aboutsummaryrefslogtreecommitdiffstats
path: root/net/sunrpc
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-04-06 16:25:56 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-04-06 16:25:56 -0400
commita63856252d2112e7c452696037a86ceb12f47f80 (patch)
treeb1ad03fe441349069f80e58de425b3f72af9e5b7 /net/sunrpc
parentb24241a09208874d5d770bee30791daae41ad762 (diff)
parent04826f43d4f0a4d56423eb8abb9f2ec9987df5b5 (diff)
Merge branch 'for-2.6.30' of git://linux-nfs.org/~bfields/linux
* 'for-2.6.30' of git://linux-nfs.org/~bfields/linux: (81 commits) nfsd41: define nfsd4_set_statp as noop for !CONFIG_NFSD_V4 nfsd41: define NFSD_DRC_SIZE_SHIFT in set_max_drc nfsd41: Documentation/filesystems/nfs41-server.txt nfsd41: CREATE_EXCLUSIVE4_1 nfsd41: SUPPATTR_EXCLCREAT attribute nfsd41: support for 3-word long attribute bitmask nfsd: dynamically skip encoded fattr bitmap in _nfsd4_verify nfsd41: pass writable attrs mask to nfsd4_decode_fattr nfsd41: provide support for minor version 1 at rpc level nfsd41: control nfsv4.1 svc via /proc/fs/nfsd/versions nfsd41: add OPEN4_SHARE_ACCESS_WANT nfs4_stateid bmap nfsd41: access_valid nfsd41: clientid handling nfsd41: check encode size for sessions maxresponse cached nfsd41: stateid handling nfsd: pass nfsd4_compound_state* to nfs4_preprocess_{state,seq}id_op nfsd41: destroy_session operation nfsd41: non-page DRC for solo sequence responses nfsd41: Add a create session replay cache nfsd41: create_session operation ...
Diffstat (limited to 'net/sunrpc')
-rw-r--r--net/sunrpc/svc.c3
-rw-r--r--net/sunrpc/svc_xprt.c127
-rw-r--r--net/sunrpc/svcsock.c35
3 files changed, 127 insertions, 38 deletions
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 9b49a6ab8ded..8847add6ca16 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1008,6 +1008,8 @@ svc_process(struct svc_rqst *rqstp)
1008 rqstp->rq_res.tail[0].iov_len = 0; 1008 rqstp->rq_res.tail[0].iov_len = 0;
1009 /* Will be turned off only in gss privacy case: */ 1009 /* Will be turned off only in gss privacy case: */
1010 rqstp->rq_splice_ok = 1; 1010 rqstp->rq_splice_ok = 1;
1011 /* Will be turned off only when NFSv4 Sessions are used */
1012 rqstp->rq_usedeferral = 1;
1011 1013
1012 /* Setup reply header */ 1014 /* Setup reply header */
1013 rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp); 1015 rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp);
@@ -1078,7 +1080,6 @@ svc_process(struct svc_rqst *rqstp)
1078 procp = versp->vs_proc + proc; 1080 procp = versp->vs_proc + proc;
1079 if (proc >= versp->vs_nproc || !procp->pc_func) 1081 if (proc >= versp->vs_nproc || !procp->pc_func)
1080 goto err_bad_proc; 1082 goto err_bad_proc;
1081 rqstp->rq_server = serv;
1082 rqstp->rq_procinfo = procp; 1083 rqstp->rq_procinfo = procp;
1083 1084
1084 /* Syntactic check complete */ 1085 /* Syntactic check complete */
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index 2819ee093f36..c200d92e57e4 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -14,6 +14,8 @@
14 14
15#define RPCDBG_FACILITY RPCDBG_SVCXPRT 15#define RPCDBG_FACILITY RPCDBG_SVCXPRT
16 16
17#define SVC_MAX_WAKING 5
18
17static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt); 19static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt);
18static int svc_deferred_recv(struct svc_rqst *rqstp); 20static int svc_deferred_recv(struct svc_rqst *rqstp);
19static struct cache_deferred_req *svc_defer(struct cache_req *req); 21static struct cache_deferred_req *svc_defer(struct cache_req *req);
@@ -301,6 +303,7 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
301 struct svc_pool *pool; 303 struct svc_pool *pool;
302 struct svc_rqst *rqstp; 304 struct svc_rqst *rqstp;
303 int cpu; 305 int cpu;
306 int thread_avail;
304 307
