aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-10-26 12:55:25 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-26 12:55:25 -0400
commit4390110fef9e5c64e10c6ca19d586932242c9a8a (patch)
treef2f26fe291c8b4e424b97ac57370b33e0c006568
parenta4dd8dce14014665862ce7911b38cb2c69e366dd (diff)
parent42d7ba3d6d56a6cbc773284896108b1e2ebcee81 (diff)
Merge branch 'for-2.6.37' of git://linux-nfs.org/~bfields/linux
* 'for-2.6.37' of git://linux-nfs.org/~bfields/linux: (99 commits) svcrpc: svc_tcp_sendto XPT_DEAD check is redundant svcrpc: no need for XPT_DEAD check in svc_xprt_enqueue svcrpc: assume svc_delete_xprt() called only once svcrpc: never clear XPT_BUSY on dead xprt nfsd4: fix connection allocation in sequence() nfsd4: only require krb5 principal for NFSv4.0 callbacks nfsd4: move minorversion to client nfsd4: delay session removal till free_client nfsd4: separate callback change and callback probe nfsd4: callback program number is per-session nfsd4: track backchannel connections nfsd4: confirm only on succesful create_session nfsd4: make backchannel sequence number per-session nfsd4: use client pointer to backchannel session nfsd4: move callback setup into session init code nfsd4: don't cache seq_misordered replies SUNRPC: Properly initialize sock_xprt.srcaddr in all cases SUNRPC: Use conventional switch statement when reclassifying sockets sunrpc/xprtrdma: clean up workqueue usage sunrpc: Turn list_for_each-s into the ..._entry-s ... Fix up trivial conflicts (two different deprecation notices added in separate branches) in Documentation/feature-removal-schedule.txt
Diffstat
-rw-r--r--Documentation/feature-removal-schedule.txt10
-rw-r--r--fs/Makefile5
-rw-r--r--fs/compat.c2
-rw-r--r--fs/lockd/host.c1
-rw-r--r--fs/lockd/mon.c1
-rw-r--r--fs/lockd/svc.c2
-rw-r--r--fs/lockd/svc4proc.c2
-rw-r--r--fs/lockd/svclock.c31
-rw-r--r--fs/lockd/svcproc.c2
-rw-r--r--fs/nfs/callback.c4
-rw-r--r--fs/nfs/client.c1
-rw-r--r--fs/nfs/dns_resolve.c6
-rw-r--r--fs/nfs/mount_clnt.c2
-rw-r--r--fs/nfsd/Kconfig12
-rw-r--r--fs/nfsd/export.c73
-rw-r--r--fs/nfsd/nfs4callback.c244
-rw-r--r--fs/nfsd/nfs4idmap.c105
-rw-r--r--fs/nfsd/nfs4proc.c7
-rw-r--r--fs/nfsd/nfs4state.c493
-rw-r--r--fs/nfsd/nfs4xdr.c18
-rw-r--r--fs/nfsd/nfsctl.c26
-rw-r--r--fs/nfsd/nfsd.h2
-rw-r--r--fs/nfsd/nfssvc.c5
-rw-r--r--fs/nfsd/state.h52
-rw-r--r--include/linux/net.h2
-rw-r--r--include/linux/nfs4.h3
-rw-r--r--include/linux/sunrpc/auth.h4
-rw-r--r--include/linux/sunrpc/cache.h37
-rw-r--r--include/linux/sunrpc/clnt.h1
-rw-r--r--include/linux/sunrpc/gss_spkm3.h55
-rw-r--r--include/linux/sunrpc/stats.h23
-rw-r--r--include/linux/sunrpc/svc_xprt.h32
-rw-r--r--include/linux/sunrpc/svcauth.h17
-rw-r--r--include/linux/sunrpc/xprt.h4
-rw-r--r--net/socket.c3
-rw-r--r--net/sunrpc/Kconfig19
-rw-r--r--net/sunrpc/auth.c2
-rw-r--r--net/sunrpc/auth_generic.c2
-rw-r--r--net/sunrpc/auth_gss/Makefile5
-rw-r--r--net/sunrpc/auth_gss/gss_spkm3_mech.c247
-rw-r--r--net/sunrpc/auth_gss/gss_spkm3_seal.c186
-rw-r--r--net/sunrpc/auth_gss/gss_spkm3_token.c267
-rw-r--r--net/sunrpc/auth_gss/gss_spkm3_unseal.c127
-rw-r--r--net/sunrpc/auth_gss/svcauth_gss.c51
-rw-r--r--net/sunrpc/cache.c288
-rw-r--r--net/sunrpc/clnt.c1
-rw-r--r--net/sunrpc/netns.h19
-rw-r--r--net/sunrpc/rpcb_clnt.c4
-rw-r--r--net/sunrpc/stats.c43
-rw-r--r--net/sunrpc/sunrpc_syms.c58
-rw-r--r--net/sunrpc/svc.c3
-rw-r--r--net/sunrpc/svc_xprt.c59
-rw-r--r--net/sunrpc/svcauth_unix.c194
-rw-r--r--net/sunrpc/svcsock.c27
-rw-r--r--net/sunrpc/xprt.c39
-rw-r--r--net/sunrpc/xprtrdma/svc_rdma.c11
-rw-r--r--net/sunrpc/xprtrdma/svc_rdma_recvfrom.c19
-rw-r--r--net/sunrpc/xprtrdma/svc_rdma_sendto.c82
-rw-r--r--net/sunrpc/xprtrdma/svc_rdma_transport.c49
-rw-r--r--net/sunrpc/xprtrdma/transport.c25
-rw-r--r--net/sunrpc/xprtsock.c358
61 files changed, 1535 insertions, 1937 deletions
diff --git a/Documentation/feature-removal-schedule.txt b/Documentation/feature-removal-schedule.txt
index e833c8c81e69..d2af87ba96e1 100644
--- a/Documentation/feature-removal-schedule.txt
+++ b/Documentation/feature-removal-schedule.txt
@@ -535,3 +535,13 @@ Why: Hareware scan is the prefer method for iwlwifi devices for
535Who: Wey-Yi Guy <wey-yi.w.guy@intel.com> 535Who: Wey-Yi Guy <wey-yi.w.guy@intel.com>
536 536
537---------------------------- 537----------------------------
538
539What: access to nfsd auth cache through sys_nfsservctl or '.' files
540 in the 'nfsd' filesystem.
541When: 2.6.40
542Why: This is a legacy interface which have been replaced by a more
543 dynamic cache. Continuing to maintain this interface is an
544 unnecessary burden.
545Who: NeilBrown <neilb@suse.de>
546
547----------------------------
diff --git a/fs/Makefile b/fs/Makefile
index e6ec1d309b1d..26956fcec917 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -29,10 +29,7 @@ obj-$(CONFIG_EVENTFD) += eventfd.o
29obj-$(CONFIG_AIO) += aio.o 29obj-$(CONFIG_AIO) += aio.o
30obj-$(CONFIG_FILE_LOCKING) += locks.o 30obj-$(CONFIG_FILE_LOCKING) += locks.o
31obj-$(CONFIG_COMPAT) += compat.o compat_ioctl.o 31obj-$(CONFIG_COMPAT) += compat.o compat_ioctl.o
32 32obj-$(CONFIG_NFSD_DEPRECATED) += nfsctl.o
33nfsd-$(CONFIG_NFSD) := nfsctl.o
34obj-y += $(nfsd-y) $(nfsd-m)
35
36obj-$(CONFIG_BINFMT_AOUT) += binfmt_aout.o 33obj-$(CONFIG_BINFMT_AOUT) += binfmt_aout.o
37obj-$(CONFIG_BINFMT_EM86) += binfmt_em86.o 34obj-$(CONFIG_BINFMT_EM86) += binfmt_em86.o
38obj-$(CONFIG_BINFMT_MISC) += binfmt_misc.o 35obj-$(CONFIG_BINFMT_MISC) += binfmt_misc.o
diff --git a/fs/compat.c b/fs/compat.c
index 0644a154672b..f03abdadc401 100644
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -1963,7 +1963,7 @@ asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1963} 1963}
1964#endif /* HAVE_SET_RESTORE_SIGMASK */ 1964#endif /* HAVE_SET_RESTORE_SIGMASK */
1965 1965
1966#if defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE) 1966#if (defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE)) && !defined(CONFIG_NFSD_DEPRECATED)
1967/* Stuff for NFS server syscalls... */ 1967/* Stuff for NFS server syscalls... */
1968struct compat_nfsctl_svc { 1968struct compat_nfsctl_svc {
1969 u16 svc32_port; 1969 u16 svc32_port;
diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index bb464d12104c..25e21e4023b2 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -353,6 +353,7 @@ nlm_bind_host(struct nlm_host *host)
353 .to_retries = 5U, 353 .to_retries = 5U,
354 }; 354 };
355 struct rpc_create_args args = { 355 struct rpc_create_args args = {
356 .net = &init_net,
356 .protocol = host->h_proto, 357 .protocol = host->h_proto,
357 .address = nlm_addr(host), 358 .address = nlm_addr(host),
358 .addrsize = host->h_addrlen, 359 .addrsize = host->h_addrlen,
diff --git a/fs/lockd/mon.c b/fs/lockd/mon.c
index e3015464fbab..e0c918949644 100644
--- a/fs/lockd/mon.c
+++ b/fs/lockd/mon.c
@@ -69,6 +69,7 @@ static struct rpc_clnt *nsm_create(void)
69 .sin_addr.s_addr = htonl(INADDR_LOOPBACK), 69 .sin_addr.s_addr = htonl(INADDR_LOOPBACK),
70 }; 70 };
71 struct rpc_create_args args = { 71 struct rpc_create_args args = {
72 .net = &init_net,
72 .protocol = XPRT_TRANSPORT_UDP, 73 .protocol = XPRT_TRANSPORT_UDP,
73 .address = (struct sockaddr *)&sin, 74 .address = (struct sockaddr *)&sin,
74 .addrsize = sizeof(sin), 75 .addrsize = sizeof(sin),
diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index f1bacf1a0391..b13aabc12298 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -206,7 +206,7 @@ static int create_lockd_listener(struct svc_serv *serv, const char *name,
206 206
207 xprt = svc_find_xprt(serv, name, family, 0); 207 xprt = svc_find_xprt(serv, name, family, 0);
208 if (xprt == NULL) 208 if (xprt == NULL)
209 return svc_create_xprt(serv, name, family, port, 209 return svc_create_xprt(serv, name, &init_net, family, port,
210 SVC_SOCK_DEFAULTS); 210 SVC_SOCK_DEFAULTS);
211 svc_xprt_put(xprt); 211 svc_xprt_put(xprt);
212 return 0; 212 return 0;
diff --git a/fs/lockd/svc4proc.c b/fs/lockd/svc4proc.c
index 031c6569a134..a336e832475d 100644
--- a/fs/lockd/svc4proc.c
+++ b/fs/lockd/svc4proc.c
@@ -230,9 +230,7 @@ static void nlm4svc_callback_exit(struct rpc_task *task, void *data)
230 230
231static void nlm4svc_callback_release(void *data) 231static void nlm4svc_callback_release(void *data)
232{ 232{
233 lock_kernel();
234 nlm_release_call(data); 233 nlm_release_call(data);
235 unlock_kernel();
236} 234}
237 235
238static const struct rpc_call_ops nlm4svc_callback_ops = { 236static const struct rpc_call_ops nlm4svc_callback_ops = {
diff --git a/fs/lockd/svclock.c b/fs/lockd/svclock.c
index 84055d31bfc5..6f1ef000975a 100644
--- a/fs/lockd/svclock.c
+++ b/fs/lockd/svclock.c
@@ -52,12 +52,13 @@ static const struct rpc_call_ops nlmsvc_grant_ops;
52 * The list of blocked locks to retry 52 * The list of blocked locks to retry
53 */ 53 */
54static LIST_HEAD(nlm_blocked); 54static LIST_HEAD(nlm_blocked);
55static DEFINE_SPINLOCK(nlm_blocked_lock);
55 56
56/* 57/*
57 * Insert a blocked lock into the global list 58 * Insert a blocked lock into the global list
58 */ 59 */
59static void 60static void
60nlmsvc_insert_block(struct nlm_block *block, unsigned long when) 61nlmsvc_insert_block_locked(struct nlm_block *block, unsigned long when)
61{ 62{
62 struct nlm_block *b; 63 struct nlm_block *b;
63 struct list_head *pos; 64 struct list_head *pos;
@@ -87,6 +88,13 @@ nlmsvc_insert_block(struct nlm_block *block, unsigned long when)
87 block->b_when = when; 88 block->b_when = when;
88} 89}
89 90
91static void nlmsvc_insert_block(struct nlm_block *block, unsigned long when)
92{
93 spin_lock(&nlm_blocked_lock);
94 nlmsvc_insert_block_locked(block, when);
95 spin_unlock(&nlm_blocked_lock);
96}
97
90/* 98/*
91 * Remove a block from the global list 99 * Remove a block from the global list
92 */ 100 */
@@ -94,7 +102,9 @@ static inline void
94nlmsvc_remove_block(struct nlm_block *block) 102nlmsvc_remove_block(struct nlm_block *block)
95{ 103{
96 if (!list_empty(&block->b_list)) { 104 if (!list_empty(&block->b_list)) {
105 spin_lock(&nlm_blocked_lock);
97 list_del_init(&block->b_list); 106 list_del_init(&block->b_list);
107 spin_unlock(&nlm_blocked_lock);
98 nlmsvc_release_block(block); 108 nlmsvc_release_block(block);
99 } 109 }
100} 110}
@@ -651,7 +661,7 @@ static int nlmsvc_grant_deferred(struct file_lock *fl, struct file_lock *conf,
651 struct nlm_block *block; 661 struct nlm_block *block;
652 int rc = -ENOENT; 662 int rc = -ENOENT;
653 663
654 lock_kernel(); 664 spin_lock(&nlm_blocked_lock);
655 list_for_each_entry(block, &nlm_blocked, b_list) { 665 list_for_each_entry(block, &nlm_blocked, b_list) {
656 if (nlm_compare_locks(&block->b_call->a_args.lock.fl, fl)) { 666 if (nlm_compare_locks(&block->b_call->a_args.lock.fl, fl)) {
657 dprintk("lockd: nlmsvc_notify_blocked block %p flags %d\n", 667 dprintk("lockd: nlmsvc_notify_blocked block %p flags %d\n",
@@ -665,13 +675,13 @@ static int nlmsvc_grant_deferred(struct file_lock *fl, struct file_lock *conf,
665 } else if (result == 0) 675 } else if (result == 0)
666 block->b_granted = 1; 676 block->b_granted = 1;
667 677
668 nlmsvc_insert_block(block, 0); 678 nlmsvc_insert_block_locked(block, 0);
669 svc_wake_up(block->b_daemon); 679 svc_wake_up(block->b_daemon);
670 rc = 0; 680 rc = 0;
671 break; 681 break;
672 } 682 }
673 } 683 }
674 unlock_kernel(); 684 spin_unlock(&nlm_blocked_lock);
675 if (rc == -ENOENT) 685 if (rc == -ENOENT)
676 printk(KERN_WARNING "lockd: grant for unknown block\n"); 686 printk(KERN_WARNING "lockd: grant for unknown block\n");
677 return rc; 687 return rc;
@@ -803,7 +813,7 @@ static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
803 813
804 dprintk("lockd: GRANT_MSG RPC callback\n"); 814 dprintk("lockd: GRANT_MSG RPC callback\n");
805 815
806 lock_kernel(); 816 spin_lock(&nlm_blocked_lock);
807 /* if the block is not on a list at this point then it has 817 /* if the block is not on a list at this point then it has
808 * been invalidated. Don't try to requeue it. 818 * been invalidated. Don't try to requeue it.
809 * 819 *
@@ -825,19 +835,20 @@ static void nlmsvc_grant_callback(struct rpc_task *task, void *data)
825 /* Call was successful, now wait for client callback */ 835 /* Call was successful, now wait for client callback */
826 timeout = 60 * HZ; 836 timeout = 60 * HZ;
827 } 837 }
828 nlmsvc_insert_block(block, timeout); 838 nlmsvc_insert_block_locked(block, timeout);
829 svc_wake_up(block->b_daemon); 839 svc_wake_up(block->b_daemon);
830out: 840out:
831 unlock_kernel(); 841 spin_unlock(&nlm_blocked_lock);
832} 842}
833 843
844/*
845 * FIXME: nlmsvc_release_block() grabs a mutex. This is not allowed for an
846 * .rpc_release rpc_call_op
847 */
834static void nlmsvc_grant_release(void *data) 848static void nlmsvc_grant_release(void *data)
835{ 849{
836 struct nlm_rqst *call = data; 850 struct nlm_rqst *call = data;
837
838 lock_kernel();
839 nlmsvc_release_block(call->a_block); 851 nlmsvc_release_block(call->a_block);
840 unlock_kernel();
841} 852}
842 853
843static const struct rpc_call_ops nlmsvc_grant_ops = { 854static const struct rpc_call_ops nlmsvc_grant_ops = {
diff --git a/fs/lockd/svcproc.c b/fs/lockd/svcproc.c
index 0f2ab741ae7c..c3069f38d602 100644
--- a/fs/lockd/svcproc.c
+++ b/fs/lockd/svcproc.c
@@ -260,9 +260,7 @@ static void nlmsvc_callback_exit(struct rpc_task *task, void *data)
260 260
261static void nlmsvc_callback_release(void *data) 261static void nlmsvc_callback_release(void *data)
262{ 262{
263 lock_kernel();
264 nlm_release_call(data); 263 nlm_release_call(data);
265 unlock_kernel();
266} 264}
267 265
268static const struct rpc_call_ops nlmsvc_callback_ops = { 266static const struct rpc_call_ops nlmsvc_callback_ops = {
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index e17b49e2eabd..aeec017fe814 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -109,7 +109,7 @@ nfs4_callback_up(struct svc_serv *serv)
109{ 109{
110 int ret; 110 int ret;
111 111
112 ret = svc_create_xprt(serv, "tcp", PF_INET, 112 ret = svc_create_xprt(serv, "tcp", &init_net, PF_INET,
113 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS); 113 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
114 if (ret <= 0) 114 if (ret <= 0)
115 goto out_err; 115 goto out_err;
@@ -117,7 +117,7 @@ nfs4_callback_up(struct svc_serv *serv)
117 dprintk("NFS: Callback listener port = %u (af %u)\n", 117 dprintk("NFS: Callback listener port = %u (af %u)\n",
118 nfs_callback_tcpport, PF_INET); 118 nfs_callback_tcpport, PF_INET);
119 119
120 ret = svc_create_xprt(serv, "tcp", PF_INET6, 120 ret = svc_create_xprt(serv, "tcp", &init_net, PF_INET6,
121 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS); 121 nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
122 if (ret > 0) { 122 if (ret > 0) {
123 nfs_callback_tcpport6 = ret; 123 nfs_callback_tcpport6 = ret;
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index fd6f0a70021b..0870d0d4efc0 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -605,6 +605,7 @@ static int nfs_create_rpc_client(struct nfs_client *clp,
605{ 605{
606 struct rpc_clnt *clnt = NULL; 606 struct rpc_clnt *clnt = NULL;
607 struct rpc_create_args args = { 607 struct rpc_create_args args = {
608 .net = &init_net,
608 .protocol = clp->cl_proto, 609 .protocol = clp->cl_proto,
609 .address = (struct sockaddr *)&clp->cl_addr, 610 .address = (struct sockaddr *)&clp->cl_addr,
610 .addrsize = clp->cl_addrlen, 611 .addrsize = clp->cl_addrlen,
diff --git a/fs/nfs/dns_resolve.c b/fs/nfs/dns_resolve.c
index dba50a5625db..a6e711ad130f 100644
--- a/fs/nfs/dns_resolve.c
+++ b/fs/nfs/dns_resolve.c
@@ -167,7 +167,7 @@ static int nfs_dns_show(struct seq_file *m, struct cache_detail *cd,
167 return 0; 167 return 0;
168 } 168 }
169 item = container_of(h, struct nfs_dns_ent, h); 169 item = container_of(h, struct nfs_dns_ent, h);
170 ttl = (long)item->h.expiry_time - (long)get_seconds(); 170 ttl = item->h.expiry_time - seconds_since_boot();
171 if (ttl < 0) 171 if (ttl < 0)
172 ttl = 0; 172 ttl = 0;
173 173
@@ -239,7 +239,7 @@ static int nfs_dns_parse(struct cache_detail *cd, char *buf, int buflen)
239 ttl = get_expiry(&buf); 239 ttl = get_expiry(&buf);
240 if (ttl == 0) 240 if (ttl == 0)
241 goto out; 241 goto out;
242 key.h.expiry_time = ttl + get_seconds(); 242 key.h.expiry_time = ttl + seconds_since_boot();
243 243
244 ret = -ENOMEM; 244 ret = -ENOMEM;
245 item = nfs_dns_lookup(cd, &key); 245 item = nfs_dns_lookup(cd, &key);
@@ -301,7 +301,7 @@ static int do_cache_lookup_nowait(struct cache_detail *cd,
301 goto out_err; 301 goto out_err;
302 ret = -ETIMEDOUT; 302 ret = -ETIMEDOUT;
303 if (!test_bit(CACHE_VALID, &(*item)->h.flags) 303 if (!test_bit(CACHE_VALID, &(*item)->h.flags)
304 || (*item)->h.expiry_time < get_seconds() 304 || (*item)->h.expiry_time < seconds_since_boot()
305 || cd->flush_time > (*item)->h.last_refresh) 305 || cd->flush_time > (*item)->h.last_refresh)
306 goto out_put; 306 goto out_put;
307 ret = -ENOENT; 307 ret = -ENOENT;
diff --git a/fs/nfs/mount_clnt.c b/fs/nfs/mount_clnt.c
index d610203d95c6..eceafe74f473 100644
--- a/fs/nfs/mount_clnt.c
+++ b/fs/nfs/mount_clnt.c
@@ -153,6 +153,7 @@ int nfs_mount(struct nfs_mount_request *info)
153 .rpc_resp = &result, 153 .rpc_resp = &result,
154 }; 154 };
155 struct rpc_create_args args = { 155 struct rpc_create_args args = {
156 .net = &init_net,
156 .protocol = info->protocol, 157 .protocol = info->protocol,
157 .address = info->sap, 158 .address = info->sap,
158 .addrsize = info->salen, 159 .addrsize = info->salen,
@@ -224,6 +225,7 @@ void nfs_umount(const struct nfs_mount_request *info)
224 .to_retries = 2, 225 .to_retries = 2,
225 }; 226 };
226 struct rpc_create_args args = { 227 struct rpc_create_args args = {
228 .net = &init_net,
227 .protocol = IPPROTO_UDP, 229 .protocol = IPPROTO_UDP,
228 .address = info->sap, 230 .address = info->sap,
229 .addrsize = info->salen, 231 .addrsize = info->salen,
diff --git a/fs/nfsd/Kconfig b/fs/nfsd/Kconfig
index 7cf4ddafb4ab..31a78fce4732 100644
--- a/fs/nfsd/Kconfig
+++ b/fs/nfsd/Kconfig
@@ -29,6 +29,18 @@ config NFSD
29 29
30 If unsure, say N. 30 If unsure, say N.
31 31
32config NFSD_DEPRECATED
33 bool "Include support for deprecated syscall interface to NFSD"
34 depends on NFSD
35 default y
36 help
37 The syscall interface to nfsd was obsoleted in 2.6.0 by a new
38 filesystem based interface. The old interface is due for removal
39 in 2.6.40. If you wish to remove the interface before then
40 say N.
41
42 In unsure, say Y.
43
32config NFSD_V2_ACL 44config NFSD_V2_ACL
33 bool 45 bool
34 depends on NFSD 46 depends on NFSD
diff --git a/fs/nfsd/export.c b/fs/nfsd/export.c
index c2a4f71d87dd..c0fcb7ab7f6d 100644
--- a/fs/nfsd/export.c
+++ b/fs/nfsd/export.c
@@ -28,9 +28,6 @@
28typedef struct auth_domain svc_client; 28typedef struct auth_domain svc_client;
29typedef struct svc_export svc_export; 29typedef struct svc_export svc_export;
30 30
31static void exp_do_unexport(svc_export *unexp);
32static int exp_verify_string(char *cp, int max);
33
34/* 31/*
35 * We have two caches. 32 * We have two caches.
36 * One maps client+vfsmnt+dentry to export options - the export map 33 * One maps client+vfsmnt+dentry to export options - the export map
@@ -802,6 +799,7 @@ exp_find_key(svc_client *clp, int fsid_type, u32 *fsidv, struct cache_req *reqp)
802 return ek; 799 return ek;
803} 800}
804 801
802#ifdef CONFIG_NFSD_DEPRECATED
805static int exp_set_key(svc_client *clp, int fsid_type, u32 *fsidv, 803static int exp_set_key(svc_client *clp, int fsid_type, u32 *fsidv,
806 struct svc_export *exp) 804 struct svc_export *exp)
807{ 805{
@@ -852,6 +850,7 @@ exp_get_fsid_key(svc_client *clp, int fsid)
852 850
853 return exp_find_key(clp, FSID_NUM, fsidv, NULL); 851 return exp_find_key(clp, FSID_NUM, fsidv, NULL);
854} 852}
853#endif
855 854
856static svc_export *exp_get_by_name(svc_client *clp, const struct path *path, 855static svc_export *exp_get_by_name(svc_client *clp, const struct path *path,
857 struct cache_req *reqp) 856 struct cache_req *reqp)
@@ -893,6 +892,7 @@ static struct svc_export *exp_parent(svc_client *clp, struct path *path)
893 return exp; 892 return exp;
894} 893}
895 894
895#ifdef CONFIG_NFSD_DEPRECATED
896/* 896/*
897 * Hashtable locking. Write locks are placed only by user processes 897 * Hashtable locking. Write locks are placed only by user processes
898 * wanting to modify export information. 898 * wanting to modify export information.
@@ -925,6 +925,19 @@ exp_writeunlock(void)
925{ 925{
926 up_write(&hash_sem); 926 up_write(&hash_sem);
927} 927}
928#else
929
930/* hash_sem not needed once deprecated interface is removed */
931void exp_readlock(void) {}
932static inline void exp_writelock(void){}
933void exp_readunlock(void) {}
934static inline void exp_writeunlock(void){}
935
936#endif
937
938#ifdef CONFIG_NFSD_DEPRECATED
939static void exp_do_unexport(svc_export *unexp);
940static int exp_verify_string(char *cp, int max);
928 941
929static void exp_fsid_unhash(struct svc_export *exp) 942static void exp_fsid_unhash(struct svc_export *exp)
930{ 943{
@@ -935,10 +948,9 @@ static void exp_fsid_unhash(struct svc_export *exp)
935 948
936 ek = exp_get_fsid_key(exp->ex_client, exp->ex_fsid); 949 ek = exp_get_fsid_key(exp->ex_client, exp->ex_fsid);
937 if (!IS_ERR(ek)) { 950 if (!IS_ERR(ek)) {
938 ek->h.expiry_time = get_seconds()-1; 951 sunrpc_invalidate(&ek->h, &svc_expkey_cache);
939 cache_put(&ek->h, &svc_expkey_cache); 952 cache_put(&ek->h, &svc_expkey_cache);
940 } 953 }
941 svc_expkey_cache.nextcheck = get_seconds();
942} 954}
943 955
944static int exp_fsid_hash(svc_client *clp, struct svc_export *exp) 956static int exp_fsid_hash(svc_client *clp, struct svc_export *exp)
@@ -973,10 +985,9 @@ static void exp_unhash(struct svc_export *exp)
973 985
974 ek = exp_get_key(exp->ex_client, inode->i_sb->s_dev, inode->i_ino); 986 ek = exp_get_key(exp->ex_client, inode->i_sb->s_dev, inode->i_ino);
975 if (!IS_ERR(ek)) { 987 if (!IS_ERR(ek)) {
976 ek->h.expiry_time = get_seconds()-1; 988 sunrpc_invalidate(&ek->h, &svc_expkey_cache);
977 cache_put(&ek->h, &svc_expkey_cache); 989 cache_put(&ek->h, &svc_expkey_cache);
978 } 990 }
979 svc_expkey_cache.nextcheck = get_seconds();
980} 991}
981 992
982/* 993/*
@@ -1097,8 +1108,7 @@ out:
1097static void 1108static void
1098exp_do_unexport(svc_export *unexp) 1109exp_do_unexport(svc_export *unexp)
1099{ 1110{
1100 unexp->h.expiry_time = get_seconds()-1; 1111 sunrpc_invalidate(&unexp->h, &svc_export_cache);
1101 svc_export_cache.nextcheck = get_seconds();
1102 exp_unhash(unexp); 1112 exp_unhash(unexp);
1103 exp_fsid_unhash(unexp); 1113 exp_fsid_unhash(unexp);
1104} 1114}
@@ -1150,6 +1160,7 @@ out_unlock:
1150 exp_writeunlock(); 1160 exp_writeunlock();
1151 return err; 1161 return err;
1152} 1162}
1163#endif /* CONFIG_NFSD_DEPRECATED */
1153 1164
1154/* 1165/*
1155 * Obtain the root fh on behalf of a client. 1166 * Obtain the root fh on behalf of a client.
@@ -1459,25 +1470,43 @@ static void show_secinfo_flags(struct seq_file *m, int flags)
1459 show_expflags(m, flags, NFSEXP_SECINFO_FLAGS); 1470 show_expflags(m, flags, NFSEXP_SECINFO_FLAGS);
1460} 1471}
1461 1472
1473static bool secinfo_flags_equal(int f, int g)
1474{
1475 f &= NFSEXP_SECINFO_FLAGS;
1476 g &= NFSEXP_SECINFO_FLAGS;
1477 return f == g;
1478}
1479
1480static int show_secinfo_run(struct seq_file *m, struct exp_flavor_info **fp, struct exp_flavor_info *end)
1481{
1482 int flags;
1483
1484 flags = (*fp)->flags;
1485 seq_printf(m, ",sec=%d", (*fp)->pseudoflavor);
1486 (*fp)++;
1487 while (*fp != end && secinfo_flags_equal(flags, (*fp)->flags)) {
1488 seq_printf(m, ":%d", (*fp)->pseudoflavor);
1489 (*fp)++;
1490 }
1491 return flags;
1492}
1493
1462static void show_secinfo(struct seq_file *m, struct svc_export *exp) 1494static void show_secinfo(struct seq_file *m, struct svc_export *exp)
1463{ 1495{
1464 struct exp_flavor_info *f; 1496 struct exp_flavor_info *f;
1465 struct exp_flavor_info *end = exp->ex_flavors + exp->ex_nflavors; 1497 struct exp_flavor_info *end = exp->ex_flavors + exp->ex_nflavors;
1466 int lastflags = 0, first = 0; 1498 int flags;
1467 1499
1468 if (exp->ex_nflavors == 0) 1500 if (exp->ex_nflavors == 0)
1469 return; 1501 return;
1470 for (f = exp->ex_flavors; f < end; f++) { 1502 f = exp->ex_flavors;
1471 if (first || f->flags != lastflags) { 1503 flags = show_secinfo_run(m, &f, end);
1472 if (!first) 1504 if (!secinfo_flags_equal(flags, exp->ex_flags))
1473 show_secinfo_flags(m, lastflags); 1505 show_secinfo_flags(m, flags);
1474 seq_printf(m, ",sec=%d", f->pseudoflavor); 1506 while (f != end) {
1475 lastflags = f->flags; 1507 flags = show_secinfo_run(m, &f, end);
1476 } else { 1508 show_secinfo_flags(m, flags);
1477 seq_printf(m, ":%d", f->pseudoflavor);
1478 }
1479 } 1509 }
1480 show_secinfo_flags(m, lastflags);
1481} 1510}
1482 1511
1483static void exp_flags(struct seq_file *m, int flag, int fsid, 1512static void exp_flags(struct seq_file *m, int flag, int fsid,
@@ -1532,6 +1561,7 @@ const struct seq_operations nfs_exports_op = {
1532 .show = e_show, 1561 .show = e_show,
1533}; 1562};
1534 1563
1564#ifdef CONFIG_NFSD_DEPRECATED
1535/* 1565/*
1536 * Add or modify a client. 1566 * Add or modify a client.
1537 * Change requests may involve the list of host addresses. The list of 1567 * Change requests may involve the list of host addresses. The list of
@@ -1563,7 +1593,7 @@ exp_addclient(struct nfsctl_client *ncp)
1563 /* Insert client into hashtable. */ 1593 /* Insert client into hashtable. */
1564 for (i = 0; i < ncp->cl_naddr; i++) { 1594 for (i = 0; i < ncp->cl_naddr; i++) {
1565 ipv6_addr_set_v4mapped(ncp->cl_addrlist[i].s_addr, &addr6); 1595 ipv6_addr_set_v4mapped(ncp->cl_addrlist[i].s_addr, &addr6);
1566 auth_unix_add_addr(&addr6, dom); 1596 auth_unix_add_addr(&init_net, &addr6, dom);
1567 } 1597 }
1568 auth_unix_forget_old(dom); 1598 auth_unix_forget_old(dom);
1569 auth_domain_put(dom); 1599 auth_domain_put(dom);
@@ -1621,6 +1651,7 @@ exp_verify_string(char *cp, int max)
1621 printk(KERN_NOTICE "nfsd: couldn't validate string %s\n", cp); 1651 printk(KERN_NOTICE "nfsd: couldn't validate string %s\n", cp);
1622 return 0; 1652 return 0;
1623} 1653}
1654#endif /* CONFIG_NFSD_DEPRECATED */
1624 1655
1625/* 1656/*
1626 * Initialize the exports module. 1657 * Initialize the exports module.
diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
index 014482c4e57d..143da2eecd7b 100644
--- a/fs/nfsd/nfs4callback.c
+++ b/fs/nfsd/nfs4callback.c
@@ -247,10 +247,11 @@ encode_cb_recall(struct xdr_stream *xdr, struct nfs4_delegation *dp,
247} 247}
248 248
249static void 249static void
250encode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_cb_sequence *args, 250encode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_callback *cb,
251 struct nfs4_cb_compound_hdr *hdr) 251 struct nfs4_cb_compound_hdr *hdr)
252{ 252{
253 __be32 *p; 253 __be32 *p;
254 struct nfsd4_session *ses = cb->cb_clp->cl_cb_session;
254 255
255 if (hdr->minorversion == 0) 256 if (hdr->minorversion == 0)
256 return; 257 return;
@@ -258,8 +259,8 @@ encode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_cb_sequence *args,
258 RESERVE_SPACE(1 + NFS4_MAX_SESSIONID_LEN + 20); 259 RESERVE_SPACE(1 + NFS4_MAX_SESSIONID_LEN + 20);
259 260
260 WRITE32(OP_CB_SEQUENCE); 261 WRITE32(OP_CB_SEQUENCE);
261 WRITEMEM(args->cbs_clp->cl_sessionid.data, NFS4_MAX_SESSIONID_LEN); 262 WRITEMEM(ses->se_sessionid.data, NFS4_MAX_SESSIONID_LEN);
262 WRITE32(args->cbs_clp->cl_cb_seq_nr); 263 WRITE32(ses->se_cb_seq_nr);
263 WRITE32(0); /* slotid, always 0 */ 264 WRITE32(0); /* slotid, always 0 */
264 WRITE32(0); /* highest slotid always 0 */ 265 WRITE32(0); /* highest slotid always 0 */
265 WRITE32(0); /* cachethis always 0 */ 266 WRITE32(0); /* cachethis always 0 */
@@ -279,18 +280,18 @@ nfs4_xdr_enc_cb_null(struct rpc_rqst *req, __be32 *p)
279 280
280static int 281static int
281nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, __be32 *p, 282nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, __be32 *p,
282 struct nfs4_rpc_args *rpc_args) 283 struct nfsd4_callback *cb)
283{ 284{
284 struct xdr_stream xdr; 285 struct xdr_stream xdr;
285 struct nfs4_delegation *args = rpc_args->args_op; 286 struct nfs4_delegation *args = cb->cb_op;
286 struct nfs4_cb_compound_hdr hdr = { 287 struct nfs4_cb_compound_hdr hdr = {
287 .ident = args->dl_ident, 288 .ident = cb->cb_clp->cl_cb_ident,
288 .minorversion = rpc_args->args_seq.cbs_minorversion, 289 .minorversion = cb->cb_minorversion,
289 }; 290 };
290 291
291 xdr_init_encode(&xdr, &req->rq_snd_buf, p); 292 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
292 encode_cb_compound_hdr(&xdr, &hdr); 293 encode_cb_compound_hdr(&xdr, &hdr);
293 encode_cb_sequence(&xdr, &rpc_args->args_seq, &hdr); 294 encode_cb_sequence(&xdr, cb, &hdr);
294 encode_cb_recall(&xdr, args, &hdr); 295 encode_cb_recall(&xdr, args, &hdr);
295 encode_cb_nops(&hdr); 296 encode_cb_nops(&hdr);
296 return 0; 297 return 0;
@@ -338,15 +339,16 @@ decode_cb_op_hdr(struct xdr_stream *xdr, enum nfs_opnum4 expected)
338 * with a single slot. 339 * with a single slot.
