1457c8996SThomas Gleixner // SPDX-License-Identifier: GPL-2.0-only
21d8206b9STom Tucker /*
31d8206b9STom Tucker * linux/net/sunrpc/svc_xprt.c
41d8206b9STom Tucker *
51d8206b9STom Tucker * Author: Tom Tucker <tom@opengridcomputing.com>
61d8206b9STom Tucker */
71d8206b9STom Tucker
81d8206b9STom Tucker #include <linux/sched.h>
94034247aSNeilBrown #include <linux/sched/mm.h>
101d8206b9STom Tucker #include <linux/errno.h>
111d8206b9STom Tucker #include <linux/freezer.h>
127086721fSJeff Layton #include <linux/kthread.h>
135a0e3ad6STejun Heo #include <linux/slab.h>
141d8206b9STom Tucker #include <net/sock.h>
15c3d4879eSScott Mayhew #include <linux/sunrpc/addr.h>
161d8206b9STom Tucker #include <linux/sunrpc/stats.h>
171d8206b9STom Tucker #include <linux/sunrpc/svc_xprt.h>
18dcf1a357SH Hartley Sweeten #include <linux/sunrpc/svcsock.h>
1999de8ea9SJ. Bruce Fields #include <linux/sunrpc/xprt.h>
203a9a231dSPaul Gortmaker #include <linux/module.h>
21c3d4879eSScott Mayhew #include <linux/netdevice.h>
22860a0d9eSJeff Layton #include <trace/events/sunrpc.h>
231d8206b9STom Tucker
241d8206b9STom Tucker #define RPCDBG_FACILITY RPCDBG_SVCXPRT
251d8206b9STom Tucker
26ff3ac5c3STrond Myklebust static unsigned int svc_rpc_per_connection_limit __read_mostly;
27ff3ac5c3STrond Myklebust module_param(svc_rpc_per_connection_limit, uint, 0644);
28ff3ac5c3STrond Myklebust
29ff3ac5c3STrond Myklebust
300f0257eaSTom Tucker static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt);
310f0257eaSTom Tucker static int svc_deferred_recv(struct svc_rqst *rqstp);
320f0257eaSTom Tucker static struct cache_deferred_req *svc_defer(struct cache_req *req);
33ff861c4dSKees Cook static void svc_age_temp_xprts(struct timer_list *t);
347710ec36SJ. Bruce Fields static void svc_delete_xprt(struct svc_xprt *xprt);
350f0257eaSTom Tucker
360f0257eaSTom Tucker /* apparently the "standard" is that clients close
370f0257eaSTom Tucker * idle connections after 5 minutes, servers after
380f0257eaSTom Tucker * 6 minutes
39855c9e76SJ. Bruce Fields * http://nfsv4bat.org/Documents/ConnectAThon/1996/nfstcp.pdf
400f0257eaSTom Tucker */
410f0257eaSTom Tucker static int svc_conn_age_period = 6*60;
420f0257eaSTom Tucker
431d8206b9STom Tucker /* List of registered transport classes */
441d8206b9STom Tucker static DEFINE_SPINLOCK(svc_xprt_class_lock);
451d8206b9STom Tucker static LIST_HEAD(svc_xprt_class_list);
461d8206b9STom Tucker
470f0257eaSTom Tucker /* SMP locking strategy:
480f0257eaSTom Tucker *
490f0257eaSTom Tucker * svc_pool->sp_lock protects most of the fields of that pool.
500f0257eaSTom Tucker * svc_serv->sv_lock protects sv_tempsocks, sv_permsocks, sv_tmpcnt.
510f0257eaSTom Tucker * when both need to be taken (rare), svc_serv->sv_lock is first.
523c519914SJeff Layton * The "service mutex" protects svc_serv->sv_nrthread.
530f0257eaSTom Tucker * svc_sock->sk_lock protects the svc_sock->sk_deferred list
540f0257eaSTom Tucker * and the ->sk_info_authunix cache.
550f0257eaSTom Tucker *
560f0257eaSTom Tucker * The XPT_BUSY bit in xprt->xpt_flags prevents a transport being
570f0257eaSTom Tucker * enqueued multiply. During normal transport processing this bit
580f0257eaSTom Tucker * is set by svc_xprt_enqueue and cleared by svc_xprt_received.
590f0257eaSTom Tucker * Providers should not manipulate this bit directly.
600f0257eaSTom Tucker *
610f0257eaSTom Tucker * Some flags can be set to certain values at any time
620f0257eaSTom Tucker * providing that certain rules are followed:
630f0257eaSTom Tucker *
640f0257eaSTom Tucker * XPT_CONN, XPT_DATA:
650f0257eaSTom Tucker * - Can be set or cleared at any time.
660f0257eaSTom Tucker * - After a set, svc_xprt_enqueue must be called to enqueue
670f0257eaSTom Tucker * the transport for processing.
680f0257eaSTom Tucker * - After a clear, the transport must be read/accepted.
690f0257eaSTom Tucker * If this succeeds, it must be set again.
700f0257eaSTom Tucker * XPT_CLOSE:
710f0257eaSTom Tucker * - Can set at any time. It is never cleared.
720f0257eaSTom Tucker * XPT_DEAD:
730f0257eaSTom Tucker * - Can only be set while XPT_BUSY is held which ensures
740f0257eaSTom Tucker * that no other thread will be using the transport or will
750f0257eaSTom Tucker * try to set XPT_DEAD.
760f0257eaSTom Tucker */
7702cea33fSChuck Lever
7802cea33fSChuck Lever /**
7902cea33fSChuck Lever * svc_reg_xprt_class - Register a server-side RPC transport class
8002cea33fSChuck Lever * @xcl: New transport class to be registered
8102cea33fSChuck Lever *
8202cea33fSChuck Lever * Returns zero on success; otherwise a negative errno is returned.
8302cea33fSChuck Lever */
svc_reg_xprt_class(struct svc_xprt_class * xcl)841d8206b9STom Tucker int svc_reg_xprt_class(struct svc_xprt_class *xcl)
851d8206b9STom Tucker {
861d8206b9STom Tucker struct svc_xprt_class *cl;
871d8206b9STom Tucker int res = -EEXIST;
881d8206b9STom Tucker
891d8206b9STom Tucker INIT_LIST_HEAD(&xcl->xcl_list);
901d8206b9STom Tucker spin_lock(&svc_xprt_class_lock);
911d8206b9STom Tucker /* Make sure there isn't already a class with the same name */
921d8206b9STom Tucker list_for_each_entry(cl, &svc_xprt_class_list, xcl_list) {
931d8206b9STom Tucker if (strcmp(xcl->xcl_name, cl->xcl_name) == 0)
941d8206b9STom Tucker goto out;
951d8206b9STom Tucker }
961d8206b9STom Tucker list_add_tail(&xcl->xcl_list, &svc_xprt_class_list);
971d8206b9STom Tucker res = 0;
981d8206b9STom Tucker out:
991d8206b9STom Tucker spin_unlock(&svc_xprt_class_lock);
1001d8206b9STom Tucker return res;
1011d8206b9STom Tucker }
1021d8206b9STom Tucker EXPORT_SYMBOL_GPL(svc_reg_xprt_class);
1031d8206b9STom Tucker
10402cea33fSChuck Lever /**
10502cea33fSChuck Lever * svc_unreg_xprt_class - Unregister a server-side RPC transport class
10602cea33fSChuck Lever * @xcl: Transport class to be unregistered
10702cea33fSChuck Lever *
10802cea33fSChuck Lever */
svc_unreg_xprt_class(struct svc_xprt_class * xcl)1091d8206b9STom Tucker void svc_unreg_xprt_class(struct svc_xprt_class *xcl)
1101d8206b9STom Tucker {
1111d8206b9STom Tucker spin_lock(&svc_xprt_class_lock);
1121d8206b9STom Tucker list_del_init(&xcl->xcl_list);
1131d8206b9STom Tucker spin_unlock(&svc_xprt_class_lock);
1141d8206b9STom Tucker }
1151d8206b9STom Tucker EXPORT_SYMBOL_GPL(svc_unreg_xprt_class);
1161d8206b9STom Tucker
117b25b60d7SChristophe JAILLET /**
118b25b60d7SChristophe JAILLET * svc_print_xprts - Format the transport list for printing
119b25b60d7SChristophe JAILLET * @buf: target buffer for formatted address
120b25b60d7SChristophe JAILLET * @maxlen: length of target buffer
121b25b60d7SChristophe JAILLET *
122b25b60d7SChristophe JAILLET * Fills in @buf with a string containing a list of transport names, each name
123b25b60d7SChristophe JAILLET * terminated with '\n'. If the buffer is too small, some entries may be
124b25b60d7SChristophe JAILLET * missing, but it is guaranteed that all lines in the output buffer are
125b25b60d7SChristophe JAILLET * complete.
