 3354 p->lsp = lsp;
3227 atomic_inc(&lsp->ls_count); 3355 atomic_inc(&lsp->ls_count);
3228 p->ctx = get_nfs_open_context(ctx); 3356 p->ctx = get_nfs_open_context(ctx);
@@ -3285,7 +3413,7 @@ static void nfs4_lock_done(struct rpc_task *task, void *calldata)
3285 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data, 3413 memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
3286 sizeof(data->lsp->ls_stateid.data)); 3414 sizeof(data->lsp->ls_stateid.data));
3287 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED; 3415 data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
3288 renew_lease(NFS_SERVER(data->ctx->dentry->d_inode), data->timestamp); 3416 renew_lease(NFS_SERVER(data->ctx->path.dentry->d_inode), data->timestamp);
3289 } 3417 }
3290 nfs_increment_lock_seqid(data->rpc_status, data->arg.lock_seqid); 3418 nfs_increment_lock_seqid(data->rpc_status, data->arg.lock_seqid);
3291out: 3419out:
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index 8ed79d5c54f9..e9662ba81d86 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -38,12 +38,14 @@
38 * subsequent patch. 38 * subsequent patch.
39 */ 39 */
40 40
41#include <linux/kernel.h>
41#include <linux/slab.h> 42#include <linux/slab.h>
42#include <linux/smp_lock.h> 43#include <linux/smp_lock.h>
43#include <linux/nfs_fs.h> 44#include <linux/nfs_fs.h>
44#include <linux/nfs_idmap.h> 45#include <linux/nfs_idmap.h>
45#include <linux/kthread.h> 46#include <linux/kthread.h>
46#include <linux/module.h> 47#include <linux/module.h>
48#include <linux/random.h>
47#include <linux/workqueue.h> 49#include <linux/workqueue.h>
48#include <linux/bitops.h> 50#include <linux/bitops.h>
49 51
@@ -69,33 +71,14 @@ static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred)
69 return status; 71 return status;
70} 72}
71 73
72u32
73nfs4_alloc_lockowner_id(struct nfs_client *clp)
74{
75 return clp->cl_lockowner_id ++;
76}
77
78static struct nfs4_state_owner *
79nfs4_client_grab_unused(struct nfs_client *clp, struct rpc_cred *cred)
80{
81 struct nfs4_state_owner *sp = NULL;
82
83 if (!list_empty(&clp->cl_unused)) {
84 sp = list_entry(clp->cl_unused.next, struct nfs4_state_owner, so_list);
85 atomic_inc(&sp->so_count);
86 sp->so_cred = cred;
87 list_move(&sp->so_list, &clp->cl_state_owners);
88 clp->cl_nunused--;
89 }
90 return sp;
91}
92
93struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp) 74struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp)
94{ 75{
95 struct nfs4_state_owner *sp; 76 struct nfs4_state_owner *sp;
77 struct rb_node *pos;
96 struct rpc_cred *cred = NULL; 78 struct rpc_cred *cred = NULL;
97 79
98 list_for_each_entry(sp, &clp->cl_state_owners, so_list) { 80 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
81 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
99 if (list_empty(&sp->so_states)) 82 if (list_empty(&sp->so_states))
100 continue; 83 continue;
101 cred = get_rpccred(sp->so_cred); 84 cred = get_rpccred(sp->so_cred);
@@ -107,32 +90,146 @@ struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp)
107static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp) 90static struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
108{ 91{
109 struct nfs4_state_owner *sp; 92 struct nfs4_state_owner *sp;
93 struct rb_node *pos;
110 94
111 if (!list_empty(&clp->cl_state_owners)) { 95 pos = rb_first(&clp->cl_state_owners);
112 sp = list_entry(clp->cl_state_owners.next, 96 if (pos != NULL) {
113 struct nfs4_state_owner, so_list); 97 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
114 return get_rpccred(sp->so_cred); 98 return get_rpccred(sp->so_cred);
115 } 99 }
116 return NULL; 100 return NULL;
117} 101}
118 102
103static void nfs_alloc_unique_id(struct rb_root *root, struct nfs_unique_id *new,
104 __u64 minval, int maxbits)
105{
106 struct rb_node **p, *parent;
107 struct nfs_unique_id *pos;
108 __u64 mask = ~0ULL;
109
110 if (maxbits < 64)
111 mask = (1ULL << maxbits) - 1ULL;
112
113 /* Ensure distribution is more or less flat */
114 get_random_bytes(&new->id, sizeof(new->id));
115 new->id &= mask;
116 if (new->id < minval)
117 new->id += minval;
118retry:
119 p = &root->rb_node;
120 parent = NULL;
121
122 while (*p != NULL) {
123 parent = *p;
124 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
125
126 if (new->id < pos->id)
127 p = &(*p)->rb_left;
128 else if (new->id > pos->id)
129 p = &(*p)->rb_right;
130 else
131 goto id_exists;
132 }
133 rb_link_node(&new->rb_node, parent, p);
134 rb_insert_color(&new->rb_node, root);
135 return;
136id_exists:
137 for (;;) {
138 new->id++;
139 if (new->id < minval || (new->id & mask) != new->id) {
140 new->id = minval;
141 break;
142 }
143 parent = rb_next(parent);
144 if (parent == NULL)
145 break;
146 pos = rb_entry(parent, struct nfs_unique_id, rb_node);
147 if (new->id < pos->id)
148 break;
149 }
150 goto retry;
151}
152
153static void nfs_free_unique_id(struct rb_root *root, struct nfs_unique_id *id)
154{
155 rb_erase(&id->rb_node, root);
156}
157
119static struct nfs4_state_owner * 158static struct nfs4_state_owner *
120nfs4_find_state_owner(struct nfs_client *clp, struct rpc_cred *cred) 159nfs4_find_state_owner(struct nfs_server *server, struct rpc_cred *cred)
121{ 160{
161 struct nfs_client *clp = server->nfs_client;
162 struct rb_node **p = &clp->cl_state_owners.rb_node,
163 *parent = NULL;
122 struct nfs4_state_owner *sp, *res = NULL; 164 struct nfs4_state_owner *sp, *res = NULL;
123 165
124 list_for_each_entry(sp, &clp->cl_state_owners, so_list) { 166 while (*p != NULL) {
125 if (sp->so_cred != cred) 167 parent = *p;
168 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
169
170 if (server < sp->so_server) {
171 p = &parent->rb_left;
126 continue; 172 continue;
127 atomic_inc(&sp->so_count); 173 }
128 /* Move to the head of the list */ 174 if (server > sp->so_server) {
129 list_move(&sp->so_list, &clp->cl_state_owners); 175 p = &parent->rb_right;
130 res = sp; 176 continue;
131 break; 177 }
178 if (cred < sp->so_cred)
179 p = &parent->rb_left;
180 else if (cred > sp->so_cred)
181 p = &parent->rb_right;
182 else {
183 atomic_inc(&sp->so_count);
184 res = sp;
185 break;
186 }
132 } 187 }
133 return res; 188 return res;
134} 189}
135 190
191static struct nfs4_state_owner *
192nfs4_insert_state_owner(struct nfs_client *clp, struct nfs4_state_owner *new)
193{
194 struct rb_node **p = &clp->cl_state_owners.rb_node,
195 *parent = NULL;
196 struct nfs4_state_owner *sp;
197
198 while (*p != NULL) {
199 parent = *p;
200 sp = rb_entry(parent, struct nfs4_state_owner, so_client_node);
201
202 if (new->so_server < sp->so_server) {
203 p = &parent->rb_left;
204 continue;
205 }
206 if (new->so_server > sp->so_server) {
207 p = &parent->rb_right;
208 continue;
209 }
210 if (new->so_cred < sp->so_cred)
211 p = &parent->rb_left;
212 else if (new->so_cred > sp->so_cred)
213 p = &parent->rb_right;
214 else {
215 atomic_inc(&sp->so_count);
216 return sp;
217 }
218 }
219 nfs_alloc_unique_id(&clp->cl_openowner_id, &new->so_owner_id, 1, 64);
220 rb_link_node(&new->so_client_node, parent, p);
221 rb_insert_color(&new->so_client_node, &clp->cl_state_owners);
222 return new;
223}
224
225static void
226nfs4_remove_state_owner(struct nfs_client *clp, struct nfs4_state_owner *sp)
227{
228 if (!RB_EMPTY_NODE(&sp->so_client_node))
229 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
230 nfs_free_unique_id(&clp->cl_openowner_id, &sp->so_owner_id);
231}
232
136/* 233/*
137 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to 234 * nfs4_alloc_state_owner(): this is called on the OPEN or CREATE path to
138 * create a new state_owner. 235 * create a new state_owner.
@@ -160,10 +257,14 @@ nfs4_alloc_state_owner(void)
160void 257void
161nfs4_drop_state_owner(struct nfs4_state_owner *sp) 258nfs4_drop_state_owner(struct nfs4_state_owner *sp)
162{ 259{
163 struct nfs_client *clp = sp->so_client; 260 if (!RB_EMPTY_NODE(&sp->so_client_node)) {
164 spin_lock(&clp->cl_lock); 261 struct nfs_client *clp = sp->so_client;
165 list_del_init(&sp->so_list); 262
166 spin_unlock(&clp->cl_lock); 263 spin_lock(&clp->cl_lock);
264 rb_erase(&sp->so_client_node, &clp->cl_state_owners);
265 RB_CLEAR_NODE(&sp->so_client_node);
266 spin_unlock(&clp->cl_lock);
267 }
167} 268}
168 269
169/* 270/*
@@ -175,26 +276,25 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct
175 struct nfs_client *clp = server->nfs_client; 276 struct nfs_client *clp = server->nfs_client;
176 struct nfs4_state_owner *sp, *new; 277 struct nfs4_state_owner *sp, *new;
177 278
178 get_rpccred(cred);
179 new = nfs4_alloc_state_owner();
180 spin_lock(&clp->cl_lock); 279 spin_lock(&clp->cl_lock);
181 sp = nfs4_find_state_owner(clp, cred); 280 sp = nfs4_find_state_owner(server, cred);
182 if (sp == NULL)
183 sp = nfs4_client_grab_unused(clp, cred);
184 if (sp == NULL && new != NULL) {
185 list_add(&new->so_list, &clp->cl_state_owners);
186 new->so_client = clp;
187 new->so_id = nfs4_alloc_lockowner_id(clp);
188 new->so_cred = cred;
189 sp = new;
190 new = NULL;
191 }
192 spin_unlock(&clp->cl_lock); 281 spin_unlock(&clp->cl_lock);
193 kfree(new);
194 if (sp != NULL) 282 if (sp != NULL)
195 return sp; 283 return sp;
196 put_rpccred(cred); 284 new = nfs4_alloc_state_owner();
197 return NULL; 285 if (new == NULL)
286 return NULL;
287 new->so_client = clp;
288 new->so_server = server;
289 new->so_cred = cred;
290 spin_lock(&clp->cl_lock);
291 sp = nfs4_insert_state_owner(clp, new);
292 spin_unlock(&clp->cl_lock);
293 if (sp == new)
294 get_rpccred(cred);
295 else
296 kfree(new);
297 return sp;
198} 298}
199 299
200/* 300/*
@@ -208,18 +308,7 @@ void nfs4_put_state_owner(struct nfs4_state_owner *sp)
208 308
209 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) 309 if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
210 return; 310 return;
211 if (clp->cl_nunused >= OPENOWNER_POOL_SIZE) 311 nfs4_remove_state_owner(clp, sp);
212 goto out_free;
213 if (list_empty(&sp->so_list))
214 goto out_free;
215 list_move(&sp->so_list, &clp->cl_unused);
216 clp->cl_nunused++;
217 spin_unlock(&clp->cl_lock);
218 put_rpccred(cred);
219 cred = NULL;
220 return;
221out_free:
222 list_del(&sp->so_list);
223 spin_unlock(&clp->cl_lock); 312 spin_unlock(&clp->cl_lock);
224 put_rpccred(cred); 313 put_rpccred(cred);
225 kfree(sp); 314 kfree(sp);
@@ -236,6 +325,7 @@ nfs4_alloc_open_state(void)
236 atomic_set(&state->count, 1); 325 atomic_set(&state->count, 1);
237 INIT_LIST_HEAD(&state->lock_states); 326 INIT_LIST_HEAD(&state->lock_states);
238 spin_lock_init(&state->state_lock); 327 spin_lock_init(&state->state_lock);
328 seqlock_init(&state->seqlock);
239 return state; 329 return state;
240} 330}
241 331
@@ -263,13 +353,10 @@ __nfs4_find_state_byowner(struct inode *inode, struct nfs4_state_owner *owner)
263 struct nfs4_state *state; 353 struct nfs4_state *state;
264 354
265 list_for_each_entry(state, &nfsi->open_states, inode_states) { 355 list_for_each_entry(state, &nfsi->open_states, inode_states) {
266 /* Is this in the process of being freed? */ 356 if (state->owner != owner)
267 if (state->state == 0)
268 continue; 357 continue;
269 if (state->owner == owner) { 358 if (atomic_inc_not_zero(&state->count))
270 atomic_inc(&state->count);
271 return state; 359 return state;
272 }
273 } 360 }
274 return NULL; 361 return NULL;
275} 362}
@@ -341,16 +428,15 @@ void nfs4_put_open_state(struct nfs4_state *state)
341/* 428/*
342 * Close the current file. 429 * Close the current file.
343 */ 430 */
344void nfs4_close_state(struct nfs4_state *state, mode_t mode) 431void nfs4_close_state(struct path *path, struct nfs4_state *state, mode_t mode)
345{ 432{
346 struct inode *inode = state->inode;
347 struct nfs4_state_owner *owner = state->owner; 433 struct nfs4_state_owner *owner = state->owner;
348 int oldstate, newstate = 0; 434 int call_close = 0;
435 int newstate;
349 436
350 atomic_inc(&owner->so_count); 437 atomic_inc(&owner->so_count);
351 /* Protect against nfs4_find_state() */ 438 /* Protect against nfs4_find_state() */
352 spin_lock(&owner->so_lock); 439 spin_lock(&owner->so_lock);
353 spin_lock(&inode->i_lock);
354 switch (mode & (FMODE_READ | FMODE_WRITE)) { 440 switch (mode & (FMODE_READ | FMODE_WRITE)) {
355 case FMODE_READ: 441 case FMODE_READ:
356 state->n_rdonly--; 442 state->n_rdonly--;
@@ -361,24 +447,29 @@ void nfs4_close_state(struct nfs4_state *state, mode_t mode)
361 case FMODE_READ|FMODE_WRITE: 447 case FMODE_READ|FMODE_WRITE:
362 state->n_rdwr--; 448 state->n_rdwr--;
363 } 449 }
364 oldstate = newstate = state->state; 450 newstate = FMODE_READ|FMODE_WRITE;
365 if (state->n_rdwr == 0) { 451 if (state->n_rdwr == 0) {
366 if (state->n_rdonly == 0) 452 if (state->n_rdonly == 0) {
367 newstate &= ~FMODE_READ; 453 newstate &= ~FMODE_READ;
368 if (state->n_wronly == 0) 454 call_close |= test_bit(NFS_O_RDONLY_STATE, &state->flags);
455 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
456 }
457 if (state->n_wronly == 0) {
369 newstate &= ~FMODE_WRITE; 458 newstate &= ~FMODE_WRITE;
459 call_close |= test_bit(NFS_O_WRONLY_STATE, &state->flags);
460 call_close |= test_bit(NFS_O_RDWR_STATE, &state->flags);
461 }
462 if (newstate == 0)
463 clear_bit(NFS_DELEGATED_STATE, &state->flags);
370 } 464 }
371 if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { 465 nfs4_state_set_mode_locked(state, newstate);
372 nfs4_state_set_mode_locked(state, newstate);
373 oldstate = newstate;
374 }
375 spin_unlock(&inode->i_lock);
376 spin_unlock(&owner->so_lock); 466 spin_unlock(&owner->so_lock);
377 467
378 if (oldstate != newstate && nfs4_do_close(inode, state) == 0) 468 if (!call_close) {
379 return; 469 nfs4_put_open_state(state);
380 nfs4_put_open_state(state); 470 nfs4_put_state_owner(owner);
381 nfs4_put_state_owner(owner); 471 } else
472 nfs4_do_close(path, state);
382} 473}
383 474
384/* 475/*
@@ -415,12 +506,22 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f
415 atomic_set(&lsp->ls_count, 1); 506 atomic_set(&lsp->ls_count, 1);
416 lsp->ls_owner = fl_owner; 507 lsp->ls_owner = fl_owner;
417 spin_lock(&clp->cl_lock); 508 spin_lock(&clp->cl_lock);
418 lsp->ls_id = nfs4_alloc_lockowner_id(clp); 509 nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64);
419 spin_unlock(&clp->cl_lock); 510 spin_unlock(&clp->cl_lock);
420 INIT_LIST_HEAD(&lsp->ls_locks); 511 INIT_LIST_HEAD(&lsp->ls_locks);
421 return lsp; 512 return lsp;
422} 513}
423 514
515static void nfs4_free_lock_state(struct nfs4_lock_state *lsp)
516{
517 struct nfs_client *clp = lsp->ls_state->owner->so_client;
518
519 spin_lock(&clp->cl_lock);
520 nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id);
521 spin_unlock(&clp->cl_lock);
522 kfree(lsp);
523}
524
424/* 525/*
425 * Return a compatible lock_state. If no initialized lock_state structure 526 * Return a compatible lock_state. If no initialized lock_state structure
426 * exists, return an uninitialized one. 527 * exists, return an uninitialized one.
@@ -450,7 +551,8 @@ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_
450 return NULL; 551 return NULL;
451 } 552 }
452 spin_unlock(&state->state_lock); 553 spin_unlock(&state->state_lock);
453 kfree(new); 554 if (new != NULL)
555 nfs4_free_lock_state(new);
454 return lsp; 556 return lsp;
455} 557}
456 558
@@ -471,7 +573,7 @@ void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
471 if (list_empty(&state->lock_states)) 573 if (list_empty(&state->lock_states))
472 clear_bit(LK_STATE_IN_USE, &state->flags); 574 clear_bit(LK_STATE_IN_USE, &state->flags);
473 spin_unlock(&state->state_lock); 575 spin_unlock(&state->state_lock);
474 kfree(lsp); 576 nfs4_free_lock_state(lsp);
475} 577}
476 578
477static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src) 579static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
@@ -513,8 +615,12 @@ int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
513void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner) 615void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
514{ 616{
515 struct nfs4_lock_state *lsp; 617 struct nfs4_lock_state *lsp;
618 int seq;
516 619
517 memcpy(dst, &state->stateid, sizeof(*dst)); 620 do {
621 seq = read_seqbegin(&state->seqlock);
622 memcpy(dst, &state->stateid, sizeof(*dst));
623 } while (read_seqretry(&state->seqlock, seq));
518 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0) 624 if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
519 return; 625 return;
520 626
@@ -557,12 +663,18 @@ void nfs_free_seqid(struct nfs_seqid *seqid)
557 * failed with a seqid incrementing error - 663 * failed with a seqid incrementing error -
558 * see comments nfs_fs.h:seqid_mutating_error() 664 * see comments nfs_fs.h:seqid_mutating_error()
559 */ 665 */
560static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid) 666static void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
561{ 667{
562 switch (status) { 668 switch (status) {
563 case 0: 669 case 0:
564 break; 670 break;
565 case -NFS4ERR_BAD_SEQID: 671 case -NFS4ERR_BAD_SEQID:
672 if (seqid->sequence->flags & NFS_SEQID_CONFIRMED)
673 return;
674 printk(KERN_WARNING "NFS: v4 server returned a bad"
675 "sequence-id error on an"
676 "unconfirmed sequence %p!\n",
677 seqid->sequence);
566 case -NFS4ERR_STALE_CLIENTID: 678 case -NFS4ERR_STALE_CLIENTID:
567 case -NFS4ERR_STALE_STATEID: 679 case -NFS4ERR_STALE_STATEID:
568 case -NFS4ERR_BAD_STATEID: 680 case -NFS4ERR_BAD_STATEID:
@@ -586,7 +698,7 @@ void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
586 struct nfs4_state_owner, so_seqid); 698 struct nfs4_state_owner, so_seqid);
587 nfs4_drop_state_owner(sp); 699 nfs4_drop_state_owner(sp);
588 } 700 }
589 return nfs_increment_seqid(status, seqid); 701 nfs_increment_seqid(status, seqid);
590} 702}
591 703
592/* 704/*
@@ -596,7 +708,7 @@ void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
596 */ 708 */
597void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid) 709void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
598{ 710{
599 return nfs_increment_seqid(status, seqid); 711 nfs_increment_seqid(status, seqid);
600} 712}
601 713
602int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) 714int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
@@ -748,15 +860,21 @@ out_err:
748static void nfs4_state_mark_reclaim(struct nfs_client *clp) 860static void nfs4_state_mark_reclaim(struct nfs_client *clp)
749{ 861{
750 struct nfs4_state_owner *sp; 862 struct nfs4_state_owner *sp;
863 struct rb_node *pos;
751 struct nfs4_state *state; 864 struct nfs4_state *state;
752 struct nfs4_lock_state *lock; 865 struct nfs4_lock_state *lock;
753 866
754 /* Reset all sequence ids to zero */ 867 /* Reset all sequence ids to zero */
755 list_for_each_entry(sp, &clp->cl_state_owners, so_list) { 868 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
869 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
756 sp->so_seqid.counter = 0; 870 sp->so_seqid.counter = 0;
757 sp->so_seqid.flags = 0; 871 sp->so_seqid.flags = 0;
758 spin_lock(&sp->so_lock); 872 spin_lock(&sp->so_lock);
759 list_for_each_entry(state, &sp->so_states, open_states) { 873 list_for_each_entry(state, &sp->so_states, open_states) {
874 clear_bit(NFS_DELEGATED_STATE, &state->flags);
875 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
876 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
877 clear_bit(NFS_O_RDWR_STATE, &state->flags);
760 list_for_each_entry(lock, &state->lock_states, ls_locks) { 878 list_for_each_entry(lock, &state->lock_states, ls_locks) {
761 lock->ls_seqid.counter = 0; 879 lock->ls_seqid.counter = 0;
762 lock->ls_seqid.flags = 0; 880 lock->ls_seqid.flags = 0;
@@ -771,6 +889,7 @@ static int reclaimer(void *ptr)
771{ 889{
772 struct nfs_client *clp = ptr; 890 struct nfs_client *clp = ptr;
773 struct nfs4_state_owner *sp; 891 struct nfs4_state_owner *sp;
892 struct rb_node *pos;
774 struct nfs4_state_recovery_ops *ops; 893 struct nfs4_state_recovery_ops *ops;
775 struct rpc_cred *cred; 894 struct rpc_cred *cred;
776 int status = 0; 895 int status = 0;
@@ -816,7 +935,8 @@ restart_loop:
816 /* Mark all delegations for reclaim */ 935 /* Mark all delegations for reclaim */
817 nfs_delegation_mark_reclaim(clp); 936 nfs_delegation_mark_reclaim(clp);
818 /* Note: list is protected by exclusive lock on cl->cl_sem */ 937 /* Note: list is protected by exclusive lock on cl->cl_sem */
819 list_for_each_entry(sp, &clp->cl_state_owners, so_list) { 938 for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) {
939 sp = rb_entry(pos, struct nfs4_state_owner, so_client_node);
820 status = nfs4_reclaim_open_state(ops, sp); 940 status = nfs4_reclaim_open_state(ops, sp);
821 if (status < 0) { 941 if (status < 0) {
822 if (status == -NFS4ERR_NO_GRACE) { 942 if (status == -NFS4ERR_NO_GRACE) {
diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c
index 8003c91ccb9a..c08738441f73 100644
--- a/fs/nfs/nfs4xdr.c
+++ b/fs/nfs/nfs4xdr.c
@@ -68,9 +68,10 @@ static int nfs4_stat_to_errno(int);
68#endif 68#endif
69 69
70/* lock,open owner id: 70/* lock,open owner id:
71 * we currently use size 1 (u32) out of (NFS4_OPAQUE_LIMIT >> 2) 71 * we currently use size 2 (u64) out of (NFS4_OPAQUE_LIMIT >> 2)
72 */ 72 */
73#define owner_id_maxsz (1 + 1) 73#define open_owner_id_maxsz (1 + 4)
74#define lock_owner_id_maxsz (1 + 4)
74#define compound_encode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2)) 75#define compound_encode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2))
75#define compound_decode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2)) 76#define compound_decode_hdr_maxsz (3 + (NFS4_MAXTAGLEN >> 2))
76#define op_encode_hdr_maxsz (1) 77#define op_encode_hdr_maxsz (1)
@@ -87,9 +88,11 @@ static int nfs4_stat_to_errno(int);
87#define encode_getattr_maxsz (op_encode_hdr_maxsz + nfs4_fattr_bitmap_maxsz) 88#define encode_getattr_maxsz (op_encode_hdr_maxsz + nfs4_fattr_bitmap_maxsz)
88#define nfs4_name_maxsz (1 + ((3 + NFS4_MAXNAMLEN) >> 2)) 89#define nfs4_name_maxsz (1 + ((3 + NFS4_MAXNAMLEN) >> 2))
89#define nfs4_path_maxsz (1 + ((3 + NFS4_MAXPATHLEN) >> 2)) 90#define nfs4_path_maxsz (1 + ((3 + NFS4_MAXPATHLEN) >> 2))
91#define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
92#define nfs4_group_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
90/* This is based on getfattr, which uses the most attributes: */ 93/* This is based on getfattr, which uses the most attributes: */
91#define nfs4_fattr_value_maxsz (1 + (1 + 2 + 2 + 4 + 2 + 1 + 1 + 2 + 2 + \ 94#define nfs4_fattr_value_maxsz (1 + (1 + 2 + 2 + 4 + 2 + 1 + 1 + 2 + 2 + \
92 3 + 3 + 3 + 2 * nfs4_name_maxsz)) 95 3 + 3 + 3 + nfs4_owner_maxsz + nfs4_group_maxsz))
93#define nfs4_fattr_maxsz (nfs4_fattr_bitmap_maxsz + \ 96#define nfs4_fattr_maxsz (nfs4_fattr_bitmap_maxsz + \
94 nfs4_fattr_value_maxsz) 97 nfs4_fattr_value_maxsz)
95#define decode_getattr_maxsz (op_decode_hdr_maxsz + nfs4_fattr_maxsz) 98#define decode_getattr_maxsz (op_decode_hdr_maxsz + nfs4_fattr_maxsz)
@@ -116,8 +119,27 @@ static int nfs4_stat_to_errno(int);
116 3 + (NFS4_VERIFIER_SIZE >> 2)) 119 3 + (NFS4_VERIFIER_SIZE >> 2))
117#define decode_setclientid_confirm_maxsz \ 120#define decode_setclientid_confirm_maxsz \
118 (op_decode_hdr_maxsz) 121 (op_decode_hdr_maxsz)
119#define encode_lookup_maxsz (op_encode_hdr_maxsz + \ 122#define encode_lookup_maxsz (op_encode_hdr_maxsz + nfs4_name_maxsz)
120 1 + ((3 + NFS4_FHSIZE) >> 2)) 123#define decode_lookup_maxsz (op_decode_hdr_maxsz)
124#define encode_share_access_maxsz \
125 (2)
126#define encode_createmode_maxsz (1 + nfs4_fattr_maxsz)
127#define encode_opentype_maxsz (1 + encode_createmode_maxsz)
128#define encode_claim_null_maxsz (1 + nfs4_name_maxsz)
129#define encode_open_maxsz (op_encode_hdr_maxsz + \
130 2 + encode_share_access_maxsz + 2 + \
131 open_owner_id_maxsz + \
132 encode_opentype_maxsz + \
133 encode_claim_null_maxsz)
134#define decode_ace_maxsz (3 + nfs4_owner_maxsz)
135#define decode_delegation_maxsz (1 + XDR_QUADLEN(NFS4_STATEID_SIZE) + 1 + \
136 decode_ace_maxsz)
137#define decode_change_info_maxsz (5)
138#define decode_open_maxsz (op_decode_hdr_maxsz + \
139 XDR_QUADLEN(NFS4_STATEID_SIZE) + \
140 decode_change_info_maxsz + 1 + \
141 nfs4_fattr_bitmap_maxsz + \
142 decode_delegation_maxsz)
121#define encode_remove_maxsz (op_encode_hdr_maxsz + \ 143#define encode_remove_maxsz (op_encode_hdr_maxsz + \
122 nfs4_name_maxsz) 144 nfs4_name_maxsz)
123#define encode_rename_maxsz (op_encode_hdr_maxsz + \ 145#define encode_rename_maxsz (op_encode_hdr_maxsz + \
@@ -134,9 +156,15 @@ static int nfs4_stat_to_errno(int);
134#define encode_create_maxsz (op_encode_hdr_maxsz + \ 156#define encode_create_maxsz (op_encode_hdr_maxsz + \
135 2 + nfs4_name_maxsz + \ 157 2 + nfs4_name_maxsz + \
136 nfs4_fattr_maxsz) 158 nfs4_fattr_maxsz)
137#define decode_create_maxsz (op_decode_hdr_maxsz + 8) 159#define decode_create_maxsz (op_decode_hdr_maxsz + \
160 decode_change_info_maxsz + \
161 nfs4_fattr_bitmap_maxsz)
138#define encode_delegreturn_maxsz (op_encode_hdr_maxsz + 4) 162#define encode_delegreturn_maxsz (op_encode_hdr_maxsz + 4)
139#define decode_delegreturn_maxsz (op_decode_hdr_maxsz) 163#define decode_delegreturn_maxsz (op_decode_hdr_maxsz)
164#define encode_fs_locations_maxsz \
165 (encode_getattr_maxsz)
166#define decode_fs_locations_maxsz \
167 (0)
140#define NFS4_enc_compound_sz (1024) /* XXX: large enough? */ 168#define NFS4_enc_compound_sz (1024) /* XXX: large enough? */
141#define NFS4_dec_compound_sz (1024) /* XXX: large enough? */ 169#define NFS4_dec_compound_sz (1024) /* XXX: large enough? */
142#define NFS4_enc_read_sz (compound_encode_hdr_maxsz + \ 170#define NFS4_enc_read_sz (compound_encode_hdr_maxsz + \
@@ -174,16 +202,21 @@ static int nfs4_stat_to_errno(int);
174 op_decode_hdr_maxsz + 2 + \ 202 op_decode_hdr_maxsz + 2 + \
175 decode_getattr_maxsz) 203 decode_getattr_maxsz)
176#define NFS4_enc_open_sz (compound_encode_hdr_maxsz + \ 204#define NFS4_enc_open_sz (compound_encode_hdr_maxsz + \
177 encode_putfh_maxsz + \ 205 encode_putfh_maxsz + \
178 op_encode_hdr_maxsz + \ 206 encode_savefh_maxsz + \
179 13 + 3 + 2 + 64 + \ 207 encode_open_maxsz + \
180 encode_getattr_maxsz + \ 208 encode_getfh_maxsz + \
181 encode_getfh_maxsz) 209 encode_getattr_maxsz + \
210 encode_restorefh_maxsz + \
211 encode_getattr_maxsz)
182#define NFS4_dec_open_sz (compound_decode_hdr_maxsz + \ 212#define NFS4_dec_open_sz (compound_decode_hdr_maxsz + \
183 decode_putfh_maxsz + \ 213 decode_putfh_maxsz + \
184 op_decode_hdr_maxsz + 4 + 5 + 2 + 3 + \ 214 decode_savefh_maxsz + \
185 decode_getattr_maxsz + \ 215 decode_open_maxsz + \
186 decode_getfh_maxsz) 216 decode_getfh_maxsz + \
217 decode_getattr_maxsz + \
218 decode_restorefh_maxsz + \
219 decode_getattr_maxsz)
187#define NFS4_enc_open_confirm_sz \ 220#define NFS4_enc_open_confirm_sz \
188 (compound_encode_hdr_maxsz + \ 221 (compound_encode_hdr_maxsz + \
189 encode_putfh_maxsz + \ 222 encode_putfh_maxsz + \
@@ -193,12 +226,12 @@ static int nfs4_stat_to_errno(int);
193 op_decode_hdr_maxsz + 4) 226 op_decode_hdr_maxsz + 4)
194#define NFS4_enc_open_noattr_sz (compound_encode_hdr_maxsz + \ 227#define NFS4_enc_open_noattr_sz (compound_encode_hdr_maxsz + \
195 encode_putfh_maxsz + \ 228 encode_putfh_maxsz + \
196 op_encode_hdr_maxsz + \ 229 encode_open_maxsz + \
197 11) 230 encode_getattr_maxsz)
198#define NFS4_dec_open_noattr_sz (compound_decode_hdr_maxsz + \ 231#define NFS4_dec_open_noattr_sz (compound_decode_hdr_maxsz + \
199 decode_putfh_maxsz + \ 232 decode_putfh_maxsz + \
200 op_decode_hdr_maxsz + \ 233 decode_open_maxsz + \
201 4 + 5 + 2 + 3) 234 decode_getattr_maxsz)
202#define NFS4_enc_open_downgrade_sz \ 235#define NFS4_enc_open_downgrade_sz \
203 (compound_encode_hdr_maxsz + \ 236 (compound_encode_hdr_maxsz + \
204 encode_putfh_maxsz + \ 237 encode_putfh_maxsz + \
@@ -256,19 +289,19 @@ static int nfs4_stat_to_errno(int);
256 op_encode_hdr_maxsz + \ 289 op_encode_hdr_maxsz + \
257 1 + 1 + 2 + 2 + \ 290 1 + 1 + 2 + 2 + \
258 1 + 4 + 1 + 2 + \ 291 1 + 4 + 1 + 2 + \
259 owner_id_maxsz) 292 lock_owner_id_maxsz)
260#define NFS4_dec_lock_sz (compound_decode_hdr_maxsz + \ 293#define NFS4_dec_lock_sz (compound_decode_hdr_maxsz + \
261 decode_putfh_maxsz + \ 294 decode_putfh_maxsz + \
262 decode_getattr_maxsz + \ 295 decode_getattr_maxsz + \
263 op_decode_hdr_maxsz + \ 296 op_decode_hdr_maxsz + \
264 2 + 2 + 1 + 2 + \ 297 2 + 2 + 1 + 2 + \
265 owner_id_maxsz) 298 lock_owner_id_maxsz)
266#define NFS4_enc_lockt_sz (compound_encode_hdr_maxsz + \ 299#define NFS4_enc_lockt_sz (compound_encode_hdr_maxsz + \
267 encode_putfh_maxsz + \ 300 encode_putfh_maxsz + \
268 encode_getattr_maxsz + \ 301 encode_getattr_maxsz + \
269 op_encode_hdr_maxsz + \ 302 op_encode_hdr_maxsz + \
270 1 + 2 + 2 + 2 + \ 303 1 + 2 + 2 + 2 + \
271 owner_id_maxsz) 304 lock_owner_id_maxsz)
272#define NFS4_dec_lockt_sz (NFS4_dec_lock_sz) 305#define NFS4_dec_lockt_sz (NFS4_dec_lock_sz)
273#define NFS4_enc_locku_sz (compound_encode_hdr_maxsz + \ 306#define NFS4_enc_locku_sz (compound_encode_hdr_maxsz + \
274 encode_putfh_maxsz + \ 307 encode_putfh_maxsz + \
@@ -298,7 +331,7 @@ static int nfs4_stat_to_errno(int);
298 encode_getfh_maxsz) 331 encode_getfh_maxsz)
299#define NFS4_dec_lookup_sz (compound_decode_hdr_maxsz + \ 332#define NFS4_dec_lookup_sz (compound_decode_hdr_maxsz + \
300 decode_putfh_maxsz + \ 333 decode_putfh_maxsz + \
301 op_decode_hdr_maxsz + \ 334 decode_lookup_maxsz + \
302 decode_getattr_maxsz + \ 335 decode_getattr_maxsz + \
303 decode_getfh_maxsz) 336 decode_getfh_maxsz)
304#define NFS4_enc_lookup_root_sz (compound_encode_hdr_maxsz + \ 337#define NFS4_enc_lookup_root_sz (compound_encode_hdr_maxsz + \
@@ -417,12 +450,13 @@ static int nfs4_stat_to_errno(int);
417#define NFS4_enc_fs_locations_sz \ 450#define NFS4_enc_fs_locations_sz \
418 (compound_encode_hdr_maxsz + \ 451 (compound_encode_hdr_maxsz + \
419 encode_putfh_maxsz + \ 452 encode_putfh_maxsz + \
420 encode_getattr_maxsz) 453 encode_lookup_maxsz + \
454 encode_fs_locations_maxsz)
421#define NFS4_dec_fs_locations_sz \ 455#define NFS4_dec_fs_locations_sz \
422 (compound_decode_hdr_maxsz + \ 456 (compound_decode_hdr_maxsz + \
423 decode_putfh_maxsz + \ 457 decode_putfh_maxsz + \
424 op_decode_hdr_maxsz + \ 458 decode_lookup_maxsz + \
425 nfs4_fattr_bitmap_maxsz) 459 decode_fs_locations_maxsz)
426 460
427static struct { 461static struct {
428 unsigned int mode; 462 unsigned int mode;
@@ -793,13 +827,14 @@ static int encode_lock(struct xdr_stream *xdr, const struct nfs_lock_args *args)
793 WRITE64(nfs4_lock_length(args->fl)); 827 WRITE64(nfs4_lock_length(args->fl));
794 WRITE32(args->new_lock_owner); 828 WRITE32(args->new_lock_owner);
795 if (args->new_lock_owner){ 829 if (args->new_lock_owner){
796 RESERVE_SPACE(4+NFS4_STATEID_SIZE+20); 830 RESERVE_SPACE(4+NFS4_STATEID_SIZE+32);
797 WRITE32(args->open_seqid->sequence->counter); 831 WRITE32(args->open_seqid->sequence->counter);
798 WRITEMEM(args->open_stateid->data, NFS4_STATEID_SIZE); 832 WRITEMEM(args->open_stateid->data, NFS4_STATEID_SIZE);
799 WRITE32(args->lock_seqid->sequence->counter); 833 WRITE32(args->lock_seqid->sequence->counter);
800 WRITE64(args->lock_owner.clientid); 834 WRITE64(args->lock_owner.clientid);
801 WRITE32(4); 835 WRITE32(16);
802 WRITE32(args->lock_owner.id); 836 WRITEMEM("lock id:", 8);
837 WRITE64(args->lock_owner.id);
803 } 838 }
804 else { 839 else {
805 RESERVE_SPACE(NFS4_STATEID_SIZE+4); 840 RESERVE_SPACE(NFS4_STATEID_SIZE+4);
@@ -814,14 +849,15 @@ static int encode_lockt(struct xdr_stream *xdr, const struct nfs_lockt_args *arg
814{ 849{
815 __be32 *p; 850 __be32 *p;
816 851
817 RESERVE_SPACE(40); 852 RESERVE_SPACE(52);
818 WRITE32(OP_LOCKT); 853 WRITE32(OP_LOCKT);
819 WRITE32(nfs4_lock_type(args->fl, 0)); 854 WRITE32(nfs4_lock_type(args->fl, 0));
820 WRITE64(args->fl->fl_start); 855 WRITE64(args->fl->fl_start);
821 WRITE64(nfs4_lock_length(args->fl)); 856 WRITE64(nfs4_lock_length(args->fl));
822 WRITE64(args->lock_owner.clientid); 857 WRITE64(args->lock_owner.clientid);
823 WRITE32(4); 858 WRITE32(16);
824 WRITE32(args->lock_owner.id); 859 WRITEMEM("lock id:", 8);
860 WRITE64(args->lock_owner.id);
825 861
826 return 0; 862 return 0;
827} 863}
@@ -886,10 +922,11 @@ static inline void encode_openhdr(struct xdr_stream *xdr, const struct nfs_opena
886 WRITE32(OP_OPEN); 922 WRITE32(OP_OPEN);
887 WRITE32(arg->seqid->sequence->counter); 923 WRITE32(arg->seqid->sequence->counter);
888 encode_share_access(xdr, arg->open_flags); 924 encode_share_access(xdr, arg->open_flags);
889 RESERVE_SPACE(16); 925 RESERVE_SPACE(28);
890 WRITE64(arg->clientid); 926 WRITE64(arg->clientid);
891 WRITE32(4); 927 WRITE32(16);
892 WRITE32(arg->id); 928 WRITEMEM("open id:", 8);
929 WRITE64(arg->id);
893} 930}
894 931
895static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_openargs *arg) 932static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_openargs *arg)
@@ -1071,7 +1108,7 @@ static int encode_read(struct xdr_stream *xdr, const struct nfs_readargs *args)
1071 1108
1072static int encode_readdir(struct xdr_stream *xdr, const struct nfs4_readdir_arg *readdir, struct rpc_rqst *req) 1109static int encode_readdir(struct xdr_stream *xdr, const struct nfs4_readdir_arg *readdir, struct rpc_rqst *req)
1073{ 1110{
1074 struct rpc_auth *auth = req->rq_task->tk_auth; 1111 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
1075 uint32_t attrs[2] = { 1112 uint32_t attrs[2] = {
1076 FATTR4_WORD0_RDATTR_ERROR|FATTR4_WORD0_FILEID, 1113 FATTR4_WORD0_RDATTR_ERROR|FATTR4_WORD0_FILEID,
1077 FATTR4_WORD1_MOUNTED_ON_FILEID, 1114 FATTR4_WORD1_MOUNTED_ON_FILEID,
@@ -1117,7 +1154,7 @@ static int encode_readdir(struct xdr_stream *xdr, const struct nfs4_readdir_arg
1117 1154
1118static int encode_readlink(struct xdr_stream *xdr, const struct nfs4_readlink *readlink, struct rpc_rqst *req) 1155static int encode_readlink(struct xdr_stream *xdr, const struct nfs4_readlink *readlink, struct rpc_rqst *req)
1119{ 1156{
1120 struct rpc_auth *auth = req->rq_task->tk_auth; 1157 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
1121 unsigned int replen; 1158 unsigned int replen;
1122 __be32 *p; 1159 __be32 *p;
1123 1160
@@ -1735,7 +1772,7 @@ out:
1735 */ 1772 */
1736static int nfs4_xdr_enc_read(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) 1773static int nfs4_xdr_enc_read(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args)
1737{ 1774{
1738 struct rpc_auth *auth = req->rq_task->tk_auth; 1775 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
1739 struct xdr_stream xdr; 1776 struct xdr_stream xdr;
1740 struct compound_hdr hdr = { 1777 struct compound_hdr hdr = {
1741 .nops = 2, 1778 .nops = 2,
@@ -1795,7 +1832,7 @@ nfs4_xdr_enc_getacl(struct rpc_rqst *req, __be32 *p,
1795 struct nfs_getaclargs *args) 1832 struct nfs_getaclargs *args)
1796{ 1833{
1797 struct xdr_stream xdr; 1834 struct xdr_stream xdr;
1798 struct rpc_auth *auth = req->rq_task->tk_auth; 1835 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
1799 struct compound_hdr hdr = { 1836 struct compound_hdr hdr = {
1800 .nops = 2, 1837 .nops = 2,
1801 }; 1838 };
@@ -2030,7 +2067,7 @@ static int nfs4_xdr_enc_fs_locations(struct rpc_rqst *req, __be32 *p, struct nfs
2030 struct compound_hdr hdr = { 2067 struct compound_hdr hdr = {
2031 .nops = 3, 2068 .nops = 3,
2032 }; 2069 };
2033 struct rpc_auth *auth = req->rq_task->tk_auth; 2070 struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth;
2034 int replen; 2071 int replen;
2035 int status; 2072 int status;
2036 2073
@@ -3269,7 +3306,7 @@ static int decode_delegation(struct xdr_stream *xdr, struct nfs_openres *res)
3269static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res) 3306static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res)
3270{ 3307{
3271 __be32 *p; 3308 __be32 *p;
3272 uint32_t bmlen; 3309 uint32_t savewords, bmlen, i;
3273 int status; 3310 int status;
3274 3311
3275 status = decode_op_hdr(xdr, OP_OPEN); 3312 status = decode_op_hdr(xdr, OP_OPEN);
@@ -3287,7 +3324,12 @@ static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res)
3287 goto xdr_error; 3324 goto xdr_error;
3288 3325
3289 READ_BUF(bmlen << 2); 3326 READ_BUF(bmlen << 2);
3290 p += bmlen; 3327 savewords = min_t(uint32_t, bmlen, NFS4_BITMAP_SIZE);
3328 for (i = 0; i < savewords; ++i)
3329 READ32(res->attrset[i]);
3330 for (; i < NFS4_BITMAP_SIZE; i++)
3331 res->attrset[i] = 0;
3332
3291 return decode_delegation(xdr, res); 3333 return decode_delegation(xdr, res);
3292xdr_error: 3334xdr_error:
3293 dprintk("%s: Bitmap too large! Length = %u\n", __FUNCTION__, bmlen); 3335 dprintk("%s: Bitmap too large! Length = %u\n", __FUNCTION__, bmlen);
diff --git a/fs/nfs/nfsroot.c b/fs/nfs/nfsroot.c
index 49d1008ce1d7..3490322d1145 100644
--- a/fs/nfs/nfsroot.c
+++ b/fs/nfs/nfsroot.c
@@ -428,7 +428,7 @@ static int __init root_nfs_getport(int program, int version, int proto)
428 printk(KERN_NOTICE "Looking up port of RPC %d/%d on %u.%u.%u.%u\n", 428 printk(KERN_NOTICE "Looking up port of RPC %d/%d on %u.%u.%u.%u\n",
429 program, version, NIPQUAD(servaddr)); 429 program, version, NIPQUAD(servaddr));
430 set_sockaddr(&sin, servaddr, 0); 430 set_sockaddr(&sin, servaddr, 0);
431 return rpcb_getport_external(&sin, program, version, proto); 431 return rpcb_getport_sync(&sin, program, version, proto);
432} 432}
433 433
434 434
@@ -496,7 +496,8 @@ static int __init root_nfs_get_handle(void)
496 NFS_MNT3_VERSION : NFS_MNT_VERSION; 496 NFS_MNT3_VERSION : NFS_MNT_VERSION;
497 497
498 set_sockaddr(&sin, servaddr, htons(mount_port)); 498 set_sockaddr(&sin, servaddr, htons(mount_port));
499 status = nfsroot_mount(&sin, nfs_path, &fh, version, protocol); 499 status = nfs_mount((struct sockaddr *) &sin, sizeof(sin), NULL,
500 nfs_path, version, protocol, &fh);
500 if (status < 0) 501 if (status < 0)
501 printk(KERN_ERR "Root-NFS: Server returned error %d " 502 printk(KERN_ERR "Root-NFS: Server returned error %d "
502 "while mounting %s\n", status, nfs_path); 503 "while mounting %s\n", status, nfs_path);
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index c5bb51a29e80..f56dae5216f4 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -85,9 +85,8 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
85 req->wb_offset = offset; 85 req->wb_offset = offset;
86 req->wb_pgbase = offset; 86 req->wb_pgbase = offset;
87 req->wb_bytes = count; 87 req->wb_bytes = count;
88 atomic_set(&req->wb_count, 1);
89 req->wb_context = get_nfs_open_context(ctx); 88 req->wb_context = get_nfs_open_context(ctx);
90 89 kref_init(&req->wb_kref);
91 return req; 90 return req;
92} 91}
93 92
@@ -109,30 +108,31 @@ void nfs_unlock_request(struct nfs_page *req)
109} 108}
110 109
111/** 110/**
112 * nfs_set_page_writeback_locked - Lock a request for writeback 111 * nfs_set_page_tag_locked - Tag a request as locked
113 * @req: 112 * @req:
114 */ 113 */
115int nfs_set_page_writeback_locked(struct nfs_page *req) 114static int nfs_set_page_tag_locked(struct nfs_page *req)
116{ 115{
117 struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode); 116 struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode);
118 117
119 if (!nfs_lock_request(req)) 118 if (!nfs_lock_request(req))
120 return 0; 119 return 0;
121 radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK); 120 radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
122 return 1; 121 return 1;
123} 122}
124 123
125/** 124/**
126 * nfs_clear_page_writeback - Unlock request and wake up sleepers 125 * nfs_clear_page_tag_locked - Clear request tag and wake up sleepers
127 */ 126 */
128void nfs_clear_page_writeback(struct nfs_page *req) 127void nfs_clear_page_tag_locked(struct nfs_page *req)
129{ 128{
130 struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode); 129 struct inode *inode = req->wb_context->path.dentry->d_inode;
130 struct nfs_inode *nfsi = NFS_I(inode);
131 131
132 if (req->wb_page != NULL) { 132 if (req->wb_page != NULL) {
133 spin_lock(&nfsi->req_lock); 133 spin_lock(&inode->i_lock);
134 radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK); 134 radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
135 spin_unlock(&nfsi->req_lock); 135 spin_unlock(&inode->i_lock);
136 } 136 }
137 nfs_unlock_request(req); 137 nfs_unlock_request(req);
138} 138}
@@ -160,11 +160,9 @@ void nfs_clear_request(struct nfs_page *req)
160 * 160 *
161 * Note: Should never be called with the spinlock held! 161 * Note: Should never be called with the spinlock held!