339 */ 340 */
340static int 341static int
341decode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_cb_sequence *res, 342decode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_callback *cb,
342 struct rpc_rqst *rqstp) 343 struct rpc_rqst *rqstp)
343{ 344{
345 struct nfsd4_session *ses = cb->cb_clp->cl_cb_session;
344 struct nfs4_sessionid id; 346 struct nfs4_sessionid id;
345 int status; 347 int status;
346 u32 dummy; 348 u32 dummy;
347 __be32 *p; 349 __be32 *p;
348 350
349 if (res->cbs_minorversion == 0) 351 if (cb->cb_minorversion == 0)
350 return 0; 352 return 0;
351 353
352 status = decode_cb_op_hdr(xdr, OP_CB_SEQUENCE); 354 status = decode_cb_op_hdr(xdr, OP_CB_SEQUENCE);
@@ -362,13 +364,12 @@ decode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_cb_sequence *res,
362 READ_BUF(NFS4_MAX_SESSIONID_LEN + 16); 364 READ_BUF(NFS4_MAX_SESSIONID_LEN + 16);
363 memcpy(id.data, p, NFS4_MAX_SESSIONID_LEN); 365 memcpy(id.data, p, NFS4_MAX_SESSIONID_LEN);
364 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN); 366 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
365 if (memcmp(id.data, res->cbs_clp->cl_sessionid.data, 367 if (memcmp(id.data, ses->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
366 NFS4_MAX_SESSIONID_LEN)) {
367 dprintk("%s Invalid session id\n", __func__); 368 dprintk("%s Invalid session id\n", __func__);
368 goto out; 369 goto out;
369 } 370 }
370 READ32(dummy); 371 READ32(dummy);
371 if (dummy != res->cbs_clp->cl_cb_seq_nr) { 372 if (dummy != ses->se_cb_seq_nr) {
372 dprintk("%s Invalid sequence number\n", __func__); 373 dprintk("%s Invalid sequence number\n", __func__);
373 goto out; 374 goto out;
374 } 375 }
@@ -392,7 +393,7 @@ nfs4_xdr_dec_cb_null(struct rpc_rqst *req, __be32 *p)
392 393
393static int 394static int
394nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, __be32 *p, 395nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, __be32 *p,
395 struct nfsd4_cb_sequence *seq) 396 struct nfsd4_callback *cb)
396{ 397{
397 struct xdr_stream xdr; 398 struct xdr_stream xdr;
398 struct nfs4_cb_compound_hdr hdr; 399 struct nfs4_cb_compound_hdr hdr;
@@ -402,8 +403,8 @@ nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, __be32 *p,
402 status = decode_cb_compound_hdr(&xdr, &hdr); 403 status = decode_cb_compound_hdr(&xdr, &hdr);
403 if (status) 404 if (status)
404 goto out; 405 goto out;
405 if (seq) { 406 if (cb) {
406 status = decode_cb_sequence(&xdr, seq, rqstp); 407 status = decode_cb_sequence(&xdr, cb, rqstp);
407 if (status) 408 if (status)
408 goto out; 409 goto out;
409 } 410 }
@@ -472,30 +473,34 @@ static int max_cb_time(void)
472/* Reference counting, callback cleanup, etc., all look racy as heck. 473/* Reference counting, callback cleanup, etc., all look racy as heck.
473 * And why is cl_cb_set an atomic? */ 474 * And why is cl_cb_set an atomic? */
474 475
475int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *cb) 476int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
476{ 477{
477 struct rpc_timeout timeparms = { 478 struct rpc_timeout timeparms = {
478 .to_initval = max_cb_time(), 479 .to_initval = max_cb_time(),
479 .to_retries = 0, 480 .to_retries = 0,
480 }; 481 };
481 struct rpc_create_args args = { 482 struct rpc_create_args args = {
482 .protocol = XPRT_TRANSPORT_TCP, 483 .net = &init_net,
483 .address = (struct sockaddr *) &cb->cb_addr, 484 .address = (struct sockaddr *) &conn->cb_addr,
484 .addrsize = cb->cb_addrlen, 485 .addrsize = conn->cb_addrlen,
485 .timeout = &timeparms, 486 .timeout = &timeparms,
486 .program = &cb_program, 487 .program = &cb_program,
487 .prognumber = cb->cb_prog,
488 .version = 0, 488 .version = 0,
489 .authflavor = clp->cl_flavor, 489 .authflavor = clp->cl_flavor,
490 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET), 490 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
491 .client_name = clp->cl_principal,
492 }; 491 };
493 struct rpc_clnt *client; 492 struct rpc_clnt *client;
494 493
495 if (!clp->cl_principal && (clp->cl_flavor >= RPC_AUTH_GSS_KRB5)) 494 if (clp->cl_minorversion == 0) {
496 return -EINVAL; 495 if (!clp->cl_principal && (clp->cl_flavor >= RPC_AUTH_GSS_KRB5))
497 if (cb->cb_minorversion) { 496 return -EINVAL;
498 args.bc_xprt = cb->cb_xprt; 497 args.client_name = clp->cl_principal;
498 args.prognumber = conn->cb_prog,
499 args.protocol = XPRT_TRANSPORT_TCP;
500 clp->cl_cb_ident = conn->cb_ident;
501 } else {
502 args.bc_xprt = conn->cb_xprt;
503 args.prognumber = clp->cl_cb_session->se_cb_prog;
499 args.protocol = XPRT_TRANSPORT_BC_TCP; 504 args.protocol = XPRT_TRANSPORT_BC_TCP;
500 } 505 }
501 /* Create RPC client */ 506 /* Create RPC client */
@@ -505,7 +510,7 @@ int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *cb)
505 PTR_ERR(client)); 510 PTR_ERR(client));
506 return PTR_ERR(client); 511 return PTR_ERR(client);
507 } 512 }
508 nfsd4_set_callback_client(clp, client); 513 clp->cl_cb_client = client;
509 return 0; 514 return 0;
510 515
511} 516}
@@ -518,7 +523,7 @@ static void warn_no_callback_path(struct nfs4_client *clp, int reason)
518 523
519static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata) 524static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
520{ 525{
521 struct nfs4_client *clp = calldata; 526 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
522 527
523 if (task->tk_status) 528 if (task->tk_status)
524 warn_no_callback_path(clp, task->tk_status); 529 warn_no_callback_path(clp, task->tk_status);
@@ -527,6 +532,8 @@ static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
527} 532}
528 533
529static const struct rpc_call_ops nfsd4_cb_probe_ops = { 534static const struct rpc_call_ops nfsd4_cb_probe_ops = {
535 /* XXX: release method to ensure we set the cb channel down if
536 * necessary on early failure? */
530 .rpc_call_done = nfsd4_cb_probe_done, 537 .rpc_call_done = nfsd4_cb_probe_done,
531}; 538};
532 539
@@ -542,38 +549,42 @@ int set_callback_cred(void)
542 return 0; 549 return 0;
543} 550}
544 551
552static struct workqueue_struct *callback_wq;
545 553
546void do_probe_callback(struct nfs4_client *clp) 554static void do_probe_callback(struct nfs4_client *clp)
547{ 555{
548 struct rpc_message msg = { 556 struct nfsd4_callback *cb = &clp->cl_cb_null;
549 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL],
550 .rpc_argp = clp,
551 .rpc_cred = callback_cred
552 };
553 int status;
554 557
555 status = rpc_call_async(clp->cl_cb_client, &msg, 558 cb->cb_op = NULL;
556 RPC_TASK_SOFT | RPC_TASK_SOFTCONN, 559 cb->cb_clp = clp;
557 &nfsd4_cb_probe_ops, (void *)clp); 560
558 if (status) 561 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL];
559 warn_no_callback_path(clp, status); 562 cb->cb_msg.rpc_argp = NULL;
563 cb->cb_msg.rpc_resp = NULL;
564 cb->cb_msg.rpc_cred = callback_cred;
565
566 cb->cb_ops = &nfsd4_cb_probe_ops;
567
568 queue_work(callback_wq, &cb->cb_work);
560} 569}
561 570
562/* 571/*
563 * Set up the callback client and put a NFSPROC4_CB_NULL on the wire... 572 * Poke the callback thread to process any updates to the callback
573 * parameters, and send a null probe.
564 */ 574 */
565void nfsd4_probe_callback(struct nfs4_client *clp, struct nfs4_cb_conn *cb) 575void nfsd4_probe_callback(struct nfs4_client *clp)
566{ 576{
567 int status; 577 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_cb_flags);
578 do_probe_callback(clp);
579}
568 580
581void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
582{
569 BUG_ON(atomic_read(&clp->cl_cb_set)); 583 BUG_ON(atomic_read(&clp->cl_cb_set));
570 584
571 status = setup_callback_client(clp, cb); 585 spin_lock(&clp->cl_lock);
572 if (status) { 586 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
573 warn_no_callback_path(clp, status); 587 spin_unlock(&clp->cl_lock);
574 return;
575 }
576 do_probe_callback(clp);
577} 588}
578 589
579/* 590/*
@@ -584,8 +595,7 @@ void nfsd4_probe_callback(struct nfs4_client *clp, struct nfs4_cb_conn *cb)
584static int nfsd41_cb_setup_sequence(struct nfs4_client *clp, 595static int nfsd41_cb_setup_sequence(struct nfs4_client *clp,
585 struct rpc_task *task) 596 struct rpc_task *task)
586{ 597{
587 struct nfs4_rpc_args *args = task->tk_msg.rpc_argp; 598 u32 *ptr = (u32 *)clp->cl_cb_session->se_sessionid.data;
588 u32 *ptr = (u32 *)clp->cl_sessionid.data;
589 int status = 0; 599 int status = 0;
590 600
591 dprintk("%s: %u:%u:%u:%u\n", __func__, 601 dprintk("%s: %u:%u:%u:%u\n", __func__,
@@ -597,14 +607,6 @@ static int nfsd41_cb_setup_sequence(struct nfs4_client *clp,
597 status = -EAGAIN; 607 status = -EAGAIN;
598 goto out; 608 goto out;
599 } 609 }
600
601 /*
602 * We'll need the clp during XDR encoding and decoding,
603 * and the sequence during decoding to verify the reply
604 */
605 args->args_seq.cbs_clp = clp;
606 task->tk_msg.rpc_resp = &args->args_seq;
607
608out: 610out:
609 dprintk("%s status=%d\n", __func__, status); 611 dprintk("%s status=%d\n", __func__, status);
610 return status; 612 return status;
@@ -616,13 +618,13 @@ out:
616 */ 618 */
617static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata) 619static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
618{ 620{
619 struct nfs4_delegation *dp = calldata; 621 struct nfsd4_callback *cb = calldata;
622 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
620 struct nfs4_client *clp = dp->dl_client; 623 struct nfs4_client *clp = dp->dl_client;
621 struct nfs4_rpc_args *args = task->tk_msg.rpc_argp; 624 u32 minorversion = clp->cl_minorversion;
622 u32 minorversion = clp->cl_cb_conn.cb_minorversion;
623 int status = 0; 625 int status = 0;
624 626
625 args->args_seq.cbs_minorversion = minorversion; 627 cb->cb_minorversion = minorversion;
626 if (minorversion) { 628 if (minorversion) {
627 status = nfsd41_cb_setup_sequence(clp, task); 629 status = nfsd41_cb_setup_sequence(clp, task);
628 if (status) { 630 if (status) {
@@ -639,19 +641,20 @@ static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
639 641
640static void nfsd4_cb_done(struct rpc_task *task, void *calldata) 642static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
641{ 643{
642 struct nfs4_delegation *dp = calldata; 644 struct nfsd4_callback *cb = calldata;
645 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
643 struct nfs4_client *clp = dp->dl_client; 646 struct nfs4_client *clp = dp->dl_client;
644 647
645 dprintk("%s: minorversion=%d\n", __func__, 648 dprintk("%s: minorversion=%d\n", __func__,
646 clp->cl_cb_conn.cb_minorversion); 649 clp->cl_minorversion);
647 650
648 if (clp->cl_cb_conn.cb_minorversion) { 651 if (clp->cl_minorversion) {
649 /* No need for lock, access serialized in nfsd4_cb_prepare */ 652 /* No need for lock, access serialized in nfsd4_cb_prepare */
650 ++clp->cl_cb_seq_nr; 653 ++clp->cl_cb_session->se_cb_seq_nr;
651 clear_bit(0, &clp->cl_cb_slot_busy); 654 clear_bit(0, &clp->cl_cb_slot_busy);
652 rpc_wake_up_next(&clp->cl_cb_waitq); 655 rpc_wake_up_next(&clp->cl_cb_waitq);
653 dprintk("%s: freed slot, new seqid=%d\n", __func__, 656 dprintk("%s: freed slot, new seqid=%d\n", __func__,
654 clp->cl_cb_seq_nr); 657 clp->cl_cb_session->se_cb_seq_nr);
655 658
656 /* We're done looking into the sequence information */ 659 /* We're done looking into the sequence information */
657 task->tk_msg.rpc_resp = NULL; 660 task->tk_msg.rpc_resp = NULL;
@@ -661,7 +664,8 @@ static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
661 664
662static void nfsd4_cb_recall_done(struct rpc_task *task, void *calldata) 665static void nfsd4_cb_recall_done(struct rpc_task *task, void *calldata)
663{ 666{
664 struct nfs4_delegation *dp = calldata; 667 struct nfsd4_callback *cb = calldata;
668 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
665 struct nfs4_client *clp = dp->dl_client; 669 struct nfs4_client *clp = dp->dl_client;
666 struct rpc_clnt *current_rpc_client = clp->cl_cb_client; 670 struct rpc_clnt *current_rpc_client = clp->cl_cb_client;
667 671
@@ -706,7 +710,8 @@ static void nfsd4_cb_recall_done(struct rpc_task *task, void *calldata)
706 710
707static void nfsd4_cb_recall_release(void *calldata) 711static void nfsd4_cb_recall_release(void *calldata)
708{ 712{
709 struct nfs4_delegation *dp = calldata; 713 struct nfsd4_callback *cb = calldata;
714 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
710 715
711 nfs4_put_delegation(dp); 716 nfs4_put_delegation(dp);
712} 717}
@@ -717,8 +722,6 @@ static const struct rpc_call_ops nfsd4_cb_recall_ops = {
717 .rpc_release = nfsd4_cb_recall_release, 722 .rpc_release = nfsd4_cb_recall_release,
718}; 723};
719 724
720static struct workqueue_struct *callback_wq;
721
722int nfsd4_create_callback_queue(void) 725int nfsd4_create_callback_queue(void)
723{ 726{
724 callback_wq = create_singlethread_workqueue("nfsd4_callbacks"); 727 callback_wq = create_singlethread_workqueue("nfsd4_callbacks");
@@ -733,57 +736,88 @@ void nfsd4_destroy_callback_queue(void)
733} 736}
734 737
735/* must be called under the state lock */ 738/* must be called under the state lock */
736void nfsd4_set_callback_client(struct nfs4_client *clp, struct rpc_clnt *new) 739void nfsd4_shutdown_callback(struct nfs4_client *clp)
737{ 740{
738 struct rpc_clnt *old = clp->cl_cb_client; 741 set_bit(NFSD4_CLIENT_KILL, &clp->cl_cb_flags);
739
740 clp->cl_cb_client = new;
741 /* 742 /*
742 * After this, any work that saw the old value of cl_cb_client will 743 * Note this won't actually result in a null callback;
743 * be gone: 744 * instead, nfsd4_do_callback_rpc() will detect the killed
745 * client, destroy the rpc client, and stop:
744 */ 746 */
747 do_probe_callback(clp);
745 flush_workqueue(callback_wq); 748 flush_workqueue(callback_wq);
746 /* So we can safely shut it down: */
747 if (old)
748 rpc_shutdown_client(old);
749} 749}
750 750
751/* 751void nfsd4_release_cb(struct nfsd4_callback *cb)
752 * called with dp->dl_count inc'ed.
753 */
754static void _nfsd4_cb_recall(struct nfs4_delegation *dp)
755{ 752{
756 struct nfs4_client *clp = dp->dl_client; 753 if (cb->cb_ops->rpc_release)
757 struct rpc_clnt *clnt = clp->cl_cb_client; 754 cb->cb_ops->rpc_release(cb);
758 struct nfs4_rpc_args *args = &dp->dl_recall.cb_args; 755}
759 struct rpc_message msg = {
760 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_RECALL],
761 .rpc_cred = callback_cred
762 };
763 756
764 if (clnt == NULL) { 757void nfsd4_process_cb_update(struct nfsd4_callback *cb)
765 nfs4_put_delegation(dp); 758{
766 return; /* Client is shutting down; give up. */ 759 struct nfs4_cb_conn conn;
760 struct nfs4_client *clp = cb->cb_clp;
761 int err;
762
763 /*
764 * This is either an update, or the client dying; in either case,
765 * kill the old client:
766 */
767 if (clp->cl_cb_client) {
768 rpc_shutdown_client(clp->cl_cb_client);
769 clp->cl_cb_client = NULL;
767 } 770 }
771 if (test_bit(NFSD4_CLIENT_KILL, &clp->cl_cb_flags))
772 return;
773 spin_lock(&clp->cl_lock);
774 /*
775 * Only serialized callback code is allowed to clear these
776 * flags; main nfsd code can only set them:
777 */
778 BUG_ON(!clp->cl_cb_flags);
779 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_cb_flags);
780 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
781 spin_unlock(&clp->cl_lock);
768 782
769 args->args_op = dp; 783 err = setup_callback_client(clp, &conn);
770 msg.rpc_argp = args; 784 if (err)
771 dp->dl_retries = 1; 785 warn_no_callback_path(clp, err);
772 rpc_call_async(clnt, &msg, RPC_TASK_SOFT, &nfsd4_cb_recall_ops, dp);
773} 786}
774 787
775void nfsd4_do_callback_rpc(struct work_struct *w) 788void nfsd4_do_callback_rpc(struct work_struct *w)
776{ 789{
777 /* XXX: for now, just send off delegation recall. */ 790 struct nfsd4_callback *cb = container_of(w, struct nfsd4_callback, cb_work);
778 /* In future, generalize to handle any sort of callback. */ 791 struct nfs4_client *clp = cb->cb_clp;
779 struct nfsd4_callback *c = container_of(w, struct nfsd4_callback, cb_work); 792 struct rpc_clnt *clnt;
780 struct nfs4_delegation *dp = container_of(c, struct nfs4_delegation, dl_recall);
781 793
782 _nfsd4_cb_recall(dp); 794 if (clp->cl_cb_flags)
783} 795 nfsd4_process_cb_update(cb);
784 796
797 clnt = clp->cl_cb_client;
798 if (!clnt) {
799 /* Callback channel broken, or client killed; give up: */
800 nfsd4_release_cb(cb);
801 return;
802 }
803 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
804 cb->cb_ops, cb);
805}
785 806
786void nfsd4_cb_recall(struct nfs4_delegation *dp) 807void nfsd4_cb_recall(struct nfs4_delegation *dp)
787{ 808{
809 struct nfsd4_callback *cb = &dp->dl_recall;
810
811 dp->dl_retries = 1;
812 cb->cb_op = dp;
813 cb->cb_clp = dp->dl_client;
814 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_RECALL];
815 cb->cb_msg.rpc_argp = cb;
816 cb->cb_msg.rpc_resp = cb;
817 cb->cb_msg.rpc_cred = callback_cred;
818
819 cb->cb_ops = &nfsd4_cb_recall_ops;
820 dp->dl_retries = 1;
821
788 queue_work(callback_wq, &dp->dl_recall.cb_work); 822 queue_work(callback_wq, &dp->dl_recall.cb_work);
789} 823}
diff --git a/fs/nfsd/nfs4idmap.c b/fs/nfsd/nfs4idmap.c
index c78dbf493424..f0695e815f0e 100644
--- a/fs/nfsd/nfs4idmap.c
+++ b/fs/nfsd/nfs4idmap.c
@@ -482,109 +482,26 @@ nfsd_idmap_shutdown(void)
482 cache_unregister(&nametoid_cache); 482 cache_unregister(&nametoid_cache);
483} 483}
484 484
485/*
486 * Deferred request handling
487 */
488
489struct idmap_defer_req {
490 struct cache_req req;
491 struct cache_deferred_req deferred_req;
492 wait_queue_head_t waitq;
493 atomic_t count;
494};
495
496static inline void
497put_mdr(struct idmap_defer_req *mdr)
498{
499 if (atomic_dec_and_test(&mdr->count))
500 kfree(mdr);
501}
502
503static inline void
504get_mdr(struct idmap_defer_req *mdr)
505{
506 atomic_inc(&mdr->count);
507}
508
509static void
510idmap_revisit(struct cache_deferred_req *dreq, int toomany)
511{
512 struct idmap_defer_req *mdr =
513 container_of(dreq, struct idmap_defer_req, deferred_req);
514
515 wake_up(&mdr->waitq);
516 put_mdr(mdr);
517}
518
519static struct cache_deferred_req *
520idmap_defer(struct cache_req *req)
521{
522 struct idmap_defer_req *mdr =
523 container_of(req, struct idmap_defer_req, req);
524
525 mdr->deferred_req.revisit = idmap_revisit;
526 get_mdr(mdr);
527 return (&mdr->deferred_req);
528}
529
530static inline int
531do_idmap_lookup(struct ent *(*lookup_fn)(struct ent *), struct ent *key,
532 struct cache_detail *detail, struct ent **item,
533 struct idmap_defer_req *mdr)
534{
535 *item = lookup_fn(key);
536 if (!*item)
537 return -ENOMEM;
538 return cache_check(detail, &(*item)->h, &mdr->req);
539}
540
541static inline int
542do_idmap_lookup_nowait(struct ent *(*lookup_fn)(struct ent *),
543 struct ent *key, struct cache_detail *detail,
544 struct ent **item)
545{
546 int ret = -ENOMEM;
547
548 *item = lookup_fn(key);
549 if (!*item)
550 goto out_err;
551 ret = -ETIMEDOUT;
552 if (!test_bit(CACHE_VALID, &(*item)->h.flags)
553 || (*item)->h.expiry_time < get_seconds()
554 || detail->flush_time > (*item)->h.last_refresh)
555 goto out_put;
556 ret = -ENOENT;
557 if (test_bit(CACHE_NEGATIVE, &(*item)->h.flags))
558 goto out_put;
559 return 0;
560out_put:
561 cache_put(&(*item)->h, detail);
562out_err:
563 *item = NULL;
564 return ret;
565}
566
567static int 485static int
568idmap_lookup(struct svc_rqst *rqstp, 486idmap_lookup(struct svc_rqst *rqstp,
569 struct ent *(*lookup_fn)(struct ent *), struct ent *key, 487 struct ent *(*lookup_fn)(struct ent *), struct ent *key,
570 struct cache_detail *detail, struct ent **item) 488 struct cache_detail *detail, struct ent **item)
571{ 489{
572 struct idmap_defer_req *mdr;
573 int ret; 490 int ret;
574 491
575 mdr = kzalloc(sizeof(*mdr), GFP_KERNEL); 492 *item = lookup_fn(key);
576 if (!mdr) 493 if (!*item)
577 return -ENOMEM; 494 return -ENOMEM;
578 atomic_set(&mdr->count, 1); 495 retry:
579 init_waitqueue_head(&mdr->waitq); 496 ret = cache_check(detail, &(*item)->h, &rqstp->rq_chandle);
580 mdr->req.defer = idmap_defer; 497
581 ret = do_idmap_lookup(lookup_fn, key, detail, item, mdr); 498 if (ret == -ETIMEDOUT) {
582 if (ret == -EAGAIN) { 499 struct ent *prev_item = *item;
583 wait_event_interruptible_timeout(mdr->waitq, 500 *item = lookup_fn(key);
584 test_bit(CACHE_VALID, &(*item)->h.flags), 1 * HZ); 501 if (*item != prev_item)
585 ret = do_idmap_lookup_nowait(lookup_fn, key, detail, item); 502 goto retry;
503 cache_put(&(*item)->h, detail);
586 } 504 }
587 put_mdr(mdr);
588 return ret; 505 return ret;
589} 506}
590 507
diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index 59ec449b0c7f..0cdfd022bb7b 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1031,8 +1031,11 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
1031 resp->cstate.session = NULL; 1031 resp->cstate.session = NULL;
1032 fh_init(&resp->cstate.current_fh, NFS4_FHSIZE); 1032 fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
1033 fh_init(&resp->cstate.save_fh, NFS4_FHSIZE); 1033 fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
1034 /* Use the deferral mechanism only for NFSv4.0 compounds */ 1034 /*
1035 rqstp->rq_usedeferral = (args->minorversion == 0); 1035 * Don't use the deferral mechanism for NFSv4; compounds make it
1036 * too hard to avoid non-idempotency problems.
1037 */
1038 rqstp->rq_usedeferral = 0;
1036 1039
1037 /* 1040 /*
1038 * According to RFC3010, this takes precedence over all other errors. 1041 * According to RFC3010, this takes precedence over all other errors.
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index a7292fcf7718..9019e8ec9dc8 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -207,7 +207,6 @@ alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_f
207{ 207{
208 struct nfs4_delegation *dp; 208 struct nfs4_delegation *dp;
209 struct nfs4_file *fp = stp->st_file; 209 struct nfs4_file *fp = stp->st_file;
210 struct nfs4_cb_conn *cb = &stp->st_stateowner->so_client->cl_cb_conn;
211 210
212 dprintk("NFSD alloc_init_deleg\n"); 211 dprintk("NFSD alloc_init_deleg\n");
213 /* 212 /*
@@ -234,7 +233,6 @@ alloc_init_deleg(struct nfs4_client *clp, struct nfs4_stateid *stp, struct svc_f
234 nfs4_file_get_access(fp, O_RDONLY); 233 nfs4_file_get_access(fp, O_RDONLY);
235 dp->dl_flock = NULL; 234 dp->dl_flock = NULL;
236 dp->dl_type = type; 235 dp->dl_type = type;
237 dp->dl_ident = cb->cb_ident;
238 dp->dl_stateid.si_boot = boot_time; 236 dp->dl_stateid.si_boot = boot_time;
239 dp->dl_stateid.si_stateownerid = current_delegid++; 237 dp->dl_stateid.si_stateownerid = current_delegid++;
240 dp->dl_stateid.si_fileid = 0; 238 dp->dl_stateid.si_fileid = 0;
@@ -535,171 +533,258 @@ gen_sessionid(struct nfsd4_session *ses)
535 */ 533 */
536#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44) 534#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
537 535
536static void
537free_session_slots(struct nfsd4_session *ses)
538{
539 int i;
540
541 for (i = 0; i < ses->se_fchannel.maxreqs; i++)
542 kfree(ses->se_slots[i]);
543}
544
538/* 545/*
539 * Give the client the number of ca_maxresponsesize_cached slots it 546 * We don't actually need to cache the rpc and session headers, so we
540 * requests, of size bounded by NFSD_SLOT_CACHE_SIZE, 547 * can allocate a little less for each slot:
541 * NFSD_MAX_MEM_PER_SESSION, and nfsd_drc_max_mem. Do not allow more 548 */
542 * than NFSD_MAX_SLOTS_PER_SESSION. 549static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
543 * 550{
544 * If we run out of reserved DRC memory we should (up to a point) 551 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
552}
553
554static int nfsd4_sanitize_slot_size(u32 size)
555{
556 size -= NFSD_MIN_HDR_SEQ_SZ; /* We don't cache the rpc header */
557 size = min_t(u32, size, NFSD_SLOT_CACHE_SIZE);
558
559 return size;
560}
561
562/*
563 * XXX: If we run out of reserved DRC memory we could (up to a point)
545 * re-negotiate active sessions and reduce their slot usage to make 564 * re-negotiate active sessions and reduce their slot usage to make
546 * rooom for new connections. For now we just fail the create session. 565 * rooom for new connections. For now we just fail the create session.
547 */ 566 */
548static int set_forechannel_drc_size(struct nfsd4_channel_attrs *fchan) 567static int nfsd4_get_drc_mem(int slotsize, u32 num)
549{ 568{
550 int mem, size = fchan->maxresp_cached; 569 int avail;
551 570
552 if (fchan->maxreqs < 1) 571 num = min_t(u32, num, NFSD_MAX_SLOTS_PER_SESSION);
553 return nfserr_inval;
554 572
555 if (size < NFSD_MIN_HDR_SEQ_SZ) 573 spin_lock(&nfsd_drc_lock);
556 size = NFSD_MIN_HDR_SEQ_SZ; 574 avail = min_t(int, NFSD_MAX_MEM_PER_SESSION,
557 size -= NFSD_MIN_HDR_SEQ_SZ; 575 nfsd_drc_max_mem - nfsd_drc_mem_used);
558 if (size > NFSD_SLOT_CACHE_SIZE) 576 num = min_t(int, num, avail / slotsize);
559 size = NFSD_SLOT_CACHE_SIZE; 577 nfsd_drc_mem_used += num * slotsize;
560 578 spin_unlock(&nfsd_drc_lock);
561 /* bound the maxreqs by NFSD_MAX_MEM_PER_SESSION */
562 mem = fchan->maxreqs * size;
563 if (mem > NFSD_MAX_MEM_PER_SESSION) {
564 fchan->maxreqs = NFSD_MAX_MEM_PER_SESSION / size;
565 if (fchan->maxreqs > NFSD_MAX_SLOTS_PER_SESSION)
566 fchan->maxreqs = NFSD_MAX_SLOTS_PER_SESSION;
567 mem = fchan->maxreqs * size;
568 }
569 579
580 return num;
581}
582
583static void nfsd4_put_drc_mem(int slotsize, int num)
584{
570 spin_lock(&nfsd_drc_lock); 585 spin_lock(&nfsd_drc_lock);
571 /* bound the total session drc memory ussage */ 586 nfsd_drc_mem_used -= slotsize * num;
572 if (mem + nfsd_drc_mem_used > nfsd_drc_max_mem) {
573 fchan->maxreqs = (nfsd_drc_max_mem - nfsd_drc_mem_used) / size;
574 mem = fchan->maxreqs * size;
575 }
576 nfsd_drc_mem_used += mem;
577 spin_unlock(&nfsd_drc_lock); 587 spin_unlock(&nfsd_drc_lock);
588}
578 589
579 if (fchan->maxreqs == 0) 590static struct nfsd4_session *alloc_session(int slotsize, int numslots)
580 return nfserr_jukebox; 591{
592 struct nfsd4_session *new;
593 int mem, i;
581 594
582 fchan->maxresp_cached = size + NFSD_MIN_HDR_SEQ_SZ; 595 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
583 return 0; 596 + sizeof(struct nfsd4_session) > PAGE_SIZE);
597 mem = numslots * sizeof(struct nfsd4_slot *);
598
599 new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
600 if (!new)
601 return NULL;
602 /* allocate each struct nfsd4_slot and data cache in one piece */
603 for (i = 0; i < numslots; i++) {
604 mem = sizeof(struct nfsd4_slot) + slotsize;
605 new->se_slots[i] = kzalloc(mem, GFP_KERNEL);
606 if (!new->se_slots[i])
607 goto out_free;
608 }
609 return new;
610out_free:
611 while (i--)
612 kfree(new->se_slots[i]);
613 kfree(new);
614 return NULL;
584} 615}
585 616
586/* 617static void init_forechannel_attrs(struct nfsd4_channel_attrs *new, struct nfsd4_channel_attrs *req, int numslots, int slotsize)
587 * fchan holds the client values on input, and the server values on output
588 * sv_max_mesg is the maximum payload plus one page for overhead.
589 */
590static int init_forechannel_attrs(struct svc_rqst *rqstp,
591 struct nfsd4_channel_attrs *session_fchan,
592 struct nfsd4_channel_attrs *fchan)
593{ 618{
594 int status = 0; 619 u32 maxrpc = nfsd_serv->sv_max_mesg;
595 __u32 maxcount = nfsd_serv->sv_max_mesg;
596 620
597 /* headerpadsz set to zero in encode routine */ 621 new->maxreqs = numslots;
622 new->maxresp_cached = slotsize + NFSD_MIN_HDR_SEQ_SZ;
623 new->maxreq_sz = min_t(u32, req->maxreq_sz, maxrpc);
624 new->maxresp_sz = min_t(u32, req->maxresp_sz, maxrpc);
625 new->maxops = min_t(u32, req->maxops, NFSD_MAX_OPS_PER_COMPOUND);
626}
598 627
599 /* Use the client's max request and max response size if possible */ 628static void free_conn(struct nfsd4_conn *c)
600 if (fchan->maxreq_sz > maxcount) 629{
601 fchan->maxreq_sz = maxcount; 630 svc_xprt_put(c->cn_xprt);
602 session_fchan->maxreq_sz = fchan->maxreq_sz; 631 kfree(c);
632}
603 633
604 if (fchan->maxresp_sz > maxcount) 634static void nfsd4_conn_lost(struct svc_xpt_user *u)
605 fchan->maxresp_sz = maxcount; 635{
606 session_fchan->maxresp_sz = fchan->maxresp_sz; 636 struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
637 struct nfs4_client *clp = c->cn_session->se_client;
607 638
608 /* Use the client's maxops if possible */ 639 spin_lock(&clp->cl_lock);
609 if (fchan->maxops > NFSD_MAX_OPS_PER_COMPOUND) 640 if (!list_empty(&c->cn_persession)) {
610 fchan->maxops = NFSD_MAX_OPS_PER_COMPOUND; 641 list_del(&c->cn_persession);
611 session_fchan->maxops = fchan->maxops; 642 free_conn(c);
643 }
644 spin_unlock(&clp->cl_lock);
645}
612 646
613 /* FIXME: Error means no more DRC pages so the server should 647static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
614 * recover pages from existing sessions. For now fail session 648{
615 * creation. 649 struct nfsd4_conn *conn;
616 */
617 status = set_forechannel_drc_size(fchan);
618 650
619 session_fchan->maxresp_cached = fchan->maxresp_cached; 651 conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
620 session_fchan->maxreqs = fchan->maxreqs; 652 if (!conn)
653 return NULL;
654 svc_xprt_get(rqstp->rq_xprt);
655 conn->cn_xprt = rqstp->rq_xprt;
656 conn->cn_flags = flags;
657 INIT_LIST_HEAD(&conn->cn_xpt_user.list);
658 return conn;
659}
621 660
622 dprintk("%s status %d\n", __func__, status); 661static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
623 return status; 662{
663 conn->cn_session = ses;
664 list_add(&conn->cn_persession, &ses->se_conns);
624} 665}
625 666
626static void 667static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
627free_session_slots(struct nfsd4_session *ses)
628{ 668{
629 int i; 669 struct nfs4_client *clp = ses->se_client;
630 670
631 for (i = 0; i < ses->se_fchannel.maxreqs; i++) 671 spin_lock(&clp->cl_lock);
632 kfree(ses->se_slots[i]); 672 __nfsd4_hash_conn(conn, ses);
673 spin_unlock(&clp->cl_lock);
633} 674}
634 675
635/* 676static void nfsd4_register_conn(struct nfsd4_conn *conn)
636 * We don't actually need to cache the rpc and session headers, so we
637 * can allocate a little less for each slot:
638 */
639static inline int slot_bytes(struct nfsd4_channel_attrs *ca)
640{ 677{
641 return ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ; 678 conn->cn_xpt_user.callback = nfsd4_conn_lost;
679 register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
642} 680}
643 681
644static int 682static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses)
645alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp,
646 struct nfsd4_create_session *cses)
647{ 683{
648 struct nfsd4_session *new, tmp; 684 struct nfsd4_conn *conn;
649 struct nfsd4_slot *sp; 685 u32 flags = NFS4_CDFC4_FORE;
650 int idx, slotsize, cachesize, i;
651 int status;
652 686
653 memset(&tmp, 0, sizeof(tmp)); 687 if (ses->se_flags & SESSION4_BACK_CHAN)
688 flags |= NFS4_CDFC4_BACK;
689 conn = alloc_conn(rqstp, flags);
690 if (!conn)
691 return nfserr_jukebox;
692 nfsd4_hash_conn(conn, ses);
693 nfsd4_register_conn(conn);
694 return nfs_ok;
695}
654 696
655 /* FIXME: For now, we just accept the client back channel attributes. */ 697static void nfsd4_del_conns(struct nfsd4_session *s)
656 tmp.se_bchannel = cses->back_channel; 698{
657 status = init_forechannel_attrs(rqstp, &tmp.se_fchannel, 699 struct nfs4_client *clp = s->se_client;
658 &cses->fore_channel); 700 struct nfsd4_conn *c;
659 if (status)
660 goto out;
661 701
662 BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot) 702 spin_lock(&clp->cl_lock);
663 + sizeof(struct nfsd4_session) > PAGE_SIZE); 703 while (!list_empty(&s->se_conns)) {
704 c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
705 list_del_init(&c->cn_persession);
706 spin_unlock(&clp->cl_lock);
664 707
665 status = nfserr_jukebox; 708 unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
666 /* allocate struct nfsd4_session and slot table pointers in one piece */ 709 free_conn(c);
667 slotsize = tmp.se_fchannel.maxreqs * sizeof(struct nfsd4_slot *);
668 new = kzalloc(sizeof(*new) + slotsize, GFP_KERNEL);
669 if (!new)
670 goto out;
671 710
672 memcpy(new, &tmp, sizeof(*new)); 711 spin_lock(&clp->cl_lock);
712 }
713 spin_unlock(&clp->cl_lock);
714}
673 715
674 /* allocate each struct nfsd4_slot and data cache in one piece */ 716void free_session(struct kref *kref)
675 cachesize = slot_bytes(&new->se_fchannel); 717{
676 for (i = 0; i < new->se_fchannel.maxreqs; i++) { 718 struct nfsd4_session *ses;
677 sp = kzalloc(sizeof(*sp) + cachesize, GFP_KERNEL); 719 int mem;
678 if (!sp) 720
679 goto out_free; 721 ses = container_of(kref, struct nfsd4_session, se_ref);
680 new->se_slots[i] = sp; 722 nfsd4_del_conns(ses);
723 spin_lock(&nfsd_drc_lock);
724 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
725 nfsd_drc_mem_used -= mem;
726 spin_unlock(&nfsd_drc_lock);
727 free_session_slots(ses);
728 kfree(ses);
729}
730
731static struct nfsd4_session *alloc_init_session(struct svc_rqst *rqstp, struct nfs4_client *clp, struct nfsd4_create_session *cses)
732{
733 struct nfsd4_session *new;
734 struct nfsd4_channel_attrs *fchan = &cses->fore_channel;
735 int numslots, slotsize;
736 int status;
737 int idx;
738
739 /*
740 * Note decreasing slot size below client's request may
741 * make it difficult for client to function correctly, whereas
742 * decreasing the number of slots will (just?) affect
743 * performance. When short on memory we therefore prefer to
744 * decrease number of slots instead of their size.