126b25b60d7SChristophe JAILLET *
127b25b60d7SChristophe JAILLET * Returns positive length of the filled-in string.
128dc9a16e4STom Tucker */
svc_print_xprts(char * buf,int maxlen)129dc9a16e4STom Tucker int svc_print_xprts(char *buf, int maxlen)
130dc9a16e4STom Tucker {
1318f3a6de3SPavel Emelyanov struct svc_xprt_class *xcl;
132dc9a16e4STom Tucker char tmpstr[80];
133dc9a16e4STom Tucker int len = 0;
134dc9a16e4STom Tucker buf[0] = '\0';
135dc9a16e4STom Tucker
136dc9a16e4STom Tucker spin_lock(&svc_xprt_class_lock);
1378f3a6de3SPavel Emelyanov list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
138dc9a16e4STom Tucker int slen;
139dc9a16e4STom Tucker
140b25b60d7SChristophe JAILLET slen = snprintf(tmpstr, sizeof(tmpstr), "%s %d\n",
141b25b60d7SChristophe JAILLET xcl->xcl_name, xcl->xcl_max_payload);
142b25b60d7SChristophe JAILLET if (slen >= sizeof(tmpstr) || len + slen >= maxlen)
143dc9a16e4STom Tucker break;
144dc9a16e4STom Tucker len += slen;
145dc9a16e4STom Tucker strcat(buf, tmpstr);
146dc9a16e4STom Tucker }
147dc9a16e4STom Tucker spin_unlock(&svc_xprt_class_lock);
148dc9a16e4STom Tucker
149dc9a16e4STom Tucker return len;
150dc9a16e4STom Tucker }
151dc9a16e4STom Tucker
152e844d307SChuck Lever /**
153e844d307SChuck Lever * svc_xprt_deferred_close - Close a transport
154e844d307SChuck Lever * @xprt: transport instance
155e844d307SChuck Lever *
156e844d307SChuck Lever * Used in contexts that need to defer the work of shutting down
157e844d307SChuck Lever * the transport to an nfsd thread.
158e844d307SChuck Lever */
svc_xprt_deferred_close(struct svc_xprt * xprt)159e844d307SChuck Lever void svc_xprt_deferred_close(struct svc_xprt *xprt)
160e844d307SChuck Lever {
161e844d307SChuck Lever if (!test_and_set_bit(XPT_CLOSE, &xprt->xpt_flags))
162e844d307SChuck Lever svc_xprt_enqueue(xprt);
163e844d307SChuck Lever }
164e844d307SChuck Lever EXPORT_SYMBOL_GPL(svc_xprt_deferred_close);
165e844d307SChuck Lever
svc_xprt_free(struct kref * kref)166e1b3157fSTom Tucker static void svc_xprt_free(struct kref *kref)
167e1b3157fSTom Tucker {
168e1b3157fSTom Tucker struct svc_xprt *xprt =
169e1b3157fSTom Tucker container_of(kref, struct svc_xprt, xpt_ref);
170e1b3157fSTom Tucker struct module *owner = xprt->xpt_class->xcl_owner;
171e3bfca01SPavel Emelyanov if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags))
172e3bfca01SPavel Emelyanov svcauth_unix_info_release(xprt);
1734df493a2STrond Myklebust put_cred(xprt->xpt_cred);
1746cdef8a6SEric Dumazet put_net_track(xprt->xpt_net, &xprt->ns_tracker);
17599de8ea9SJ. Bruce Fields /* See comment on corresponding get in xs_setup_bc_tcp(): */
17699de8ea9SJ. Bruce Fields if (xprt->xpt_bc_xprt)
17799de8ea9SJ. Bruce Fields xprt_put(xprt->xpt_bc_xprt);
17839a9beabSJ. Bruce Fields if (xprt->xpt_bc_xps)
17939a9beabSJ. Bruce Fields xprt_switch_put(xprt->xpt_bc_xps);
18011bbb0f7SChuck Lever trace_svc_xprt_free(xprt);
181e1b3157fSTom Tucker xprt->xpt_ops->xpo_free(xprt);
182e1b3157fSTom Tucker module_put(owner);
183e1b3157fSTom Tucker }
184e1b3157fSTom Tucker
svc_xprt_put(struct svc_xprt * xprt)185e1b3157fSTom Tucker void svc_xprt_put(struct svc_xprt *xprt)
186e1b3157fSTom Tucker {
187e1b3157fSTom Tucker kref_put(&xprt->xpt_ref, svc_xprt_free);
188e1b3157fSTom Tucker }
189e1b3157fSTom Tucker EXPORT_SYMBOL_GPL(svc_xprt_put);
190e1b3157fSTom Tucker
1911d8206b9STom Tucker /*
1921d8206b9STom Tucker * Called by transport drivers to initialize the transport independent
1931d8206b9STom Tucker * portion of the transport instance.