162 */ 162 */
163void 163static void nfs_free_request(struct kref *kref)
164nfs_release_request(struct nfs_page *req)
165{ 164{
166 if (!atomic_dec_and_test(&req->wb_count)) 165 struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
167 return;
168 166
169 /* Release struct file or cached credential */ 167 /* Release struct file or cached credential */
170 nfs_clear_request(req); 168 nfs_clear_request(req);
@@ -172,6 +170,11 @@ nfs_release_request(struct nfs_page *req)
172 nfs_page_free(req); 170 nfs_page_free(req);
173} 171}
174 172
173void nfs_release_request(struct nfs_page *req)
174{
175 kref_put(&req->wb_kref, nfs_free_request);
176}
177
175static int nfs_wait_bit_interruptible(void *word) 178static int nfs_wait_bit_interruptible(void *word)
176{ 179{
177 int ret = 0; 180 int ret = 0;
@@ -193,7 +196,7 @@ static int nfs_wait_bit_interruptible(void *word)
193int 196int
194nfs_wait_on_request(struct nfs_page *req) 197nfs_wait_on_request(struct nfs_page *req)
195{ 198{
196 struct rpc_clnt *clnt = NFS_CLIENT(req->wb_context->dentry->d_inode); 199 struct rpc_clnt *clnt = NFS_CLIENT(req->wb_context->path.dentry->d_inode);
197 sigset_t oldmask; 200 sigset_t oldmask;
198 int ret = 0; 201 int ret = 0;
199 202
@@ -379,20 +382,20 @@ void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index)
379/** 382/**
380 * nfs_scan_list - Scan a list for matching requests 383 * nfs_scan_list - Scan a list for matching requests
381 * @nfsi: NFS inode 384 * @nfsi: NFS inode
382 * @head: One of the NFS inode request lists
383 * @dst: Destination list 385 * @dst: Destination list
384 * @idx_start: lower bound of page->index to scan 386 * @idx_start: lower bound of page->index to scan
385 * @npages: idx_start + npages sets the upper bound to scan. 387 * @npages: idx_start + npages sets the upper bound to scan.
388 * @tag: tag to scan for
386 * 389 *
387 * Moves elements from one of the inode request lists. 390 * Moves elements from one of the inode request lists.
388 * If the number of requests is set to 0, the entire address_space 391 * If the number of requests is set to 0, the entire address_space
389 * starting at index idx_start, is scanned. 392 * starting at index idx_start, is scanned.
390 * The requests are *not* checked to ensure that they form a contiguous set. 393 * The requests are *not* checked to ensure that they form a contiguous set.
391 * You must be holding the inode's req_lock when calling this function 394 * You must be holding the inode's i_lock when calling this function
392 */ 395 */
393int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, 396int nfs_scan_list(struct nfs_inode *nfsi,
394 struct list_head *dst, pgoff_t idx_start, 397 struct list_head *dst, pgoff_t idx_start,
395 unsigned int npages) 398 unsigned int npages, int tag)
396{ 399{
397 struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES]; 400 struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES];
398 struct nfs_page *req; 401 struct nfs_page *req;
@@ -407,9 +410,9 @@ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head,
407 idx_end = idx_start + npages - 1; 410 idx_end = idx_start + npages - 1;
408 411
409 for (;;) { 412 for (;;) {
410 found = radix_tree_gang_lookup(&nfsi->nfs_page_tree, 413 found = radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree,
411 (void **)&pgvec[0], idx_start, 414 (void **)&pgvec[0], idx_start,
412 NFS_SCAN_MAXENTRIES); 415 NFS_SCAN_MAXENTRIES, tag);
413 if (found <= 0) 416 if (found <= 0)
414 break; 417 break;
415 for (i = 0; i < found; i++) { 418 for (i = 0; i < found; i++) {
@@ -417,15 +420,18 @@ int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head,
417 if (req->wb_index > idx_end) 420 if (req->wb_index > idx_end)
418 goto out; 421 goto out;
419 idx_start = req->wb_index + 1; 422 idx_start = req->wb_index + 1;
420 if (req->wb_list_head != head) 423 if (nfs_set_page_tag_locked(req)) {
421 continue;
422 if (nfs_set_page_writeback_locked(req)) {
423 nfs_list_remove_request(req); 424 nfs_list_remove_request(req);
425 radix_tree_tag_clear(&nfsi->nfs_page_tree,
426 req->wb_index, tag);
424 nfs_list_add_request(req, dst); 427 nfs_list_add_request(req, dst);
425 res++; 428 res++;
429 if (res == INT_MAX)
430 goto out;
426 } 431 }
427 } 432 }
428 433 /* for latency reduction */
434 cond_resched_lock(&nfsi->vfs_inode.i_lock);
429 } 435 }
430out: 436out:
431 return res; 437 return res;
diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index 7bd7cb95c034..6ae2e58ed05a 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -145,8 +145,8 @@ static void nfs_readpage_release(struct nfs_page *req)
145 unlock_page(req->wb_page); 145 unlock_page(req->wb_page);
146 146
147 dprintk("NFS: read done (%s/%Ld %d@%Ld)\n", 147 dprintk("NFS: read done (%s/%Ld %d@%Ld)\n",
148 req->wb_context->dentry->d_inode->i_sb->s_id, 148 req->wb_context->path.dentry->d_inode->i_sb->s_id,
149 (long long)NFS_FILEID(req->wb_context->dentry->d_inode), 149 (long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
150 req->wb_bytes, 150 req->wb_bytes,
151 (long long)req_offset(req)); 151 (long long)req_offset(req));
152 nfs_clear_request(req); 152 nfs_clear_request(req);
@@ -164,7 +164,7 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
164 int flags; 164 int flags;
165 165
166 data->req = req; 166 data->req = req;
167 data->inode = inode = req->wb_context->dentry->d_inode; 167 data->inode = inode = req->wb_context->path.dentry->d_inode;
168 data->cred = req->wb_context->cred; 168 data->cred = req->wb_context->cred;
169 169
170 data->args.fh = NFS_FH(inode); 170 data->args.fh = NFS_FH(inode);
@@ -483,17 +483,19 @@ int nfs_readpage(struct file *file, struct page *page)
483 */ 483 */
484 error = nfs_wb_page(inode, page); 484 error = nfs_wb_page(inode, page);
485 if (error) 485 if (error)
486 goto out_error; 486 goto out_unlock;
487 if (PageUptodate(page))
488 goto out_unlock;
487 489
488 error = -ESTALE; 490 error = -ESTALE;
489 if (NFS_STALE(inode)) 491 if (NFS_STALE(inode))
490 goto out_error; 492 goto out_unlock;
491 493
492 if (file == NULL) { 494 if (file == NULL) {
493 error = -EBADF; 495 error = -EBADF;
494 ctx = nfs_find_open_context(inode, NULL, FMODE_READ); 496 ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
495 if (ctx == NULL) 497 if (ctx == NULL)
496 goto out_error; 498 goto out_unlock;
497 } else 499 } else
498 ctx = get_nfs_open_context((struct nfs_open_context *) 500 ctx = get_nfs_open_context((struct nfs_open_context *)
499 file->private_data); 501 file->private_data);
@@ -502,8 +504,7 @@ int nfs_readpage(struct file *file, struct page *page)
502 504
503 put_nfs_open_context(ctx); 505 put_nfs_open_context(ctx);
504 return error; 506 return error;
505 507out_unlock:
506out_error:
507 unlock_page(page); 508 unlock_page(page);
508 return error; 509 return error;
509} 510}
@@ -520,21 +521,32 @@ readpage_async_filler(void *data, struct page *page)
520 struct inode *inode = page->mapping->host; 521 struct inode *inode = page->mapping->host;
521 struct nfs_page *new; 522 struct nfs_page *new;
522 unsigned int len; 523 unsigned int len;
524 int error;
525
526 error = nfs_wb_page(inode, page);
527 if (error)
528 goto out_unlock;
529 if (PageUptodate(page))
530 goto out_unlock;
523 531
524 nfs_wb_page(inode, page);
525 len = nfs_page_length(page); 532 len = nfs_page_length(page);
526 if (len == 0) 533 if (len == 0)
527 return nfs_return_empty_page(page); 534 return nfs_return_empty_page(page);
535
528 new = nfs_create_request(desc->ctx, inode, page, 0, len); 536 new = nfs_create_request(desc->ctx, inode, page, 0, len);
529 if (IS_ERR(new)) { 537 if (IS_ERR(new))
530 SetPageError(page); 538 goto out_error;
531 unlock_page(page); 539
532 return PTR_ERR(new);
533 }
534 if (len < PAGE_CACHE_SIZE) 540 if (len < PAGE_CACHE_SIZE)
535 zero_user_page(page, len, PAGE_CACHE_SIZE - len, KM_USER0); 541 zero_user_page(page, len, PAGE_CACHE_SIZE - len, KM_USER0);
536 nfs_pageio_add_request(desc->pgio, new); 542 nfs_pageio_add_request(desc->pgio, new);
537 return 0; 543 return 0;
544out_error:
545 error = PTR_ERR(new);
546 SetPageError(page);
547out_unlock:
548 unlock_page(page);
549 return error;
538} 550}
539 551
540int nfs_readpages(struct file *filp, struct address_space *mapping, 552int nfs_readpages(struct file *filp, struct address_space *mapping,
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index ca20d3cc2609..a2b1af89ca1a 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -45,6 +45,7 @@
45#include <linux/inet.h> 45#include <linux/inet.h>
46#include <linux/nfs_xdr.h> 46#include <linux/nfs_xdr.h>
47#include <linux/magic.h> 47#include <linux/magic.h>
48#include <linux/parser.h>
48 49
49#include <asm/system.h> 50#include <asm/system.h>
50#include <asm/uaccess.h> 51#include <asm/uaccess.h>
@@ -57,6 +58,167 @@
57 58
58#define NFSDBG_FACILITY NFSDBG_VFS 59#define NFSDBG_FACILITY NFSDBG_VFS
59 60
61
62struct nfs_parsed_mount_data {
63 int flags;
64 int rsize, wsize;
65 int timeo, retrans;
66 int acregmin, acregmax,
67 acdirmin, acdirmax;
68 int namlen;
69 unsigned int bsize;
70 unsigned int auth_flavor_len;
71 rpc_authflavor_t auth_flavors[1];
72 char *client_address;
73
74 struct {
75 struct sockaddr_in address;
76 unsigned int program;
77 unsigned int version;
78 unsigned short port;
79 int protocol;
80 } mount_server;
81
82 struct {
83 struct sockaddr_in address;
84 char *hostname;
85 char *export_path;
86 unsigned int program;
87 int protocol;
88 } nfs_server;
89};
90
91enum {
92 /* Mount options that take no arguments */
93 Opt_soft, Opt_hard,
94 Opt_intr, Opt_nointr,
95 Opt_posix, Opt_noposix,
96 Opt_cto, Opt_nocto,
97 Opt_ac, Opt_noac,
98 Opt_lock, Opt_nolock,
99 Opt_v2, Opt_v3,
100 Opt_udp, Opt_tcp,
101 Opt_acl, Opt_noacl,
102 Opt_rdirplus, Opt_nordirplus,
103 Opt_sharecache, Opt_nosharecache,
104
105 /* Mount options that take integer arguments */
106 Opt_port,
107 Opt_rsize, Opt_wsize, Opt_bsize,
108 Opt_timeo, Opt_retrans,
109 Opt_acregmin, Opt_acregmax,
110 Opt_acdirmin, Opt_acdirmax,
111 Opt_actimeo,
112 Opt_namelen,
113 Opt_mountport,
114 Opt_mountprog, Opt_mountvers,
115 Opt_nfsprog, Opt_nfsvers,
116
117 /* Mount options that take string arguments */
118 Opt_sec, Opt_proto, Opt_mountproto,
119 Opt_addr, Opt_mounthost, Opt_clientaddr,
120
121 /* Mount options that are ignored */
122 Opt_userspace, Opt_deprecated,
123
124 Opt_err
125};
126
127static match_table_t nfs_mount_option_tokens = {
128 { Opt_userspace, "bg" },
129 { Opt_userspace, "fg" },
130 { Opt_soft, "soft" },
131 { Opt_hard, "hard" },
132 { Opt_intr, "intr" },
133 { Opt_nointr, "nointr" },
134 { Opt_posix, "posix" },
135 { Opt_noposix, "noposix" },
136 { Opt_cto, "cto" },
137 { Opt_nocto, "nocto" },
138 { Opt_ac, "ac" },
139 { Opt_noac, "noac" },
140 { Opt_lock, "lock" },
141 { Opt_nolock, "nolock" },
142 { Opt_v2, "v2" },
143 { Opt_v3, "v3" },
144 { Opt_udp, "udp" },
145 { Opt_tcp, "tcp" },
146 { Opt_acl, "acl" },
147 { Opt_noacl, "noacl" },
148 { Opt_rdirplus, "rdirplus" },
149 { Opt_nordirplus, "nordirplus" },
150 { Opt_sharecache, "sharecache" },
151 { Opt_nosharecache, "nosharecache" },
152
153 { Opt_port, "port=%u" },
154 { Opt_rsize, "rsize=%u" },
155 { Opt_wsize, "wsize=%u" },
156 { Opt_bsize, "bsize=%u" },
157 { Opt_timeo, "timeo=%u" },
158 { Opt_retrans, "retrans=%u" },
159 { Opt_acregmin, "acregmin=%u" },
160 { Opt_acregmax, "acregmax=%u" },
161 { Opt_acdirmin, "acdirmin=%u" },
162 { Opt_acdirmax, "acdirmax=%u" },
163 { Opt_actimeo, "actimeo=%u" },
164 { Opt_userspace, "retry=%u" },
165 { Opt_namelen, "namlen=%u" },
166 { Opt_mountport, "mountport=%u" },
167 { Opt_mountprog, "mountprog=%u" },
168 { Opt_mountvers, "mountvers=%u" },
169 { Opt_nfsprog, "nfsprog=%u" },
170 { Opt_nfsvers, "nfsvers=%u" },
171 { Opt_nfsvers, "vers=%u" },
172
173 { Opt_sec, "sec=%s" },
174 { Opt_proto, "proto=%s" },
175 { Opt_mountproto, "mountproto=%s" },
176 { Opt_addr, "addr=%s" },
177 { Opt_clientaddr, "clientaddr=%s" },
178 { Opt_mounthost, "mounthost=%s" },
179
180 { Opt_err, NULL }
181};
182
183enum {
184 Opt_xprt_udp, Opt_xprt_tcp,
185
186 Opt_xprt_err
187};
188
189static match_table_t nfs_xprt_protocol_tokens = {
190 { Opt_xprt_udp, "udp" },
191 { Opt_xprt_tcp, "tcp" },
192
193 { Opt_xprt_err, NULL }
194};
195
196enum {
197 Opt_sec_none, Opt_sec_sys,
198 Opt_sec_krb5, Opt_sec_krb5i, Opt_sec_krb5p,
199 Opt_sec_lkey, Opt_sec_lkeyi, Opt_sec_lkeyp,
200 Opt_sec_spkm, Opt_sec_spkmi, Opt_sec_spkmp,
201
202 Opt_sec_err
203};
204
205static match_table_t nfs_secflavor_tokens = {
206 { Opt_sec_none, "none" },
207 { Opt_sec_none, "null" },
208 { Opt_sec_sys, "sys" },
209
210 { Opt_sec_krb5, "krb5" },
211 { Opt_sec_krb5i, "krb5i" },
212 { Opt_sec_krb5p, "krb5p" },
213
214 { Opt_sec_lkey, "lkey" },
215 { Opt_sec_lkeyi, "lkeyi" },
216 { Opt_sec_lkeyp, "lkeyp" },
217
218 { Opt_sec_err, NULL }
219};
220
221
60static void nfs_umount_begin(struct vfsmount *, int); 222static void nfs_umount_begin(struct vfsmount *, int);
61static int nfs_statfs(struct dentry *, struct kstatfs *); 223static int nfs_statfs(struct dentry *, struct kstatfs *);
62static int nfs_show_options(struct seq_file *, struct vfsmount *); 224static int nfs_show_options(struct seq_file *, struct vfsmount *);
@@ -263,11 +425,11 @@ static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
263 { RPC_AUTH_GSS_SPKM, "spkm" }, 425 { RPC_AUTH_GSS_SPKM, "spkm" },
264 { RPC_AUTH_GSS_SPKMI, "spkmi" }, 426 { RPC_AUTH_GSS_SPKMI, "spkmi" },
265 { RPC_AUTH_GSS_SPKMP, "spkmp" }, 427 { RPC_AUTH_GSS_SPKMP, "spkmp" },
266 { -1, "unknown" } 428 { UINT_MAX, "unknown" }
267 }; 429 };
268 int i; 430 int i;
269 431
270 for (i=0; sec_flavours[i].flavour != -1; i++) { 432 for (i = 0; sec_flavours[i].flavour != UINT_MAX; i++) {
271 if (sec_flavours[i].flavour == flavour) 433 if (sec_flavours[i].flavour == flavour)
272 break; 434 break;
273 } 435 }
@@ -291,6 +453,7 @@ static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss,
291 { NFS_MOUNT_NONLM, ",nolock", "" }, 453 { NFS_MOUNT_NONLM, ",nolock", "" },
292 { NFS_MOUNT_NOACL, ",noacl", "" }, 454 { NFS_MOUNT_NOACL, ",noacl", "" },
293 { NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" }, 455 { NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" },
456 { NFS_MOUNT_UNSHARED, ",nosharecache", ""},
294 { 0, NULL, NULL } 457 { 0, NULL, NULL }
295 }; 458 };
296 const struct proc_nfs_info *nfs_infop; 459 const struct proc_nfs_info *nfs_infop;
@@ -430,87 +593,641 @@ static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
430 */ 593 */
431static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags) 594static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags)
432{ 595{
596 struct nfs_server *server = NFS_SB(vfsmnt->mnt_sb);
597 struct rpc_clnt *rpc;
598
433 shrink_submounts(vfsmnt, &nfs_automount_list); 599 shrink_submounts(vfsmnt, &nfs_automount_list);
600
601 if (!(flags & MNT_FORCE))
602 return;
603 /* -EIO all pending I/O */
604 rpc = server->client_acl;
605 if (!IS_ERR(rpc))
606 rpc_killall_tasks(rpc);
607 rpc = server->client;
608 if (!IS_ERR(rpc))
609 rpc_killall_tasks(rpc);
434} 610}
435 611
436/* 612/*
437 * Validate the NFS2/NFS3 mount data 613 * Sanity-check a server address provided by the mount command
438 * - fills in the mount root filehandle
439 */ 614 */
440static int nfs_validate_mount_data(struct nfs_mount_data *data, 615static int nfs_verify_server_address(struct sockaddr *addr)
441 struct nfs_fh *mntfh)
442{ 616{
443 if (data == NULL) { 617 switch (addr->sa_family) {
444 dprintk("%s: missing data argument\n", __FUNCTION__); 618 case AF_INET: {
445 return -EINVAL; 619 struct sockaddr_in *sa = (struct sockaddr_in *) addr;
620 if (sa->sin_addr.s_addr != INADDR_ANY)
621 return 1;
622 break;
623 }
446 } 624 }
447 625
448 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) { 626 return 0;
449 dprintk("%s: bad mount version\n", __FUNCTION__); 627}
450 return -EINVAL; 628
629/*
630 * Error-check and convert a string of mount options from user space into
631 * a data structure
632 */
633static int nfs_parse_mount_options(char *raw,
634 struct nfs_parsed_mount_data *mnt)
635{
636 char *p, *string;
637
638 if (!raw) {
639 dfprintk(MOUNT, "NFS: mount options string was NULL.\n");
640 return 1;
451 } 641 }
642 dfprintk(MOUNT, "NFS: nfs mount opts='%s'\n", raw);
452 643
453 switch (data->version) { 644 while ((p = strsep(&raw, ",")) != NULL) {
454 case 1: 645 substring_t args[MAX_OPT_ARGS];
455 data->namlen = 0; 646 int option, token;
456 case 2: 647
457 data->bsize = 0; 648 if (!*p)
458 case 3: 649 continue;
459 if (data->flags & NFS_MOUNT_VER3) { 650
460 dprintk("%s: mount structure version %d does not support NFSv3\n", 651 dfprintk(MOUNT, "NFS: parsing nfs mount option '%s'\n", p);
461 __FUNCTION__, 652
462 data->version); 653 token = match_token(p, nfs_mount_option_tokens, args);
463 return -EINVAL; 654 switch (token) {
655 case Opt_soft:
656 mnt->flags |= NFS_MOUNT_SOFT;
657 break;
658 case Opt_hard:
659 mnt->flags &= ~NFS_MOUNT_SOFT;
660 break;
661 case Opt_intr:
662 mnt->flags |= NFS_MOUNT_INTR;
663 break;
664 case Opt_nointr:
665 mnt->flags &= ~NFS_MOUNT_INTR;
666 break;
667 case Opt_posix:
668 mnt->flags |= NFS_MOUNT_POSIX;
669 break;
670 case Opt_noposix:
671 mnt->flags &= ~NFS_MOUNT_POSIX;
672 break;
673 case Opt_cto:
674 mnt->flags &= ~NFS_MOUNT_NOCTO;
675 break;
676 case Opt_nocto:
677 mnt->flags |= NFS_MOUNT_NOCTO;
678 break;
679 case Opt_ac:
680 mnt->flags &= ~NFS_MOUNT_NOAC;
681 break;
682 case Opt_noac:
683 mnt->flags |= NFS_MOUNT_NOAC;
684 break;
685 case Opt_lock:
686 mnt->flags &= ~NFS_MOUNT_NONLM;
687 break;
688 case Opt_nolock:
689 mnt->flags |= NFS_MOUNT_NONLM;
690 break;
691 case Opt_v2:
692 mnt->flags &= ~NFS_MOUNT_VER3;
693 break;
694 case Opt_v3:
695 mnt->flags |= NFS_MOUNT_VER3;
696 break;
697 case Opt_udp:
698 mnt->flags &= ~NFS_MOUNT_TCP;
699 mnt->nfs_server.protocol = IPPROTO_UDP;
700 mnt->timeo = 7;
701 mnt->retrans = 5;
702 break;
703 case Opt_tcp:
704 mnt->flags |= NFS_MOUNT_TCP;
705 mnt->nfs_server.protocol = IPPROTO_TCP;
706 mnt->timeo = 600;
707 mnt->retrans = 2;
708 break;
709 case Opt_acl:
710 mnt->flags &= ~NFS_MOUNT_NOACL;
711 break;
712 case Opt_noacl:
713 mnt->flags |= NFS_MOUNT_NOACL;
714 break;
715 case Opt_rdirplus:
716 mnt->flags &= ~NFS_MOUNT_NORDIRPLUS;
717 break;
718 case Opt_nordirplus:
719 mnt->flags |= NFS_MOUNT_NORDIRPLUS;
720 break;
721 case Opt_sharecache:
722 mnt->flags &= ~NFS_MOUNT_UNSHARED;
723 break;
724 case Opt_nosharecache:
725 mnt->flags |= NFS_MOUNT_UNSHARED;
726 break;
727
728 case Opt_port:
729 if (match_int(args, &option))
730 return 0;
731 if (option < 0 || option > 65535)
732 return 0;
733 mnt->nfs_server.address.sin_port = htonl(option);
734 break;
735 case Opt_rsize:
736 if (match_int(args, &mnt->rsize))
737 return 0;
738 break;
739 case Opt_wsize:
740 if (match_int(args, &mnt->wsize))
741 return 0;
742 break;
743 case Opt_bsize:
744 if (match_int(args, &option))
745 return 0;
746 if (option < 0)
747 return 0;
748 mnt->bsize = option;
749 break;
750 case Opt_timeo:
751 if (match_int(args, &mnt->timeo))
752 return 0;
753 break;
754 case Opt_retrans:
755 if (match_int(args, &mnt->retrans))
756 return 0;
757 break;
758 case Opt_acregmin:
759 if (match_int(args, &mnt->acregmin))
760 return 0;
761 break;
762 case Opt_acregmax:
763 if (match_int(args, &mnt->acregmax))
764 return 0;
765 break;
766 case Opt_acdirmin:
767 if (match_int(args, &mnt->acdirmin))
768 return 0;
769 break;
770 case Opt_acdirmax:
771 if (match_int(args, &mnt->acdirmax))
772 return 0;
773 break;
774 case Opt_actimeo:
775 if (match_int(args, &option))
776 return 0;
777 if (option < 0)
778 return 0;
779 mnt->acregmin =
780 mnt->acregmax =
781 mnt->acdirmin =
782 mnt->acdirmax = option;
783 break;
784 case Opt_namelen:
785 if (match_int(args, &mnt->namlen))
786 return 0;
787 break;
788 case Opt_mountport:
789 if (match_int(args, &option))
790 return 0;
791 if (option < 0 || option > 65535)
792 return 0;
793 mnt->mount_server.port = option;
794 break;
795 case Opt_mountprog:
796 if (match_int(args, &option))
797 return 0;
798 if (option < 0)
799 return 0;
800 mnt->mount_server.program = option;
801 break;
802 case Opt_mountvers:
803 if (match_int(args, &option))
804 return 0;
805 if (option < 0)
806 return 0;
807 mnt->mount_server.version = option;
808 break;
809 case Opt_nfsprog:
810 if (match_int(args, &option))
811 return 0;
812 if (option < 0)
813 return 0;
814 mnt->nfs_server.program = option;
815 break;
816 case Opt_nfsvers:
817 if (match_int(args, &option))
818 return 0;
819 switch (option) {
820 case 2:
821 mnt->flags &= ~NFS_MOUNT_VER3;
822 break;
823 case 3:
824 mnt->flags |= NFS_MOUNT_VER3;
825 break;
826 default:
827 goto out_unrec_vers;
464 } 828 }
465 data->root.size = NFS2_FHSIZE; 829 break;
466 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE); 830
467 case 4: 831 case Opt_sec:
468 if (data->flags & NFS_MOUNT_SECFLAVOUR) { 832 string = match_strdup(args);
469 dprintk("%s: mount structure version %d does not support strong security\n", 833 if (string == NULL)
470 __FUNCTION__, 834 goto out_nomem;
471 data->version); 835 token = match_token(string, nfs_secflavor_tokens, args);
472 return -EINVAL; 836 kfree(string);
837
838 /*
839 * The flags setting is for v2/v3. The flavor_len
840 * setting is for v4. v2/v3 also need to know the
841 * difference between NULL and UNIX.
842 */
843 switch (token) {
844 case Opt_sec_none:
845 mnt->flags &= ~NFS_MOUNT_SECFLAVOUR;
846 mnt->auth_flavor_len = 0;
847 mnt->auth_flavors[0] = RPC_AUTH_NULL;
848 break;
849 case Opt_sec_sys:
850 mnt->flags &= ~NFS_MOUNT_SECFLAVOUR;
851 mnt->auth_flavor_len = 0;
852 mnt->auth_flavors[0] = RPC_AUTH_UNIX;
853 break;
854 case Opt_sec_krb5:
855 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
856 mnt->auth_flavor_len = 1;
857 mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5;
858 break;
859 case Opt_sec_krb5i:
860 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
861 mnt->auth_flavor_len = 1;
862 mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5I;
863 break;
864 case Opt_sec_krb5p:
865 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
866 mnt->auth_flavor_len = 1;
867 mnt->auth_flavors[0] = RPC_AUTH_GSS_KRB5P;
868 break;
869 case Opt_sec_lkey:
870 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
871 mnt->auth_flavor_len = 1;
872 mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEY;
873 break;
874 case Opt_sec_lkeyi:
875 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
876 mnt->auth_flavor_len = 1;
877 mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEYI;
878 break;
879 case Opt_sec_lkeyp:
880 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
881 mnt->auth_flavor_len = 1;
882 mnt->auth_flavors[0] = RPC_AUTH_GSS_LKEYP;
883 break;
884 case Opt_sec_spkm:
885 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
886 mnt->auth_flavor_len = 1;
887 mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKM;
888 break;
889 case Opt_sec_spkmi:
890 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
891 mnt->auth_flavor_len = 1;
892 mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKMI;
893 break;
894 case Opt_sec_spkmp:
895 mnt->flags |= NFS_MOUNT_SECFLAVOUR;
896 mnt->auth_flavor_len = 1;
897 mnt->auth_flavors[0] = RPC_AUTH_GSS_SPKMP;
898 break;
899 default:
900 goto out_unrec_sec;
473 } 901 }
474 case 5: 902 break;
475 memset(data->context, 0, sizeof(data->context)); 903 case Opt_proto:
476 } 904 string = match_strdup(args);
905 if (string == NULL)
906 goto out_nomem;
907 token = match_token(string,
908 nfs_xprt_protocol_tokens, args);
909 kfree(string);
910
911 switch (token) {
912 case Opt_udp:
913 mnt->flags &= ~NFS_MOUNT_TCP;
914 mnt->nfs_server.protocol = IPPROTO_UDP;
915 mnt->timeo = 7;
916 mnt->retrans = 5;
917 break;
918 case Opt_tcp:
919 mnt->flags |= NFS_MOUNT_TCP;
920 mnt->nfs_server.protocol = IPPROTO_TCP;
921 mnt->timeo = 600;
922 mnt->retrans = 2;
923 break;
924 default:
925 goto out_unrec_xprt;
926 }
927 break;
928 case Opt_mountproto:
929 string = match_strdup(args);
930 if (string == NULL)
931 goto out_nomem;
932 token = match_token(string,
933 nfs_xprt_protocol_tokens, args);
934 kfree(string);
935
936 switch (token) {
937 case Opt_udp:
938 mnt->mount_server.protocol = IPPROTO_UDP;
939 break;
940 case Opt_tcp:
941 mnt->mount_server.protocol = IPPROTO_TCP;
942 break;
943 default:
944 goto out_unrec_xprt;
945 }
946 break;
947 case Opt_addr:
948 string = match_strdup(args);
949 if (string == NULL)
950 goto out_nomem;
951 mnt->nfs_server.address.sin_family = AF_INET;
952 mnt->nfs_server.address.sin_addr.s_addr =
953 in_aton(string);
954 kfree(string);
955 break;
956 case Opt_clientaddr:
957 string = match_strdup(args);
958 if (string == NULL)
959 goto out_nomem;
960 mnt->client_address = string;
961 break;
962 case Opt_mounthost:
963 string = match_strdup(args);
964 if (string == NULL)
965 goto out_nomem;
966 mnt->mount_server.address.sin_family = AF_INET;
967 mnt->mount_server.address.sin_addr.s_addr =
968 in_aton(string);
969 kfree(string);
970 break;
477 971
478 /* Set the pseudoflavor */ 972 case Opt_userspace:
479 if (!(data->flags & NFS_MOUNT_SECFLAVOUR)) 973 case Opt_deprecated:
480 data->pseudoflavor = RPC_AUTH_UNIX; 974 break;
481 975
482#ifndef CONFIG_NFS_V3 976 default:
483 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */ 977 goto out_unknown;
484 if (data->flags & NFS_MOUNT_VER3) { 978 }
485 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
486 return -EPROTONOSUPPORT;
487 } 979 }
488#endif /* CONFIG_NFS_V3 */
489 980
490 /* We now require that the mount process passes the remote address */ 981 return 1;
491 if (data->addr.sin_addr.s_addr == INADDR_ANY) { 982
492 dprintk("%s: mount program didn't pass remote address!\n", 983out_nomem:
493 __FUNCTION__); 984 printk(KERN_INFO "NFS: not enough memory to parse option\n");
494 return -EINVAL; 985 return 0;
986
987out_unrec_vers:
988 printk(KERN_INFO "NFS: unrecognized NFS version number\n");
989 return 0;
990
991out_unrec_xprt:
992 printk(KERN_INFO "NFS: unrecognized transport protocol\n");
993 return 0;
994
995out_unrec_sec:
996 printk(KERN_INFO "NFS: unrecognized security flavor\n");
997 return 0;
998
999out_unknown:
1000 printk(KERN_INFO "NFS: unknown mount option: %s\n", p);
1001 return 0;
1002}
1003
1004/*
1005 * Use the remote server's MOUNT service to request the NFS file handle
1006 * corresponding to the provided path.
1007 */
1008static int nfs_try_mount(struct nfs_parsed_mount_data *args,
1009 struct nfs_fh *root_fh)
1010{
1011 struct sockaddr_in sin;
1012 int status;
1013
1014 if (args->mount_server.version == 0) {
1015 if (args->flags & NFS_MOUNT_VER3)
1016 args->mount_server.version = NFS_MNT3_VERSION;
1017 else
1018 args->mount_server.version = NFS_MNT_VERSION;
495 } 1019 }
496 1020
497 /* Prepare the root filehandle */ 1021 /*
498 if (data->flags & NFS_MOUNT_VER3) 1022 * Construct the mount server's address.
499 mntfh->size = data->root.size; 1023 */
1024 if (args->mount_server.address.sin_addr.s_addr != INADDR_ANY)
1025 sin = args->mount_server.address;
500 else 1026 else
501 mntfh->size = NFS2_FHSIZE; 1027 sin = args->nfs_server.address;
1028 if (args->mount_server.port == 0) {
1029 status = rpcb_getport_sync(&sin,
1030 args->mount_server.program,
1031 args->mount_server.version,
1032 args->mount_server.protocol);
1033 if (status < 0)
1034 goto out_err;
1035 sin.sin_port = htons(status);
1036 } else
1037 sin.sin_port = htons(args->mount_server.port);
1038
1039 /*
1040 * Now ask the mount server to map our export path
1041 * to a file handle.
1042 */
1043 status = nfs_mount((struct sockaddr *) &sin,
1044 sizeof(sin),
1045 args->nfs_server.hostname,
1046 args->nfs_server.export_path,
1047 args->mount_server.version,
1048 args->mount_server.protocol,
1049 root_fh);
1050 if (status < 0)
1051 goto out_err;
1052
1053 return status;
502 1054
503 if (mntfh->size > sizeof(mntfh->data)) { 1055out_err:
504 dprintk("%s: invalid root filehandle\n", __FUNCTION__); 1056 dfprintk(MOUNT, "NFS: unable to contact server on host "
505 return -EINVAL; 1057 NIPQUAD_FMT "\n", NIPQUAD(sin.sin_addr.s_addr));
1058 return status;
1059}
1060
1061/*
1062 * Validate the NFS2/NFS3 mount data
1063 * - fills in the mount root filehandle
1064 *
1065 * For option strings, user space handles the following behaviors:
1066 *
1067 * + DNS: mapping server host name to IP address ("addr=" option)
1068 *
1069 * + failure mode: how to behave if a mount request can't be handled
1070 * immediately ("fg/bg" option)
1071 *
1072 * + retry: how often to retry a mount request ("retry=" option)
1073 *
1074 * + breaking back: trying proto=udp after proto=tcp, v2 after v3,
1075 * mountproto=tcp after mountproto=udp, and so on
1076 *
1077 * XXX: as far as I can tell, changing the NFS program number is not
1078 * supported in the NFS client.
1079 */
1080static int nfs_validate_mount_data(struct nfs_mount_data **options,
1081 struct nfs_fh *mntfh,
1082 const char *dev_name)
1083{
1084 struct nfs_mount_data *data = *options;
1085
1086 if (data == NULL)
1087 goto out_no_data;
1088
1089 switch (data->version) {
1090 case 1:
1091 data->namlen = 0;
1092 case 2:
1093 data->bsize = 0;
1094 case 3:
1095 if (data->flags & NFS_MOUNT_VER3)
1096 goto out_no_v3;
1097 data->root.size = NFS2_FHSIZE;
1098 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1099 case 4:
1100 if (data->flags & NFS_MOUNT_SECFLAVOUR)
1101 goto out_no_sec;
1102 case 5:
1103 memset(data->context, 0, sizeof(data->context));
1104 case 6:
1105 if (data->flags & NFS_MOUNT_VER3)
1106 mntfh->size = data->root.size;
1107 else
1108 mntfh->size = NFS2_FHSIZE;
1109
1110 if (mntfh->size > sizeof(mntfh->data))
1111 goto out_invalid_fh;
1112
1113 memcpy(mntfh->data, data->root.data, mntfh->size);
1114 if (mntfh->size < sizeof(mntfh->data))
1115 memset(mntfh->data + mntfh->size, 0,
1116 sizeof(mntfh->data) - mntfh->size);
1117 break;
1118 default: {
1119 unsigned int len;
1120 char *c;
1121 int status;
1122 struct nfs_parsed_mount_data args = {
1123 .flags = (NFS_MOUNT_VER3 | NFS_MOUNT_TCP),
1124 .rsize = NFS_MAX_FILE_IO_SIZE,
1125 .wsize = NFS_MAX_FILE_IO_SIZE,
1126 .timeo = 600,
1127 .retrans = 2,
1128 .acregmin = 3,
1129 .acregmax = 60,
1130 .acdirmin = 30,
1131 .acdirmax = 60,
1132 .mount_server.protocol = IPPROTO_UDP,
1133 .mount_server.program = NFS_MNT_PROGRAM,
1134 .nfs_server.protocol = IPPROTO_TCP,
1135 .nfs_server.program = NFS_PROGRAM,
1136 };
1137
1138 if (nfs_parse_mount_options((char *) *options, &args) == 0)
1139 return -EINVAL;
1140
1141 data = kzalloc(sizeof(*data), GFP_KERNEL);
1142 if (data == NULL)
1143 return -ENOMEM;
1144
1145 /*
1146 * NB: after this point, caller will free "data"
1147 * if we return an error
1148 */
1149 *options = data;
1150
1151 c = strchr(dev_name, ':');
1152 if (c == NULL)
1153 return -EINVAL;
1154 len = c - dev_name - 1;
1155 if (len > sizeof(data->hostname))
1156 return -EINVAL;
1157 strncpy(data->hostname, dev_name, len);
1158 args.nfs_server.hostname = data->hostname;
1159
1160 c++;
1161 if (strlen(c) > NFS_MAXPATHLEN)
1162 return -EINVAL;
1163 args.nfs_server.export_path = c;
1164
1165 status = nfs_try_mount(&args, mntfh);
1166 if (status)
1167 return -EINVAL;
1168
1169 /*
1170 * Translate to nfs_mount_data, which nfs_fill_super
1171 * can deal with.