745 */
746 slotsize = nfsd4_sanitize_slot_size(fchan->maxresp_cached);
747 numslots = nfsd4_get_drc_mem(slotsize, fchan->maxreqs);
748
749 new = alloc_session(slotsize, numslots);
750 if (!new) {
751 nfsd4_put_drc_mem(slotsize, fchan->maxreqs);
752 return NULL;
681 } 753 }
754 init_forechannel_attrs(&new->se_fchannel, fchan, numslots, slotsize);
682 755
683 new->se_client = clp; 756 new->se_client = clp;
684 gen_sessionid(new); 757 gen_sessionid(new);
685 idx = hash_sessionid(&new->se_sessionid);
686 memcpy(clp->cl_sessionid.data, new->se_sessionid.data,
687 NFS4_MAX_SESSIONID_LEN);
688 758
759 INIT_LIST_HEAD(&new->se_conns);
760
761 new->se_cb_seq_nr = 1;
689 new->se_flags = cses->flags; 762 new->se_flags = cses->flags;
763 new->se_cb_prog = cses->callback_prog;
690 kref_init(&new->se_ref); 764 kref_init(&new->se_ref);
765 idx = hash_sessionid(&new->se_sessionid);
691 spin_lock(&client_lock); 766 spin_lock(&client_lock);
692 list_add(&new->se_hash, &sessionid_hashtbl[idx]); 767 list_add(&new->se_hash, &sessionid_hashtbl[idx]);
693 list_add(&new->se_perclnt, &clp->cl_sessions); 768 list_add(&new->se_perclnt, &clp->cl_sessions);
694 spin_unlock(&client_lock); 769 spin_unlock(&client_lock);
695 770
696 status = nfs_ok; 771 status = nfsd4_new_conn(rqstp, new);
697out: 772 /* whoops: benny points out, status is ignored! (err, or bogus) */
698 return status; 773 if (status) {
699out_free: 774 free_session(&new->se_ref);
700 free_session_slots(new); 775 return NULL;
701 kfree(new); 776 }
702 goto out; 777 if (!clp->cl_cb_session && (cses->flags & SESSION4_BACK_CHAN)) {
778 struct sockaddr *sa = svc_addr(rqstp);
779
780 clp->cl_cb_session = new;
781 clp->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
782 svc_xprt_get(rqstp->rq_xprt);
783 rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
784 clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
785 nfsd4_probe_callback(clp);
786 }
787 return new;
703} 788}
704 789
705/* caller must hold client_lock */ 790/* caller must hold client_lock */
@@ -731,21 +816,6 @@ unhash_session(struct nfsd4_session *ses)
731 list_del(&ses->se_perclnt); 816 list_del(&ses->se_perclnt);
732} 817}
733 818
734void
735free_session(struct kref *kref)
736{
737 struct nfsd4_session *ses;
738 int mem;
739
740 ses = container_of(kref, struct nfsd4_session, se_ref);
741 spin_lock(&nfsd_drc_lock);
742 mem = ses->se_fchannel.maxreqs * slot_bytes(&ses->se_fchannel);
743 nfsd_drc_mem_used -= mem;
744 spin_unlock(&nfsd_drc_lock);
745 free_session_slots(ses);
746 kfree(ses);
747}
748
749/* must be called under the client_lock */ 819/* must be called under the client_lock */
750static inline void 820static inline void
751renew_client_locked(struct nfs4_client *clp) 821renew_client_locked(struct nfs4_client *clp)
@@ -812,6 +882,13 @@ static struct nfs4_client *alloc_client(struct xdr_netobj name)
812static inline void 882static inline void
813free_client(struct nfs4_client *clp) 883free_client(struct nfs4_client *clp)
814{ 884{
885 while (!list_empty(&clp->cl_sessions)) {
886 struct nfsd4_session *ses;
887 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
888 se_perclnt);
889 list_del(&ses->se_perclnt);
890 nfsd4_put_session(ses);
891 }
815 if (clp->cl_cred.cr_group_info) 892 if (clp->cl_cred.cr_group_info)
816 put_group_info(clp->cl_cred.cr_group_info); 893 put_group_info(clp->cl_cred.cr_group_info);
817 kfree(clp->cl_principal); 894 kfree(clp->cl_principal);
@@ -838,15 +915,12 @@ release_session_client(struct nfsd4_session *session)
838static inline void 915static inline void
839unhash_client_locked(struct nfs4_client *clp) 916unhash_client_locked(struct nfs4_client *clp)
840{ 917{
918 struct nfsd4_session *ses;
919
841 mark_client_expired(clp); 920 mark_client_expired(clp);
842 list_del(&clp->cl_lru); 921 list_del(&clp->cl_lru);
843 while (!list_empty(&clp->cl_sessions)) { 922 list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
844 struct nfsd4_session *ses; 923 list_del_init(&ses->se_hash);
845 ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
846 se_perclnt);
847 unhash_session(ses);
848 nfsd4_put_session(ses);
849 }
850} 924}
851 925
852static void 926static void
@@ -875,7 +949,7 @@ expire_client(struct nfs4_client *clp)
875 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient); 949 sop = list_entry(clp->cl_openowners.next, struct nfs4_stateowner, so_perclient);
876 release_openowner(sop); 950 release_openowner(sop);
877 } 951 }
878 nfsd4_set_callback_client(clp, NULL); 952 nfsd4_shutdown_callback(clp);
879 if (clp->cl_cb_conn.cb_xprt) 953 if (clp->cl_cb_conn.cb_xprt)
880 svc_xprt_put(clp->cl_cb_conn.cb_xprt); 954 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
881 list_del(&clp->cl_idhash); 955 list_del(&clp->cl_idhash);
@@ -960,6 +1034,8 @@ static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
960 if (clp == NULL) 1034 if (clp == NULL)
961 return NULL; 1035 return NULL;
962 1036
1037 INIT_LIST_HEAD(&clp->cl_sessions);
1038
963 princ = svc_gss_principal(rqstp); 1039 princ = svc_gss_principal(rqstp);
964 if (princ) { 1040 if (princ) {
965 clp->cl_principal = kstrdup(princ, GFP_KERNEL); 1041 clp->cl_principal = kstrdup(princ, GFP_KERNEL);
@@ -976,8 +1052,9 @@ static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
976 INIT_LIST_HEAD(&clp->cl_strhash); 1052 INIT_LIST_HEAD(&clp->cl_strhash);
977 INIT_LIST_HEAD(&clp->cl_openowners); 1053 INIT_LIST_HEAD(&clp->cl_openowners);
978 INIT_LIST_HEAD(&clp->cl_delegations); 1054 INIT_LIST_HEAD(&clp->cl_delegations);
979 INIT_LIST_HEAD(&clp->cl_sessions);
980 INIT_LIST_HEAD(&clp->cl_lru); 1055 INIT_LIST_HEAD(&clp->cl_lru);
1056 spin_lock_init(&clp->cl_lock);
1057 INIT_WORK(&clp->cl_cb_null.cb_work, nfsd4_do_callback_rpc);
981 clp->cl_time = get_seconds(); 1058 clp->cl_time = get_seconds();
982 clear_bit(0, &clp->cl_cb_slot_busy); 1059 clear_bit(0, &clp->cl_cb_slot_busy);
983 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table"); 1060 rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
@@ -986,7 +1063,7 @@ static struct nfs4_client *create_client(struct xdr_netobj name, char *recdir,
986 clp->cl_flavor = rqstp->rq_flavor; 1063 clp->cl_flavor = rqstp->rq_flavor;
987 copy_cred(&clp->cl_cred, &rqstp->rq_cred); 1064 copy_cred(&clp->cl_cred, &rqstp->rq_cred);
988 gen_confirm(clp); 1065 gen_confirm(clp);
989 1066 clp->cl_cb_session = NULL;
990 return clp; 1067 return clp;
991} 1068}
992 1069
@@ -1098,7 +1175,7 @@ find_unconfirmed_client_by_str(const char *dname, unsigned int hashval,
1098static void 1175static void
1099gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid) 1176gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1100{ 1177{
1101 struct nfs4_cb_conn *cb = &clp->cl_cb_conn; 1178 struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
1102 unsigned short expected_family; 1179 unsigned short expected_family;
1103 1180
1104 /* Currently, we only support tcp and tcp6 for the callback channel */ 1181 /* Currently, we only support tcp and tcp6 for the callback channel */
@@ -1111,24 +1188,23 @@ gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, u32 scopeid)
1111 else 1188 else
1112 goto out_err; 1189 goto out_err;
1113 1190
1114 cb->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val, 1191 conn->cb_addrlen = rpc_uaddr2sockaddr(se->se_callback_addr_val,
1115 se->se_callback_addr_len, 1192 se->se_callback_addr_len,
1116 (struct sockaddr *) &cb->cb_addr, 1193 (struct sockaddr *)&conn->cb_addr,
1117 sizeof(cb->cb_addr)); 1194 sizeof(conn->cb_addr));
1118 1195
1119 if (!cb->cb_addrlen || cb->cb_addr.ss_family != expected_family) 1196 if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
1120 goto out_err; 1197 goto out_err;
1121 1198
1122 if (cb->cb_addr.ss_family == AF_INET6) 1199 if (conn->cb_addr.ss_family == AF_INET6)
1123 ((struct sockaddr_in6 *) &cb->cb_addr)->sin6_scope_id = scopeid; 1200 ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
1124 1201
1125 cb->cb_minorversion = 0; 1202 conn->cb_prog = se->se_callback_prog;
1126 cb->cb_prog = se->se_callback_prog; 1203 conn->cb_ident = se->se_callback_ident;
1127 cb->cb_ident = se->se_callback_ident;
1128 return; 1204 return;
1129out_err: 1205out_err:
1130 cb->cb_addr.ss_family = AF_UNSPEC; 1206 conn->cb_addr.ss_family = AF_UNSPEC;
1131 cb->cb_addrlen = 0; 1207 conn->cb_addrlen = 0;
1132 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) " 1208 dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
1133 "will not receive delegations\n", 1209 "will not receive delegations\n",
1134 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id); 1210 clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
@@ -1415,7 +1491,9 @@ nfsd4_create_session(struct svc_rqst *rqstp,
1415{ 1491{
1416 struct sockaddr *sa = svc_addr(rqstp); 1492 struct sockaddr *sa = svc_addr(rqstp);
1417 struct nfs4_client *conf, *unconf; 1493 struct nfs4_client *conf, *unconf;
1494 struct nfsd4_session *new;
1418 struct nfsd4_clid_slot *cs_slot = NULL; 1495 struct nfsd4_clid_slot *cs_slot = NULL;
1496 bool confirm_me = false;
1419 int status = 0; 1497 int status = 0;
1420 1498
1421 nfs4_lock_state(); 1499 nfs4_lock_state();
@@ -1438,7 +1516,6 @@ nfsd4_create_session(struct svc_rqst *rqstp,
1438 cs_slot->sl_seqid, cr_ses->seqid); 1516 cs_slot->sl_seqid, cr_ses->seqid);
1439 goto out; 1517 goto out;
1440 } 1518 }
1441 cs_slot->sl_seqid++;
1442 } else if (unconf) { 1519 } else if (unconf) {
1443 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) || 1520 if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
1444 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) { 1521 !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
@@ -1451,25 +1528,10 @@ nfsd4_create_session(struct svc_rqst *rqstp,
1451 if (status) { 1528 if (status) {
1452 /* an unconfirmed replay returns misordered */ 1529 /* an unconfirmed replay returns misordered */
1453 status = nfserr_seq_misordered; 1530 status = nfserr_seq_misordered;
1454 goto out_cache; 1531 goto out;
1455 } 1532 }
1456 1533
1457 cs_slot->sl_seqid++; /* from 0 to 1 */ 1534 confirm_me = true;
1458 move_to_confirmed(unconf);
1459
1460 if (cr_ses->flags & SESSION4_BACK_CHAN) {
1461 unconf->cl_cb_conn.cb_xprt = rqstp->rq_xprt;
1462 svc_xprt_get(rqstp->rq_xprt);
1463 rpc_copy_addr(
1464 (struct sockaddr *)&unconf->cl_cb_conn.cb_addr,
1465 sa);
1466 unconf->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1467 unconf->cl_cb_conn.cb_minorversion =
1468 cstate->minorversion;
1469 unconf->cl_cb_conn.cb_prog = cr_ses->callback_prog;
1470 unconf->cl_cb_seq_nr = 1;
1471 nfsd4_probe_callback(unconf, &unconf->cl_cb_conn);
1472 }
1473 conf = unconf; 1535 conf = unconf;
1474 } else { 1536 } else {
1475 status = nfserr_stale_clientid; 1537 status = nfserr_stale_clientid;
@@ -1477,22 +1539,30 @@ nfsd4_create_session(struct svc_rqst *rqstp,
1477 } 1539 }
1478 1540
1479 /* 1541 /*
1542 * XXX: we should probably set this at creation time, and check
1543 * for consistent minorversion use throughout:
1544 */
1545 conf->cl_minorversion = 1;
1546 /*
1480 * We do not support RDMA or persistent sessions 1547 * We do not support RDMA or persistent sessions
1481 */ 1548 */
1482 cr_ses->flags &= ~SESSION4_PERSIST; 1549 cr_ses->flags &= ~SESSION4_PERSIST;
1483 cr_ses->flags &= ~SESSION4_RDMA; 1550 cr_ses->flags &= ~SESSION4_RDMA;
1484 1551
1485 status = alloc_init_session(rqstp, conf, cr_ses); 1552 status = nfserr_jukebox;
1486 if (status) 1553 new = alloc_init_session(rqstp, conf, cr_ses);
1554 if (!new)
1487 goto out; 1555 goto out;
1488 1556 status = nfs_ok;
1489 memcpy(cr_ses->sessionid.data, conf->cl_sessionid.data, 1557 memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
1490 NFS4_MAX_SESSIONID_LEN); 1558 NFS4_MAX_SESSIONID_LEN);
1559 cs_slot->sl_seqid++;
1491 cr_ses->seqid = cs_slot->sl_seqid; 1560 cr_ses->seqid = cs_slot->sl_seqid;
1492 1561
1493out_cache:
1494 /* cache solo and embedded create sessions under the state lock */ 1562 /* cache solo and embedded create sessions under the state lock */
1495 nfsd4_cache_create_session(cr_ses, cs_slot, status); 1563 nfsd4_cache_create_session(cr_ses, cs_slot, status);
1564 if (confirm_me)
1565 move_to_confirmed(conf);
1496out: 1566out:
1497 nfs4_unlock_state(); 1567 nfs4_unlock_state();
1498 dprintk("%s returns %d\n", __func__, ntohl(status)); 1568 dprintk("%s returns %d\n", __func__, ntohl(status));
@@ -1546,8 +1616,11 @@ nfsd4_destroy_session(struct svc_rqst *r,
1546 1616
1547 nfs4_lock_state(); 1617 nfs4_lock_state();
1548 /* wait for callbacks */ 1618 /* wait for callbacks */
1549 nfsd4_set_callback_client(ses->se_client, NULL); 1619 nfsd4_shutdown_callback(ses->se_client);
1550 nfs4_unlock_state(); 1620 nfs4_unlock_state();
1621
1622 nfsd4_del_conns(ses);
1623
1551 nfsd4_put_session(ses); 1624 nfsd4_put_session(ses);
1552 status = nfs_ok; 1625 status = nfs_ok;
1553out: 1626out:
@@ -1555,6 +1628,36 @@ out:
1555 return status; 1628 return status;
1556} 1629}
1557 1630
1631static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
1632{
1633 struct nfsd4_conn *c;
1634
1635 list_for_each_entry(c, &s->se_conns, cn_persession) {
1636 if (c->cn_xprt == xpt) {
1637 return c;
1638 }
1639 }
1640 return NULL;
1641}
1642
1643static void nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
1644{
1645 struct nfs4_client *clp = ses->se_client;
1646 struct nfsd4_conn *c;
1647
1648 spin_lock(&clp->cl_lock);
1649 c = __nfsd4_find_conn(new->cn_xprt, ses);
1650 if (c) {
1651 spin_unlock(&clp->cl_lock);
1652 free_conn(new);
1653 return;
1654 }
1655 __nfsd4_hash_conn(new, ses);
1656 spin_unlock(&clp->cl_lock);
1657 nfsd4_register_conn(new);
1658 return;
1659}
1660
1558__be32 1661__be32
1559nfsd4_sequence(struct svc_rqst *rqstp, 1662nfsd4_sequence(struct svc_rqst *rqstp,
1560 struct nfsd4_compound_state *cstate, 1663 struct nfsd4_compound_state *cstate,
@@ -1563,11 +1666,20 @@ nfsd4_sequence(struct svc_rqst *rqstp,
1563 struct nfsd4_compoundres *resp = rqstp->rq_resp; 1666 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1564 struct nfsd4_session *session; 1667 struct nfsd4_session *session;
1565 struct nfsd4_slot *slot; 1668 struct nfsd4_slot *slot;
1669 struct nfsd4_conn *conn;
1566 int status; 1670 int status;
1567 1671
1568 if (resp->opcnt != 1) 1672 if (resp->opcnt != 1)
1569 return nfserr_sequence_pos; 1673 return nfserr_sequence_pos;
1570 1674
1675 /*
1676 * Will be either used or freed by nfsd4_sequence_check_conn
1677 * below.
1678 */
1679 conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
1680 if (!conn)
1681 return nfserr_jukebox;
1682
1571 spin_lock(&client_lock); 1683 spin_lock(&client_lock);
1572 status = nfserr_badsession; 1684 status = nfserr_badsession;
1573 session = find_in_sessionid_hashtbl(&seq->sessionid); 1685 session = find_in_sessionid_hashtbl(&seq->sessionid);
@@ -1599,6 +1711,9 @@ nfsd4_sequence(struct svc_rqst *rqstp,
1599 if (status) 1711 if (status)
1600 goto out; 1712 goto out;
1601 1713
1714 nfsd4_sequence_check_conn(conn, session);
1715 conn = NULL;
1716
1602 /* Success! bump slot seqid */ 1717 /* Success! bump slot seqid */
1603 slot->sl_inuse = true; 1718 slot->sl_inuse = true;
1604 slot->sl_seqid = seq->seqid; 1719 slot->sl_seqid = seq->seqid;
@@ -1613,6 +1728,7 @@ out:
1613 nfsd4_get_session(cstate->session); 1728 nfsd4_get_session(cstate->session);
1614 atomic_inc(&session->se_client->cl_refcount); 1729 atomic_inc(&session->se_client->cl_refcount);
1615 } 1730 }
1731 kfree(conn);
1616 spin_unlock(&client_lock); 1732 spin_unlock(&client_lock);
1617 dprintk("%s: return %d\n", __func__, ntohl(status)); 1733 dprintk("%s: return %d\n", __func__, ntohl(status));
1618 return status; 1734 return status;
@@ -1747,6 +1863,11 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1747 goto out; 1863 goto out;
1748 gen_clid(new); 1864 gen_clid(new);
1749 } 1865 }
1866 /*
1867 * XXX: we should probably set this at creation time, and check
1868 * for consistent minorversion use throughout:
1869 */
1870 new->cl_minorversion = 0;
1750 gen_callback(new, setclid, rpc_get_scope_id(sa)); 1871 gen_callback(new, setclid, rpc_get_scope_id(sa));
1751 add_to_unconfirmed(new, strhashval); 1872 add_to_unconfirmed(new, strhashval);
1752 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot; 1873 setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
@@ -1807,7 +1928,8 @@ nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1807 status = nfserr_clid_inuse; 1928 status = nfserr_clid_inuse;
1808 else { 1929 else {
1809 atomic_set(&conf->cl_cb_set, 0); 1930 atomic_set(&conf->cl_cb_set, 0);
1810 nfsd4_probe_callback(conf, &unconf->cl_cb_conn); 1931 nfsd4_change_callback(conf, &unconf->cl_cb_conn);
1932 nfsd4_probe_callback(conf);
1811 expire_client(unconf); 1933 expire_client(unconf);
1812 status = nfs_ok; 1934 status = nfs_ok;
1813 1935
@@ -1841,7 +1963,7 @@ nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
1841 } 1963 }
1842 move_to_confirmed(unconf); 1964 move_to_confirmed(unconf);
1843 conf = unconf; 1965 conf = unconf;
1844 nfsd4_probe_callback(conf, &conf->cl_cb_conn); 1966 nfsd4_probe_callback(conf);
1845 status = nfs_ok; 1967 status = nfs_ok;
1846 } 1968 }
1847 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm))) 1969 } else if ((!conf || (conf && !same_verf(&conf->cl_confirm, &confirm)))
@@ -2944,7 +3066,11 @@ nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2944 if (STALE_STATEID(stateid)) 3066 if (STALE_STATEID(stateid))
2945 goto out; 3067 goto out;
2946 3068
2947 status = nfserr_bad_stateid; 3069 /*
3070 * We assume that any stateid that has the current boot time,
3071 * but that we can't find, is expired:
3072 */
3073 status = nfserr_expired;
2948 if (is_delegation_stateid(stateid)) { 3074 if (is_delegation_stateid(stateid)) {
2949 dp = find_delegation_stateid(ino, stateid); 3075 dp = find_delegation_stateid(ino, stateid);
2950 if (!dp) 3076 if (!dp)
@@ -2964,6 +3090,7 @@ nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
2964 stp = find_stateid(stateid, flags); 3090 stp = find_stateid(stateid, flags);
2965 if (!stp) 3091 if (!stp)
2966 goto out; 3092 goto out;
3093 status = nfserr_bad_stateid;
2967 if (nfs4_check_fh(current_fh, stp)) 3094 if (nfs4_check_fh(current_fh, stp))
2968 goto out; 3095 goto out;
2969 if (!stp->st_stateowner->so_confirmed) 3096 if (!stp->st_stateowner->so_confirmed)
@@ -3038,8 +3165,9 @@ nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
3038 * a replayed close: 3165 * a replayed close:
3039 */ 3166 */
3040 sop = search_close_lru(stateid->si_stateownerid, flags); 3167 sop = search_close_lru(stateid->si_stateownerid, flags);
3168 /* It's not stale; let's assume it's expired: */
3041 if (sop == NULL) 3169 if (sop == NULL)
3042 return nfserr_bad_stateid; 3170 return nfserr_expired;
3043 *sopp = sop; 3171 *sopp = sop;
3044 goto check_replay; 3172 goto check_replay;
3045 } 3173 }
@@ -3304,6 +3432,7 @@ nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3304 status = nfserr_bad_stateid; 3432 status = nfserr_bad_stateid;
3305 if (!is_delegation_stateid(stateid)) 3433 if (!is_delegation_stateid(stateid))
3306 goto out; 3434 goto out;
3435 status = nfserr_expired;
3307 dp = find_delegation_stateid(inode, stateid); 3436 dp = find_delegation_stateid(inode, stateid);
3308 if (!dp) 3437 if (!dp)
3309 goto out; 3438 goto out;
diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index 1a468bbd330f..f35a94a04026 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -1805,19 +1805,23 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
1805 goto out_nfserr; 1805 goto out_nfserr;
1806 } 1806 }
1807 } 1807 }
1808 if ((buflen -= 16) < 0)
1809 goto out_resource;
1810 1808
1811 if (unlikely(bmval2)) { 1809 if (bmval2) {
1810 if ((buflen -= 16) < 0)
1811 goto out_resource;
1812 WRITE32(3); 1812 WRITE32(3);
1813 WRITE32(bmval0); 1813 WRITE32(bmval0);
1814 WRITE32(bmval1); 1814 WRITE32(bmval1);
1815 WRITE32(bmval2); 1815 WRITE32(bmval2);
1816 } else if (likely(bmval1)) { 1816 } else if (bmval1) {
1817 if ((buflen -= 12) < 0)
1818 goto out_resource;
1817 WRITE32(2); 1819 WRITE32(2);
1818 WRITE32(bmval0); 1820 WRITE32(bmval0);
1819 WRITE32(bmval1); 1821 WRITE32(bmval1);
1820 } else { 1822 } else {
1823 if ((buflen -= 8) < 0)
1824 goto out_resource;
1821 WRITE32(1); 1825 WRITE32(1);
1822 WRITE32(bmval0); 1826 WRITE32(bmval0);
1823 } 1827 }
@@ -1828,15 +1832,17 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
1828 u32 word1 = nfsd_suppattrs1(minorversion); 1832 u32 word1 = nfsd_suppattrs1(minorversion);
1829 u32 word2 = nfsd_suppattrs2(minorversion); 1833 u32 word2 = nfsd_suppattrs2(minorversion);
1830 1834
1831 if ((buflen -= 12) < 0)
1832 goto out_resource;
1833 if (!aclsupport) 1835 if (!aclsupport)
1834 word0 &= ~FATTR4_WORD0_ACL; 1836 word0 &= ~FATTR4_WORD0_ACL;
1835 if (!word2) { 1837 if (!word2) {
1838 if ((buflen -= 12) < 0)
1839 goto out_resource;
1836 WRITE32(2); 1840 WRITE32(2);
1837 WRITE32(word0); 1841 WRITE32(word0);
1838 WRITE32(word1); 1842 WRITE32(word1);
1839 } else { 1843 } else {
1844 if ((buflen -= 16) < 0)
1845 goto out_resource;
1840 WRITE32(3); 1846 WRITE32(3);
1841 WRITE32(word0); 1847 WRITE32(word0);
1842 WRITE32(word1); 1848 WRITE32(word1);
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index 06fa87e52e82..d6dc3f61f8ba 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -22,6 +22,7 @@
22 */ 22 */
23enum { 23enum {
24 NFSD_Root = 1, 24 NFSD_Root = 1,
25#ifdef CONFIG_NFSD_DEPRECATED
25 NFSD_Svc, 26 NFSD_Svc,
26 NFSD_Add, 27 NFSD_Add,
27 NFSD_Del, 28 NFSD_Del,
@@ -29,6 +30,7 @@ enum {
29 NFSD_Unexport, 30 NFSD_Unexport,
30 NFSD_Getfd, 31 NFSD_Getfd,
31 NFSD_Getfs, 32 NFSD_Getfs,
33#endif
32 NFSD_List, 34 NFSD_List,
33 NFSD_Export_features, 35 NFSD_Export_features,
34 NFSD_Fh, 36 NFSD_Fh,
@@ -54,6 +56,7 @@ enum {
54/* 56/*
55 * write() for these nodes. 57 * write() for these nodes.
56 */ 58 */
59#ifdef CONFIG_NFSD_DEPRECATED
57static ssize_t write_svc(struct file *file, char *buf, size_t size); 60static ssize_t write_svc(struct file *file, char *buf, size_t size);
58static ssize_t write_add(struct file *file, char *buf, size_t size); 61static ssize_t write_add(struct file *file, char *buf, size_t size);
59static ssize_t write_del(struct file *file, char *buf, size_t size); 62static ssize_t write_del(struct file *file, char *buf, size_t size);
@@ -61,6 +64,7 @@ static ssize_t write_export(struct file *file, char *buf, size_t size);
61static ssize_t write_unexport(struct file *file, char *buf, size_t size); 64static ssize_t write_unexport(struct file *file, char *buf, size_t size);
62static ssize_t write_getfd(struct file *file, char *buf, size_t size); 65static ssize_t write_getfd(struct file *file, char *buf, size_t size);
63static ssize_t write_getfs(struct file *file, char *buf, size_t size); 66static ssize_t write_getfs(struct file *file, char *buf, size_t size);
67#endif
64static ssize_t write_filehandle(struct file *file, char *buf, size_t size); 68static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
65static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size); 69static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
66static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size); 70static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
@@ -76,6 +80,7 @@ static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
76#endif 80#endif
77 81
78static ssize_t (*write_op[])(struct file *, char *, size_t) = { 82static ssize_t (*write_op[])(struct file *, char *, size_t) = {
83#ifdef CONFIG_NFSD_DEPRECATED
79 [NFSD_Svc] = write_svc, 84 [NFSD_Svc] = write_svc,
80 [NFSD_Add] = write_add, 85 [NFSD_Add] = write_add,
81 [NFSD_Del] = write_del, 86 [NFSD_Del] = write_del,
@@ -83,6 +88,7 @@ static ssize_t (*write_op[])(struct file *, char *, size_t) = {
83 [NFSD_Unexport] = write_unexport, 88 [NFSD_Unexport] = write_unexport,
84 [NFSD_Getfd] = write_getfd, 89 [NFSD_Getfd] = write_getfd,
85 [NFSD_Getfs] = write_getfs, 90 [NFSD_Getfs] = write_getfs,
91#endif
86 [NFSD_Fh] = write_filehandle, 92 [NFSD_Fh] = write_filehandle,
87 [NFSD_FO_UnlockIP] = write_unlock_ip, 93 [NFSD_FO_UnlockIP] = write_unlock_ip,
88 [NFSD_FO_UnlockFS] = write_unlock_fs, 94 [NFSD_FO_UnlockFS] = write_unlock_fs,
@@ -121,6 +127,14 @@ static ssize_t nfsctl_transaction_write(struct file *file, const char __user *bu
121 127
122static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) 128static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
123{ 129{
130 static int warned;
131 if (file->f_dentry->d_name.name[0] == '.' && !warned) {
132 printk(KERN_INFO
133 "Warning: \"%s\" uses deprecated NFSD interface: %s."
134 " This will be removed in 2.6.40\n",
135 current->comm, file->f_dentry->d_name.name);
136 warned = 1;
137 }
124 if (! file->private_data) { 138 if (! file->private_data) {
125 /* An attempt to read a transaction file without writing 139 /* An attempt to read a transaction file without writing
126 * causes a 0-byte write so that the file can return 140 * causes a 0-byte write so that the file can return
@@ -187,6 +201,7 @@ static const struct file_operations pool_stats_operations = {
187 * payload - write methods 201 * payload - write methods
188 */ 202 */
189 203
204#ifdef CONFIG_NFSD_DEPRECATED
190/** 205/**
191 * write_svc - Start kernel's NFSD server 206 * write_svc - Start kernel's NFSD server
192 * 207 *
@@ -402,7 +417,7 @@ static ssize_t write_getfs(struct file *file, char *buf, size_t size)
402 417
403 ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, &in6); 418 ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, &in6);
404 419
405 clp = auth_unix_lookup(&in6); 420 clp = auth_unix_lookup(&init_net, &in6);
406 if (!clp) 421 if (!clp)
407 err = -EPERM; 422 err = -EPERM;
408 else { 423 else {
@@ -465,7 +480,7 @@ static ssize_t write_getfd(struct file *file, char *buf, size_t size)
465 480
466 ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, &in6); 481 ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, &in6);
467 482
468 clp = auth_unix_lookup(&in6); 483 clp = auth_unix_lookup(&init_net, &in6);
469 if (!clp) 484 if (!clp)
470 err = -EPERM; 485 err = -EPERM;
471 else { 486 else {
@@ -482,6 +497,7 @@ static ssize_t write_getfd(struct file *file, char *buf, size_t size)
482 out: 497 out:
483 return err; 498 return err;
484} 499}
500#endif /* CONFIG_NFSD_DEPRECATED */
485 501
486/** 502/**
487 * write_unlock_ip - Release all locks used by a client 503 * write_unlock_ip - Release all locks used by a client
@@ -1000,12 +1016,12 @@ static ssize_t __write_ports_addxprt(char *buf)
1000 if (err != 0) 1016 if (err != 0)
1001 return err; 1017 return err;
1002 1018
1003 err = svc_create_xprt(nfsd_serv, transport, 1019 err = svc_create_xprt(nfsd_serv, transport, &init_net,
1004 PF_INET, port, SVC_SOCK_ANONYMOUS); 1020 PF_INET, port, SVC_SOCK_ANONYMOUS);
1005 if (err < 0) 1021 if (err < 0)
1006 goto out_err; 1022 goto out_err;
1007 1023
1008 err = svc_create_xprt(nfsd_serv, transport, 1024 err = svc_create_xprt(nfsd_serv, transport, &init_net,
1009 PF_INET6, port, SVC_SOCK_ANONYMOUS); 1025 PF_INET6, port, SVC_SOCK_ANONYMOUS);
1010 if (err < 0 && err != -EAFNOSUPPORT) 1026 if (err < 0 && err != -EAFNOSUPPORT)
1011 goto out_close; 1027 goto out_close;
@@ -1356,6 +1372,7 @@ static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1356static int nfsd_fill_super(struct super_block * sb, void * data, int silent) 1372static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1357{ 1373{
1358 static struct tree_descr nfsd_files[] = { 1374 static struct tree_descr nfsd_files[] = {
1375#ifdef CONFIG_NFSD_DEPRECATED
1359 [NFSD_Svc] = {".svc", &transaction_ops, S_IWUSR}, 1376 [NFSD_Svc] = {".svc", &transaction_ops, S_IWUSR},
1360 [NFSD_Add] = {".add", &transaction_ops, S_IWUSR}, 1377 [NFSD_Add] = {".add", &transaction_ops, S_IWUSR},
1361 [NFSD_Del] = {".del", &transaction_ops, S_IWUSR}, 1378 [NFSD_Del] = {".del", &transaction_ops, S_IWUSR},
@@ -1363,6 +1380,7 @@ static int nfsd_fill_super(struct super_block * sb, void * data, int silent)
1363 [NFSD_Unexport] = {".unexport", &transaction_ops, S_IWUSR}, 1380 [NFSD_Unexport] = {".unexport", &transaction_ops, S_IWUSR},
1364 [NFSD_Getfd] = {".getfd", &transaction_ops, S_IWUSR|S_IRUSR}, 1381 [NFSD_Getfd] = {".getfd", &transaction_ops, S_IWUSR|S_IRUSR},
1365 [NFSD_Getfs] = {".getfs", &transaction_ops, S_IWUSR|S_IRUSR}, 1382 [NFSD_Getfs] = {".getfs", &transaction_ops, S_IWUSR|S_IRUSR},
1383#endif
1366 [NFSD_List] = {"exports", &exports_operations, S_IRUGO}, 1384 [NFSD_List] = {"exports", &exports_operations, S_IRUGO},
1367 [NFSD_Export_features] = {"export_features", 1385 [NFSD_Export_features] = {"export_features",
1368 &export_features_operations, S_IRUGO}, 1386 &export_features_operations, S_IRUGO},
diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h
index b76ac3a82e39..6b641cf2c19a 100644
--- a/fs/nfsd/nfsd.h
+++ b/fs/nfsd/nfsd.h
@@ -249,7 +249,7 @@ extern time_t nfsd4_grace;
249#define COMPOUND_SLACK_SPACE 140 /* OP_GETFH */ 249#define COMPOUND_SLACK_SPACE 140 /* OP_GETFH */
250#define COMPOUND_ERR_SLACK_SPACE 12 /* OP_SETATTR */ 250#define COMPOUND_ERR_SLACK_SPACE 12 /* OP_SETATTR */
251 251
252#define NFSD_LAUNDROMAT_MINTIMEOUT 10 /* seconds */ 252#define NFSD_LAUNDROMAT_MINTIMEOUT 1 /* seconds */
253 253
254/* 254/*
255 * The following attributes are currently not supported by the NFSv4 server: 255 * The following attributes are currently not supported by the NFSv4 server:
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index e2c43464f237..2bae1d86f5f2 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -16,6 +16,7 @@
16#include <linux/lockd/bind.h> 16#include <linux/lockd/bind.h>
17#include <linux/nfsacl.h> 17#include <linux/nfsacl.h>
18#include <linux/seq_file.h> 18#include <linux/seq_file.h>
19#include <net/net_namespace.h>
19#include "nfsd.h" 20#include "nfsd.h"
20#include "cache.h" 21#include "cache.h"
21#include "vfs.h" 22#include "vfs.h"
@@ -186,12 +187,12 @@ static int nfsd_init_socks(int port)
186 if (!list_empty(&nfsd_serv->sv_permsocks)) 187 if (!list_empty(&nfsd_serv->sv_permsocks))
187 return 0; 188 return 0;
188 189
189 error = svc_create_xprt(nfsd_serv, "udp", PF_INET, port, 190 error = svc_create_xprt(nfsd_serv, "udp", &init_net, PF_INET, port,
190 SVC_SOCK_DEFAULTS); 191 SVC_SOCK_DEFAULTS);
191 if (error < 0) 192 if (error < 0)
192 return error; 193 return error;
193 194
194 error = svc_create_xprt(nfsd_serv, "tcp", PF_INET, port, 195 error = svc_create_xprt(nfsd_serv, "tcp", &init_net, PF_INET, port,
195 SVC_SOCK_DEFAULTS); 196 SVC_SOCK_DEFAULTS);
196 if (error < 0) 197 if (error < 0)
197 return error; 198 return error;
diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
index 322518c88e4b..39adc27b0685 100644
--- a/fs/nfsd/state.h
+++ b/fs/nfsd/state.h
@@ -35,6 +35,7 @@
35#ifndef _NFSD4_STATE_H 35#ifndef _NFSD4_STATE_H
36#define _NFSD4_STATE_H 36#define _NFSD4_STATE_H
37 37
38#include <linux/sunrpc/svc_xprt.h>
38#include <linux/nfsd/nfsfh.h> 39#include <linux/nfsd/nfsfh.h>
39#include "nfsfh.h" 40#include "nfsfh.h"
40 41
@@ -64,19 +65,12 @@ typedef struct {
64 (s)->si_fileid, \ 65 (s)->si_fileid, \
65 (s)->si_generation 66 (s)->si_generation
66 67
67struct nfsd4_cb_sequence {
68 /* args/res */
69 u32 cbs_minorversion;
70 struct nfs4_client *cbs_clp;
71};
72
73struct nfs4_rpc_args {
74 void *args_op;
75 struct nfsd4_cb_sequence args_seq;
76};
77
78struct nfsd4_callback { 68struct nfsd4_callback {
79 struct nfs4_rpc_args cb_args; 69 void *cb_op;
70 struct nfs4_client *cb_clp;
71 u32 cb_minorversion;
72 struct rpc_message cb_msg;
73 const struct rpc_call_ops *cb_ops;
80 struct work_struct cb_work; 74 struct work_struct cb_work;
81}; 75};
82 76
@@ -91,7 +85,6 @@ struct nfs4_delegation {
91 u32 dl_type; 85 u32 dl_type;
92 time_t dl_time; 86 time_t dl_time;
93/* For recall: */ 87/* For recall: */
94 u32 dl_ident;
95 stateid_t dl_stateid; 88 stateid_t dl_stateid;
96 struct knfsd_fh dl_fh; 89 struct knfsd_fh dl_fh;
97 int dl_retries; 90 int dl_retries;
@@ -103,8 +96,8 @@ struct nfs4_cb_conn {
103 /* SETCLIENTID info */ 96 /* SETCLIENTID info */
104 struct sockaddr_storage cb_addr; 97 struct sockaddr_storage cb_addr;
105 size_t cb_addrlen; 98 size_t cb_addrlen;
106 u32 cb_prog; 99 u32 cb_prog; /* used only in 4.0 case;
107 u32 cb_minorversion; 100 per-session otherwise */
108 u32 cb_ident; /* minorversion 0 only */ 101 u32 cb_ident; /* minorversion 0 only */
109 struct svc_xprt *cb_xprt; /* minorversion 1 only */ 102 struct svc_xprt *cb_xprt; /* minorversion 1 only */
110}; 103};
@@ -160,6 +153,15 @@ struct nfsd4_clid_slot {
160 struct nfsd4_create_session sl_cr_ses; 153 struct nfsd4_create_session sl_cr_ses;
161}; 154};
162 155
156struct nfsd4_conn {
157 struct list_head cn_persession;
158 struct svc_xprt *cn_xprt;
159 struct svc_xpt_user cn_xpt_user;
160 struct nfsd4_session *cn_session;
161/* CDFC4_FORE, CDFC4_BACK: */
162 unsigned char cn_flags;
163};
164
163struct nfsd4_session { 165struct nfsd4_session {
164 struct kref se_ref; 166 struct kref se_ref;
165 struct list_head se_hash; /* hash by sessionid */ 167 struct list_head se_hash; /* hash by sessionid */
@@ -169,6 +171,9 @@ struct nfsd4_session {
169 struct nfs4_sessionid se_sessionid; 171 struct nfs4_sessionid se_sessionid;
170 struct nfsd4_channel_attrs se_fchannel; 172 struct nfsd4_channel_attrs se_fchannel;
171 struct nfsd4_channel_attrs se_bchannel; 173 struct nfsd4_channel_attrs se_bchannel;
174 struct list_head se_conns;
175 u32 se_cb_prog;
176 u32 se_cb_seq_nr;
172 struct nfsd4_slot *se_slots[]; /* forward channel slots */ 177 struct nfsd4_slot *se_slots[]; /* forward channel slots */
173}; 178};
174 179
@@ -221,24 +226,32 @@ struct nfs4_client {
221 clientid_t cl_clientid; /* generated by server */ 226 clientid_t cl_clientid; /* generated by server */
222 nfs4_verifier cl_confirm; /* generated by server */ 227 nfs4_verifier cl_confirm; /* generated by server */
223 u32 cl_firststate; /* recovery dir creation */ 228 u32 cl_firststate; /* recovery dir creation */
229 u32 cl_minorversion;
224 230
225 /* for v4.0 and v4.1 callbacks: */ 231 /* for v4.0 and v4.1 callbacks: */
226 struct nfs4_cb_conn cl_cb_conn; 232 struct nfs4_cb_conn cl_cb_conn;
233#define NFSD4_CLIENT_CB_UPDATE 1
234#define NFSD4_CLIENT_KILL 2
235 unsigned long cl_cb_flags;
227 struct rpc_clnt *cl_cb_client; 236 struct rpc_clnt *cl_cb_client;
237 u32 cl_cb_ident;
228 atomic_t cl_cb_set; 238 atomic_t cl_cb_set;
239 struct nfsd4_callback cl_cb_null;
240 struct nfsd4_session *cl_cb_session;
241
242 /* for all client information that callback code might need: */
243 spinlock_t cl_lock;
229 244
230 /* for nfs41 */ 245 /* for nfs41 */
231 struct list_head cl_sessions; 246 struct list_head cl_sessions;
232 struct nfsd4_clid_slot cl_cs_slot; /* create_session slot */ 247 struct nfsd4_clid_slot cl_cs_slot; /* create_session slot */
233 u32 cl_exchange_flags; 248 u32 cl_exchange_flags;
234 struct nfs4_sessionid cl_sessionid;
235 /* number of rpc's in progress over an associated session: */ 249 /* number of rpc's in progress over an associated session: */
236 atomic_t cl_refcount; 250 atomic_t cl_refcount;
237 251
238 /* for nfs41 callbacks */ 252 /* for nfs41 callbacks */
239 /* We currently support a single back channel with a single slot */ 253 /* We currently support a single back channel with a single slot */
240 unsigned long cl_cb_slot_busy; 254 unsigned long cl_cb_slot_busy;
241 u32 cl_cb_seq_nr;
242 struct rpc_wait_queue cl_cb_waitq; /* backchannel callers may */ 255 struct rpc_wait_queue cl_cb_waitq; /* backchannel callers may */
243 /* wait here for slots */ 256 /* wait here for slots */
244}; 257};
@@ -440,12 +453,13 @@ extern int nfs4_in_grace(void);
440extern __be32 nfs4_check_open_reclaim(clientid_t *clid); 453extern __be32 nfs4_check_open_reclaim(clientid_t *clid);
441extern void nfs4_free_stateowner(struct kref *kref); 454extern void nfs4_free_stateowner(struct kref *kref);
442extern int set_callback_cred(void); 455extern int set_callback_cred(void);
443extern void nfsd4_probe_callback(struct nfs4_client *clp, struct nfs4_cb_conn *); 456extern void nfsd4_probe_callback(struct nfs4_client *clp);
457extern void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *);
444extern void nfsd4_do_callback_rpc(struct work_struct *); 458extern void nfsd4_do_callback_rpc(struct work_struct *);
445extern void nfsd4_cb_recall(struct nfs4_delegation *dp); 459extern void nfsd4_cb_recall(struct nfs4_delegation *dp);
446extern int nfsd4_create_callback_queue(void); 460extern int nfsd4_create_callback_queue(void);
447extern void nfsd4_destroy_callback_queue(void); 461extern void nfsd4_destroy_callback_queue(void);
448extern void nfsd4_set_callback_client(struct nfs4_client *, struct rpc_clnt *); 462extern void nfsd4_shutdown_callback(struct nfs4_client *);
449extern void nfs4_put_delegation(struct nfs4_delegation *dp); 463extern void nfs4_put_delegation(struct nfs4_delegation *dp);
450extern __be32 nfs4_make_rec_clidname(char *clidname, struct xdr_netobj *clname); 464extern __be32 nfs4_make_rec_clidname(char *clidname, struct xdr_netobj *clname);
451extern void nfsd4_init_recdir(char *recdir_name); 465extern void nfsd4_init_recdir(char *recdir_name);
diff --git a/include/linux/net.h b/include/linux/net.h
index dee0b11a8759..16faa130088c 100644
--- a/include/linux/net.h
+++ b/include/linux/net.h
@@ -229,6 +229,8 @@ enum {
229extern int sock_wake_async(struct socket *sk, int how, int band); 229extern int sock_wake_async(struct socket *sk, int how, int band);
230extern int sock_register(const struct net_proto_family *fam); 230extern int sock_register(const struct net_proto_family *fam);
231extern void sock_unregister(int family); 231extern void sock_unregister(int family);
232extern int __sock_create(struct net *net, int family, int type, int proto,
233 struct socket **res, int kern);
232extern int sock_create(int family, int type, int proto, 234extern int sock_create(int family, int type, int proto,
233 struct socket **res); 235 struct socket **res);
234extern int sock_create_kern(int family, int type, int proto, 236extern int sock_create_kern(int family, int type, int proto,
diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h
index a9683d6acaa4..4925b22219d2 100644
--- a/include/linux/nfs4.h
+++ b/include/linux/nfs4.h
@@ -63,6 +63,9 @@
63#define NFS4_SHARE_SIGNAL_DELEG_WHEN_RESRC_AVAIL 0x10000 63#define NFS4_SHARE_SIGNAL_DELEG_WHEN_RESRC_AVAIL 0x10000
64#define NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED 0x20000 64#define NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED 0x20000
65 65
66#define NFS4_CDFC4_FORE 0x1
67#define NFS4_CDFC4_BACK 0x2
68
66#define NFS4_SET_TO_SERVER_TIME 0 69#define NFS4_SET_TO_SERVER_TIME 0
67#define NFS4_SET_TO_CLIENT_TIME 1 70#define NFS4_SET_TO_CLIENT_TIME 1
68 71
diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h
index 5bbc447175dc..b2024757edd5 100644
--- a/include/linux/sunrpc/auth.h
+++ b/include/linux/sunrpc/auth.h
@@ -122,8 +122,8 @@ extern const struct rpc_authops authnull_ops;
122int __init rpc_init_authunix(void); 122int __init rpc_init_authunix(void);
123int __init rpc_init_generic_auth(void); 123int __init rpc_init_generic_auth(void);
124int __init rpcauth_init_module(void); 124int __init rpcauth_init_module(void);
125void __exit rpcauth_remove_module(void); 125void rpcauth_remove_module(void);
126void __exit rpc_destroy_generic_auth(void); 126void rpc_destroy_generic_auth(void);
127void rpc_destroy_authunix(void); 127void rpc_destroy_authunix(void);
128 128
129struct rpc_cred * rpc_lookup_cred(void); 129struct rpc_cred * rpc_lookup_cred(void);
diff --git a/include/linux/sunrpc/cache.h b/include/linux/sunrpc/cache.h
index 7bf3e84b92f4..6950c981882d 100644
--- a/include/linux/sunrpc/cache.h
+++ b/include/linux/sunrpc/cache.h
@@ -125,12 +125,15 @@ struct cache_detail {
125 */ 125 */
126struct cache_req { 126struct cache_req {
127 struct cache_deferred_req *(*defer)(struct cache_req *req); 127 struct cache_deferred_req *(*defer)(struct cache_req *req);
128 int thread_wait; /* How long (jiffies) we can block the
129 * current thread to wait for updates.