1941d8206b9STom Tucker */
svc_xprt_init(struct net * net,struct svc_xprt_class * xcl,struct svc_xprt * xprt,struct svc_serv * serv)195bd4620ddSStanislav Kinsbursky void svc_xprt_init(struct net *net, struct svc_xprt_class *xcl,
196bd4620ddSStanislav Kinsbursky struct svc_xprt *xprt, struct svc_serv *serv)
1971d8206b9STom Tucker {
1981d8206b9STom Tucker memset(xprt, 0, sizeof(*xprt));
1991d8206b9STom Tucker xprt->xpt_class = xcl;
2001d8206b9STom Tucker xprt->xpt_ops = xcl->xcl_ops;
201e1b3157fSTom Tucker kref_init(&xprt->xpt_ref);
202bb5cf160STom Tucker xprt->xpt_server = serv;
2037a182083STom Tucker INIT_LIST_HEAD(&xprt->xpt_list);
2047a182083STom Tucker INIT_LIST_HEAD(&xprt->xpt_ready);
2058c7b0172STom Tucker INIT_LIST_HEAD(&xprt->xpt_deferred);
206edc7a894SJ. Bruce Fields INIT_LIST_HEAD(&xprt->xpt_users);
207a50fea26STom Tucker mutex_init(&xprt->xpt_mutex);
208def13d74STom Tucker spin_lock_init(&xprt->xpt_lock);
2094e5caaa5STom Tucker set_bit(XPT_BUSY, &xprt->xpt_flags);
2106cdef8a6SEric Dumazet xprt->xpt_net = get_net_track(net, &xprt->ns_tracker, GFP_ATOMIC);
211ece200ddSChuck Lever strcpy(xprt->xpt_remotebuf, "uninitialized");
2121d8206b9STom Tucker }
2131d8206b9STom Tucker EXPORT_SYMBOL_GPL(svc_xprt_init);
214b700cbb1STom Tucker
__svc_xpo_create(struct svc_xprt_class * xcl,struct svc_serv * serv,struct net * net,const int family,const unsigned short port,int flags)2155dd248f6SChuck Lever static struct svc_xprt *__svc_xpo_create(struct svc_xprt_class *xcl,
2165dd248f6SChuck Lever struct svc_serv *serv,
21762832c03SPavel Emelyanov struct net *net,
2189652ada3SChuck Lever const int family,
2199652ada3SChuck Lever const unsigned short port,
2209652ada3SChuck Lever int flags)
221b700cbb1STom Tucker {
222b700cbb1STom Tucker struct sockaddr_in sin = {
223b700cbb1STom Tucker .sin_family = AF_INET,
224e6f1cebfSAl Viro .sin_addr.s_addr = htonl(INADDR_ANY),
225b700cbb1STom Tucker .sin_port = htons(port),
226b700cbb1STom Tucker };
227dfd56b8bSEric Dumazet #if IS_ENABLED(CONFIG_IPV6)
2285dd248f6SChuck Lever struct sockaddr_in6 sin6 = {
2295dd248f6SChuck Lever .sin6_family = AF_INET6,
2305dd248f6SChuck Lever .sin6_addr = IN6ADDR_ANY_INIT,
2315dd248f6SChuck Lever .sin6_port = htons(port),
2325dd248f6SChuck Lever };
233dfd56b8bSEric Dumazet #endif
2344b8f380eSChuck Lever struct svc_xprt *xprt;
2355dd248f6SChuck Lever struct sockaddr *sap;
2365dd248f6SChuck Lever size_t len;
2375dd248f6SChuck Lever
2389652ada3SChuck Lever switch (family) {
2399652ada3SChuck Lever case PF_INET:
2405dd248f6SChuck Lever sap = (struct sockaddr *)&sin;
2415dd248f6SChuck Lever len = sizeof(sin);
2425dd248f6SChuck Lever break;
243dfd56b8bSEric Dumazet #if IS_ENABLED(CONFIG_IPV6)
2449652ada3SChuck Lever case PF_INET6:
2455dd248f6SChuck Lever sap = (struct sockaddr *)&sin6;
2465dd248f6SChuck Lever len = sizeof(sin6);
2475dd248f6SChuck Lever break;
248dfd56b8bSEric Dumazet #endif
2495dd248f6SChuck Lever default:
2505dd248f6SChuck Lever return ERR_PTR(-EAFNOSUPPORT);
2515dd248f6SChuck Lever }
2525dd248f6SChuck Lever
2534b8f380eSChuck Lever xprt = xcl->xcl_ops->xpo_create(serv, net, sap, len, flags);
2544b8f380eSChuck Lever if (IS_ERR(xprt))
2554b8f380eSChuck Lever trace_svc_xprt_create_err(serv->sv_program->pg_name,
256dc6c6fb3SChuck Lever xcl->xcl_name, sap, len, xprt);
2574b8f380eSChuck Lever return xprt;
2585dd248f6SChuck Lever }
2595dd248f6SChuck Lever
2607dcfbd86SChuck Lever /**
2617dcfbd86SChuck Lever * svc_xprt_received - start next receiver thread
2627dcfbd86SChuck Lever * @xprt: controlling transport
2637dcfbd86SChuck Lever *
2647dcfbd86SChuck Lever * The caller must hold the XPT_BUSY bit and must
2656741019cSJ. Bruce Fields * not thereafter touch transport data.
2666741019cSJ. Bruce Fields *
2676741019cSJ. Bruce Fields * Note: XPT_DATA only gets cleared when a read-attempt finds no (or
2686741019cSJ. Bruce Fields * insufficient) data.
2696741019cSJ. Bruce Fields */
svc_xprt_received(struct svc_xprt * xprt)2707dcfbd86SChuck Lever void svc_xprt_received(struct svc_xprt *xprt)
2716741019cSJ. Bruce Fields {
272acf06a7fSJeff Layton if (!test_bit(XPT_BUSY, &xprt->xpt_flags)) {
273acf06a7fSJeff Layton WARN_ONCE(1, "xprt=0x%p already busy!", xprt);
274ff1fdb9bSWeston Andros Adamson return;
275acf06a7fSJeff Layton }
276acf06a7fSJeff Layton
2776741019cSJ. Bruce Fields /* As soon as we clear busy, the xprt could be closed and
278c0219c49SChuck Lever * 'put', so we need a reference to call svc_xprt_enqueue with:
2796741019cSJ. Bruce Fields */
2806741019cSJ. Bruce Fields svc_xprt_get(xprt);
2810971374eSTrond Myklebust smp_mb__before_atomic();
2826741019cSJ. Bruce Fields clear_bit(XPT_BUSY, &xprt->xpt_flags);
283c0219c49SChuck Lever svc_xprt_enqueue(xprt);
2846741019cSJ. Bruce Fields svc_xprt_put(xprt);
2856741019cSJ. Bruce Fields }
2867dcfbd86SChuck Lever EXPORT_SYMBOL_GPL(svc_xprt_received);
2876741019cSJ. Bruce Fields
svc_add_new_perm_xprt(struct svc_serv * serv,struct svc_xprt * new)28839b55301SJ. Bruce Fields void svc_add_new_perm_xprt(struct svc_serv *serv, struct svc_xprt *new)
28939b55301SJ. Bruce Fields {
29039b55301SJ. Bruce Fields clear_bit(XPT_TEMP, &new->xpt_flags);
29139b55301SJ. Bruce Fields spin_lock_bh(&serv->sv_lock);
29239b55301SJ. Bruce Fields list_add(&new->xpt_list, &serv->sv_permsocks);
29339b55301SJ. Bruce Fields spin_unlock_bh(&serv->sv_lock);
29439b55301SJ. Bruce Fields svc_xprt_received(new);
29539b55301SJ. Bruce Fields }
29639b55301SJ. Bruce Fields
_svc_xprt_create(struct svc_serv * serv,const char * xprt_name,struct net * net,const int family,const unsigned short port,int flags,const struct cred * cred)297352ad314SChuck Lever static int _svc_xprt_create(struct svc_serv *serv, const char *xprt_name,
298fc5d00b0SPavel Emelyanov struct net *net, const int family,