305 if (!(xprt->xpt_flags & 308 if (!(xprt->xpt_flags &
306 ((1<<XPT_CONN)|(1<<XPT_DATA)|(1<<XPT_CLOSE)|(1<<XPT_DEFERRED)))) 309 ((1<<XPT_CONN)|(1<<XPT_DATA)|(1<<XPT_CLOSE)|(1<<XPT_DEFERRED))))
@@ -312,18 +315,14 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
312 315
313 spin_lock_bh(&pool->sp_lock); 316 spin_lock_bh(&pool->sp_lock);
314 317
315 if (!list_empty(&pool->sp_threads) &&
316 !list_empty(&pool->sp_sockets))
317 printk(KERN_ERR
318 "svc_xprt_enqueue: "
319 "threads and transports both waiting??\n");
320
321 if (test_bit(XPT_DEAD, &xprt->xpt_flags)) { 318 if (test_bit(XPT_DEAD, &xprt->xpt_flags)) {
322 /* Don't enqueue dead transports */ 319 /* Don't enqueue dead transports */
323 dprintk("svc: transport %p is dead, not enqueued\n", xprt); 320 dprintk("svc: transport %p is dead, not enqueued\n", xprt);
324 goto out_unlock; 321 goto out_unlock;
325 } 322 }
326 323
324 pool->sp_stats.packets++;
325
327 /* Mark transport as busy. It will remain in this state until 326 /* Mark transport as busy. It will remain in this state until
328 * the provider calls svc_xprt_received. We update XPT_BUSY 327 * the provider calls svc_xprt_received. We update XPT_BUSY
329 * atomically because it also guards against trying to enqueue 328 * atomically because it also guards against trying to enqueue
@@ -356,7 +355,15 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
356 } 355 }
357 356
358 process: 357 process:
359 if (!list_empty(&pool->sp_threads)) { 358 /* Work out whether threads are available */
359 thread_avail = !list_empty(&pool->sp_threads); /* threads are asleep */
360 if (pool->sp_nwaking >= SVC_MAX_WAKING) {
361 /* too many threads are runnable and trying to wake up */
362 thread_avail = 0;
363 pool->sp_stats.overloads_avoided++;
364 }
365
366 if (thread_avail) {
360 rqstp = list_entry(pool->sp_threads.next, 367 rqstp = list_entry(pool->sp_threads.next,
361 struct svc_rqst, 368 struct svc_rqst,
362 rq_list); 369 rq_list);
@@ -371,11 +378,15 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
371 svc_xprt_get(xprt); 378 svc_xprt_get(xprt);
372 rqstp->rq_reserved = serv->sv_max_mesg; 379 rqstp->rq_reserved = serv->sv_max_mesg;
373 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved); 380 atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
381 rqstp->rq_waking = 1;
382 pool->sp_nwaking++;
383 pool->sp_stats.threads_woken++;
374 BUG_ON(xprt->xpt_pool != pool); 384 BUG_ON(xprt->xpt_pool != pool);
375 wake_up(&rqstp->rq_wait); 385 wake_up(&rqstp->rq_wait);
376 } else { 386 } else {
377 dprintk("svc: transport %p put into queue\n", xprt); 387 dprintk("svc: transport %p put into queue\n", xprt);
378 list_add_tail(&xprt->xpt_ready, &pool->sp_sockets); 388 list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
389 pool->sp_stats.sockets_queued++;
379 BUG_ON(xprt->xpt_pool != pool); 390 BUG_ON(xprt->xpt_pool != pool);
380 } 391 }
381 392
@@ -588,6 +599,7 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
588 int pages; 599 int pages;
589 struct xdr_buf *arg; 600 struct xdr_buf *arg;
590 DECLARE_WAITQUEUE(wait, current); 601 DECLARE_WAITQUEUE(wait, current);
602 long time_left;
591 603
592 dprintk("svc: server %p waiting for data (to = %ld)\n", 604 dprintk("svc: server %p waiting for data (to = %ld)\n",
593 rqstp, timeout); 605 rqstp, timeout);
@@ -636,6 +648,11 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
636 return -EINTR; 648 return -EINTR;
637 649
638 spin_lock_bh(&pool->sp_lock); 650 spin_lock_bh(&pool->sp_lock);
651 if (rqstp->rq_waking) {
652 rqstp->rq_waking = 0;
653 pool->sp_nwaking--;
654 BUG_ON(pool->sp_nwaking < 0);
655 }
639 xprt = svc_xprt_dequeue(pool); 656 xprt = svc_xprt_dequeue(pool);
640 if (xprt) { 657 if (xprt) {
641 rqstp->rq_xprt = xprt; 658 rqstp->rq_xprt = xprt;