130 */
128}; 131};
129/* this must be embedded in a deferred_request that is being 132/* this must be embedded in a deferred_request that is being
130 * delayed awaiting cache-fill 133 * delayed awaiting cache-fill
131 */ 134 */
132struct cache_deferred_req { 135struct cache_deferred_req {
133 struct list_head hash; /* on hash chain */ 136 struct hlist_node hash; /* on hash chain */
134 struct list_head recent; /* on fifo */ 137 struct list_head recent; /* on fifo */
135 struct cache_head *item; /* cache item we wait on */ 138 struct cache_head *item; /* cache item we wait on */
136 void *owner; /* we might need to discard all defered requests 139 void *owner; /* we might need to discard all defered requests
@@ -194,7 +197,9 @@ extern void cache_purge(struct cache_detail *detail);
194#define NEVER (0x7FFFFFFF) 197#define NEVER (0x7FFFFFFF)
195extern void __init cache_initialize(void); 198extern void __init cache_initialize(void);
196extern int cache_register(struct cache_detail *cd); 199extern int cache_register(struct cache_detail *cd);
200extern int cache_register_net(struct cache_detail *cd, struct net *net);
197extern void cache_unregister(struct cache_detail *cd); 201extern void cache_unregister(struct cache_detail *cd);
202extern void cache_unregister_net(struct cache_detail *cd, struct net *net);
198 203
199extern int sunrpc_cache_register_pipefs(struct dentry *parent, const char *, 204extern int sunrpc_cache_register_pipefs(struct dentry *parent, const char *,
200 mode_t, struct cache_detail *); 205 mode_t, struct cache_detail *);
@@ -218,14 +223,42 @@ static inline int get_int(char **bpp, int *anint)
218 return 0; 223 return 0;
219} 224}
220 225
226/*
227 * timestamps kept in the cache are expressed in seconds
228 * since boot. This is the best for measuring differences in
229 * real time.
230 */
231static inline time_t seconds_since_boot(void)
232{
233 struct timespec boot;
234 getboottime(&boot);
235 return get_seconds() - boot.tv_sec;
236}
237
238static inline time_t convert_to_wallclock(time_t sinceboot)
239{
240 struct timespec boot;
241 getboottime(&boot);
242 return boot.tv_sec + sinceboot;
243}
244
221static inline time_t get_expiry(char **bpp) 245static inline time_t get_expiry(char **bpp)
222{ 246{
223 int rv; 247 int rv;
248 struct timespec boot;
249
224 if (get_int(bpp, &rv)) 250 if (get_int(bpp, &rv))
225 return 0; 251 return 0;
226 if (rv < 0) 252 if (rv < 0)
227 return 0; 253 return 0;
228 return rv; 254 getboottime(&boot);
255 return rv - boot.tv_sec;
229} 256}
230 257
258static inline void sunrpc_invalidate(struct cache_head *h,
259 struct cache_detail *detail)
260{
261 h->expiry_time = seconds_since_boot() - 1;
262 detail->nextcheck = seconds_since_boot();
263}
231#endif /* _LINUX_SUNRPC_CACHE_H_ */ 264#endif /* _LINUX_SUNRPC_CACHE_H_ */
diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index c83df09a8e2b..a5a55f284b7d 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -102,6 +102,7 @@ struct rpc_procinfo {
102#ifdef __KERNEL__ 102#ifdef __KERNEL__
103 103
104struct rpc_create_args { 104struct rpc_create_args {
105 struct net *net;
105 int protocol; 106 int protocol;
106 struct sockaddr *address; 107 struct sockaddr *address;
107 size_t addrsize; 108 size_t addrsize;
diff --git a/include/linux/sunrpc/gss_spkm3.h b/include/linux/sunrpc/gss_spkm3.h
deleted file mode 100644
index e3e6a3437f8b..000000000000
--- a/include/linux/sunrpc/gss_spkm3.h
+++ /dev/null
@@ -1,55 +0,0 @@
1/*
2 * linux/include/linux/sunrpc/gss_spkm3.h
3 *
4 * Copyright (c) 2000 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Andy Adamson <andros@umich.edu>
8 */
9
10#include <linux/sunrpc/auth_gss.h>
11#include <linux/sunrpc/gss_err.h>
12#include <linux/sunrpc/gss_asn1.h>
13
14struct spkm3_ctx {
15 struct xdr_netobj ctx_id; /* per message context id */
16 int endtime; /* endtime of the context */
17 struct xdr_netobj mech_used;
18 unsigned int ret_flags ;
19 struct xdr_netobj conf_alg;
20 struct xdr_netobj derived_conf_key;
21 struct xdr_netobj intg_alg;
22 struct xdr_netobj derived_integ_key;
23};
24
25/* OIDs declarations for K-ALG, I-ALG, C-ALG, and OWF-ALG */
26extern const struct xdr_netobj hmac_md5_oid;
27extern const struct xdr_netobj cast5_cbc_oid;
28
29/* SPKM InnerContext Token types */
30
31#define SPKM_ERROR_TOK 3
32#define SPKM_MIC_TOK 4
33#define SPKM_WRAP_TOK 5
34#define SPKM_DEL_TOK 6
35
36u32 spkm3_make_token(struct spkm3_ctx *ctx, struct xdr_buf * text, struct xdr_netobj * token, int toktype);
37
38u32 spkm3_read_token(struct spkm3_ctx *ctx, struct xdr_netobj *read_token, struct xdr_buf *message_buffer, int toktype);
39
40#define CKSUMTYPE_RSA_MD5 0x0007
41#define CKSUMTYPE_HMAC_MD5 0x0008
42
43s32 make_spkm3_checksum(s32 cksumtype, struct xdr_netobj *key, char *header,
44 unsigned int hdrlen, struct xdr_buf *body,
45 unsigned int body_offset, struct xdr_netobj *cksum);
46void asn1_bitstring_len(struct xdr_netobj *in, int *enclen, int *zerobits);
47int decode_asn1_bitstring(struct xdr_netobj *out, char *in, int enclen,
48 int explen);
49void spkm3_mic_header(unsigned char **hdrbuf, unsigned int *hdrlen,
50 unsigned char *ctxhdr, int elen, int zbit);
51void spkm3_make_mic_token(unsigned char **tokp, int toklen,
52 struct xdr_netobj *mic_hdr,
53 struct xdr_netobj *md5cksum, int md5elen, int md5zbit);
54u32 spkm3_verify_mic_token(unsigned char **tokp, int *mic_hdrlen,
55 unsigned char **cksum);
diff --git a/include/linux/sunrpc/stats.h b/include/linux/sunrpc/stats.h
index 5fa0f2084307..680471d1f28a 100644
--- a/include/linux/sunrpc/stats.h
+++ b/include/linux/sunrpc/stats.h
@@ -38,8 +38,21 @@ struct svc_stat {
38 rpcbadclnt; 38 rpcbadclnt;
39}; 39};
40 40
41void rpc_proc_init(void); 41struct net;
42void rpc_proc_exit(void); 42#ifdef CONFIG_PROC_FS
43int rpc_proc_init(struct net *);
44void rpc_proc_exit(struct net *);
45#else
46static inline int rpc_proc_init(struct net *net)
47{
48 return 0;
49}
50
51static inline void rpc_proc_exit(struct net *net)
52{
53}
54#endif
55
43#ifdef MODULE 56#ifdef MODULE
44void rpc_modcount(struct inode *, int); 57void rpc_modcount(struct inode *, int);
45#endif 58#endif
@@ -54,9 +67,6 @@ void svc_proc_unregister(const char *);
54 67
55void svc_seq_show(struct seq_file *, 68void svc_seq_show(struct seq_file *,
56 const struct svc_stat *); 69 const struct svc_stat *);
57
58extern struct proc_dir_entry *proc_net_rpc;
59
60#else 70#else
61 71
62static inline struct proc_dir_entry *rpc_proc_register(struct rpc_stat *s) { return NULL; } 72static inline struct proc_dir_entry *rpc_proc_register(struct rpc_stat *s) { return NULL; }
@@ -69,9 +79,6 @@ static inline void svc_proc_unregister(const char *p) {}
69 79
70static inline void svc_seq_show(struct seq_file *seq, 80static inline void svc_seq_show(struct seq_file *seq,
71 const struct svc_stat *st) {} 81 const struct svc_stat *st) {}
72
73#define proc_net_rpc NULL
74
75#endif 82#endif
76 83
77#endif /* _LINUX_SUNRPC_STATS_H */ 84#endif /* _LINUX_SUNRPC_STATS_H */
diff --git a/include/linux/sunrpc/svc_xprt.h b/include/linux/sunrpc/svc_xprt.h
index 5f4e18b3ce73..bbdb680ffbe9 100644
--- a/include/linux/sunrpc/svc_xprt.h
+++ b/include/linux/sunrpc/svc_xprt.h
@@ -12,6 +12,7 @@
12 12
13struct svc_xprt_ops { 13struct svc_xprt_ops {
14 struct svc_xprt *(*xpo_create)(struct svc_serv *, 14 struct svc_xprt *(*xpo_create)(struct svc_serv *,
15 struct net *net,
15 struct sockaddr *, int, 16 struct sockaddr *, int,
16 int); 17 int);
17 struct svc_xprt *(*xpo_accept)(struct svc_xprt *); 18 struct svc_xprt *(*xpo_accept)(struct svc_xprt *);
@@ -32,6 +33,16 @@ struct svc_xprt_class {
32 u32 xcl_max_payload; 33 u32 xcl_max_payload;
33}; 34};
34 35
36/*
37 * This is embedded in an object that wants a callback before deleting
38 * an xprt; intended for use by NFSv4.1, which needs to know when a
39 * client's tcp connection (and hence possibly a backchannel) goes away.
40 */
41struct svc_xpt_user {
42 struct list_head list;
43 void (*callback)(struct svc_xpt_user *);
44};
45
35struct svc_xprt { 46struct svc_xprt {
36 struct svc_xprt_class *xpt_class; 47 struct svc_xprt_class *xpt_class;
37 struct svc_xprt_ops *xpt_ops; 48 struct svc_xprt_ops *xpt_ops;
@@ -66,14 +77,31 @@ struct svc_xprt {
66 struct sockaddr_storage xpt_remote; /* remote peer's address */ 77 struct sockaddr_storage xpt_remote; /* remote peer's address */
67 size_t xpt_remotelen; /* length of address */ 78 size_t xpt_remotelen; /* length of address */
68 struct rpc_wait_queue xpt_bc_pending; /* backchannel wait queue */ 79 struct rpc_wait_queue xpt_bc_pending; /* backchannel wait queue */
80 struct list_head xpt_users; /* callbacks on free */
81
82 struct net *xpt_net;
69}; 83};
70 84
85static inline void register_xpt_user(struct svc_xprt *xpt, struct svc_xpt_user *u)
86{
87 spin_lock(&xpt->xpt_lock);
88 list_add(&u->list, &xpt->xpt_users);
89 spin_unlock(&xpt->xpt_lock);
90}
91
92static inline void unregister_xpt_user(struct svc_xprt *xpt, struct svc_xpt_user *u)
93{
94 spin_lock(&xpt->xpt_lock);
95 list_del_init(&u->list);
96 spin_unlock(&xpt->xpt_lock);
97}
98
71int svc_reg_xprt_class(struct svc_xprt_class *); 99int svc_reg_xprt_class(struct svc_xprt_class *);
72void svc_unreg_xprt_class(struct svc_xprt_class *); 100void svc_unreg_xprt_class(struct svc_xprt_class *);
73void svc_xprt_init(struct svc_xprt_class *, struct svc_xprt *, 101void svc_xprt_init(struct svc_xprt_class *, struct svc_xprt *,
74 struct svc_serv *); 102 struct svc_serv *);
75int svc_create_xprt(struct svc_serv *, const char *, const int, 103int svc_create_xprt(struct svc_serv *, const char *, struct net *,
76 const unsigned short, int); 104 const int, const unsigned short, int);
77void svc_xprt_enqueue(struct svc_xprt *xprt); 105void svc_xprt_enqueue(struct svc_xprt *xprt);
78void svc_xprt_received(struct svc_xprt *); 106void svc_xprt_received(struct svc_xprt *);
79void svc_xprt_put(struct svc_xprt *xprt); 107void svc_xprt_put(struct svc_xprt *xprt);
diff --git a/include/linux/sunrpc/svcauth.h b/include/linux/sunrpc/svcauth.h
index d39dbdc7b10f..25d333c1b571 100644
--- a/include/linux/sunrpc/svcauth.h
+++ b/include/linux/sunrpc/svcauth.h
@@ -108,10 +108,15 @@ struct auth_ops {
108#define SVC_NEGATIVE 4 108#define SVC_NEGATIVE 4
109#define SVC_OK 5 109#define SVC_OK 5
110#define SVC_DROP 6 110#define SVC_DROP 6
111#define SVC_DENIED 7 111#define SVC_CLOSE 7 /* Like SVC_DROP, but request is definitely
112#define SVC_PENDING 8 112 * lost so if there is a tcp connection, it
113#define SVC_COMPLETE 9 113 * should be closed
114 */
115#define SVC_DENIED 8
116#define SVC_PENDING 9
117#define SVC_COMPLETE 10
114 118
119struct svc_xprt;
115 120
116extern int svc_authenticate(struct svc_rqst *rqstp, __be32 *authp); 121extern int svc_authenticate(struct svc_rqst *rqstp, __be32 *authp);
117extern int svc_authorise(struct svc_rqst *rqstp); 122extern int svc_authorise(struct svc_rqst *rqstp);
@@ -121,13 +126,13 @@ extern void svc_auth_unregister(rpc_authflavor_t flavor);
121 126
122extern struct auth_domain *unix_domain_find(char *name); 127extern struct auth_domain *unix_domain_find(char *name);
123extern void auth_domain_put(struct auth_domain *item); 128extern void auth_domain_put(struct auth_domain *item);
124extern int auth_unix_add_addr(struct in6_addr *addr, struct auth_domain *dom); 129extern int auth_unix_add_addr(struct net *net, struct in6_addr *addr, struct auth_domain *dom);
125extern struct auth_domain *auth_domain_lookup(char *name, struct auth_domain *new); 130extern struct auth_domain *auth_domain_lookup(char *name, struct auth_domain *new);
126extern struct auth_domain *auth_domain_find(char *name); 131extern struct auth_domain *auth_domain_find(char *name);
127extern struct auth_domain *auth_unix_lookup(struct in6_addr *addr); 132extern struct auth_domain *auth_unix_lookup(struct net *net, struct in6_addr *addr);
128extern int auth_unix_forget_old(struct auth_domain *dom); 133extern int auth_unix_forget_old(struct auth_domain *dom);
129extern void svcauth_unix_purge(void); 134extern void svcauth_unix_purge(void);
130extern void svcauth_unix_info_release(void *); 135extern void svcauth_unix_info_release(struct svc_xprt *xpt);
131extern int svcauth_unix_set_client(struct svc_rqst *rqstp); 136extern int svcauth_unix_set_client(struct svc_rqst *rqstp);
132 137
133static inline unsigned long hash_str(char *name, int bits) 138static inline unsigned long hash_str(char *name, int bits)
diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index ff5a77b28c50..89d10d279a20 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -224,6 +224,7 @@ struct rpc_xprt {
224 bklog_u; /* backlog queue utilization */ 224 bklog_u; /* backlog queue utilization */
225 } stat; 225 } stat;
226 226
227 struct net *xprt_net;
227 const char *address_strings[RPC_DISPLAY_MAX]; 228 const char *address_strings[RPC_DISPLAY_MAX];
228}; 229};
229 230
@@ -249,6 +250,7 @@ static inline int bc_prealloc(struct rpc_rqst *req)
249 250
250struct xprt_create { 251struct xprt_create {
251 int ident; /* XPRT_TRANSPORT identifier */ 252 int ident; /* XPRT_TRANSPORT identifier */
253 struct net * net;
252 struct sockaddr * srcaddr; /* optional local address */ 254 struct sockaddr * srcaddr; /* optional local address */
253 struct sockaddr * dstaddr; /* remote peer address */ 255 struct sockaddr * dstaddr; /* remote peer address */
254 size_t addrlen; 256 size_t addrlen;
@@ -280,6 +282,8 @@ void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task);
280void xprt_release(struct rpc_task *task); 282void xprt_release(struct rpc_task *task);
281struct rpc_xprt * xprt_get(struct rpc_xprt *xprt); 283struct rpc_xprt * xprt_get(struct rpc_xprt *xprt);
282void xprt_put(struct rpc_xprt *xprt); 284void xprt_put(struct rpc_xprt *xprt);
285struct rpc_xprt * xprt_alloc(struct net *net, int size, int max_req);
286void xprt_free(struct rpc_xprt *);
283 287
284static inline __be32 *xprt_skip_transport_header(struct rpc_xprt *xprt, __be32 *p) 288static inline __be32 *xprt_skip_transport_header(struct rpc_xprt *xprt, __be32 *p)
285{ 289{
diff --git a/net/socket.c b/net/socket.c
index abf3e2561521..7f67c072d496 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -1145,7 +1145,7 @@ call_kill:
1145} 1145}
1146EXPORT_SYMBOL(sock_wake_async); 1146EXPORT_SYMBOL(sock_wake_async);
1147 1147
1148static int __sock_create(struct net *net, int family, int type, int protocol, 1148int __sock_create(struct net *net, int family, int type, int protocol,
1149 struct socket **res, int kern) 1149 struct socket **res, int kern)
1150{ 1150{
1151 int err; 1151 int err;
@@ -1257,6 +1257,7 @@ out_release:
1257 rcu_read_unlock(); 1257 rcu_read_unlock();
1258 goto out_sock_release; 1258 goto out_sock_release;
1259} 1259}
1260EXPORT_SYMBOL(__sock_create);
1260 1261
1261int sock_create(int family, int type, int protocol, struct socket **res) 1262int sock_create(int family, int type, int protocol, struct socket **res)
1262{ 1263{
diff --git a/net/sunrpc/Kconfig b/net/sunrpc/Kconfig
index 3376d7657185..8873fd8ddacd 100644
--- a/net/sunrpc/Kconfig
+++ b/net/sunrpc/Kconfig
@@ -36,22 +36,3 @@ config RPCSEC_GSS_KRB5
36 Kerberos support should be installed. 36 Kerberos support should be installed.
37 37
38 If unsure, say Y. 38 If unsure, say Y.
39
40config RPCSEC_GSS_SPKM3
41 tristate "Secure RPC: SPKM3 mechanism (EXPERIMENTAL)"
42 depends on SUNRPC && EXPERIMENTAL
43 select SUNRPC_GSS
44 select CRYPTO
45 select CRYPTO_MD5
46 select CRYPTO_DES
47 select CRYPTO_CAST5
48 select CRYPTO_CBC
49 help
50 Choose Y here to enable Secure RPC using the SPKM3 public key
51 GSS-API mechanism (RFC 2025).
52
53 Secure RPC calls with SPKM3 require an auxiliary userspace
54 daemon which may be found in the Linux nfs-utils package
55 available from http://linux-nfs.org/.
56
57 If unsure, say N.
diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c
index 68192e562749..afe67849269f 100644
--- a/net/sunrpc/auth.c
+++ b/net/sunrpc/auth.c
@@ -658,7 +658,7 @@ out1:
658 return err; 658 return err;
659} 659}
660 660
661void __exit rpcauth_remove_module(void) 661void rpcauth_remove_module(void)
662{ 662{
663 rpc_destroy_authunix(); 663 rpc_destroy_authunix();
664 rpc_destroy_generic_auth(); 664 rpc_destroy_generic_auth();
diff --git a/net/sunrpc/auth_generic.c b/net/sunrpc/auth_generic.c
index 43162bb3b78f..e010a015d996 100644
--- a/net/sunrpc/auth_generic.c
+++ b/net/sunrpc/auth_generic.c
@@ -158,7 +158,7 @@ int __init rpc_init_generic_auth(void)
158 return rpcauth_init_credcache(&generic_auth); 158 return rpcauth_init_credcache(&generic_auth);
159} 159}
160 160
161void __exit rpc_destroy_generic_auth(void) 161void rpc_destroy_generic_auth(void)
162{ 162{
163 rpcauth_destroy_credcache(&generic_auth); 163 rpcauth_destroy_credcache(&generic_auth);
164} 164}
diff --git a/net/sunrpc/auth_gss/Makefile b/net/sunrpc/auth_gss/Makefile
index 74a231735f67..7350d86a32ee 100644
--- a/net/sunrpc/auth_gss/Makefile
+++ b/net/sunrpc/auth_gss/Makefile
@@ -11,8 +11,3 @@ obj-$(CONFIG_RPCSEC_GSS_KRB5) += rpcsec_gss_krb5.o
11 11
12rpcsec_gss_krb5-objs := gss_krb5_mech.o gss_krb5_seal.o gss_krb5_unseal.o \ 12rpcsec_gss_krb5-objs := gss_krb5_mech.o gss_krb5_seal.o gss_krb5_unseal.o \
13 gss_krb5_seqnum.o gss_krb5_wrap.o gss_krb5_crypto.o gss_krb5_keys.o 13 gss_krb5_seqnum.o gss_krb5_wrap.o gss_krb5_crypto.o gss_krb5_keys.o
14
15obj-$(CONFIG_RPCSEC_GSS_SPKM3) += rpcsec_gss_spkm3.o
16
17rpcsec_gss_spkm3-objs := gss_spkm3_mech.o gss_spkm3_seal.o gss_spkm3_unseal.o \
18 gss_spkm3_token.o
diff --git a/net/sunrpc/auth_gss/gss_spkm3_mech.c b/net/sunrpc/auth_gss/gss_spkm3_mech.c
deleted file mode 100644
index adade3d313f2..000000000000
--- a/net/sunrpc/auth_gss/gss_spkm3_mech.c
+++ /dev/null
@@ -1,247 +0,0 @@
1/*
2 * linux/net/sunrpc/gss_spkm3_mech.c
3 *
4 * Copyright (c) 2003 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Andy Adamson <andros@umich.edu>
8 * J. Bruce Fields <bfields@umich.edu>
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 *
35 */
36
37#include <linux/err.h>
38#include <linux/module.h>
39#include <linux/init.h>
40#include <linux/types.h>
41#include <linux/slab.h>
42#include <linux/sunrpc/auth.h>
43#include <linux/in.h>
44#include <linux/sunrpc/svcauth_gss.h>
45#include <linux/sunrpc/gss_spkm3.h>
46#include <linux/sunrpc/xdr.h>
47#include <linux/crypto.h>
48
49#ifdef RPC_DEBUG
50# define RPCDBG_FACILITY RPCDBG_AUTH
51#endif
52
53static const void *
54simple_get_bytes(const void *p, const void *end, void *res, int len)
55{
56 const void *q = (const void *)((const char *)p + len);
57 if (unlikely(q > end || q < p))
58 return ERR_PTR(-EFAULT);
59 memcpy(res, p, len);
60 return q;
61}
62
63static const void *
64simple_get_netobj(const void *p, const void *end, struct xdr_netobj *res)
65{
66 const void *q;
67 unsigned int len;
68 p = simple_get_bytes(p, end, &len, sizeof(len));
69 if (IS_ERR(p))
70 return p;
71 res->len = len;
72 if (len == 0) {
73 res->data = NULL;
74 return p;
75 }
76 q = (const void *)((const char *)p + len);
77 if (unlikely(q > end || q < p))
78 return ERR_PTR(-EFAULT);
79 res->data = kmemdup(p, len, GFP_NOFS);
80 if (unlikely(res->data == NULL))
81 return ERR_PTR(-ENOMEM);
82 return q;
83}
84
85static int
86gss_import_sec_context_spkm3(const void *p, size_t len,
87 struct gss_ctx *ctx_id,
88 gfp_t gfp_mask)
89{
90 const void *end = (const void *)((const char *)p + len);
91 struct spkm3_ctx *ctx;
92 int version;
93
94 if (!(ctx = kzalloc(sizeof(*ctx), gfp_mask)))
95 goto out_err;
96
97 p = simple_get_bytes(p, end, &version, sizeof(version));
98 if (IS_ERR(p))
99 goto out_err_free_ctx;
100 if (version != 1) {
101 dprintk("RPC: unknown spkm3 token format: "
102 "obsolete nfs-utils?\n");
103 p = ERR_PTR(-EINVAL);
104 goto out_err_free_ctx;
105 }
106
107 p = simple_get_netobj(p, end, &ctx->ctx_id);
108 if (IS_ERR(p))
109 goto out_err_free_ctx;
110
111 p = simple_get_bytes(p, end, &ctx->endtime, sizeof(ctx->endtime));
112 if (IS_ERR(p))
113 goto out_err_free_ctx_id;
114
115 p = simple_get_netobj(p, end, &ctx->mech_used);
116 if (IS_ERR(p))
117 goto out_err_free_ctx_id;
118
119 p = simple_get_bytes(p, end, &ctx->ret_flags, sizeof(ctx->ret_flags));
120 if (IS_ERR(p))
121 goto out_err_free_mech;
122
123 p = simple_get_netobj(p, end, &ctx->conf_alg);
124 if (IS_ERR(p))
125 goto out_err_free_mech;
126
127 p = simple_get_netobj(p, end, &ctx->derived_conf_key);
128 if (IS_ERR(p))
129 goto out_err_free_conf_alg;
130
131 p = simple_get_netobj(p, end, &ctx->intg_alg);
132 if (IS_ERR(p))
133 goto out_err_free_conf_key;
134
135 p = simple_get_netobj(p, end, &ctx->derived_integ_key);
136 if (IS_ERR(p))
137 goto out_err_free_intg_alg;
138
139 if (p != end) {
140 p = ERR_PTR(-EFAULT);
141 goto out_err_free_intg_key;
142 }
143
144 ctx_id->internal_ctx_id = ctx;
145
146 dprintk("RPC: Successfully imported new spkm context.\n");
147 return 0;
148
149out_err_free_intg_key:
150 kfree(ctx->derived_integ_key.data);
151out_err_free_intg_alg:
152 kfree(ctx->intg_alg.data);
153out_err_free_conf_key:
154 kfree(ctx->derived_conf_key.data);
155out_err_free_conf_alg:
156 kfree(ctx->conf_alg.data);
157out_err_free_mech:
158 kfree(ctx->mech_used.data);
159out_err_free_ctx_id:
160 kfree(ctx->ctx_id.data);
161out_err_free_ctx:
162 kfree(ctx);
163out_err:
164 return PTR_ERR(p);
165}
166
167static void
168gss_delete_sec_context_spkm3(void *internal_ctx)
169{
170 struct spkm3_ctx *sctx = internal_ctx;
171
172 kfree(sctx->derived_integ_key.data);
173 kfree(sctx->intg_alg.data);
174 kfree(sctx->derived_conf_key.data);
175 kfree(sctx->conf_alg.data);
176 kfree(sctx->mech_used.data);
177 kfree(sctx->ctx_id.data);
178 kfree(sctx);
179}
180
181static u32
182gss_verify_mic_spkm3(struct gss_ctx *ctx,
183 struct xdr_buf *signbuf,
184 struct xdr_netobj *checksum)
185{
186 u32 maj_stat = 0;
187 struct spkm3_ctx *sctx = ctx->internal_ctx_id;
188
189 maj_stat = spkm3_read_token(sctx, checksum, signbuf, SPKM_MIC_TOK);
190
191 dprintk("RPC: gss_verify_mic_spkm3 returning %d\n", maj_stat);
192 return maj_stat;
193}
194
195static u32
196gss_get_mic_spkm3(struct gss_ctx *ctx,
197 struct xdr_buf *message_buffer,
198 struct xdr_netobj *message_token)
199{
200 u32 err = 0;
201 struct spkm3_ctx *sctx = ctx->internal_ctx_id;
202
203 err = spkm3_make_token(sctx, message_buffer,
204 message_token, SPKM_MIC_TOK);
205 dprintk("RPC: gss_get_mic_spkm3 returning %d\n", err);
206 return err;
207}
208
209static const struct gss_api_ops gss_spkm3_ops = {
210 .gss_import_sec_context = gss_import_sec_context_spkm3,
211 .gss_get_mic = gss_get_mic_spkm3,
212 .gss_verify_mic = gss_verify_mic_spkm3,
213 .gss_delete_sec_context = gss_delete_sec_context_spkm3,
214};
215
216static struct pf_desc gss_spkm3_pfs[] = {
217 {RPC_AUTH_GSS_SPKM, RPC_GSS_SVC_NONE, "spkm3"},
218 {RPC_AUTH_GSS_SPKMI, RPC_GSS_SVC_INTEGRITY, "spkm3i"},
219};
220
221static struct gss_api_mech gss_spkm3_mech = {
222 .gm_name = "spkm3",
223 .gm_owner = THIS_MODULE,
224 .gm_oid = {7, "\053\006\001\005\005\001\003"},
225 .gm_ops = &gss_spkm3_ops,
226 .gm_pf_num = ARRAY_SIZE(gss_spkm3_pfs),
227 .gm_pfs = gss_spkm3_pfs,
228};
229
230static int __init init_spkm3_module(void)
231{
232 int status;
233
234 status = gss_mech_register(&gss_spkm3_mech);
235 if (status)
236 printk("Failed to register spkm3 gss mechanism!\n");
237 return status;
238}
239
240static void __exit cleanup_spkm3_module(void)
241{
242 gss_mech_unregister(&gss_spkm3_mech);
243}
244
245MODULE_LICENSE("GPL");
246module_init(init_spkm3_module);
247module_exit(cleanup_spkm3_module);
diff --git a/net/sunrpc/auth_gss/gss_spkm3_seal.c b/net/sunrpc/auth_gss/gss_spkm3_seal.c
deleted file mode 100644
index 5a3a65a0e2b4..000000000000
--- a/net/sunrpc/auth_gss/gss_spkm3_seal.c
+++ /dev/null
@@ -1,186 +0,0 @@
1/*
2 * linux/net/sunrpc/gss_spkm3_seal.c
3 *
4 * Copyright (c) 2003 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Andy Adamson <andros@umich.edu>
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
23 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
24 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
25 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
29 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
30 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
31 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
32 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 *
34 */
35
36#include <linux/types.h>
37#include <linux/jiffies.h>
38#include <linux/sunrpc/gss_spkm3.h>
39#include <linux/random.h>
40#include <linux/crypto.h>
41#include <linux/pagemap.h>
42#include <linux/scatterlist.h>
43#include <linux/sunrpc/xdr.h>
44
45#ifdef RPC_DEBUG
46# define RPCDBG_FACILITY RPCDBG_AUTH
47#endif
48
49const struct xdr_netobj hmac_md5_oid = { 8, "\x2B\x06\x01\x05\x05\x08\x01\x01"};
50const struct xdr_netobj cast5_cbc_oid = {9, "\x2A\x86\x48\x86\xF6\x7D\x07\x42\x0A"};
51
52/*
53 * spkm3_make_token()
54 *
55 * Only SPKM_MIC_TOK with md5 intg-alg is supported
56 */
57
58u32
59spkm3_make_token(struct spkm3_ctx *ctx,
60 struct xdr_buf * text, struct xdr_netobj * token,
61 int toktype)
62{
63 s32 checksum_type;
64 char tokhdrbuf[25];
65 char cksumdata[16];
66 struct xdr_netobj md5cksum = {.len = 0, .data = cksumdata};
67 struct xdr_netobj mic_hdr = {.len = 0, .data = tokhdrbuf};
68 int tokenlen = 0;
69 unsigned char *ptr;
70 s32 now;
71 int ctxelen = 0, ctxzbit = 0;
72 int md5elen = 0, md5zbit = 0;
73
74 now = jiffies;
75
76 if (ctx->ctx_id.len != 16) {
77 dprintk("RPC: spkm3_make_token BAD ctx_id.len %d\n",
78 ctx->ctx_id.len);
79 goto out_err;
80 }
81
82 if (!g_OID_equal(&ctx->intg_alg, &hmac_md5_oid)) {
83 dprintk("RPC: gss_spkm3_seal: unsupported I-ALG "
84 "algorithm. only support hmac-md5 I-ALG.\n");
85 goto out_err;
86 } else
87 checksum_type = CKSUMTYPE_HMAC_MD5;
88
89 if (!g_OID_equal(&ctx->conf_alg, &cast5_cbc_oid)) {
90 dprintk("RPC: gss_spkm3_seal: unsupported C-ALG "
91 "algorithm\n");
92 goto out_err;
93 }
94
95 if (toktype == SPKM_MIC_TOK) {
96 /* Calculate checksum over the mic-header */
97 asn1_bitstring_len(&ctx->ctx_id, &ctxelen, &ctxzbit);
98 spkm3_mic_header(&mic_hdr.data, &mic_hdr.len, ctx->ctx_id.data,
99 ctxelen, ctxzbit);
100 if (make_spkm3_checksum(checksum_type, &ctx->derived_integ_key,
101 (char *)mic_hdr.data, mic_hdr.len,
102 text, 0, &md5cksum))
103 goto out_err;
104
105 asn1_bitstring_len(&md5cksum, &md5elen, &md5zbit);
106 tokenlen = 10 + ctxelen + 1 + md5elen + 1;
107
108 /* Create token header using generic routines */
109 token->len = g_token_size(&ctx->mech_used, tokenlen + 2);
110
111 ptr = token->data;
112 g_make_token_header(&ctx->mech_used, tokenlen + 2, &ptr);
113
114 spkm3_make_mic_token(&ptr, tokenlen, &mic_hdr, &md5cksum, md5elen, md5zbit);
115 } else if (toktype == SPKM_WRAP_TOK) { /* Not Supported */
116 dprintk("RPC: gss_spkm3_seal: SPKM_WRAP_TOK "
117 "not supported\n");
118 goto out_err;
119 }
120
121 /* XXX need to implement sequence numbers, and ctx->expired */
122
123 return GSS_S_COMPLETE;
124out_err:
125 token->data = NULL;
126 token->len = 0;
127 return GSS_S_FAILURE;
128}
129
130static int
131spkm3_checksummer(struct scatterlist *sg, void *data)
132{
133 struct hash_desc *desc = data;
134
135 return crypto_hash_update(desc, sg, sg->length);
136}
137
138/* checksum the plaintext data and hdrlen bytes of the token header */
139s32
140make_spkm3_checksum(s32 cksumtype, struct xdr_netobj *key, char *header,
141 unsigned int hdrlen, struct xdr_buf *body,
142 unsigned int body_offset, struct xdr_netobj *cksum)
143{
144 char *cksumname;
145 struct hash_desc desc; /* XXX add to ctx? */
146 struct scatterlist sg[1];
147 int err;
148
149 switch (cksumtype) {
150 case CKSUMTYPE_HMAC_MD5:
151 cksumname = "hmac(md5)";
152 break;
153 default:
154 dprintk("RPC: spkm3_make_checksum:"
155 " unsupported checksum %d", cksumtype);
156 return GSS_S_FAILURE;
157 }
158
159 if (key->data == NULL || key->len <= 0) return GSS_S_FAILURE;
160
161 desc.tfm = crypto_alloc_hash(cksumname, 0, CRYPTO_ALG_ASYNC);
162 if (IS_ERR(desc.tfm))
163 return GSS_S_FAILURE;
164 cksum->len = crypto_hash_digestsize(desc.tfm);
165 desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
166
167 err = crypto_hash_setkey(desc.tfm, key->data, key->len);
168 if (err)
169 goto out;
170
171 err = crypto_hash_init(&desc);
172 if (err)
173 goto out;
174
175 sg_init_one(sg, header, hdrlen);
176 crypto_hash_update(&desc, sg, sg->length);
177
178 xdr_process_buf(body, body_offset, body->len - body_offset,
179 spkm3_checksummer, &desc);
180 crypto_hash_final(&desc, cksum->data);
181
182out:
183 crypto_free_hash(desc.tfm);
184
185 return err ? GSS_S_FAILURE : 0;
186}
diff --git a/net/sunrpc/auth_gss/gss_spkm3_token.c b/net/sunrpc/auth_gss/gss_spkm3_token.c
deleted file mode 100644
index a99825d7caa0..000000000000
--- a/net/sunrpc/auth_gss/gss_spkm3_token.c
+++ /dev/null
@@ -1,267 +0,0 @@
1/*
2 * linux/net/sunrpc/gss_spkm3_token.c
3 *
4 * Copyright (c) 2003 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Andy Adamson <andros@umich.edu>
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
23 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
24 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
25 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
29 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
30 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
31 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
32 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 *
34 */
35
36#include <linux/types.h>
37#include <linux/slab.h>
38#include <linux/jiffies.h>
39#include <linux/sunrpc/gss_spkm3.h>
40#include <linux/random.h>
41#include <linux/crypto.h>
42
43#ifdef RPC_DEBUG
44# define RPCDBG_FACILITY RPCDBG_AUTH
45#endif
46
47/*
48 * asn1_bitstring_len()
49 *
50 * calculate the asn1 bitstring length of the xdr_netobject
51 */
52void
53asn1_bitstring_len(struct xdr_netobj *in, int *enclen, int *zerobits)
54{
55 int i, zbit = 0,elen = in->len;
56 char *ptr;
57
58 ptr = &in->data[in->len -1];
59
60 /* count trailing 0's */
61 for(i = in->len; i > 0; i--) {
62 if (*ptr == 0) {
63 ptr--;
64 elen--;
65 } else
66 break;
67 }
68
69 /* count number of 0 bits in final octet */
70 ptr = &in->data[elen - 1];
71 for(i = 0; i < 8; i++) {
72 short mask = 0x01;
73
74 if (!((mask << i) & *ptr))
75 zbit++;
76 else
77 break;
78 }
79 *enclen = elen;
80 *zerobits = zbit;
81}
82
83/*
84 * decode_asn1_bitstring()
85 *
86 * decode a bitstring into a buffer of the expected length.