2994df493a2STrond Myklebust const unsigned short port, int flags,
3004df493a2STrond Myklebust const struct cred *cred)
3015dd248f6SChuck Lever {
3025dd248f6SChuck Lever struct svc_xprt_class *xcl;
3035dd248f6SChuck Lever
304b700cbb1STom Tucker spin_lock(&svc_xprt_class_lock);
305b700cbb1STom Tucker list_for_each_entry(xcl, &svc_xprt_class_list, xcl_list) {
306b700cbb1STom Tucker struct svc_xprt *newxprt;
307ed2849d3SNeilBrown unsigned short newport;
3084e5caaa5STom Tucker
3094e5caaa5STom Tucker if (strcmp(xprt_name, xcl->xcl_name))
3104e5caaa5STom Tucker continue;
3114e5caaa5STom Tucker
3124e5caaa5STom Tucker if (!try_module_get(xcl->xcl_owner))
3134e5caaa5STom Tucker goto err;
3144e5caaa5STom Tucker
3154e5caaa5STom Tucker spin_unlock(&svc_xprt_class_lock);
31662832c03SPavel Emelyanov newxprt = __svc_xpo_create(xcl, serv, net, family, port, flags);
317b700cbb1STom Tucker if (IS_ERR(newxprt)) {
318b700cbb1STom Tucker module_put(xcl->xcl_owner);
3194e5caaa5STom Tucker return PTR_ERR(newxprt);
320b700cbb1STom Tucker }
3214df493a2STrond Myklebust newxprt->xpt_cred = get_cred(cred);
32239b55301SJ. Bruce Fields svc_add_new_perm_xprt(serv, newxprt);
323ed2849d3SNeilBrown newport = svc_xprt_local_port(newxprt);
324ed2849d3SNeilBrown return newport;
325b700cbb1STom Tucker }
3264e5caaa5STom Tucker err:
327b700cbb1STom Tucker spin_unlock(&svc_xprt_class_lock);
32868717908SChuck Lever /* This errno is exposed to user space. Provide a reasonable
32968717908SChuck Lever * perror msg for a bad transport. */
33068717908SChuck Lever return -EPROTONOSUPPORT;
331b700cbb1STom Tucker }
332d96b9c93SJ. Bruce Fields
333352ad314SChuck Lever /**
334352ad314SChuck Lever * svc_xprt_create - Add a new listener to @serv
335352ad314SChuck Lever * @serv: target RPC service
336352ad314SChuck Lever * @xprt_name: transport class name
337352ad314SChuck Lever * @net: network namespace
338352ad314SChuck Lever * @family: network address family
339352ad314SChuck Lever * @port: listener port
340352ad314SChuck Lever * @flags: SVC_SOCK flags
341352ad314SChuck Lever * @cred: credential to bind to this transport
342352ad314SChuck Lever *
343352ad314SChuck Lever * Return values:
344352ad314SChuck Lever * %0: New listener added successfully
345352ad314SChuck Lever * %-EPROTONOSUPPORT: Requested transport type not supported
346352ad314SChuck Lever */
svc_xprt_create(struct svc_serv * serv,const char * xprt_name,struct net * net,const int family,const unsigned short port,int flags,const struct cred * cred)347352ad314SChuck Lever int svc_xprt_create(struct svc_serv *serv, const char *xprt_name,
348d96b9c93SJ. Bruce Fields struct net *net, const int family,
3494df493a2STrond Myklebust const unsigned short port, int flags,
3504df493a2STrond Myklebust const struct cred *cred)
351d96b9c93SJ. Bruce Fields {
352d96b9c93SJ. Bruce Fields int err;
353d96b9c93SJ. Bruce Fields
354352ad314SChuck Lever err = _svc_xprt_create(serv, xprt_name, net, family, port, flags, cred);
355d96b9c93SJ. Bruce Fields if (err == -EPROTONOSUPPORT) {
356d96b9c93SJ. Bruce Fields request_module("svc%s", xprt_name);
357352ad314SChuck Lever err = _svc_xprt_create(serv, xprt_name, net, family, port, flags, cred);
358d96b9c93SJ. Bruce Fields }
359d96b9c93SJ. Bruce Fields return err;
360d96b9c93SJ. Bruce Fields }
361352ad314SChuck Lever EXPORT_SYMBOL_GPL(svc_xprt_create);
3629dbc240fSTom Tucker
3639dbc240fSTom Tucker /*
3649dbc240fSTom Tucker * Copy the local and remote xprt addresses to the rqstp structure
3659dbc240fSTom Tucker */
svc_xprt_copy_addrs(struct svc_rqst * rqstp,struct svc_xprt * xprt)3669dbc240fSTom Tucker void svc_xprt_copy_addrs(struct svc_rqst *rqstp, struct svc_xprt *xprt)
3679dbc240fSTom Tucker {
3689dbc240fSTom Tucker memcpy(&rqstp->rq_addr, &xprt->xpt_remote, xprt->xpt_remotelen);
3699dbc240fSTom Tucker rqstp->rq_addrlen = xprt->xpt_remotelen;
3709dbc240fSTom Tucker
3719dbc240fSTom Tucker /*
3729dbc240fSTom Tucker * Destination address in request is needed for binding the
3739dbc240fSTom Tucker * source address in RPC replies/callbacks later.
3749dbc240fSTom Tucker */
375849a1cf1SMi Jinlong memcpy(&rqstp->rq_daddr, &xprt->xpt_local, xprt->xpt_locallen);
376849a1cf1SMi Jinlong rqstp->rq_daddrlen = xprt->xpt_locallen;
3779dbc240fSTom Tucker }
3789dbc240fSTom Tucker EXPORT_SYMBOL_GPL(svc_xprt_copy_addrs);
3799dbc240fSTom Tucker
3800f0257eaSTom Tucker /**
3810f0257eaSTom Tucker * svc_print_addr - Format rq_addr field for printing
3820f0257eaSTom Tucker * @rqstp: svc_rqst struct containing address to print
3830f0257eaSTom Tucker * @buf: target buffer for formatted address
3840f0257eaSTom Tucker * @len: length of target buffer
3850f0257eaSTom Tucker *
3860f0257eaSTom Tucker */
svc_print_addr(struct svc_rqst * rqstp,char * buf,size_t len)3870f0257eaSTom Tucker char *svc_print_addr(struct svc_rqst *rqstp, char *buf, size_t len)
3880f0257eaSTom Tucker {
3890f0257eaSTom Tucker return __svc_print_addr(svc_addr(rqstp), buf, len);
3900f0257eaSTom Tucker }
3910f0257eaSTom Tucker EXPORT_SYMBOL_GPL(svc_print_addr);
3920f0257eaSTom Tucker
svc_xprt_slots_in_range(struct svc_xprt * xprt)393ff3ac5c3STrond Myklebust static bool svc_xprt_slots_in_range(struct svc_xprt *xprt)
394ff3ac5c3STrond Myklebust {
395ff3ac5c3STrond Myklebust unsigned int limit = svc_rpc_per_connection_limit;
396ff3ac5c3STrond Myklebust int nrqsts = atomic_read(&xprt->xpt_nr_rqsts);
397ff3ac5c3STrond Myklebust
398ff3ac5c3STrond Myklebust return limit == 0 || (nrqsts >= 0 && nrqsts < limit);
399ff3ac5c3STrond Myklebust }
400ff3ac5c3STrond Myklebust
svc_xprt_reserve_slot(struct svc_rqst * rqstp,struct svc_xprt * xprt)401ff3ac5c3STrond Myklebust static bool svc_xprt_reserve_slot(struct svc_rqst *rqstp, struct svc_xprt *xprt)
402ff3ac5c3STrond Myklebust {
403ff3ac5c3STrond Myklebust if (!test_bit(RQ_DATA, &rqstp->rq_flags)) {