1172 */
1173 data->version = 6;
1174 data->flags = args.flags;
1175 data->rsize = args.rsize;
1176 data->wsize = args.wsize;
1177 data->timeo = args.timeo;
1178 data->retrans = args.retrans;
1179 data->acregmin = args.acregmin;
1180 data->acregmax = args.acregmax;
1181 data->acdirmin = args.acdirmin;
1182 data->acdirmax = args.acdirmax;
1183 data->addr = args.nfs_server.address;
1184 data->namlen = args.namlen;
1185 data->bsize = args.bsize;
1186 data->pseudoflavor = args.auth_flavors[0];
1187
1188 break;
1189 }
506 } 1190 }
507 1191
508 memcpy(mntfh->data, data->root.data, mntfh->size); 1192 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
509 if (mntfh->size < sizeof(mntfh->data)) 1193 data->pseudoflavor = RPC_AUTH_UNIX;
510 memset(mntfh->data + mntfh->size, 0, 1194
511 sizeof(mntfh->data) - mntfh->size); 1195#ifndef CONFIG_NFS_V3
1196 if (data->flags & NFS_MOUNT_VER3)
1197 goto out_v3_not_compiled;
1198#endif /* !CONFIG_NFS_V3 */
1199
1200 if (!nfs_verify_server_address((struct sockaddr *) &data->addr))
1201 goto out_no_address;
512 1202
513 return 0; 1203 return 0;
1204
1205out_no_data:
1206 dfprintk(MOUNT, "NFS: mount program didn't pass any mount data\n");
1207 return -EINVAL;
1208
1209out_no_v3:
1210 dfprintk(MOUNT, "NFS: nfs_mount_data version %d does not support v3\n",
1211 data->version);
1212 return -EINVAL;
1213
1214out_no_sec:
1215 dfprintk(MOUNT, "NFS: nfs_mount_data version supports only AUTH_SYS\n");
1216 return -EINVAL;
1217
1218#ifndef CONFIG_NFS_V3
1219out_v3_not_compiled:
1220 dfprintk(MOUNT, "NFS: NFSv3 is not compiled into kernel\n");
1221 return -EPROTONOSUPPORT;
1222#endif /* !CONFIG_NFS_V3 */
1223
1224out_no_address:
1225 dfprintk(MOUNT, "NFS: mount program didn't pass remote address\n");
1226 return -EINVAL;
1227
1228out_invalid_fh:
1229 dfprintk(MOUNT, "NFS: invalid root filehandle\n");
1230 return -EINVAL;
514} 1231}
515 1232
516/* 1233/*
@@ -600,13 +1317,51 @@ static int nfs_compare_super(struct super_block *sb, void *data)
600{ 1317{
601 struct nfs_server *server = data, *old = NFS_SB(sb); 1318 struct nfs_server *server = data, *old = NFS_SB(sb);
602 1319
603 if (old->nfs_client != server->nfs_client) 1320 if (memcmp(&old->nfs_client->cl_addr,
1321 &server->nfs_client->cl_addr,
1322 sizeof(old->nfs_client->cl_addr)) != 0)
1323 return 0;
1324 /* Note: NFS_MOUNT_UNSHARED == NFS4_MOUNT_UNSHARED */
1325 if (old->flags & NFS_MOUNT_UNSHARED)
604 return 0; 1326 return 0;
605 if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0) 1327 if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0)
606 return 0; 1328 return 0;
607 return 1; 1329 return 1;
608} 1330}
609 1331
1332#define NFS_MS_MASK (MS_RDONLY|MS_NOSUID|MS_NODEV|MS_NOEXEC|MS_SYNCHRONOUS)
1333
1334static int nfs_compare_mount_options(const struct super_block *s, const struct nfs_server *b, int flags)
1335{
1336 const struct nfs_server *a = s->s_fs_info;
1337 const struct rpc_clnt *clnt_a = a->client;
1338 const struct rpc_clnt *clnt_b = b->client;
1339
1340 if ((s->s_flags & NFS_MS_MASK) != (flags & NFS_MS_MASK))
1341 goto Ebusy;
1342 if (a->nfs_client != b->nfs_client)
1343 goto Ebusy;
1344 if (a->flags != b->flags)
1345 goto Ebusy;
1346 if (a->wsize != b->wsize)
1347 goto Ebusy;
1348 if (a->rsize != b->rsize)
1349 goto Ebusy;
1350 if (a->acregmin != b->acregmin)
1351 goto Ebusy;
1352 if (a->acregmax != b->acregmax)
1353 goto Ebusy;
1354 if (a->acdirmin != b->acdirmin)
1355 goto Ebusy;
1356 if (a->acdirmax != b->acdirmax)
1357 goto Ebusy;
1358 if (clnt_a->cl_auth->au_flavor != clnt_b->cl_auth->au_flavor)
1359 goto Ebusy;
1360 return 0;
1361Ebusy:
1362 return -EBUSY;
1363}
1364
610static int nfs_get_sb(struct file_system_type *fs_type, 1365static int nfs_get_sb(struct file_system_type *fs_type,
611 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt) 1366 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
612{ 1367{
@@ -615,30 +1370,37 @@ static int nfs_get_sb(struct file_system_type *fs_type,
615 struct nfs_fh mntfh; 1370 struct nfs_fh mntfh;
616 struct nfs_mount_data *data = raw_data; 1371 struct nfs_mount_data *data = raw_data;
617 struct dentry *mntroot; 1372 struct dentry *mntroot;
1373 int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
618 int error; 1374 int error;
619 1375
620 /* Validate the mount data */ 1376 /* Validate the mount data */
621 error = nfs_validate_mount_data(data, &mntfh); 1377 error = nfs_validate_mount_data(&data, &mntfh, dev_name);
622 if (error < 0) 1378 if (error < 0)
623 return error; 1379 goto out;
624 1380
625 /* Get a volume representation */ 1381 /* Get a volume representation */
626 server = nfs_create_server(data, &mntfh); 1382 server = nfs_create_server(data, &mntfh);
627 if (IS_ERR(server)) { 1383 if (IS_ERR(server)) {
628 error = PTR_ERR(server); 1384 error = PTR_ERR(server);
629 goto out_err_noserver; 1385 goto out;
630 } 1386 }
631 1387
1388 if (server->flags & NFS_MOUNT_UNSHARED)
1389 compare_super = NULL;
1390
632 /* Get a superblock - note that we may end up sharing one that already exists */ 1391 /* Get a superblock - note that we may end up sharing one that already exists */
633 s = sget(fs_type, nfs_compare_super, nfs_set_super, server); 1392 s = sget(fs_type, compare_super, nfs_set_super, server);
634 if (IS_ERR(s)) { 1393 if (IS_ERR(s)) {
635 error = PTR_ERR(s); 1394 error = PTR_ERR(s);
636 goto out_err_nosb; 1395 goto out_err_nosb;
637 } 1396 }
638 1397
639 if (s->s_fs_info != server) { 1398 if (s->s_fs_info != server) {
1399 error = nfs_compare_mount_options(s, server, flags);
640 nfs_free_server(server); 1400 nfs_free_server(server);
641 server = NULL; 1401 server = NULL;
1402 if (error < 0)
1403 goto error_splat_super;
642 } 1404 }
643 1405
644 if (!s->s_root) { 1406 if (!s->s_root) {
@@ -656,17 +1418,21 @@ static int nfs_get_sb(struct file_system_type *fs_type,
656 s->s_flags |= MS_ACTIVE; 1418 s->s_flags |= MS_ACTIVE;
657 mnt->mnt_sb = s; 1419 mnt->mnt_sb = s;
658 mnt->mnt_root = mntroot; 1420 mnt->mnt_root = mntroot;
659 return 0; 1421 error = 0;
1422
1423out:
1424 if (data != raw_data)
1425 kfree(data);
1426 return error;
660 1427
661out_err_nosb: 1428out_err_nosb:
662 nfs_free_server(server); 1429 nfs_free_server(server);
663out_err_noserver: 1430 goto out;
664 return error;
665 1431
666error_splat_super: 1432error_splat_super:
667 up_write(&s->s_umount); 1433 up_write(&s->s_umount);
668 deactivate_super(s); 1434 deactivate_super(s);
669 return error; 1435 goto out;
670} 1436}
671 1437
672/* 1438/*
@@ -691,6 +1457,7 @@ static int nfs_xdev_get_sb(struct file_system_type *fs_type, int flags,
691 struct super_block *s; 1457 struct super_block *s;
692 struct nfs_server *server; 1458 struct nfs_server *server;
693 struct dentry *mntroot; 1459 struct dentry *mntroot;
1460 int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
694 int error; 1461 int error;
695 1462
696 dprintk("--> nfs_xdev_get_sb()\n"); 1463 dprintk("--> nfs_xdev_get_sb()\n");
@@ -702,16 +1469,22 @@ static int nfs_xdev_get_sb(struct file_system_type *fs_type, int flags,
702 goto out_err_noserver; 1469 goto out_err_noserver;
703 } 1470 }
704 1471
1472 if (server->flags & NFS_MOUNT_UNSHARED)
1473 compare_super = NULL;
1474
705 /* Get a superblock - note that we may end up sharing one that already exists */ 1475 /* Get a superblock - note that we may end up sharing one that already exists */
706 s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); 1476 s = sget(&nfs_fs_type, compare_super, nfs_set_super, server);
707 if (IS_ERR(s)) { 1477 if (IS_ERR(s)) {
708 error = PTR_ERR(s); 1478 error = PTR_ERR(s);
709 goto out_err_nosb; 1479 goto out_err_nosb;
710 } 1480 }
711 1481
712 if (s->s_fs_info != server) { 1482 if (s->s_fs_info != server) {
1483 error = nfs_compare_mount_options(s, server, flags);
713 nfs_free_server(server); 1484 nfs_free_server(server);
714 server = NULL; 1485 server = NULL;
1486 if (error < 0)
1487 goto error_splat_super;
715 } 1488 }
716 1489
717 if (!s->s_root) { 1490 if (!s->s_root) {
@@ -772,25 +1545,164 @@ static void nfs4_fill_super(struct super_block *sb)
772 nfs_initialise_sb(sb); 1545 nfs_initialise_sb(sb);
773} 1546}
774 1547
775static void *nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen) 1548/*
1549 * Validate NFSv4 mount options
1550 */
1551static int nfs4_validate_mount_data(struct nfs4_mount_data **options,
1552 const char *dev_name,
1553 struct sockaddr_in *addr,
1554 rpc_authflavor_t *authflavour,
1555 char **hostname,
1556 char **mntpath,
1557 char **ip_addr)
776{ 1558{
777 void *p = NULL; 1559 struct nfs4_mount_data *data = *options;
778 1560 char *c;
779 if (!src->len) 1561
780 return ERR_PTR(-EINVAL); 1562 if (data == NULL)
781 if (src->len < maxlen) 1563 goto out_no_data;
782 maxlen = src->len; 1564
783 if (dst == NULL) { 1565 switch (data->version) {
784 p = dst = kmalloc(maxlen + 1, GFP_KERNEL); 1566 case 1:
785 if (p == NULL) 1567 if (data->host_addrlen != sizeof(*addr))
786 return ERR_PTR(-ENOMEM); 1568 goto out_no_address;
787 } 1569 if (copy_from_user(addr, data->host_addr, sizeof(*addr)))
788 if (copy_from_user(dst, src->data, maxlen)) { 1570 return -EFAULT;
789 kfree(p); 1571 if (addr->sin_port == 0)
790 return ERR_PTR(-EFAULT); 1572 addr->sin_port = htons(NFS_PORT);
1573 if (!nfs_verify_server_address((struct sockaddr *) addr))
1574 goto out_no_address;
1575
1576 switch (data->auth_flavourlen) {
1577 case 0:
1578 *authflavour = RPC_AUTH_UNIX;
1579 break;
1580 case 1:
1581 if (copy_from_user(authflavour, data->auth_flavours,
1582 sizeof(*authflavour)))
1583 return -EFAULT;
1584 break;
1585 default:
1586 goto out_inval_auth;
1587 }
1588
1589 c = strndup_user(data->hostname.data, NFS4_MAXNAMLEN);
1590 if (IS_ERR(c))
1591 return PTR_ERR(c);
1592 *hostname = c;
1593
1594 c = strndup_user(data->mnt_path.data, NFS4_MAXPATHLEN);
1595 if (IS_ERR(c))
1596 return PTR_ERR(c);
1597 *mntpath = c;
1598 dfprintk(MOUNT, "NFS: MNTPATH: '%s'\n", *mntpath);
1599
1600 c = strndup_user(data->client_addr.data, 16);
1601 if (IS_ERR(c))
1602 return PTR_ERR(c);
1603 *ip_addr = c;
1604
1605 break;
1606 default: {
1607 unsigned int len;
1608 struct nfs_parsed_mount_data args = {
1609 .rsize = NFS_MAX_FILE_IO_SIZE,
1610 .wsize = NFS_MAX_FILE_IO_SIZE,
1611 .timeo = 600,
1612 .retrans = 2,
1613 .acregmin = 3,
1614 .acregmax = 60,
1615 .acdirmin = 30,
1616 .acdirmax = 60,
1617 .nfs_server.protocol = IPPROTO_TCP,
1618 };
1619
1620 if (nfs_parse_mount_options((char *) *options, &args) == 0)
1621 return -EINVAL;
1622
1623 if (!nfs_verify_server_address((struct sockaddr *)
1624 &args.nfs_server.address))
1625 return -EINVAL;
1626 *addr = args.nfs_server.address;
1627
1628 switch (args.auth_flavor_len) {
1629 case 0:
1630 *authflavour = RPC_AUTH_UNIX;
1631 break;
1632 case 1:
1633 *authflavour = (rpc_authflavor_t) args.auth_flavors[0];
1634 break;
1635 default:
1636 goto out_inval_auth;
1637 }
1638
1639 /*
1640 * Translate to nfs4_mount_data, which nfs4_fill_super
1641 * can deal with.
1642 */
1643 data = kzalloc(sizeof(*data), GFP_KERNEL);
1644 if (data == NULL)
1645 return -ENOMEM;
1646 *options = data;
1647
1648 data->version = 1;
1649 data->flags = args.flags & NFS4_MOUNT_FLAGMASK;
1650 data->rsize = args.rsize;
1651 data->wsize = args.wsize;
1652 data->timeo = args.timeo;
1653 data->retrans = args.retrans;
1654 data->acregmin = args.acregmin;
1655 data->acregmax = args.acregmax;
1656 data->acdirmin = args.acdirmin;
1657 data->acdirmax = args.acdirmax;
1658 data->proto = args.nfs_server.protocol;
1659
1660 /*
1661 * Split "dev_name" into "hostname:mntpath".
1662 */
1663 c = strchr(dev_name, ':');
1664 if (c == NULL)
1665 return -EINVAL;
1666 /* while calculating len, pretend ':' is '\0' */
1667 len = c - dev_name;
1668 if (len > NFS4_MAXNAMLEN)
1669 return -EINVAL;
1670 *hostname = kzalloc(len, GFP_KERNEL);
1671 if (*hostname == NULL)
1672 return -ENOMEM;
1673 strncpy(*hostname, dev_name, len - 1);
1674
1675 c++; /* step over the ':' */
1676 len = strlen(c);
1677 if (len > NFS4_MAXPATHLEN)
1678 return -EINVAL;
1679 *mntpath = kzalloc(len + 1, GFP_KERNEL);
1680 if (*mntpath == NULL)
1681 return -ENOMEM;
1682 strncpy(*mntpath, c, len);
1683
1684 dprintk("MNTPATH: %s\n", *mntpath);
1685
1686 *ip_addr = args.client_address;
1687
1688 break;
1689 }
791 } 1690 }
792 dst[maxlen] = '\0'; 1691
793 return dst; 1692 return 0;
1693
1694out_no_data:
1695 dfprintk(MOUNT, "NFS4: mount program didn't pass any mount data\n");
1696 return -EINVAL;
1697
1698out_inval_auth:
1699 dfprintk(MOUNT, "NFS4: Invalid number of RPC auth flavours %d\n",
1700 data->auth_flavourlen);
1701 return -EINVAL;
1702
1703out_no_address:
1704 dfprintk(MOUNT, "NFS4: mount program didn't pass remote address\n");
1705 return -EINVAL;
794} 1706}
795 1707
796/* 1708/*
@@ -806,81 +1718,29 @@ static int nfs4_get_sb(struct file_system_type *fs_type,
806 rpc_authflavor_t authflavour; 1718 rpc_authflavor_t authflavour;
807 struct nfs_fh mntfh; 1719 struct nfs_fh mntfh;
808 struct dentry *mntroot; 1720 struct dentry *mntroot;
809 char *mntpath = NULL, *hostname = NULL, ip_addr[16]; 1721 char *mntpath = NULL, *hostname = NULL, *ip_addr = NULL;
810 void *p; 1722 int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
811 int error; 1723 int error;
812 1724
813 if (data == NULL) { 1725 /* Validate the mount data */
814 dprintk("%s: missing data argument\n", __FUNCTION__); 1726 error = nfs4_validate_mount_data(&data, dev_name, &addr, &authflavour,
815 return -EINVAL; 1727 &hostname, &mntpath, &ip_addr);
816 } 1728 if (error < 0)
817 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) { 1729 goto out;
818 dprintk("%s: bad mount version\n", __FUNCTION__);
819 return -EINVAL;
820 }
821
822 /* We now require that the mount process passes the remote address */
823 if (data->host_addrlen != sizeof(addr))
824 return -EINVAL;
825
826 if (copy_from_user(&addr, data->host_addr, sizeof(addr)))
827 return -EFAULT;
828
829 if (addr.sin_family != AF_INET ||
830 addr.sin_addr.s_addr == INADDR_ANY
831 ) {
832 dprintk("%s: mount program didn't pass remote IP address!\n",
833 __FUNCTION__);
834 return -EINVAL;
835 }
836 /* RFC3530: The default port for NFS is 2049 */
837 if (addr.sin_port == 0)
838 addr.sin_port = htons(NFS_PORT);
839
840 /* Grab the authentication type */
841 authflavour = RPC_AUTH_UNIX;
842 if (data->auth_flavourlen != 0) {
843 if (data->auth_flavourlen != 1) {
844 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
845 __FUNCTION__, data->auth_flavourlen);
846 error = -EINVAL;
847 goto out_err_noserver;
848 }
849
850 if (copy_from_user(&authflavour, data->auth_flavours,
851 sizeof(authflavour))) {
852 error = -EFAULT;
853 goto out_err_noserver;
854 }
855 }
856
857 p = nfs_copy_user_string(NULL, &data->hostname, 256);
858 if (IS_ERR(p))
859 goto out_err;
860 hostname = p;
861
862 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
863 if (IS_ERR(p))
864 goto out_err;
865 mntpath = p;
866
867 dprintk("MNTPATH: %s\n", mntpath);
868
869 p = nfs_copy_user_string(ip_addr, &data->client_addr,
870 sizeof(ip_addr) - 1);
871 if (IS_ERR(p))
872 goto out_err;
873 1730
874 /* Get a volume representation */ 1731 /* Get a volume representation */
875 server = nfs4_create_server(data, hostname, &addr, mntpath, ip_addr, 1732 server = nfs4_create_server(data, hostname, &addr, mntpath, ip_addr,
876 authflavour, &mntfh); 1733 authflavour, &mntfh);
877 if (IS_ERR(server)) { 1734 if (IS_ERR(server)) {
878 error = PTR_ERR(server); 1735 error = PTR_ERR(server);
879 goto out_err_noserver; 1736 goto out;
880 } 1737 }
881 1738
1739 if (server->flags & NFS4_MOUNT_UNSHARED)
1740 compare_super = NULL;
1741
882 /* Get a superblock - note that we may end up sharing one that already exists */ 1742 /* Get a superblock - note that we may end up sharing one that already exists */
883 s = sget(fs_type, nfs_compare_super, nfs_set_super, server); 1743 s = sget(fs_type, compare_super, nfs_set_super, server);
884 if (IS_ERR(s)) { 1744 if (IS_ERR(s)) {
885 error = PTR_ERR(s); 1745 error = PTR_ERR(s);
886 goto out_free; 1746 goto out_free;
@@ -906,25 +1766,22 @@ static int nfs4_get_sb(struct file_system_type *fs_type,
906 s->s_flags |= MS_ACTIVE; 1766 s->s_flags |= MS_ACTIVE;
907 mnt->mnt_sb = s; 1767 mnt->mnt_sb = s;
908 mnt->mnt_root = mntroot; 1768 mnt->mnt_root = mntroot;
1769 error = 0;
1770
1771out:
1772 kfree(ip_addr);
909 kfree(mntpath); 1773 kfree(mntpath);
910 kfree(hostname); 1774 kfree(hostname);
911 return 0; 1775 return error;
912
913out_err:
914 error = PTR_ERR(p);
915 goto out_err_noserver;
916 1776
917out_free: 1777out_free:
918 nfs_free_server(server); 1778 nfs_free_server(server);
919out_err_noserver: 1779 goto out;
920 kfree(mntpath);
921 kfree(hostname);
922 return error;
923 1780
924error_splat_super: 1781error_splat_super:
925 up_write(&s->s_umount); 1782 up_write(&s->s_umount);
926 deactivate_super(s); 1783 deactivate_super(s);
927 goto out_err_noserver; 1784 goto out;
928} 1785}
929 1786
930static void nfs4_kill_super(struct super_block *sb) 1787static void nfs4_kill_super(struct super_block *sb)
@@ -949,6 +1806,7 @@ static int nfs4_xdev_get_sb(struct file_system_type *fs_type, int flags,
949 struct super_block *s; 1806 struct super_block *s;
950 struct nfs_server *server; 1807 struct nfs_server *server;
951 struct dentry *mntroot; 1808 struct dentry *mntroot;
1809 int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
952 int error; 1810 int error;
953 1811
954 dprintk("--> nfs4_xdev_get_sb()\n"); 1812 dprintk("--> nfs4_xdev_get_sb()\n");
@@ -960,8 +1818,11 @@ static int nfs4_xdev_get_sb(struct file_system_type *fs_type, int flags,
960 goto out_err_noserver; 1818 goto out_err_noserver;
961 } 1819 }
962 1820
1821 if (server->flags & NFS4_MOUNT_UNSHARED)
1822 compare_super = NULL;
1823
963 /* Get a superblock - note that we may end up sharing one that already exists */ 1824 /* Get a superblock - note that we may end up sharing one that already exists */
964 s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); 1825 s = sget(&nfs_fs_type, compare_super, nfs_set_super, server);
965 if (IS_ERR(s)) { 1826 if (IS_ERR(s)) {
966 error = PTR_ERR(s); 1827 error = PTR_ERR(s);
967 goto out_err_nosb; 1828 goto out_err_nosb;
@@ -1016,6 +1877,7 @@ static int nfs4_referral_get_sb(struct file_system_type *fs_type, int flags,
1016 struct nfs_server *server; 1877 struct nfs_server *server;
1017 struct dentry *mntroot; 1878 struct dentry *mntroot;
1018 struct nfs_fh mntfh; 1879 struct nfs_fh mntfh;
1880 int (*compare_super)(struct super_block *, void *) = nfs_compare_super;
1019 int error; 1881 int error;
1020 1882
1021 dprintk("--> nfs4_referral_get_sb()\n"); 1883 dprintk("--> nfs4_referral_get_sb()\n");
@@ -1027,8 +1889,11 @@ static int nfs4_referral_get_sb(struct file_system_type *fs_type, int flags,
1027 goto out_err_noserver; 1889 goto out_err_noserver;
1028 } 1890 }
1029 1891
1892 if (server->flags & NFS4_MOUNT_UNSHARED)
1893 compare_super = NULL;
1894
1030 /* Get a superblock - note that we may end up sharing one that already exists */ 1895 /* Get a superblock - note that we may end up sharing one that already exists */
1031 s = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server); 1896 s = sget(&nfs_fs_type, compare_super, nfs_set_super, server);
1032 if (IS_ERR(s)) { 1897 if (IS_ERR(s)) {
1033 error = PTR_ERR(s); 1898 error = PTR_ERR(s);
1034 goto out_err_nosb; 1899 goto out_err_nosb;
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index af344a158e01..73ac992ece85 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -117,19 +117,19 @@ static struct nfs_page *nfs_page_find_request_locked(struct page *page)
117 if (PagePrivate(page)) { 117 if (PagePrivate(page)) {
118 req = (struct nfs_page *)page_private(page); 118 req = (struct nfs_page *)page_private(page);
119 if (req != NULL) 119 if (req != NULL)
120 atomic_inc(&req->wb_count); 120 kref_get(&req->wb_kref);
121 } 121 }
122 return req; 122 return req;
123} 123}
124 124
125static struct nfs_page *nfs_page_find_request(struct page *page) 125static struct nfs_page *nfs_page_find_request(struct page *page)
126{ 126{
127 struct inode *inode = page->mapping->host;
127 struct nfs_page *req = NULL; 128 struct nfs_page *req = NULL;
128 spinlock_t *req_lock = &NFS_I(page->mapping->host)->req_lock;
129 129
130 spin_lock(req_lock); 130 spin_lock(&inode->i_lock);
131 req = nfs_page_find_request_locked(page); 131 req = nfs_page_find_request_locked(page);
132 spin_unlock(req_lock); 132 spin_unlock(&inode->i_lock);
133 return req; 133 return req;
134} 134}
135 135
@@ -191,8 +191,6 @@ static int nfs_writepage_setup(struct nfs_open_context *ctx, struct page *page,
191 } 191 }
192 /* Update file length */ 192 /* Update file length */
193 nfs_grow_file(page, offset, count); 193 nfs_grow_file(page, offset, count);
194 /* Set the PG_uptodate flag? */
195 nfs_mark_uptodate(page, offset, count);
196 nfs_unlock_request(req); 194 nfs_unlock_request(req);
197 return 0; 195 return 0;
198} 196}
@@ -253,16 +251,16 @@ static void nfs_end_page_writeback(struct page *page)
253static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, 251static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
254 struct page *page) 252 struct page *page)
255{ 253{
254 struct inode *inode = page->mapping->host;
255 struct nfs_inode *nfsi = NFS_I(inode);
256 struct nfs_page *req; 256 struct nfs_page *req;
257 struct nfs_inode *nfsi = NFS_I(page->mapping->host);
258 spinlock_t *req_lock = &nfsi->req_lock;
259 int ret; 257 int ret;
260 258
261 spin_lock(req_lock); 259 spin_lock(&inode->i_lock);
262 for(;;) { 260 for(;;) {
263 req = nfs_page_find_request_locked(page); 261 req = nfs_page_find_request_locked(page);
264 if (req == NULL) { 262 if (req == NULL) {
265 spin_unlock(req_lock); 263 spin_unlock(&inode->i_lock);
266 return 1; 264 return 1;
267 } 265 }
268 if (nfs_lock_request_dontget(req)) 266 if (nfs_lock_request_dontget(req))
@@ -272,28 +270,28 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
272 * succeed provided that someone hasn't already marked the 270 * succeed provided that someone hasn't already marked the
273 * request as dirty (in which case we don't care). 271 * request as dirty (in which case we don't care).
274 */ 272 */
275 spin_unlock(req_lock); 273 spin_unlock(&inode->i_lock);
276 ret = nfs_wait_on_request(req); 274 ret = nfs_wait_on_request(req);
277 nfs_release_request(req); 275 nfs_release_request(req);
278 if (ret != 0) 276 if (ret != 0)
279 return ret; 277 return ret;
280 spin_lock(req_lock); 278 spin_lock(&inode->i_lock);
281 } 279 }
282 if (test_bit(PG_NEED_COMMIT, &req->wb_flags)) { 280 if (test_bit(PG_NEED_COMMIT, &req->wb_flags)) {
283 /* This request is marked for commit */ 281 /* This request is marked for commit */
284 spin_unlock(req_lock); 282 spin_unlock(&inode->i_lock);
285 nfs_unlock_request(req); 283 nfs_unlock_request(req);
286 nfs_pageio_complete(pgio); 284 nfs_pageio_complete(pgio);
287 return 1; 285 return 1;
288 } 286 }
289 if (nfs_set_page_writeback(page) != 0) { 287 if (nfs_set_page_writeback(page) != 0) {
290 spin_unlock(req_lock); 288 spin_unlock(&inode->i_lock);
291 BUG(); 289 BUG();
292 } 290 }
293 radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, 291 radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index,
294 NFS_PAGE_TAG_WRITEBACK); 292 NFS_PAGE_TAG_LOCKED);
295 ret = test_bit(PG_NEED_FLUSH, &req->wb_flags); 293 ret = test_bit(PG_NEED_FLUSH, &req->wb_flags);
296 spin_unlock(req_lock); 294 spin_unlock(&inode->i_lock);
297 nfs_pageio_add_request(pgio, req); 295 nfs_pageio_add_request(pgio, req);
298 return ret; 296 return ret;
299} 297}
@@ -400,7 +398,7 @@ static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
400 if (PageDirty(req->wb_page)) 398 if (PageDirty(req->wb_page))
401 set_bit(PG_NEED_FLUSH, &req->wb_flags); 399 set_bit(PG_NEED_FLUSH, &req->wb_flags);
402 nfsi->npages++; 400 nfsi->npages++;
403 atomic_inc(&req->wb_count); 401 kref_get(&req->wb_kref);
404 return 0; 402 return 0;
405} 403}
406 404
@@ -409,12 +407,12 @@ static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
409 */ 407 */
410static void nfs_inode_remove_request(struct nfs_page *req) 408static void nfs_inode_remove_request(struct nfs_page *req)
411{ 409{
412 struct inode *inode = req->wb_context->dentry->d_inode; 410 struct inode *inode = req->wb_context->path.dentry->d_inode;
413 struct nfs_inode *nfsi = NFS_I(inode); 411 struct nfs_inode *nfsi = NFS_I(inode);
414 412
415 BUG_ON (!NFS_WBACK_BUSY(req)); 413 BUG_ON (!NFS_WBACK_BUSY(req));
416 414
417 spin_lock(&nfsi->req_lock); 415 spin_lock(&inode->i_lock);
418 set_page_private(req->wb_page, 0); 416 set_page_private(req->wb_page, 0);
419 ClearPagePrivate(req->wb_page); 417 ClearPagePrivate(req->wb_page);
420 radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index); 418 radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
@@ -422,11 +420,11 @@ static void nfs_inode_remove_request(struct nfs_page *req)
422 __set_page_dirty_nobuffers(req->wb_page); 420 __set_page_dirty_nobuffers(req->wb_page);
423 nfsi->npages--; 421 nfsi->npages--;
424 if (!nfsi->npages) { 422 if (!nfsi->npages) {
425 spin_unlock(&nfsi->req_lock); 423 spin_unlock(&inode->i_lock);
426 nfs_end_data_update(inode); 424 nfs_end_data_update(inode);
427 iput(inode); 425 iput(inode);
428 } else 426 } else
429 spin_unlock(&nfsi->req_lock); 427 spin_unlock(&inode->i_lock);
430 nfs_clear_request(req); 428 nfs_clear_request(req);
431 nfs_release_request(req); 429 nfs_release_request(req);
432} 430}
@@ -457,14 +455,16 @@ nfs_dirty_request(struct nfs_page *req)
457static void 455static void
458nfs_mark_request_commit(struct nfs_page *req) 456nfs_mark_request_commit(struct nfs_page *req)
459{ 457{
460 struct inode *inode = req->wb_context->dentry->d_inode; 458 struct inode *inode = req->wb_context->path.dentry->d_inode;
461 struct nfs_inode *nfsi = NFS_I(inode); 459 struct nfs_inode *nfsi = NFS_I(inode);
462 460
463 spin_lock(&nfsi->req_lock); 461 spin_lock(&inode->i_lock);
464 nfs_list_add_request(req, &nfsi->commit);
465 nfsi->ncommit++; 462 nfsi->ncommit++;
466 set_bit(PG_NEED_COMMIT, &(req)->wb_flags); 463 set_bit(PG_NEED_COMMIT, &(req)->wb_flags);
467 spin_unlock(&nfsi->req_lock); 464 radix_tree_tag_set(&nfsi->nfs_page_tree,
465 req->wb_index,
466 NFS_PAGE_TAG_COMMIT);
467 spin_unlock(&inode->i_lock);
468 inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); 468 inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
469 __mark_inode_dirty(inode, I_DIRTY_DATASYNC); 469 __mark_inode_dirty(inode, I_DIRTY_DATASYNC);
470} 470}
@@ -526,18 +526,18 @@ static int nfs_wait_on_requests_locked(struct inode *inode, pgoff_t idx_start, u
526 idx_end = idx_start + npages - 1; 526 idx_end = idx_start + npages - 1;
527 527
528 next = idx_start; 528 next = idx_start;
529 while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_WRITEBACK)) { 529 while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_LOCKED)) {
530 if (req->wb_index > idx_end) 530 if (req->wb_index > idx_end)
531 break; 531 break;
532 532
533 next = req->wb_index + 1; 533 next = req->wb_index + 1;
534 BUG_ON(!NFS_WBACK_BUSY(req)); 534 BUG_ON(!NFS_WBACK_BUSY(req));
535 535
536 atomic_inc(&req->wb_count); 536 kref_get(&req->wb_kref);
537 spin_unlock(&nfsi->req_lock); 537 spin_unlock(&inode->i_lock);
538 error = nfs_wait_on_request(req); 538 error = nfs_wait_on_request(req);
539 nfs_release_request(req); 539 nfs_release_request(req);
540 spin_lock(&nfsi->req_lock); 540 spin_lock(&inode->i_lock);
541 if (error < 0) 541 if (error < 0)
542 return error; 542 return error;
543 res++; 543 res++;
@@ -577,10 +577,9 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst, pgoff_t idx_start, u
577 int res = 0; 577 int res = 0;
578 578
579 if (nfsi->ncommit != 0) { 579 if (nfsi->ncommit != 0) {
580 res = nfs_scan_list(nfsi, &nfsi->commit, dst, idx_start, npages); 580 res = nfs_scan_list(nfsi, dst, idx_start, npages,
581 NFS_PAGE_TAG_COMMIT);
581 nfsi->ncommit -= res; 582 nfsi->ncommit -= res;
582 if ((nfsi->ncommit == 0) != list_empty(&nfsi->commit))
583 printk(KERN_ERR "NFS: desynchronized value of nfs_i.ncommit.\n");
584 } 583 }
585 return res; 584 return res;
586} 585}
@@ -603,7 +602,6 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
603{ 602{
604 struct address_space *mapping = page->mapping; 603 struct address_space *mapping = page->mapping;
605 struct inode *inode = mapping->host; 604 struct inode *inode = mapping->host;
606 struct nfs_inode *nfsi = NFS_I(inode);
607 struct nfs_page *req, *new = NULL; 605 struct nfs_page *req, *new = NULL;
608 pgoff_t rqend, end; 606 pgoff_t rqend, end;
609 607
@@ -613,13 +611,13 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
613 /* Loop over all inode entries and see if we find 611 /* Loop over all inode entries and see if we find
614 * A request for the page we wish to update 612 * A request for the page we wish to update
615 */ 613 */
616 spin_lock(&nfsi->req_lock); 614 spin_lock(&inode->i_lock);
617 req = nfs_page_find_request_locked(page); 615 req = nfs_page_find_request_locked(page);
618 if (req) { 616 if (req) {
619 if (!nfs_lock_request_dontget(req)) { 617 if (!nfs_lock_request_dontget(req)) {
620 int error; 618 int error;
621 619
622 spin_unlock(&nfsi->req_lock); 620 spin_unlock(&inode->i_lock);
623 error = nfs_wait_on_request(req); 621 error = nfs_wait_on_request(req);
624 nfs_release_request(req); 622 nfs_release_request(req);
625 if (error < 0) { 623 if (error < 0) {
@@ -629,7 +627,7 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
629 } 627 }
630 continue; 628 continue;
631 } 629 }
632 spin_unlock(&nfsi->req_lock); 630 spin_unlock(&inode->i_lock);
633 if (new) 631 if (new)
634 nfs_release_request(new); 632 nfs_release_request(new);
635 break; 633 break;
@@ -640,14 +638,14 @@ static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
640 nfs_lock_request_dontget(new); 638 nfs_lock_request_dontget(new);
641 error = nfs_inode_add_request(inode, new); 639 error = nfs_inode_add_request(inode, new);
642 if (error) { 640 if (error) {
643 spin_unlock(&nfsi->req_lock); 641 spin_unlock(&inode->i_lock);
644 nfs_unlock_request(new); 642 nfs_unlock_request(new);
645 return ERR_PTR(error); 643 return ERR_PTR(error);
646 } 644 }
647 spin_unlock(&nfsi->req_lock); 645 spin_unlock(&inode->i_lock);
648 return new; 646 return new;
649 } 647 }
650 spin_unlock(&nfsi->req_lock); 648 spin_unlock(&inode->i_lock);
651 649
652 new = nfs_create_request(ctx, inode, page, offset, bytes); 650 new = nfs_create_request(ctx, inode, page, offset, bytes);
653 if (IS_ERR(new)) 651 if (IS_ERR(new))
@@ -751,12 +749,17 @@ int nfs_updatepage(struct file *file, struct page *page,
751static void nfs_writepage_release(struct nfs_page *req) 749static void nfs_writepage_release(struct nfs_page *req)
752{ 750{
753 751
754 if (PageError(req->wb_page) || !nfs_reschedule_unstable_write(req)) { 752 if (PageError(req->wb_page)) {
753 nfs_end_page_writeback(req->wb_page);
754 nfs_inode_remove_request(req);
755 } else if (!nfs_reschedule_unstable_write(req)) {
756 /* Set the PG_uptodate flag */
757 nfs_mark_uptodate(req->wb_page, req->wb_pgbase, req->wb_bytes);
755 nfs_end_page_writeback(req->wb_page); 758 nfs_end_page_writeback(req->wb_page);
756 nfs_inode_remove_request(req); 759 nfs_inode_remove_request(req);
757 } else 760 } else
758 nfs_end_page_writeback(req->wb_page); 761 nfs_end_page_writeback(req->wb_page);
759 nfs_clear_page_writeback(req); 762 nfs_clear_page_tag_locked(req);
760} 763}
761 764
762static inline int flush_task_priority(int how) 765static inline int flush_task_priority(int how)
@@ -786,7 +789,7 @@ static void nfs_write_rpcsetup(struct nfs_page *req,
786 * NB: take care not to mess about with data->commit et al. */ 789 * NB: take care not to mess about with data->commit et al. */
787 790
788 data->req = req; 791 data->req = req;
789 data->inode = inode = req->wb_context->dentry->d_inode; 792 data->inode = inode = req->wb_context->path.dentry->d_inode;
790 data->cred = req->wb_context->cred; 793 data->cred = req->wb_context->cred;
791 794
792 data->args.fh = NFS_FH(inode); 795 data->args.fh = NFS_FH(inode);
@@ -885,7 +888,7 @@ out_bad:
885 } 888 }
886 nfs_redirty_request(req); 889 nfs_redirty_request(req);
887 nfs_end_page_writeback(req->wb_page); 890 nfs_end_page_writeback(req->wb_page);
888 nfs_clear_page_writeback(req); 891 nfs_clear_page_tag_locked(req);
889 return -ENOMEM; 892 return -ENOMEM;
890} 893}
891 894
@@ -928,7 +931,7 @@ static int nfs_flush_one(struct inode *inode, struct list_head *head, unsigned i
928 nfs_list_remove_request(req); 931 nfs_list_remove_request(req);
929 nfs_redirty_request(req); 932 nfs_redirty_request(req);
930 nfs_end_page_writeback(req->wb_page); 933 nfs_end_page_writeback(req->wb_page);
931 nfs_clear_page_writeback(req); 934 nfs_clear_page_tag_locked(req);
932 } 935 }
933 return -ENOMEM; 936 return -ENOMEM;
934} 937}
@@ -954,8 +957,8 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
954 struct page *page = req->wb_page; 957 struct page *page = req->wb_page;
955 958
956 dprintk("NFS: write (%s/%Ld %d@%Ld)", 959 dprintk("NFS: write (%s/%Ld %d@%Ld)",
957 req->wb_context->dentry->d_inode->i_sb->s_id, 960 req->wb_context->path.dentry->d_inode->i_sb->s_id,
958 (long long)NFS_FILEID(req->wb_context->dentry->d_inode), 961 (long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
959 req->wb_bytes, 962 req->wb_bytes,
960 (long long)req_offset(req)); 963 (long long)req_offset(req));
961 964
@@ -970,9 +973,9 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
970 } 973 }
971 974
972 if (nfs_write_need_commit(data)) { 975 if (nfs_write_need_commit(data)) {
973 spinlock_t *req_lock = &NFS_I(page->mapping->host)->req_lock; 976 struct inode *inode = page->mapping->host;
974 977
975 spin_lock(req_lock); 978 spin_lock(&inode->i_lock);
976 if (test_bit(PG_NEED_RESCHED, &req->wb_flags)) { 979 if (test_bit(PG_NEED_RESCHED, &req->wb_flags)) {
977 /* Do nothing we need to resend the writes */ 980 /* Do nothing we need to resend the writes */
978 } else if (!test_and_set_bit(PG_NEED_COMMIT, &req->wb_flags)) { 981 } else if (!test_and_set_bit(PG_NEED_COMMIT, &req->wb_flags)) {
@@ -983,7 +986,7 @@ static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
983 clear_bit(PG_NEED_COMMIT, &req->wb_flags); 986 clear_bit(PG_NEED_COMMIT, &req->wb_flags);
984 dprintk(" server reboot detected\n"); 987 dprintk(" server reboot detected\n");
985 } 988 }
986 spin_unlock(req_lock); 989 spin_unlock(&inode->i_lock);
987 } else 990 } else
988 dprintk(" OK\n"); 991 dprintk(" OK\n");
989 992
@@ -1020,8 +1023,8 @@ static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
1020 page = req->wb_page; 1023 page = req->wb_page;
1021 1024
1022 dprintk("NFS: write (%s/%Ld %d@%Ld)", 1025 dprintk("NFS: write (%s/%Ld %d@%Ld)",
1023 req->wb_context->dentry->d_inode->i_sb->s_id, 1026 req->wb_context->path.dentry->d_inode->i_sb->s_id,
1024 (long long)NFS_FILEID(req->wb_context->dentry->d_inode), 1027 (long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
1025 req->wb_bytes, 1028 req->wb_bytes,
1026 (long long)req_offset(req)); 1029 (long long)req_offset(req));
1027 1030
@@ -1039,12 +1042,14 @@ static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
1039 dprintk(" marked for commit\n"); 1042 dprintk(" marked for commit\n");
1040 goto next; 1043 goto next;
1041 } 1044 }
1045 /* Set the PG_uptodate flag? */
1046 nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
1042 dprintk(" OK\n"); 1047 dprintk(" OK\n");
1043remove_request: 1048remove_request:
1044 nfs_end_page_writeback(page); 1049 nfs_end_page_writeback(page);
1045 nfs_inode_remove_request(req); 1050 nfs_inode_remove_request(req);
1046 next: 1051 next:
1047 nfs_clear_page_writeback(req); 1052 nfs_clear_page_tag_locked(req);
1048 } 1053 }
1049} 1054}
1050 1055
@@ -1157,7 +1162,7 @@ static void nfs_commit_rpcsetup(struct list_head *head,
1157 1162
1158 list_splice_init(head, &data->pages); 1163 list_splice_init(head, &data->pages);
1159 first = nfs_list_entry(data->pages.next); 1164 first = nfs_list_entry(data->pages.next);
1160 inode = first->wb_context->dentry->d_inode; 1165 inode = first->wb_context->path.dentry->d_inode;
1161 1166
1162 data->inode = inode; 1167 data->inode = inode;
1163 data->cred = first->wb_context->cred; 1168 data->cred = first->wb_context->cred;
@@ -1207,7 +1212,7 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how)
1207 nfs_list_remove_request(req); 1212 nfs_list_remove_request(req);
1208 nfs_mark_request_commit(req); 1213 nfs_mark_request_commit(req);
1209 dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); 1214 dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1210 nfs_clear_page_writeback(req); 1215 nfs_clear_page_tag_locked(req);
1211 } 1216 }
1212 return -ENOMEM; 1217 return -ENOMEM;
1213} 1218}
@@ -1234,8 +1239,8 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
1234 dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS); 1239 dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1235 1240
1236 dprintk("NFS: commit (%s/%Ld %d@%Ld)", 1241 dprintk("NFS: commit (%s/%Ld %d@%Ld)",
1237 req->wb_context->dentry->d_inode->i_sb->s_id, 1242 req->wb_context->path.dentry->d_inode->i_sb->s_id,
1238 (long long)NFS_FILEID(req->wb_context->dentry->d_inode), 1243 (long long)NFS_FILEID(req->wb_context->path.dentry->d_inode),
1239 req->wb_bytes, 1244 req->wb_bytes,
1240 (long long)req_offset(req)); 1245 (long long)req_offset(req));
1241 if (task->tk_status < 0) { 1246 if (task->tk_status < 0) {
@@ -1249,6 +1254,9 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
1249 * returned by the server against all stored verfs. */ 1254 * returned by the server against all stored verfs. */
1250 if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) { 1255 if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) {
1251 /* We have a match */ 1256 /* We have a match */
1257 /* Set the PG_uptodate flag */
1258 nfs_mark_uptodate(req->wb_page, req->wb_pgbase,
1259 req->wb_bytes);
1252 nfs_inode_remove_request(req); 1260 nfs_inode_remove_request(req);
1253 dprintk(" OK\n"); 1261 dprintk(" OK\n");
1254 goto next; 1262 goto next;
@@ -1257,7 +1265,7 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata)
1257 dprintk(" mismatch\n"); 1265 dprintk(" mismatch\n");
1258 nfs_redirty_request(req); 1266 nfs_redirty_request(req);
1259 next: 1267 next:
1260 nfs_clear_page_writeback(req); 1268 nfs_clear_page_tag_locked(req);
1261 } 1269 }
1262} 1270}
1263 1271
@@ -1268,13 +1276,12 @@ static const struct rpc_call_ops nfs_commit_ops = {
1268 1276
1269int nfs_commit_inode(struct inode *inode, int how) 1277int nfs_commit_inode(struct inode *inode, int how)
1270{ 1278{
1271 struct nfs_inode *nfsi = NFS_I(inode);
1272 LIST_HEAD(head); 1279 LIST_HEAD(head);
1273 int res; 1280 int res;
1274 1281
1275 spin_lock(&nfsi->req_lock); 1282 spin_lock(&inode->i_lock);
1276 res = nfs_scan_commit(inode, &head, 0, 0); 1283 res = nfs_scan_commit(inode, &head, 0, 0);
1277 spin_unlock(&nfsi->req_lock); 1284 spin_unlock(&inode->i_lock);
1278 if (res) { 1285 if (res) {
1279 int error = nfs_commit_list(inode, &head, how); 1286 int error = nfs_commit_list(inode, &head, how);
1280 if (error < 0) 1287 if (error < 0)
@@ -1292,7 +1299,6 @@ static inline int nfs_commit_list(struct inode *inode, struct list_head *head, i
1292long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_control *wbc, int how) 1299long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_control *wbc, int how)
1293{ 1300{
1294 struct inode *inode = mapping->host; 1301 struct inode *inode = mapping->host;
1295 struct nfs_inode *nfsi = NFS_I(inode);
1296 pgoff_t idx_start, idx_end; 1302 pgoff_t idx_start, idx_end;
1297 unsigned int npages = 0; 1303 unsigned int npages = 0;
1298 LIST_HEAD(head); 1304 LIST_HEAD(head);
@@ -1314,7 +1320,7 @@ long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_contr
1314 } 1320 }
1315 } 1321 }
1316 how &= ~FLUSH_NOCOMMIT; 1322 how &= ~FLUSH_NOCOMMIT;
1317 spin_lock(&nfsi->req_lock); 1323 spin_lock(&inode->i_lock);
1318 do { 1324 do {
1319 ret = nfs_wait_on_requests_locked(inode, idx_start, npages); 1325 ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
1320 if (ret != 0) 1326 if (ret != 0)
@@ -1325,18 +1331,19 @@ long nfs_sync_mapping_wait(struct address_space *mapping, struct writeback_contr
1325 if (pages == 0) 1331 if (pages == 0)
1326 break; 1332 break;
1327 if (how & FLUSH_INVALIDATE) { 1333 if (how & FLUSH_INVALIDATE) {
1328 spin_unlock(&nfsi->req_lock); 1334 spin_unlock(&inode->i_lock);
1329 nfs_cancel_commit_list(&head); 1335 nfs_cancel_commit_list(&head);
1330 ret = pages; 1336 ret = pages;
1331 spin_lock(&nfsi->req_lock); 1337 spin_lock(&inode->i_lock);
1332 continue; 1338 continue;
1333 } 1339 }
1334 pages += nfs_scan_commit(inode, &head, 0, 0); 1340 pages += nfs_scan_commit(inode, &head, 0, 0);
1335 spin_unlock(&nfsi->req_lock); 1341 spin_unlock(&inode->i_lock);
1336 ret = nfs_commit_list(inode, &head, how); 1342 ret = nfs_commit_list(inode, &head, how);
1337 spin_lock(&nfsi->req_lock); 1343 spin_lock(&inode->i_lock);
1344
1338 } while (ret >= 0); 1345 } while (ret >= 0);
1339 spin_unlock(&nfsi->req_lock); 1346 spin_unlock(&inode->i_lock);
1340 return ret; 1347 return ret;
1341} 1348}
1342 1349
@@ -1430,7 +1437,6 @@ int nfs_set_page_dirty(struct page *page)
1430{ 1437{
1431 struct address_space *mapping = page->mapping; 1438 struct address_space *mapping = page->mapping;
1432 struct inode *inode; 1439 struct inode *inode;
1433 spinlock_t *req_lock;
1434 struct nfs_page *req; 1440 struct nfs_page *req;
1435 int ret; 1441 int ret;
1436 1442
@@ -1439,18 +1445,17 @@ int nfs_set_page_dirty(struct page *page)
1439 inode = mapping->host; 1445 inode = mapping->host;
1440 if (!inode) 1446 if (!inode)
1441 goto out_raced; 1447 goto out_raced;
1442 req_lock = &NFS_I(inode)->req_lock; 1448 spin_lock(&inode->i_lock);
1443 spin_lock(req_lock);
1444 req = nfs_page_find_request_locked(page); 1449 req = nfs_page_find_request_locked(page);
1445 if (req != NULL) { 1450 if (req != NULL) {
1446 /* Mark any existing write requests for flushing */ 1451 /* Mark any existing write requests for flushing */
1447 ret = !test_and_set_bit(PG_NEED_FLUSH, &req->wb_flags); 1452 ret = !test_and_set_bit(PG_NEED_FLUSH, &req->wb_flags);
1448 spin_unlock(req_lock); 1453 spin_unlock(&inode->i_lock);
1449 nfs_release_request(req); 1454 nfs_release_request(req);
1450 return ret; 1455 return ret;
1451 } 1456 }
1452 ret = __set_page_dirty_nobuffers(page); 1457 ret = __set_page_dirty_nobuffers(page);
1453 spin_unlock(req_lock); 1458 spin_unlock(&inode->i_lock);
1454 return ret; 1459 return ret;
1455out_raced: 1460out_raced:
1456 return !TestSetPageDirty(page); 1461 return !TestSetPageDirty(page);
diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
index 864090edc28b..5443c52b57aa 100644
--- a/fs/nfsd/nfs4callback.c
+++ b/fs/nfsd/nfs4callback.c
@@ -394,7 +394,6 @@ nfsd4_probe_callback(struct nfs4_client *clp)
394 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL], 394 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL],
395 .rpc_argp = clp, 395 .rpc_argp = clp,
396 }; 396 };
397 char clientname[16];
398 int status; 397 int status;
399 398
400 if (atomic_read(&cb->cb_set)) 399 if (atomic_read(&cb->cb_set))
@@ -417,11 +416,6 @@ nfsd4_probe_callback(struct nfs4_client *clp)
417 memset(program->stats, 0, sizeof(cb->cb_stat)); 416 memset(program->stats, 0, sizeof(cb->cb_stat));
418 program->stats->program = program; 417 program->stats->program = program;
419 418
420 /* Just here to make some printk's more useful: */
421 snprintf(clientname, sizeof(clientname),
422 "%u.%u.%u.%u", NIPQUAD(addr.sin_addr));
423 args.servername = clientname;
424
425 /* Create RPC client */ 419 /* Create RPC client */
426 cb->cb_client = rpc_create(&args); 420 cb->cb_client = rpc_create(&args);
427 if (IS_ERR(cb->cb_client)) { 421 if (IS_ERR(cb->cb_client)) {
@@ -429,29 +423,23 @@ nfsd4_probe_callback(struct nfs4_client *clp)
429 goto out_err; 423 goto out_err;
430 } 424 }
431 425
432 /* Kick rpciod, put the call on the wire. */
433 if (rpciod_up() != 0)
434 goto out_clnt;
435
436 /* the task holds a reference to the nfs4_client struct */ 426 /* the task holds a reference to the nfs4_client struct */
437 atomic_inc(&clp->cl_count); 427 atomic_inc(&clp->cl_count);
438 428
439 msg.rpc_cred = nfsd4_lookupcred(clp,0); 429 msg.rpc_cred = nfsd4_lookupcred(clp,0);
440 if (IS_ERR(msg.rpc_cred)) 430 if (IS_ERR(msg.rpc_cred))
441 goto out_rpciod; 431 goto out_release_clp;
442 status = rpc_call_async(cb->cb_client, &msg, RPC_TASK_ASYNC, &nfs4_cb_null_ops, NULL); 432 status = rpc_call_async(cb->cb_client, &msg, RPC_TASK_ASYNC, &nfs4_cb_null_ops, NULL);
443 put_rpccred(msg.rpc_cred); 433 put_rpccred(msg.rpc_cred);
444 434
445 if (status != 0) { 435 if (status != 0) {
446 dprintk("NFSD: asynchronous NFSPROC4_CB_NULL failed!\n"); 436 dprintk("NFSD: asynchronous NFSPROC4_CB_NULL failed!\n");
447 goto out_rpciod; 437 goto out_release_clp;
448 } 438 }
449 return; 439 return;
450 440
451out_rpciod: 441out_release_clp:
452 atomic_dec(&clp->cl_count); 442 atomic_dec(&clp->cl_count);
453 rpciod_down();
454out_clnt:
455 rpc_shutdown_client(cb->cb_client); 443 rpc_shutdown_client(cb->cb_client);
456out_err: 444out_err:
457 cb->cb_client = NULL; 445 cb->cb_client = NULL;
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 3cc8ce422ab1..8c52913d7cb6 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -378,7 +378,6 @@ shutdown_callback_client(struct nfs4_client *clp)
378 if (clnt) { 378 if (clnt) {
379 clp->cl_callback.cb_client = NULL; 379 clp->cl_callback.cb_client = NULL;
380 rpc_shutdown_client(clnt); 380 rpc_shutdown_client(clnt);
381 rpciod_down();
382 } 381 }
383} 382}
384 383
diff --git a/include/linux/lockd/lockd.h b/include/linux/lockd/lockd.h
index 05707e2fccae..e2d1ce36b367 100644
--- a/include/linux/lockd/lockd.h
+++ b/include/linux/lockd/lockd.h
@@ -39,6 +39,7 @@
39struct nlm_host { 39struct nlm_host {
40 struct hlist_node h_hash; /* doubly linked list */ 40 struct hlist_node h_hash; /* doubly linked list */
41 struct sockaddr_in h_addr; /* peer address */ 41 struct sockaddr_in h_addr; /* peer address */
42 struct sockaddr_in h_saddr; /* our address (optional) */
42 struct rpc_clnt * h_rpcclnt; /* RPC client to talk to peer */ 43 struct rpc_clnt * h_rpcclnt; /* RPC client to talk to peer */
43 char * h_name; /* remote hostname */ 44 char * h_name; /* remote hostname */
44 u32 h_version; /* interface version */ 45 u32 h_version; /* interface version */
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index 7e7f33a38fc0..8726491de154 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -15,6 +15,7 @@
15 15
16#include <linux/types.h> 16#include <linux/types.h>
17 17
18#define NFS4_BITMAP_SIZE 2
18#define NFS4_VERIFIER_SIZE 8 19#define NFS4_VERIFIER_SIZE 8
19#define NFS4_STATEID_SIZE 16 20#define NFS4_STATEID_SIZE 16
20#define NFS4_FHSIZE 128 21#define NFS4_FHSIZE 128
diff --git a/include/linux/nfs4_mount.h b/include/linux/nfs4_mount.h
index 26b4c83f831d..a0dcf6655657 100644
--- a/include/linux/nfs4_mount.h
+++ b/include/linux/nfs4_mount.h
@@ -65,6 +65,7 @@ struct nfs4_mount_data {
65#define NFS4_MOUNT_NOCTO 0x0010 /* 1 */ 65#define NFS4_MOUNT_NOCTO 0x0010 /* 1 */
66#define NFS4_MOUNT_NOAC 0x0020 /* 1 */ 66#define NFS4_MOUNT_NOAC 0x0020 /* 1 */
67#define NFS4_MOUNT_STRICTLOCK 0x1000 /* 1 */ 67#define NFS4_MOUNT_STRICTLOCK 0x1000 /* 1 */
68#define NFS4_MOUNT_FLAGMASK 0xFFFF 68#define NFS4_MOUNT_UNSHARED 0x8000 /* 1 */
69#define NFS4_MOUNT_FLAGMASK 0x9033
69 70
70#endif 71#endif
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 0543439a97af..c098ae194f79 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -30,7 +30,9 @@
30#ifdef __KERNEL__ 30#ifdef __KERNEL__
31 31
32#include <linux/in.h> 32#include <linux/in.h>
33#include <linux/kref.h>
33#include <linux/mm.h> 34#include <linux/mm.h>
35#include <linux/namei.h>
34#include <linux/pagemap.h> 36#include <linux/pagemap.h>
35#include <linux/rbtree.h> 37#include <linux/rbtree.h>
36#include <linux/rwsem.h> 38#include <linux/rwsem.h>
@@ -69,9 +71,8 @@ struct nfs_access_entry {
69 71
70struct nfs4_state; 72struct nfs4_state;
71struct nfs_open_context { 73struct nfs_open_context {
72 atomic_t count; 74 struct kref kref;
73 struct vfsmount *vfsmnt; 75 struct path path;
74 struct dentry *dentry;
75 struct rpc_cred *cred; 76 struct rpc_cred *cred;
76 struct nfs4_state *state; 77 struct nfs4_state *state;
77 fl_owner_t lockowner; 78 fl_owner_t lockowner;
@@ -155,13 +156,9 @@ struct nfs_inode {
155 /* 156 /*
156 * This is the list of dirty unwritten pages. 157 * This is the list of dirty unwritten pages.