87 * enclen = bit string length
88 * explen = expected length (define in rfc)
89 */
90int
91decode_asn1_bitstring(struct xdr_netobj *out, char *in, int enclen, int explen)
92{
93 if (!(out->data = kzalloc(explen,GFP_NOFS)))
94 return 0;
95 out->len = explen;
96 memcpy(out->data, in, enclen);
97 return 1;
98}
99
100/*
101 * SPKMInnerContextToken choice SPKM_MIC asn1 token layout
102 *
103 * contextid is always 16 bytes plain data. max asn1 bitstring len = 17.
104 *
105 * tokenlen = pos[0] to end of token (max pos[45] with MD5 cksum)
106 *
107 * pos value
108 * ----------
109 * [0] a4 SPKM-MIC tag
110 * [1] ?? innertoken length (max 44)
111 *
112 *
113 * tok_hdr piece of checksum data starts here
114 *
115 * the maximum mic-header len = 9 + 17 = 26
116 * mic-header
117 * ----------
118 * [2] 30 SEQUENCE tag
119 * [3] ?? mic-header length: (max 23) = TokenID + ContextID
120 *
121 * TokenID - all fields constant and can be hardcoded
122 * -------
123 * [4] 02 Type 2
124 * [5] 02 Length 2
125 * [6][7] 01 01 TokenID (SPKM_MIC_TOK)
126 *
127 * ContextID - encoded length not constant, calculated
128 * ---------
129 * [8] 03 Type 3
130 * [9] ?? encoded length
131 * [10] ?? ctxzbit
132 * [11] contextid
133 *
134 * mic_header piece of checksum data ends here.
135 *
136 * int-cksum - encoded length not constant, calculated
137 * ---------
138 * [??] 03 Type 3
139 * [??] ?? encoded length
140 * [??] ?? md5zbit
141 * [??] int-cksum (NID_md5 = 16)
142 *
143 * maximum SPKM-MIC innercontext token length =
144 * 10 + encoded contextid_size(17 max) + 2 + encoded
145 * cksum_size (17 maxfor NID_md5) = 46
146 */
147
148/*
149 * spkm3_mic_header()
150 *
151 * Prepare the SPKM_MIC_TOK mic-header for check-sum calculation
152 * elen: 16 byte context id asn1 bitstring encoded length
153 */
154void
155spkm3_mic_header(unsigned char **hdrbuf, unsigned int *hdrlen, unsigned char *ctxdata, int elen, int zbit)
156{
157 char *hptr = *hdrbuf;
158 char *top = *hdrbuf;
159
160 *(u8 *)hptr++ = 0x30;
161 *(u8 *)hptr++ = elen + 7; /* on the wire header length */
162
163 /* tokenid */
164 *(u8 *)hptr++ = 0x02;
165 *(u8 *)hptr++ = 0x02;
166 *(u8 *)hptr++ = 0x01;
167 *(u8 *)hptr++ = 0x01;
168
169 /* coniextid */
170 *(u8 *)hptr++ = 0x03;
171 *(u8 *)hptr++ = elen + 1; /* add 1 to include zbit */
172 *(u8 *)hptr++ = zbit;
173 memcpy(hptr, ctxdata, elen);
174 hptr += elen;
175 *hdrlen = hptr - top;
176}
177
178/*
179 * spkm3_mic_innercontext_token()
180 *
181 * *tokp points to the beginning of the SPKM_MIC token described
182 * in rfc 2025, section 3.2.1:
183 *
184 * toklen is the inner token length
185 */
186void
187spkm3_make_mic_token(unsigned char **tokp, int toklen, struct xdr_netobj *mic_hdr, struct xdr_netobj *md5cksum, int md5elen, int md5zbit)
188{
189 unsigned char *ict = *tokp;
190
191 *(u8 *)ict++ = 0xa4;
192 *(u8 *)ict++ = toklen;
193 memcpy(ict, mic_hdr->data, mic_hdr->len);
194 ict += mic_hdr->len;
195
196 *(u8 *)ict++ = 0x03;
197 *(u8 *)ict++ = md5elen + 1; /* add 1 to include zbit */
198 *(u8 *)ict++ = md5zbit;
199 memcpy(ict, md5cksum->data, md5elen);
200}
201
202u32
203spkm3_verify_mic_token(unsigned char **tokp, int *mic_hdrlen, unsigned char **cksum)
204{
205 struct xdr_netobj spkm3_ctx_id = {.len =0, .data = NULL};
206 unsigned char *ptr = *tokp;
207 int ctxelen;
208 u32 ret = GSS_S_DEFECTIVE_TOKEN;
209
210 /* spkm3 innercontext token preamble */
211 if ((ptr[0] != 0xa4) || (ptr[2] != 0x30)) {
212 dprintk("RPC: BAD SPKM ictoken preamble\n");
213 goto out;
214 }
215
216 *mic_hdrlen = ptr[3];
217
218 /* token type */
219 if ((ptr[4] != 0x02) || (ptr[5] != 0x02)) {
220 dprintk("RPC: BAD asn1 SPKM3 token type\n");
221 goto out;
222 }
223
224 /* only support SPKM_MIC_TOK */
225 if((ptr[6] != 0x01) || (ptr[7] != 0x01)) {
226 dprintk("RPC: ERROR unsupported SPKM3 token\n");
227 goto out;
228 }
229
230 /* contextid */
231 if (ptr[8] != 0x03) {
232 dprintk("RPC: BAD SPKM3 asn1 context-id type\n");
233 goto out;
234 }
235
236 ctxelen = ptr[9];
237 if (ctxelen > 17) { /* length includes asn1 zbit octet */
238 dprintk("RPC: BAD SPKM3 contextid len %d\n", ctxelen);
239 goto out;
240 }
241
242 /* ignore ptr[10] */
243
244 if(!decode_asn1_bitstring(&spkm3_ctx_id, &ptr[11], ctxelen - 1, 16))
245 goto out;
246
247 /*
248 * in the current implementation: the optional int-alg is not present
249 * so the default int-alg (md5) is used the optional snd-seq field is
250 * also not present
251 */
252
253 if (*mic_hdrlen != 6 + ctxelen) {
254 dprintk("RPC: BAD SPKM_ MIC_TOK header len %d: we only "
255 "support default int-alg (should be absent) "
256 "and do not support snd-seq\n", *mic_hdrlen);
257 goto out;
258 }
259 /* checksum */
260 *cksum = (&ptr[10] + ctxelen); /* ctxelen includes ptr[10] */
261
262 ret = GSS_S_COMPLETE;
263out:
264 kfree(spkm3_ctx_id.data);
265 return ret;
266}
267
diff --git a/net/sunrpc/auth_gss/gss_spkm3_unseal.c b/net/sunrpc/auth_gss/gss_spkm3_unseal.c
deleted file mode 100644
index cc21ee860bb6..000000000000
--- a/net/sunrpc/auth_gss/gss_spkm3_unseal.c
+++ /dev/null
@@ -1,127 +0,0 @@
1/*
2 * linux/net/sunrpc/gss_spkm3_unseal.c
3 *
4 * Copyright (c) 2003 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Andy Adamson <andros@umich.edu>
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
23 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
24 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
25 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
27 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
28 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
29 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
30 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
31 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
32 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 *
34 */
35
36#include <linux/types.h>
37#include <linux/slab.h>
38#include <linux/jiffies.h>
39#include <linux/sunrpc/gss_spkm3.h>
40#include <linux/crypto.h>
41
42#ifdef RPC_DEBUG
43# define RPCDBG_FACILITY RPCDBG_AUTH
44#endif
45
46/*
47 * spkm3_read_token()
48 *
49 * only SPKM_MIC_TOK with md5 intg-alg is supported
50 */
51u32
52spkm3_read_token(struct spkm3_ctx *ctx,
53 struct xdr_netobj *read_token, /* checksum */
54 struct xdr_buf *message_buffer, /* signbuf */
55 int toktype)
56{
57 s32 checksum_type;
58 s32 code;
59 struct xdr_netobj wire_cksum = {.len =0, .data = NULL};
60 char cksumdata[16];
61 struct xdr_netobj md5cksum = {.len = 0, .data = cksumdata};
62 unsigned char *ptr = (unsigned char *)read_token->data;
63 unsigned char *cksum;
64 int bodysize, md5elen;
65 int mic_hdrlen;
66 u32 ret = GSS_S_DEFECTIVE_TOKEN;
67
68 if (g_verify_token_header((struct xdr_netobj *) &ctx->mech_used,
69 &bodysize, &ptr, read_token->len))
70 goto out;
71
72 /* decode the token */
73
74 if (toktype != SPKM_MIC_TOK) {
75 dprintk("RPC: BAD SPKM3 token type: %d\n", toktype);
76 goto out;
77 }
78
79 if ((ret = spkm3_verify_mic_token(&ptr, &mic_hdrlen, &cksum)))
80 goto out;
81
82 if (*cksum++ != 0x03) {
83 dprintk("RPC: spkm3_read_token BAD checksum type\n");
84 goto out;
85 }
86 md5elen = *cksum++;
87 cksum++; /* move past the zbit */
88
89 if (!decode_asn1_bitstring(&wire_cksum, cksum, md5elen - 1, 16))
90 goto out;
91
92 /* HARD CODED FOR MD5 */
93
94 /* compute the checksum of the message.
95 * ptr + 2 = start of header piece of checksum
96 * mic_hdrlen + 2 = length of header piece of checksum
97 */
98 ret = GSS_S_DEFECTIVE_TOKEN;
99 if (!g_OID_equal(&ctx->intg_alg, &hmac_md5_oid)) {
100 dprintk("RPC: gss_spkm3_seal: unsupported I-ALG "
101 "algorithm\n");
102 goto out;
103 }
104
105 checksum_type = CKSUMTYPE_HMAC_MD5;
106
107 code = make_spkm3_checksum(checksum_type,
108 &ctx->derived_integ_key, ptr + 2, mic_hdrlen + 2,
109 message_buffer, 0, &md5cksum);
110
111 if (code)
112 goto out;
113
114 ret = GSS_S_BAD_SIG;
115 code = memcmp(md5cksum.data, wire_cksum.data, wire_cksum.len);
116 if (code) {
117 dprintk("RPC: bad MIC checksum\n");
118 goto out;
119 }
120
121
122 /* XXX: need to add expiration and sequencing */
123 ret = GSS_S_COMPLETE;
124out:
125 kfree(wire_cksum.data);
126 return ret;
127}
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index cc385b3a59c2..dec2a6fc7c12 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -964,7 +964,7 @@ svcauth_gss_set_client(struct svc_rqst *rqstp)
964 if (rqstp->rq_gssclient == NULL) 964 if (rqstp->rq_gssclient == NULL)
965 return SVC_DENIED; 965 return SVC_DENIED;
966 stat = svcauth_unix_set_client(rqstp); 966 stat = svcauth_unix_set_client(rqstp);
967 if (stat == SVC_DROP) 967 if (stat == SVC_DROP || stat == SVC_CLOSE)
968 return stat; 968 return stat;
969 return SVC_OK; 969 return SVC_OK;
970} 970}
@@ -1018,7 +1018,7 @@ static int svcauth_gss_handle_init(struct svc_rqst *rqstp,
1018 return SVC_DENIED; 1018 return SVC_DENIED;
1019 memset(&rsikey, 0, sizeof(rsikey)); 1019 memset(&rsikey, 0, sizeof(rsikey));
1020 if (dup_netobj(&rsikey.in_handle, &gc->gc_ctx)) 1020 if (dup_netobj(&rsikey.in_handle, &gc->gc_ctx))
1021 return SVC_DROP; 1021 return SVC_CLOSE;
1022 *authp = rpc_autherr_badverf; 1022 *authp = rpc_autherr_badverf;
1023 if (svc_safe_getnetobj(argv, &tmpobj)) { 1023 if (svc_safe_getnetobj(argv, &tmpobj)) {
1024 kfree(rsikey.in_handle.data); 1024 kfree(rsikey.in_handle.data);
@@ -1026,38 +1026,35 @@ static int svcauth_gss_handle_init(struct svc_rqst *rqstp,
1026 } 1026 }
1027 if (dup_netobj(&rsikey.in_token, &tmpobj)) { 1027 if (dup_netobj(&rsikey.in_token, &tmpobj)) {
1028 kfree(rsikey.in_handle.data); 1028 kfree(rsikey.in_handle.data);
1029 return SVC_DROP; 1029 return SVC_CLOSE;
1030 } 1030 }
1031 1031
1032 /* Perform upcall, or find upcall result: */ 1032 /* Perform upcall, or find upcall result: */
1033 rsip = rsi_lookup(&rsikey); 1033 rsip = rsi_lookup(&rsikey);
1034 rsi_free(&rsikey); 1034 rsi_free(&rsikey);
1035 if (!rsip) 1035 if (!rsip)
1036 return SVC_DROP; 1036 return SVC_CLOSE;
1037 switch (cache_check(&rsi_cache, &rsip->h, &rqstp->rq_chandle)) { 1037 if (cache_check(&rsi_cache, &rsip->h, &rqstp->rq_chandle) < 0)
1038 case -EAGAIN:
1039 case -ETIMEDOUT:
1040 case -ENOENT:
1041 /* No upcall result: */ 1038 /* No upcall result: */
1042 return SVC_DROP; 1039 return SVC_CLOSE;
1043 case 0: 1040
1044 ret = SVC_DROP; 1041 ret = SVC_CLOSE;
1045 /* Got an answer to the upcall; use it: */ 1042 /* Got an answer to the upcall; use it: */
1046 if (gss_write_init_verf(rqstp, rsip)) 1043 if (gss_write_init_verf(rqstp, rsip))
1047 goto out; 1044 goto out;
1048 if (resv->iov_len + 4 > PAGE_SIZE) 1045 if (resv->iov_len + 4 > PAGE_SIZE)
1049 goto out; 1046 goto out;
1050 svc_putnl(resv, RPC_SUCCESS); 1047 svc_putnl(resv, RPC_SUCCESS);
1051 if (svc_safe_putnetobj(resv, &rsip->out_handle)) 1048 if (svc_safe_putnetobj(resv, &rsip->out_handle))
1052 goto out; 1049 goto out;
1053 if (resv->iov_len + 3 * 4 > PAGE_SIZE) 1050 if (resv->iov_len + 3 * 4 > PAGE_SIZE)
1054 goto out; 1051 goto out;
1055 svc_putnl(resv, rsip->major_status); 1052 svc_putnl(resv, rsip->major_status);
1056 svc_putnl(resv, rsip->minor_status); 1053 svc_putnl(resv, rsip->minor_status);
1057 svc_putnl(resv, GSS_SEQ_WIN); 1054 svc_putnl(resv, GSS_SEQ_WIN);
1058 if (svc_safe_putnetobj(resv, &rsip->out_token)) 1055 if (svc_safe_putnetobj(resv, &rsip->out_token))
1059 goto out; 1056 goto out;
1060 } 1057
1061 ret = SVC_COMPLETE; 1058 ret = SVC_COMPLETE;
1062out: 1059out:
1063 cache_put(&rsip->h, &rsi_cache); 1060 cache_put(&rsip->h, &rsi_cache);
diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c
index 7dce81a926c5..e433e7580e27 100644
--- a/net/sunrpc/cache.c
+++ b/net/sunrpc/cache.c
@@ -33,15 +33,16 @@
33#include <linux/sunrpc/cache.h> 33#include <linux/sunrpc/cache.h>
34#include <linux/sunrpc/stats.h> 34#include <linux/sunrpc/stats.h>
35#include <linux/sunrpc/rpc_pipe_fs.h> 35#include <linux/sunrpc/rpc_pipe_fs.h>
36#include "netns.h"
36 37
37#define RPCDBG_FACILITY RPCDBG_CACHE 38#define RPCDBG_FACILITY RPCDBG_CACHE
38 39
39static int cache_defer_req(struct cache_req *req, struct cache_head *item); 40static void cache_defer_req(struct cache_req *req, struct cache_head *item);
40static void cache_revisit_request(struct cache_head *item); 41static void cache_revisit_request(struct cache_head *item);
41 42
42static void cache_init(struct cache_head *h) 43static void cache_init(struct cache_head *h)
43{ 44{
44 time_t now = get_seconds(); 45 time_t now = seconds_since_boot();
45 h->next = NULL; 46 h->next = NULL;
46 h->flags = 0; 47 h->flags = 0;
47 kref_init(&h->ref); 48 kref_init(&h->ref);
@@ -51,7 +52,7 @@ static void cache_init(struct cache_head *h)
51 52
52static inline int cache_is_expired(struct cache_detail *detail, struct cache_head *h) 53static inline int cache_is_expired(struct cache_detail *detail, struct cache_head *h)
53{ 54{
54 return (h->expiry_time < get_seconds()) || 55 return (h->expiry_time < seconds_since_boot()) ||
55 (detail->flush_time > h->last_refresh); 56 (detail->flush_time > h->last_refresh);
56} 57}
57 58
@@ -126,7 +127,7 @@ static void cache_dequeue(struct cache_detail *detail, struct cache_head *ch);
126static void cache_fresh_locked(struct cache_head *head, time_t expiry) 127static void cache_fresh_locked(struct cache_head *head, time_t expiry)
127{ 128{
128 head->expiry_time = expiry; 129 head->expiry_time = expiry;
129 head->last_refresh = get_seconds(); 130 head->last_refresh = seconds_since_boot();
130 set_bit(CACHE_VALID, &head->flags); 131 set_bit(CACHE_VALID, &head->flags);
131} 132}
132 133
@@ -237,7 +238,7 @@ int cache_check(struct cache_detail *detail,
237 238
238 /* now see if we want to start an upcall */ 239 /* now see if we want to start an upcall */
239 refresh_age = (h->expiry_time - h->last_refresh); 240 refresh_age = (h->expiry_time - h->last_refresh);
240 age = get_seconds() - h->last_refresh; 241 age = seconds_since_boot() - h->last_refresh;
241 242
242 if (rqstp == NULL) { 243 if (rqstp == NULL) {
243 if (rv == -EAGAIN) 244 if (rv == -EAGAIN)
@@ -252,7 +253,7 @@ int cache_check(struct cache_detail *detail,
252 cache_revisit_request(h); 253 cache_revisit_request(h);
253 if (rv == -EAGAIN) { 254 if (rv == -EAGAIN) {
254 set_bit(CACHE_NEGATIVE, &h->flags); 255 set_bit(CACHE_NEGATIVE, &h->flags);
255 cache_fresh_locked(h, get_seconds()+CACHE_NEW_EXPIRY); 256 cache_fresh_locked(h, seconds_since_boot()+CACHE_NEW_EXPIRY);
256 cache_fresh_unlocked(h, detail); 257 cache_fresh_unlocked(h, detail);
257 rv = -ENOENT; 258 rv = -ENOENT;
258 } 259 }
@@ -267,7 +268,8 @@ int cache_check(struct cache_detail *detail,
267 } 268 }
268 269
269 if (rv == -EAGAIN) { 270 if (rv == -EAGAIN) {
270 if (cache_defer_req(rqstp, h) < 0) { 271 cache_defer_req(rqstp, h);
272 if (!test_bit(CACHE_PENDING, &h->flags)) {
271 /* Request is not deferred */ 273 /* Request is not deferred */
272 rv = cache_is_valid(detail, h); 274 rv = cache_is_valid(detail, h);
273 if (rv == -EAGAIN) 275 if (rv == -EAGAIN)
@@ -387,11 +389,11 @@ static int cache_clean(void)
387 return -1; 389 return -1;
388 } 390 }
389 current_detail = list_entry(next, struct cache_detail, others); 391 current_detail = list_entry(next, struct cache_detail, others);
390 if (current_detail->nextcheck > get_seconds()) 392 if (current_detail->nextcheck > seconds_since_boot())
391 current_index = current_detail->hash_size; 393 current_index = current_detail->hash_size;
392 else { 394 else {
393 current_index = 0; 395 current_index = 0;
394 current_detail->nextcheck = get_seconds()+30*60; 396 current_detail->nextcheck = seconds_since_boot()+30*60;
395 } 397 }
396 } 398 }
397 399
@@ -476,7 +478,7 @@ EXPORT_SYMBOL_GPL(cache_flush);
476void cache_purge(struct cache_detail *detail) 478void cache_purge(struct cache_detail *detail)
477{ 479{
478 detail->flush_time = LONG_MAX; 480 detail->flush_time = LONG_MAX;
479 detail->nextcheck = get_seconds(); 481 detail->nextcheck = seconds_since_boot();
480 cache_flush(); 482 cache_flush();
481 detail->flush_time = 1; 483 detail->flush_time = 1;
482} 484}
@@ -505,81 +507,155 @@ EXPORT_SYMBOL_GPL(cache_purge);
505 507
506static DEFINE_SPINLOCK(cache_defer_lock); 508static DEFINE_SPINLOCK(cache_defer_lock);
507static LIST_HEAD(cache_defer_list); 509static LIST_HEAD(cache_defer_list);
508static struct list_head cache_defer_hash[DFR_HASHSIZE]; 510static struct hlist_head cache_defer_hash[DFR_HASHSIZE];
509static int cache_defer_cnt; 511static int cache_defer_cnt;
510 512
511static int cache_defer_req(struct cache_req *req, struct cache_head *item) 513static void __unhash_deferred_req(struct cache_deferred_req *dreq)
514{
515 hlist_del_init(&dreq->hash);
516 if (!list_empty(&dreq->recent)) {
517 list_del_init(&dreq->recent);
518 cache_defer_cnt--;
519 }
520}
521
522static void __hash_deferred_req(struct cache_deferred_req *dreq, struct cache_head *item)
512{ 523{
513 struct cache_deferred_req *dreq, *discard;
514 int hash = DFR_HASH(item); 524 int hash = DFR_HASH(item);
515 525
516 if (cache_defer_cnt >= DFR_MAX) { 526 INIT_LIST_HEAD(&dreq->recent);
517 /* too much in the cache, randomly drop this one, 527 hlist_add_head(&dreq->hash, &cache_defer_hash[hash]);
518 * or continue and drop the oldest below 528}
519 */ 529
520 if (net_random()&1) 530static void setup_deferral(struct cache_deferred_req *dreq,
521 return -ENOMEM; 531 struct cache_head *item,
522 } 532 int count_me)
523 dreq = req->defer(req); 533{
524 if (dreq == NULL)
525 return -ENOMEM;
526 534
527 dreq->item = item; 535 dreq->item = item;
528 536
529 spin_lock(&cache_defer_lock); 537 spin_lock(&cache_defer_lock);
530 538
531 list_add(&dreq->recent, &cache_defer_list); 539 __hash_deferred_req(dreq, item);
532
533 if (cache_defer_hash[hash].next == NULL)
534 INIT_LIST_HEAD(&cache_defer_hash[hash]);
535 list_add(&dreq->hash, &cache_defer_hash[hash]);
536 540
537 /* it is in, now maybe clean up */ 541 if (count_me) {
538 discard = NULL; 542 cache_defer_cnt++;
539 if (++cache_defer_cnt > DFR_MAX) { 543 list_add(&dreq->recent, &cache_defer_list);
540 discard = list_entry(cache_defer_list.prev,
541 struct cache_deferred_req, recent);
542 list_del_init(&discard->recent);
543 list_del_init(&discard->hash);
544 cache_defer_cnt--;
545 } 544 }
545
546 spin_unlock(&cache_defer_lock); 546 spin_unlock(&cache_defer_lock);
547 547
548}
549
550struct thread_deferred_req {
551 struct cache_deferred_req handle;
552 struct completion completion;
553};
554
555static void cache_restart_thread(struct cache_deferred_req *dreq, int too_many)
556{
557 struct thread_deferred_req *dr =
558 container_of(dreq, struct thread_deferred_req, handle);
559 complete(&dr->completion);
560}
561
562static void cache_wait_req(struct cache_req *req, struct cache_head *item)
563{
564 struct thread_deferred_req sleeper;
565 struct cache_deferred_req *dreq = &sleeper.handle;
566
567 sleeper.completion = COMPLETION_INITIALIZER_ONSTACK(sleeper.completion);
568 dreq->revisit = cache_restart_thread;
569
570 setup_deferral(dreq, item, 0);
571
572 if (!test_bit(CACHE_PENDING, &item->flags) ||
573 wait_for_completion_interruptible_timeout(
574 &sleeper.completion, req->thread_wait) <= 0) {
575 /* The completion wasn't completed, so we need
576 * to clean up
577 */
578 spin_lock(&cache_defer_lock);
579 if (!hlist_unhashed(&sleeper.handle.hash)) {
580 __unhash_deferred_req(&sleeper.handle);
581 spin_unlock(&cache_defer_lock);
582 } else {
583 /* cache_revisit_request already removed
584 * this from the hash table, but hasn't
585 * called ->revisit yet. It will very soon
586 * and we need to wait for it.
587 */
588 spin_unlock(&cache_defer_lock);
589 wait_for_completion(&sleeper.completion);
590 }
591 }
592}
593
594static void cache_limit_defers(void)
595{
596 /* Make sure we haven't exceed the limit of allowed deferred
597 * requests.
598 */
599 struct cache_deferred_req *discard = NULL;
600
601 if (cache_defer_cnt <= DFR_MAX)
602 return;
603
604 spin_lock(&cache_defer_lock);
605
606 /* Consider removing either the first or the last */
607 if (cache_defer_cnt > DFR_MAX) {
608 if (net_random() & 1)
609 discard = list_entry(cache_defer_list.next,
610 struct cache_deferred_req, recent);
611 else
612 discard = list_entry(cache_defer_list.prev,
613 struct cache_deferred_req, recent);
614 __unhash_deferred_req(discard);
615 }
616 spin_unlock(&cache_defer_lock);
548 if (discard) 617 if (discard)
549 /* there was one too many */
550 discard->revisit(discard, 1); 618 discard->revisit(discard, 1);
619}
551 620
552 if (!test_bit(CACHE_PENDING, &item->flags)) { 621static void cache_defer_req(struct cache_req *req, struct cache_head *item)
553 /* must have just been validated... */ 622{
554 cache_revisit_request(item); 623 struct cache_deferred_req *dreq;
555 return -EAGAIN; 624
625 if (req->thread_wait) {
626 cache_wait_req(req, item);
627 if (!test_bit(CACHE_PENDING, &item->flags))
628 return;
556 } 629 }
557 return 0; 630 dreq = req->defer(req);
631 if (dreq == NULL)
632 return;
633 setup_deferral(dreq, item, 1);
634 if (!test_bit(CACHE_PENDING, &item->flags))
635 /* Bit could have been cleared before we managed to
636 * set up the deferral, so need to revisit just in case
637 */
638 cache_revisit_request(item);
639
640 cache_limit_defers();
558} 641}
559 642
560static void cache_revisit_request(struct cache_head *item) 643static void cache_revisit_request(struct cache_head *item)
561{ 644{
562 struct cache_deferred_req *dreq; 645 struct cache_deferred_req *dreq;
563 struct list_head pending; 646 struct list_head pending;
564 647 struct hlist_node *lp, *tmp;
565 struct list_head *lp;
566 int hash = DFR_HASH(item); 648 int hash = DFR_HASH(item);
567 649
568 INIT_LIST_HEAD(&pending); 650 INIT_LIST_HEAD(&pending);
569 spin_lock(&cache_defer_lock); 651 spin_lock(&cache_defer_lock);
570 652
571 lp = cache_defer_hash[hash].next; 653 hlist_for_each_entry_safe(dreq, lp, tmp, &cache_defer_hash[hash], hash)
572 if (lp) { 654 if (dreq->item == item) {
573 while (lp != &cache_defer_hash[hash]) { 655 __unhash_deferred_req(dreq);
574 dreq = list_entry(lp, struct cache_deferred_req, hash); 656 list_add(&dreq->recent, &pending);
575 lp = lp->next;
576 if (dreq->item == item) {
577 list_del_init(&dreq->hash);
578 list_move(&dreq->recent, &pending);
579 cache_defer_cnt--;
580 }
581 } 657 }
582 } 658
583 spin_unlock(&cache_defer_lock); 659 spin_unlock(&cache_defer_lock);
584 660
585 while (!list_empty(&pending)) { 661 while (!list_empty(&pending)) {
@@ -600,9 +676,8 @@ void cache_clean_deferred(void *owner)
600 676
601 list_for_each_entry_safe(dreq, tmp, &cache_defer_list, recent) { 677 list_for_each_entry_safe(dreq, tmp, &cache_defer_list, recent) {
602 if (dreq->owner == owner) { 678 if (dreq->owner == owner) {
603 list_del_init(&dreq->hash); 679 __unhash_deferred_req(dreq);
604 list_move(&dreq->recent, &pending); 680 list_add(&dreq->recent, &pending);
605 cache_defer_cnt--;
606 } 681 }
607 } 682 }
608 spin_unlock(&cache_defer_lock); 683 spin_unlock(&cache_defer_lock);
@@ -901,7 +976,7 @@ static int cache_release(struct inode *inode, struct file *filp,
901 filp->private_data = NULL; 976 filp->private_data = NULL;
902 kfree(rp); 977 kfree(rp);
903 978
904 cd->last_close = get_seconds(); 979 cd->last_close = seconds_since_boot();
905 atomic_dec(&cd->readers); 980 atomic_dec(&cd->readers);
906 } 981 }
907 module_put(cd->owner); 982 module_put(cd->owner);
@@ -1014,6 +1089,23 @@ static void warn_no_listener(struct cache_detail *detail)
1014 } 1089 }
1015} 1090}
1016 1091
1092static bool cache_listeners_exist(struct cache_detail *detail)
1093{
1094 if (atomic_read(&detail->readers))
1095 return true;
1096 if (detail->last_close == 0)
1097 /* This cache was never opened */
1098 return false;
1099 if (detail->last_close < seconds_since_boot() - 30)
1100 /*
1101 * We allow for the possibility that someone might
1102 * restart a userspace daemon without restarting the
1103 * server; but after 30 seconds, we give up.
1104 */
1105 return false;
1106 return true;
1107}
1108
1017/* 1109/*
1018 * register an upcall request to user-space and queue it up for read() by the 1110 * register an upcall request to user-space and queue it up for read() by the
1019 * upcall daemon. 1111 * upcall daemon.