404ff3ac5c3STrond Myklebust if (!svc_xprt_slots_in_range(xprt))
405ff3ac5c3STrond Myklebust return false;
406ff3ac5c3STrond Myklebust atomic_inc(&xprt->xpt_nr_rqsts);
407ff3ac5c3STrond Myklebust set_bit(RQ_DATA, &rqstp->rq_flags);
408ff3ac5c3STrond Myklebust }
409ff3ac5c3STrond Myklebust return true;
410ff3ac5c3STrond Myklebust }
411ff3ac5c3STrond Myklebust
svc_xprt_release_slot(struct svc_rqst * rqstp)412ff3ac5c3STrond Myklebust static void svc_xprt_release_slot(struct svc_rqst *rqstp)
413ff3ac5c3STrond Myklebust {
414ff3ac5c3STrond Myklebust struct svc_xprt *xprt = rqstp->rq_xprt;
415ff3ac5c3STrond Myklebust if (test_and_clear_bit(RQ_DATA, &rqstp->rq_flags)) {
416ff3ac5c3STrond Myklebust atomic_dec(&xprt->xpt_nr_rqsts);
41795503d29SJ. Bruce Fields smp_wmb(); /* See smp_rmb() in svc_xprt_ready() */
418ff3ac5c3STrond Myklebust svc_xprt_enqueue(xprt);
419ff3ac5c3STrond Myklebust }
420ff3ac5c3STrond Myklebust }
421ff3ac5c3STrond Myklebust
svc_xprt_ready(struct svc_xprt * xprt)42266c898caSJ. Bruce Fields static bool svc_xprt_ready(struct svc_xprt *xprt)
4239c335c0bSJ. Bruce Fields {
4241602a7b7STrond Myklebust unsigned long xpt_flags;
4251602a7b7STrond Myklebust
42695503d29SJ. Bruce Fields /*
42795503d29SJ. Bruce Fields * If another cpu has recently updated xpt_flags,
42895503d29SJ. Bruce Fields * sk_sock->flags, xpt_reserved, or xpt_nr_rqsts, we need to
42995503d29SJ. Bruce Fields * know about it; otherwise it's possible that both that cpu and
43095503d29SJ. Bruce Fields * this one could call svc_xprt_enqueue() without either
43195503d29SJ. Bruce Fields * svc_xprt_enqueue() recognizing that the conditions below
43295503d29SJ. Bruce Fields * are satisfied, and we could stall indefinitely:
43395503d29SJ. Bruce Fields */
43495503d29SJ. Bruce Fields smp_rmb();
4351602a7b7STrond Myklebust xpt_flags = READ_ONCE(xprt->xpt_flags);
4361602a7b7STrond Myklebust
43782e5d82aSChuck Lever trace_svc_xprt_enqueue(xprt, xpt_flags);
438c0219c49SChuck Lever if (xpt_flags & BIT(XPT_BUSY))
439c0219c49SChuck Lever return false;
440b3cbf98eSChuck Lever if (xpt_flags & (BIT(XPT_CONN) | BIT(XPT_CLOSE) | BIT(XPT_HANDSHAKE)))
4419c335c0bSJ. Bruce Fields return true;
4421602a7b7STrond Myklebust if (xpt_flags & (BIT(XPT_DATA) | BIT(XPT_DEFERRED))) {
443ff3ac5c3STrond Myklebust if (xprt->xpt_ops->xpo_has_wspace(xprt) &&
444ff3ac5c3STrond Myklebust svc_xprt_slots_in_range(xprt))
44582ea2d76STrond Myklebust return true;
44682ea2d76STrond Myklebust trace_svc_xprt_no_write_space(xprt);
44782ea2d76STrond Myklebust return false;
44882ea2d76STrond Myklebust }
4499c335c0bSJ. Bruce Fields return false;
4509c335c0bSJ. Bruce Fields }
4519c335c0bSJ. Bruce Fields
452c0219c49SChuck Lever /**
453c0219c49SChuck Lever * svc_xprt_enqueue - Queue a transport on an idle nfsd thread
454c0219c49SChuck Lever * @xprt: transport with data pending
455c0219c49SChuck Lever *
456c0219c49SChuck Lever */
svc_xprt_enqueue(struct svc_xprt * xprt)457c0219c49SChuck Lever void svc_xprt_enqueue(struct svc_xprt *xprt)
4580f0257eaSTom Tucker {
4590f0257eaSTom Tucker struct svc_pool *pool;
4600f0257eaSTom Tucker
46166c898caSJ. Bruce Fields if (!svc_xprt_ready(xprt))
4627dbb53baSChuck Lever return;
4630f0257eaSTom Tucker
4640f0257eaSTom Tucker /* Mark transport as busy. It will remain in this state until
4650f0257eaSTom Tucker * the provider calls svc_xprt_received. We update XPT_BUSY
4660f0257eaSTom Tucker * atomically because it also guards against trying to enqueue
4670f0257eaSTom Tucker * the transport twice.
4680f0257eaSTom Tucker */
4697dbb53baSChuck Lever if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
4707dbb53baSChuck Lever return;
4710f0257eaSTom Tucker
4722059b698SChuck Lever pool = svc_pool_for_cpu(xprt->xpt_server);
4730c0746d0STrond Myklebust
474ccf08bedSChuck Lever percpu_counter_inc(&pool->sp_sockets_queued);
475b1691bc0SJeff Layton spin_lock_bh(&pool->sp_lock);
476b1691bc0SJeff Layton list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
4770f0257eaSTom Tucker spin_unlock_bh(&pool->sp_lock);
47822700f3cSTrond Myklebust
479850bac3aSChuck Lever svc_pool_wake_idle_thread(pool);
4800f0257eaSTom Tucker }
4810f0257eaSTom Tucker EXPORT_SYMBOL_GPL(svc_xprt_enqueue);
4820f0257eaSTom Tucker
4830f0257eaSTom Tucker /*
484b1691bc0SJeff Layton * Dequeue the first transport, if there is one.
4850f0257eaSTom Tucker */
svc_xprt_dequeue(struct svc_pool * pool)4860f0257eaSTom Tucker static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
4870f0257eaSTom Tucker {
488b1691bc0SJeff Layton struct svc_xprt *xprt = NULL;
4890f0257eaSTom Tucker
4900f0257eaSTom Tucker if (list_empty(&pool->sp_sockets))
49183a712e0SJeff Layton goto out;
4920f0257eaSTom Tucker
493b1691bc0SJeff Layton spin_lock_bh(&pool->sp_lock);
494b1691bc0SJeff Layton if (likely(!list_empty(&pool->sp_sockets))) {
495b1691bc0SJeff Layton xprt = list_first_entry(&pool->sp_sockets,
4960f0257eaSTom Tucker struct svc_xprt, xpt_ready);
4970f0257eaSTom Tucker list_del_init(&xprt->xpt_ready);
498b1691bc0SJeff Layton svc_xprt_get(xprt);
499b1691bc0SJeff Layton }
500b1691bc0SJeff Layton spin_unlock_bh(&pool->sp_lock);
50183a712e0SJeff Layton out:
5020f0257eaSTom Tucker return xprt;
5030f0257eaSTom Tucker }
5040f0257eaSTom Tucker
5050f0257eaSTom Tucker /**
5060f0257eaSTom Tucker * svc_reserve - change the space reserved for the reply to a request.
5070f0257eaSTom Tucker * @rqstp: The request in question
5080f0257eaSTom Tucker * @space: new max space to reserve
5090f0257eaSTom Tucker *
5100f0257eaSTom Tucker * Each request reserves some space on the output queue of the transport
5110f0257eaSTom Tucker * to make sure the reply fits. This function reduces that reserved
5120f0257eaSTom Tucker * space to be the amount of space used already, plus @space.