157 */ 158 */
158 spinlock_t req_lock;
159 struct list_head dirty;
160 struct list_head commit;
161 struct radix_tree_root nfs_page_tree; 159 struct radix_tree_root nfs_page_tree;
162 160
163 unsigned int ndirty, 161 unsigned long ncommit,
164 ncommit,
165 npages; 162 npages;
166 163
167 /* Open contexts for shared mmap writes */ 164 /* Open contexts for shared mmap writes */
@@ -187,6 +184,7 @@ struct nfs_inode {
187#define NFS_INO_INVALID_ACCESS 0x0008 /* cached access cred invalid */ 184#define NFS_INO_INVALID_ACCESS 0x0008 /* cached access cred invalid */
188#define NFS_INO_INVALID_ACL 0x0010 /* cached acls are invalid */ 185#define NFS_INO_INVALID_ACL 0x0010 /* cached acls are invalid */
189#define NFS_INO_REVAL_PAGECACHE 0x0020 /* must revalidate pagecache */ 186#define NFS_INO_REVAL_PAGECACHE 0x0020 /* must revalidate pagecache */
187#define NFS_INO_REVAL_FORCED 0x0040 /* force revalidation ignoring a delegation */
190 188
191/* 189/*
192 * Bit offsets in flags field 190 * Bit offsets in flags field
@@ -496,21 +494,18 @@ static inline void nfs3_forget_cached_acls(struct inode *inode)
496 494
497/* 495/*
498 * linux/fs/mount_clnt.c 496 * linux/fs/mount_clnt.c
499 * (Used only by nfsroot module)
500 */ 497 */
501extern int nfsroot_mount(struct sockaddr_in *, char *, struct nfs_fh *, 498extern int nfs_mount(struct sockaddr *, size_t, char *, char *,
502 int, int); 499 int, int, struct nfs_fh *);
503 500
504/* 501/*
505 * inline functions 502 * inline functions
506 */ 503 */
507 504
508static inline loff_t 505static inline loff_t nfs_size_to_loff_t(__u64 size)
509nfs_size_to_loff_t(__u64 size)
510{ 506{
511 loff_t maxsz = (((loff_t) ULONG_MAX) << PAGE_CACHE_SHIFT) + PAGE_CACHE_SIZE - 1; 507 if (size > (__u64) OFFSET_MAX - 1)
512 if (size > maxsz) 508 return OFFSET_MAX - 1;
513 return maxsz;
514 return (loff_t) size; 509 return (loff_t) size;
515} 510}
516 511
@@ -557,6 +552,7 @@ extern void * nfs_root_data(void);
557#define NFSDBG_ROOT 0x0080 552#define NFSDBG_ROOT 0x0080
558#define NFSDBG_CALLBACK 0x0100 553#define NFSDBG_CALLBACK 0x0100
559#define NFSDBG_CLIENT 0x0200 554#define NFSDBG_CLIENT 0x0200
555#define NFSDBG_MOUNT 0x0400
560#define NFSDBG_ALL 0xFFFF 556#define NFSDBG_ALL 0xFFFF
561 557
562#ifdef __KERNEL__ 558#ifdef __KERNEL__
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 52b4378311c8..0cac49bc0955 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -16,7 +16,6 @@ struct nfs_client {
16#define NFS_CS_INITING 1 /* busy initialising */ 16#define NFS_CS_INITING 1 /* busy initialising */
17 int cl_nfsversion; /* NFS protocol version */ 17 int cl_nfsversion; /* NFS protocol version */
18 unsigned long cl_res_state; /* NFS resources state */ 18 unsigned long cl_res_state; /* NFS resources state */
19#define NFS_CS_RPCIOD 0 /* - rpciod started */
20#define NFS_CS_CALLBACK 1 /* - callback started */ 19#define NFS_CS_CALLBACK 1 /* - callback started */
21#define NFS_CS_IDMAP 2 /* - idmap started */ 20#define NFS_CS_IDMAP 2 /* - idmap started */
22#define NFS_CS_RENEWD 3 /* - renewd started */ 21#define NFS_CS_RENEWD 3 /* - renewd started */
@@ -35,7 +34,8 @@ struct nfs_client {
35 nfs4_verifier cl_confirm; 34 nfs4_verifier cl_confirm;
36 unsigned long cl_state; 35 unsigned long cl_state;
37 36
38 u32 cl_lockowner_id; 37 struct rb_root cl_openowner_id;
38 struct rb_root cl_lockowner_id;
39 39
40 /* 40 /*
41 * The following rwsem ensures exclusive access to the server 41 * The following rwsem ensures exclusive access to the server
@@ -44,9 +44,7 @@ struct nfs_client {
44 struct rw_semaphore cl_sem; 44 struct rw_semaphore cl_sem;
45 45
46 struct list_head cl_delegations; 46 struct list_head cl_delegations;
47 struct list_head cl_state_owners; 47 struct rb_root cl_state_owners;
48 struct list_head cl_unused;
49 int cl_nunused;
50 spinlock_t cl_lock; 48 spinlock_t cl_lock;
51 49
52 unsigned long cl_lease_time; 50 unsigned long cl_lease_time;
diff --git a/include/linux/nfs_mount.h b/include/linux/nfs_mount.h
index cc8b9c59acb8..a3ade89a64d2 100644
--- a/include/linux/nfs_mount.h
+++ b/include/linux/nfs_mount.h
@@ -37,7 +37,7 @@ struct nfs_mount_data {
37 int acdirmin; /* 1 */ 37 int acdirmin; /* 1 */
38 int acdirmax; /* 1 */ 38 int acdirmax; /* 1 */
39 struct sockaddr_in addr; /* 1 */ 39 struct sockaddr_in addr; /* 1 */
40 char hostname[256]; /* 1 */ 40 char hostname[NFS_MAXNAMLEN + 1]; /* 1 */
41 int namlen; /* 2 */ 41 int namlen; /* 2 */
42 unsigned int bsize; /* 3 */ 42 unsigned int bsize; /* 3 */
43 struct nfs3_fh root; /* 4 */ 43 struct nfs3_fh root; /* 4 */
@@ -62,6 +62,7 @@ struct nfs_mount_data {
62#define NFS_MOUNT_STRICTLOCK 0x1000 /* reserved for NFSv4 */ 62#define NFS_MOUNT_STRICTLOCK 0x1000 /* reserved for NFSv4 */
63#define NFS_MOUNT_SECFLAVOUR 0x2000 /* 5 */ 63#define NFS_MOUNT_SECFLAVOUR 0x2000 /* 5 */
64#define NFS_MOUNT_NORDIRPLUS 0x4000 /* 5 */ 64#define NFS_MOUNT_NORDIRPLUS 0x4000 /* 5 */
65#define NFS_MOUNT_UNSHARED 0x8000 /* 5 */
65#define NFS_MOUNT_FLAGMASK 0xFFFF 66#define NFS_MOUNT_FLAGMASK 0xFFFF
66 67
67#endif 68#endif
diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h
index bd193af80162..78e60798d10e 100644
--- a/include/linux/nfs_page.h
+++ b/include/linux/nfs_page.h
@@ -16,12 +16,13 @@
16#include <linux/sunrpc/auth.h> 16#include <linux/sunrpc/auth.h>
17#include <linux/nfs_xdr.h> 17#include <linux/nfs_xdr.h>
18 18
19#include <asm/atomic.h> 19#include <linux/kref.h>
20 20
21/* 21/*
22 * Valid flags for the radix tree 22 * Valid flags for the radix tree
23 */ 23 */
24#define NFS_PAGE_TAG_WRITEBACK 0 24#define NFS_PAGE_TAG_LOCKED 0
25#define NFS_PAGE_TAG_COMMIT 1
25 26
26/* 27/*
27 * Valid flags for a dirty buffer 28 * Valid flags for a dirty buffer
@@ -33,8 +34,7 @@
33 34
34struct nfs_inode; 35struct nfs_inode;
35struct nfs_page { 36struct nfs_page {
36 struct list_head wb_list, /* Defines state of page: */ 37 struct list_head wb_list; /* Defines state of page: */
37 *wb_list_head; /* read/write/commit */
38 struct page *wb_page; /* page to read in/write out */ 38 struct page *wb_page; /* page to read in/write out */
39 struct nfs_open_context *wb_context; /* File state context info */ 39 struct nfs_open_context *wb_context; /* File state context info */
40 atomic_t wb_complete; /* i/os we're waiting for */ 40 atomic_t wb_complete; /* i/os we're waiting for */
@@ -42,7 +42,7 @@ struct nfs_page {
42 unsigned int wb_offset, /* Offset & ~PAGE_CACHE_MASK */ 42 unsigned int wb_offset, /* Offset & ~PAGE_CACHE_MASK */
43 wb_pgbase, /* Start of page data */ 43 wb_pgbase, /* Start of page data */
44 wb_bytes; /* Length of request */ 44 wb_bytes; /* Length of request */
45 atomic_t wb_count; /* reference count */ 45 struct kref wb_kref; /* reference count */
46 unsigned long wb_flags; 46 unsigned long wb_flags;
47 struct nfs_writeverf wb_verf; /* Commit cookie */ 47 struct nfs_writeverf wb_verf; /* Commit cookie */
48}; 48};
@@ -71,8 +71,8 @@ extern void nfs_clear_request(struct nfs_page *req);
71extern void nfs_release_request(struct nfs_page *req); 71extern void nfs_release_request(struct nfs_page *req);
72 72
73 73
74extern int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head, struct list_head *dst, 74extern int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *dst,
75 pgoff_t idx_start, unsigned int npages); 75 pgoff_t idx_start, unsigned int npages, int tag);
76extern void nfs_pageio_init(struct nfs_pageio_descriptor *desc, 76extern void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
77 struct inode *inode, 77 struct inode *inode,
78 int (*doio)(struct inode *, struct list_head *, unsigned int, size_t, int), 78 int (*doio)(struct inode *, struct list_head *, unsigned int, size_t, int),
@@ -84,12 +84,11 @@ extern void nfs_pageio_complete(struct nfs_pageio_descriptor *desc);
84extern void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *, pgoff_t); 84extern void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *, pgoff_t);
85extern int nfs_wait_on_request(struct nfs_page *); 85extern int nfs_wait_on_request(struct nfs_page *);
86extern void nfs_unlock_request(struct nfs_page *req); 86extern void nfs_unlock_request(struct nfs_page *req);
87extern int nfs_set_page_writeback_locked(struct nfs_page *req); 87extern void nfs_clear_page_tag_locked(struct nfs_page *req);
88extern void nfs_clear_page_writeback(struct nfs_page *req);
89 88
90 89
91/* 90/*
92 * Lock the page of an asynchronous request without incrementing the wb_count 91 * Lock the page of an asynchronous request without getting a new reference
93 */ 92 */
94static inline int 93static inline int
95nfs_lock_request_dontget(struct nfs_page *req) 94nfs_lock_request_dontget(struct nfs_page *req)
@@ -98,14 +97,14 @@ nfs_lock_request_dontget(struct nfs_page *req)
98} 97}
99 98
100/* 99/*
101 * Lock the page of an asynchronous request 100 * Lock the page of an asynchronous request and take a reference
102 */ 101 */
103static inline int 102static inline int
104nfs_lock_request(struct nfs_page *req) 103nfs_lock_request(struct nfs_page *req)
105{ 104{
106 if (test_and_set_bit(PG_BUSY, &req->wb_flags)) 105 if (test_and_set_bit(PG_BUSY, &req->wb_flags))
107 return 0; 106 return 0;
108 atomic_inc(&req->wb_count); 107 kref_get(&req->wb_kref);
109 return 1; 108 return 1;
110} 109}
111 110
@@ -118,7 +117,6 @@ static inline void
118nfs_list_add_request(struct nfs_page *req, struct list_head *head) 117nfs_list_add_request(struct nfs_page *req, struct list_head *head)
119{ 118{
120 list_add_tail(&req->wb_list, head); 119 list_add_tail(&req->wb_list, head);
121 req->wb_list_head = head;
122} 120}
123 121
124 122
@@ -132,7 +130,6 @@ nfs_list_remove_request(struct nfs_page *req)
132 if (list_empty(&req->wb_list)) 130 if (list_empty(&req->wb_list))
133 return; 131 return;
134 list_del_init(&req->wb_list); 132 list_del_init(&req->wb_list);
135 req->wb_list_head = NULL;
136} 133}
137 134
138static inline struct nfs_page * 135static inline struct nfs_page *
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 10c26ed0db71..38d77681cf27 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -119,7 +119,7 @@ struct nfs_openargs {
119 struct nfs_seqid * seqid; 119 struct nfs_seqid * seqid;
120 int open_flags; 120 int open_flags;
121 __u64 clientid; 121 __u64 clientid;
122 __u32 id; 122 __u64 id;
123 union { 123 union {
124 struct iattr * attrs; /* UNCHECKED, GUARDED */ 124 struct iattr * attrs; /* UNCHECKED, GUARDED */
125 nfs4_verifier verifier; /* EXCLUSIVE */ 125 nfs4_verifier verifier; /* EXCLUSIVE */
@@ -144,6 +144,7 @@ struct nfs_openres {
144 nfs4_stateid delegation; 144 nfs4_stateid delegation;
145 __u32 do_recall; 145 __u32 do_recall;
146 __u64 maxsize; 146 __u64 maxsize;
147 __u32 attrset[NFS4_BITMAP_SIZE];
147}; 148};
148 149
149/* 150/*
@@ -180,7 +181,7 @@ struct nfs_closeres {
180 * */ 181 * */
181struct nfs_lowner { 182struct nfs_lowner {
182 __u64 clientid; 183 __u64 clientid;
183 u32 id; 184 __u64 id;
184}; 185};
185 186
186struct nfs_lock_args { 187struct nfs_lock_args {
diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h
index 534cdc7be58d..7a69ca3bebaf 100644
--- a/include/linux/sunrpc/auth.h
+++ b/include/linux/sunrpc/auth.h
@@ -16,6 +16,7 @@
16#include <linux/sunrpc/xdr.h> 16#include <linux/sunrpc/xdr.h>
17 17
18#include <asm/atomic.h> 18#include <asm/atomic.h>
19#include <linux/rcupdate.h>
19 20
20/* size of the nodename buffer */ 21/* size of the nodename buffer */
21#define UNX_MAXNODENAME 32 22#define UNX_MAXNODENAME 32
@@ -30,22 +31,28 @@ struct auth_cred {
30/* 31/*
31 * Client user credentials 32 * Client user credentials
32 */ 33 */
34struct rpc_auth;
35struct rpc_credops;
33struct rpc_cred { 36struct rpc_cred {
34 struct hlist_node cr_hash; /* hash chain */ 37 struct hlist_node cr_hash; /* hash chain */
35 struct rpc_credops * cr_ops; 38 struct list_head cr_lru; /* lru garbage collection */
36 unsigned long cr_expire; /* when to gc */ 39 struct rcu_head cr_rcu;
37 atomic_t cr_count; /* ref count */ 40 struct rpc_auth * cr_auth;
38 unsigned short cr_flags; /* various flags */ 41 const struct rpc_credops *cr_ops;
39#ifdef RPC_DEBUG 42#ifdef RPC_DEBUG
40 unsigned long cr_magic; /* 0x0f4aa4f0 */ 43 unsigned long cr_magic; /* 0x0f4aa4f0 */
41#endif 44#endif
45 unsigned long cr_expire; /* when to gc */
46 unsigned long cr_flags; /* various flags */
47 atomic_t cr_count; /* ref count */
42 48
43 uid_t cr_uid; 49 uid_t cr_uid;
44 50
45 /* per-flavor data */ 51 /* per-flavor data */
46}; 52};
47#define RPCAUTH_CRED_NEW 0x0001 53#define RPCAUTH_CRED_NEW 0
48#define RPCAUTH_CRED_UPTODATE 0x0002 54#define RPCAUTH_CRED_UPTODATE 1
55#define RPCAUTH_CRED_HASHED 2
49 56
50#define RPCAUTH_CRED_MAGIC 0x0f4aa4f0 57#define RPCAUTH_CRED_MAGIC 0x0f4aa4f0
51 58
@@ -56,10 +63,10 @@ struct rpc_cred {
56#define RPC_CREDCACHE_MASK (RPC_CREDCACHE_NR - 1) 63#define RPC_CREDCACHE_MASK (RPC_CREDCACHE_NR - 1)
57struct rpc_cred_cache { 64struct rpc_cred_cache {
58 struct hlist_head hashtable[RPC_CREDCACHE_NR]; 65 struct hlist_head hashtable[RPC_CREDCACHE_NR];
59 unsigned long nextgc; /* next garbage collection */ 66 spinlock_t lock;
60 unsigned long expire; /* cache expiry interval */
61}; 67};
62 68
69struct rpc_authops;
63struct rpc_auth { 70struct rpc_auth {
64 unsigned int au_cslack; /* call cred size estimate */ 71 unsigned int au_cslack; /* call cred size estimate */
65 /* guess at number of u32's auth adds before 72 /* guess at number of u32's auth adds before
@@ -69,7 +76,7 @@ struct rpc_auth {
69 unsigned int au_verfsize; 76 unsigned int au_verfsize;
70 77
71 unsigned int au_flags; /* various flags */ 78 unsigned int au_flags; /* various flags */
72 struct rpc_authops * au_ops; /* operations */ 79 const struct rpc_authops *au_ops; /* operations */
73 rpc_authflavor_t au_flavor; /* pseudoflavor (note may 80 rpc_authflavor_t au_flavor; /* pseudoflavor (note may
74 * differ from the flavor in 81 * differ from the flavor in
75 * au_ops->au_flavor in gss 82 * au_ops->au_flavor in gss
@@ -115,17 +122,19 @@ struct rpc_credops {
115 void *, __be32 *, void *); 122 void *, __be32 *, void *);
116}; 123};
117 124
118extern struct rpc_authops authunix_ops; 125extern const struct rpc_authops authunix_ops;
119extern struct rpc_authops authnull_ops; 126extern const struct rpc_authops authnull_ops;
120#ifdef CONFIG_SUNRPC_SECURE 127
121extern struct rpc_authops authdes_ops; 128void __init rpc_init_authunix(void);
122#endif 129void __init rpcauth_init_module(void);
130void __exit rpcauth_remove_module(void);
123 131
124int rpcauth_register(struct rpc_authops *); 132int rpcauth_register(const struct rpc_authops *);
125int rpcauth_unregister(struct rpc_authops *); 133int rpcauth_unregister(const struct rpc_authops *);
126struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *); 134struct rpc_auth * rpcauth_create(rpc_authflavor_t, struct rpc_clnt *);
127void rpcauth_destroy(struct rpc_auth *); 135void rpcauth_release(struct rpc_auth *);
128struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); 136struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int);
137void rpcauth_init_cred(struct rpc_cred *, const struct auth_cred *, struct rpc_auth *, const struct rpc_credops *);
129struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int); 138struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int);
130struct rpc_cred * rpcauth_bindcred(struct rpc_task *); 139struct rpc_cred * rpcauth_bindcred(struct rpc_task *);
131void rpcauth_holdcred(struct rpc_task *); 140void rpcauth_holdcred(struct rpc_task *);
@@ -138,8 +147,9 @@ int rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp,
138int rpcauth_refreshcred(struct rpc_task *); 147int rpcauth_refreshcred(struct rpc_task *);
139void rpcauth_invalcred(struct rpc_task *); 148void rpcauth_invalcred(struct rpc_task *);
140int rpcauth_uptodatecred(struct rpc_task *); 149int rpcauth_uptodatecred(struct rpc_task *);
141int rpcauth_init_credcache(struct rpc_auth *, unsigned long); 150int rpcauth_init_credcache(struct rpc_auth *);
142void rpcauth_free_credcache(struct rpc_auth *); 151void rpcauth_destroy_credcache(struct rpc_auth *);
152void rpcauth_clear_credcache(struct rpc_cred_cache *);
143 153
144static inline 154static inline
145struct rpc_cred * get_rpccred(struct rpc_cred *cred) 155struct rpc_cred * get_rpccred(struct rpc_cred *cred)
diff --git a/include/linux/sunrpc/auth_gss.h b/include/linux/sunrpc/auth_gss.h
index 2db2fbf34947..67658e17a375 100644
--- a/include/linux/sunrpc/auth_gss.h
+++ b/include/linux/sunrpc/auth_gss.h
@@ -75,6 +75,7 @@ struct gss_cl_ctx {
75 struct xdr_netobj gc_wire_ctx; 75 struct xdr_netobj gc_wire_ctx;
76 u32 gc_win; 76 u32 gc_win;
77 unsigned long gc_expiry; 77 unsigned long gc_expiry;
78 struct rcu_head gc_rcu;
78}; 79};
79 80
80struct gss_upcall_msg; 81struct gss_upcall_msg;
@@ -85,11 +86,6 @@ struct gss_cred {
85 struct gss_upcall_msg *gc_upcall; 86 struct gss_upcall_msg *gc_upcall;
86}; 87};
87 88
88#define gc_uid gc_base.cr_uid
89#define gc_count gc_base.cr_count
90#define gc_flags gc_base.cr_flags
91#define gc_expire gc_base.cr_expire
92
93#endif /* __KERNEL__ */ 89#endif /* __KERNEL__ */
94#endif /* _LINUX_SUNRPC_AUTH_GSS_H */ 90#endif /* _LINUX_SUNRPC_AUTH_GSS_H */
95 91
diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index 66611423c8ee..c0d9d14983b3 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -24,8 +24,10 @@ struct rpc_inode;
24 * The high-level client handle 24 * The high-level client handle
25 */ 25 */
26struct rpc_clnt { 26struct rpc_clnt {
27 atomic_t cl_count; /* Number of clones */ 27 struct kref cl_kref; /* Number of references */
28 atomic_t cl_users; /* number of references */ 28 struct list_head cl_clients; /* Global list of clients */
29 struct list_head cl_tasks; /* List of tasks */
30 spinlock_t cl_lock; /* spinlock */
29 struct rpc_xprt * cl_xprt; /* transport */ 31 struct rpc_xprt * cl_xprt; /* transport */
30 struct rpc_procinfo * cl_procinfo; /* procedure info */ 32 struct rpc_procinfo * cl_procinfo; /* procedure info */
31 u32 cl_prog, /* RPC program number */ 33 u32 cl_prog, /* RPC program number */
@@ -41,9 +43,7 @@ struct rpc_clnt {
41 unsigned int cl_softrtry : 1,/* soft timeouts */ 43 unsigned int cl_softrtry : 1,/* soft timeouts */
42 cl_intr : 1,/* interruptible */ 44 cl_intr : 1,/* interruptible */
43 cl_discrtry : 1,/* disconnect before retry */ 45 cl_discrtry : 1,/* disconnect before retry */
44 cl_autobind : 1,/* use getport() */ 46 cl_autobind : 1;/* use getport() */
45 cl_oneshot : 1,/* dispose after use */
46 cl_dead : 1;/* abandoned */
47 47
48 struct rpc_rtt * cl_rtt; /* RTO estimator data */ 48 struct rpc_rtt * cl_rtt; /* RTO estimator data */
49 49
@@ -98,6 +98,7 @@ struct rpc_create_args {
98 int protocol; 98 int protocol;
99 struct sockaddr *address; 99 struct sockaddr *address;
100 size_t addrsize; 100 size_t addrsize;
101 struct sockaddr *saddress;
101 struct rpc_timeout *timeout; 102 struct rpc_timeout *timeout;
102 char *servername; 103 char *servername;
103 struct rpc_program *program; 104 struct rpc_program *program;
@@ -110,20 +111,20 @@ struct rpc_create_args {
110#define RPC_CLNT_CREATE_HARDRTRY (1UL << 0) 111#define RPC_CLNT_CREATE_HARDRTRY (1UL << 0)
111#define RPC_CLNT_CREATE_INTR (1UL << 1) 112#define RPC_CLNT_CREATE_INTR (1UL << 1)
112#define RPC_CLNT_CREATE_AUTOBIND (1UL << 2) 113#define RPC_CLNT_CREATE_AUTOBIND (1UL << 2)
113#define RPC_CLNT_CREATE_ONESHOT (1UL << 3) 114#define RPC_CLNT_CREATE_NONPRIVPORT (1UL << 3)
114#define RPC_CLNT_CREATE_NONPRIVPORT (1UL << 4) 115#define RPC_CLNT_CREATE_NOPING (1UL << 4)
115#define RPC_CLNT_CREATE_NOPING (1UL << 5) 116#define RPC_CLNT_CREATE_DISCRTRY (1UL << 5)
116#define RPC_CLNT_CREATE_DISCRTRY (1UL << 6)
117 117
118struct rpc_clnt *rpc_create(struct rpc_create_args *args); 118struct rpc_clnt *rpc_create(struct rpc_create_args *args);
119struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, 119struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *,
120 struct rpc_program *, int); 120 struct rpc_program *, int);
121struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); 121struct rpc_clnt *rpc_clone_client(struct rpc_clnt *);
122int rpc_shutdown_client(struct rpc_clnt *); 122void rpc_shutdown_client(struct rpc_clnt *);
123int rpc_destroy_client(struct rpc_clnt *);
124void rpc_release_client(struct rpc_clnt *); 123void rpc_release_client(struct rpc_clnt *);
124
125int rpcb_register(u32, u32, int, unsigned short, int *); 125int rpcb_register(u32, u32, int, unsigned short, int *);
126void rpcb_getport(struct rpc_task *); 126int rpcb_getport_sync(struct sockaddr_in *, __u32, __u32, int);
127void rpcb_getport_async(struct rpc_task *);
127 128
128void rpc_call_setup(struct rpc_task *, struct rpc_message *, int); 129void rpc_call_setup(struct rpc_task *, struct rpc_message *, int);
129 130
@@ -132,20 +133,16 @@ int rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg,
132 void *calldata); 133 void *calldata);
133int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg, 134int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg,
134 int flags); 135 int flags);
136struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred,
137 int flags);
135void rpc_restart_call(struct rpc_task *); 138void rpc_restart_call(struct rpc_task *);
136void rpc_clnt_sigmask(struct rpc_clnt *clnt, sigset_t *oldset); 139void rpc_clnt_sigmask(struct rpc_clnt *clnt, sigset_t *oldset);
137void rpc_clnt_sigunmask(struct rpc_clnt *clnt, sigset_t *oldset); 140void rpc_clnt_sigunmask(struct rpc_clnt *clnt, sigset_t *oldset);
138void rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int); 141void rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int);
139size_t rpc_max_payload(struct rpc_clnt *); 142size_t rpc_max_payload(struct rpc_clnt *);
140void rpc_force_rebind(struct rpc_clnt *); 143void rpc_force_rebind(struct rpc_clnt *);
141int rpc_ping(struct rpc_clnt *clnt, int flags);
142size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t); 144size_t rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t);
143char * rpc_peeraddr2str(struct rpc_clnt *, enum rpc_display_format_t); 145char * rpc_peeraddr2str(struct rpc_clnt *, enum rpc_display_format_t);
144 146
145/*
146 * Helper function for NFSroot support
147 */
148int rpcb_getport_external(struct sockaddr_in *, __u32, __u32, int);
149
150#endif /* __KERNEL__ */ 147#endif /* __KERNEL__ */
151#endif /* _LINUX_SUNRPC_CLNT_H */ 148#endif /* _LINUX_SUNRPC_CLNT_H */
diff --git a/include/linux/sunrpc/gss_api.h b/include/linux/sunrpc/gss_api.h
index 5eca9e442051..bbac101ac372 100644
--- a/include/linux/sunrpc/gss_api.h
+++ b/include/linux/sunrpc/gss_api.h
@@ -77,7 +77,7 @@ struct gss_api_mech {
77 struct module *gm_owner; 77 struct module *gm_owner;
78 struct xdr_netobj gm_oid; 78 struct xdr_netobj gm_oid;
79 char *gm_name; 79 char *gm_name;
80 struct gss_api_ops *gm_ops; 80 const struct gss_api_ops *gm_ops;
81 /* pseudoflavors supported by this mechanism: */ 81 /* pseudoflavors supported by this mechanism: */
82 int gm_pf_num; 82 int gm_pf_num;
83 struct pf_desc * gm_pfs; 83 struct pf_desc * gm_pfs;
diff --git a/include/linux/sunrpc/rpc_pipe_fs.h b/include/linux/sunrpc/rpc_pipe_fs.h
index ad293760f6eb..51b977a4ca20 100644
--- a/include/linux/sunrpc/rpc_pipe_fs.h
+++ b/include/linux/sunrpc/rpc_pipe_fs.h
@@ -23,9 +23,11 @@ struct rpc_inode {
23 void *private; 23 void *private;
24 struct list_head pipe; 24 struct list_head pipe;
25 struct list_head in_upcall; 25 struct list_head in_upcall;
26 struct list_head in_downcall;
26 int pipelen; 27 int pipelen;
27 int nreaders; 28 int nreaders;
28 int nwriters; 29 int nwriters;
30 int nkern_readwriters;
29 wait_queue_head_t waitq; 31 wait_queue_head_t waitq;
30#define RPC_PIPE_WAIT_FOR_OPEN 1 32#define RPC_PIPE_WAIT_FOR_OPEN 1
31 int flags; 33 int flags;
diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h
index 2047fb202a13..8ea077db0099 100644
--- a/include/linux/sunrpc/sched.h
+++ b/include/linux/sunrpc/sched.h
@@ -98,7 +98,6 @@ struct rpc_task {
98 unsigned short tk_pid; /* debugging aid */ 98 unsigned short tk_pid; /* debugging aid */
99#endif 99#endif
100}; 100};
101#define tk_auth tk_client->cl_auth
102#define tk_xprt tk_client->cl_xprt 101#define tk_xprt tk_client->cl_xprt
103 102
104/* support walking a list of tasks on a wait queue */ 103/* support walking a list of tasks on a wait queue */
@@ -110,11 +109,6 @@ struct rpc_task {
110 if (!list_empty(head) && \ 109 if (!list_empty(head) && \
111 ((task=list_entry((head)->next, struct rpc_task, u.tk_wait.list)),1)) 110 ((task=list_entry((head)->next, struct rpc_task, u.tk_wait.list)),1))
112 111
113/* .. and walking list of all tasks */
114#define alltask_for_each(task, pos, head) \
115 list_for_each(pos, head) \
116 if ((task=list_entry(pos, struct rpc_task, tk_task)),1)
117
118typedef void (*rpc_action)(struct rpc_task *); 112typedef void (*rpc_action)(struct rpc_task *);
119 113
120struct rpc_call_ops { 114struct rpc_call_ops {
diff --git a/include/linux/sunrpc/svcsock.h b/include/linux/sunrpc/svcsock.h
index e21dd93ac4b7..a53e0fa855d2 100644
--- a/include/linux/sunrpc/svcsock.h
+++ b/include/linux/sunrpc/svcsock.h
@@ -59,6 +59,7 @@ struct svc_sock {
59 /* cache of various info for TCP sockets */ 59 /* cache of various info for TCP sockets */
60 void *sk_info_authunix; 60 void *sk_info_authunix;
61 61
62 struct sockaddr_storage sk_local; /* local address */
62 struct sockaddr_storage sk_remote; /* remote peer's address */ 63 struct sockaddr_storage sk_remote; /* remote peer's address */
63 int sk_remotelen; /* length of address */ 64 int sk_remotelen; /* length of address */
64}; 65};
diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index 34f7590506fa..d11cedd14f0f 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -17,6 +17,8 @@
17#include <linux/sunrpc/xdr.h> 17#include <linux/sunrpc/xdr.h>
18#include <linux/sunrpc/msg_prot.h> 18#include <linux/sunrpc/msg_prot.h>
19 19
20#ifdef __KERNEL__
21
20extern unsigned int xprt_udp_slot_table_entries; 22extern unsigned int xprt_udp_slot_table_entries;
21extern unsigned int xprt_tcp_slot_table_entries; 23extern unsigned int xprt_tcp_slot_table_entries;
22 24
@@ -194,7 +196,13 @@ struct rpc_xprt {
194 char * address_strings[RPC_DISPLAY_MAX]; 196 char * address_strings[RPC_DISPLAY_MAX];
195}; 197};
196 198
197#ifdef __KERNEL__ 199struct rpc_xprtsock_create {
200 int proto; /* IPPROTO_UDP or IPPROTO_TCP */
201 struct sockaddr * srcaddr; /* optional local address */
202 struct sockaddr * dstaddr; /* remote peer address */
203 size_t addrlen;
204 struct rpc_timeout * timeout; /* optional timeout parameters */
205};
198 206
199/* 207/*
200 * Transport operations used by ULPs 208 * Transport operations used by ULPs
@@ -204,7 +212,7 @@ void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long
204/* 212/*
205 * Generic internal transport functions 213 * Generic internal transport functions
206 */ 214 */
207struct rpc_xprt * xprt_create_transport(int proto, struct sockaddr *addr, size_t size, struct rpc_timeout *toparms); 215struct rpc_xprt * xprt_create_transport(struct rpc_xprtsock_create *args);
208void xprt_connect(struct rpc_task *task); 216void xprt_connect(struct rpc_task *task);
209void xprt_reserve(struct rpc_task *task); 217void xprt_reserve(struct rpc_task *task);
210int xprt_reserve_xprt(struct rpc_task *task); 218int xprt_reserve_xprt(struct rpc_task *task);
@@ -242,8 +250,8 @@ void xprt_disconnect(struct rpc_xprt *xprt);
242/* 250/*
243 * Socket transport setup operations 251 * Socket transport setup operations
244 */ 252 */
245struct rpc_xprt * xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to); 253struct rpc_xprt * xs_setup_udp(struct rpc_xprtsock_create *args);
246struct rpc_xprt * xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to); 254struct rpc_xprt * xs_setup_tcp(struct rpc_xprtsock_create *args);
247int init_socket_xprt(void); 255int init_socket_xprt(void);
248void cleanup_socket_xprt(void); 256void cleanup_socket_xprt(void);
249 257
diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c
index 9527f2bb1744..aa55d0a03e6f 100644
--- a/net/sunrpc/auth.c
+++ b/net/sunrpc/auth.c
@@ -13,17 +13,22 @@
13#include <linux/errno.h> 13#include <linux/errno.h>
14#include <linux/sunrpc/clnt.h> 14#include <linux/sunrpc/clnt.h>
15#include <linux/spinlock.h> 15#include <linux/spinlock.h>
16#include <linux/smp_lock.h>
16 17
17#ifdef RPC_DEBUG 18#ifdef RPC_DEBUG
18# define RPCDBG_FACILITY RPCDBG_AUTH 19# define RPCDBG_FACILITY RPCDBG_AUTH
19#endif 20#endif
20 21
21static struct rpc_authops * auth_flavors[RPC_AUTH_MAXFLAVOR] = { 22static DEFINE_SPINLOCK(rpc_authflavor_lock);
23static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = {
22 &authnull_ops, /* AUTH_NULL */ 24 &authnull_ops, /* AUTH_NULL */
23 &authunix_ops, /* AUTH_UNIX */ 25 &authunix_ops, /* AUTH_UNIX */
24 NULL, /* others can be loadable modules */ 26 NULL, /* others can be loadable modules */
25}; 27};
26 28
29static LIST_HEAD(cred_unused);
30static unsigned long number_cred_unused;
31
27static u32 32static u32
28pseudoflavor_to_flavor(u32 flavor) { 33pseudoflavor_to_flavor(u32 flavor) {
29 if (flavor >= RPC_AUTH_MAXFLAVOR) 34 if (flavor >= RPC_AUTH_MAXFLAVOR)
@@ -32,55 +37,67 @@ pseudoflavor_to_flavor(u32 flavor) {
32} 37}
33 38
34int 39int
35rpcauth_register(struct rpc_authops *ops) 40rpcauth_register(const struct rpc_authops *ops)
36{ 41{
37 rpc_authflavor_t flavor; 42 rpc_authflavor_t flavor;
43 int ret = -EPERM;
38 44
39 if ((flavor = ops->au_flavor) >= RPC_AUTH_MAXFLAVOR) 45 if ((flavor = ops->au_flavor) >= RPC_AUTH_MAXFLAVOR)
40 return -EINVAL; 46 return -EINVAL;
41 if (auth_flavors[flavor] != NULL) 47 spin_lock(&rpc_authflavor_lock);
42 return -EPERM; /* what else? */ 48 if (auth_flavors[flavor] == NULL) {
43 auth_flavors[flavor] = ops; 49 auth_flavors[flavor] = ops;
44 return 0; 50 ret = 0;
51 }
52 spin_unlock(&rpc_authflavor_lock);
53 return ret;
45} 54}
46 55
47int 56int
48rpcauth_unregister(struct rpc_authops *ops) 57rpcauth_unregister(const struct rpc_authops *ops)
49{ 58{
50 rpc_authflavor_t flavor; 59 rpc_authflavor_t flavor;
60 int ret = -EPERM;
51 61
52 if ((flavor = ops->au_flavor) >= RPC_AUTH_MAXFLAVOR) 62 if ((flavor = ops->au_flavor) >= RPC_AUTH_MAXFLAVOR)
53 return -EINVAL; 63 return -EINVAL;
54 if (auth_flavors[flavor] != ops) 64 spin_lock(&rpc_authflavor_lock);
55 return -EPERM; /* what else? */ 65 if (auth_flavors[flavor] == ops) {
56 auth_flavors[flavor] = NULL; 66 auth_flavors[flavor] = NULL;
57 return 0; 67 ret = 0;
68 }
69 spin_unlock(&rpc_authflavor_lock);
70 return ret;
58} 71}
59 72
60struct rpc_auth * 73struct rpc_auth *
61rpcauth_create(rpc_authflavor_t pseudoflavor, struct rpc_clnt *clnt) 74rpcauth_create(rpc_authflavor_t pseudoflavor, struct rpc_clnt *clnt)
62{ 75{
63 struct rpc_auth *auth; 76 struct rpc_auth *auth;
64 struct rpc_authops *ops; 77 const struct rpc_authops *ops;
65 u32 flavor = pseudoflavor_to_flavor(pseudoflavor); 78 u32 flavor = pseudoflavor_to_flavor(pseudoflavor);
66 79
67 auth = ERR_PTR(-EINVAL); 80 auth = ERR_PTR(-EINVAL);
68 if (flavor >= RPC_AUTH_MAXFLAVOR) 81 if (flavor >= RPC_AUTH_MAXFLAVOR)
69 goto out; 82 goto out;
70 83
71 /* FIXME - auth_flavors[] really needs an rw lock,
72 * and module refcounting. */
73#ifdef CONFIG_KMOD 84#ifdef CONFIG_KMOD
74 if ((ops = auth_flavors[flavor]) == NULL) 85 if ((ops = auth_flavors[flavor]) == NULL)
75 request_module("rpc-auth-%u", flavor); 86 request_module("rpc-auth-%u", flavor);
76#endif 87#endif
77 if ((ops = auth_flavors[flavor]) == NULL) 88 spin_lock(&rpc_authflavor_lock);
89 ops = auth_flavors[flavor];
90 if (ops == NULL || !try_module_get(ops->owner)) {
91 spin_unlock(&rpc_authflavor_lock);
78 goto out; 92 goto out;
93 }
94 spin_unlock(&rpc_authflavor_lock);
79 auth = ops->create(clnt, pseudoflavor); 95 auth = ops->create(clnt, pseudoflavor);
96 module_put(ops->owner);
80 if (IS_ERR(auth)) 97 if (IS_ERR(auth))
81 return auth; 98 return auth;
82 if (clnt->cl_auth) 99 if (clnt->cl_auth)
83 rpcauth_destroy(clnt->cl_auth); 100 rpcauth_release(clnt->cl_auth);
84 clnt->cl_auth = auth; 101 clnt->cl_auth = auth;
85 102
86out: 103out:
@@ -88,7 +105,7 @@ out:
88} 105}
89 106
90void 107void
91rpcauth_destroy(struct rpc_auth *auth) 108rpcauth_release(struct rpc_auth *auth)
92{ 109{
93 if (!atomic_dec_and_test(&auth->au_count)) 110 if (!atomic_dec_and_test(&auth->au_count))
94 return; 111 return;
@@ -97,11 +114,31 @@ rpcauth_destroy(struct rpc_auth *auth)
97 114
98static DEFINE_SPINLOCK(rpc_credcache_lock); 115static DEFINE_SPINLOCK(rpc_credcache_lock);
99 116
117static void
118rpcauth_unhash_cred_locked(struct rpc_cred *cred)
119{
120 hlist_del_rcu(&cred->cr_hash);
121 smp_mb__before_clear_bit();
122 clear_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags);
123}
124
125static void
126rpcauth_unhash_cred(struct rpc_cred *cred)
127{
128 spinlock_t *cache_lock;
129
130 cache_lock = &cred->cr_auth->au_credcache->lock;
131 spin_lock(cache_lock);
132 if (atomic_read(&cred->cr_count) == 0)
133 rpcauth_unhash_cred_locked(cred);
134 spin_unlock(cache_lock);
135}
136
100/* 137/*
101 * Initialize RPC credential cache 138 * Initialize RPC credential cache
102 */ 139 */
103int 140int
104rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire) 141rpcauth_init_credcache(struct rpc_auth *auth)
105{ 142{
106 struct rpc_cred_cache *new; 143 struct rpc_cred_cache *new;
107 int i; 144 int i;
@@ -111,8 +148,7 @@ rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire)
111 return -ENOMEM; 148 return -ENOMEM;
112 for (i = 0; i < RPC_CREDCACHE_NR; i++) 149 for (i = 0; i < RPC_CREDCACHE_NR; i++)
113 INIT_HLIST_HEAD(&new->hashtable[i]); 150 INIT_HLIST_HEAD(&new->hashtable[i]);
114 new->expire = expire; 151 spin_lock_init(&new->lock);
115 new->nextgc = jiffies + (expire >> 1);
116 auth->au_credcache = new; 152 auth->au_credcache = new;
117 return 0; 153 return 0;
118} 154}
@@ -121,13 +157,13 @@ rpcauth_init_credcache(struct rpc_auth *auth, unsigned long expire)
121 * Destroy a list of credentials 157 * Destroy a list of credentials
122 */ 158 */
123static inline 159static inline
124void rpcauth_destroy_credlist(struct hlist_head *head) 160void rpcauth_destroy_credlist(struct list_head *head)
125{ 161{
126 struct rpc_cred *cred; 162 struct rpc_cred *cred;
127 163
128 while (!hlist_empty(head)) { 164 while (!list_empty(head)) {
129 cred = hlist_entry(head->first, struct rpc_cred, cr_hash); 165 cred = list_entry(head->next, struct rpc_cred, cr_lru);
130 hlist_del_init(&cred->cr_hash); 166 list_del_init(&cred->cr_lru);
131 put_rpccred(cred); 167 put_rpccred(cred);
132 } 168 }
133} 169}
@@ -137,58 +173,95 @@ void rpcauth_destroy_credlist(struct hlist_head *head)
137 * that are not referenced. 173 * that are not referenced.