@@ -1032,10 +1124,9 @@ int sunrpc_cache_pipe_upcall(struct cache_detail *detail, struct cache_head *h,
1032 char *bp; 1124 char *bp;
1033 int len; 1125 int len;
1034 1126
1035 if (atomic_read(&detail->readers) == 0 && 1127 if (!cache_listeners_exist(detail)) {
1036 detail->last_close < get_seconds() - 30) { 1128 warn_no_listener(detail);
1037 warn_no_listener(detail); 1129 return -EINVAL;
1038 return -EINVAL;
1039 } 1130 }
1040 1131
1041 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 1132 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
@@ -1094,13 +1185,19 @@ int qword_get(char **bpp, char *dest, int bufsize)
1094 if (bp[0] == '\\' && bp[1] == 'x') { 1185 if (bp[0] == '\\' && bp[1] == 'x') {
1095 /* HEX STRING */ 1186 /* HEX STRING */
1096 bp += 2; 1187 bp += 2;
1097 while (isxdigit(bp[0]) && isxdigit(bp[1]) && len < bufsize) { 1188 while (len < bufsize) {
1098 int byte = isdigit(*bp) ? *bp-'0' : toupper(*bp)-'A'+10; 1189 int h, l;
1099 bp++; 1190
1100 byte <<= 4; 1191 h = hex_to_bin(bp[0]);
1101 byte |= isdigit(*bp) ? *bp-'0' : toupper(*bp)-'A'+10; 1192 if (h < 0)
1102 *dest++ = byte; 1193 break;
1103 bp++; 1194
1195 l = hex_to_bin(bp[1]);
1196 if (l < 0)
1197 break;
1198
1199 *dest++ = (h << 4) | l;
1200 bp += 2;
1104 len++; 1201 len++;
1105 } 1202 }
1106 } else { 1203 } else {
@@ -1218,7 +1315,8 @@ static int c_show(struct seq_file *m, void *p)
1218 1315
1219 ifdebug(CACHE) 1316 ifdebug(CACHE)
1220 seq_printf(m, "# expiry=%ld refcnt=%d flags=%lx\n", 1317 seq_printf(m, "# expiry=%ld refcnt=%d flags=%lx\n",
1221 cp->expiry_time, atomic_read(&cp->ref.refcount), cp->flags); 1318 convert_to_wallclock(cp->expiry_time),
1319 atomic_read(&cp->ref.refcount), cp->flags);
1222 cache_get(cp); 1320 cache_get(cp);
1223 if (cache_check(cd, cp, NULL)) 1321 if (cache_check(cd, cp, NULL))
1224 /* cache_check does a cache_put on failure */ 1322 /* cache_check does a cache_put on failure */
@@ -1284,7 +1382,7 @@ static ssize_t read_flush(struct file *file, char __user *buf,
1284 unsigned long p = *ppos; 1382 unsigned long p = *ppos;
1285 size_t len; 1383 size_t len;
1286 1384
1287 sprintf(tbuf, "%lu\n", cd->flush_time); 1385 sprintf(tbuf, "%lu\n", convert_to_wallclock(cd->flush_time));
1288 len = strlen(tbuf); 1386 len = strlen(tbuf);
1289 if (p >= len) 1387 if (p >= len)
1290 return 0; 1388 return 0;
@@ -1302,19 +1400,20 @@ static ssize_t write_flush(struct file *file, const char __user *buf,
1302 struct cache_detail *cd) 1400 struct cache_detail *cd)
1303{ 1401{
1304 char tbuf[20]; 1402 char tbuf[20];
1305 char *ep; 1403 char *bp, *ep;
1306 long flushtime; 1404
1307 if (*ppos || count > sizeof(tbuf)-1) 1405 if (*ppos || count > sizeof(tbuf)-1)
1308 return -EINVAL; 1406 return -EINVAL;
1309 if (copy_from_user(tbuf, buf, count)) 1407 if (copy_from_user(tbuf, buf, count))
1310 return -EFAULT; 1408 return -EFAULT;
1311 tbuf[count] = 0; 1409 tbuf[count] = 0;
1312 flushtime = simple_strtoul(tbuf, &ep, 0); 1410 simple_strtoul(tbuf, &ep, 0);
1313 if (*ep && *ep != '\n') 1411 if (*ep && *ep != '\n')
1314 return -EINVAL; 1412 return -EINVAL;
1315 1413
1316 cd->flush_time = flushtime; 1414 bp = tbuf;
1317 cd->nextcheck = get_seconds(); 1415 cd->flush_time = get_expiry(&bp);
1416 cd->nextcheck = seconds_since_boot();
1318 cache_flush(); 1417 cache_flush();
1319 1418
1320 *ppos += count; 1419 *ppos += count;
@@ -1438,8 +1537,10 @@ static const struct file_operations cache_flush_operations_procfs = {
1438 .llseek = no_llseek, 1537 .llseek = no_llseek,
1439}; 1538};
1440 1539
1441static void remove_cache_proc_entries(struct cache_detail *cd) 1540static void remove_cache_proc_entries(struct cache_detail *cd, struct net *net)
1442{ 1541{
1542 struct sunrpc_net *sn;
1543
1443 if (cd->u.procfs.proc_ent == NULL) 1544 if (cd->u.procfs.proc_ent == NULL)
1444 return; 1545 return;
1445 if (cd->u.procfs.flush_ent) 1546 if (cd->u.procfs.flush_ent)
@@ -1449,15 +1550,18 @@ static void remove_cache_proc_entries(struct cache_detail *cd)
1449 if (cd->u.procfs.content_ent) 1550 if (cd->u.procfs.content_ent)
1450 remove_proc_entry("content", cd->u.procfs.proc_ent); 1551 remove_proc_entry("content", cd->u.procfs.proc_ent);
1451 cd->u.procfs.proc_ent = NULL; 1552 cd->u.procfs.proc_ent = NULL;
1452 remove_proc_entry(cd->name, proc_net_rpc); 1553 sn = net_generic(net, sunrpc_net_id);
1554 remove_proc_entry(cd->name, sn->proc_net_rpc);
1453} 1555}
1454 1556
1455#ifdef CONFIG_PROC_FS 1557#ifdef CONFIG_PROC_FS
1456static int create_cache_proc_entries(struct cache_detail *cd) 1558static int create_cache_proc_entries(struct cache_detail *cd, struct net *net)
1457{ 1559{
1458 struct proc_dir_entry *p; 1560 struct proc_dir_entry *p;
1561 struct sunrpc_net *sn;
1459 1562
1460 cd->u.procfs.proc_ent = proc_mkdir(cd->name, proc_net_rpc); 1563 sn = net_generic(net, sunrpc_net_id);
1564 cd->u.procfs.proc_ent = proc_mkdir(cd->name, sn->proc_net_rpc);
1461 if (cd->u.procfs.proc_ent == NULL) 1565 if (cd->u.procfs.proc_ent == NULL)
1462 goto out_nomem; 1566 goto out_nomem;
1463 cd->u.procfs.channel_ent = NULL; 1567 cd->u.procfs.channel_ent = NULL;
@@ -1488,11 +1592,11 @@ static int create_cache_proc_entries(struct cache_detail *cd)
1488 } 1592 }
1489 return 0; 1593 return 0;
1490out_nomem: 1594out_nomem:
1491 remove_cache_proc_entries(cd); 1595 remove_cache_proc_entries(cd, net);
1492 return -ENOMEM; 1596 return -ENOMEM;
1493} 1597}
1494#else /* CONFIG_PROC_FS */ 1598#else /* CONFIG_PROC_FS */
1495static int create_cache_proc_entries(struct cache_detail *cd) 1599static int create_cache_proc_entries(struct cache_detail *cd, struct net *net)
1496{ 1600{
1497 return 0; 1601 return 0;
1498} 1602}
@@ -1503,23 +1607,33 @@ void __init cache_initialize(void)
1503 INIT_DELAYED_WORK_DEFERRABLE(&cache_cleaner, do_cache_clean); 1607 INIT_DELAYED_WORK_DEFERRABLE(&cache_cleaner, do_cache_clean);
1504} 1608}
1505 1609
1506int cache_register(struct cache_detail *cd) 1610int cache_register_net(struct cache_detail *cd, struct net *net)
1507{ 1611{
1508 int ret; 1612 int ret;
1509 1613
1510 sunrpc_init_cache_detail(cd); 1614 sunrpc_init_cache_detail(cd);
1511 ret = create_cache_proc_entries(cd); 1615 ret = create_cache_proc_entries(cd, net);
1512 if (ret) 1616 if (ret)
1513 sunrpc_destroy_cache_detail(cd); 1617 sunrpc_destroy_cache_detail(cd);
1514 return ret; 1618 return ret;
1515} 1619}
1620
1621int cache_register(struct cache_detail *cd)
1622{
1623 return cache_register_net(cd, &init_net);
1624}
1516EXPORT_SYMBOL_GPL(cache_register); 1625EXPORT_SYMBOL_GPL(cache_register);
1517 1626
1518void cache_unregister(struct cache_detail *cd) 1627void cache_unregister_net(struct cache_detail *cd, struct net *net)
1519{ 1628{
1520 remove_cache_proc_entries(cd); 1629 remove_cache_proc_entries(cd, net);
1521 sunrpc_destroy_cache_detail(cd); 1630 sunrpc_destroy_cache_detail(cd);
1522} 1631}
1632
1633void cache_unregister(struct cache_detail *cd)
1634{
1635 cache_unregister_net(cd, &init_net);
1636}
1523EXPORT_SYMBOL_GPL(cache_unregister); 1637EXPORT_SYMBOL_GPL(cache_unregister);
1524 1638
1525static ssize_t cache_read_pipefs(struct file *filp, char __user *buf, 1639static ssize_t cache_read_pipefs(struct file *filp, char __user *buf,
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index cbc5b8ccc8be..9dab9573be41 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -284,6 +284,7 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args)
284 struct rpc_xprt *xprt; 284 struct rpc_xprt *xprt;
285 struct rpc_clnt *clnt; 285 struct rpc_clnt *clnt;
286 struct xprt_create xprtargs = { 286 struct xprt_create xprtargs = {
287 .net = args->net,
287 .ident = args->protocol, 288 .ident = args->protocol,
288 .srcaddr = args->saddress, 289 .srcaddr = args->saddress,
289 .dstaddr = args->address, 290 .dstaddr = args->address,
diff --git a/net/sunrpc/netns.h b/net/sunrpc/netns.h
new file mode 100644
index 000000000000..d013bf211cae
--- /dev/null
+++ b/net/sunrpc/netns.h
@@ -0,0 +1,19 @@
1#ifndef __SUNRPC_NETNS_H__
2#define __SUNRPC_NETNS_H__
3
4#include <net/net_namespace.h>
5#include <net/netns/generic.h>
6
7struct cache_detail;
8
9struct sunrpc_net {
10 struct proc_dir_entry *proc_net_rpc;
11 struct cache_detail *ip_map_cache;
12};
13
14extern int sunrpc_net_id;
15
16int ip_map_cache_create(struct net *);
17void ip_map_cache_destroy(struct net *);
18
19#endif
diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c
index 63ec116b4dd4..fa6d7ca2c851 100644
--- a/net/sunrpc/rpcb_clnt.c
+++ b/net/sunrpc/rpcb_clnt.c
@@ -177,6 +177,7 @@ static DEFINE_MUTEX(rpcb_create_local_mutex);
177static int rpcb_create_local(void) 177static int rpcb_create_local(void)
178{ 178{
179 struct rpc_create_args args = { 179 struct rpc_create_args args = {
180 .net = &init_net,
180 .protocol = XPRT_TRANSPORT_TCP, 181 .protocol = XPRT_TRANSPORT_TCP,
181 .address = (struct sockaddr *)&rpcb_inaddr_loopback, 182 .address = (struct sockaddr *)&rpcb_inaddr_loopback,
182 .addrsize = sizeof(rpcb_inaddr_loopback), 183 .addrsize = sizeof(rpcb_inaddr_loopback),
@@ -229,6 +230,7 @@ static struct rpc_clnt *rpcb_create(char *hostname, struct sockaddr *srvaddr,
229 size_t salen, int proto, u32 version) 230 size_t salen, int proto, u32 version)
230{ 231{
231 struct rpc_create_args args = { 232 struct rpc_create_args args = {
233 .net = &init_net,
232 .protocol = proto, 234 .protocol = proto,
233 .address = srvaddr, 235 .address = srvaddr,
234 .addrsize = salen, 236 .addrsize = salen,
@@ -248,7 +250,7 @@ static struct rpc_clnt *rpcb_create(char *hostname, struct sockaddr *srvaddr,
248 ((struct sockaddr_in6 *)srvaddr)->sin6_port = htons(RPCBIND_PORT); 250 ((struct sockaddr_in6 *)srvaddr)->sin6_port = htons(RPCBIND_PORT);
249 break; 251 break;
250 default: 252 default:
251 return NULL; 253 return ERR_PTR(-EAFNOSUPPORT);
252 } 254 }
253 255
254 return rpc_create(&args); 256 return rpc_create(&args);
diff --git a/net/sunrpc/stats.c b/net/sunrpc/stats.c
index ea1046f3f9a3..f71a73107ae9 100644
--- a/net/sunrpc/stats.c
+++ b/net/sunrpc/stats.c
@@ -22,11 +22,10 @@
22#include <linux/sunrpc/clnt.h> 22#include <linux/sunrpc/clnt.h>
23#include <linux/sunrpc/svcsock.h> 23#include <linux/sunrpc/svcsock.h>
24#include <linux/sunrpc/metrics.h> 24#include <linux/sunrpc/metrics.h>
25#include <net/net_namespace.h>
26 25
27#define RPCDBG_FACILITY RPCDBG_MISC 26#include "netns.h"
28 27
29struct proc_dir_entry *proc_net_rpc = NULL; 28#define RPCDBG_FACILITY RPCDBG_MISC
30 29
31/* 30/*
32 * Get RPC client stats 31 * Get RPC client stats
@@ -218,10 +217,11 @@ EXPORT_SYMBOL_GPL(rpc_print_iostats);
218static inline struct proc_dir_entry * 217static inline struct proc_dir_entry *
219do_register(const char *name, void *data, const struct file_operations *fops) 218do_register(const char *name, void *data, const struct file_operations *fops)
220{ 219{
221 rpc_proc_init(); 220 struct sunrpc_net *sn;
222 dprintk("RPC: registering /proc/net/rpc/%s\n", name);
223 221
224 return proc_create_data(name, 0, proc_net_rpc, fops, data); 222 dprintk("RPC: registering /proc/net/rpc/%s\n", name);
223 sn = net_generic(&init_net, sunrpc_net_id);
224 return proc_create_data(name, 0, sn->proc_net_rpc, fops, data);
225} 225}
226 226
227struct proc_dir_entry * 227struct proc_dir_entry *
@@ -234,7 +234,10 @@ EXPORT_SYMBOL_GPL(rpc_proc_register);
234void 234void
235rpc_proc_unregister(const char *name) 235rpc_proc_unregister(const char *name)
236{ 236{
237 remove_proc_entry(name, proc_net_rpc); 237 struct sunrpc_net *sn;
238
239 sn = net_generic(&init_net, sunrpc_net_id);
240 remove_proc_entry(name, sn->proc_net_rpc);
238} 241}
239EXPORT_SYMBOL_GPL(rpc_proc_unregister); 242EXPORT_SYMBOL_GPL(rpc_proc_unregister);
240 243
@@ -248,25 +251,29 @@ EXPORT_SYMBOL_GPL(svc_proc_register);
248void 251void
249svc_proc_unregister(const char *name) 252svc_proc_unregister(const char *name)
250{ 253{
251 remove_proc_entry(name, proc_net_rpc); 254 struct sunrpc_net *sn;
255
256 sn = net_generic(&init_net, sunrpc_net_id);
257 remove_proc_entry(name, sn->proc_net_rpc);
252} 258}
253EXPORT_SYMBOL_GPL(svc_proc_unregister); 259EXPORT_SYMBOL_GPL(svc_proc_unregister);
254 260
255void 261int rpc_proc_init(struct net *net)
256rpc_proc_init(void)
257{ 262{
263 struct sunrpc_net *sn;
264
258 dprintk("RPC: registering /proc/net/rpc\n"); 265 dprintk("RPC: registering /proc/net/rpc\n");
259 if (!proc_net_rpc) 266 sn = net_generic(net, sunrpc_net_id);
260 proc_net_rpc = proc_mkdir("rpc", init_net.proc_net); 267 sn->proc_net_rpc = proc_mkdir("rpc", net->proc_net);
268 if (sn->proc_net_rpc == NULL)
269 return -ENOMEM;
270
271 return 0;
261} 272}
262 273
263void 274void rpc_proc_exit(struct net *net)
264rpc_proc_exit(void)
265{ 275{
266 dprintk("RPC: unregistering /proc/net/rpc\n"); 276 dprintk("RPC: unregistering /proc/net/rpc\n");
267 if (proc_net_rpc) { 277 remove_proc_entry("rpc", net->proc_net);
268 proc_net_rpc = NULL;
269 remove_proc_entry("rpc", init_net.proc_net);
270 }
271} 278}
272 279
diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c
index c0d085013a2b..9d0809160994 100644
--- a/net/sunrpc/sunrpc_syms.c
+++ b/net/sunrpc/sunrpc_syms.c
@@ -22,7 +22,44 @@
22#include <linux/sunrpc/rpc_pipe_fs.h> 22#include <linux/sunrpc/rpc_pipe_fs.h>
23#include <linux/sunrpc/xprtsock.h> 23#include <linux/sunrpc/xprtsock.h>
24 24
25extern struct cache_detail ip_map_cache, unix_gid_cache; 25#include "netns.h"
26
27int sunrpc_net_id;
28
29static __net_init int sunrpc_init_net(struct net *net)
30{
31 int err;
32
33 err = rpc_proc_init(net);
34 if (err)
35 goto err_proc;
36
37 err = ip_map_cache_create(net);
38 if (err)
39 goto err_ipmap;
40
41 return 0;
42
43err_ipmap:
44 rpc_proc_exit(net);
45err_proc:
46 return err;
47}
48
49static __net_exit void sunrpc_exit_net(struct net *net)
50{
51 ip_map_cache_destroy(net);
52 rpc_proc_exit(net);
53}
54
55static struct pernet_operations sunrpc_net_ops = {
56 .init = sunrpc_init_net,
57 .exit = sunrpc_exit_net,
58 .id = &sunrpc_net_id,
59 .size = sizeof(struct sunrpc_net),
60};
61
62extern struct cache_detail unix_gid_cache;
26 63
27extern void cleanup_rpcb_clnt(void); 64extern void cleanup_rpcb_clnt(void);
28 65
@@ -38,18 +75,22 @@ init_sunrpc(void)
38 err = rpcauth_init_module(); 75 err = rpcauth_init_module();
39 if (err) 76 if (err)
40 goto out3; 77 goto out3;
78
79 cache_initialize();
80
81 err = register_pernet_subsys(&sunrpc_net_ops);
82 if (err)
83 goto out4;
41#ifdef RPC_DEBUG 84#ifdef RPC_DEBUG
42 rpc_register_sysctl(); 85 rpc_register_sysctl();
43#endif 86#endif
44#ifdef CONFIG_PROC_FS
45 rpc_proc_init();
46#endif
47 cache_initialize();
48 cache_register(&ip_map_cache);
49 cache_register(&unix_gid_cache); 87 cache_register(&unix_gid_cache);
50 svc_init_xprt_sock(); /* svc sock transport */ 88 svc_init_xprt_sock(); /* svc sock transport */
51 init_socket_xprt(); /* clnt sock transport */ 89 init_socket_xprt(); /* clnt sock transport */
52 return 0; 90 return 0;
91
92out4:
93 rpcauth_remove_module();
53out3: 94out3:
54 rpc_destroy_mempool(); 95 rpc_destroy_mempool();
55out2: 96out2:
@@ -67,14 +108,11 @@ cleanup_sunrpc(void)
67 svc_cleanup_xprt_sock(); 108 svc_cleanup_xprt_sock();
68 unregister_rpc_pipefs(); 109 unregister_rpc_pipefs();
69 rpc_destroy_mempool(); 110 rpc_destroy_mempool();
70 cache_unregister(&ip_map_cache);
71 cache_unregister(&unix_gid_cache); 111 cache_unregister(&unix_gid_cache);
112 unregister_pernet_subsys(&sunrpc_net_ops);
72#ifdef RPC_DEBUG 113#ifdef RPC_DEBUG
73 rpc_unregister_sysctl(); 114 rpc_unregister_sysctl();
74#endif 115#endif
75#ifdef CONFIG_PROC_FS
76 rpc_proc_exit();
77#endif
78 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 116 rcu_barrier(); /* Wait for completion of call_rcu()'s */
79} 117}
80MODULE_LICENSE("GPL"); 118MODULE_LICENSE("GPL");
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index d9017d64597e..6359c42c4941 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1055,6 +1055,9 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
1055 goto err_bad; 1055 goto err_bad;
1056 case SVC_DENIED: 1056 case SVC_DENIED:
1057 goto err_bad_auth; 1057 goto err_bad_auth;
1058 case SVC_CLOSE:
1059 if (test_bit(XPT_TEMP, &rqstp->rq_xprt->xpt_flags))
1060 svc_close_xprt(rqstp->rq_xprt);
1058 case SVC_DROP: 1061 case SVC_DROP:
1059 goto dropit; 1062 goto dropit;
1060 case SVC_COMPLETE: 1063 case SVC_COMPLETE:
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index cbc084939dd8..c82fe739fbdc 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -100,16 +100,14 @@ EXPORT_SYMBOL_GPL(svc_unreg_xprt_class);
100 */ 100 */
101int svc_print_xprts(char *buf, int maxlen) 101int svc_print_xprts(char *buf, int maxlen)
102{ 102{
103 struct list_head *le; 103 struct svc_xprt_class *xcl;
104 char tmpstr[80]; 104 char tmpstr[80];
105 int len = 0; 105 int len = 0;
106 buf[0] = '\0'; 106 buf[0] = '\0';
107 107
108 spin_lock(&svc_xprt_class_lock); 108 spin_lock(&svc_xprt_class_lock);
109 list_for_each(le, &svc_xprt_class_list) { 109 list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
110 int slen; 110 int slen;
111 struct svc_xprt_class *xcl =
112 list_entry(le, struct svc_xprt_class, xcl_list);
113 111
114 sprintf(tmpstr, "%s %d\n", xcl->xcl_name, xcl->xcl_max_payload); 112 sprintf(tmpstr, "%s %d\n", xcl->xcl_name, xcl->xcl_max_payload);
115 slen = strlen(tmpstr); 113 slen = strlen(tmpstr);
@@ -128,9 +126,9 @@ static void svc_xprt_free(struct kref *kref)
128 struct svc_xprt *xprt = 126 struct svc_xprt *xprt =
129 container_of(kref, struct svc_xprt, xpt_ref); 127 container_of(kref, struct svc_xprt, xpt_ref);
130 struct module *owner = xprt->xpt_class->xcl_owner; 128 struct module *owner = xprt->xpt_class->xcl_owner;
131 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags) && 129 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags))
132 xprt->xpt_auth_cache != NULL) 130 svcauth_unix_info_release(xprt);
133 svcauth_unix_info_release(xprt->xpt_auth_cache); 131 put_net(xprt->xpt_net);
134 xprt->xpt_ops->xpo_free(xprt); 132 xprt->xpt_ops->xpo_free(xprt);
135 module_put(owner); 133 module_put(owner);
136} 134}
@@ -156,15 +154,18 @@ void svc_xprt_init(struct svc_xprt_class *xcl, struct svc_xprt *xprt,
156 INIT_LIST_HEAD(&xprt->xpt_list); 154 INIT_LIST_HEAD(&xprt->xpt_list);
157 INIT_LIST_HEAD(&xprt->xpt_ready); 155 INIT_LIST_HEAD(&xprt->xpt_ready);
158 INIT_LIST_HEAD(&xprt->xpt_deferred); 156 INIT_LIST_HEAD(&xprt->xpt_deferred);
157 INIT_LIST_HEAD(&xprt->xpt_users);
159 mutex_init(&xprt->xpt_mutex); 158 mutex_init(&xprt->xpt_mutex);
160 spin_lock_init(&xprt->xpt_lock); 159 spin_lock_init(&xprt->xpt_lock);
161 set_bit(XPT_BUSY, &xprt->xpt_flags); 160 set_bit(XPT_BUSY, &xprt->xpt_flags);
162 rpc_init_wait_queue(&xprt->xpt_bc_pending, "xpt_bc_pending"); 161 rpc_init_wait_queue(&xprt->xpt_bc_pending, "xpt_bc_pending");
162 xprt->xpt_net = get_net(&init_net);
163} 163}
164EXPORT_SYMBOL_GPL(svc_xprt_init); 164EXPORT_SYMBOL_GPL(svc_xprt_init);
165 165
166static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl, 166static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
167 struct svc_serv *serv, 167 struct svc_serv *serv,
168 struct net *net,
168 const int family, 169 const int family,
169 const unsigned short port, 170 const unsigned short port,
170 int flags) 171 int flags)
@@ -199,12 +200,12 @@ static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
199 return ERR_PTR(-EAFNOSUPPORT); 200 return ERR_PTR(-EAFNOSUPPORT);
200 } 201 }
201 202
202 return xcl->xcl_ops->xpo_create(serv, sap, len, flags); 203 return xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
203} 204}
204 205
205int svc_create_xprt(struct svc_serv *serv, const char *xprt_name, 206int svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
206 const int family, const unsigned short port, 207 struct net *net, const int family,
207 int flags) 208 const unsigned short port, int flags)
208{ 209{
209 struct svc_xprt_class *xcl; 210 struct svc_xprt_class *xcl;
210 211
@@ -220,7 +221,7 @@ int svc_create_xprt(struct svc_serv *serv, const char *xprt_name,
220 goto err; 221 goto err;
221 222
222 spin_unlock(&svc_xprt_class_lock); 223 spin_unlock(&svc_xprt_class_lock);
223 newxprt = __svc_xpo_create(xcl, serv, family, port, flags); 224 newxprt = __svc_xpo_create(xcl, serv, net, family, port, flags);
224 if (IS_ERR(newxprt)) { 225 if (IS_ERR(newxprt)) {
225 module_put(xcl->xcl_owner); 226 module_put(xcl->xcl_owner);
226 return PTR_ERR(newxprt); 227 return PTR_ERR(newxprt);
@@ -329,12 +330,6 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
329 "svc_xprt_enqueue: " 330 "svc_xprt_enqueue: "
330 "threads and transports both waiting??\n"); 331 "threads and transports both waiting??\n");
331 332
332 if (test_bit(XPT_DEAD, &xprt->xpt_flags)) {
333 /* Don't enqueue dead transports */
334 dprintk("svc: transport %p is dead, not enqueued\n", xprt);
335 goto out_unlock;
336 }
337
338 pool->sp_stats.packets++; 333 pool->sp_stats.packets++;
339 334
340 /* Mark transport as busy. It will remain in this state until 335 /* Mark transport as busy. It will remain in this state until
@@ -651,6 +646,11 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
651 if (signalled() || kthread_should_stop()) 646 if (signalled() || kthread_should_stop())
652 return -EINTR; 647 return -EINTR;
653 648
649 /* Normally we will wait up to 5 seconds for any required
650 * cache information to be provided.
651 */
652 rqstp->rq_chandle.thread_wait = 5*HZ;
653
654 spin_lock_bh(&pool->sp_lock); 654 spin_lock_bh(&pool->sp_lock);
655 xprt = svc_xprt_dequeue(pool); 655 xprt = svc_xprt_dequeue(pool);
656 if (xprt) { 656 if (xprt) {
@@ -658,6 +658,12 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
658 svc_xprt_get(xprt); 658 svc_xprt_get(xprt);
659 rqstp->rq_reserved = serv->sv_max_mesg; 659 rqstp->rq_reserved = serv->sv_max_mesg;
660 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved); 660 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
661
662 /* As there is a shortage of threads and this request
663 * had to be queued, don't allow the thread to wait so
664 * long for cache updates.
665 */
666 rqstp->rq_chandle.thread_wait = 1*HZ;
661 } else { 667 } else {
662 /* No data pending. Go to sleep */ 668 /* No data pending. Go to sleep */
663 svc_thread_enqueue(pool, rqstp); 669 svc_thread_enqueue(pool, rqstp);
@@ -868,6 +874,19 @@ static void svc_age_temp_xprts(unsigned long closure)
868 mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ); 874 mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ);
869} 875}
870 876
877static void call_xpt_users(struct svc_xprt *xprt)
878{
879 struct svc_xpt_user *u;
880
881 spin_lock(&xprt->xpt_lock);
882 while (!list_empty(&xprt->xpt_users)) {
883 u = list_first_entry(&xprt->xpt_users, struct svc_xpt_user, list);
884 list_del(&u->list);
885 u->callback(u);
886 }
887 spin_unlock(&xprt->xpt_lock);
888}
889
871/* 890/*
872 * Remove a dead transport 891 * Remove a dead transport
873 */ 892 */
@@ -878,7 +897,7 @@ void svc_delete_xprt(struct svc_xprt *xprt)
878 897
879 /* Only do this once */ 898 /* Only do this once */
880 if (test_and_set_bit(XPT_DEAD, &xprt->xpt_flags)) 899 if (test_and_set_bit(XPT_DEAD, &xprt->xpt_flags))
881 return; 900 BUG();
882 901
883 dprintk("svc: svc_delete_xprt(%p)\n", xprt); 902 dprintk("svc: svc_delete_xprt(%p)\n", xprt);
884 xprt->xpt_ops->xpo_detach(xprt); 903 xprt->xpt_ops->xpo_detach(xprt);
@@ -900,6 +919,7 @@ void svc_delete_xprt(struct svc_xprt *xprt)
900 while ((dr = svc_deferred_dequeue(xprt)) != NULL) 919 while ((dr = svc_deferred_dequeue(xprt)) != NULL)
901 kfree(dr); 920 kfree(dr);
902 921
922 call_xpt_users(xprt);
903 svc_xprt_put(xprt); 923 svc_xprt_put(xprt);
904} 924}
905 925
@@ -910,10 +930,7 @@ void svc_close_xprt(struct svc_xprt *xprt)
910 /* someone else will have to effect the close */ 930 /* someone else will have to effect the close */
911 return; 931 return;
912 932
913 svc_xprt_get(xprt);
914 svc_delete_xprt(xprt); 933 svc_delete_xprt(xprt);
915 clear_bit(XPT_BUSY, &xprt->xpt_flags);
916 svc_xprt_put(xprt);
917} 934}
918EXPORT_SYMBOL_GPL(svc_close_xprt); 935EXPORT_SYMBOL_GPL(svc_close_xprt);
919 936
diff --git a/net/sunrpc/svcauth_unix.c b/net/sunrpc/svcauth_unix.c
index 207311610988..560677d187f1 100644
--- a/net/sunrpc/svcauth_unix.c
+++ b/net/sunrpc/svcauth_unix.c
@@ -18,6 +18,8 @@
18 18
19#include <linux/sunrpc/clnt.h> 19#include <linux/sunrpc/clnt.h>
20 20
21#include "netns.h"
22
21/* 23/*
22 * AUTHUNIX and AUTHNULL credentials are both handled here. 24 * AUTHUNIX and AUTHNULL credentials are both handled here.
23 * AUTHNULL is treated just like AUTHUNIX except that the uid/gid 25 * AUTHNULL is treated just like AUTHUNIX except that the uid/gid
@@ -92,7 +94,6 @@ struct ip_map {
92 struct unix_domain *m_client; 94 struct unix_domain *m_client;
93 int m_add_change; 95 int m_add_change;
94}; 96};
95static struct cache_head *ip_table[IP_HASHMAX];
96 97
97static void ip_map_put(struct kref *kref) 98static void ip_map_put(struct kref *kref)
98{ 99{
@@ -178,8 +179,8 @@ static int ip_map_upcall(struct cache_detail *cd, struct cache_head *h)
178 return sunrpc_cache_pipe_upcall(cd, h, ip_map_request); 179 return sunrpc_cache_pipe_upcall(cd, h, ip_map_request);
179} 180}
180 181
181static struct ip_map *ip_map_lookup(char *class, struct in6_addr *addr); 182static struct ip_map *__ip_map_lookup(struct cache_detail *cd, char *class, struct in6_addr *addr);
182static int ip_map_update(struct ip_map *ipm, struct unix_domain *udom, time_t expiry); 183static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm, struct unix_domain *udom, time_t expiry);
183 184
184static int ip_map_parse(struct cache_detail *cd, 185static int ip_map_parse(struct cache_detail *cd,
185 char *mesg, int mlen) 186 char *mesg, int mlen)
@@ -219,10 +220,9 @@ static int ip_map_parse(struct cache_detail *cd,
219 switch (address.sa.sa_family) { 220 switch (address.sa.sa_family) {
220 case AF_INET: 221 case AF_INET:
221 /* Form a mapped IPv4 address in sin6 */ 222 /* Form a mapped IPv4 address in sin6 */
222 memset(&sin6, 0, sizeof(sin6));
223 sin6.sin6_family = AF_INET6; 223 sin6.sin6_family = AF_INET6;
224 sin6.sin6_addr.s6_addr32[2] = htonl(0xffff); 224 ipv6_addr_set_v4mapped(address.s4.sin_addr.s_addr,
225 sin6.sin6_addr.s6_addr32[3] = address.s4.sin_addr.s_addr; 225 &sin6.sin6_addr);
226 break; 226 break;
227#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 227#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
228 case AF_INET6: 228 case AF_INET6:
@@ -249,9 +249,9 @@ static int ip_map_parse(struct cache_detail *cd,
249 dom = NULL; 249 dom = NULL;
250 250
251 /* IPv6 scope IDs are ignored for now */ 251 /* IPv6 scope IDs are ignored for now */
252 ipmp = ip_map_lookup(class, &sin6.sin6_addr); 252 ipmp = __ip_map_lookup(cd, class, &sin6.sin6_addr);
253 if (ipmp) { 253 if (ipmp) {
254 err = ip_map_update(ipmp, 254 err = __ip_map_update(cd, ipmp,
255 container_of(dom, struct unix_domain, h), 255 container_of(dom, struct unix_domain, h),
256 expiry); 256 expiry);
257 } else 257 } else
@@ -294,29 +294,15 @@ static int ip_map_show(struct seq_file *m,
294} 294}
295 295
296 296
297struct cache_detail ip_map_cache = { 297static struct ip_map *__ip_map_lookup(struct cache_detail *cd, char *class,
298 .owner = THIS_MODULE, 298 struct in6_addr *addr)
299 .hash_size = IP_HASHMAX,
300 .hash_table = ip_table,
301 .name = "auth.unix.ip",
302 .cache_put = ip_map_put,
303 .cache_upcall = ip_map_upcall,
304 .cache_parse = ip_map_parse,
305 .cache_show = ip_map_show,
306 .match = ip_map_match,
307 .init = ip_map_init,
308 .update = update,
309 .alloc = ip_map_alloc,
310};
311
312static struct ip_map *ip_map_lookup(char *class, struct in6_addr *addr)
313{ 299{
314 struct ip_map ip; 300 struct ip_map ip;
315 struct cache_head *ch; 301 struct cache_head *ch;
316 302
317 strcpy(ip.m_class, class); 303 strcpy(ip.m_class, class);
318 ipv6_addr_copy(&ip.m_addr, addr); 304 ipv6_addr_copy(&ip.m_addr, addr);
319 ch = sunrpc_cache_lookup(&ip_map_cache, &ip.h, 305 ch = sunrpc_cache_lookup(cd, &ip.h,
320 hash_str(class, IP_HASHBITS) ^ 306 hash_str(class, IP_HASHBITS) ^
321 hash_ip6(*addr)); 307 hash_ip6(*addr));
322 308
@@ -326,7 +312,17 @@ static struct ip_map *ip_map_lookup(char *class, struct in6_addr *addr)
326 return NULL; 312 return NULL;
327} 313}
328 314
329static int ip_map_update(struct ip_map *ipm, struct unix_domain *udom, time_t expiry) 315static inline struct ip_map *ip_map_lookup(struct net *net, char *class,
316 struct in6_addr *addr)
317{
318 struct sunrpc_net *sn;
319
320 sn = net_generic(net, sunrpc_net_id);
321 return __ip_map_lookup(sn->ip_map_cache, class, addr);
322}
323
324static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm,
325 struct unix_domain *udom, time_t expiry)
330{ 326{
331 struct ip_map ip; 327 struct ip_map ip;
332 struct cache_head *ch; 328 struct cache_head *ch;
@@ -344,17 +340,25 @@ static int ip_map_update(struct ip_map *ipm, struct unix_domain *udom, time_t ex
344 ip.m_add_change++; 340 ip.m_add_change++;
345 } 341 }
346 ip.h.expiry_time = expiry; 342 ip.h.expiry_time = expiry;
347 ch = sunrpc_cache_update(&ip_map_cache, 343 ch = sunrpc_cache_update(cd, &ip.h, &ipm->h,
348 &ip.h, &ipm->h,
349 hash_str(ipm->m_class, IP_HASHBITS) ^ 344 hash_str(ipm->m_class, IP_HASHBITS) ^
350 hash_ip6(ipm->m_addr)); 345 hash_ip6(ipm->m_addr));
351 if (!ch) 346 if (!ch)
352 return -ENOMEM; 347 return -ENOMEM;
353 cache_put(ch, &ip_map_cache); 348 cache_put(ch, cd);
354 return 0; 349 return 0;
355} 350}
356 351
357int auth_unix_add_addr(struct in6_addr *addr, struct auth_domain *dom) 352static inline int ip_map_update(struct net *net, struct ip_map *ipm,
353 struct unix_domain *udom, time_t expiry)
354{
355 struct sunrpc_net *sn;
356
357 sn = net_generic(net, sunrpc_net_id);
358 return __ip_map_update(sn->ip_map_cache, ipm, udom, expiry);
359}
360
361int auth_unix_add_addr(struct net *net, struct in6_addr *addr, struct auth_domain *dom)
358{ 362{
359 struct unix_domain *udom; 363 struct unix_domain *udom;
360 struct ip_map *ipmp; 364 struct ip_map *ipmp;
@@ -362,10 +366,10 @@ int auth_unix_add_addr(struct in6_addr *addr, struct auth_domain *dom)
362 if (dom->flavour != &svcauth_unix) 366 if (dom->flavour != &svcauth_unix)
363 return -EINVAL; 367 return -EINVAL;
364 udom = container_of(dom, struct unix_domain, h); 368 udom = container_of(dom, struct unix_domain, h);
365 ipmp = ip_map_lookup("nfsd", addr); 369 ipmp = ip_map_lookup(net, "nfsd", addr);
366 370
367 if (ipmp) 371 if (ipmp)
368 return ip_map_update(ipmp, udom, NEVER); 372 return ip_map_update(net, ipmp, udom, NEVER);
369 else 373 else
370 return -ENOMEM; 374 return -ENOMEM;
371} 375}
@@ -383,16 +387,18 @@ int auth_unix_forget_old(struct auth_domain *dom)
383} 387}
384EXPORT_SYMBOL_GPL(auth_unix_forget_old); 388EXPORT_SYMBOL_GPL(auth_unix_forget_old);
385 389
386struct auth_domain *auth_unix_lookup(struct in6_addr *addr) 390struct auth_domain *auth_unix_lookup(struct net *net, struct in6_addr *addr)
387{ 391{
388 struct ip_map *ipm; 392 struct ip_map *ipm;
389 struct auth_domain *rv; 393 struct auth_domain *rv;
394 struct sunrpc_net *sn;
390 395
391 ipm = ip_map_lookup("nfsd", addr); 396 sn = net_generic(net, sunrpc_net_id);
397 ipm = ip_map_lookup(net, "nfsd", addr);
392 398
393 if (!ipm) 399 if (!ipm)
394 return NULL; 400 return NULL;
395 if (cache_check(&ip_map_cache, &ipm->h, NULL)) 401 if (cache_check(sn->ip_map_cache, &ipm->h, NULL))
396 return NULL; 402 return NULL;
397 403
398 if ((ipm->m_client->addr_changes - ipm->m_add_change) >0) { 404 if ((ipm->m_client->addr_changes - ipm->m_add_change) >0) {
@@ -403,22 +409,29 @@ struct auth_domain *auth_unix_lookup(struct in6_addr *addr)
403 rv = &ipm->m_client->h; 409 rv = &ipm->m_client->h;
404 kref_get(&rv->ref); 410 kref_get(&rv->ref);
405 } 411 }
406 cache_put(&ipm->h, &ip_map_cache); 412 cache_put(&ipm->h, sn->ip_map_cache);
407 return rv; 413 return rv;
408} 414}
409EXPORT_SYMBOL_GPL(auth_unix_lookup); 415EXPORT_SYMBOL_GPL(auth_unix_lookup);
410 416
411void svcauth_unix_purge(void) 417void svcauth_unix_purge(void)
412{ 418{
413 cache_purge(&ip_map_cache); 419 struct net *net;
420
421 for_each_net(net) {
422 struct sunrpc_net *sn;
423
424 sn = net_generic(net, sunrpc_net_id);
425 cache_purge(sn->ip_map_cache);
426 }
414} 427}
415EXPORT_SYMBOL_GPL(svcauth_unix_purge); 428EXPORT_SYMBOL_GPL(svcauth_unix_purge);
416 429
417static inline struct ip_map * 430static inline struct ip_map *
418ip_map_cached_get(struct svc_rqst *rqstp) 431ip_map_cached_get(struct svc_xprt *xprt)
419{ 432{
420 struct ip_map *ipm = NULL; 433 struct ip_map *ipm = NULL;
421 struct svc_xprt *xprt = rqstp->rq_xprt; 434 struct sunrpc_net *sn;
422 435
423 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) { 436 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) {
424 spin_lock(&xprt->xpt_lock); 437 spin_lock(&xprt->xpt_lock);
@@ -430,9 +443,10 @@ ip_map_cached_get(struct svc_rqst *rqstp)
430 * remembered, e.g. by a second mount from the 443 * remembered, e.g. by a second mount from the
431 * same IP address. 444 * same IP address.