5130f0257eaSTom Tucker *
5140f0257eaSTom Tucker */
svc_reserve(struct svc_rqst * rqstp,int space)5150f0257eaSTom Tucker void svc_reserve(struct svc_rqst *rqstp, int space)
5160f0257eaSTom Tucker {
517d4b09acfSVasily Averin struct svc_xprt *xprt = rqstp->rq_xprt;
518d4b09acfSVasily Averin
5190f0257eaSTom Tucker space += rqstp->rq_res.head[0].iov_len;
5200f0257eaSTom Tucker
521d4b09acfSVasily Averin if (xprt && space < rqstp->rq_reserved) {
5220f0257eaSTom Tucker atomic_sub((rqstp->rq_reserved - space), &xprt->xpt_reserved);
5230f0257eaSTom Tucker rqstp->rq_reserved = space;
52495503d29SJ. Bruce Fields smp_wmb(); /* See smp_rmb() in svc_xprt_ready() */
5250f0257eaSTom Tucker svc_xprt_enqueue(xprt);
5260f0257eaSTom Tucker }
5270f0257eaSTom Tucker }
52824c3767eSTrond Myklebust EXPORT_SYMBOL_GPL(svc_reserve);
5290f0257eaSTom Tucker
free_deferred(struct svc_xprt * xprt,struct svc_deferred_req * dr)530948f072aSNeilBrown static void free_deferred(struct svc_xprt *xprt, struct svc_deferred_req *dr)
531948f072aSNeilBrown {
532948f072aSNeilBrown if (!dr)
533948f072aSNeilBrown return;
534948f072aSNeilBrown
535948f072aSNeilBrown xprt->xpt_ops->xpo_release_ctxt(xprt, dr->xprt_ctxt);
536948f072aSNeilBrown kfree(dr);
537948f072aSNeilBrown }
538948f072aSNeilBrown
svc_xprt_release(struct svc_rqst * rqstp)5390f0257eaSTom Tucker static void svc_xprt_release(struct svc_rqst *rqstp)
5400f0257eaSTom Tucker {
5410f0257eaSTom Tucker struct svc_xprt *xprt = rqstp->rq_xprt;
5420f0257eaSTom Tucker
543948f072aSNeilBrown xprt->xpt_ops->xpo_release_ctxt(xprt, rqstp->rq_xprt_ctxt);
544948f072aSNeilBrown rqstp->rq_xprt_ctxt = NULL;
5450f0257eaSTom Tucker
546948f072aSNeilBrown free_deferred(xprt, rqstp->rq_deferred);
5472779e3aeSTom Tucker rqstp->rq_deferred = NULL;
5482779e3aeSTom Tucker
549b20cb39dSChuck Lever svc_rqst_release_pages(rqstp);
5500f0257eaSTom Tucker rqstp->rq_res.page_len = 0;
5510f0257eaSTom Tucker rqstp->rq_res.page_base = 0;
5520f0257eaSTom Tucker
5530f0257eaSTom Tucker /* Reset response buffer and release
5540f0257eaSTom Tucker * the reservation.
5550f0257eaSTom Tucker * But first, check that enough space was reserved
5560f0257eaSTom Tucker * for the reply, otherwise we have a bug!
5570f0257eaSTom Tucker */
5580f0257eaSTom Tucker if ((rqstp->rq_res.len) > rqstp->rq_reserved)
5590f0257eaSTom Tucker printk(KERN_ERR "RPC request reserved %d but used %d\n",
5600f0257eaSTom Tucker rqstp->rq_reserved,
5610f0257eaSTom Tucker rqstp->rq_res.len);
5620f0257eaSTom Tucker
5630f0257eaSTom Tucker rqstp->rq_res.head[0].iov_len = 0;
5640f0257eaSTom Tucker svc_reserve(rqstp, 0);
565ff3ac5c3STrond Myklebust svc_xprt_release_slot(rqstp);
5660f0257eaSTom Tucker rqstp->rq_xprt = NULL;
5670f0257eaSTom Tucker svc_xprt_put(xprt);
5680f0257eaSTom Tucker }
5690f0257eaSTom Tucker
570850bac3aSChuck Lever /**
571850bac3aSChuck Lever * svc_wake_up - Wake up a service thread for non-transport work
572850bac3aSChuck Lever * @serv: RPC service
573850bac3aSChuck Lever *
574ceff739cSJeff Layton * Some svc_serv's will have occasional work to do, even when a xprt is not
575ceff739cSJeff Layton * waiting to be serviced. This function is there to "kick" a task in one of
576ceff739cSJeff Layton * those services so that it can wake up and do that work. Note that we only
577ceff739cSJeff Layton * bother with pool 0 as we don't need to wake up more than one thread for
578ceff739cSJeff Layton * this purpose.
5790f0257eaSTom Tucker */
svc_wake_up(struct svc_serv * serv)5800f0257eaSTom Tucker void svc_wake_up(struct svc_serv *serv)
5810f0257eaSTom Tucker {
582850bac3aSChuck Lever struct svc_pool *pool = &serv->sv_pools[0];
5830f0257eaSTom Tucker
5844d5db3f5SJeff Layton set_bit(SP_TASK_PENDING, &pool->sp_flags);
5856859d1f2SNeilBrown svc_pool_wake_idle_thread(pool);
5860f0257eaSTom Tucker }
58724c3767eSTrond Myklebust EXPORT_SYMBOL_GPL(svc_wake_up);
5880f0257eaSTom Tucker
svc_port_is_privileged(struct sockaddr * sin)5890f0257eaSTom Tucker int svc_port_is_privileged(struct sockaddr *sin)
5900f0257eaSTom Tucker {
5910f0257eaSTom Tucker switch (sin->sa_family) {
5920f0257eaSTom Tucker case AF_INET:
5930f0257eaSTom Tucker return ntohs(((struct sockaddr_in *)sin)->sin_port)
5940f0257eaSTom Tucker < PROT_SOCK;
5950f0257eaSTom Tucker case AF_INET6:
5960f0257eaSTom Tucker return ntohs(((struct sockaddr_in6 *)sin)->sin6_port)
5970f0257eaSTom Tucker < PROT_SOCK;
5980f0257eaSTom Tucker default:
5990f0257eaSTom Tucker return 0;
6000f0257eaSTom Tucker }
6010f0257eaSTom Tucker }
6020f0257eaSTom Tucker
6030f0257eaSTom Tucker /*
604c9233eb7SJeff Layton * Make sure that we don't have too many active connections. If we have,
605c9233eb7SJeff Layton * something must be dropped. It's not clear what will happen if we allow
606c9233eb7SJeff Layton * "too many" connections, but when dealing with network-facing software,
607c9233eb7SJeff Layton * we have to code defensively. Here we do that by imposing hard limits.
6080f0257eaSTom Tucker *
6090f0257eaSTom Tucker * There's no point in trying to do random drop here for DoS
6100f0257eaSTom Tucker * prevention. The NFS clients does 1 reconnect in 15 seconds. An
6110f0257eaSTom Tucker * attacker can easily beat that.
6120f0257eaSTom Tucker *
6130f0257eaSTom Tucker * The only somewhat efficient mechanism would be if drop old
6140f0257eaSTom Tucker * connections from the same IP first. But right now we don't even
6150f0257eaSTom Tucker * record the client IP in svc_sock.