@@ -668,12 +685,14 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
668 add_wait_queue(&rqstp->rq_wait, &wait); 685 add_wait_queue(&rqstp->rq_wait, &wait);
669 spin_unlock_bh(&pool->sp_lock); 686 spin_unlock_bh(&pool->sp_lock);
670 687
671 schedule_timeout(timeout); 688 time_left = schedule_timeout(timeout);
672 689
673 try_to_freeze(); 690 try_to_freeze();
674 691
675 spin_lock_bh(&pool->sp_lock); 692 spin_lock_bh(&pool->sp_lock);
676 remove_wait_queue(&rqstp->rq_wait, &wait); 693 remove_wait_queue(&rqstp->rq_wait, &wait);
694 if (!time_left)
695 pool->sp_stats.threads_timedout++;
677 696
678 xprt = rqstp->rq_xprt; 697 xprt = rqstp->rq_xprt;
679 if (!xprt) { 698 if (!xprt) {
@@ -958,7 +977,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req)
958 struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle); 977 struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle);
959 struct svc_deferred_req *dr; 978 struct svc_deferred_req *dr;
960 979
961 if (rqstp->rq_arg.page_len) 980 if (rqstp->rq_arg.page_len || !rqstp->rq_usedeferral)
962 return NULL; /* if more than a page, give up FIXME */ 981 return NULL; /* if more than a page, give up FIXME */
963 if (rqstp->rq_deferred) { 982 if (rqstp->rq_deferred) {
964 dr = rqstp->rq_deferred; 983 dr = rqstp->rq_deferred;
@@ -1112,3 +1131,93 @@ int svc_xprt_names(struct svc_serv *serv, char *buf, int buflen)
1112 return totlen; 1131 return totlen;
1113} 1132}
1114EXPORT_SYMBOL_GPL(svc_xprt_names); 1133EXPORT_SYMBOL_GPL(svc_xprt_names);
1134
1135
1136/*----------------------------------------------------------------------------*/
1137
1138static void *svc_pool_stats_start(struct seq_file *m, loff_t *pos)
1139{
1140 unsigned int pidx = (unsigned int)*pos;
1141 struct svc_serv *serv = m->private;
1142
1143 dprintk("svc_pool_stats_start, *pidx=%u\n", pidx);
1144
1145 lock_kernel();
1146 /* bump up the pseudo refcount while traversing */
1147 svc_get(serv);
1148 unlock_kernel();
1149
1150 if (!pidx)
1151 return SEQ_START_TOKEN;
1152 return (pidx > serv->sv_nrpools ? NULL : &serv->sv_pools[pidx-1]);
1153}
1154
1155static void *svc_pool_stats_next(struct seq_file *m, void *p, loff_t *pos)
1156{
1157 struct svc_pool *pool = p;
1158 struct svc_serv *serv = m->private;
1159
1160 dprintk("svc_pool_stats_next, *pos=%llu\n", *pos);
1161
1162 if (p == SEQ_START_TOKEN) {
1163 pool = &serv->sv_pools[0];
1164 } else {
1165 unsigned int pidx = (pool - &serv->sv_pools[0]);
1166 if (pidx < serv->sv_nrpools-1)
1167 pool = &serv->sv_pools[pidx+1];
1168 else
1169 pool = NULL;
1170 }
1171 ++*pos;
1172 return pool;
1173}
1174
1175static void svc_pool_stats_stop(struct seq_file *m, void *p)
1176{
1177 struct svc_serv *serv = m->private;
1178
1179 lock_kernel();
1180 /* this function really, really should have been called svc_put() */
1181 svc_destroy(serv);
1182 unlock_kernel();
1183}
1184
1185static int svc_pool_stats_show(struct seq_file *m, void *p)
1186{
1187 struct svc_pool *pool = p;
1188
1189 if (p == SEQ_START_TOKEN) {
1190 seq_puts(m, "# pool packets-arrived sockets-enqueued threads-woken overloads-avoided threads-timedout\n");
1191 return 0;
1192 }
1193
1194 seq_printf(m, "%u %lu %lu %lu %lu %lu\n",
1195 pool->sp_id,
1196 pool->sp_stats.packets,
1197 pool->sp_stats.sockets_queued,
1198 pool->sp_stats.threads_woken,
1199 pool->sp_stats.overloads_avoided,
1200 pool->sp_stats.threads_timedout);
1201
1202 return 0;
1203}
1204
1205static const struct seq_operations svc_pool_stats_seq_ops = {
1206 .start = svc_pool_stats_start,
1207 .next = svc_pool_stats_next,
1208 .stop = svc_pool_stats_stop,
1209 .show = svc_pool_stats_show,
1210};
1211
1212int svc_pool_stats_open(struct svc_serv *serv, struct file *file)
1213{
1214 int err;
1215
1216 err = seq_open(file, &svc_pool_stats_seq_ops);