432 */ 445 */
446 sn = net_generic(xprt->xpt_net, sunrpc_net_id);
433 xprt->xpt_auth_cache = NULL; 447 xprt->xpt_auth_cache = NULL;
434 spin_unlock(&xprt->xpt_lock); 448 spin_unlock(&xprt->xpt_lock);
435 cache_put(&ipm->h, &ip_map_cache); 449 cache_put(&ipm->h, sn->ip_map_cache);
436 return NULL; 450 return NULL;
437 } 451 }
438 cache_get(&ipm->h); 452 cache_get(&ipm->h);
@@ -443,10 +457,8 @@ ip_map_cached_get(struct svc_rqst *rqstp)
443} 457}
444 458
445static inline void 459static inline void
446ip_map_cached_put(struct svc_rqst *rqstp, struct ip_map *ipm) 460ip_map_cached_put(struct svc_xprt *xprt, struct ip_map *ipm)
447{ 461{
448 struct svc_xprt *xprt = rqstp->rq_xprt;
449
450 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) { 462 if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) {
451 spin_lock(&xprt->xpt_lock); 463 spin_lock(&xprt->xpt_lock);
452 if (xprt->xpt_auth_cache == NULL) { 464 if (xprt->xpt_auth_cache == NULL) {
@@ -456,15 +468,26 @@ ip_map_cached_put(struct svc_rqst *rqstp, struct ip_map *ipm)
456 } 468 }
457 spin_unlock(&xprt->xpt_lock); 469 spin_unlock(&xprt->xpt_lock);
458 } 470 }
459 if (ipm) 471 if (ipm) {
460 cache_put(&ipm->h, &ip_map_cache); 472 struct sunrpc_net *sn;
473
474 sn = net_generic(xprt->xpt_net, sunrpc_net_id);
475 cache_put(&ipm->h, sn->ip_map_cache);
476 }
461} 477}
462 478
463void 479void
464svcauth_unix_info_release(void *info) 480svcauth_unix_info_release(struct svc_xprt *xpt)
465{ 481{
466 struct ip_map *ipm = info; 482 struct ip_map *ipm;
467 cache_put(&ipm->h, &ip_map_cache); 483
484 ipm = xpt->xpt_auth_cache;
485 if (ipm != NULL) {
486 struct sunrpc_net *sn;
487
488 sn = net_generic(xpt->xpt_net, sunrpc_net_id);
489 cache_put(&ipm->h, sn->ip_map_cache);
490 }
468} 491}
469 492
470/**************************************************************************** 493/****************************************************************************
@@ -674,6 +697,8 @@ static struct group_info *unix_gid_find(uid_t uid, struct svc_rqst *rqstp)
674 switch (ret) { 697 switch (ret) {
675 case -ENOENT: 698 case -ENOENT:
676 return ERR_PTR(-ENOENT); 699 return ERR_PTR(-ENOENT);
700 case -ETIMEDOUT:
701 return ERR_PTR(-ESHUTDOWN);
677 case 0: 702 case 0:
678 gi = get_group_info(ug->gi); 703 gi = get_group_info(ug->gi);
679 cache_put(&ug->h, &unix_gid_cache); 704 cache_put(&ug->h, &unix_gid_cache);
@@ -691,6 +716,9 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
691 struct ip_map *ipm; 716 struct ip_map *ipm;
692 struct group_info *gi; 717 struct group_info *gi;
693 struct svc_cred *cred = &rqstp->rq_cred; 718 struct svc_cred *cred = &rqstp->rq_cred;
719 struct svc_xprt *xprt = rqstp->rq_xprt;
720 struct net *net = xprt->xpt_net;
721 struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
694 722
695 switch (rqstp->rq_addr.ss_family) { 723 switch (rqstp->rq_addr.ss_family) {
696 case AF_INET: 724 case AF_INET:
@@ -709,26 +737,27 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
709 if (rqstp->rq_proc == 0) 737 if (rqstp->rq_proc == 0)
710 return SVC_OK; 738 return SVC_OK;
711 739
712 ipm = ip_map_cached_get(rqstp); 740 ipm = ip_map_cached_get(xprt);
713 if (ipm == NULL) 741 if (ipm == NULL)
714 ipm = ip_map_lookup(rqstp->rq_server->sv_program->pg_class, 742 ipm = __ip_map_lookup(sn->ip_map_cache, rqstp->rq_server->sv_program->pg_class,
715 &sin6->sin6_addr); 743 &sin6->sin6_addr);
716 744
717 if (ipm == NULL) 745 if (ipm == NULL)
718 return SVC_DENIED; 746 return SVC_DENIED;
719 747
720 switch (cache_check(&ip_map_cache, &ipm->h, &rqstp->rq_chandle)) { 748 switch (cache_check(sn->ip_map_cache, &ipm->h, &rqstp->rq_chandle)) {
721 default: 749 default:
722 BUG(); 750 BUG();
723 case -EAGAIN:
724 case -ETIMEDOUT: 751 case -ETIMEDOUT:
752 return SVC_CLOSE;
753 case -EAGAIN:
725 return SVC_DROP; 754 return SVC_DROP;
726 case -ENOENT: 755 case -ENOENT:
727 return SVC_DENIED; 756 return SVC_DENIED;
728 case 0: 757 case 0:
729 rqstp->rq_client = &ipm->m_client->h; 758 rqstp->rq_client = &ipm->m_client->h;
730 kref_get(&rqstp->rq_client->ref); 759 kref_get(&rqstp->rq_client->ref);
731 ip_map_cached_put(rqstp, ipm); 760 ip_map_cached_put(xprt, ipm);
732 break; 761 break;
733 } 762 }
734 763
@@ -736,6 +765,8 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
736 switch (PTR_ERR(gi)) { 765 switch (PTR_ERR(gi)) {
737 case -EAGAIN: 766 case -EAGAIN:
738 return SVC_DROP; 767 return SVC_DROP;
768 case -ESHUTDOWN:
769 return SVC_CLOSE;
739 case -ENOENT: 770 case -ENOENT:
740 break; 771 break;
741 default: 772 default:
@@ -776,7 +807,7 @@ svcauth_null_accept(struct svc_rqst *rqstp, __be32 *authp)
776 cred->cr_gid = (gid_t) -1; 807 cred->cr_gid = (gid_t) -1;
777 cred->cr_group_info = groups_alloc(0); 808 cred->cr_group_info = groups_alloc(0);
778 if (cred->cr_group_info == NULL) 809 if (cred->cr_group_info == NULL)
779 return SVC_DROP; /* kmalloc failure - client must retry */ 810 return SVC_CLOSE; /* kmalloc failure - client must retry */
780 811
781 /* Put NULL verifier */ 812 /* Put NULL verifier */
782 svc_putnl(resv, RPC_AUTH_NULL); 813 svc_putnl(resv, RPC_AUTH_NULL);
@@ -840,7 +871,7 @@ svcauth_unix_accept(struct svc_rqst *rqstp, __be32 *authp)
840 goto badcred; 871 goto badcred;
841 cred->cr_group_info = groups_alloc(slen); 872 cred->cr_group_info = groups_alloc(slen);
842 if (cred->cr_group_info == NULL) 873 if (cred->cr_group_info == NULL)
843 return SVC_DROP; 874 return SVC_CLOSE;
844 for (i = 0; i < slen; i++) 875 for (i = 0; i < slen; i++)
845 GROUP_AT(cred->cr_group_info, i) = svc_getnl(argv); 876 GROUP_AT(cred->cr_group_info, i) = svc_getnl(argv);
846 if (svc_getu32(argv) != htonl(RPC_AUTH_NULL) || svc_getu32(argv) != 0) { 877 if (svc_getu32(argv) != htonl(RPC_AUTH_NULL) || svc_getu32(argv) != 0) {
@@ -886,3 +917,56 @@ struct auth_ops svcauth_unix = {
886 .set_client = svcauth_unix_set_client, 917 .set_client = svcauth_unix_set_client,
887}; 918};
888 919
920int ip_map_cache_create(struct net *net)
921{
922 int err = -ENOMEM;
923 struct cache_detail *cd;
924 struct cache_head **tbl;
925 struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
926
927 cd = kzalloc(sizeof(struct cache_detail), GFP_KERNEL);
928 if (cd == NULL)
929 goto err_cd;
930
931 tbl = kzalloc(IP_HASHMAX * sizeof(struct cache_head *), GFP_KERNEL);
932 if (tbl == NULL)
933 goto err_tbl;
934
935 cd->owner = THIS_MODULE,
936 cd->hash_size = IP_HASHMAX,
937 cd->hash_table = tbl,
938 cd->name = "auth.unix.ip",
939 cd->cache_put = ip_map_put,
940 cd->cache_upcall = ip_map_upcall,
941 cd->cache_parse = ip_map_parse,
942 cd->cache_show = ip_map_show,
943 cd->match = ip_map_match,
944 cd->init = ip_map_init,
945 cd->update = update,
946 cd->alloc = ip_map_alloc,
947
948 err = cache_register_net(cd, net);
949 if (err)
950 goto err_reg;
951
952 sn->ip_map_cache = cd;
953 return 0;
954
955err_reg:
956 kfree(tbl);
957err_tbl:
958 kfree(cd);
959err_cd:
960 return err;
961}
962
963void ip_map_cache_destroy(struct net *net)
964{
965 struct sunrpc_net *sn;
966
967 sn = net_generic(net, sunrpc_net_id);
968 cache_purge(sn->ip_map_cache);
969 cache_unregister_net(sn->ip_map_cache, net);
970 kfree(sn->ip_map_cache->hash_table);
971 kfree(sn->ip_map_cache);
972}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7e534dd09077..07919e16be3e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -64,7 +64,8 @@ static void svc_tcp_sock_detach(struct svc_xprt *);
64static void svc_sock_free(struct svc_xprt *); 64static void svc_sock_free(struct svc_xprt *);
65 65
66static struct svc_xprt *svc_create_socket(struct svc_serv *, int, 66static struct svc_xprt *svc_create_socket(struct svc_serv *, int,
67 struct sockaddr *, int, int); 67 struct net *, struct sockaddr *,
68 int, int);
68#ifdef CONFIG_DEBUG_LOCK_ALLOC 69#ifdef CONFIG_DEBUG_LOCK_ALLOC
69static struct lock_class_key svc_key[2]; 70static struct lock_class_key svc_key[2];
70static struct lock_class_key svc_slock_key[2]; 71static struct lock_class_key svc_slock_key[2];
@@ -657,10 +658,11 @@ static struct svc_xprt *svc_udp_accept(struct svc_xprt *xprt)
657} 658}
658 659
659static struct svc_xprt *svc_udp_create(struct svc_serv *serv, 660static struct svc_xprt *svc_udp_create(struct svc_serv *serv,
661 struct net *net,
660 struct sockaddr *sa, int salen, 662 struct sockaddr *sa, int salen,
661 int flags) 663 int flags)
662{ 664{
663 return svc_create_socket(serv, IPPROTO_UDP, sa, salen, flags); 665 return svc_create_socket(serv, IPPROTO_UDP, net, sa, salen, flags);
664} 666}
665 667
666static struct svc_xprt_ops svc_udp_ops = { 668static struct svc_xprt_ops svc_udp_ops = {
@@ -1133,9 +1135,6 @@ static int svc_tcp_sendto(struct svc_rqst *rqstp)
1133 reclen = htonl(0x80000000|((xbufp->len ) - 4)); 1135 reclen = htonl(0x80000000|((xbufp->len ) - 4));
1134 memcpy(xbufp->head[0].iov_base, &reclen, 4); 1136 memcpy(xbufp->head[0].iov_base, &reclen, 4);
1135 1137
1136 if (test_bit(XPT_DEAD, &rqstp->rq_xprt->xpt_flags))
1137 return -ENOTCONN;
1138
1139 sent = svc_sendto(rqstp, &rqstp->rq_res); 1138 sent = svc_sendto(rqstp, &rqstp->rq_res);
1140 if (sent != xbufp->len) { 1139 if (sent != xbufp->len) {
1141 printk(KERN_NOTICE 1140 printk(KERN_NOTICE
@@ -1178,10 +1177,11 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
1178} 1177}
1179 1178
1180static struct svc_xprt *svc_tcp_create(struct svc_serv *serv, 1179static struct svc_xprt *svc_tcp_create(struct svc_serv *serv,
1180 struct net *net,
1181 struct sockaddr *sa, int salen, 1181 struct sockaddr *sa, int salen,
1182 int flags) 1182 int flags)
1183{ 1183{
1184 return svc_create_socket(serv, IPPROTO_TCP, sa, salen, flags); 1184 return svc_create_socket(serv, IPPROTO_TCP, net, sa, salen, flags);
1185} 1185}
1186 1186
1187static struct svc_xprt_ops svc_tcp_ops = { 1187static struct svc_xprt_ops svc_tcp_ops = {
@@ -1258,19 +1258,13 @@ void svc_sock_update_bufs(struct svc_serv *serv)
1258 * The number of server threads has changed. Update 1258 * The number of server threads has changed. Update
1259 * rcvbuf and sndbuf accordingly on all sockets 1259 * rcvbuf and sndbuf accordingly on all sockets
1260 */ 1260 */
1261 struct list_head *le; 1261 struct svc_sock *svsk;
1262 1262
1263 spin_lock_bh(&serv->sv_lock); 1263 spin_lock_bh(&serv->sv_lock);
1264 list_for_each(le, &serv->sv_permsocks) { 1264 list_for_each_entry(svsk, &serv->sv_permsocks, sk_xprt.xpt_list)
1265 struct svc_sock *svsk =
1266 list_entry(le, struct svc_sock, sk_xprt.xpt_list);
1267 set_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags); 1265 set_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags);
1268 } 1266 list_for_each_entry(svsk, &serv->sv_tempsocks, sk_xprt.xpt_list)
1269 list_for_each(le, &serv->sv_tempsocks) {
1270 struct svc_sock *svsk =
1271 list_entry(le, struct svc_sock, sk_xprt.xpt_list);
1272 set_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags); 1267 set_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags);
1273 }
1274 spin_unlock_bh(&serv->sv_lock); 1268 spin_unlock_bh(&serv->sv_lock);
1275} 1269}
1276EXPORT_SYMBOL_GPL(svc_sock_update_bufs); 1270EXPORT_SYMBOL_GPL(svc_sock_update_bufs);
@@ -1385,6 +1379,7 @@ EXPORT_SYMBOL_GPL(svc_addsock);
1385 */ 1379 */
1386static struct svc_xprt *svc_create_socket(struct svc_serv *serv, 1380static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
1387 int protocol, 1381 int protocol,
1382 struct net *net,
1388 struct sockaddr *sin, int len, 1383 struct sockaddr *sin, int len,
1389 int flags) 1384 int flags)
1390{ 1385{
@@ -1421,7 +1416,7 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
1421 return ERR_PTR(-EINVAL); 1416 return ERR_PTR(-EINVAL);
1422 } 1417 }
1423 1418
1424 error = sock_create_kern(family, type, protocol, &sock); 1419 error = __sock_create(net, family, type, protocol, &sock, 1);
1425 if (error < 0) 1420 if (error < 0)
1426 return ERR_PTR(error); 1421 return ERR_PTR(error);
1427 1422
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 970fb00f388c..4c8f18aff7c3 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -199,8 +199,6 @@ int xprt_reserve_xprt(struct rpc_task *task)
199 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) { 199 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
200 if (task == xprt->snd_task) 200 if (task == xprt->snd_task)
201 return 1; 201 return 1;
202 if (task == NULL)
203 return 0;
204 goto out_sleep; 202 goto out_sleep;
205 } 203 }
206 xprt->snd_task = task; 204 xprt->snd_task = task;
@@ -757,13 +755,11 @@ static void xprt_connect_status(struct rpc_task *task)
757 */ 755 */
758struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid) 756struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
759{ 757{
760 struct list_head *pos; 758 struct rpc_rqst *entry;
761 759
762 list_for_each(pos, &xprt->recv) { 760 list_for_each_entry(entry, &xprt->recv, rq_list)
763 struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
764 if (entry->rq_xid == xid) 761 if (entry->rq_xid == xid)
765 return entry; 762 return entry;
766 }
767 763
768 dprintk("RPC: xprt_lookup_rqst did not find xid %08x\n", 764 dprintk("RPC: xprt_lookup_rqst did not find xid %08x\n",
769 ntohl(xid)); 765 ntohl(xid));
@@ -962,6 +958,37 @@ static void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
962 spin_unlock(&xprt->reserve_lock); 958 spin_unlock(&xprt->reserve_lock);
963} 959}
964 960
961struct rpc_xprt *xprt_alloc(struct net *net, int size, int max_req)
962{
963 struct rpc_xprt *xprt;
964
965 xprt = kzalloc(size, GFP_KERNEL);
966 if (xprt == NULL)
967 goto out;
968
969 xprt->max_reqs = max_req;
970 xprt->slot = kcalloc(max_req, sizeof(struct rpc_rqst), GFP_KERNEL);
971 if (xprt->slot == NULL)
972 goto out_free;
973
974 xprt->xprt_net = get_net(net);
975 return xprt;
976
977out_free:
978 kfree(xprt);
979out:
980 return NULL;
981}
982EXPORT_SYMBOL_GPL(xprt_alloc);
983
984void xprt_free(struct rpc_xprt *xprt)
985{
986 put_net(xprt->xprt_net);
987 kfree(xprt->slot);
988 kfree(xprt);
989}
990EXPORT_SYMBOL_GPL(xprt_free);
991
965/** 992/**
966 * xprt_reserve - allocate an RPC request slot 993 * xprt_reserve - allocate an RPC request slot
967 * @task: RPC task requesting a slot allocation 994 * @task: RPC task requesting a slot allocation
diff --git a/net/sunrpc/xprtrdma/svc_rdma.c b/net/sunrpc/xprtrdma/svc_rdma.c
index d718b8fa9525..09af4fab1a45 100644
--- a/net/sunrpc/xprtrdma/svc_rdma.c
+++ b/net/sunrpc/xprtrdma/svc_rdma.c
@@ -43,6 +43,7 @@
43#include <linux/slab.h> 43#include <linux/slab.h>
44#include <linux/fs.h> 44#include <linux/fs.h>
45#include <linux/sysctl.h> 45#include <linux/sysctl.h>
46#include <linux/workqueue.h>
46#include <linux/sunrpc/clnt.h> 47#include <linux/sunrpc/clnt.h>
47#include <linux/sunrpc/sched.h> 48#include <linux/sunrpc/sched.h>
48#include <linux/sunrpc/svc_rdma.h> 49#include <linux/sunrpc/svc_rdma.h>
@@ -74,6 +75,8 @@ atomic_t rdma_stat_sq_prod;
74struct kmem_cache *svc_rdma_map_cachep; 75struct kmem_cache *svc_rdma_map_cachep;
75struct kmem_cache *svc_rdma_ctxt_cachep; 76struct kmem_cache *svc_rdma_ctxt_cachep;
76 77
78struct workqueue_struct *svc_rdma_wq;
79
77/* 80/*
78 * This function implements reading and resetting an atomic_t stat 81 * This function implements reading and resetting an atomic_t stat
79 * variable through read/write to a proc file. Any write to the file 82 * variable through read/write to a proc file. Any write to the file
@@ -231,7 +234,7 @@ static ctl_table svcrdma_root_table[] = {
231void svc_rdma_cleanup(void) 234void svc_rdma_cleanup(void)
232{ 235{
233 dprintk("SVCRDMA Module Removed, deregister RPC RDMA transport\n"); 236 dprintk("SVCRDMA Module Removed, deregister RPC RDMA transport\n");
234 flush_scheduled_work(); 237 destroy_workqueue(svc_rdma_wq);
235 if (svcrdma_table_header) { 238 if (svcrdma_table_header) {
236 unregister_sysctl_table(svcrdma_table_header); 239 unregister_sysctl_table(svcrdma_table_header);
237 svcrdma_table_header = NULL; 240 svcrdma_table_header = NULL;
@@ -249,6 +252,11 @@ int svc_rdma_init(void)
249 dprintk("\tsq_depth : %d\n", 252 dprintk("\tsq_depth : %d\n",
250 svcrdma_max_requests * RPCRDMA_SQ_DEPTH_MULT); 253 svcrdma_max_requests * RPCRDMA_SQ_DEPTH_MULT);
251 dprintk("\tmax_inline : %d\n", svcrdma_max_req_size); 254 dprintk("\tmax_inline : %d\n", svcrdma_max_req_size);
255
256 svc_rdma_wq = alloc_workqueue("svc_rdma", 0, 0);
257 if (!svc_rdma_wq)
258 return -ENOMEM;
259
252 if (!svcrdma_table_header) 260 if (!svcrdma_table_header)
253 svcrdma_table_header = 261 svcrdma_table_header =
254 register_sysctl_table(svcrdma_root_table); 262 register_sysctl_table(svcrdma_root_table);
@@ -283,6 +291,7 @@ int svc_rdma_init(void)
283 kmem_cache_destroy(svc_rdma_map_cachep); 291 kmem_cache_destroy(svc_rdma_map_cachep);
284 err0: 292 err0:
285 unregister_sysctl_table(svcrdma_table_header); 293 unregister_sysctl_table(svcrdma_table_header);
294 destroy_workqueue(svc_rdma_wq);
286 return -ENOMEM; 295 return -ENOMEM;
287} 296}
288MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>"); 297MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>");
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 0194de814933..df67211c4baf 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -263,9 +263,9 @@ static int fast_reg_read_chunks(struct svcxprt_rdma *xprt,
263 frmr->page_list_len = PAGE_ALIGN(byte_count) >> PAGE_SHIFT; 263 frmr->page_list_len = PAGE_ALIGN(byte_count) >> PAGE_SHIFT;
264 for (page_no = 0; page_no < frmr->page_list_len; page_no++) { 264 for (page_no = 0; page_no < frmr->page_list_len; page_no++) {
265 frmr->page_list->page_list[page_no] = 265 frmr->page_list->page_list[page_no] =
266 ib_dma_map_single(xprt->sc_cm_id->device, 266 ib_dma_map_page(xprt->sc_cm_id->device,
267 page_address(rqstp->rq_arg.pages[page_no]), 267 rqstp->rq_arg.pages[page_no], 0,
268 PAGE_SIZE, DMA_FROM_DEVICE); 268 PAGE_SIZE, DMA_FROM_DEVICE);
269 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 269 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
270 frmr->page_list->page_list[page_no])) 270 frmr->page_list->page_list[page_no]))
271 goto fatal_err; 271 goto fatal_err;
@@ -309,17 +309,21 @@ static int rdma_set_ctxt_sge(struct svcxprt_rdma *xprt,
309 int count) 309 int count)
310{ 310{
311 int i; 311 int i;
312 unsigned long off;
312 313
313 ctxt->count = count; 314 ctxt->count = count;
314 ctxt->direction = DMA_FROM_DEVICE; 315 ctxt->direction = DMA_FROM_DEVICE;
315 for (i = 0; i < count; i++) { 316 for (i = 0; i < count; i++) {
316 ctxt->sge[i].length = 0; /* in case map fails */ 317 ctxt->sge[i].length = 0; /* in case map fails */
317 if (!frmr) { 318 if (!frmr) {
319 BUG_ON(0 == virt_to_page(vec[i].iov_base));
320 off = (unsigned long)vec[i].iov_base & ~PAGE_MASK;
318 ctxt->sge[i].addr = 321 ctxt->sge[i].addr =
319 ib_dma_map_single(xprt->sc_cm_id->device, 322 ib_dma_map_page(xprt->sc_cm_id->device,
320 vec[i].iov_base, 323 virt_to_page(vec[i].iov_base),
321 vec[i].iov_len, 324 off,
322 DMA_FROM_DEVICE); 325 vec[i].iov_len,
326 DMA_FROM_DEVICE);
323 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 327 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
324 ctxt->sge[i].addr)) 328 ctxt->sge[i].addr))
325 return -EINVAL; 329 return -EINVAL;
@@ -491,6 +495,7 @@ next_sge:
491 printk(KERN_ERR "svcrdma: Error %d posting RDMA_READ\n", 495 printk(KERN_ERR "svcrdma: Error %d posting RDMA_READ\n",
492 err); 496 err);
493 set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags); 497 set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
498 svc_rdma_unmap_dma(ctxt);
494 svc_rdma_put_context(ctxt, 0); 499 svc_rdma_put_context(ctxt, 0);
495 goto out; 500 goto out;
496 } 501 }
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index b15e1ebb2bfa..249a835b703f 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -70,8 +70,8 @@
70 * on extra page for the RPCRMDA header. 70 * on extra page for the RPCRMDA header.
71 */ 71 */
72static int fast_reg_xdr(struct svcxprt_rdma *xprt, 72static int fast_reg_xdr(struct svcxprt_rdma *xprt,
73 struct xdr_buf *xdr, 73 struct xdr_buf *xdr,
74 struct svc_rdma_req_map *vec) 74 struct svc_rdma_req_map *vec)
75{ 75{
76 int sge_no; 76 int sge_no;
77 u32 sge_bytes; 77 u32 sge_bytes;
@@ -96,21 +96,25 @@ static int fast_reg_xdr(struct svcxprt_rdma *xprt,
96 vec->count = 2; 96 vec->count = 2;
97 sge_no++; 97 sge_no++;
98 98
99 /* Build the FRMR */ 99 /* Map the XDR head */
100 frmr->kva = frva; 100 frmr->kva = frva;
101 frmr->direction = DMA_TO_DEVICE; 101 frmr->direction = DMA_TO_DEVICE;
102 frmr->access_flags = 0; 102 frmr->access_flags = 0;
103 frmr->map_len = PAGE_SIZE; 103 frmr->map_len = PAGE_SIZE;
104 frmr->page_list_len = 1; 104 frmr->page_list_len = 1;
105 page_off = (unsigned long)xdr->head[0].iov_base & ~PAGE_MASK;
105 frmr->page_list->page_list[page_no] = 106 frmr->page_list->page_list[page_no] =
106 ib_dma_map_single(xprt->sc_cm_id->device, 107 ib_dma_map_page(xprt->sc_cm_id->device,
107 (void *)xdr->head[0].iov_base, 108 virt_to_page(xdr->head[0].iov_base),
108 PAGE_SIZE, DMA_TO_DEVICE); 109 page_off,
110 PAGE_SIZE - page_off,
111 DMA_TO_DEVICE);
109 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 112 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
110 frmr->page_list->page_list[page_no])) 113 frmr->page_list->page_list[page_no]))
111 goto fatal_err; 114 goto fatal_err;
112 atomic_inc(&xprt->sc_dma_used); 115 atomic_inc(&xprt->sc_dma_used);
113 116
117 /* Map the XDR page list */
114 page_off = xdr->page_base; 118 page_off = xdr->page_base;
115 page_bytes = xdr->page_len + page_off; 119 page_bytes = xdr->page_len + page_off;
116 if (!page_bytes) 120 if (!page_bytes)
@@ -128,9 +132,9 @@ static int fast_reg_xdr(struct svcxprt_rdma *xprt,
128 page_bytes -= sge_bytes; 132 page_bytes -= sge_bytes;
129 133
130 frmr->page_list->page_list[page_no] = 134 frmr->page_list->page_list[page_no] =
131 ib_dma_map_single(xprt->sc_cm_id->device, 135 ib_dma_map_page(xprt->sc_cm_id->device,
132 page_address(page), 136 page, page_off,
133 PAGE_SIZE, DMA_TO_DEVICE); 137 sge_bytes, DMA_TO_DEVICE);
134 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 138 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
135 frmr->page_list->page_list[page_no])) 139 frmr->page_list->page_list[page_no]))
136 goto fatal_err; 140 goto fatal_err;
@@ -166,8 +170,10 @@ static int fast_reg_xdr(struct svcxprt_rdma *xprt,
166 vec->sge[sge_no].iov_base = frva + frmr->map_len + page_off; 170 vec->sge[sge_no].iov_base = frva + frmr->map_len + page_off;
167 171
168 frmr->page_list->page_list[page_no] = 172 frmr->page_list->page_list[page_no] =
169 ib_dma_map_single(xprt->sc_cm_id->device, va, PAGE_SIZE, 173 ib_dma_map_page(xprt->sc_cm_id->device, virt_to_page(va),
170 DMA_TO_DEVICE); 174 page_off,
175 PAGE_SIZE,
176 DMA_TO_DEVICE);
171 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 177 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
172 frmr->page_list->page_list[page_no])) 178 frmr->page_list->page_list[page_no]))
173 goto fatal_err; 179 goto fatal_err;
@@ -245,6 +251,35 @@ static int map_xdr(struct svcxprt_rdma *xprt,
245 return 0; 251 return 0;
246} 252}
247 253
254static dma_addr_t dma_map_xdr(struct svcxprt_rdma *xprt,
255 struct xdr_buf *xdr,
256 u32 xdr_off, size_t len, int dir)
257{
258 struct page *page;
259 dma_addr_t dma_addr;
260 if (xdr_off < xdr->head[0].iov_len) {
261 /* This offset is in the head */
262 xdr_off += (unsigned long)xdr->head[0].iov_base & ~PAGE_MASK;
263 page = virt_to_page(xdr->head[0].iov_base);
264 } else {
265 xdr_off -= xdr->head[0].iov_len;
266 if (xdr_off < xdr->page_len) {
267 /* This offset is in the page list */
268 page = xdr->pages[xdr_off >> PAGE_SHIFT];
269 xdr_off &= ~PAGE_MASK;
270 } else {
271 /* This offset is in the tail */
272 xdr_off -= xdr->page_len;
273 xdr_off += (unsigned long)
274 xdr->tail[0].iov_base & ~PAGE_MASK;
275 page = virt_to_page(xdr->tail[0].iov_base);
276 }
277 }
278 dma_addr = ib_dma_map_page(xprt->sc_cm_id->device, page, xdr_off,
279 min_t(size_t, PAGE_SIZE, len), dir);
280 return dma_addr;
281}
282
248/* Assumptions: 283/* Assumptions:
249 * - We are using FRMR 284 * - We are using FRMR
250 * - or - 285 * - or -
@@ -293,10 +328,9 @@ static int send_write(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp,
293 sge[sge_no].length = sge_bytes; 328 sge[sge_no].length = sge_bytes;
294 if (!vec->frmr) { 329 if (!vec->frmr) {
295 sge[sge_no].addr = 330 sge[sge_no].addr =
296 ib_dma_map_single(xprt->sc_cm_id->device, 331 dma_map_xdr(xprt, &rqstp->rq_res, xdr_off,
297 (void *) 332 sge_bytes, DMA_TO_DEVICE);
298 vec->sge[xdr_sge_no].iov_base + sge_off, 333 xdr_off += sge_bytes;
299 sge_bytes, DMA_TO_DEVICE);
300 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 334 if (ib_dma_mapping_error(xprt->sc_cm_id->device,
301 sge[sge_no].addr)) 335 sge[sge_no].addr))
302 goto err; 336 goto err;
@@ -333,6 +367,8 @@ static int send_write(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp,
333 goto err; 367 goto err;
334 return 0; 368 return 0;
335 err: 369 err:
370 svc_rdma_unmap_dma(ctxt);
371 svc_rdma_put_frmr(xprt, vec->frmr);
336 svc_rdma_put_context(ctxt, 0); 372 svc_rdma_put_context(ctxt, 0);
337 /* Fatal error, close transport */ 373 /* Fatal error, close transport */
338 return -EIO; 374 return -EIO;
@@ -494,7 +530,8 @@ static int send_reply_chunks(struct svcxprt_rdma *xprt,
494 * In all three cases, this function prepares the RPCRDMA header in 530 * In all three cases, this function prepares the RPCRDMA header in
495 * sge[0], the 'type' parameter indicates the type to place in the 531 * sge[0], the 'type' parameter indicates the type to place in the
496 * RPCRDMA header, and the 'byte_count' field indicates how much of 532 * RPCRDMA header, and the 'byte_count' field indicates how much of
497 * the XDR to include in this RDMA_SEND. 533 * the XDR to include in this RDMA_SEND. NB: The offset of the payload
534 * to send is zero in the XDR.
498 */ 535 */
499static int send_reply(struct svcxprt_rdma *rdma, 536static int send_reply(struct svcxprt_rdma *rdma,
500 struct svc_rqst *rqstp, 537 struct svc_rqst *rqstp,
@@ -536,23 +573,24 @@ static int send_reply(struct svcxprt_rdma *rdma,
536 ctxt->sge[0].lkey = rdma->sc_dma_lkey; 573 ctxt->sge[0].lkey = rdma->sc_dma_lkey;
537 ctxt->sge[0].length = svc_rdma_xdr_get_reply_hdr_len(rdma_resp); 574 ctxt->sge[0].length = svc_rdma_xdr_get_reply_hdr_len(rdma_resp);
538 ctxt->sge[0].addr = 575 ctxt->sge[0].addr =
539 ib_dma_map_single(rdma->sc_cm_id->device, page_address(page), 576 ib_dma_map_page(rdma->sc_cm_id->device, page, 0,
540 ctxt->sge[0].length, DMA_TO_DEVICE); 577 ctxt->sge[0].length, DMA_TO_DEVICE);
541 if (ib_dma_mapping_error(rdma->sc_cm_id->device, ctxt->sge[0].addr)) 578 if (ib_dma_mapping_error(rdma->sc_cm_id->device, ctxt->sge[0].addr))
542 goto err; 579 goto err;
543 atomic_inc(&rdma->sc_dma_used); 580 atomic_inc(&rdma->sc_dma_used);
544 581
545 ctxt->direction = DMA_TO_DEVICE; 582 ctxt->direction = DMA_TO_DEVICE;
546 583
547 /* Determine how many of our SGE are to be transmitted */ 584 /* Map the payload indicated by 'byte_count' */
548 for (sge_no = 1; byte_count && sge_no < vec->count; sge_no++) { 585 for (sge_no = 1; byte_count && sge_no < vec->count; sge_no++) {
586 int xdr_off = 0;
549 sge_bytes = min_t(size_t, vec->sge[sge_no].iov_len, byte_count); 587 sge_bytes = min_t(size_t, vec->sge[sge_no].iov_len, byte_count);
550 byte_count -= sge_bytes; 588 byte_count -= sge_bytes;
551 if (!vec->frmr) { 589 if (!vec->frmr) {
552 ctxt->sge[sge_no].addr = 590 ctxt->sge[sge_no].addr =
553 ib_dma_map_single(rdma->sc_cm_id->device, 591 dma_map_xdr(rdma, &rqstp->rq_res, xdr_off,
554 vec->sge[sge_no].iov_base, 592 sge_bytes, DMA_TO_DEVICE);
555 sge_bytes, DMA_TO_DEVICE); 593 xdr_off += sge_bytes;
556 if (ib_dma_mapping_error(rdma->sc_cm_id->device, 594 if (ib_dma_mapping_error(rdma->sc_cm_id->device,
557 ctxt->sge[sge_no].addr)) 595 ctxt->sge[sge_no].addr))
558 goto err; 596 goto err;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index edea15a54e51..9df1eadc912a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -45,6 +45,7 @@
45#include <linux/sched.h> 45#include <linux/sched.h>
46#include <linux/slab.h> 46#include <linux/slab.h>
47#include <linux/spinlock.h> 47#include <linux/spinlock.h>
48#include <linux/workqueue.h>
48#include <rdma/ib_verbs.h> 49#include <rdma/ib_verbs.h>
49#include <rdma/rdma_cm.h> 50#include <rdma/rdma_cm.h>
50#include <linux/sunrpc/svc_rdma.h> 51#include <linux/sunrpc/svc_rdma.h>
@@ -52,6 +53,7 @@
52#define RPCDBG_FACILITY RPCDBG_SVCXPRT 53#define RPCDBG_FACILITY RPCDBG_SVCXPRT
53 54
54static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, 55static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
56 struct net *net,
55 struct sockaddr *sa, int salen, 57 struct sockaddr *sa, int salen,
56 int flags); 58 int flags);
57static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt); 59static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt);
@@ -89,6 +91,9 @@ struct svc_xprt_class svc_rdma_class = {
89/* WR context cache. Created in svc_rdma.c */ 91/* WR context cache. Created in svc_rdma.c */
90extern struct kmem_cache *svc_rdma_ctxt_cachep; 92extern struct kmem_cache *svc_rdma_ctxt_cachep;
91 93
94/* Workqueue created in svc_rdma.c */
95extern struct workqueue_struct *svc_rdma_wq;
96
92struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *xprt) 97struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *xprt)
93{ 98{
94 struct svc_rdma_op_ctxt *ctxt; 99 struct svc_rdma_op_ctxt *ctxt;
@@ -120,7 +125,7 @@ void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt)
120 */ 125 */
121 if (ctxt->sge[i].lkey == xprt->sc_dma_lkey) { 126 if (ctxt->sge[i].lkey == xprt->sc_dma_lkey) {
122 atomic_dec(&xprt->sc_dma_used); 127 atomic_dec(&xprt->sc_dma_used);
123 ib_dma_unmap_single(xprt->sc_cm_id->device, 128 ib_dma_unmap_page(xprt->sc_cm_id->device,
124 ctxt->sge[i].addr, 129 ctxt->sge[i].addr,
125 ctxt->sge[i].length, 130 ctxt->sge[i].length,
126 ctxt->direction); 131 ctxt->direction);
@@ -502,8 +507,8 @@ int svc_rdma_post_recv(struct svcxprt_rdma *xprt)
502 BUG_ON(sge_no >= xprt->sc_max_sge); 507 BUG_ON(sge_no >= xprt->sc_max_sge);
503 page = svc_rdma_get_page(); 508 page = svc_rdma_get_page();
504 ctxt->pages[sge_no] = page; 509 ctxt->pages[sge_no] = page;
505 pa = ib_dma_map_single(xprt->sc_cm_id->device, 510 pa = ib_dma_map_page(xprt->sc_cm_id->device,
506 page_address(page), PAGE_SIZE, 511 page, 0, PAGE_SIZE,
507 DMA_FROM_DEVICE); 512 DMA_FROM_DEVICE);
508 if (ib_dma_mapping_error(xprt->sc_cm_id->device, pa)) 513 if (ib_dma_mapping_error(xprt->sc_cm_id->device, pa))
509 goto err_put_ctxt; 514 goto err_put_ctxt;
@@ -511,9 +516,9 @@ int svc_rdma_post_recv(struct svcxprt_rdma *xprt)
511 ctxt->sge[sge_no].addr = pa; 516 ctxt->sge[sge_no].addr = pa;
512 ctxt->sge[sge_no].length = PAGE_SIZE; 517 ctxt->sge[sge_no].length = PAGE_SIZE;
513 ctxt->sge[sge_no].lkey = xprt->sc_dma_lkey; 518 ctxt->sge[sge_no].lkey = xprt->sc_dma_lkey;
519 ctxt->count = sge_no + 1;
514 buflen += PAGE_SIZE; 520 buflen += PAGE_SIZE;
515 } 521 }
516 ctxt->count = sge_no;
517 recv_wr.next = NULL; 522 recv_wr.next = NULL;
518 recv_wr.sg_list = &ctxt->sge[0]; 523 recv_wr.sg_list = &ctxt->sge[0];
519 recv_wr.num_sge = ctxt->count; 524 recv_wr.num_sge = ctxt->count;
@@ -529,6 +534,7 @@ int svc_rdma_post_recv(struct svcxprt_rdma *xprt)
529 return ret; 534 return ret;
530 535
531 err_put_ctxt: 536 err_put_ctxt:
537 svc_rdma_unmap_dma(ctxt);
532 svc_rdma_put_context(ctxt, 1); 538 svc_rdma_put_context(ctxt, 1);
533 return -ENOMEM; 539 return -ENOMEM;
534} 540}
@@ -670,6 +676,7 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
670 * Create a listening RDMA service endpoint. 676 * Create a listening RDMA service endpoint.