616c9233eb7SJeff Layton *
617c9233eb7SJeff Layton * single-threaded services that expect a lot of clients will probably
618c9233eb7SJeff Layton * need to set sv_maxconn to override the default value which is based
619c9233eb7SJeff Layton * on the number of threads
6200f0257eaSTom Tucker */
svc_check_conn_limits(struct svc_serv * serv)6210f0257eaSTom Tucker static void svc_check_conn_limits(struct svc_serv *serv)
6220f0257eaSTom Tucker {
623c9233eb7SJeff Layton unsigned int limit = serv->sv_maxconn ? serv->sv_maxconn :
624c9233eb7SJeff Layton (serv->sv_nrthreads+3) * 20;
625c9233eb7SJeff Layton
626c9233eb7SJeff Layton if (serv->sv_tmpcnt > limit) {
6270f0257eaSTom Tucker struct svc_xprt *xprt = NULL;
6280f0257eaSTom Tucker spin_lock_bh(&serv->sv_lock);
6290f0257eaSTom Tucker if (!list_empty(&serv->sv_tempsocks)) {
6300f0257eaSTom Tucker /* Try to help the admin */
631e87cc472SJoe Perches net_notice_ratelimited("%s: too many open connections, consider increasing the %s\n",
632c9233eb7SJeff Layton serv->sv_name, serv->sv_maxconn ?
633e87cc472SJoe Perches "max number of connections" :
634e87cc472SJoe Perches "number of threads");
6350f0257eaSTom Tucker /*
6360f0257eaSTom Tucker * Always select the oldest connection. It's not fair,
6370f0257eaSTom Tucker * but so is life
6380f0257eaSTom Tucker */
6390f0257eaSTom Tucker xprt = list_entry(serv->sv_tempsocks.prev,
6400f0257eaSTom Tucker struct svc_xprt,
6410f0257eaSTom Tucker xpt_list);
6420f0257eaSTom Tucker set_bit(XPT_CLOSE, &xprt->xpt_flags);
6430f0257eaSTom Tucker svc_xprt_get(xprt);
6440f0257eaSTom Tucker }
6450f0257eaSTom Tucker spin_unlock_bh(&serv->sv_lock);
6460f0257eaSTom Tucker
6470f0257eaSTom Tucker if (xprt) {
6480f0257eaSTom Tucker svc_xprt_enqueue(xprt);
6490f0257eaSTom Tucker svc_xprt_put(xprt);
6500f0257eaSTom Tucker }
6510f0257eaSTom Tucker }
6520f0257eaSTom Tucker }
6530f0257eaSTom Tucker
svc_alloc_arg(struct svc_rqst * rqstp)6547b719e2bSNeilBrown static bool svc_alloc_arg(struct svc_rqst *rqstp)
6550f0257eaSTom Tucker {
6560f0257eaSTom Tucker struct svc_serv *serv = rqstp->rq_server;
657ab836264SChuck Lever struct xdr_buf *arg = &rqstp->rq_arg;
658e38b3f20SNeilBrown unsigned long pages, filled, ret;
6590f0257eaSTom Tucker
6608c6ae498SChuck Lever pages = (serv->sv_max_mesg + 2 * PAGE_SIZE) >> PAGE_SHIFT;
6618c6ae498SChuck Lever if (pages > RPCSVC_MAXPAGES) {
662f6e70aabSChuck Lever pr_warn_once("svc: warning: pages=%lu > RPCSVC_MAXPAGES=%lu\n",
6638c6ae498SChuck Lever pages, RPCSVC_MAXPAGES);
664b25cd058SWeston Andros Adamson /* use as many pages as possible */
6658c6ae498SChuck Lever pages = RPCSVC_MAXPAGES;
6668c6ae498SChuck Lever }
667f6e70aabSChuck Lever
668e38b3f20SNeilBrown for (filled = 0; filled < pages; filled = ret) {
669*f685ef2cSChuck Lever ret = alloc_pages_bulk_array(GFP_KERNEL, pages,
670*f685ef2cSChuck Lever rqstp->rq_pages);
671e38b3f20SNeilBrown if (ret > filled)
672e38b3f20SNeilBrown /* Made progress, don't sleep yet */
673e38b3f20SNeilBrown continue;
674f6e70aabSChuck Lever
67539039024SNeilBrown set_current_state(TASK_IDLE);
67639039024SNeilBrown if (kthread_should_stop()) {
6777b54fe61SJeff Layton set_current_state(TASK_RUNNING);
6787b719e2bSNeilBrown return false;
6797b54fe61SJeff Layton }
68028fffa6cSChuck Lever trace_svc_alloc_arg_err(pages, ret);
6814034247aSNeilBrown memalloc_retry_wait(GFP_KERNEL);
6820f0257eaSTom Tucker }
683ab836264SChuck Lever rqstp->rq_page_end = &rqstp->rq_pages[pages];
684ab836264SChuck Lever rqstp->rq_pages[pages] = NULL; /* this might be seen in nfsd_splice_actor() */
6850f0257eaSTom Tucker
6860f0257eaSTom Tucker /* Make arg->head point to first page and arg->pages point to rest */
6870f0257eaSTom Tucker arg->head[0].iov_base = page_address(rqstp->rq_pages[0]);
6880f0257eaSTom Tucker arg->head[0].iov_len = PAGE_SIZE;
6890f0257eaSTom Tucker arg->pages = rqstp->rq_pages + 1;
6900f0257eaSTom Tucker arg->page_base = 0;
6910f0257eaSTom Tucker /* save at least one page for response */
6920f0257eaSTom Tucker arg->page_len = (pages-2)*PAGE_SIZE;
6930f0257eaSTom Tucker arg->len = (pages-1)*PAGE_SIZE;
6940f0257eaSTom Tucker arg->tail[0].iov_len = 0;
695695bc1f3SChuck Lever
696695bc1f3SChuck Lever rqstp->rq_xid = xdr_zero;
6977b719e2bSNeilBrown return true;
6986797fa5aSJ. Bruce Fields }
6990f0257eaSTom Tucker
700b1691bc0SJeff Layton static bool
rqst_should_sleep(struct svc_rqst * rqstp)701b1691bc0SJeff Layton rqst_should_sleep(struct svc_rqst *rqstp)
702b1691bc0SJeff Layton {
703b1691bc0SJeff Layton struct svc_pool *pool = rqstp->rq_pool;
704b1691bc0SJeff Layton
705b1691bc0SJeff Layton /* did someone call svc_wake_up? */
7066859d1f2SNeilBrown if (test_bit(SP_TASK_PENDING, &pool->sp_flags))
707b1691bc0SJeff Layton return false;
708b1691bc0SJeff Layton
709b1691bc0SJeff Layton /* was a socket queued? */
710b1691bc0SJeff Layton if (!list_empty(&pool->sp_sockets))
711b1691bc0SJeff Layton return false;
712b1691bc0SJeff Layton
713b1691bc0SJeff Layton /* are we shutting down? */
71439039024SNeilBrown if (kthread_should_stop())
715b1691bc0SJeff Layton return false;
716b1691bc0SJeff Layton
717b1691bc0SJeff Layton /* are we freezing? */
718b1691bc0SJeff Layton if (freezing(current))
719b1691bc0SJeff Layton return false;
720b1691bc0SJeff Layton
721b1691bc0SJeff Layton return true;
722b1691bc0SJeff Layton }
723b1691bc0SJeff Layton