138 */ 174 */
139void 175void
140rpcauth_free_credcache(struct rpc_auth *auth) 176rpcauth_clear_credcache(struct rpc_cred_cache *cache)
141{ 177{
142 struct rpc_cred_cache *cache = auth->au_credcache; 178 LIST_HEAD(free);
143 HLIST_HEAD(free); 179 struct hlist_head *head;
144 struct hlist_node *pos, *next;
145 struct rpc_cred *cred; 180 struct rpc_cred *cred;
146 int i; 181 int i;
147 182
148 spin_lock(&rpc_credcache_lock); 183 spin_lock(&rpc_credcache_lock);
184 spin_lock(&cache->lock);
149 for (i = 0; i < RPC_CREDCACHE_NR; i++) { 185 for (i = 0; i < RPC_CREDCACHE_NR; i++) {
150 hlist_for_each_safe(pos, next, &cache->hashtable[i]) { 186 head = &cache->hashtable[i];
151 cred = hlist_entry(pos, struct rpc_cred, cr_hash); 187 while (!hlist_empty(head)) {
152 __hlist_del(&cred->cr_hash); 188 cred = hlist_entry(head->first, struct rpc_cred, cr_hash);
153 hlist_add_head(&cred->cr_hash, &free); 189 get_rpccred(cred);
190 if (!list_empty(&cred->cr_lru)) {
191 list_del(&cred->cr_lru);
192 number_cred_unused--;
193 }
194 list_add_tail(&cred->cr_lru, &free);
195 rpcauth_unhash_cred_locked(cred);
154 } 196 }
155 } 197 }
198 spin_unlock(&cache->lock);
156 spin_unlock(&rpc_credcache_lock); 199 spin_unlock(&rpc_credcache_lock);
157 rpcauth_destroy_credlist(&free); 200 rpcauth_destroy_credlist(&free);
158} 201}
159 202
160static void 203/*
161rpcauth_prune_expired(struct rpc_auth *auth, struct rpc_cred *cred, struct hlist_head *free) 204 * Destroy the RPC credential cache
205 */
206void
207rpcauth_destroy_credcache(struct rpc_auth *auth)
162{ 208{
163 if (atomic_read(&cred->cr_count) != 1) 209 struct rpc_cred_cache *cache = auth->au_credcache;
164 return; 210
165 if (time_after(jiffies, cred->cr_expire + auth->au_credcache->expire)) 211 if (cache) {
166 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 212 auth->au_credcache = NULL;
167 if (!(cred->cr_flags & RPCAUTH_CRED_UPTODATE)) { 213 rpcauth_clear_credcache(cache);
168 __hlist_del(&cred->cr_hash); 214 kfree(cache);
169 hlist_add_head(&cred->cr_hash, free);
170 } 215 }
171} 216}
172 217
173/* 218/*
174 * Remove stale credentials. Avoid sleeping inside the loop. 219 * Remove stale credentials. Avoid sleeping inside the loop.
175 */ 220 */
176static void 221static int
177rpcauth_gc_credcache(struct rpc_auth *auth, struct hlist_head *free) 222rpcauth_prune_expired(struct list_head *free, int nr_to_scan)
178{ 223{
179 struct rpc_cred_cache *cache = auth->au_credcache; 224 spinlock_t *cache_lock;
180 struct hlist_node *pos, *next; 225 struct rpc_cred *cred;
181 struct rpc_cred *cred;
182 int i;
183 226
184 dprintk("RPC: gc'ing RPC credentials for auth %p\n", auth); 227 while (!list_empty(&cred_unused)) {
185 for (i = 0; i < RPC_CREDCACHE_NR; i++) { 228 cred = list_entry(cred_unused.next, struct rpc_cred, cr_lru);
186 hlist_for_each_safe(pos, next, &cache->hashtable[i]) { 229 list_del_init(&cred->cr_lru);
187 cred = hlist_entry(pos, struct rpc_cred, cr_hash); 230 number_cred_unused--;
188 rpcauth_prune_expired(auth, cred, free); 231 if (atomic_read(&cred->cr_count) != 0)
232 continue;
233 cache_lock = &cred->cr_auth->au_credcache->lock;
234 spin_lock(cache_lock);
235 if (atomic_read(&cred->cr_count) == 0) {
236 get_rpccred(cred);
237 list_add_tail(&cred->cr_lru, free);
238 rpcauth_unhash_cred_locked(cred);
239 nr_to_scan--;
189 } 240 }
241 spin_unlock(cache_lock);
242 if (nr_to_scan == 0)
243 break;
190 } 244 }
191 cache->nextgc = jiffies + cache->expire; 245 return nr_to_scan;
246}
247
248/*
249 * Run memory cache shrinker.
250 */
251static int
252rpcauth_cache_shrinker(int nr_to_scan, gfp_t gfp_mask)
253{
254 LIST_HEAD(free);
255 int res;
256
257 if (list_empty(&cred_unused))
258 return 0;
259 spin_lock(&rpc_credcache_lock);
260 nr_to_scan = rpcauth_prune_expired(&free, nr_to_scan);
261 res = (number_cred_unused / 100) * sysctl_vfs_cache_pressure;
262 spin_unlock(&rpc_credcache_lock);
263 rpcauth_destroy_credlist(&free);
264 return res;
192} 265}
193 266
194/* 267/*
@@ -198,53 +271,56 @@ struct rpc_cred *
198rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, 271rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred,
199 int flags) 272 int flags)
200{ 273{
274 LIST_HEAD(free);
201 struct rpc_cred_cache *cache = auth->au_credcache; 275 struct rpc_cred_cache *cache = auth->au_credcache;
202 HLIST_HEAD(free); 276 struct hlist_node *pos;
203 struct hlist_node *pos, *next; 277 struct rpc_cred *cred = NULL,
204 struct rpc_cred *new = NULL, 278 *entry, *new;
205 *cred = NULL;
206 int nr = 0; 279 int nr = 0;
207 280
208 if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS)) 281 if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS))
209 nr = acred->uid & RPC_CREDCACHE_MASK; 282 nr = acred->uid & RPC_CREDCACHE_MASK;
210retry: 283
211 spin_lock(&rpc_credcache_lock); 284 rcu_read_lock();
212 if (time_before(cache->nextgc, jiffies)) 285 hlist_for_each_entry_rcu(entry, pos, &cache->hashtable[nr], cr_hash) {
213 rpcauth_gc_credcache(auth, &free); 286 if (!entry->cr_ops->crmatch(acred, entry, flags))
214 hlist_for_each_safe(pos, next, &cache->hashtable[nr]) { 287 continue;
215 struct rpc_cred *entry; 288 spin_lock(&cache->lock);
216 entry = hlist_entry(pos, struct rpc_cred, cr_hash); 289 if (test_bit(RPCAUTH_CRED_HASHED, &entry->cr_flags) == 0) {
217 if (entry->cr_ops->crmatch(acred, entry, flags)) { 290 spin_unlock(&cache->lock);
218 hlist_del(&entry->cr_hash); 291 continue;
219 cred = entry;
220 break;
221 } 292 }
222 rpcauth_prune_expired(auth, entry, &free); 293 cred = get_rpccred(entry);
294 spin_unlock(&cache->lock);
295 break;
223 } 296 }
224 if (new) { 297 rcu_read_unlock();
225 if (cred)
226 hlist_add_head(&new->cr_hash, &free);
227 else
228 cred = new;
229 }
230 if (cred) {
231 hlist_add_head(&cred->cr_hash, &cache->hashtable[nr]);
232 get_rpccred(cred);
233 }
234 spin_unlock(&rpc_credcache_lock);
235 298
236 rpcauth_destroy_credlist(&free); 299 if (cred != NULL)
300 goto found;
237 301
238 if (!cred) { 302 new = auth->au_ops->crcreate(auth, acred, flags);
239 new = auth->au_ops->crcreate(auth, acred, flags); 303 if (IS_ERR(new)) {
240 if (!IS_ERR(new)) { 304 cred = new;
241#ifdef RPC_DEBUG 305 goto out;
242 new->cr_magic = RPCAUTH_CRED_MAGIC; 306 }
243#endif 307
244 goto retry; 308 spin_lock(&cache->lock);
245 } else 309 hlist_for_each_entry(entry, pos, &cache->hashtable[nr], cr_hash) {
246 cred = new; 310 if (!entry->cr_ops->crmatch(acred, entry, flags))
247 } else if ((cred->cr_flags & RPCAUTH_CRED_NEW) 311 continue;
312 cred = get_rpccred(entry);
313 break;
314 }
315 if (cred == NULL) {
316 cred = new;
317 set_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags);
318 hlist_add_head_rcu(&cred->cr_hash, &cache->hashtable[nr]);
319 } else
320 list_add_tail(&new->cr_lru, &free);
321 spin_unlock(&cache->lock);
322found:
323 if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags)
248 && cred->cr_ops->cr_init != NULL 324 && cred->cr_ops->cr_init != NULL
249 && !(flags & RPCAUTH_LOOKUP_NEW)) { 325 && !(flags & RPCAUTH_LOOKUP_NEW)) {
250 int res = cred->cr_ops->cr_init(auth, cred); 326 int res = cred->cr_ops->cr_init(auth, cred);
@@ -253,8 +329,9 @@ retry:
253 cred = ERR_PTR(res); 329 cred = ERR_PTR(res);
254 } 330 }
255 } 331 }
256 332 rpcauth_destroy_credlist(&free);
257 return (struct rpc_cred *) cred; 333out:
334 return cred;
258} 335}
259 336
260struct rpc_cred * 337struct rpc_cred *
@@ -275,10 +352,27 @@ rpcauth_lookupcred(struct rpc_auth *auth, int flags)
275 return ret; 352 return ret;
276} 353}
277 354
355void
356rpcauth_init_cred(struct rpc_cred *cred, const struct auth_cred *acred,
357 struct rpc_auth *auth, const struct rpc_credops *ops)
358{
359 INIT_HLIST_NODE(&cred->cr_hash);
360 INIT_LIST_HEAD(&cred->cr_lru);
361 atomic_set(&cred->cr_count, 1);
362 cred->cr_auth = auth;
363 cred->cr_ops = ops;
364 cred->cr_expire = jiffies;
365#ifdef RPC_DEBUG
366 cred->cr_magic = RPCAUTH_CRED_MAGIC;
367#endif
368 cred->cr_uid = acred->uid;
369}
370EXPORT_SYMBOL(rpcauth_init_cred);
371
278struct rpc_cred * 372struct rpc_cred *
279rpcauth_bindcred(struct rpc_task *task) 373rpcauth_bindcred(struct rpc_task *task)
280{ 374{
281 struct rpc_auth *auth = task->tk_auth; 375 struct rpc_auth *auth = task->tk_client->cl_auth;
282 struct auth_cred acred = { 376 struct auth_cred acred = {
283 .uid = current->fsuid, 377 .uid = current->fsuid,
284 .gid = current->fsgid, 378 .gid = current->fsgid,
@@ -288,7 +382,7 @@ rpcauth_bindcred(struct rpc_task *task)
288 int flags = 0; 382 int flags = 0;
289 383
290 dprintk("RPC: %5u looking up %s cred\n", 384 dprintk("RPC: %5u looking up %s cred\n",
291 task->tk_pid, task->tk_auth->au_ops->au_name); 385 task->tk_pid, task->tk_client->cl_auth->au_ops->au_name);
292 get_group_info(acred.group_info); 386 get_group_info(acred.group_info);
293 if (task->tk_flags & RPC_TASK_ROOTCREDS) 387 if (task->tk_flags & RPC_TASK_ROOTCREDS)
294 flags |= RPCAUTH_LOOKUP_ROOTCREDS; 388 flags |= RPCAUTH_LOOKUP_ROOTCREDS;
@@ -304,19 +398,42 @@ rpcauth_bindcred(struct rpc_task *task)
304void 398void
305rpcauth_holdcred(struct rpc_task *task) 399rpcauth_holdcred(struct rpc_task *task)
306{ 400{
307 dprintk("RPC: %5u holding %s cred %p\n", 401 struct rpc_cred *cred = task->tk_msg.rpc_cred;
308 task->tk_pid, task->tk_auth->au_ops->au_name, 402 if (cred != NULL) {
309 task->tk_msg.rpc_cred); 403 get_rpccred(cred);
310 if (task->tk_msg.rpc_cred) 404 dprintk("RPC: %5u holding %s cred %p\n", task->tk_pid,
311 get_rpccred(task->tk_msg.rpc_cred); 405 cred->cr_auth->au_ops->au_name, cred);
406 }
312} 407}
313 408
314void 409void
315put_rpccred(struct rpc_cred *cred) 410put_rpccred(struct rpc_cred *cred)
316{ 411{
317 cred->cr_expire = jiffies; 412 /* Fast path for unhashed credentials */
413 if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0)
414 goto need_lock;
415
318 if (!atomic_dec_and_test(&cred->cr_count)) 416 if (!atomic_dec_and_test(&cred->cr_count))
319 return; 417 return;
418 goto out_destroy;
419need_lock:
420 if (!atomic_dec_and_lock(&cred->cr_count, &rpc_credcache_lock))
421 return;
422 if (!list_empty(&cred->cr_lru)) {
423 number_cred_unused--;
424 list_del_init(&cred->cr_lru);
425 }
426 if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) == 0)
427 rpcauth_unhash_cred(cred);
428 else if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) != 0) {
429 cred->cr_expire = jiffies;
430 list_add_tail(&cred->cr_lru, &cred_unused);
431 number_cred_unused++;
432 spin_unlock(&rpc_credcache_lock);
433 return;
434 }
435 spin_unlock(&rpc_credcache_lock);
436out_destroy:
320 cred->cr_ops->crdestroy(cred); 437 cred->cr_ops->crdestroy(cred);
321} 438}
322 439
@@ -326,7 +443,7 @@ rpcauth_unbindcred(struct rpc_task *task)
326 struct rpc_cred *cred = task->tk_msg.rpc_cred; 443 struct rpc_cred *cred = task->tk_msg.rpc_cred;
327 444
328 dprintk("RPC: %5u releasing %s cred %p\n", 445 dprintk("RPC: %5u releasing %s cred %p\n",
329 task->tk_pid, task->tk_auth->au_ops->au_name, cred); 446 task->tk_pid, cred->cr_auth->au_ops->au_name, cred);
330 447
331 put_rpccred(cred); 448 put_rpccred(cred);
332 task->tk_msg.rpc_cred = NULL; 449 task->tk_msg.rpc_cred = NULL;
@@ -338,7 +455,7 @@ rpcauth_marshcred(struct rpc_task *task, __be32 *p)
338 struct rpc_cred *cred = task->tk_msg.rpc_cred; 455 struct rpc_cred *cred = task->tk_msg.rpc_cred;
339 456
340 dprintk("RPC: %5u marshaling %s cred %p\n", 457 dprintk("RPC: %5u marshaling %s cred %p\n",
341 task->tk_pid, task->tk_auth->au_ops->au_name, cred); 458 task->tk_pid, cred->cr_auth->au_ops->au_name, cred);
342 459
343 return cred->cr_ops->crmarshal(task, p); 460 return cred->cr_ops->crmarshal(task, p);
344} 461}
@@ -349,7 +466,7 @@ rpcauth_checkverf(struct rpc_task *task, __be32 *p)
349 struct rpc_cred *cred = task->tk_msg.rpc_cred; 466 struct rpc_cred *cred = task->tk_msg.rpc_cred;
350 467
351 dprintk("RPC: %5u validating %s cred %p\n", 468 dprintk("RPC: %5u validating %s cred %p\n",
352 task->tk_pid, task->tk_auth->au_ops->au_name, cred); 469 task->tk_pid, cred->cr_auth->au_ops->au_name, cred);
353 470
354 return cred->cr_ops->crvalidate(task, p); 471 return cred->cr_ops->crvalidate(task, p);
355} 472}
@@ -359,13 +476,17 @@ rpcauth_wrap_req(struct rpc_task *task, kxdrproc_t encode, void *rqstp,
359 __be32 *data, void *obj) 476 __be32 *data, void *obj)
360{ 477{
361 struct rpc_cred *cred = task->tk_msg.rpc_cred; 478 struct rpc_cred *cred = task->tk_msg.rpc_cred;
479 int ret;
362 480
363 dprintk("RPC: %5u using %s cred %p to wrap rpc data\n", 481 dprintk("RPC: %5u using %s cred %p to wrap rpc data\n",
364 task->tk_pid, cred->cr_ops->cr_name, cred); 482 task->tk_pid, cred->cr_ops->cr_name, cred);
365 if (cred->cr_ops->crwrap_req) 483 if (cred->cr_ops->crwrap_req)
366 return cred->cr_ops->crwrap_req(task, encode, rqstp, data, obj); 484 return cred->cr_ops->crwrap_req(task, encode, rqstp, data, obj);
367 /* By default, we encode the arguments normally. */ 485 /* By default, we encode the arguments normally. */
368 return encode(rqstp, data, obj); 486 lock_kernel();
487 ret = encode(rqstp, data, obj);
488 unlock_kernel();
489 return ret;
369} 490}
370 491
371int 492int
@@ -373,6 +494,7 @@ rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp,
373 __be32 *data, void *obj) 494 __be32 *data, void *obj)
374{ 495{
375 struct rpc_cred *cred = task->tk_msg.rpc_cred; 496 struct rpc_cred *cred = task->tk_msg.rpc_cred;
497 int ret;
376 498
377 dprintk("RPC: %5u using %s cred %p to unwrap rpc data\n", 499 dprintk("RPC: %5u using %s cred %p to unwrap rpc data\n",
378 task->tk_pid, cred->cr_ops->cr_name, cred); 500 task->tk_pid, cred->cr_ops->cr_name, cred);
@@ -380,7 +502,10 @@ rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp,
380 return cred->cr_ops->crunwrap_resp(task, decode, rqstp, 502 return cred->cr_ops->crunwrap_resp(task, decode, rqstp,
381 data, obj); 503 data, obj);
382 /* By default, we decode the arguments normally. */ 504 /* By default, we decode the arguments normally. */
383 return decode(rqstp, data, obj); 505 lock_kernel();
506 ret = decode(rqstp, data, obj);
507 unlock_kernel();
508 return ret;
384} 509}
385 510
386int 511int
@@ -390,7 +515,7 @@ rpcauth_refreshcred(struct rpc_task *task)
390 int err; 515 int err;
391 516
392 dprintk("RPC: %5u refreshing %s cred %p\n", 517 dprintk("RPC: %5u refreshing %s cred %p\n",
393 task->tk_pid, task->tk_auth->au_ops->au_name, cred); 518 task->tk_pid, cred->cr_auth->au_ops->au_name, cred);
394 519
395 err = cred->cr_ops->crrefresh(task); 520 err = cred->cr_ops->crrefresh(task);
396 if (err < 0) 521 if (err < 0)
@@ -401,17 +526,34 @@ rpcauth_refreshcred(struct rpc_task *task)
401void 526void
402rpcauth_invalcred(struct rpc_task *task) 527rpcauth_invalcred(struct rpc_task *task)
403{ 528{
529 struct rpc_cred *cred = task->tk_msg.rpc_cred;
530
404 dprintk("RPC: %5u invalidating %s cred %p\n", 531 dprintk("RPC: %5u invalidating %s cred %p\n",
405 task->tk_pid, task->tk_auth->au_ops->au_name, task->tk_msg.rpc_cred); 532 task->tk_pid, cred->cr_auth->au_ops->au_name, cred);
406 spin_lock(&rpc_credcache_lock); 533 if (cred)
407 if (task->tk_msg.rpc_cred) 534 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
408 task->tk_msg.rpc_cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE;
409 spin_unlock(&rpc_credcache_lock);
410} 535}
411 536
412int 537int
413rpcauth_uptodatecred(struct rpc_task *task) 538rpcauth_uptodatecred(struct rpc_task *task)
414{ 539{
415 return !(task->tk_msg.rpc_cred) || 540 struct rpc_cred *cred = task->tk_msg.rpc_cred;
416 (task->tk_msg.rpc_cred->cr_flags & RPCAUTH_CRED_UPTODATE); 541
542 return cred == NULL ||
543 test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) != 0;
544}
545
546
547static struct shrinker *rpc_cred_shrinker;
548
549void __init rpcauth_init_module(void)
550{
551 rpc_init_authunix();
552 rpc_cred_shrinker = set_shrinker(DEFAULT_SEEKS, rpcauth_cache_shrinker);
553}
554
555void __exit rpcauth_remove_module(void)
556{
557 if (rpc_cred_shrinker != NULL)
558 remove_shrinker(rpc_cred_shrinker);
417} 559}
diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c
index 4e4ccc5b6fea..baf4096d52d4 100644
--- a/net/sunrpc/auth_gss/auth_gss.c
+++ b/net/sunrpc/auth_gss/auth_gss.c
@@ -54,9 +54,10 @@
54#include <linux/sunrpc/gss_api.h> 54#include <linux/sunrpc/gss_api.h>
55#include <asm/uaccess.h> 55#include <asm/uaccess.h>
56 56
57static struct rpc_authops authgss_ops; 57static const struct rpc_authops authgss_ops;
58 58
59static struct rpc_credops gss_credops; 59static const struct rpc_credops gss_credops;
60static const struct rpc_credops gss_nullops;
60 61
61#ifdef RPC_DEBUG 62#ifdef RPC_DEBUG
62# define RPCDBG_FACILITY RPCDBG_AUTH 63# define RPCDBG_FACILITY RPCDBG_AUTH
@@ -64,7 +65,6 @@ static struct rpc_credops gss_credops;
64 65
65#define NFS_NGROUPS 16 66#define NFS_NGROUPS 16
66 67
67#define GSS_CRED_EXPIRE (60 * HZ) /* XXX: reasonable? */
68#define GSS_CRED_SLACK 1024 /* XXX: unused */ 68#define GSS_CRED_SLACK 1024 /* XXX: unused */
69/* length of a krb5 verifier (48), plus data added before arguments when 69/* length of a krb5 verifier (48), plus data added before arguments when
70 * using integrity (two 4-byte integers): */ 70 * using integrity (two 4-byte integers): */
@@ -79,19 +79,16 @@ static struct rpc_credops gss_credops;
79/* dump the buffer in `emacs-hexl' style */ 79/* dump the buffer in `emacs-hexl' style */
80#define isprint(c) ((c > 0x1f) && (c < 0x7f)) 80#define isprint(c) ((c > 0x1f) && (c < 0x7f))
81 81
82static DEFINE_RWLOCK(gss_ctx_lock);
83
84struct gss_auth { 82struct gss_auth {
83 struct kref kref;
85 struct rpc_auth rpc_auth; 84 struct rpc_auth rpc_auth;
86 struct gss_api_mech *mech; 85 struct gss_api_mech *mech;
87 enum rpc_gss_svc service; 86 enum rpc_gss_svc service;
88 struct list_head upcalls;
89 struct rpc_clnt *client; 87 struct rpc_clnt *client;
90 struct dentry *dentry; 88 struct dentry *dentry;
91 spinlock_t lock;
92}; 89};
93 90
94static void gss_destroy_ctx(struct gss_cl_ctx *); 91static void gss_free_ctx(struct gss_cl_ctx *);
95static struct rpc_pipe_ops gss_upcall_ops; 92static struct rpc_pipe_ops gss_upcall_ops;
96 93
97static inline struct gss_cl_ctx * 94static inline struct gss_cl_ctx *
@@ -105,20 +102,24 @@ static inline void
105gss_put_ctx(struct gss_cl_ctx *ctx) 102gss_put_ctx(struct gss_cl_ctx *ctx)
106{ 103{
107 if (atomic_dec_and_test(&ctx->count)) 104 if (atomic_dec_and_test(&ctx->count))
108 gss_destroy_ctx(ctx); 105 gss_free_ctx(ctx);
109} 106}
110 107
108/* gss_cred_set_ctx:
109 * called by gss_upcall_callback and gss_create_upcall in order
110 * to set the gss context. The actual exchange of an old context
111 * and a new one is protected by the inode->i_lock.