1217 if (!err)
1218 ((struct seq_file *) file->private_data)->private = serv;
1219 return err;
1220}
1221EXPORT_SYMBOL(svc_pool_stats_open);
1222
1223/*----------------------------------------------------------------------------*/
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 9d504234af4a..af3198814c15 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -345,7 +345,6 @@ static void svc_sock_setbufsize(struct socket *sock, unsigned int snd,
345 lock_sock(sock->sk); 345 lock_sock(sock->sk);
346 sock->sk->sk_sndbuf = snd * 2; 346 sock->sk->sk_sndbuf = snd * 2;
347 sock->sk->sk_rcvbuf = rcv * 2; 347 sock->sk->sk_rcvbuf = rcv * 2;
348 sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK|SOCK_RCVBUF_LOCK;
349 release_sock(sock->sk); 348 release_sock(sock->sk);
350#endif 349#endif
351} 350}
@@ -797,23 +796,6 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
797 test_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags), 796 test_bit(XPT_CONN, &svsk->sk_xprt.xpt_flags),
798 test_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags)); 797 test_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags));
799 798
800 if (test_and_clear_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags))
801 /* sndbuf needs to have room for one request
802 * per thread, otherwise we can stall even when the
803 * network isn't a bottleneck.
804 *
805 * We count all threads rather than threads in a
806 * particular pool, which provides an upper bound
807 * on the number of threads which will access the socket.
808 *
809 * rcvbuf just needs to be able to hold a few requests.
810 * Normally they will be removed from the queue
811 * as soon a a complete request arrives.
812 */
813 svc_sock_setbufsize(svsk->sk_sock,
814 (serv->sv_nrthreads+3) * serv->sv_max_mesg,
815 3 * serv->sv_max_mesg);
816
817 clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags); 799 clear_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
818 800
819 /* Receive data. If we haven't got the record length yet, get 801 /* Receive data. If we haven't got the record length yet, get
@@ -1061,15 +1043,6 @@ static void svc_tcp_init(struct svc_sock *svsk, struct svc_serv *serv)
1061 1043
1062 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; 1044 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
1063 1045
1064 /* initialise setting must have enough space to
1065 * receive and respond to one request.
1066 * svc_tcp_recvfrom will re-adjust if necessary
1067 */
1068 svc_sock_setbufsize(svsk->sk_sock,
1069 3 * svsk->sk_xprt.xpt_server->sv_max_mesg,
1070 3 * svsk->sk_xprt.xpt_server->sv_max_mesg);
1071
1072 set_bit(XPT_CHNGBUF, &svsk->sk_xprt.xpt_flags);
1073 set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags); 1046 set_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags);
1074 if (sk->sk_state != TCP_ESTABLISHED) 1047 if (sk->sk_state != TCP_ESTABLISHED)
1075 set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags); 1048 set_bit(XPT_CLOSE, &svsk->sk_xprt.xpt_flags);
@@ -1139,8 +1112,14 @@ static struct svc_sock *svc_setup_socket(struct svc_serv *serv,
1139 /* Initialize the socket */ 1112 /* Initialize the socket */
1140 if (sock->type == SOCK_DGRAM) 1113 if (sock->type == SOCK_DGRAM)
1141 svc_udp_init(svsk, serv); 1114 svc_udp_init(svsk, serv);
1142 else 1115 else {
1116 /* initialise setting must have enough space to
1117 * receive and respond to one request.
1118 */
1119 svc_sock_setbufsize(svsk->sk_sock, 4 * serv->sv_max_mesg,
1120 4 * serv->sv_max_mesg);
1143 svc_tcp_init(svsk, serv); 1121 svc_tcp_init(svsk, serv);
1122 }
1144 1123
1145 dprintk("svc: svc_setup_socket created %p (inet %p)\n", 1124 dprintk("svc: svc_setup_socket created %p (inet %p)\n",
1146 svsk, svsk->sk_sk); 1125 svsk, svsk->sk_sk);