671 */ 677 */
672static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, 678static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
679 struct net *net,
673 struct sockaddr *sa, int salen, 680 struct sockaddr *sa, int salen,
674 int flags) 681 int flags)
675{ 682{
@@ -798,8 +805,8 @@ static void frmr_unmap_dma(struct svcxprt_rdma *xprt,
798 if (ib_dma_mapping_error(frmr->mr->device, addr)) 805 if (ib_dma_mapping_error(frmr->mr->device, addr))
799 continue; 806 continue;
800 atomic_dec(&xprt->sc_dma_used); 807 atomic_dec(&xprt->sc_dma_used);
801 ib_dma_unmap_single(frmr->mr->device, addr, PAGE_SIZE, 808 ib_dma_unmap_page(frmr->mr->device, addr, PAGE_SIZE,
802 frmr->direction); 809 frmr->direction);
803 } 810 }
804} 811}
805 812
@@ -1184,7 +1191,7 @@ static void svc_rdma_free(struct svc_xprt *xprt)
1184 struct svcxprt_rdma *rdma = 1191 struct svcxprt_rdma *rdma =
1185 container_of(xprt, struct svcxprt_rdma, sc_xprt); 1192 container_of(xprt, struct svcxprt_rdma, sc_xprt);
1186 INIT_WORK(&rdma->sc_work, __svc_rdma_free); 1193 INIT_WORK(&rdma->sc_work, __svc_rdma_free);
1187 schedule_work(&rdma->sc_work); 1194 queue_work(svc_rdma_wq, &rdma->sc_work);
1188} 1195}
1189 1196
1190static int svc_rdma_has_wspace(struct svc_xprt *xprt) 1197static int svc_rdma_has_wspace(struct svc_xprt *xprt)
@@ -1274,7 +1281,7 @@ int svc_rdma_send(struct svcxprt_rdma *xprt, struct ib_send_wr *wr)
1274 atomic_read(&xprt->sc_sq_count) < 1281 atomic_read(&xprt->sc_sq_count) <
1275 xprt->sc_sq_depth); 1282 xprt->sc_sq_depth);
1276 if (test_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags)) 1283 if (test_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags))
1277 return 0; 1284 return -ENOTCONN;
1278 continue; 1285 continue;
1279 } 1286 }
1280 /* Take a transport ref for each WR posted */ 1287 /* Take a transport ref for each WR posted */
@@ -1306,7 +1313,6 @@ void svc_rdma_send_error(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rmsgp,
1306 enum rpcrdma_errcode err) 1313 enum rpcrdma_errcode err)
1307{ 1314{
1308 struct ib_send_wr err_wr; 1315 struct ib_send_wr err_wr;
1309 struct ib_sge sge;
1310 struct page *p; 1316 struct page *p;
1311 struct svc_rdma_op_ctxt *ctxt; 1317 struct svc_rdma_op_ctxt *ctxt;
1312 u32 *va; 1318 u32 *va;
@@ -1319,26 +1325,27 @@ void svc_rdma_send_error(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rmsgp,
1319 /* XDR encode error */ 1325 /* XDR encode error */
1320 length = svc_rdma_xdr_encode_error(xprt, rmsgp, err, va); 1326 length = svc_rdma_xdr_encode_error(xprt, rmsgp, err, va);
1321 1327
1328 ctxt = svc_rdma_get_context(xprt);
1329 ctxt->direction = DMA_FROM_DEVICE;
1330 ctxt->count = 1;
1331 ctxt->pages[0] = p;
1332
1322 /* Prepare SGE for local address */ 1333 /* Prepare SGE for local address */
1323 sge.addr = ib_dma_map_single(xprt->sc_cm_id->device, 1334 ctxt->sge[0].addr = ib_dma_map_page(xprt->sc_cm_id->device,
1324 page_address(p), PAGE_SIZE, DMA_FROM_DEVICE); 1335 p, 0, length, DMA_FROM_DEVICE);
1325 if (ib_dma_mapping_error(xprt->sc_cm_id->device, sge.addr)) { 1336 if (ib_dma_mapping_error(xprt->sc_cm_id->device, ctxt->sge[0].addr)) {
1326 put_page(p); 1337 put_page(p);
1327 return; 1338 return;
1328 } 1339 }
1329 atomic_inc(&xprt->sc_dma_used); 1340 atomic_inc(&xprt->sc_dma_used);
1330 sge.lkey = xprt->sc_dma_lkey; 1341 ctxt->sge[0].lkey = xprt->sc_dma_lkey;
1331 sge.length = length; 1342 ctxt->sge[0].length = length;
1332
1333 ctxt = svc_rdma_get_context(xprt);
1334 ctxt->count = 1;
1335 ctxt->pages[0] = p;
1336 1343
1337 /* Prepare SEND WR */ 1344 /* Prepare SEND WR */
1338 memset(&err_wr, 0, sizeof err_wr); 1345 memset(&err_wr, 0, sizeof err_wr);
1339 ctxt->wr_op = IB_WR_SEND; 1346 ctxt->wr_op = IB_WR_SEND;
1340 err_wr.wr_id = (unsigned long)ctxt; 1347 err_wr.wr_id = (unsigned long)ctxt;
1341 err_wr.sg_list = &sge; 1348 err_wr.sg_list = ctxt->sge;
1342 err_wr.num_sge = 1; 1349 err_wr.num_sge = 1;
1343 err_wr.opcode = IB_WR_SEND; 1350 err_wr.opcode = IB_WR_SEND;
1344 err_wr.send_flags = IB_SEND_SIGNALED; 1351 err_wr.send_flags = IB_SEND_SIGNALED;
@@ -1348,9 +1355,7 @@ void svc_rdma_send_error(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rmsgp,
1348 if (ret) { 1355 if (ret) {
1349 dprintk("svcrdma: Error %d posting send for protocol error\n", 1356 dprintk("svcrdma: Error %d posting send for protocol error\n",
1350 ret); 1357 ret);
1351 ib_dma_unmap_single(xprt->sc_cm_id->device, 1358 svc_rdma_unmap_dma(ctxt);
1352 sge.addr, PAGE_SIZE,
1353 DMA_FROM_DEVICE);
1354 svc_rdma_put_context(ctxt, 1); 1359 svc_rdma_put_context(ctxt, 1);
1355 } 1360 }
1356} 1361}
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index a85e866a77f7..0867070bb5ca 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -237,8 +237,7 @@ xprt_rdma_destroy(struct rpc_xprt *xprt)
237 237
238 dprintk("RPC: %s: called\n", __func__); 238 dprintk("RPC: %s: called\n", __func__);
239 239
240 cancel_delayed_work(&r_xprt->rdma_connect); 240 cancel_delayed_work_sync(&r_xprt->rdma_connect);
241 flush_scheduled_work();
242 241
243 xprt_clear_connected(xprt); 242 xprt_clear_connected(xprt);
244 243
@@ -251,9 +250,7 @@ xprt_rdma_destroy(struct rpc_xprt *xprt)
251 250
252 xprt_rdma_free_addresses(xprt); 251 xprt_rdma_free_addresses(xprt);
253 252
254 kfree(xprt->slot); 253 xprt_free(xprt);
255 xprt->slot = NULL;
256 kfree(xprt);
257 254
258 dprintk("RPC: %s: returning\n", __func__); 255 dprintk("RPC: %s: returning\n", __func__);
259 256
@@ -285,23 +282,14 @@ xprt_setup_rdma(struct xprt_create *args)
285 return ERR_PTR(-EBADF); 282 return ERR_PTR(-EBADF);
286 } 283 }
287 284
288 xprt = kzalloc(sizeof(struct rpcrdma_xprt), GFP_KERNEL); 285 xprt = xprt_alloc(args->net, sizeof(struct rpcrdma_xprt),
286 xprt_rdma_slot_table_entries);
289 if (xprt == NULL) { 287 if (xprt == NULL) {
290 dprintk("RPC: %s: couldn't allocate rpcrdma_xprt\n", 288 dprintk("RPC: %s: couldn't allocate rpcrdma_xprt\n",
291 __func__); 289 __func__);
292 return ERR_PTR(-ENOMEM); 290 return ERR_PTR(-ENOMEM);
293 } 291 }
294 292
295 xprt->max_reqs = xprt_rdma_slot_table_entries;
296 xprt->slot = kcalloc(xprt->max_reqs,
297 sizeof(struct rpc_rqst), GFP_KERNEL);
298 if (xprt->slot == NULL) {
299 dprintk("RPC: %s: couldn't allocate %d slots\n",
300 __func__, xprt->max_reqs);
301 kfree(xprt);
302 return ERR_PTR(-ENOMEM);
303 }
304
305 /* 60 second timeout, no retries */ 293 /* 60 second timeout, no retries */
306 xprt->timeout = &xprt_rdma_default_timeout; 294 xprt->timeout = &xprt_rdma_default_timeout;
307 xprt->bind_timeout = (60U * HZ); 295 xprt->bind_timeout = (60U * HZ);
@@ -410,8 +398,7 @@ out3:
410out2: 398out2:
411 rpcrdma_ia_close(&new_xprt->rx_ia); 399 rpcrdma_ia_close(&new_xprt->rx_ia);
412out1: 400out1:
413 kfree(xprt->slot); 401 xprt_free(xprt);
414 kfree(xprt);
415 return ERR_PTR(rc); 402 return ERR_PTR(rc);
416} 403}
417 404
@@ -460,7 +447,7 @@ xprt_rdma_connect(struct rpc_task *task)
460 } else { 447 } else {
461 schedule_delayed_work(&r_xprt->rdma_connect, 0); 448 schedule_delayed_work(&r_xprt->rdma_connect, 0);
462 if (!RPC_IS_ASYNC(task)) 449 if (!RPC_IS_ASYNC(task))
463 flush_scheduled_work(); 450 flush_delayed_work(&r_xprt->rdma_connect);
464 } 451 }
465} 452}
466 453
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index fe9306bf10cc..dfcab5ac65af 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -774,8 +774,7 @@ static void xs_destroy(struct rpc_xprt *xprt)
774 774
775 xs_close(xprt); 775 xs_close(xprt);
776 xs_free_peer_addresses(xprt); 776 xs_free_peer_addresses(xprt);
777 kfree(xprt->slot); 777 xprt_free(xprt);
778 kfree(xprt);
779 module_put(THIS_MODULE); 778 module_put(THIS_MODULE);
780} 779}
781 780
@@ -1516,7 +1515,7 @@ static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1516 xs_update_peer_port(xprt); 1515 xs_update_peer_port(xprt);
1517} 1516}
1518 1517
1519static unsigned short xs_get_srcport(struct sock_xprt *transport, struct socket *sock) 1518static unsigned short xs_get_srcport(struct sock_xprt *transport)
1520{ 1519{
1521 unsigned short port = transport->srcport; 1520 unsigned short port = transport->srcport;
1522 1521
@@ -1525,7 +1524,7 @@ static unsigned short xs_get_srcport(struct sock_xprt *transport, struct socket
1525 return port; 1524 return port;
1526} 1525}
1527 1526
1528static unsigned short xs_next_srcport(struct sock_xprt *transport, struct socket *sock, unsigned short port) 1527static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
1529{ 1528{
1530 if (transport->srcport != 0) 1529 if (transport->srcport != 0)
1531 transport->srcport = 0; 1530 transport->srcport = 0;
@@ -1535,23 +1534,18 @@ static unsigned short xs_next_srcport(struct sock_xprt *transport, struct socket
1535 return xprt_max_resvport; 1534 return xprt_max_resvport;
1536 return --port; 1535 return --port;
1537} 1536}
1538 1537static int xs_bind(struct sock_xprt *transport, struct socket *sock)
1539static int xs_bind4(struct sock_xprt *transport, struct socket *sock)
1540{ 1538{
1541 struct sockaddr_in myaddr = { 1539 struct sockaddr_storage myaddr;
1542 .sin_family = AF_INET,
1543 };
1544 struct sockaddr_in *sa;
1545 int err, nloop = 0; 1540 int err, nloop = 0;
1546 unsigned short port = xs_get_srcport(transport, sock); 1541 unsigned short port = xs_get_srcport(transport);
1547 unsigned short last; 1542 unsigned short last;
1548 1543
1549 sa = (struct sockaddr_in *)&transport->srcaddr; 1544 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
1550 myaddr.sin_addr = sa->sin_addr;
1551 do { 1545 do {
1552 myaddr.sin_port = htons(port); 1546 rpc_set_port((struct sockaddr *)&myaddr, port);
1553 err = kernel_bind(sock, (struct sockaddr *) &myaddr, 1547 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1554 sizeof(myaddr)); 1548 transport->xprt.addrlen);
1555 if (port == 0) 1549 if (port == 0)
1556 break; 1550 break;
1557 if (err == 0) { 1551 if (err == 0) {
@@ -1559,48 +1553,23 @@ static int xs_bind4(struct sock_xprt *transport, struct socket *sock)
1559 break; 1553 break;
1560 } 1554 }
1561 last = port; 1555 last = port;
1562 port = xs_next_srcport(transport, sock, port); 1556 port = xs_next_srcport(transport, port);
1563 if (port > last) 1557 if (port > last)
1564 nloop++; 1558 nloop++;
1565 } while (err == -EADDRINUSE && nloop != 2); 1559 } while (err == -EADDRINUSE && nloop != 2);
1566 dprintk("RPC: %s %pI4:%u: %s (%d)\n",
1567 __func__, &myaddr.sin_addr,
1568 port, err ? "failed" : "ok", err);
1569 return err;
1570}
1571
1572static int xs_bind6(struct sock_xprt *transport, struct socket *sock)
1573{
1574 struct sockaddr_in6 myaddr = {
1575 .sin6_family = AF_INET6,
1576 };
1577 struct sockaddr_in6 *sa;
1578 int err, nloop = 0;
1579 unsigned short port = xs_get_srcport(transport, sock);
1580 unsigned short last;
1581 1560
1582 sa = (struct sockaddr_in6 *)&transport->srcaddr; 1561 if (myaddr.ss_family == AF_INET)
1583 myaddr.sin6_addr = sa->sin6_addr; 1562 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1584 do { 1563 &((struct sockaddr_in *)&myaddr)->sin_addr,
1585 myaddr.sin6_port = htons(port); 1564 port, err ? "failed" : "ok", err);
1586 err = kernel_bind(sock, (struct sockaddr *) &myaddr, 1565 else
1587 sizeof(myaddr)); 1566 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1588 if (port == 0) 1567 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1589 break; 1568 port, err ? "failed" : "ok", err);
1590 if (err == 0) {
1591 transport->srcport = port;
1592 break;
1593 }
1594 last = port;
1595 port = xs_next_srcport(transport, sock, port);
1596 if (port > last)
1597 nloop++;
1598 } while (err == -EADDRINUSE && nloop != 2);
1599 dprintk("RPC: xs_bind6 %pI6:%u: %s (%d)\n",
1600 &myaddr.sin6_addr, port, err ? "failed" : "ok", err);
1601 return err; 1569 return err;
1602} 1570}
1603 1571
1572
1604#ifdef CONFIG_DEBUG_LOCK_ALLOC 1573#ifdef CONFIG_DEBUG_LOCK_ALLOC
1605static struct lock_class_key xs_key[2]; 1574static struct lock_class_key xs_key[2];
1606static struct lock_class_key xs_slock_key[2]; 1575static struct lock_class_key xs_slock_key[2];
@@ -1622,6 +1591,18 @@ static inline void xs_reclassify_socket6(struct socket *sock)
1622 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", 1591 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1623 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); 1592 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
1624} 1593}
1594
1595static inline void xs_reclassify_socket(int family, struct socket *sock)
1596{
1597 switch (family) {
1598 case AF_INET:
1599 xs_reclassify_socket4(sock);
1600 break;
1601 case AF_INET6:
1602 xs_reclassify_socket6(sock);
1603 break;
1604 }
1605}
1625#else 1606#else
1626static inline void xs_reclassify_socket4(struct socket *sock) 1607static inline void xs_reclassify_socket4(struct socket *sock)
1627{ 1608{
@@ -1630,8 +1611,36 @@ static inline void xs_reclassify_socket4(struct socket *sock)
1630static inline void xs_reclassify_socket6(struct socket *sock) 1611static inline void xs_reclassify_socket6(struct socket *sock)
1631{ 1612{
1632} 1613}
1614
1615static inline void xs_reclassify_socket(int family, struct socket *sock)
1616{
1617}
1633#endif 1618#endif
1634 1619
1620static struct socket *xs_create_sock(struct rpc_xprt *xprt,
1621 struct sock_xprt *transport, int family, int type, int protocol)
1622{
1623 struct socket *sock;
1624 int err;
1625
1626 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
1627 if (err < 0) {
1628 dprintk("RPC: can't create %d transport socket (%d).\n",
1629 protocol, -err);
1630 goto out;
1631 }
1632 xs_reclassify_socket(family, sock);
1633
1634 if (xs_bind(transport, sock)) {
1635 sock_release(sock);
1636 goto out;
1637 }
1638
1639 return sock;
1640out:
1641 return ERR_PTR(err);
1642}
1643
1635static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1644static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1636{ 1645{
1637 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); 1646 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
@@ -1661,82 +1670,23 @@ static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1661 xs_udp_do_set_buffer_size(xprt); 1670 xs_udp_do_set_buffer_size(xprt);
1662} 1671}
1663 1672
1664/** 1673static void xs_udp_setup_socket(struct work_struct *work)
1665 * xs_udp_connect_worker4 - set up a UDP socket
1666 * @work: RPC transport to connect
1667 *
1668 * Invoked by a work queue tasklet.
1669 */
1670static void xs_udp_connect_worker4(struct work_struct *work)
1671{ 1674{
1672 struct sock_xprt *transport = 1675 struct sock_xprt *transport =
1673 container_of(work, struct sock_xprt, connect_worker.work); 1676 container_of(work, struct sock_xprt, connect_worker.work);
1674 struct rpc_xprt *xprt = &transport->xprt; 1677 struct rpc_xprt *xprt = &transport->xprt;
1675 struct socket *sock = transport->sock; 1678 struct socket *sock = transport->sock;
1676 int err, status = -EIO; 1679 int status = -EIO;
1677 1680
1678 if (xprt->shutdown) 1681 if (xprt->shutdown)
1679 goto out; 1682 goto out;
1680 1683
1681 /* Start by resetting any existing state */ 1684 /* Start by resetting any existing state */
1682 xs_reset_transport(transport); 1685 xs_reset_transport(transport);
1683 1686 sock = xs_create_sock(xprt, transport,
1684 err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock); 1687 xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
1685 if (err < 0) { 1688 if (IS_ERR(sock))
1686 dprintk("RPC: can't create UDP transport socket (%d).\n", -err);
1687 goto out; 1689 goto out;
1688 }
1689 xs_reclassify_socket4(sock);
1690
1691 if (xs_bind4(transport, sock)) {
1692 sock_release(sock);
1693 goto out;
1694 }
1695
1696 dprintk("RPC: worker connecting xprt %p via %s to "
1697 "%s (port %s)\n", xprt,
1698 xprt->address_strings[RPC_DISPLAY_PROTO],
1699 xprt->address_strings[RPC_DISPLAY_ADDR],
1700 xprt->address_strings[RPC_DISPLAY_PORT]);
1701
1702 xs_udp_finish_connecting(xprt, sock);
1703 status = 0;
1704out:
1705 xprt_clear_connecting(xprt);
1706 xprt_wake_pending_tasks(xprt, status);
1707}
1708
1709/**
1710 * xs_udp_connect_worker6 - set up a UDP socket
1711 * @work: RPC transport to connect
1712 *
1713 * Invoked by a work queue tasklet.
1714 */
1715static void xs_udp_connect_worker6(struct work_struct *work)
1716{
1717 struct sock_xprt *transport =
1718 container_of(work, struct sock_xprt, connect_worker.work);
1719 struct rpc_xprt *xprt = &transport->xprt;
1720 struct socket *sock = transport->sock;
1721 int err, status = -EIO;
1722
1723 if (xprt->shutdown)
1724 goto out;
1725
1726 /* Start by resetting any existing state */
1727 xs_reset_transport(transport);
1728
1729 err = sock_create_kern(PF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
1730 if (err < 0) {
1731 dprintk("RPC: can't create UDP transport socket (%d).\n", -err);
1732 goto out;
1733 }
1734 xs_reclassify_socket6(sock);
1735
1736 if (xs_bind6(transport, sock) < 0) {
1737 sock_release(sock);
1738 goto out;
1739 }
1740 1690
1741 dprintk("RPC: worker connecting xprt %p via %s to " 1691 dprintk("RPC: worker connecting xprt %p via %s to "
1742 "%s (port %s)\n", xprt, 1692 "%s (port %s)\n", xprt,
@@ -1755,12 +1705,12 @@ out:
1755 * We need to preserve the port number so the reply cache on the server can 1705 * We need to preserve the port number so the reply cache on the server can
1756 * find our cached RPC replies when we get around to reconnecting. 1706 * find our cached RPC replies when we get around to reconnecting.
1757 */ 1707 */
1758static void xs_abort_connection(struct rpc_xprt *xprt, struct sock_xprt *transport) 1708static void xs_abort_connection(struct sock_xprt *transport)
1759{ 1709{
1760 int result; 1710 int result;
1761 struct sockaddr any; 1711 struct sockaddr any;
1762 1712
1763 dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); 1713 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport);
1764 1714
1765 /* 1715 /*
1766 * Disconnect the transport socket by doing a connect operation 1716 * Disconnect the transport socket by doing a connect operation
@@ -1770,13 +1720,13 @@ static void xs_abort_connection(struct rpc_xprt *xprt, struct sock_xprt *transpo
1770 any.sa_family = AF_UNSPEC; 1720 any.sa_family = AF_UNSPEC;
1771 result = kernel_connect(transport->sock, &any, sizeof(any), 0); 1721 result = kernel_connect(transport->sock, &any, sizeof(any), 0);
1772 if (!result) 1722 if (!result)
1773 xs_sock_mark_closed(xprt); 1723 xs_sock_mark_closed(&transport->xprt);
1774 else 1724 else
1775 dprintk("RPC: AF_UNSPEC connect return code %d\n", 1725 dprintk("RPC: AF_UNSPEC connect return code %d\n",
1776 result); 1726 result);
1777} 1727}
1778 1728
1779static void xs_tcp_reuse_connection(struct rpc_xprt *xprt, struct sock_xprt *transport) 1729static void xs_tcp_reuse_connection(struct sock_xprt *transport)
1780{ 1730{
1781 unsigned int state = transport->inet->sk_state; 1731 unsigned int state = transport->inet->sk_state;
1782 1732
@@ -1799,7 +1749,7 @@ static void xs_tcp_reuse_connection(struct rpc_xprt *xprt, struct sock_xprt *tra
1799 "sk_shutdown set to %d\n", 1749 "sk_shutdown set to %d\n",
1800 __func__, transport->inet->sk_shutdown); 1750 __func__, transport->inet->sk_shutdown);
1801 } 1751 }
1802 xs_abort_connection(xprt, transport); 1752 xs_abort_connection(transport);
1803} 1753}
1804 1754
1805static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) 1755static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
@@ -1852,12 +1802,12 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
1852 * 1802 *
1853 * Invoked by a work queue tasklet. 1803 * Invoked by a work queue tasklet.
1854 */ 1804 */
1855static void xs_tcp_setup_socket(struct rpc_xprt *xprt, 1805static void xs_tcp_setup_socket(struct work_struct *work)
1856 struct sock_xprt *transport,
1857 struct socket *(*create_sock)(struct rpc_xprt *,
1858 struct sock_xprt *))
1859{ 1806{
1807 struct sock_xprt *transport =
1808 container_of(work, struct sock_xprt, connect_worker.work);
1860 struct socket *sock = transport->sock; 1809 struct socket *sock = transport->sock;
1810 struct rpc_xprt *xprt = &transport->xprt;
1861 int status = -EIO; 1811 int status = -EIO;
1862 1812
1863 if (xprt->shutdown) 1813 if (xprt->shutdown)
@@ -1865,7 +1815,8 @@ static void xs_tcp_setup_socket(struct rpc_xprt *xprt,
1865 1815
1866 if (!sock) { 1816 if (!sock) {
1867 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); 1817 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1868 sock = create_sock(xprt, transport); 1818 sock = xs_create_sock(xprt, transport,
1819 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
1869 if (IS_ERR(sock)) { 1820 if (IS_ERR(sock)) {
1870 status = PTR_ERR(sock); 1821 status = PTR_ERR(sock);
1871 goto out; 1822 goto out;
@@ -1876,7 +1827,7 @@ static void xs_tcp_setup_socket(struct rpc_xprt *xprt,
1876 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT, 1827 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
1877 &xprt->state); 1828 &xprt->state);
1878 /* "close" the socket, preserving the local port */ 1829 /* "close" the socket, preserving the local port */
1879 xs_tcp_reuse_connection(xprt, transport); 1830 xs_tcp_reuse_connection(transport);
1880 1831
1881 if (abort_and_exit) 1832 if (abort_and_exit)
1882 goto out_eagain; 1833 goto out_eagain;
@@ -1925,84 +1876,6 @@ out:
1925 xprt_wake_pending_tasks(xprt, status); 1876 xprt_wake_pending_tasks(xprt, status);
1926} 1877}
1927 1878
1928static struct socket *xs_create_tcp_sock4(struct rpc_xprt *xprt,
1929 struct sock_xprt *transport)
1930{
1931 struct socket *sock;
1932 int err;
1933
1934 /* start from scratch */
1935 err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
1936 if (err < 0) {
1937 dprintk("RPC: can't create TCP transport socket (%d).\n",
1938 -err);
1939 goto out_err;
1940 }
1941 xs_reclassify_socket4(sock);
1942
1943 if (xs_bind4(transport, sock) < 0) {
1944 sock_release(sock);
1945 goto out_err;
1946 }
1947 return sock;
1948out_err:
1949 return ERR_PTR(-EIO);
1950}
1951
1952/**
1953 * xs_tcp_connect_worker4 - connect a TCP socket to a remote endpoint
1954 * @work: RPC transport to connect
1955 *
1956 * Invoked by a work queue tasklet.
1957 */
1958static void xs_tcp_connect_worker4(struct work_struct *work)
1959{
1960 struct sock_xprt *transport =
1961 container_of(work, struct sock_xprt, connect_worker.work);
1962 struct rpc_xprt *xprt = &transport->xprt;
1963
1964 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock4);
1965}
1966
1967static struct socket *xs_create_tcp_sock6(struct rpc_xprt *xprt,
1968 struct sock_xprt *transport)
1969{
1970 struct socket *sock;
1971 int err;
1972
1973 /* start from scratch */
1974 err = sock_create_kern(PF_INET6, SOCK_STREAM, IPPROTO_TCP, &sock);
1975 if (err < 0) {
1976 dprintk("RPC: can't create TCP transport socket (%d).\n",
1977 -err);
1978 goto out_err;
1979 }
1980 xs_reclassify_socket6(sock);
1981
1982 if (xs_bind6(transport, sock) < 0) {
1983 sock_release(sock);
1984 goto out_err;
1985 }
1986 return sock;
1987out_err:
1988 return ERR_PTR(-EIO);
1989}
1990
1991/**
1992 * xs_tcp_connect_worker6 - connect a TCP socket to a remote endpoint
1993 * @work: RPC transport to connect
1994 *
1995 * Invoked by a work queue tasklet.
1996 */
1997static void xs_tcp_connect_worker6(struct work_struct *work)
1998{
1999 struct sock_xprt *transport =
2000 container_of(work, struct sock_xprt, connect_worker.work);
2001 struct rpc_xprt *xprt = &transport->xprt;
2002
2003 xs_tcp_setup_socket(xprt, transport, xs_create_tcp_sock6);
2004}
2005
2006/** 1879/**
2007 * xs_connect - connect a socket to a remote endpoint 1880 * xs_connect - connect a socket to a remote endpoint
2008 * @task: address of RPC task that manages state of connect request 1881 * @task: address of RPC task that manages state of connect request
@@ -2262,6 +2135,31 @@ static struct rpc_xprt_ops bc_tcp_ops = {
2262 .print_stats = xs_tcp_print_stats, 2135 .print_stats = xs_tcp_print_stats,
2263}; 2136};
2264 2137
2138static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2139{
2140 static const struct sockaddr_in sin = {
2141 .sin_family = AF_INET,
2142 .sin_addr.s_addr = htonl(INADDR_ANY),
2143 };
2144 static const struct sockaddr_in6 sin6 = {
2145 .sin6_family = AF_INET6,
2146 .sin6_addr = IN6ADDR_ANY_INIT,
2147 };
2148
2149 switch (family) {
2150 case AF_INET:
2151 memcpy(sap, &sin, sizeof(sin));
2152 break;
2153 case AF_INET6:
2154 memcpy(sap, &sin6, sizeof(sin6));
2155 break;
2156 default:
2157 dprintk("RPC: %s: Bad address family\n", __func__);
2158 return -EAFNOSUPPORT;
2159 }
2160 return 0;
2161}
2162
2265static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, 2163static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
2266 unsigned int slot_table_size) 2164 unsigned int slot_table_size)
2267{ 2165{
@@ -2273,27 +2171,25 @@ static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
2273 return ERR_PTR(-EBADF); 2171 return ERR_PTR(-EBADF);
2274 } 2172 }
2275 2173
2276 new = kzalloc(sizeof(*new), GFP_KERNEL); 2174 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size);
2277 if (new == NULL) { 2175 if (xprt == NULL) {
2278 dprintk("RPC: xs_setup_xprt: couldn't allocate " 2176 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2279 "rpc_xprt\n"); 2177 "rpc_xprt\n");
2280 return ERR_PTR(-ENOMEM); 2178 return ERR_PTR(-ENOMEM);
2281 } 2179 }
2282 xprt = &new->xprt;
2283
2284 xprt->max_reqs = slot_table_size;
2285 xprt->slot = kcalloc(xprt->max_reqs, sizeof(struct rpc_rqst), GFP_KERNEL);
2286 if (xprt->slot == NULL) {
2287 kfree(xprt);
2288 dprintk("RPC: xs_setup_xprt: couldn't allocate slot "
2289 "table\n");
2290 return ERR_PTR(-ENOMEM);
2291 }
2292 2180
2181 new = container_of(xprt, struct sock_xprt, xprt);
2293 memcpy(&xprt->addr, args->dstaddr, args->addrlen); 2182 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2294 xprt->addrlen = args->addrlen; 2183 xprt->addrlen = args->addrlen;
2295 if (args->srcaddr) 2184 if (args->srcaddr)
2296 memcpy(&new->srcaddr, args->srcaddr, args->addrlen); 2185 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
2186 else {
2187 int err;
2188 err = xs_init_anyaddr(args->dstaddr->sa_family,
2189 (struct sockaddr *)&new->srcaddr);
2190 if (err != 0)
2191 return ERR_PTR(err);
2192 }
2297 2193
2298 return xprt; 2194 return xprt;
2299} 2195}
@@ -2341,7 +2237,7 @@ static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
2341 xprt_set_bound(xprt); 2237 xprt_set_bound(xprt);
2342 2238
2343 INIT_DELAYED_WORK(&transport->connect_worker, 2239 INIT_DELAYED_WORK(&transport->connect_worker,
2344 xs_udp_connect_worker4); 2240 xs_udp_setup_socket);
2345 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); 2241 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
2346 break; 2242 break;
2347 case AF_INET6: 2243 case AF_INET6:
@@ -2349,7 +2245,7 @@ static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
2349 xprt_set_bound(xprt); 2245 xprt_set_bound(xprt);
2350 2246
2351 INIT_DELAYED_WORK(&transport->connect_worker, 2247 INIT_DELAYED_WORK(&transport->connect_worker,
2352 xs_udp_connect_worker6); 2248 xs_udp_setup_socket);
2353 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); 2249 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
2354 break; 2250 break;
2355 default: 2251 default:
@@ -2371,8 +2267,7 @@ static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
2371 return xprt; 2267 return xprt;
2372 ret = ERR_PTR(-EINVAL); 2268 ret = ERR_PTR(-EINVAL);
2373out_err: 2269out_err:
2374 kfree(xprt->slot); 2270 xprt_free(xprt);
2375 kfree(xprt);
2376 return ret; 2271 return ret;
2377} 2272}
2378 2273
@@ -2416,7 +2311,7 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
2416 xprt_set_bound(xprt); 2311 xprt_set_bound(xprt);
2417 2312
2418 INIT_DELAYED_WORK(&transport->connect_worker, 2313 INIT_DELAYED_WORK(&transport->connect_worker,
2419 xs_tcp_connect_worker4); 2314 xs_tcp_setup_socket);
2420 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); 2315 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
2421 break; 2316 break;
2422 case AF_INET6: 2317 case AF_INET6:
@@ -2424,7 +2319,7 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
2424 xprt_set_bound(xprt); 2319 xprt_set_bound(xprt);
2425 2320
2426 INIT_DELAYED_WORK(&transport->connect_worker, 2321 INIT_DELAYED_WORK(&transport->connect_worker,
2427 xs_tcp_connect_worker6); 2322 xs_tcp_setup_socket);
2428 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); 2323 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
2429 break; 2324 break;
2430 default: 2325 default:
@@ -2447,8 +2342,7 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
2447 return xprt; 2342 return xprt;
2448 ret = ERR_PTR(-EINVAL); 2343 ret = ERR_PTR(-EINVAL);
2449out_err: 2344out_err:
2450 kfree(xprt->slot); 2345 xprt_free(xprt);
2451 kfree(xprt);
2452 return ret; 2346 return ret;
2453} 2347}
2454 2348
@@ -2507,15 +2401,10 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2507 goto out_err; 2401 goto out_err;
2508 } 2402 }
2509 2403
2510 if (xprt_bound(xprt)) 2404 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2511 dprintk("RPC: set up xprt to %s (port %s) via %s\n", 2405 xprt->address_strings[RPC_DISPLAY_ADDR],
2512 xprt->address_strings[RPC_DISPLAY_ADDR], 2406 xprt->address_strings[RPC_DISPLAY_PORT],
2513 xprt->address_strings[RPC_DISPLAY_PORT], 2407 xprt->address_strings[RPC_DISPLAY_PROTO]);
2514 xprt->address_strings[RPC_DISPLAY_PROTO]);
2515 else
2516 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2517 xprt->address_strings[RPC_DISPLAY_ADDR],
2518 xprt->address_strings[RPC_DISPLAY_PROTO]);
2519 2408
2520 /* 2409 /*
2521 * Since we don't want connections for the backchannel, we set 2410 * Since we don't want connections for the backchannel, we set
@@ -2528,8 +2417,7 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2528 return xprt; 2417 return xprt;
2529 ret = ERR_PTR(-EINVAL); 2418 ret = ERR_PTR(-EINVAL);
2530out_err: 2419out_err:
2531 kfree(xprt->slot); 2420 xprt_free(xprt);
2532 kfree(xprt);
2533 return ret; 2421 return ret;
2534} 2422}
2535 2423
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-02 05:34:11 -0400