112 */
111static void 113static void
112gss_cred_set_ctx(struct rpc_cred *cred, struct gss_cl_ctx *ctx) 114gss_cred_set_ctx(struct rpc_cred *cred, struct gss_cl_ctx *ctx)
113{ 115{
114 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); 116 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
115 struct gss_cl_ctx *old; 117 struct gss_cl_ctx *old;
116 write_lock(&gss_ctx_lock); 118
117 old = gss_cred->gc_ctx; 119 old = gss_cred->gc_ctx;
118 gss_cred->gc_ctx = ctx; 120 rcu_assign_pointer(gss_cred->gc_ctx, ctx);
119 cred->cr_flags |= RPCAUTH_CRED_UPTODATE; 121 set_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
120 cred->cr_flags &= ~RPCAUTH_CRED_NEW; 122 clear_bit(RPCAUTH_CRED_NEW, &cred->cr_flags);
121 write_unlock(&gss_ctx_lock);
122 if (old) 123 if (old)
123 gss_put_ctx(old); 124 gss_put_ctx(old);
124} 125}
@@ -129,10 +130,10 @@ gss_cred_is_uptodate_ctx(struct rpc_cred *cred)
129 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); 130 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
130 int res = 0; 131 int res = 0;
131 132
132 read_lock(&gss_ctx_lock); 133 rcu_read_lock();
133 if ((cred->cr_flags & RPCAUTH_CRED_UPTODATE) && gss_cred->gc_ctx) 134 if (test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) && gss_cred->gc_ctx)
134 res = 1; 135 res = 1;
135 read_unlock(&gss_ctx_lock); 136 rcu_read_unlock();
136 return res; 137 return res;
137} 138}
138 139
@@ -171,10 +172,10 @@ gss_cred_get_ctx(struct rpc_cred *cred)
171 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); 172 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
172 struct gss_cl_ctx *ctx = NULL; 173 struct gss_cl_ctx *ctx = NULL;
173 174
174 read_lock(&gss_ctx_lock); 175 rcu_read_lock();
175 if (gss_cred->gc_ctx) 176 if (gss_cred->gc_ctx)
176 ctx = gss_get_ctx(gss_cred->gc_ctx); 177 ctx = gss_get_ctx(gss_cred->gc_ctx);
177 read_unlock(&gss_ctx_lock); 178 rcu_read_unlock();
178 return ctx; 179 return ctx;
179} 180}
180 181
@@ -269,10 +270,10 @@ gss_release_msg(struct gss_upcall_msg *gss_msg)
269} 270}
270 271
271static struct gss_upcall_msg * 272static struct gss_upcall_msg *
272__gss_find_upcall(struct gss_auth *gss_auth, uid_t uid) 273__gss_find_upcall(struct rpc_inode *rpci, uid_t uid)
273{ 274{
274 struct gss_upcall_msg *pos; 275 struct gss_upcall_msg *pos;
275 list_for_each_entry(pos, &gss_auth->upcalls, list) { 276 list_for_each_entry(pos, &rpci->in_downcall, list) {
276 if (pos->uid != uid) 277 if (pos->uid != uid)
277 continue; 278 continue;
278 atomic_inc(&pos->count); 279 atomic_inc(&pos->count);
@@ -290,24 +291,24 @@ __gss_find_upcall(struct gss_auth *gss_auth, uid_t uid)
290static inline struct gss_upcall_msg * 291static inline struct gss_upcall_msg *
291gss_add_msg(struct gss_auth *gss_auth, struct gss_upcall_msg *gss_msg) 292gss_add_msg(struct gss_auth *gss_auth, struct gss_upcall_msg *gss_msg)
292{ 293{
294 struct inode *inode = gss_auth->dentry->d_inode;
295 struct rpc_inode *rpci = RPC_I(inode);
293 struct gss_upcall_msg *old; 296 struct gss_upcall_msg *old;
294 297
295 spin_lock(&gss_auth->lock); 298 spin_lock(&inode->i_lock);
296 old = __gss_find_upcall(gss_auth, gss_msg->uid); 299 old = __gss_find_upcall(rpci, gss_msg->uid);
297 if (old == NULL) { 300 if (old == NULL) {
298 atomic_inc(&gss_msg->count); 301 atomic_inc(&gss_msg->count);
299 list_add(&gss_msg->list, &gss_auth->upcalls); 302 list_add(&gss_msg->list, &rpci->in_downcall);
300 } else 303 } else
301 gss_msg = old; 304 gss_msg = old;
302 spin_unlock(&gss_auth->lock); 305 spin_unlock(&inode->i_lock);
303 return gss_msg; 306 return gss_msg;
304} 307}
305 308
306static void 309static void
307__gss_unhash_msg(struct gss_upcall_msg *gss_msg) 310__gss_unhash_msg(struct gss_upcall_msg *gss_msg)
308{ 311{
309 if (list_empty(&gss_msg->list))
310 return;
311 list_del_init(&gss_msg->list); 312 list_del_init(&gss_msg->list);
312 rpc_wake_up_status(&gss_msg->rpc_waitqueue, gss_msg->msg.errno); 313 rpc_wake_up_status(&gss_msg->rpc_waitqueue, gss_msg->msg.errno);
313 wake_up_all(&gss_msg->waitqueue); 314 wake_up_all(&gss_msg->waitqueue);
@@ -318,10 +319,14 @@ static void
318gss_unhash_msg(struct gss_upcall_msg *gss_msg) 319gss_unhash_msg(struct gss_upcall_msg *gss_msg)
319{ 320{
320 struct gss_auth *gss_auth = gss_msg->auth; 321 struct gss_auth *gss_auth = gss_msg->auth;
322 struct inode *inode = gss_auth->dentry->d_inode;
321 323
322 spin_lock(&gss_auth->lock); 324 if (list_empty(&gss_msg->list))
323 __gss_unhash_msg(gss_msg); 325 return;
324 spin_unlock(&gss_auth->lock); 326 spin_lock(&inode->i_lock);
327 if (!list_empty(&gss_msg->list))
328 __gss_unhash_msg(gss_msg);
329 spin_unlock(&inode->i_lock);
325} 330}
326 331
327static void 332static void
@@ -330,16 +335,16 @@ gss_upcall_callback(struct rpc_task *task)
330 struct gss_cred *gss_cred = container_of(task->tk_msg.rpc_cred, 335 struct gss_cred *gss_cred = container_of(task->tk_msg.rpc_cred,
331 struct gss_cred, gc_base); 336 struct gss_cred, gc_base);
332 struct gss_upcall_msg *gss_msg = gss_cred->gc_upcall; 337 struct gss_upcall_msg *gss_msg = gss_cred->gc_upcall;
338 struct inode *inode = gss_msg->auth->dentry->d_inode;
333 339
334 BUG_ON(gss_msg == NULL); 340 spin_lock(&inode->i_lock);
335 if (gss_msg->ctx) 341 if (gss_msg->ctx)
336 gss_cred_set_ctx(task->tk_msg.rpc_cred, gss_get_ctx(gss_msg->ctx)); 342 gss_cred_set_ctx(task->tk_msg.rpc_cred, gss_get_ctx(gss_msg->ctx));
337 else 343 else
338 task->tk_status = gss_msg->msg.errno; 344 task->tk_status = gss_msg->msg.errno;
339 spin_lock(&gss_msg->auth->lock);
340 gss_cred->gc_upcall = NULL; 345 gss_cred->gc_upcall = NULL;
341 rpc_wake_up_status(&gss_msg->rpc_waitqueue, gss_msg->msg.errno); 346 rpc_wake_up_status(&gss_msg->rpc_waitqueue, gss_msg->msg.errno);
342 spin_unlock(&gss_msg->auth->lock); 347 spin_unlock(&inode->i_lock);
343 gss_release_msg(gss_msg); 348 gss_release_msg(gss_msg);
344} 349}
345 350
@@ -386,11 +391,12 @@ static inline int
386gss_refresh_upcall(struct rpc_task *task) 391gss_refresh_upcall(struct rpc_task *task)
387{ 392{
388 struct rpc_cred *cred = task->tk_msg.rpc_cred; 393 struct rpc_cred *cred = task->tk_msg.rpc_cred;
389 struct gss_auth *gss_auth = container_of(task->tk_client->cl_auth, 394 struct gss_auth *gss_auth = container_of(cred->cr_auth,
390 struct gss_auth, rpc_auth); 395 struct gss_auth, rpc_auth);
391 struct gss_cred *gss_cred = container_of(cred, 396 struct gss_cred *gss_cred = container_of(cred,
392 struct gss_cred, gc_base); 397 struct gss_cred, gc_base);
393 struct gss_upcall_msg *gss_msg; 398 struct gss_upcall_msg *gss_msg;
399 struct inode *inode = gss_auth->dentry->d_inode;
394 int err = 0; 400 int err = 0;
395 401
396 dprintk("RPC: %5u gss_refresh_upcall for uid %u\n", task->tk_pid, 402 dprintk("RPC: %5u gss_refresh_upcall for uid %u\n", task->tk_pid,
@@ -400,7 +406,7 @@ gss_refresh_upcall(struct rpc_task *task)
400 err = PTR_ERR(gss_msg); 406 err = PTR_ERR(gss_msg);
401 goto out; 407 goto out;
402 } 408 }
403 spin_lock(&gss_auth->lock); 409 spin_lock(&inode->i_lock);
404 if (gss_cred->gc_upcall != NULL) 410 if (gss_cred->gc_upcall != NULL)
405 rpc_sleep_on(&gss_cred->gc_upcall->rpc_waitqueue, task, NULL, NULL); 411 rpc_sleep_on(&gss_cred->gc_upcall->rpc_waitqueue, task, NULL, NULL);
406 else if (gss_msg->ctx == NULL && gss_msg->msg.errno >= 0) { 412 else if (gss_msg->ctx == NULL && gss_msg->msg.errno >= 0) {
@@ -411,7 +417,7 @@ gss_refresh_upcall(struct rpc_task *task)
411 rpc_sleep_on(&gss_msg->rpc_waitqueue, task, gss_upcall_callback, NULL); 417 rpc_sleep_on(&gss_msg->rpc_waitqueue, task, gss_upcall_callback, NULL);
412 } else 418 } else
413 err = gss_msg->msg.errno; 419 err = gss_msg->msg.errno;
414 spin_unlock(&gss_auth->lock); 420 spin_unlock(&inode->i_lock);
415 gss_release_msg(gss_msg); 421 gss_release_msg(gss_msg);
416out: 422out:
417 dprintk("RPC: %5u gss_refresh_upcall for uid %u result %d\n", 423 dprintk("RPC: %5u gss_refresh_upcall for uid %u result %d\n",
@@ -422,6 +428,7 @@ out:
422static inline int 428static inline int
423gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred) 429gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred)
424{ 430{
431 struct inode *inode = gss_auth->dentry->d_inode;
425 struct rpc_cred *cred = &gss_cred->gc_base; 432 struct rpc_cred *cred = &gss_cred->gc_base;
426 struct gss_upcall_msg *gss_msg; 433 struct gss_upcall_msg *gss_msg;
427 DEFINE_WAIT(wait); 434 DEFINE_WAIT(wait);
@@ -435,12 +442,11 @@ gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred)
435 } 442 }
436 for (;;) { 443 for (;;) {
437 prepare_to_wait(&gss_msg->waitqueue, &wait, TASK_INTERRUPTIBLE); 444 prepare_to_wait(&gss_msg->waitqueue, &wait, TASK_INTERRUPTIBLE);
438 spin_lock(&gss_auth->lock); 445 spin_lock(&inode->i_lock);
439 if (gss_msg->ctx != NULL || gss_msg->msg.errno < 0) { 446 if (gss_msg->ctx != NULL || gss_msg->msg.errno < 0) {
440 spin_unlock(&gss_auth->lock);
441 break; 447 break;
442 } 448 }
443 spin_unlock(&gss_auth->lock); 449 spin_unlock(&inode->i_lock);
444 if (signalled()) { 450 if (signalled()) {
445 err = -ERESTARTSYS; 451 err = -ERESTARTSYS;
446 goto out_intr; 452 goto out_intr;
@@ -451,6 +457,7 @@ gss_create_upcall(struct gss_auth *gss_auth, struct gss_cred *gss_cred)
451 gss_cred_set_ctx(cred, gss_get_ctx(gss_msg->ctx)); 457 gss_cred_set_ctx(cred, gss_get_ctx(gss_msg->ctx));
452 else 458 else
453 err = gss_msg->msg.errno; 459 err = gss_msg->msg.errno;
460 spin_unlock(&inode->i_lock);
454out_intr: 461out_intr:
455 finish_wait(&gss_msg->waitqueue, &wait); 462 finish_wait(&gss_msg->waitqueue, &wait);
456 gss_release_msg(gss_msg); 463 gss_release_msg(gss_msg);
@@ -489,12 +496,11 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
489 const void *p, *end; 496 const void *p, *end;
490 void *buf; 497 void *buf;
491 struct rpc_clnt *clnt; 498 struct rpc_clnt *clnt;
492 struct gss_auth *gss_auth;
493 struct rpc_cred *cred;
494 struct gss_upcall_msg *gss_msg; 499 struct gss_upcall_msg *gss_msg;
500 struct inode *inode = filp->f_path.dentry->d_inode;
495 struct gss_cl_ctx *ctx; 501 struct gss_cl_ctx *ctx;
496 uid_t uid; 502 uid_t uid;
497 int err = -EFBIG; 503 ssize_t err = -EFBIG;
498 504
499 if (mlen > MSG_BUF_MAXSIZE) 505 if (mlen > MSG_BUF_MAXSIZE)
500 goto out; 506 goto out;
@@ -503,7 +509,7 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
503 if (!buf) 509 if (!buf)
504 goto out; 510 goto out;
505 511
506 clnt = RPC_I(filp->f_path.dentry->d_inode)->private; 512 clnt = RPC_I(inode)->private;
507 err = -EFAULT; 513 err = -EFAULT;
508 if (copy_from_user(buf, src, mlen)) 514 if (copy_from_user(buf, src, mlen))
509 goto err; 515 goto err;
@@ -519,43 +525,38 @@ gss_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
519 ctx = gss_alloc_context(); 525 ctx = gss_alloc_context();
520 if (ctx == NULL) 526 if (ctx == NULL)
521 goto err; 527 goto err;
522 err = 0; 528
523 gss_auth = container_of(clnt->cl_auth, struct gss_auth, rpc_auth); 529 err = -ENOENT;
524 p = gss_fill_context(p, end, ctx, gss_auth->mech); 530 /* Find a matching upcall */
531 spin_lock(&inode->i_lock);
532 gss_msg = __gss_find_upcall(RPC_I(inode), uid);
533 if (gss_msg == NULL) {
534 spin_unlock(&inode->i_lock);
535 goto err_put_ctx;
536 }
537 list_del_init(&gss_msg->list);
538 spin_unlock(&inode->i_lock);
539
540 p = gss_fill_context(p, end, ctx, gss_msg->auth->mech);
525 if (IS_ERR(p)) { 541 if (IS_ERR(p)) {
526 err = PTR_ERR(p); 542 err = PTR_ERR(p);
527 if (err != -EACCES) 543 gss_msg->msg.errno = (err == -EACCES) ? -EACCES : -EAGAIN;
528 goto err_put_ctx; 544 goto err_release_msg;
529 }
530 spin_lock(&gss_auth->lock);
531 gss_msg = __gss_find_upcall(gss_auth, uid);
532 if (gss_msg) {
533 if (err == 0 && gss_msg->ctx == NULL)
534 gss_msg->ctx = gss_get_ctx(ctx);
535 gss_msg->msg.errno = err;
536 __gss_unhash_msg(gss_msg);
537 spin_unlock(&gss_auth->lock);
538 gss_release_msg(gss_msg);
539 } else {
540 struct auth_cred acred = { .uid = uid };
541 spin_unlock(&gss_auth->lock);
542 cred = rpcauth_lookup_credcache(clnt->cl_auth, &acred, RPCAUTH_LOOKUP_NEW);
543 if (IS_ERR(cred)) {
544 err = PTR_ERR(cred);
545 goto err_put_ctx;
546 }
547 gss_cred_set_ctx(cred, gss_get_ctx(ctx));
548 } 545 }
549 gss_put_ctx(ctx); 546 gss_msg->ctx = gss_get_ctx(ctx);
550 kfree(buf); 547 err = mlen;
551 dprintk("RPC: gss_pipe_downcall returning length %Zu\n", mlen); 548
552 return mlen; 549err_release_msg:
550 spin_lock(&inode->i_lock);
551 __gss_unhash_msg(gss_msg);
552 spin_unlock(&inode->i_lock);
553 gss_release_msg(gss_msg);
553err_put_ctx: 554err_put_ctx:
554 gss_put_ctx(ctx); 555 gss_put_ctx(ctx);
555err: 556err:
556 kfree(buf); 557 kfree(buf);
557out: 558out:
558 dprintk("RPC: gss_pipe_downcall returning %d\n", err); 559 dprintk("RPC: gss_pipe_downcall returning %Zd\n", err);
559 return err; 560 return err;
560} 561}
561 562
@@ -563,27 +564,21 @@ static void
563gss_pipe_release(struct inode *inode) 564gss_pipe_release(struct inode *inode)
564{ 565{
565 struct rpc_inode *rpci = RPC_I(inode); 566 struct rpc_inode *rpci = RPC_I(inode);
566 struct rpc_clnt *clnt; 567 struct gss_upcall_msg *gss_msg;
567 struct rpc_auth *auth;
568 struct gss_auth *gss_auth;
569 568
570 clnt = rpci->private; 569 spin_lock(&inode->i_lock);
571 auth = clnt->cl_auth; 570 while (!list_empty(&rpci->in_downcall)) {
572 gss_auth = container_of(auth, struct gss_auth, rpc_auth);
573 spin_lock(&gss_auth->lock);
574 while (!list_empty(&gss_auth->upcalls)) {
575 struct gss_upcall_msg *gss_msg;
576 571
577 gss_msg = list_entry(gss_auth->upcalls.next, 572 gss_msg = list_entry(rpci->in_downcall.next,
578 struct gss_upcall_msg, list); 573 struct gss_upcall_msg, list);
579 gss_msg->msg.errno = -EPIPE; 574 gss_msg->msg.errno = -EPIPE;
580 atomic_inc(&gss_msg->count); 575 atomic_inc(&gss_msg->count);
581 __gss_unhash_msg(gss_msg); 576 __gss_unhash_msg(gss_msg);
582 spin_unlock(&gss_auth->lock); 577 spin_unlock(&inode->i_lock);
583 gss_release_msg(gss_msg); 578 gss_release_msg(gss_msg);
584 spin_lock(&gss_auth->lock); 579 spin_lock(&inode->i_lock);
585 } 580 }
586 spin_unlock(&gss_auth->lock); 581 spin_unlock(&inode->i_lock);
587} 582}
588 583
589static void 584static void
@@ -637,18 +632,13 @@ gss_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor)
637 gss_auth->service = gss_pseudoflavor_to_service(gss_auth->mech, flavor); 632 gss_auth->service = gss_pseudoflavor_to_service(gss_auth->mech, flavor);
638 if (gss_auth->service == 0) 633 if (gss_auth->service == 0)
639 goto err_put_mech; 634 goto err_put_mech;
640 INIT_LIST_HEAD(&gss_auth->upcalls);
641 spin_lock_init(&gss_auth->lock);
642 auth = &gss_auth->rpc_auth; 635 auth = &gss_auth->rpc_auth;
643 auth->au_cslack = GSS_CRED_SLACK >> 2; 636 auth->au_cslack = GSS_CRED_SLACK >> 2;
644 auth->au_rslack = GSS_VERF_SLACK >> 2; 637 auth->au_rslack = GSS_VERF_SLACK >> 2;
645 auth->au_ops = &authgss_ops; 638 auth->au_ops = &authgss_ops;
646 auth->au_flavor = flavor; 639 auth->au_flavor = flavor;
647 atomic_set(&auth->au_count, 1); 640 atomic_set(&auth->au_count, 1);
648 641 kref_init(&gss_auth->kref);
649 err = rpcauth_init_credcache(auth, GSS_CRED_EXPIRE);
650 if (err)
651 goto err_put_mech;
652 642
653 gss_auth->dentry = rpc_mkpipe(clnt->cl_dentry, gss_auth->mech->gm_name, 643 gss_auth->dentry = rpc_mkpipe(clnt->cl_dentry, gss_auth->mech->gm_name,
654 clnt, &gss_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); 644 clnt, &gss_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN);
@@ -657,7 +647,13 @@ gss_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor)
657 goto err_put_mech; 647 goto err_put_mech;
658 } 648 }
659 649
650 err = rpcauth_init_credcache(auth);
651 if (err)
652 goto err_unlink_pipe;
653
660 return auth; 654 return auth;
655err_unlink_pipe:
656 rpc_unlink(gss_auth->dentry);
661err_put_mech: 657err_put_mech:
662 gss_mech_put(gss_auth->mech); 658 gss_mech_put(gss_auth->mech);
663err_free: 659err_free:
@@ -668,6 +664,25 @@ out_dec:
668} 664}
669 665
670static void 666static void
667gss_free(struct gss_auth *gss_auth)
668{
669 rpc_unlink(gss_auth->dentry);
670 gss_auth->dentry = NULL;
671 gss_mech_put(gss_auth->mech);
672
673 kfree(gss_auth);
674 module_put(THIS_MODULE);
675}
676
677static void
678gss_free_callback(struct kref *kref)
679{
680 struct gss_auth *gss_auth = container_of(kref, struct gss_auth, kref);
681
682 gss_free(gss_auth);
683}
684
685static void
671gss_destroy(struct rpc_auth *auth) 686gss_destroy(struct rpc_auth *auth)
672{ 687{
673 struct gss_auth *gss_auth; 688 struct gss_auth *gss_auth;
@@ -675,23 +690,51 @@ gss_destroy(struct rpc_auth *auth)
675 dprintk("RPC: destroying GSS authenticator %p flavor %d\n", 690 dprintk("RPC: destroying GSS authenticator %p flavor %d\n",
676 auth, auth->au_flavor); 691 auth, auth->au_flavor);
677 692
693 rpcauth_destroy_credcache(auth);
694
678 gss_auth = container_of(auth, struct gss_auth, rpc_auth); 695 gss_auth = container_of(auth, struct gss_auth, rpc_auth);
679 rpc_unlink(gss_auth->dentry); 696 kref_put(&gss_auth->kref, gss_free_callback);
680 gss_auth->dentry = NULL; 697}
681 gss_mech_put(gss_auth->mech);
682 698
683 rpcauth_free_credcache(auth); 699/*
684 kfree(gss_auth); 700 * gss_destroying_context will cause the RPCSEC_GSS to send a NULL RPC call
685 module_put(THIS_MODULE); 701 * to the server with the GSS control procedure field set to
702 * RPC_GSS_PROC_DESTROY. This should normally cause the server to release
703 * all RPCSEC_GSS state associated with that context.
704 */
705static int
706gss_destroying_context(struct rpc_cred *cred)
707{
708 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
709 struct gss_auth *gss_auth = container_of(cred->cr_auth, struct gss_auth, rpc_auth);
710 struct rpc_task *task;
711
712 if (gss_cred->gc_ctx == NULL ||
713 gss_cred->gc_ctx->gc_proc == RPC_GSS_PROC_DESTROY)
714 return 0;
715
716 gss_cred->gc_ctx->gc_proc = RPC_GSS_PROC_DESTROY;
717 cred->cr_ops = &gss_nullops;
718
719 /* Take a reference to ensure the cred will be destroyed either
720 * by the RPC call or by the put_rpccred() below */
721 get_rpccred(cred);
722
723 task = rpc_call_null(gss_auth->client, cred, RPC_TASK_ASYNC);
724 if (!IS_ERR(task))
725 rpc_put_task(task);
726
727 put_rpccred(cred);
728 return 1;
686} 729}
687 730
688/* gss_destroy_cred (and gss_destroy_ctx) are used to clean up after failure 731/* gss_destroy_cred (and gss_free_ctx) are used to clean up after failure
689 * to create a new cred or context, so they check that things have been 732 * to create a new cred or context, so they check that things have been
690 * allocated before freeing them. */ 733 * allocated before freeing them. */
691static void 734static void
692gss_destroy_ctx(struct gss_cl_ctx *ctx) 735gss_do_free_ctx(struct gss_cl_ctx *ctx)
693{ 736{
694 dprintk("RPC: gss_destroy_ctx\n"); 737 dprintk("RPC: gss_free_ctx\n");
695 738
696 if (ctx->gc_gss_ctx) 739 if (ctx->gc_gss_ctx)
697 gss_delete_sec_context(&ctx->gc_gss_ctx); 740 gss_delete_sec_context(&ctx->gc_gss_ctx);
@@ -701,15 +744,46 @@ gss_destroy_ctx(struct gss_cl_ctx *ctx)
701} 744}
702 745
703static void 746static void
704gss_destroy_cred(struct rpc_cred *rc) 747gss_free_ctx_callback(struct rcu_head *head)
705{ 748{
706 struct gss_cred *cred = container_of(rc, struct gss_cred, gc_base); 749 struct gss_cl_ctx *ctx = container_of(head, struct gss_cl_ctx, gc_rcu);
750 gss_do_free_ctx(ctx);
751}
707 752
708 dprintk("RPC: gss_destroy_cred \n"); 753static void
754gss_free_ctx(struct gss_cl_ctx *ctx)
755{
756 call_rcu(&ctx->gc_rcu, gss_free_ctx_callback);
757}
709 758
710 if (cred->gc_ctx) 759static void
711 gss_put_ctx(cred->gc_ctx); 760gss_free_cred(struct gss_cred *gss_cred)
712 kfree(cred); 761{
762 dprintk("RPC: gss_free_cred %p\n", gss_cred);
763 kfree(gss_cred);
764}
765
766static void
767gss_free_cred_callback(struct rcu_head *head)
768{
769 struct gss_cred *gss_cred = container_of(head, struct gss_cred, gc_base.cr_rcu);
770 gss_free_cred(gss_cred);
771}
772
773static void
774gss_destroy_cred(struct rpc_cred *cred)
775{
776 struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base);
777 struct gss_auth *gss_auth = container_of(cred->cr_auth, struct gss_auth, rpc_auth);
778 struct gss_cl_ctx *ctx = gss_cred->gc_ctx;
779
780 if (gss_destroying_context(cred))
781 return;
782 rcu_assign_pointer(gss_cred->gc_ctx, NULL);
783 call_rcu(&cred->cr_rcu, gss_free_cred_callback);
784 if (ctx)
785 gss_put_ctx(ctx);
786 kref_put(&gss_auth->kref, gss_free_callback);
713} 787}
714 788
715/* 789/*
@@ -734,16 +808,14 @@ gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
734 if (!(cred = kzalloc(sizeof(*cred), GFP_KERNEL))) 808 if (!(cred = kzalloc(sizeof(*cred), GFP_KERNEL)))
735 goto out_err; 809 goto out_err;
736 810
737 atomic_set(&cred->gc_count, 1); 811 rpcauth_init_cred(&cred->gc_base, acred, auth, &gss_credops);
738 cred->gc_uid = acred->uid;
739 /* 812 /*
740 * Note: in order to force a call to call_refresh(), we deliberately 813 * Note: in order to force a call to call_refresh(), we deliberately
741 * fail to flag the credential as RPCAUTH_CRED_UPTODATE. 814 * fail to flag the credential as RPCAUTH_CRED_UPTODATE.
742 */ 815 */
743 cred->gc_flags = 0; 816 cred->gc_base.cr_flags = 1UL << RPCAUTH_CRED_NEW;
744 cred->gc_base.cr_ops = &gss_credops;
745 cred->gc_base.cr_flags = RPCAUTH_CRED_NEW;
746 cred->gc_service = gss_auth->service; 817 cred->gc_service = gss_auth->service;
818 kref_get(&gss_auth->kref);
747 return &cred->gc_base; 819 return &cred->gc_base;
748 820
749out_err: 821out_err:
@@ -774,7 +846,7 @@ gss_match(struct auth_cred *acred, struct rpc_cred *rc, int flags)
774 * we don't really care if the credential has expired or not, 846 * we don't really care if the credential has expired or not,
775 * since the caller should be prepared to reinitialise it. 847 * since the caller should be prepared to reinitialise it.
776 */ 848 */
777 if ((flags & RPCAUTH_LOOKUP_NEW) && (rc->cr_flags & RPCAUTH_CRED_NEW)) 849 if ((flags & RPCAUTH_LOOKUP_NEW) && test_bit(RPCAUTH_CRED_NEW, &rc->cr_flags))
778 goto out; 850 goto out;
779 /* Don't match with creds that have expired. */ 851 /* Don't match with creds that have expired. */
780 if (gss_cred->gc_ctx && time_after(jiffies, gss_cred->gc_ctx->gc_expiry)) 852 if (gss_cred->gc_ctx && time_after(jiffies, gss_cred->gc_ctx->gc_expiry))
@@ -830,7 +902,7 @@ gss_marshal(struct rpc_task *task, __be32 *p)
830 mic.data = (u8 *)(p + 1); 902 mic.data = (u8 *)(p + 1);
831 maj_stat = gss_get_mic(ctx->gc_gss_ctx, &verf_buf, &mic); 903 maj_stat = gss_get_mic(ctx->gc_gss_ctx, &verf_buf, &mic);
832 if (maj_stat == GSS_S_CONTEXT_EXPIRED) { 904 if (maj_stat == GSS_S_CONTEXT_EXPIRED) {
833 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 905 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
834 } else if (maj_stat != 0) { 906 } else if (maj_stat != 0) {
835 printk("gss_marshal: gss_get_mic FAILED (%d)\n", maj_stat); 907 printk("gss_marshal: gss_get_mic FAILED (%d)\n", maj_stat);
836 goto out_put_ctx; 908 goto out_put_ctx;
@@ -855,6 +927,13 @@ gss_refresh(struct rpc_task *task)
855 return 0; 927 return 0;
856} 928}
857 929
930/* Dummy refresh routine: used only when destroying the context */
931static int
932gss_refresh_null(struct rpc_task *task)
933{
934 return -EACCES;
935}
936
858static __be32 * 937static __be32 *
859gss_validate(struct rpc_task *task, __be32 *p) 938gss_validate(struct rpc_task *task, __be32 *p)
860{ 939{
@@ -883,12 +962,15 @@ gss_validate(struct rpc_task *task, __be32 *p)
883 962
884 maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &verf_buf, &mic); 963 maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &verf_buf, &mic);
885 if (maj_stat == GSS_S_CONTEXT_EXPIRED) 964 if (maj_stat == GSS_S_CONTEXT_EXPIRED)
886 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 965 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
887 if (maj_stat) 966 if (maj_stat) {
967 dprintk("RPC: %5u gss_validate: gss_verify_mic returned"
968 "error 0x%08x\n", task->tk_pid, maj_stat);
888 goto out_bad; 969 goto out_bad;
970 }
889 /* We leave it to unwrap to calculate au_rslack. For now we just 971 /* We leave it to unwrap to calculate au_rslack. For now we just
890 * calculate the length of the verifier: */ 972 * calculate the length of the verifier: */
891 task->tk_auth->au_verfsize = XDR_QUADLEN(len) + 2; 973 cred->cr_auth->au_verfsize = XDR_QUADLEN(len) + 2;
892 gss_put_ctx(ctx); 974 gss_put_ctx(ctx);
893 dprintk("RPC: %5u gss_validate: gss_verify_mic succeeded.\n", 975 dprintk("RPC: %5u gss_validate: gss_verify_mic succeeded.\n",
894 task->tk_pid); 976 task->tk_pid);
@@ -917,7 +999,9 @@ gss_wrap_req_integ(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
917 offset = (u8 *)p - (u8 *)snd_buf->head[0].iov_base; 999 offset = (u8 *)p - (u8 *)snd_buf->head[0].iov_base;
918 *p++ = htonl(rqstp->rq_seqno); 1000 *p++ = htonl(rqstp->rq_seqno);
919 1001
1002 lock_kernel();
920 status = encode(rqstp, p, obj); 1003 status = encode(rqstp, p, obj);
1004 unlock_kernel();
921 if (status) 1005 if (status)
922 return status; 1006 return status;
923 1007
@@ -937,7 +1021,7 @@ gss_wrap_req_integ(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
937 maj_stat = gss_get_mic(ctx->gc_gss_ctx, &integ_buf, &mic); 1021 maj_stat = gss_get_mic(ctx->gc_gss_ctx, &integ_buf, &mic);
938 status = -EIO; /* XXX? */ 1022 status = -EIO; /* XXX? */
939 if (maj_stat == GSS_S_CONTEXT_EXPIRED) 1023 if (maj_stat == GSS_S_CONTEXT_EXPIRED)
940 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 1024 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
941 else if (maj_stat) 1025 else if (maj_stat)
942 return status; 1026 return status;
943 q = xdr_encode_opaque(p, NULL, mic.len); 1027 q = xdr_encode_opaque(p, NULL, mic.len);
@@ -1011,7 +1095,9 @@ gss_wrap_req_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
1011 offset = (u8 *)p - (u8 *)snd_buf->head[0].iov_base; 1095 offset = (u8 *)p - (u8 *)snd_buf->head[0].iov_base;
1012 *p++ = htonl(rqstp->rq_seqno); 1096 *p++ = htonl(rqstp->rq_seqno);
1013 1097
1098 lock_kernel();
1014 status = encode(rqstp, p, obj); 1099 status = encode(rqstp, p, obj);
1100 unlock_kernel();
1015 if (status) 1101 if (status)
1016 return status; 1102 return status;
1017 1103
@@ -1036,7 +1122,7 @@ gss_wrap_req_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
1036 /* We're assuming that when GSS_S_CONTEXT_EXPIRED, the encryption was 1122 /* We're assuming that when GSS_S_CONTEXT_EXPIRED, the encryption was
1037 * done anyway, so it's safe to put the request on the wire: */ 1123 * done anyway, so it's safe to put the request on the wire: */
1038 if (maj_stat == GSS_S_CONTEXT_EXPIRED) 1124 if (maj_stat == GSS_S_CONTEXT_EXPIRED)
1039 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 1125 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
1040 else if (maj_stat) 1126 else if (maj_stat)
1041 return status; 1127 return status;
1042 1128
@@ -1070,12 +1156,16 @@ gss_wrap_req(struct rpc_task *task,
1070 /* The spec seems a little ambiguous here, but I think that not 1156 /* The spec seems a little ambiguous here, but I think that not
1071 * wrapping context destruction requests makes the most sense. 1157 * wrapping context destruction requests makes the most sense.
1072 */ 1158 */
1159 lock_kernel();
1073 status = encode(rqstp, p, obj); 1160 status = encode(rqstp, p, obj);
1161 unlock_kernel();
1074 goto out; 1162 goto out;
1075 } 1163 }
1076 switch (gss_cred->gc_service) { 1164 switch (gss_cred->gc_service) {
1077 case RPC_GSS_SVC_NONE: 1165 case RPC_GSS_SVC_NONE:
1166 lock_kernel();
1078 status = encode(rqstp, p, obj); 1167 status = encode(rqstp, p, obj);
1168 unlock_kernel();
1079 break; 1169 break;
1080 case RPC_GSS_SVC_INTEGRITY: 1170 case RPC_GSS_SVC_INTEGRITY:
1081 status = gss_wrap_req_integ(cred, ctx, encode, 1171 status = gss_wrap_req_integ(cred, ctx, encode,
@@ -1123,7 +1213,7 @@ gss_unwrap_resp_integ(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
1123 1213
1124 maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &integ_buf, &mic); 1214 maj_stat = gss_verify_mic(ctx->gc_gss_ctx, &integ_buf, &mic);
1125 if (maj_stat == GSS_S_CONTEXT_EXPIRED) 1215 if (maj_stat == GSS_S_CONTEXT_EXPIRED)
1126 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 1216 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
1127 if (maj_stat != GSS_S_COMPLETE) 1217 if (maj_stat != GSS_S_COMPLETE)
1128 return status; 1218 return status;
1129 return 0; 1219 return 0;
@@ -1148,7 +1238,7 @@ gss_unwrap_resp_priv(struct rpc_cred *cred, struct gss_cl_ctx *ctx,
1148 1238
1149 maj_stat = gss_unwrap(ctx->gc_gss_ctx, offset, rcv_buf); 1239 maj_stat = gss_unwrap(ctx->gc_gss_ctx, offset, rcv_buf);
1150 if (maj_stat == GSS_S_CONTEXT_EXPIRED) 1240 if (maj_stat == GSS_S_CONTEXT_EXPIRED)
1151 cred->cr_flags &= ~RPCAUTH_CRED_UPTODATE; 1241 clear_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags);
1152 if (maj_stat != GSS_S_COMPLETE) 1242 if (maj_stat != GSS_S_COMPLETE)
1153 return status; 1243 return status;
1154 if (ntohl(*(*p)++) != rqstp->rq_seqno) 1244 if (ntohl(*(*p)++) != rqstp->rq_seqno)
@@ -1188,10 +1278,12 @@ gss_unwrap_resp(struct rpc_task *task,
1188 break; 1278 break;
1189 } 1279 }
1190 /* take into account extra slack for integrity and privacy cases: */ 1280 /* take into account extra slack for integrity and privacy cases: */
1191 task->tk_auth->au_rslack = task->tk_auth->au_verfsize + (p - savedp) 1281 cred->cr_auth->au_rslack = cred->cr_auth->au_verfsize + (p - savedp)
1192 + (savedlen - head->iov_len); 1282 + (savedlen - head->iov_len);
1193out_decode: 1283out_decode:
1284 lock_kernel();
1194 status = decode(rqstp, p, obj); 1285 status = decode(rqstp, p, obj);
1286 unlock_kernel();
1195out: 1287out:
1196 gss_put_ctx(ctx); 1288 gss_put_ctx(ctx);
1197 dprintk("RPC: %5u gss_unwrap_resp returning %d\n", task->tk_pid, 1289 dprintk("RPC: %5u gss_unwrap_resp returning %d\n", task->tk_pid,
@@ -1199,7 +1291,7 @@ out:
1199 return status; 1291 return status;
1200} 1292}
1201 1293
1202static struct rpc_authops authgss_ops = { 1294static const struct rpc_authops authgss_ops = {
1203 .owner = THIS_MODULE, 1295 .owner = THIS_MODULE,
1204 .au_flavor = RPC_AUTH_GSS, 1296 .au_flavor = RPC_AUTH_GSS,
1205#ifdef RPC_DEBUG 1297#ifdef RPC_DEBUG
@@ -1211,7 +1303,7 @@ static struct rpc_authops authgss_ops = {
1211 .crcreate = gss_create_cred 1303 .crcreate = gss_create_cred
1212}; 1304};
1213 1305
1214static struct rpc_credops gss_credops = { 1306static const struct rpc_credops gss_credops = {
1215 .cr_name = "AUTH_GSS", 1307 .cr_name = "AUTH_GSS",
1216 .crdestroy = gss_destroy_cred, 1308 .crdestroy = gss_destroy_cred,
1217 .cr_init = gss_cred_init, 1309 .cr_init = gss_cred_init,
@@ -1223,6 +1315,17 @@ static struct rpc_credops gss_credops = {
1223 .crunwrap_resp = gss_unwrap_resp, 1315 .crunwrap_resp = gss_unwrap_resp,
1224}; 1316};
1225 1317
1318static const struct rpc_credops gss_nullops = {
1319 .cr_name = "AUTH_GSS",
1320 .crdestroy = gss_destroy_cred,
1321 .crmatch = gss_match,
1322 .crmarshal = gss_marshal,
1323 .crrefresh = gss_refresh_null,
1324 .crvalidate = gss_validate,
1325 .crwrap_req = gss_wrap_req,
1326 .crunwrap_resp = gss_unwrap_resp,
1327};
1328
1226static struct rpc_pipe_ops gss_upcall_ops = { 1329static struct rpc_pipe_ops gss_upcall_ops = {
1227 .upcall = gss_pipe_upcall, 1330 .upcall = gss_pipe_upcall,
1228 .downcall = gss_pipe_downcall, 1331 .downcall = gss_pipe_downcall,
diff --git a/net/sunrpc/auth_gss/gss_krb5_mech.c b/net/sunrpc/auth_gss/gss_krb5_mech.c
index 7b1943217053..71b9daefdff3 100644
--- a/net/sunrpc/auth_gss/gss_krb5_mech.c
+++ b/net/sunrpc/auth_gss/gss_krb5_mech.c
@@ -201,7 +201,7 @@ gss_delete_sec_context_kerberos(void *internal_ctx) {
201 kfree(kctx); 201 kfree(kctx);
202} 202}
203 203
204static struct gss_api_ops gss_kerberos_ops = { 204static const struct gss_api_ops gss_kerberos_ops = {
205 .gss_import_sec_context = gss_import_sec_context_kerberos, 205 .gss_import_sec_context = gss_import_sec_context_kerberos,
206 .gss_get_mic = gss_get_mic_kerberos, 206 .gss_get_mic = gss_get_mic_kerberos,
207 .gss_verify_mic = gss_verify_mic_kerberos, 207 .gss_verify_mic = gss_verify_mic_kerberos,
diff --git a/net/sunrpc/auth_gss/gss_spkm3_mech.c b/net/sunrpc/auth_gss/gss_spkm3_mech.c
index 7e15aa68ae64..577d590e755f 100644
--- a/net/sunrpc/auth_gss/gss_spkm3_mech.c
+++ b/net/sunrpc/auth_gss/gss_spkm3_mech.c
@@ -202,7 +202,7 @@ gss_get_mic_spkm3(struct gss_ctx *ctx,
202 return err; 202 return err;
203} 203}
204 204
205static struct gss_api_ops gss_spkm3_ops = { 205static const struct gss_api_ops gss_spkm3_ops = {
206 .gss_import_sec_context = gss_import_sec_context_spkm3, 206 .gss_import_sec_context = gss_import_sec_context_spkm3,
207 .gss_get_mic = gss_get_mic_spkm3, 207 .gss_get_mic = gss_get_mic_spkm3,
208 .gss_verify_mic = gss_verify_mic_spkm3, 208 .gss_verify_mic = gss_verify_mic_spkm3,
diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c
index 3df9fccab2f8..537d0e8589dd 100644
--- a/net/sunrpc/auth_null.c
+++ b/net/sunrpc/auth_null.c
@@ -76,7 +76,7 @@ nul_marshal(struct rpc_task *task, __be32 *p)
76static int 76static int
77nul_refresh(struct rpc_task *task) 77nul_refresh(struct rpc_task *task)
78{ 78{
79 task->tk_msg.rpc_cred->cr_flags |= RPCAUTH_CRED_UPTODATE; 79 set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags);
80 return 0; 80 return 0;
81} 81}
82 82
@@ -101,7 +101,7 @@ nul_validate(struct rpc_task *task, __be32 *p)
101 return p; 101 return p;
102} 102}
103 103
104struct rpc_authops authnull_ops = { 104const struct rpc_authops authnull_ops = {
105 .owner = THIS_MODULE, 105 .owner = THIS_MODULE,
106 .au_flavor = RPC_AUTH_NULL, 106 .au_flavor = RPC_AUTH_NULL,
107#ifdef RPC_DEBUG 107#ifdef RPC_DEBUG
@@ -122,7 +122,7 @@ struct rpc_auth null_auth = {
122}; 122};
123 123
124static 124static
125struct rpc_credops null_credops = { 125const struct rpc_credops null_credops = {
126 .cr_name = "AUTH_NULL", 126 .cr_name = "AUTH_NULL",
127 .crdestroy = nul_destroy_cred, 127 .crdestroy = nul_destroy_cred,
128 .crmatch = nul_match, 128 .crmatch = nul_match,
@@ -133,9 +133,11 @@ struct rpc_credops null_credops = {
133 133
134static 134static
135struct rpc_cred null_cred = { 135struct rpc_cred null_cred = {
136 .cr_lru = LIST_HEAD_INIT(null_cred.cr_lru),
137 .cr_auth = &null_auth,
136 .cr_ops = &null_credops, 138 .cr_ops = &null_credops,
137 .cr_count = ATOMIC_INIT(1), 139 .cr_count = ATOMIC_INIT(1),
138 .cr_flags = RPCAUTH_CRED_UPTODATE, 140 .cr_flags = 1UL << RPCAUTH_CRED_UPTODATE,
139#ifdef RPC_DEBUG 141#ifdef RPC_DEBUG
140 .cr_magic = RPCAUTH_CRED_MAGIC, 142 .cr_magic = RPCAUTH_CRED_MAGIC,
141#endif 143#endif
diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c
index 4e7733aee36e..5ed91e5bcee4 100644
--- a/net/sunrpc/auth_unix.c
+++ b/net/sunrpc/auth_unix.c
@@ -20,11 +20,6 @@ struct unx_cred {
20 gid_t uc_gids[NFS_NGROUPS]; 20 gid_t uc_gids[NFS_NGROUPS];
21}; 21};
22#define uc_uid uc_base.cr_uid 22#define uc_uid uc_base.cr_uid
23#define uc_count uc_base.cr_count
24#define uc_flags uc_base.cr_flags
25#define uc_expire uc_base.cr_expire
26
27#define UNX_CRED_EXPIRE (60 * HZ)
28 23
29#define UNX_WRITESLACK (21 + (UNX_MAXNODENAME >> 2)) 24#define UNX_WRITESLACK (21 + (UNX_MAXNODENAME >> 2))
30 25
@@ -34,15 +29,14 @@ struct unx_cred {
34 29
35static struct rpc_auth unix_auth; 30static struct rpc_auth unix_auth;
36static struct rpc_cred_cache unix_cred_cache; 31static struct rpc_cred_cache unix_cred_cache;
37static struct rpc_credops unix_credops; 32static const struct rpc_credops unix_credops;
38 33
39static struct rpc_auth * 34static struct rpc_auth *
40unx_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor) 35unx_create(struct rpc_clnt *clnt, rpc_authflavor_t flavor)
41{ 36{
42 dprintk("RPC: creating UNIX authenticator for client %p\n", 37 dprintk("RPC: creating UNIX authenticator for client %p\n",
43 clnt); 38 clnt);
44 if (atomic_inc_return(&unix_auth.au_count) == 0) 39 atomic_inc(&unix_auth.au_count);
45 unix_cred_cache.nextgc = jiffies + (unix_cred_cache.expire >> 1);
46 return &unix_auth; 40 return &unix_auth;
47} 41}
48 42
@@ -50,7 +44,7 @@ static void
50unx_destroy(struct rpc_auth *auth) 44unx_destroy(struct rpc_auth *auth)
51{ 45{
52 dprintk("RPC: destroying UNIX authenticator %p\n", auth); 46 dprintk("RPC: destroying UNIX authenticator %p\n", auth);
53 rpcauth_free_credcache(auth); 47 rpcauth_clear_credcache(auth->au_credcache);
54} 48}
55 49
56/* 50/*
@@ -74,8 +68,8 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
74 if (!(cred = kmalloc(sizeof(*cred), GFP_KERNEL))) 68 if (!(cred = kmalloc(sizeof(*cred), GFP_KERNEL)))
75 return ERR_PTR(-ENOMEM); 69 return ERR_PTR(-ENOMEM);
76 70
77 atomic_set(&cred->uc_count, 1); 71 rpcauth_init_cred(&cred->uc_base, acred, auth, &unix_credops);
78 cred->uc_flags = RPCAUTH_CRED_UPTODATE; 72 cred->uc_base.cr_flags = 1UL << RPCAUTH_CRED_UPTODATE;
79 if (flags & RPCAUTH_LOOKUP_ROOTCREDS) { 73 if (flags & RPCAUTH_LOOKUP_ROOTCREDS) {
80 cred->uc_uid = 0; 74 cred->uc_uid = 0;
81 cred->uc_gid = 0; 75 cred->uc_gid = 0;
@@ -85,22 +79,34 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
85 if (groups > NFS_NGROUPS) 79 if (groups > NFS_NGROUPS)
86 groups = NFS_NGROUPS; 80 groups = NFS_NGROUPS;
87 81
88 cred->uc_uid = acred->uid;
89 cred->uc_gid = acred->gid; 82 cred->uc_gid = acred->gid;
90 for (i = 0; i < groups; i++) 83 for (i = 0; i < groups; i++)
91 cred->uc_gids[i] = GROUP_AT(acred->group_info, i); 84 cred->uc_gids[i] = GROUP_AT(acred->group_info, i);
92 if (i < NFS_NGROUPS) 85 if (i < NFS_NGROUPS)
93 cred->uc_gids[i] = NOGROUP; 86 cred->uc_gids[i] = NOGROUP;
94 } 87 }
95 cred->uc_base.cr_ops = &unix_credops;
96 88
97 return (struct rpc_cred *) cred; 89 return &cred->uc_base;
90}
91
92static void
93unx_free_cred(struct unx_cred *unx_cred)
94{
95 dprintk("RPC: unx_free_cred %p\n", unx_cred);
96 kfree(unx_cred);
97}
98
99static void
100unx_free_cred_callback(struct rcu_head *head)
101{
102 struct unx_cred *unx_cred = container_of(head, struct unx_cred, uc_base.cr_rcu);
103 unx_free_cred(unx_cred);
98} 104}
99 105
100static void 106static void
101unx_destroy_cred(struct rpc_cred *cred) 107unx_destroy_cred(struct rpc_cred *cred)
102{ 108{
103 kfree(cred); 109 call_rcu(&cred->cr_rcu, unx_free_cred_callback);
104} 110}
105 111
106/* 112/*
@@ -111,7 +117,7 @@ unx_destroy_cred(struct rpc_cred *cred)
111static int 117static int
112unx_match(struct auth_cred *acred, struct rpc_cred *rcred, int flags) 118unx_match(struct auth_cred *acred, struct rpc_cred *rcred, int flags)
113{ 119{
114 struct unx_cred *cred = (struct unx_cred *) rcred; 120 struct unx_cred *cred = container_of(rcred, struct unx_cred, uc_base);
115 int i; 121 int i;
116 122
117 if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS)) { 123 if (!(flags & RPCAUTH_LOOKUP_ROOTCREDS)) {
@@ -142,7 +148,7 @@ static __be32 *
142unx_marshal(struct rpc_task *task, __be32 *p) 148unx_marshal(struct rpc_task *task, __be32 *p)
143{ 149{
144 struct rpc_clnt *clnt = task->tk_client; 150 struct rpc_clnt *clnt = task->tk_client;
145 struct unx_cred *cred = (struct unx_cred *) task->tk_msg.rpc_cred; 151 struct unx_cred *cred = container_of(task->tk_msg.rpc_cred, struct unx_cred, uc_base);
146 __be32 *base, *hold; 152 __be32 *base, *hold;
147 int i; 153 int i;
148 154
@@ -175,7 +181,7 @@ unx_marshal(struct rpc_task *task, __be32 *p)
175static int 181static int
176unx_refresh(struct rpc_task *task) 182unx_refresh(struct rpc_task *task)
177{ 183{
178 task->tk_msg.rpc_cred->cr_flags |= RPCAUTH_CRED_UPTODATE; 184 set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags);
179 return 0; 185 return 0;
180} 186}
181 187
@@ -198,13 +204,18 @@ unx_validate(struct rpc_task *task, __be32 *p)
198 printk("RPC: giant verf size: %u\n", size); 204 printk("RPC: giant verf size: %u\n", size);
199 return NULL; 205 return NULL;
200 } 206 }
201 task->tk_auth->au_rslack = (size >> 2) + 2; 207 task->tk_msg.rpc_cred->cr_auth->au_rslack = (size >> 2) + 2;
202 p += (size >> 2); 208 p += (size >> 2);
203 209
204 return p; 210 return p;
205} 211}
206 212
207struct rpc_authops authunix_ops = { 213void __init rpc_init_authunix(void)
214{
215 spin_lock_init(&unix_cred_cache.lock);
216}
217
218const struct rpc_authops authunix_ops = {
208 .owner = THIS_MODULE, 219 .owner = THIS_MODULE,
209 .au_flavor = RPC_AUTH_UNIX, 220 .au_flavor = RPC_AUTH_UNIX,
210#ifdef RPC_DEBUG 221#ifdef RPC_DEBUG
@@ -218,7 +229,6 @@ struct rpc_authops authunix_ops = {
218 229
219static 230static
220struct rpc_cred_cache unix_cred_cache = { 231struct rpc_cred_cache unix_cred_cache = {
221 .expire = UNX_CRED_EXPIRE,
222}; 232};
223 233
224static 234static
@@ -232,7 +242,7 @@ struct rpc_auth unix_auth = {
232}; 242};
233 243
234static 244static
235struct rpc_credops unix_credops = { 245const struct rpc_credops unix_credops = {
236 .cr_name = "AUTH_UNIX", 246 .cr_name = "AUTH_UNIX",
237 .crdestroy = unx_destroy_cred, 247 .crdestroy = unx_destroy_cred,
238 .crmatch = unx_match, 248 .crmatch = unx_match,
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index d8fbee40a19c..52429b1ffcc1 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -44,6 +44,12 @@
44 dprintk("RPC: %5u %s (status %d)\n", t->tk_pid, \ 44 dprintk("RPC: %5u %s (status %d)\n", t->tk_pid, \
45 __FUNCTION__, t->tk_status) 45 __FUNCTION__, t->tk_status)
46 46
47/*
48 * All RPC clients are linked into this list
49 */
50static LIST_HEAD(all_clients);
51static DEFINE_SPINLOCK(rpc_client_lock);
52
47static DECLARE_WAIT_QUEUE_HEAD(destroy_wait); 53static DECLARE_WAIT_QUEUE_HEAD(destroy_wait);
48 54
49 55
@@ -66,6 +72,21 @@ static void call_connect_status(struct rpc_task *task);
66static __be32 * call_header(struct rpc_task *task); 72static __be32 * call_header(struct rpc_task *task);
67static __be32 * call_verify(struct rpc_task *task); 73static __be32 * call_verify(struct rpc_task *task);
68 74
75static int rpc_ping(struct rpc_clnt *clnt, int flags);
76
77static void rpc_register_client(struct rpc_clnt *clnt)
78{
79 spin_lock(&rpc_client_lock);
80 list_add(&clnt->cl_clients, &all_clients);
81 spin_unlock(&rpc_client_lock);
82}
83
84static void rpc_unregister_client(struct rpc_clnt *clnt)
85{
86 spin_lock(&rpc_client_lock);
87 list_del(&clnt->cl_clients);
88 spin_unlock(&rpc_client_lock);
89}
69 90
70static int 91static int
71rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name) 92rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name)
@@ -111,6 +132,9 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s
111 dprintk("RPC: creating %s client for %s (xprt %p)\n", 132 dprintk("RPC: creating %s client for %s (xprt %p)\n",
112 program->name, servname, xprt); 133 program->name, servname, xprt);
113 134
135 err = rpciod_up();
136 if (err)
137 goto out_no_rpciod;
114 err = -EINVAL; 138 err = -EINVAL;
115 if (!xprt) 139 if (!xprt)
116 goto out_no_xprt; 140 goto out_no_xprt;
@@ -121,8 +145,6 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s
121 clnt = kzalloc(sizeof(*clnt), GFP_KERNEL); 145 clnt = kzalloc(sizeof(*clnt), GFP_KERNEL);
122 if (!clnt) 146 if (!clnt)
123 goto out_err; 147 goto out_err;
124 atomic_set(&clnt->cl_users, 0);
125 atomic_set(&clnt->cl_count, 1);
126 clnt->cl_parent = clnt; 148 clnt->cl_parent = clnt;
127 149
128 clnt->cl_server = clnt->cl_inline_name; 150 clnt->cl_server = clnt->cl_inline_name;
@@ -148,6 +170,8 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s
148 if (clnt->cl_metrics == NULL) 170 if (clnt->cl_metrics == NULL)
149 goto out_no_stats; 171 goto out_no_stats;
150 clnt->cl_program = program; 172 clnt->cl_program = program;
173 INIT_LIST_HEAD(&clnt->cl_tasks);
174 spin_lock_init(&clnt->cl_lock);
151 175
152 if (!xprt_bound(clnt->cl_xprt)) 176 if (!xprt_bound(clnt->cl_xprt))
153 clnt->cl_autobind = 1; 177 clnt->cl_autobind = 1;
@@ -155,6 +179,8 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s
155 clnt->cl_rtt = &clnt->cl_rtt_default; 179 clnt->cl_rtt = &clnt->cl_rtt_default;
156 rpc_init_rtt(&clnt->cl_rtt_default, xprt->timeout.to_initval); 180 rpc_init_rtt(&clnt->cl_rtt_default, xprt->timeout.to_initval);
157 181
182 kref_init(&clnt->cl_kref);
183
158 err = rpc_setup_pipedir(clnt, program->pipe_dir_name); 184 err = rpc_setup_pipedir(clnt, program->pipe_dir_name);
159 if (err < 0) 185 if (err < 0)
160 goto out_no_path; 186 goto out_no_path;
@@ -172,6 +198,7 @@ static struct rpc_clnt * rpc_new_client(struct rpc_xprt *xprt, char *servname, s
172 if (clnt->cl_nodelen > UNX_MAXNODENAME) 198 if (clnt->cl_nodelen > UNX_MAXNODENAME)
173 clnt->cl_nodelen = UNX_MAXNODENAME; 199 clnt->cl_nodelen = UNX_MAXNODENAME;
174 memcpy(clnt->cl_nodename, utsname()->nodename, clnt->cl_nodelen); 200 memcpy(clnt->cl_nodename, utsname()->nodename, clnt->cl_nodelen);
201 rpc_register_client(clnt);
175 return clnt; 202 return clnt;
176 203
177out_no_auth: 204out_no_auth:
@@ -188,6 +215,8 @@ out_no_stats:
188out_err: 215out_err:
189 xprt_put(xprt); 216 xprt_put(xprt);
190out_no_xprt: 217out_no_xprt:
218 rpciod_down();
219out_no_rpciod:
191 return ERR_PTR(err); 220 return ERR_PTR(err);
192} 221}
193 222
@@ -205,13 +234,32 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args)
205{ 234{
206 struct rpc_xprt *xprt; 235 struct rpc_xprt *xprt;
207 struct rpc_clnt *clnt; 236 struct rpc_clnt *clnt;
237 struct rpc_xprtsock_create xprtargs = {
238 .proto = args->protocol,
239 .srcaddr = args->saddress,
240 .dstaddr = args->address,
241 .addrlen = args->addrsize,
242 .timeout = args->timeout
243 };
244 char servername[20];
208 245
209 xprt = xprt_create_transport(args->protocol, args->address, 246 xprt = xprt_create_transport(&xprtargs);
210 args->addrsize, args->timeout);
211 if (IS_ERR(xprt)) 247 if (IS_ERR(xprt))
212 return (struct rpc_clnt *)xprt; 248 return (struct rpc_clnt *)xprt;
213 249
214 /* 250 /*
251 * If the caller chooses not to specify a hostname, whip
252 * up a string representation of the passed-in address.
253 */
254 if (args->servername == NULL) {
255 struct sockaddr_in *addr =
256 (struct sockaddr_in *) &args->address;
257 snprintf(servername, sizeof(servername), NIPQUAD_FMT,
258 NIPQUAD(addr->sin_addr.s_addr));
259 args->servername = servername;
260 }
261
262 /*
215 * By default, kernel RPC client connects from a reserved port. 263 * By default, kernel RPC client connects from a reserved port.
216 * CAP_NET_BIND_SERVICE will not be set for unprivileged requesters, 264 * CAP_NET_BIND_SERVICE will not be set for unprivileged requesters,
217 * but it is always enabled for rpciod, which handles the connect 265 * but it is always enabled for rpciod, which handles the connect
@@ -245,8 +293,6 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args)
245 clnt->cl_intr = 1; 293 clnt->cl_intr = 1;
246 if (args->flags & RPC_CLNT_CREATE_AUTOBIND) 294 if (args->flags & RPC_CLNT_CREATE_AUTOBIND)
247 clnt->cl_autobind = 1; 295 clnt->cl_autobind = 1;
248 if (args->flags & RPC_CLNT_CREATE_ONESHOT)
249 clnt->cl_oneshot = 1;
250 if (args->flags & RPC_CLNT_CREATE_DISCRTRY) 296 if (args->flags & RPC_CLNT_CREATE_DISCRTRY)
251 clnt->cl_discrtry = 1; 297 clnt->cl_discrtry = 1;
252 298
@@ -268,24 +314,25 @@ rpc_clone_client(struct rpc_clnt *clnt)
268 new = kmemdup(clnt, sizeof(*new), GFP_KERNEL); 314 new = kmemdup(clnt, sizeof(*new), GFP_KERNEL);
269 if (!new) 315 if (!new)
270 goto out_no_clnt; 316 goto out_no_clnt;
271 atomic_set(&new->cl_count, 1); 317 new->cl_parent = clnt;
272 atomic_set(&new->cl_users, 0); 318 /* Turn off autobind on clones */
319 new->cl_autobind = 0;
320 INIT_LIST_HEAD(&new->cl_tasks);
321 spin_lock_init(&new->cl_lock);
322 rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval);
273 new->cl_metrics = rpc_alloc_iostats(clnt); 323 new->cl_metrics = rpc_alloc_iostats(clnt);
274 if (new->cl_metrics == NULL) 324 if (new->cl_metrics == NULL)
275 goto out_no_stats; 325 goto out_no_stats;
326 kref_init(&new->cl_kref);
276 err = rpc_setup_pipedir(new, clnt->cl_program->pipe_dir_name); 327 err = rpc_setup_pipedir(new, clnt->cl_program->pipe_dir_name);
277 if (err != 0) 328 if (err != 0)
278 goto out_no_path; 329 goto out_no_path;
279 new->cl_parent = clnt;
280 atomic_inc(&clnt->cl_count);
281 new->cl_xprt = xprt_get(clnt->cl_xprt);
282 /* Turn off autobind on clones */
283 new->cl_autobind = 0;
284 new->cl_oneshot = 0;
285 new->cl_dead = 0;
286 rpc_init_rtt(&new->cl_rtt_default, clnt->cl_xprt->timeout.to_initval);
287 if (new->cl_auth) 330 if (new->cl_auth)
288 atomic_inc(&new->cl_auth->au_count); 331 atomic_inc(&new->cl_auth->au_count);
332 xprt_get(clnt->cl_xprt);
333 kref_get(&clnt->cl_kref);
334 rpc_register_client(new);
335 rpciod_up();
289 return new; 336 return new;
290out_no_path: 337out_no_path:
291 rpc_free_iostats(new->cl_metrics); 338 rpc_free_iostats(new->cl_metrics);
@@ -298,86 +345,86 @@ out_no_clnt:
298 345
299/* 346/*
300 * Properly shut down an RPC client, terminating all outstanding 347 * Properly shut down an RPC client, terminating all outstanding
301 * requests. Note that we must be certain that cl_oneshot and 348 * requests.
302 * cl_dead are cleared, or else the client would be destroyed
303 * when the last task releases it.
304 */ 349 */
305int 350void rpc_shutdown_client(struct rpc_clnt *clnt)
306rpc_shutdown_client(struct rpc_clnt *clnt)
307{ 351{
308 dprintk("RPC: shutting down %s client for %s, tasks=%d\n", 352 dprintk("RPC: shutting down %s client for %s\n",
309 clnt->cl_protname, clnt->cl_server, 353 clnt->cl_protname, clnt->cl_server);
310 atomic_read(&clnt->cl_users)); 354
311 355 while (!list_empty(&clnt->cl_tasks)) {
312 while (atomic_read(&clnt->cl_users) > 0) {
313 /* Don't let rpc_release_client destroy us */
314 clnt->cl_oneshot = 0;
315 clnt->cl_dead = 0;
316 rpc_killall_tasks(clnt); 356 rpc_killall_tasks(clnt);
317 wait_event_timeout(destroy_wait, 357 wait_event_timeout(destroy_wait,
318 !atomic_read(&clnt->cl_users), 1*HZ); 358 list_empty(&clnt->cl_tasks), 1*HZ);
319 }
320
321 if (atomic_read(&clnt->cl_users) < 0) {
322 printk(KERN_ERR "RPC: rpc_shutdown_client clnt %p tasks=%d\n",
323 clnt, atomic_read(&clnt->cl_users));
324#ifdef RPC_DEBUG
325 rpc_show_tasks();
326#endif
327 BUG();
328 } 359 }
329 360
330 return rpc_destroy_client(clnt); 361 rpc_release_client(clnt);
331} 362}
332 363
333/* 364/*
334 * Delete an RPC client 365 * Free an RPC client
335 */ 366 */
336int 367static void
337rpc_destroy_client(struct rpc_clnt *clnt) 368rpc_free_client(struct kref *kref)
338{ 369{
339 if (!atomic_dec_and_test(&clnt->cl_count)) 370 struct rpc_clnt *clnt = container_of(kref, struct rpc_clnt, cl_kref);
340 return 1;
341 BUG_ON(atomic_read(&clnt->cl_users) != 0);
342 371
343 dprintk("RPC: destroying %s client for %s\n", 372 dprintk("RPC: destroying %s client for %s\n",
344 clnt->cl_protname, clnt->cl_server); 373 clnt->cl_protname, clnt->cl_server);
345 if (clnt->cl_auth) {
346 rpcauth_destroy(clnt->cl_auth);
347 clnt->cl_auth = NULL;
348 }
349 if (!IS_ERR(clnt->cl_dentry)) { 374 if (!IS_ERR(clnt->cl_dentry)) {
350 rpc_rmdir(clnt->cl_dentry); 375 rpc_rmdir(clnt->cl_dentry);
351 rpc_put_mount(); 376 rpc_put_mount();
352 } 377 }
353 if (clnt->cl_parent != clnt) { 378 if (clnt->cl_parent != clnt) {
354 rpc_destroy_client(clnt->cl_parent); 379 rpc_release_client(clnt->cl_parent);
355 goto out_free; 380 goto out_free;
356 } 381 }
357 if (clnt->cl_server != clnt->cl_inline_name) 382 if (clnt->cl_server != clnt->cl_inline_name)
358 kfree(clnt->cl_server); 383 kfree(clnt->cl_server);
359out_free: 384out_free:
385 rpc_unregister_client(clnt);
360 rpc_free_iostats(clnt->cl_metrics); 386 rpc_free_iostats(clnt->cl_metrics);
361 clnt->cl_metrics = NULL; 387 clnt->cl_metrics = NULL;
362 xprt_put(clnt->cl_xprt); 388 xprt_put(clnt->cl_xprt);
389 rpciod_down();
363 kfree(clnt); 390 kfree(clnt);
364 return 0;
365} 391}
366 392
367/* 393/*
368 * Release an RPC client 394 * Free an RPC client
395 */
396static void
397rpc_free_auth(struct kref *kref)
398{
399 struct rpc_clnt *clnt = container_of(kref, struct rpc_clnt, cl_kref);
400
401 if (clnt->cl_auth == NULL) {
402 rpc_free_client(kref);
403 return;
404 }
405
406 /*
407 * Note: RPCSEC_GSS may need to send NULL RPC calls in order to
408 * release remaining GSS contexts. This mechanism ensures
409 * that it can do so safely.
410 */
411 kref_init(kref);
412 rpcauth_release(clnt->cl_auth);
413 clnt->cl_auth = NULL;
414 kref_put(kref, rpc_free_client);
415}
416
417/*
418 * Release reference to the RPC client
369 */ 419 */
370void 420void
371rpc_release_client(struct rpc_clnt *clnt) 421rpc_release_client(struct rpc_clnt *clnt)
372{ 422{
373 dprintk("RPC: rpc_release_client(%p, %d)\n", 423 dprintk("RPC: rpc_release_client(%p)\n", clnt);
374 clnt, atomic_read(&clnt->cl_users));
375 424
376 if (!atomic_dec_and_test(&clnt->cl_users)) 425 if (list_empty(&clnt->cl_tasks))
377 return; 426 wake_up(&destroy_wait);
378 wake_up(&destroy_wait); 427 kref_put(&clnt->cl_kref, rpc_free_auth);
379 if (clnt->cl_oneshot || clnt->cl_dead)
380 rpc_destroy_client(clnt);
381} 428}
382 429
383/** 430/**
@@ -468,82 +515,96 @@ void rpc_clnt_sigunmask(struct rpc_clnt *clnt, sigset_t *oldset)
468 rpc_restore_sigmask(oldset); 515 rpc_restore_sigmask(oldset);
469} 516}
470 517
471/* 518static
472 * New rpc_call implementation 519struct rpc_task *rpc_do_run_task(struct rpc_clnt *clnt,
520 struct rpc_message *msg,
521 int flags,
522 const struct rpc_call_ops *ops,
523 void *data)
524{
525 struct rpc_task *task, *ret;
526 sigset_t oldset;
527
528 task = rpc_new_task(clnt, flags, ops, data);
529 if (task == NULL) {
530 rpc_release_calldata(ops, data);
531 return ERR_PTR(-ENOMEM);
532 }
533
534 /* Mask signals on synchronous RPC calls and RPCSEC_GSS upcalls */
535 rpc_task_sigmask(task, &oldset);
536 if (msg != NULL) {
537 rpc_call_setup(task, msg, 0);
538 if (task->tk_status != 0) {
539 ret = ERR_PTR(task->tk_status);
540 rpc_put_task(task);
541 goto out;
542 }
543 }
544 atomic_inc(&task->tk_count);
545 rpc_execute(task);
546 ret = task;
547out:
548 rpc_restore_sigmask(&oldset);
549 return ret;
550}
551
552/**
553 * rpc_call_sync - Perform a synchronous RPC call
554 * @clnt: pointer to RPC client
555 * @msg: RPC call parameters
556 * @flags: RPC call flags
473 */ 557 */
474int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg, int flags) 558int rpc_call_sync(struct rpc_clnt *clnt, struct rpc_message *msg, int flags)
475{ 559{
476 struct rpc_task *task; 560 struct rpc_task *task;
477 sigset_t oldset; 561 int status;
478 int status;
479
480 /* If this client is slain all further I/O fails */
481 if (clnt->cl_dead)
482 return -EIO;
483 562
484 BUG_ON(flags & RPC_TASK_ASYNC); 563 BUG_ON(flags & RPC_TASK_ASYNC);
485 564
486 task = rpc_new_task(clnt, flags, &rpc_default_ops, NULL); 565 task = rpc_do_run_task(clnt, msg, flags, &rpc_default_ops, NULL);
487 if (task == NULL) 566 if (IS_ERR(task))
488 return -ENOMEM; 567 return PTR_ERR(task);
489
490 /* Mask signals on RPC calls _and_ GSS_AUTH upcalls */
491 rpc_task_sigmask(task, &oldset);
492
493 /* Set up the call info struct and execute the task */
494 rpc_call_setup(task, msg, 0);
495 if (task->tk_status == 0) {
496 atomic_inc(&task->tk_count);
497 rpc_execute(task);
498 }
499 status = task->tk_status; 568 status = task->tk_status;
500 rpc_put_task(task); 569 rpc_put_task(task);
501 rpc_restore_sigmask(&oldset);
502 return status; 570 return status;
503} 571}
504 572
505/* 573/**
506 * New rpc_call implementation 574 * rpc_call_async - Perform an asynchronous RPC call
575 * @clnt: pointer to RPC client
576 * @msg: RPC call parameters
577 * @flags: RPC call flags
578 * @ops: RPC call ops
579 * @data: user call data
507 */ 580 */
508int 581int
509rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, int flags, 582rpc_call_async(struct rpc_clnt *clnt, struct rpc_message *msg, int flags,
510 const struct rpc_call_ops *tk_ops, void *data) 583 const struct rpc_call_ops *tk_ops, void *data)
511{ 584{
512 struct rpc_task *task; 585 struct rpc_task *task;
513 sigset_t oldset;
514 int status;
515
516 /* If this client is slain all further I/O fails */
517 status = -EIO;
518 if (clnt->cl_dead)
519 goto out_release;
520
521 flags |= RPC_TASK_ASYNC;
522
523 /* Create/initialize a new RPC task */
524 status = -ENOMEM;
525 if (!(task = rpc_new_task(clnt, flags, tk_ops, data)))
526 goto out_release;
527
528 /* Mask signals on GSS_AUTH upcalls */
529 rpc_task_sigmask(task, &oldset);
530 586
531 rpc_call_setup(task, msg, 0); 587 task = rpc_do_run_task(clnt, msg, flags|RPC_TASK_ASYNC, tk_ops, data);
532 588 if (IS_ERR(task))
533 /* Set up the call info struct and execute the task */ 589 return PTR_ERR(task);
534 status = task->tk_status; 590 rpc_put_task(task);
535 if (status == 0) 591 return 0;
536 rpc_execute(task);
537 else
538 rpc_put_task(task);
539
540 rpc_restore_sigmask(&oldset);
541 return status;
542out_release:
543 rpc_release_calldata(tk_ops, data);
544 return status;
545} 592}
546 593
594/**
595 * rpc_run_task - Allocate a new RPC task, then run rpc_execute against it
596 * @clnt: pointer to RPC client
597 * @flags: RPC flags
598 * @ops: RPC call ops
599 * @data: user call data
600 */
601struct rpc_task *rpc_run_task(struct rpc_clnt *clnt, int flags,
602 const struct rpc_call_ops *tk_ops,
603 void *data)
604{
605 return rpc_do_run_task(clnt, NULL, flags, tk_ops, data);
606}
607EXPORT_SYMBOL(rpc_run_task);
547 608
548void 609void
549rpc_call_setup(struct rpc_task *task, struct rpc_message *msg, int flags) 610rpc_call_setup(struct rpc_task *task, struct rpc_message *msg, int flags)
@@ -745,7 +806,7 @@ call_reserveresult(struct rpc_task *task)
745static void 806static void
746call_allocate(struct rpc_task *task) 807call_allocate(struct rpc_task *task)
747{ 808{
748 unsigned int slack = task->tk_auth->au_cslack; 809 unsigned int slack = task->tk_msg.rpc_cred->cr_auth->au_cslack;
749 struct rpc_rqst *req = task->tk_rqstp; 810 struct rpc_rqst *req = task->tk_rqstp;
750 struct rpc_xprt *xprt = task->tk_xprt; 811 struct rpc_xprt *xprt = task->tk_xprt;
751 struct rpc_procinfo *proc = task->tk_msg.rpc_proc; 812 struct rpc_procinfo *proc = task->tk_msg.rpc_proc;
@@ -843,10 +904,8 @@ call_encode(struct rpc_task *task)
843 if (encode == NULL) 904 if (encode == NULL)
844 return; 905 return;
845 906
846 lock_kernel();
847 task->tk_status = rpcauth_wrap_req(task, encode, req, p, 907 task->tk_status = rpcauth_wrap_req(task, encode, req, p,
848 task->tk_msg.rpc_argp); 908 task->tk_msg.rpc_argp);
849 unlock_kernel();
850 if (task->tk_status == -ENOMEM) { 909 if (task->tk_status == -ENOMEM) {
851 /* XXX: Is this sane? */ 910 /* XXX: Is this sane? */
852 rpc_delay(task, 3*HZ); 911 rpc_delay(task, 3*HZ);
@@ -1177,10 +1236,8 @@ call_decode(struct rpc_task *task)
1177 task->tk_action = rpc_exit_task; 1236 task->tk_action = rpc_exit_task;
1178 1237
1179 if (decode) { 1238 if (decode) {
1180 lock_kernel();
1181 task->tk_status = rpcauth_unwrap_resp(task, decode, req, p, 1239 task->tk_status = rpcauth_unwrap_resp(task, decode, req, p,
1182 task->tk_msg.rpc_resp); 1240 task->tk_msg.rpc_resp);
1183 unlock_kernel();
1184 } 1241 }
1185 dprintk("RPC: %5u call_decode result %d\n", task->tk_pid, 1242 dprintk("RPC: %5u call_decode result %d\n", task->tk_pid,
1186 task->tk_status); 1243 task->tk_status);
@@ -1273,9 +1330,9 @@ call_verify(struct rpc_task *task)
1273 * - if it isn't pointer subtraction in the NFS client may give 1330 * - if it isn't pointer subtraction in the NFS client may give
1274 * undefined results 1331 * undefined results
1275 */ 1332 */
1276 printk(KERN_WARNING 1333 dprintk("RPC: %5u %s: XDR representation not a multiple of"
1277 "call_verify: XDR representation not a multiple of" 1334 " 4 bytes: 0x%x\n", task->tk_pid, __FUNCTION__,
1278 " 4 bytes: 0x%x\n", task->tk_rqstp->rq_rcv_buf.len); 1335 task->tk_rqstp->rq_rcv_buf.len);
1279 goto out_eio; 1336 goto out_eio;
1280 } 1337 }
1281 if ((len -= 3) < 0) 1338 if ((len -= 3) < 0)
@@ -1283,7 +1340,8 @@ call_verify(struct rpc_task *task)
1283 p += 1; /* skip XID */ 1340 p += 1; /* skip XID */
1284 1341
1285 if ((n = ntohl(*p++)) != RPC_REPLY) { 1342 if ((n = ntohl(*p++)) != RPC_REPLY) {
1286 printk(KERN_WARNING "call_verify: not an RPC reply: %x\n", n); 1343 dprintk("RPC: %5u %s: not an RPC reply: %x\n",
1344 task->tk_pid, __FUNCTION__, n);
1287 goto out_garbage; 1345 goto out_garbage;
1288 } 1346 }
1289 if ((n = ntohl(*p++)) != RPC_MSG_ACCEPTED) { 1347 if ((n = ntohl(*p++)) != RPC_MSG_ACCEPTED) {
@@ -1334,7 +1392,8 @@ call_verify(struct rpc_task *task)
1334 "authentication.\n", task->tk_client->cl_server); 1392 "authentication.\n", task->tk_client->cl_server);
1335 break; 1393 break;
1336 default: 1394 default:
1337 printk(KERN_WARNING "call_verify: unknown auth error: %x\n", n); 1395 dprintk("RPC: %5u %s: unknown auth error: %x\n",
1396 task->tk_pid, __FUNCTION__, n);
1338 error = -EIO; 1397 error = -EIO;
1339 } 1398 }
1340 dprintk("RPC: %5u %s: call rejected %d\n", 1399 dprintk("RPC: %5u %s: call rejected %d\n",
@@ -1342,7 +1401,8 @@ call_verify(struct rpc_task *task)
1342 goto out_err; 1401 goto out_err;
1343 } 1402 }
1344 if (!(p = rpcauth_checkverf(task, p))) { 1403 if (!(p = rpcauth_checkverf(task, p))) {
1345 printk(KERN_WARNING "call_verify: auth check failed\n"); 1404 dprintk("RPC: %5u %s: auth check failed\n",
1405 task->tk_pid, __FUNCTION__);
1346 goto out_garbage; /* bad verifier, retry */ 1406 goto out_garbage; /* bad verifier, retry */
1347 } 1407 }
1348 len = p - (__be32 *)iov->iov_base - 1; 1408 len = p - (__be32 *)iov->iov_base - 1;
@@ -1381,7 +1441,8 @@ call_verify(struct rpc_task *task)
1381 task->tk_pid, __FUNCTION__); 1441 task->tk_pid, __FUNCTION__);
1382 break; /* retry */ 1442 break; /* retry */
1383 default: 1443 default:
1384 printk(KERN_WARNING "call_verify: server accept status: %x\n", n); 1444 dprintk("RPC: %5u %s: server accept status: %x\n",
1445 task->tk_pid, __FUNCTION__, n);
1385 /* Also retry */ 1446 /* Also retry */
1386 } 1447 }
1387 1448
@@ -1395,14 +1456,16 @@ out_garbage:
1395out_retry: 1456out_retry:
1396 return ERR_PTR(-EAGAIN); 1457 return ERR_PTR(-EAGAIN);
1397 } 1458 }
1398 printk(KERN_WARNING "RPC %s: retry failed, exit EIO\n", __FUNCTION__);
1399out_eio: 1459out_eio:
1400 error = -EIO; 1460 error = -EIO;
1401out_err: 1461out_err:
1402 rpc_exit(task, error); 1462 rpc_exit(task, error);
1463 dprintk("RPC: %5u %s: call failed with error %d\n", task->tk_pid,
1464 __FUNCTION__, error);
1403 return ERR_PTR(error); 1465 return ERR_PTR(error);
1404out_overflow: 1466out_overflow:
1405 printk(KERN_WARNING "RPC %s: server reply was truncated.\n", __FUNCTION__); 1467 dprintk("RPC: %5u %s: server reply was truncated.\n", task->tk_pid,
1468 __FUNCTION__);
1406 goto out_garbage; 1469 goto out_garbage;
1407} 1470}
1408 1471
@@ -1421,7 +1484,7 @@ static struct rpc_procinfo rpcproc_null = {
1421 .p_decode = rpcproc_decode_null, 1484 .p_decode = rpcproc_decode_null,
1422}; 1485};
1423 1486
1424int rpc_ping(struct rpc_clnt *clnt, int flags) 1487static int rpc_ping(struct rpc_clnt *clnt, int flags)
1425{ 1488{
1426 struct rpc_message msg = { 1489 struct rpc_message msg = {
1427 .rpc_proc = &rpcproc_null, 1490 .rpc_proc = &rpcproc_null,
@@ -1432,3 +1495,51 @@ int rpc_ping(struct rpc_clnt *clnt, int flags)
1432 put_rpccred(msg.rpc_cred); 1495 put_rpccred(msg.rpc_cred);
1433 return err; 1496 return err;
1434} 1497}
1498
1499struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred, int flags)
1500{
1501 struct rpc_message msg = {
1502 .rpc_proc = &rpcproc_null,
1503 .rpc_cred = cred,
1504 };
1505 return rpc_do_run_task(clnt, &msg, flags, &rpc_default_ops, NULL);
1506}
1507EXPORT_SYMBOL(rpc_call_null);
1508
1509#ifdef RPC_DEBUG
1510void rpc_show_tasks(void)
1511{
1512 struct rpc_clnt *clnt;
1513 struct rpc_task *t;
1514
1515 spin_lock(&rpc_client_lock);
1516 if (list_empty(&all_clients))
1517 goto out;
1518 printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout "
1519 "-rpcwait -action- ---ops--\n");
1520 list_for_each_entry(clnt, &all_clients, cl_clients) {
1521 if (list_empty(&clnt->cl_tasks))
1522 continue;
1523 spin_lock(&clnt->cl_lock);
1524 list_for_each_entry(t, &clnt->cl_tasks, tk_task) {
1525 const char *rpc_waitq = "none";
1526
1527 if (RPC_IS_QUEUED(t))
1528 rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq);
1529
1530 printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n",
1531 t->tk_pid,
1532 (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1),
1533 t->tk_flags, t->tk_status,
1534 t->tk_client,
1535 (t->tk_client ? t->tk_client->cl_prog : 0),
1536 t->tk_rqstp, t->tk_timeout,
1537 rpc_waitq,
1538 t->tk_action, t->tk_ops);
1539 }
1540 spin_unlock(&clnt->cl_lock);
1541 }
1542out:
1543 spin_unlock(&rpc_client_lock);
1544}
1545#endif
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 5887457dc936..e787b6a43eee 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -344,7 +344,7 @@ rpc_info_open(struct inode *inode, struct file *file)
344 mutex_lock(&inode->i_mutex); 344 mutex_lock(&inode->i_mutex);
345 clnt = RPC_I(inode)->private; 345 clnt = RPC_I(inode)->private;
346 if (clnt) { 346 if (clnt) {
347 atomic_inc(&clnt->cl_users); 347 kref_get(&clnt->cl_kref);
348 m->private = clnt; 348 m->private = clnt;
349 } else { 349 } else {
350 single_release(inode, file); 350 single_release(inode, file);
@@ -448,6 +448,15 @@ void rpc_put_mount(void)
448 simple_release_fs(&rpc_mount, &rpc_mount_count); 448 simple_release_fs(&rpc_mount, &rpc_mount_count);
449} 449}
450 450
451static int rpc_delete_dentry(struct dentry *dentry)
452{
453 return 1;
454}
455
456static struct dentry_operations rpc_dentry_operations = {
457 .d_delete = rpc_delete_dentry,
458};
459
451static int 460static int
452rpc_lookup_parent(char *path, struct nameidata *nd) 461rpc_lookup_parent(char *path, struct nameidata *nd)
453{ 462{
@@ -506,7 +515,7 @@ rpc_get_inode(struct super_block *sb, int mode)
506 * FIXME: This probably has races. 515 * FIXME: This probably has races.
507 */ 516 */
508static void 517static void
509rpc_depopulate(struct dentry *parent) 518rpc_depopulate(struct dentry *parent, int start, int eof)
510{ 519{
511 struct inode *dir = parent->d_inode; 520 struct inode *dir = parent->d_inode;
512 struct list_head *pos, *next; 521 struct list_head *pos, *next;
@@ -518,6 +527,10 @@ repeat:
518 spin_lock(&dcache_lock); 527 spin_lock(&dcache_lock);
519 list_for_each_safe(pos, next, &parent->d_subdirs) { 528 list_for_each_safe(pos, next, &parent->d_subdirs) {
520 dentry = list_entry(pos, struct dentry, d_u.d_child); 529 dentry = list_entry(pos, struct dentry, d_u.d_child);
530 if (!dentry->d_inode ||
531 dentry->d_inode->i_ino < start ||
532 dentry->d_inode->i_ino >= eof)
533 continue;
521 spin_lock(&dentry->d_lock); 534 spin_lock(&dentry->d_lock);
522 if (!d_unhashed(dentry)) { 535 if (!d_unhashed(dentry)) {
523 dget_locked(dentry); 536 dget_locked(dentry);
@@ -533,11 +546,11 @@ repeat:
533 if (n) { 546 if (n) {
534 do { 547 do {
535 dentry = dvec[--n]; 548 dentry = dvec[--n];
536 if (dentry->d_inode) { 549 if (S_ISREG(dentry->d_inode->i_mode))
537 rpc_close_pipes(dentry->d_inode);
538 simple_unlink(dir, dentry); 550 simple_unlink(dir, dentry);
539 } 551 else if (S_ISDIR(dentry->d_inode->i_mode))
540 inode_dir_notify(dir, DN_DELETE); 552 simple_rmdir(dir, dentry);
553 d_delete(dentry);
541 dput(dentry); 554 dput(dentry);
542 } while (n); 555 } while (n);
543 goto repeat; 556 goto repeat;
@@ -560,6 +573,7 @@ rpc_populate(struct dentry *parent,
560 dentry = d_alloc_name(parent, files[i].name); 573 dentry = d_alloc_name(parent, files[i].name);
561 if (!dentry) 574 if (!dentry)
562 goto out_bad; 575 goto out_bad;
576 dentry->d_op = &rpc_dentry_operations;
563 mode = files[i].mode; 577 mode = files[i].mode;
564 inode = rpc_get_inode(dir->i_sb, mode); 578 inode = rpc_get_inode(dir->i_sb, mode);
565 if (!inode) { 579 if (!inode) {
@@ -607,21 +621,14 @@ static int
607__rpc_rmdir(struct inode *dir, struct dentry *dentry) 621__rpc_rmdir(struct inode *dir, struct dentry *dentry)
608{ 622{
609 int error; 623 int error;
610 624 error = simple_rmdir(dir, dentry);
611 shrink_dcache_parent(dentry); 625 if (!error)
612 if (d_unhashed(dentry)) 626 d_delete(dentry);
613 return 0; 627 return error;
614 if ((error = simple_rmdir(dir, dentry)) != 0)
615 return error;
616 if (!error) {
617 inode_dir_notify(dir, DN_DELETE);
618 d_drop(dentry);
619 }
620 return 0;
621} 628}
622 629
623static struct dentry * 630static struct dentry *
624rpc_lookup_create(struct dentry *parent, const char *name, int len) 631rpc_lookup_create(struct dentry *parent, const char *name, int len, int exclusive)
625{ 632{
626 struct inode *dir = parent->d_inode; 633 struct inode *dir = parent->d_inode;
627 struct dentry *dentry; 634 struct dentry *dentry;
@@ -630,7 +637,9 @@ rpc_lookup_create(struct dentry *parent, const char *name, int len)
630 dentry = lookup_one_len(name, parent, len); 637 dentry = lookup_one_len(name, parent, len);
631 if (IS_ERR(dentry)) 638 if (IS_ERR(dentry))
632 goto out_err; 639 goto out_err;
633 if (dentry->d_inode) { 640 if (!dentry->d_inode)
641 dentry->d_op = &rpc_dentry_operations;
642 else if (exclusive) {
634 dput(dentry); 643 dput(dentry);
635 dentry = ERR_PTR(-EEXIST); 644 dentry = ERR_PTR(-EEXIST);
636 goto out_err; 645 goto out_err;
@@ -649,7 +658,7 @@ rpc_lookup_negative(char *path, struct nameidata *nd)
649 658
650 if ((error = rpc_lookup_parent(path, nd)) != 0) 659 if ((error = rpc_lookup_parent(path, nd)) != 0)
651 return ERR_PTR(error); 660 return ERR_PTR(error);
652 dentry = rpc_lookup_create(nd->dentry, nd->last.name, nd->last.len); 661 dentry = rpc_lookup_create(nd->dentry, nd->last.name, nd->last.len, 1);
653 if (IS_ERR(dentry)) 662 if (IS_ERR(dentry))
654 rpc_release_path(nd); 663 rpc_release_path(nd);
655 return dentry; 664 return dentry;
@@ -681,7 +690,7 @@ out:
681 rpc_release_path(&nd); 690 rpc_release_path(&nd);
682 return dentry; 691 return dentry;
683err_depopulate: 692err_depopulate:
684 rpc_depopulate(dentry); 693 rpc_depopulate(dentry, RPCAUTH_info, RPCAUTH_EOF);
685 __rpc_rmdir(dir, dentry); 694 __rpc_rmdir(dir, dentry);
686err_dput: 695err_dput:
687 dput(dentry); 696 dput(dentry);
@@ -701,7 +710,7 @@ rpc_rmdir(struct dentry *dentry)
701 parent = dget_parent(dentry); 710 parent = dget_parent(dentry);
702 dir = parent->d_inode; 711 dir = parent->d_inode;
703 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 712 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
704 rpc_depopulate(dentry); 713 rpc_depopulate(dentry, RPCAUTH_info, RPCAUTH_EOF);
705 error = __rpc_rmdir(dir, dentry); 714 error = __rpc_rmdir(dir, dentry);
706 dput(dentry); 715 dput(dentry);
707 mutex_unlock(&dir->i_mutex); 716 mutex_unlock(&dir->i_mutex);
@@ -716,10 +725,21 @@ rpc_mkpipe(struct dentry *parent, const char *name, void *private, struct rpc_pi
716 struct inode *dir, *inode; 725 struct inode *dir, *inode;
717 struct rpc_inode *rpci; 726 struct rpc_inode *rpci;
718 727
719 dentry = rpc_lookup_create(parent, name, strlen(name)); 728 dentry = rpc_lookup_create(parent, name, strlen(name), 0);
720 if (IS_ERR(dentry)) 729 if (IS_ERR(dentry))
721 return dentry; 730 return dentry;
722 dir = parent->d_inode; 731 dir = parent->d_inode;
732 if (dentry->d_inode) {
733 rpci = RPC_I(dentry->d_inode);
734 if (rpci->private != private ||
735 rpci->ops != ops ||
736 rpci->flags != flags) {
737 dput (dentry);
738 dentry = ERR_PTR(-EBUSY);
739 }
740 rpci->nkern_readwriters++;
741 goto out;
742 }
723 inode = rpc_get_inode(dir->i_sb, S_IFIFO | S_IRUSR | S_IWUSR); 743 inode = rpc_get_inode(dir->i_sb, S_IFIFO | S_IRUSR | S_IWUSR);
724 if (!inode) 744 if (!inode)
725 goto err_dput; 745 goto err_dput;
@@ -730,6 +750,7 @@ rpc_mkpipe(struct dentry *parent, const char *name, void *private, struct rpc_pi
730 rpci->private = private; 750 rpci->private = private;
731 rpci->flags = flags; 751 rpci->flags = flags;
732 rpci->ops = ops; 752 rpci->ops = ops;
753 rpci->nkern_readwriters = 1;
733 inode_dir_notify(dir, DN_CREATE); 754 inode_dir_notify(dir, DN_CREATE);
734 dget(dentry); 755 dget(dentry);
735out: 756out:
@@ -754,13 +775,11 @@ rpc_unlink(struct dentry *dentry)
754 parent = dget_parent(dentry); 775 parent = dget_parent(dentry);
755 dir = parent->d_inode; 776 dir = parent->d_inode;
756 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT); 777 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
757 if (!d_unhashed(dentry)) { 778 if (--RPC_I(dentry->d_inode)->nkern_readwriters == 0) {
758 d_drop(dentry); 779 rpc_close_pipes(dentry->d_inode);
759 if (dentry->d_inode) { 780 error = simple_unlink(dir, dentry);
760 rpc_close_pipes(dentry->d_inode); 781 if (!error)
761 error = simple_unlink(dir, dentry); 782 d_delete(dentry);
762 }
763 inode_dir_notify(dir, DN_DELETE);
764 } 783 }
765 dput(dentry); 784 dput(dentry);
766 mutex_unlock(&dir->i_mutex); 785 mutex_unlock(&dir->i_mutex);
@@ -833,6 +852,7 @@ init_once(void * foo, struct kmem_cache * cachep, unsigned long flags)
833 rpci->nreaders = 0; 852 rpci->nreaders = 0;
834 rpci->nwriters = 0; 853 rpci->nwriters = 0;
835 INIT_LIST_HEAD(&rpci->in_upcall); 854 INIT_LIST_HEAD(&rpci->in_upcall);
855 INIT_LIST_HEAD(&rpci->in_downcall);
836 INIT_LIST_HEAD(&rpci->pipe); 856 INIT_LIST_HEAD(&rpci->pipe);
837 rpci->pipelen = 0; 857 rpci->pipelen = 0;
838 init_waitqueue_head(&rpci->waitq); 858 init_waitqueue_head(&rpci->waitq);
diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c
index 6c7aa8a1f0c6..d1740dbab991 100644
--- a/net/sunrpc/rpcb_clnt.c
+++ b/net/sunrpc/rpcb_clnt.c
@@ -12,6 +12,8 @@
12 * Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de> 12 * Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de>
13 */ 13 */
14 14
15#include <linux/module.h>
16
15#include <linux/types.h> 17#include <linux/types.h>
16#include <linux/socket.h> 18#include <linux/socket.h>
17#include <linux/kernel.h> 19#include <linux/kernel.h>
@@ -184,8 +186,8 @@ static struct rpc_clnt *rpcb_create(char *hostname, struct sockaddr *srvaddr,
184 .program = &rpcb_program, 186 .program = &rpcb_program,
185 .version = version, 187 .version = version,
186 .authflavor = RPC_AUTH_UNIX, 188 .authflavor = RPC_AUTH_UNIX,
187 .flags = (RPC_CLNT_CREATE_ONESHOT | 189 .flags = (RPC_CLNT_CREATE_NOPING |
188 RPC_CLNT_CREATE_NOPING), 190 RPC_CLNT_CREATE_INTR),
189 }; 191 };
190 192
191 ((struct sockaddr_in *)srvaddr)->sin_port = htons(RPCBIND_PORT); 193 ((struct sockaddr_in *)srvaddr)->sin_port = htons(RPCBIND_PORT);
@@ -238,6 +240,7 @@ int rpcb_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay)
238 240
239 error = rpc_call_sync(rpcb_clnt, &msg, 0); 241 error = rpc_call_sync(rpcb_clnt, &msg, 0);
240 242
243 rpc_shutdown_client(rpcb_clnt);
241 if (error < 0) 244 if (error < 0)
242 printk(KERN_WARNING "RPC: failed to contact local rpcbind " 245 printk(KERN_WARNING "RPC: failed to contact local rpcbind "
243 "server (errno %d).\n", -error); 246 "server (errno %d).\n", -error);
@@ -246,21 +249,20 @@ int rpcb_register(u32 prog, u32 vers, int prot, unsigned short port, int *okay)
246 return error; 249 return error;
247} 250}
248 251
249#ifdef CONFIG_ROOT_NFS
250/** 252/**
251 * rpcb_getport_external - obtain the port for an RPC service on a given host 253 * rpcb_getport_sync - obtain the port for an RPC service on a given host
252 * @sin: address of remote peer 254 * @sin: address of remote peer
253 * @prog: RPC program number to bind 255 * @prog: RPC program number to bind
254 * @vers: RPC version number to bind 256 * @vers: RPC version number to bind
255 * @prot: transport protocol to use to make this request 257 * @prot: transport protocol to use to make this request
256 * 258 *
257 * Called from outside the RPC client in a synchronous task context. 259 * Called from outside the RPC client in a synchronous task context.
260 * Uses default timeout parameters specified by underlying transport.
258 * 261 *
259 * For now, this supports only version 2 queries, but is used only by 262 * XXX: Needs to support IPv6, and rpcbind versions 3 and 4
260 * mount_clnt for NFS_ROOT.
261 */ 263 */
262int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog, 264int rpcb_getport_sync(struct sockaddr_in *sin, __u32 prog,
263 __u32 vers, int prot) 265 __u32 vers, int prot)
264{ 266{
265 struct rpcbind_args map = { 267 struct rpcbind_args map = {
266 .r_prog = prog, 268 .r_prog = prog,
@@ -277,15 +279,16 @@ int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog,
277 char hostname[40]; 279 char hostname[40];
278 int status; 280 int status;
279 281
280 dprintk("RPC: rpcb_getport_external(%u.%u.%u.%u, %u, %u, %d)\n", 282 dprintk("RPC: %s(" NIPQUAD_FMT ", %u, %u, %d)\n",
281 NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot); 283 __FUNCTION__, NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot);
282 284
283 sprintf(hostname, "%u.%u.%u.%u", NIPQUAD(sin->sin_addr.s_addr)); 285 sprintf(hostname, NIPQUAD_FMT, NIPQUAD(sin->sin_addr.s_addr));
284 rpcb_clnt = rpcb_create(hostname, (struct sockaddr *)sin, prot, 2, 0); 286 rpcb_clnt = rpcb_create(hostname, (struct sockaddr *)sin, prot, 2, 0);
285 if (IS_ERR(rpcb_clnt)) 287 if (IS_ERR(rpcb_clnt))
286 return PTR_ERR(rpcb_clnt); 288 return PTR_ERR(rpcb_clnt);
287 289
288 status = rpc_call_sync(rpcb_clnt, &msg, 0); 290 status = rpc_call_sync(rpcb_clnt, &msg, 0);
291 rpc_shutdown_client(rpcb_clnt);
289 292
290 if (status >= 0) { 293 if (status >= 0) {
291 if (map.r_port != 0) 294 if (map.r_port != 0)
@@ -294,16 +297,16 @@ int rpcb_getport_external(struct sockaddr_in *sin, __u32 prog,
294 } 297 }
295 return status; 298 return status;
296} 299}
297#endif 300EXPORT_SYMBOL_GPL(rpcb_getport_sync);
298 301
299/** 302/**
300 * rpcb_getport - obtain the port for a given RPC service on a given host 303 * rpcb_getport_async - obtain the port for a given RPC service on a given host
301 * @task: task that is waiting for portmapper request 304 * @task: task that is waiting for portmapper request
302 * 305 *
303 * This one can be called for an ongoing RPC request, and can be used in 306 * This one can be called for an ongoing RPC request, and can be used in
304 * an async (rpciod) context. 307 * an async (rpciod) context.
305 */ 308 */
306void rpcb_getport(struct rpc_task *task) 309void rpcb_getport_async(struct rpc_task *task)
307{ 310{
308 struct rpc_clnt *clnt = task->tk_client; 311 struct rpc_clnt *clnt = task->tk_client;
309 int bind_version; 312 int bind_version;
@@ -314,17 +317,17 @@ void rpcb_getport(struct rpc_task *task)
314 struct sockaddr addr; 317 struct sockaddr addr;
315 int status; 318 int status;
316 319
317 dprintk("RPC: %5u rpcb_getport(%s, %u, %u, %d)\n", 320 dprintk("RPC: %5u %s(%s, %u, %u, %d)\n",
318 task->tk_pid, clnt->cl_server, 321 task->tk_pid, __FUNCTION__,
319 clnt->cl_prog, clnt->cl_vers, xprt->prot); 322 clnt->cl_server, clnt->cl_prog, clnt->cl_vers, xprt->prot);
320 323
321 /* Autobind on cloned rpc clients is discouraged */ 324 /* Autobind on cloned rpc clients is discouraged */
322 BUG_ON(clnt->cl_parent != clnt); 325 BUG_ON(clnt->cl_parent != clnt);
323 326
324 if (xprt_test_and_set_binding(xprt)) { 327 if (xprt_test_and_set_binding(xprt)) {
325 status = -EACCES; /* tell caller to check again */ 328 status = -EACCES; /* tell caller to check again */
326 dprintk("RPC: %5u rpcb_getport waiting for another binder\n", 329 dprintk("RPC: %5u %s: waiting for another binder\n",
327 task->tk_pid); 330 task->tk_pid, __FUNCTION__);
328 goto bailout_nowake; 331 goto bailout_nowake;
329 } 332 }
330 333
@@ -335,27 +338,28 @@ void rpcb_getport(struct rpc_task *task)
335 /* Someone else may have bound if we slept */ 338 /* Someone else may have bound if we slept */
336 if (xprt_bound(xprt)) { 339 if (xprt_bound(xprt)) {
337 status = 0; 340 status = 0;
338 dprintk("RPC: %5u rpcb_getport already bound\n", task->tk_pid); 341 dprintk("RPC: %5u %s: already bound\n",
342 task->tk_pid, __FUNCTION__);
339 goto bailout_nofree; 343 goto bailout_nofree;
340 } 344 }
341 345
342 if (rpcb_next_version[xprt->bind_index].rpc_proc == NULL) { 346 if (rpcb_next_version[xprt->bind_index].rpc_proc == NULL) {
343 xprt->bind_index = 0; 347 xprt->bind_index = 0;
344 status = -EACCES; /* tell caller to try again later */ 348 status = -EACCES; /* tell caller to try again later */
345 dprintk("RPC: %5u rpcb_getport no more getport versions " 349 dprintk("RPC: %5u %s: no more getport versions available\n",
346 "available\n", task->tk_pid); 350 task->tk_pid, __FUNCTION__);
347 goto bailout_nofree; 351 goto bailout_nofree;
348 } 352 }
349 bind_version = rpcb_next_version[xprt->bind_index].rpc_vers; 353 bind_version = rpcb_next_version[xprt->bind_index].rpc_vers;
350 354
351 dprintk("RPC: %5u rpcb_getport trying rpcbind version %u\n", 355 dprintk("RPC: %5u %s: trying rpcbind version %u\n",
352 task->tk_pid, bind_version); 356 task->tk_pid, __FUNCTION__, bind_version);
353 357
354 map = kzalloc(sizeof(struct rpcbind_args), GFP_ATOMIC); 358 map = kzalloc(sizeof(struct rpcbind_args), GFP_ATOMIC);
355 if (!map) { 359 if (!map) {
356 status = -ENOMEM; 360 status = -ENOMEM;
357 dprintk("RPC: %5u rpcb_getport no memory available\n", 361 dprintk("RPC: %5u %s: no memory available\n",
358 task->tk_pid); 362 task->tk_pid, __FUNCTION__);
359 goto bailout_nofree; 363 goto bailout_nofree;
360 } 364 }
361 map->r_prog = clnt->cl_prog; 365 map->r_prog = clnt->cl_prog;
@@ -373,16 +377,17 @@ void rpcb_getport(struct rpc_task *task)
373 rpcb_clnt = rpcb_create(clnt->cl_server, &addr, xprt->prot, bind_version, 0); 377 rpcb_clnt = rpcb_create(clnt->cl_server, &addr, xprt->prot, bind_version, 0);
374 if (IS_ERR(rpcb_clnt)) { 378 if (IS_ERR(rpcb_clnt)) {
375 status = PTR_ERR(rpcb_clnt); 379 status = PTR_ERR(rpcb_clnt);
376 dprintk("RPC: %5u rpcb_getport rpcb_create failed, error %ld\n", 380 dprintk("RPC: %5u %s: rpcb_create failed, error %ld\n",
377 task->tk_pid, PTR_ERR(rpcb_clnt)); 381 task->tk_pid, __FUNCTION__, PTR_ERR(rpcb_clnt));
378 goto bailout; 382 goto bailout;
379 } 383 }
380 384
381 child = rpc_run_task(rpcb_clnt, RPC_TASK_ASYNC, &rpcb_getport_ops, map); 385 child = rpc_run_task(rpcb_clnt, RPC_TASK_ASYNC, &rpcb_getport_ops, map);
386 rpc_release_client(rpcb_clnt);
382 if (IS_ERR(child)) { 387 if (IS_ERR(child)) {
383 status = -EIO; 388 status = -EIO;
384 dprintk("RPC: %5u rpcb_getport rpc_run_task failed\n", 389 dprintk("RPC: %5u %s: rpc_run_task failed\n",
385 task->tk_pid); 390 task->tk_pid, __FUNCTION__);
386 goto bailout_nofree; 391 goto bailout_nofree;
387 } 392 }
388 rpc_put_task(child); 393 rpc_put_task(child);
diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 944d75396fb3..2ac43c41c3a9 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -25,7 +25,6 @@
25#ifdef RPC_DEBUG 25#ifdef RPC_DEBUG
26#define RPCDBG_FACILITY RPCDBG_SCHED 26#define RPCDBG_FACILITY RPCDBG_SCHED
27#define RPC_TASK_MAGIC_ID 0xf00baa 27#define RPC_TASK_MAGIC_ID 0xf00baa
28static int rpc_task_id;
29#endif 28#endif
30 29
31/* 30/*
@@ -40,7 +39,6 @@ static mempool_t *rpc_task_mempool __read_mostly;
40static mempool_t *rpc_buffer_mempool __read_mostly; 39static mempool_t *rpc_buffer_mempool __read_mostly;
41 40
42static void __rpc_default_timer(struct rpc_task *task); 41static void __rpc_default_timer(struct rpc_task *task);
43static void rpciod_killall(void);
44static void rpc_async_schedule(struct work_struct *); 42static void rpc_async_schedule(struct work_struct *);
45static void rpc_release_task(struct rpc_task *task); 43static void rpc_release_task(struct rpc_task *task);
46 44
@@ -50,23 +48,13 @@ static void rpc_release_task(struct rpc_task *task);
50static RPC_WAITQ(delay_queue, "delayq"); 48static RPC_WAITQ(delay_queue, "delayq");
51 49
52/* 50/*
53 * All RPC tasks are linked into this list
54 */
55static LIST_HEAD(all_tasks);
56
57/*
58 * rpciod-related stuff 51 * rpciod-related stuff
59 */ 52 */
60static DEFINE_MUTEX(rpciod_mutex); 53static DEFINE_MUTEX(rpciod_mutex);
61static unsigned int rpciod_users; 54static atomic_t rpciod_users = ATOMIC_INIT(0);
62struct workqueue_struct *rpciod_workqueue; 55struct workqueue_struct *rpciod_workqueue;
63 56
64/* 57/*
65 * Spinlock for other critical sections of code.
66 */
67static DEFINE_SPINLOCK(rpc_sched_lock);
68
69/*
70 * Disable the timer for a given RPC task. Should be called with 58 * Disable the timer for a given RPC task. Should be called with
71 * queue->lock and bh_disabled in order to avoid races within 59 * queue->lock and bh_disabled in order to avoid races within
72 * rpc_run_timer(). 60 * rpc_run_timer().
@@ -267,18 +255,33 @@ static int rpc_wait_bit_interruptible(void *word)
267 return 0; 255 return 0;
268} 256}
269 257
258#ifdef RPC_DEBUG
259static void rpc_task_set_debuginfo(struct rpc_task *task)
260{
261 static atomic_t rpc_pid;
262
263 task->tk_magic = RPC_TASK_MAGIC_ID;
264 task->tk_pid = atomic_inc_return(&rpc_pid);
265}
266#else
267static inline void rpc_task_set_debuginfo(struct rpc_task *task)
268{
269}
270#endif
271
270static void rpc_set_active(struct rpc_task *task) 272static void rpc_set_active(struct rpc_task *task)
271{ 273{
274 struct rpc_clnt *clnt;
272 if (test_and_set_bit(RPC_TASK_ACTIVE, &task->tk_runstate) != 0) 275 if (test_and_set_bit(RPC_TASK_ACTIVE, &task->tk_runstate) != 0)
273 return; 276 return;
274 spin_lock(&rpc_sched_lock); 277 rpc_task_set_debuginfo(task);
275#ifdef RPC_DEBUG
276 task->tk_magic = RPC_TASK_MAGIC_ID;
277 task->tk_pid = rpc_task_id++;
278#endif
279 /* Add to global list of all tasks */ 278 /* Add to global list of all tasks */
280 list_add_tail(&task->tk_task, &all_tasks); 279 clnt = task->tk_client;
281 spin_unlock(&rpc_sched_lock); 280 if (clnt != NULL) {
281 spin_lock(&clnt->cl_lock);
282 list_add_tail(&task->tk_task, &clnt->cl_tasks);
283 spin_unlock(&clnt->cl_lock);
284 }
282} 285}
283 286
284/* 287/*
@@ -818,6 +821,7 @@ void rpc_init_task(struct rpc_task *task, struct rpc_clnt *clnt, int flags, cons
818 if (tk_ops->rpc_call_prepare != NULL) 821 if (tk_ops->rpc_call_prepare != NULL)
819 task->tk_action = rpc_prepare_task; 822 task->tk_action = rpc_prepare_task;
820 task->tk_calldata = calldata; 823 task->tk_calldata = calldata;
824 INIT_LIST_HEAD(&task->tk_task);
821 825
822 /* Initialize retry counters */ 826 /* Initialize retry counters */
823 task->tk_garb_retry = 2; 827 task->tk_garb_retry = 2;
@@ -830,7 +834,7 @@ void rpc_init_task(struct rpc_task *task, struct rpc_clnt *clnt, int flags, cons
830 task->tk_workqueue = rpciod_workqueue; 834 task->tk_workqueue = rpciod_workqueue;
831 835
832 if (clnt) { 836 if (clnt) {
833 atomic_inc(&clnt->cl_users); 837 kref_get(&clnt->cl_kref);
834 if (clnt->cl_softrtry) 838 if (clnt->cl_softrtry)
835 task->tk_flags |= RPC_TASK_SOFT; 839 task->tk_flags |= RPC_TASK_SOFT;
836 if (!clnt->cl_intr) 840 if (!clnt->cl_intr)
@@ -860,9 +864,7 @@ static void rpc_free_task(struct rcu_head *rcu)
860} 864}
861 865
862/* 866/*
863 * Create a new task for the specified client. We have to 867 * Create a new task for the specified client.
864 * clean up after an allocation failure, as the client may
865 * have specified "oneshot".
866 */ 868 */
867struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc_call_ops *tk_ops, void *calldata) 869struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc_call_ops *tk_ops, void *calldata)
868{ 870{
@@ -870,7 +872,7 @@ struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc
870 872
871 task = rpc_alloc_task(); 873 task = rpc_alloc_task();
872 if (!task) 874 if (!task)
873 goto cleanup; 875 goto out;
874 876
875 rpc_init_task(task, clnt, flags, tk_ops, calldata); 877 rpc_init_task(task, clnt, flags, tk_ops, calldata);
876 878
@@ -878,16 +880,6 @@ struct rpc_task *rpc_new_task(struct rpc_clnt *clnt, int flags, const struct rpc
878 task->tk_flags |= RPC_TASK_DYNAMIC; 880 task->tk_flags |= RPC_TASK_DYNAMIC;
879out: 881out:
880 return task; 882 return task;
881
882cleanup:
883 /* Check whether to release the client */
884 if (clnt) {
885 printk("rpc_new_task: failed, users=%d, oneshot=%d\n",
886 atomic_read(&clnt->cl_users), clnt->cl_oneshot);
887 atomic_inc(&clnt->cl_users); /* pretend we were used ... */
888 rpc_release_client(clnt);
889 }
890 goto out;
891} 883}
892 884
893 885
@@ -920,11 +912,13 @@ static void rpc_release_task(struct rpc_task *task)
920#endif 912#endif
921 dprintk("RPC: %5u release task\n", task->tk_pid); 913 dprintk("RPC: %5u release task\n", task->tk_pid);
922 914
923 /* Remove from global task list */ 915 if (!list_empty(&task->tk_task)) {
924 spin_lock(&rpc_sched_lock); 916 struct rpc_clnt *clnt = task->tk_client;
925 list_del(&task->tk_task); 917 /* Remove from client task list */
926 spin_unlock(&rpc_sched_lock); 918 spin_lock(&clnt->cl_lock);
927 919 list_del(&task->tk_task);
920 spin_unlock(&clnt->cl_lock);
921 }
928 BUG_ON (RPC_IS_QUEUED(task)); 922 BUG_ON (RPC_IS_QUEUED(task));
929 923
930 /* Synchronously delete any running timer */ 924 /* Synchronously delete any running timer */
@@ -939,29 +933,6 @@ static void rpc_release_task(struct rpc_task *task)
939 rpc_put_task(task); 933 rpc_put_task(task);
940} 934}
941 935
942/**
943 * rpc_run_task - Allocate a new RPC task, then run rpc_execute against it
944 * @clnt: pointer to RPC client
945 * @flags: RPC flags
946 * @ops: RPC call ops
947 * @data: user call data
948 */
949struct rpc_task *rpc_run_task(struct rpc_clnt *clnt, int flags,
950 const struct rpc_call_ops *ops,
951 void *data)
952{
953 struct rpc_task *task;
954 task = rpc_new_task(clnt, flags, ops, data);
955 if (task == NULL) {
956 rpc_release_calldata(ops, data);
957 return ERR_PTR(-ENOMEM);
958 }
959 atomic_inc(&task->tk_count);
960 rpc_execute(task);
961 return task;
962}
963EXPORT_SYMBOL(rpc_run_task);
964
965/* 936/*
966 * Kill all tasks for the given client. 937 * Kill all tasks for the given client.
967 * XXX: kill their descendants as well? 938 * XXX: kill their descendants as well?
@@ -969,44 +940,25 @@ EXPORT_SYMBOL(rpc_run_task);
969void rpc_killall_tasks(struct rpc_clnt *clnt) 940void rpc_killall_tasks(struct rpc_clnt *clnt)
970{ 941{
971 struct rpc_task *rovr; 942 struct rpc_task *rovr;
972 struct list_head *le;
973 943
974 dprintk("RPC: killing all tasks for client %p\n", clnt);
975 944
945 if (list_empty(&clnt->cl_tasks))
946 return;
947 dprintk("RPC: killing all tasks for client %p\n", clnt);
976 /* 948 /*
977 * Spin lock all_tasks to prevent changes... 949 * Spin lock all_tasks to prevent changes...
978 */ 950 */
979 spin_lock(&rpc_sched_lock); 951 spin_lock(&clnt->cl_lock);
980 alltask_for_each(rovr, le, &all_tasks) { 952 list_for_each_entry(rovr, &clnt->cl_tasks, tk_task) {
981 if (! RPC_IS_ACTIVATED(rovr)) 953 if (! RPC_IS_ACTIVATED(rovr))
982 continue; 954 continue;
983 if (!clnt || rovr->tk_client == clnt) { 955 if (!(rovr->tk_flags & RPC_TASK_KILLED)) {
984 rovr->tk_flags |= RPC_TASK_KILLED; 956 rovr->tk_flags |= RPC_TASK_KILLED;
985 rpc_exit(rovr, -EIO); 957 rpc_exit(rovr, -EIO);
986 rpc_wake_up_task(rovr); 958 rpc_wake_up_task(rovr);
987 } 959 }
988 } 960 }
989 spin_unlock(&rpc_sched_lock); 961 spin_unlock(&clnt->cl_lock);
990}
991
992static void rpciod_killall(void)
993{
994 unsigned long flags;
995
996 while (!list_empty(&all_tasks)) {
997 clear_thread_flag(TIF_SIGPENDING);
998 rpc_killall_tasks(NULL);
999 flush_workqueue(rpciod_workqueue);
1000 if (!list_empty(&all_tasks)) {
1001 dprintk("RPC: rpciod_killall: waiting for tasks "
1002 "to exit\n");
1003 yield();
1004 }
1005 }
1006
1007 spin_lock_irqsave(&current->sighand->siglock, flags);
1008 recalc_sigpending();
1009 spin_unlock_irqrestore(&current->sighand->siglock, flags);
1010} 962}
1011 963
1012/* 964/*
@@ -1018,28 +970,27 @@ rpciod_up(void)
1018 struct workqueue_struct *wq; 970 struct workqueue_struct *wq;
1019 int error = 0; 971 int error = 0;
1020 972
973 if (atomic_inc_not_zero(&rpciod_users))
974 return 0;
975
1021 mutex_lock(&rpciod_mutex); 976 mutex_lock(&rpciod_mutex);
1022 dprintk("RPC: rpciod_up: users %u\n", rpciod_users); 977
1023 rpciod_users++; 978 /* Guard against races with rpciod_down() */
1024 if (rpciod_workqueue) 979 if (rpciod_workqueue != NULL)
1025 goto out; 980 goto out_ok;
1026 /*
1027 * If there's no pid, we should be the first user.
1028 */
1029 if (rpciod_users > 1)
1030 printk(KERN_WARNING "rpciod_up: no workqueue, %u users??\n", rpciod_users);
1031 /* 981 /*
1032 * Create the rpciod thread and wait for it to start. 982 * Create the rpciod thread and wait for it to start.
1033 */ 983 */
984 dprintk("RPC: creating workqueue rpciod\n");
1034 error = -ENOMEM; 985 error = -ENOMEM;
1035 wq = create_workqueue("rpciod"); 986 wq = create_workqueue("rpciod");
1036 if (wq == NULL) { 987 if (wq == NULL)
1037 printk(KERN_WARNING "rpciod_up: create workqueue failed, error=%d\n", error);
1038 rpciod_users--;
1039 goto out; 988 goto out;
1040 } 989
1041 rpciod_workqueue = wq; 990 rpciod_workqueue = wq;
1042 error = 0; 991 error = 0;
992out_ok:
993 atomic_inc(&rpciod_users);
1043out: 994out:
1044 mutex_unlock(&rpciod_mutex); 995 mutex_unlock(&rpciod_mutex);
1045 return error; 996 return error;
@@ -1048,59 +999,19 @@ out:
1048void 999void
1049rpciod_down(void) 1000rpciod_down(void)
1050{ 1001{
1002 if (!atomic_dec_and_test(&rpciod_users))
1003 return;
1004
1051 mutex_lock(&rpciod_mutex); 1005 mutex_lock(&rpciod_mutex);
1052 dprintk("RPC: rpciod_down sema %u\n", rpciod_users); 1006 dprintk("RPC: destroying workqueue rpciod\n");
1053 if (rpciod_users) {
1054 if (--rpciod_users)
1055 goto out;
1056 } else
1057 printk(KERN_WARNING "rpciod_down: no users??\n");
1058 1007
1059 if (!rpciod_workqueue) { 1008 if (atomic_read(&rpciod_users) == 0 && rpciod_workqueue != NULL) {
1060 dprintk("RPC: rpciod_down: Nothing to do!\n"); 1009 destroy_workqueue(rpciod_workqueue);
1061 goto out; 1010 rpciod_workqueue = NULL;
1062 } 1011 }
1063 rpciod_killall();
1064
1065 destroy_workqueue(rpciod_workqueue);
1066 rpciod_workqueue = NULL;
1067 out:
1068 mutex_unlock(&rpciod_mutex); 1012 mutex_unlock(&rpciod_mutex);
1069} 1013}
1070 1014
1071#ifdef RPC_DEBUG
1072void rpc_show_tasks(void)
1073{
1074 struct list_head *le;
1075 struct rpc_task *t;
1076
1077 spin_lock(&rpc_sched_lock);
1078 if (list_empty(&all_tasks)) {
1079 spin_unlock(&rpc_sched_lock);
1080 return;
1081 }
1082 printk("-pid- proc flgs status -client- -prog- --rqstp- -timeout "
1083 "-rpcwait -action- ---ops--\n");
1084 alltask_for_each(t, le, &all_tasks) {
1085 const char *rpc_waitq = "none";
1086
1087 if (RPC_IS_QUEUED(t))
1088 rpc_waitq = rpc_qname(t->u.tk_wait.rpc_waitq);
1089
1090 printk("%5u %04d %04x %6d %8p %6d %8p %8ld %8s %8p %8p\n",
1091 t->tk_pid,
1092 (t->tk_msg.rpc_proc ? t->tk_msg.rpc_proc->p_proc : -1),
1093 t->tk_flags, t->tk_status,
1094 t->tk_client,
1095 (t->tk_client ? t->tk_client->cl_prog : 0),
1096 t->tk_rqstp, t->tk_timeout,
1097 rpc_waitq,
1098 t->tk_action, t->tk_ops);
1099 }
1100 spin_unlock(&rpc_sched_lock);
1101}
1102#endif
1103
1104void 1015void
1105rpc_destroy_mempool(void) 1016rpc_destroy_mempool(void)
1106{ 1017{
diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c
index 73075dec83c0..384c4ad5ab86 100644
--- a/net/sunrpc/sunrpc_syms.c
+++ b/net/sunrpc/sunrpc_syms.c
@@ -28,15 +28,11 @@ EXPORT_SYMBOL(rpc_init_task);
28EXPORT_SYMBOL(rpc_sleep_on); 28EXPORT_SYMBOL(rpc_sleep_on);
29EXPORT_SYMBOL(rpc_wake_up_next); 29EXPORT_SYMBOL(rpc_wake_up_next);
30EXPORT_SYMBOL(rpc_wake_up_task); 30EXPORT_SYMBOL(rpc_wake_up_task);
31EXPORT_SYMBOL(rpciod_down);
32EXPORT_SYMBOL(rpciod_up);
33EXPORT_SYMBOL(rpc_new_task);
34EXPORT_SYMBOL(rpc_wake_up_status); 31EXPORT_SYMBOL(rpc_wake_up_status);
35 32
36/* RPC client functions */ 33/* RPC client functions */
37EXPORT_SYMBOL(rpc_clone_client); 34EXPORT_SYMBOL(rpc_clone_client);
38EXPORT_SYMBOL(rpc_bind_new_program); 35EXPORT_SYMBOL(rpc_bind_new_program);
39EXPORT_SYMBOL(rpc_destroy_client);
40EXPORT_SYMBOL(rpc_shutdown_client); 36EXPORT_SYMBOL(rpc_shutdown_client);
41EXPORT_SYMBOL(rpc_killall_tasks); 37EXPORT_SYMBOL(rpc_killall_tasks);
42EXPORT_SYMBOL(rpc_call_sync); 38EXPORT_SYMBOL(rpc_call_sync);
@@ -61,7 +57,7 @@ EXPORT_SYMBOL(rpcauth_unregister);
61EXPORT_SYMBOL(rpcauth_create); 57EXPORT_SYMBOL(rpcauth_create);
62EXPORT_SYMBOL(rpcauth_lookupcred); 58EXPORT_SYMBOL(rpcauth_lookupcred);
63EXPORT_SYMBOL(rpcauth_lookup_credcache); 59EXPORT_SYMBOL(rpcauth_lookup_credcache);
64EXPORT_SYMBOL(rpcauth_free_credcache); 60EXPORT_SYMBOL(rpcauth_destroy_credcache);
65EXPORT_SYMBOL(rpcauth_init_credcache); 61EXPORT_SYMBOL(rpcauth_init_credcache);
66EXPORT_SYMBOL(put_rpccred); 62EXPORT_SYMBOL(put_rpccred);
67 63
@@ -156,6 +152,7 @@ init_sunrpc(void)
156 cache_register(&ip_map_cache); 152 cache_register(&ip_map_cache);
157 cache_register(&unix_gid_cache); 153 cache_register(&unix_gid_cache);
158 init_socket_xprt(); 154 init_socket_xprt();
155 rpcauth_init_module();
159out: 156out:
160 return err; 157 return err;
161} 158}
@@ -163,6 +160,7 @@ out:
163static void __exit 160static void __exit
164cleanup_sunrpc(void) 161cleanup_sunrpc(void)
165{ 162{
163 rpcauth_remove_module();
166 cleanup_socket_xprt(); 164 cleanup_socket_xprt();
167 unregister_rpc_pipefs(); 165 unregister_rpc_pipefs();
168 rpc_destroy_mempool(); 166 rpc_destroy_mempool();
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 5baf48de2558..64b9b8c743c4 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -644,6 +644,7 @@ svc_recvfrom(struct svc_rqst *rqstp, struct kvec *iov, int nr, int buflen)
644 struct msghdr msg = { 644 struct msghdr msg = {
645 .msg_flags = MSG_DONTWAIT, 645 .msg_flags = MSG_DONTWAIT,
646 }; 646 };
647 struct sockaddr *sin;
647 int len; 648 int len;
648 649
649 len = kernel_recvmsg(svsk->sk_sock, &msg, iov, nr, buflen, 650 len = kernel_recvmsg(svsk->sk_sock, &msg, iov, nr, buflen,
@@ -654,6 +655,19 @@ svc_recvfrom(struct svc_rqst *rqstp, struct kvec *iov, int nr, int buflen)
654 memcpy(&rqstp->rq_addr, &svsk->sk_remote, svsk->sk_remotelen); 655 memcpy(&rqstp->rq_addr, &svsk->sk_remote, svsk->sk_remotelen);
655 rqstp->rq_addrlen = svsk->sk_remotelen; 656 rqstp->rq_addrlen = svsk->sk_remotelen;
656 657
658 /* Destination address in request is needed for binding the
659 * source address in RPC callbacks later.
660 */
661 sin = (struct sockaddr *)&svsk->sk_local;
662 switch (sin->sa_family) {
663 case AF_INET:
664 rqstp->rq_daddr.addr = ((struct sockaddr_in *)sin)->sin_addr;
665 break;
666 case AF_INET6:
667 rqstp->rq_daddr.addr6 = ((struct sockaddr_in6 *)sin)->sin6_addr;
668 break;
669 }
670
657 dprintk("svc: socket %p recvfrom(%p, %Zu) = %d\n", 671 dprintk("svc: socket %p recvfrom(%p, %Zu) = %d\n",
658 svsk, iov[0].iov_base, iov[0].iov_len, len); 672 svsk, iov[0].iov_base, iov[0].iov_len, len);
659 673
@@ -1064,6 +1078,12 @@ svc_tcp_accept(struct svc_sock *svsk)
1064 goto failed; 1078 goto failed;
1065 memcpy(&newsvsk->sk_remote, sin, slen); 1079 memcpy(&newsvsk->sk_remote, sin, slen);
1066 newsvsk->sk_remotelen = slen; 1080 newsvsk->sk_remotelen = slen;
1081 err = kernel_getsockname(newsock, sin, &slen);
1082 if (unlikely(err < 0)) {
1083 dprintk("svc_tcp_accept: kernel_getsockname error %d\n", -err);
1084 slen = offsetof(struct sockaddr, sa_data);
1085 }
1086 memcpy(&newsvsk->sk_local, sin, slen);
1067 1087
1068 svc_sock_received(newsvsk); 1088 svc_sock_received(newsvsk);
1069 1089
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 5b05b73e4c1d..c8c2edccad7e 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -127,7 +127,7 @@ static void xprt_clear_locked(struct rpc_xprt *xprt)
127 clear_bit(XPRT_LOCKED, &xprt->state); 127 clear_bit(XPRT_LOCKED, &xprt->state);
128 smp_mb__after_clear_bit(); 128 smp_mb__after_clear_bit();
129 } else 129 } else
130 schedule_work(&xprt->task_cleanup); 130 queue_work(rpciod_workqueue, &xprt->task_cleanup);
131} 131}
132 132
133/* 133/*
@@ -515,7 +515,7 @@ xprt_init_autodisconnect(unsigned long data)
515 if (xprt_connecting(xprt)) 515 if (xprt_connecting(xprt))
516 xprt_release_write(xprt, NULL); 516 xprt_release_write(xprt, NULL);
517 else 517 else
518 schedule_work(&xprt->task_cleanup); 518 queue_work(rpciod_workqueue, &xprt->task_cleanup);
519 return; 519 return;
520out_abort: 520out_abort:
521 spin_unlock(&xprt->transport_lock); 521 spin_unlock(&xprt->transport_lock);
@@ -886,27 +886,24 @@ void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long i
886 886
887/** 887/**
888 * xprt_create_transport - create an RPC transport 888 * xprt_create_transport - create an RPC transport
889 * @proto: requested transport protocol 889 * @args: rpc transport creation arguments
890 * @ap: remote peer address
891 * @size: length of address
892 * @to: timeout parameters
893 * 890 *
894 */ 891 */
895struct rpc_xprt *xprt_create_transport(int proto, struct sockaddr *ap, size_t size, struct rpc_timeout *to) 892struct rpc_xprt *xprt_create_transport(struct rpc_xprtsock_create *args)
896{ 893{
897 struct rpc_xprt *xprt; 894 struct rpc_xprt *xprt;
898 struct rpc_rqst *req; 895 struct rpc_rqst *req;
899 896
900 switch (proto) { 897 switch (args->proto) {
901 case IPPROTO_UDP: 898 case IPPROTO_UDP:
902 xprt = xs_setup_udp(ap, size, to); 899 xprt = xs_setup_udp(args);
903 break; 900 break;
904 case IPPROTO_TCP: 901 case IPPROTO_TCP:
905 xprt = xs_setup_tcp(ap, size, to); 902 xprt = xs_setup_tcp(args);
906 break; 903 break;
907 default: 904 default:
908 printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n", 905 printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n",
909 proto); 906 args->proto);
910 return ERR_PTR(-EIO); 907 return ERR_PTR(-EIO);
911 } 908 }
912 if (IS_ERR(xprt)) { 909 if (IS_ERR(xprt)) {
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index cc33c5880abb..4ae7eed7f617 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -235,6 +235,7 @@ struct sock_xprt {
235 * Connection of transports 235 * Connection of transports
236 */ 236 */
237 struct delayed_work connect_worker; 237 struct delayed_work connect_worker;
238 struct sockaddr_storage addr;
238 unsigned short port; 239 unsigned short port;
239 240
240 /* 241 /*
@@ -653,8 +654,7 @@ static void xs_destroy(struct rpc_xprt *xprt)
653 654
654 dprintk("RPC: xs_destroy xprt %p\n", xprt); 655 dprintk("RPC: xs_destroy xprt %p\n", xprt);
655 656
656 cancel_delayed_work(&transport->connect_worker); 657 cancel_rearming_delayed_work(&transport->connect_worker);
657 flush_scheduled_work();
658 658
659 xprt_disconnect(xprt); 659 xprt_disconnect(xprt);
660 xs_close(xprt); 660 xs_close(xprt);
@@ -1001,7 +1001,7 @@ static void xs_tcp_state_change(struct sock *sk)
1001 /* Try to schedule an autoclose RPC calls */ 1001 /* Try to schedule an autoclose RPC calls */
1002 set_bit(XPRT_CLOSE_WAIT, &xprt->state); 1002 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
1003 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0) 1003 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
1004 schedule_work(&xprt->task_cleanup); 1004 queue_work(rpciod_workqueue, &xprt->task_cleanup);
1005 default: 1005 default:
1006 xprt_disconnect(xprt); 1006 xprt_disconnect(xprt);
1007 } 1007 }
@@ -1146,31 +1146,36 @@ static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1146 sap->sin_port = htons(port); 1146 sap->sin_port = htons(port);
1147} 1147}
1148 1148
1149static int xs_bindresvport(struct sock_xprt *transport, struct socket *sock) 1149static int xs_bind(struct sock_xprt *transport, struct socket *sock)
1150{ 1150{
1151 struct sockaddr_in myaddr = { 1151 struct sockaddr_in myaddr = {
1152 .sin_family = AF_INET, 1152 .sin_family = AF_INET,
1153 }; 1153 };
1154 struct sockaddr_in *sa;
1154 int err; 1155 int err;
1155 unsigned short port = transport->port; 1156 unsigned short port = transport->port;
1156 1157
1158 if (!transport->xprt.resvport)
1159 port = 0;
1160 sa = (struct sockaddr_in *)&transport->addr;
1161 myaddr.sin_addr = sa->sin_addr;
1157 do { 1162 do {
1158 myaddr.sin_port = htons(port); 1163 myaddr.sin_port = htons(port);
1159 err = kernel_bind(sock, (struct sockaddr *) &myaddr, 1164 err = kernel_bind(sock, (struct sockaddr *) &myaddr,
1160 sizeof(myaddr)); 1165 sizeof(myaddr));
1166 if (!transport->xprt.resvport)
1167 break;
1161 if (err == 0) { 1168 if (err == 0) {
1162 transport->port = port; 1169 transport->port = port;
1163 dprintk("RPC: xs_bindresvport bound to port %u\n", 1170 break;
1164 port);
1165 return 0;
1166 } 1171 }
1167 if (port <= xprt_min_resvport) 1172 if (port <= xprt_min_resvport)
1168 port = xprt_max_resvport; 1173 port = xprt_max_resvport;
1169 else 1174 else
1170 port--; 1175 port--;
1171 } while (err == -EADDRINUSE && port != transport->port); 1176 } while (err == -EADDRINUSE && port != transport->port);
1172 1177 dprintk("RPC: xs_bind "NIPQUAD_FMT":%u: %s (%d)\n",
1173 dprintk("RPC: can't bind to reserved port (%d).\n", -err); 1178 NIPQUAD(myaddr.sin_addr), port, err ? "failed" : "ok", err);
1174 return err; 1179 return err;
1175} 1180}
1176 1181
@@ -1229,7 +1234,7 @@ static void xs_udp_connect_worker(struct work_struct *work)
1229 } 1234 }
1230 xs_reclassify_socket(sock); 1235 xs_reclassify_socket(sock);
1231 1236
1232 if (xprt->resvport && xs_bindresvport(transport, sock) < 0) { 1237 if (xs_bind(transport, sock)) {
1233 sock_release(sock); 1238 sock_release(sock);
1234 goto out; 1239 goto out;
1235 } 1240 }
@@ -1316,7 +1321,7 @@ static void xs_tcp_connect_worker(struct work_struct *work)
1316 } 1321 }
1317 xs_reclassify_socket(sock); 1322 xs_reclassify_socket(sock);
1318 1323
1319 if (xprt->resvport && xs_bindresvport(transport, sock) < 0) { 1324 if (xs_bind(transport, sock)) {
1320 sock_release(sock); 1325 sock_release(sock);
1321 goto out; 1326 goto out;
1322 } 1327 }
@@ -1410,18 +1415,16 @@ static void xs_connect(struct rpc_task *task)
1410 dprintk("RPC: xs_connect delayed xprt %p for %lu " 1415 dprintk("RPC: xs_connect delayed xprt %p for %lu "
1411 "seconds\n", 1416 "seconds\n",
1412 xprt, xprt->reestablish_timeout / HZ); 1417 xprt, xprt->reestablish_timeout / HZ);
1413 schedule_delayed_work(&transport->connect_worker, 1418 queue_delayed_work(rpciod_workqueue,
1414 xprt->reestablish_timeout); 1419 &transport->connect_worker,
1420 xprt->reestablish_timeout);
1415 xprt->reestablish_timeout <<= 1; 1421 xprt->reestablish_timeout <<= 1;
1416 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) 1422 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
1417 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; 1423 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
1418 } else { 1424 } else {
1419 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); 1425 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
1420 schedule_delayed_work(&transport->connect_worker, 0); 1426 queue_delayed_work(rpciod_workqueue,
1421 1427 &transport->connect_worker, 0);
1422 /* flush_scheduled_work can sleep... */
1423 if (!RPC_IS_ASYNC(task))
1424 flush_scheduled_work();
1425 } 1428 }
1426} 1429}
1427 1430
@@ -1476,7 +1479,7 @@ static struct rpc_xprt_ops xs_udp_ops = {
1476 .set_buffer_size = xs_udp_set_buffer_size, 1479 .set_buffer_size = xs_udp_set_buffer_size,
1477 .reserve_xprt = xprt_reserve_xprt_cong, 1480 .reserve_xprt = xprt_reserve_xprt_cong,
1478 .release_xprt = xprt_release_xprt_cong, 1481 .release_xprt = xprt_release_xprt_cong,
1479 .rpcbind = rpcb_getport, 1482 .rpcbind = rpcb_getport_async,
1480 .set_port = xs_set_port, 1483 .set_port = xs_set_port,
1481 .connect = xs_connect, 1484 .connect = xs_connect,
1482 .buf_alloc = rpc_malloc, 1485 .buf_alloc = rpc_malloc,
@@ -1493,7 +1496,7 @@ static struct rpc_xprt_ops xs_udp_ops = {
1493static struct rpc_xprt_ops xs_tcp_ops = { 1496static struct rpc_xprt_ops xs_tcp_ops = {
1494 .reserve_xprt = xprt_reserve_xprt, 1497 .reserve_xprt = xprt_reserve_xprt,
1495 .release_xprt = xs_tcp_release_xprt, 1498 .release_xprt = xs_tcp_release_xprt,
1496 .rpcbind = rpcb_getport, 1499 .rpcbind = rpcb_getport_async,
1497 .set_port = xs_set_port, 1500 .set_port = xs_set_port,
1498 .connect = xs_connect, 1501 .connect = xs_connect,
1499 .buf_alloc = rpc_malloc, 1502 .buf_alloc = rpc_malloc,
@@ -1505,12 +1508,12 @@ static struct rpc_xprt_ops xs_tcp_ops = {
1505 .print_stats = xs_tcp_print_stats, 1508 .print_stats = xs_tcp_print_stats,
1506}; 1509};
1507 1510
1508static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, unsigned int slot_table_size) 1511static struct rpc_xprt *xs_setup_xprt(struct rpc_xprtsock_create *args, unsigned int slot_table_size)
1509{ 1512{
1510 struct rpc_xprt *xprt; 1513 struct rpc_xprt *xprt;
1511 struct sock_xprt *new; 1514 struct sock_xprt *new;
1512 1515
1513 if (addrlen > sizeof(xprt->addr)) { 1516 if (args->addrlen > sizeof(xprt->addr)) {
1514 dprintk("RPC: xs_setup_xprt: address too large\n"); 1517 dprintk("RPC: xs_setup_xprt: address too large\n");
1515 return ERR_PTR(-EBADF); 1518 return ERR_PTR(-EBADF);
1516 } 1519 }
@@ -1532,8 +1535,10 @@ static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, uns
1532 return ERR_PTR(-ENOMEM); 1535 return ERR_PTR(-ENOMEM);
1533 } 1536 }
1534 1537
1535 memcpy(&xprt->addr, addr, addrlen); 1538 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
1536 xprt->addrlen = addrlen; 1539 xprt->addrlen = args->addrlen;
1540 if (args->srcaddr)
1541 memcpy(&new->addr, args->srcaddr, args->addrlen);
1537 new->port = xs_get_random_port(); 1542 new->port = xs_get_random_port();
1538 1543
1539 return xprt; 1544 return xprt;
@@ -1541,22 +1546,20 @@ static struct rpc_xprt *xs_setup_xprt(struct sockaddr *addr, size_t addrlen, uns
1541 1546
1542/** 1547/**
1543 * xs_setup_udp - Set up transport to use a UDP socket 1548 * xs_setup_udp - Set up transport to use a UDP socket
1544 * @addr: address of remote server 1549 * @args: rpc transport creation arguments
1545 * @addrlen: length of address in bytes
1546 * @to: timeout parameters
1547 * 1550 *
1548 */ 1551 */
1549struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to) 1552struct rpc_xprt *xs_setup_udp(struct rpc_xprtsock_create *args)
1550{ 1553{
1551 struct rpc_xprt *xprt; 1554 struct rpc_xprt *xprt;
1552 struct sock_xprt *transport; 1555 struct sock_xprt *transport;
1553 1556
1554 xprt = xs_setup_xprt(addr, addrlen, xprt_udp_slot_table_entries); 1557 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries);
1555 if (IS_ERR(xprt)) 1558 if (IS_ERR(xprt))
1556 return xprt; 1559 return xprt;
1557 transport = container_of(xprt, struct sock_xprt, xprt); 1560 transport = container_of(xprt, struct sock_xprt, xprt);
1558 1561
1559 if (ntohs(((struct sockaddr_in *)addr)->sin_port) != 0) 1562 if (ntohs(((struct sockaddr_in *)args->dstaddr)->sin_port) != 0)
1560 xprt_set_bound(xprt); 1563 xprt_set_bound(xprt);
1561 1564
1562 xprt->prot = IPPROTO_UDP; 1565 xprt->prot = IPPROTO_UDP;
@@ -1572,8 +1575,8 @@ struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_
1572 1575
1573 xprt->ops = &xs_udp_ops; 1576 xprt->ops = &xs_udp_ops;
1574 1577
1575 if (to) 1578 if (args->timeout)
1576 xprt->timeout = *to; 1579 xprt->timeout = *args->timeout;
1577 else 1580 else
1578 xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); 1581 xprt_set_timeout(&xprt->timeout, 5, 5 * HZ);
1579 1582
@@ -1586,22 +1589,20 @@ struct rpc_xprt *xs_setup_udp(struct sockaddr *addr, size_t addrlen, struct rpc_
1586 1589
1587/** 1590/**
1588 * xs_setup_tcp - Set up transport to use a TCP socket 1591 * xs_setup_tcp - Set up transport to use a TCP socket
1589 * @addr: address of remote server 1592 * @args: rpc transport creation arguments
1590 * @addrlen: length of address in bytes
1591 * @to: timeout parameters
1592 * 1593 *
1593 */ 1594 */
1594struct rpc_xprt *xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_timeout *to) 1595struct rpc_xprt *xs_setup_tcp(struct rpc_xprtsock_create *args)
1595{ 1596{
1596 struct rpc_xprt *xprt; 1597 struct rpc_xprt *xprt;
1597 struct sock_xprt *transport; 1598 struct sock_xprt *transport;
1598 1599
1599 xprt = xs_setup_xprt(addr, addrlen, xprt_tcp_slot_table_entries); 1600 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries);
1600 if (IS_ERR(xprt)) 1601 if (IS_ERR(xprt))
1601 return xprt; 1602 return xprt;
1602 transport = container_of(xprt, struct sock_xprt, xprt); 1603 transport = container_of(xprt, struct sock_xprt, xprt);
1603 1604
1604 if (ntohs(((struct sockaddr_in *)addr)->sin_port) != 0) 1605 if (ntohs(((struct sockaddr_in *)args->dstaddr)->sin_port) != 0)
1605 xprt_set_bound(xprt); 1606 xprt_set_bound(xprt);
1606 1607
1607 xprt->prot = IPPROTO_TCP; 1608 xprt->prot = IPPROTO_TCP;
@@ -1616,8 +1617,8 @@ struct rpc_xprt *xs_setup_tcp(struct sockaddr *addr, size_t addrlen, struct rpc_
1616 1617
1617 xprt->ops = &xs_tcp_ops; 1618 xprt->ops = &xs_tcp_ops;
1618 1619
1619 if (to) 1620 if (args->timeout)
1620 xprt->timeout = *to; 1621 xprt->timeout = *args->timeout;
1621 else 1622 else
1622 xprt_set_timeout(&xprt->timeout, 2, 60 * HZ); 1623 xprt_set_timeout(&xprt->timeout, 2, 60 * HZ);
1623 1624
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-07 19:59:27 -0500