svc_get_next_xprt(struct svc_rqst * rqstp)724c743b425SNeilBrown static struct svc_xprt *svc_get_next_xprt(struct svc_rqst *rqstp)
7256797fa5aSJ. Bruce Fields {
7266797fa5aSJ. Bruce Fields struct svc_pool *pool = rqstp->rq_pool;
7270f0257eaSTom Tucker
728b1691bc0SJeff Layton /* rq_xprt should be clear on entry */
729b1691bc0SJeff Layton WARN_ON_ONCE(rqstp->rq_xprt);
730b1691bc0SJeff Layton
73122700f3cSTrond Myklebust rqstp->rq_xprt = svc_xprt_dequeue(pool);
73222700f3cSTrond Myklebust if (rqstp->rq_xprt)
73322700f3cSTrond Myklebust goto out_found;
734b1691bc0SJeff Layton
73539039024SNeilBrown set_current_state(TASK_IDLE);
73622700f3cSTrond Myklebust smp_mb__before_atomic();
73722700f3cSTrond Myklebust clear_bit(SP_CONGESTED, &pool->sp_flags);
738b1691bc0SJeff Layton clear_bit(RQ_BUSY, &rqstp->rq_flags);
73922700f3cSTrond Myklebust smp_mb__after_atomic();
7407086721fSJeff Layton
741b1691bc0SJeff Layton if (likely(rqst_should_sleep(rqstp)))
742c743b425SNeilBrown schedule();
743b1691bc0SJeff Layton else
744983c6844STrond Myklebust __set_current_state(TASK_RUNNING);
7450f0257eaSTom Tucker
7460f0257eaSTom Tucker try_to_freeze();
7470f0257eaSTom Tucker
748b1691bc0SJeff Layton set_bit(RQ_BUSY, &rqstp->rq_flags);
74922700f3cSTrond Myklebust smp_mb__after_atomic();
7506859d1f2SNeilBrown clear_bit(SP_TASK_PENDING, &pool->sp_flags);
75122700f3cSTrond Myklebust rqstp->rq_xprt = svc_xprt_dequeue(pool);
75222700f3cSTrond Myklebust if (rqstp->rq_xprt)
75322700f3cSTrond Myklebust goto out_found;
754106f359cSTrond Myklebust
75539039024SNeilBrown if (kthread_should_stop())
7567b719e2bSNeilBrown return NULL;
7577b719e2bSNeilBrown return NULL;
75822700f3cSTrond Myklebust out_found:
7596859d1f2SNeilBrown clear_bit(SP_TASK_PENDING, &pool->sp_flags);
76022700f3cSTrond Myklebust /* Normally we will wait up to 5 seconds for any required
76122700f3cSTrond Myklebust * cache information to be provided.
76222700f3cSTrond Myklebust */
76322700f3cSTrond Myklebust if (!test_bit(SP_CONGESTED, &pool->sp_flags))
76422700f3cSTrond Myklebust rqstp->rq_chandle.thread_wait = 5*HZ;
76522700f3cSTrond Myklebust else
76622700f3cSTrond Myklebust rqstp->rq_chandle.thread_wait = 1*HZ;
76755f5088cSChuck Lever trace_svc_xprt_dequeue(rqstp);
76822700f3cSTrond Myklebust return rqstp->rq_xprt;
7690f0257eaSTom Tucker }
7700f0257eaSTom Tucker
svc_add_new_temp_xprt(struct svc_serv * serv,struct svc_xprt * newxpt)771e1d83ee6SRashika Kheria static void svc_add_new_temp_xprt(struct svc_serv *serv, struct svc_xprt *newxpt)
77265b2e665SJ. Bruce Fields {
77365b2e665SJ. Bruce Fields spin_lock_bh(&serv->sv_lock);
77465b2e665SJ. Bruce Fields set_bit(XPT_TEMP, &newxpt->xpt_flags);
77565b2e665SJ. Bruce Fields list_add(&newxpt->xpt_list, &serv->sv_tempsocks);
77665b2e665SJ. Bruce Fields serv->sv_tmpcnt++;
77765b2e665SJ. Bruce Fields if (serv->sv_temptimer.function == NULL) {
77865b2e665SJ. Bruce Fields /* setup timer to age temp transports */
779841b86f3SKees Cook serv->sv_temptimer.function = svc_age_temp_xprts;
78065b2e665SJ. Bruce Fields mod_timer(&serv->sv_temptimer,
78165b2e665SJ. Bruce Fields jiffies + svc_conn_age_period * HZ);
78265b2e665SJ. Bruce Fields }
78365b2e665SJ. Bruce Fields spin_unlock_bh(&serv->sv_lock);
78465b2e665SJ. Bruce Fields svc_xprt_received(newxpt);
78565b2e665SJ. Bruce Fields }
78665b2e665SJ. Bruce Fields
svc_handle_xprt(struct svc_rqst * rqstp,struct svc_xprt * xprt)7876797fa5aSJ. Bruce Fields static int svc_handle_xprt(struct svc_rqst *rqstp, struct svc_xprt *xprt)
7886797fa5aSJ. Bruce Fields {
7896797fa5aSJ. Bruce Fields struct svc_serv *serv = rqstp->rq_server;
7906797fa5aSJ. Bruce Fields int len = 0;
7916797fa5aSJ. Bruce Fields
7921b644b6eSJ. Bruce Fields if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
793546125d1SScott Mayhew if (test_and_clear_bit(XPT_KILL_TEMP, &xprt->xpt_flags))
794546125d1SScott Mayhew xprt->xpt_ops->xpo_kill_temp_xprt(xprt);
7951b644b6eSJ. Bruce Fields svc_delete_xprt(xprt);
796ca7896cdSJ. Bruce Fields /* Leave XPT_BUSY set on the dead xprt: */
79783a712e0SJeff Layton goto out;
798ca7896cdSJ. Bruce Fields }
799ca7896cdSJ. Bruce Fields if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
8000f0257eaSTom Tucker struct svc_xprt *newxpt;
8010f0257eaSTom Tucker /*
8020f0257eaSTom Tucker * We know this module_get will succeed because the
8030f0257eaSTom Tucker * listener holds a reference too
8040f0257eaSTom Tucker */
80565b2e665SJ. Bruce Fields __module_get(xprt->xpt_class->xcl_owner);
8060f0257eaSTom Tucker svc_check_conn_limits(xprt->xpt_server);
80765b2e665SJ. Bruce Fields newxpt = xprt->xpt_ops->xpo_accept(xprt);
8081237d354STrond Myklebust if (newxpt) {
8091237d354STrond Myklebust newxpt->xpt_cred = get_cred(xprt->xpt_cred);
81065b2e665SJ. Bruce Fields svc_add_new_temp_xprt(serv, newxpt);
81111bbb0f7SChuck Lever trace_svc_xprt_accept(newxpt, serv->sv_name);
81282011c80SChuck Lever } else {
813c789102cSTrond Myklebust module_put(xprt->xpt_class->xcl_owner);
81482011c80SChuck Lever }
81582011c80SChuck Lever svc_xprt_received(xprt);
816b3cbf98eSChuck Lever } else if (test_bit(XPT_HANDSHAKE, &xprt->xpt_flags)) {
817b3cbf98eSChuck Lever xprt->xpt_ops->xpo_handshake(xprt);
818b3cbf98eSChuck Lever svc_xprt_received(xprt);
819ff3ac5c3STrond Myklebust } else if (svc_xprt_reserve_slot(rqstp, xprt)) {
8206797fa5aSJ. Bruce Fields /* XPT_DATA|XPT_DEFERRED case: */
8210f0257eaSTom Tucker rqstp->rq_deferred = svc_deferred_dequeue(xprt);
822ca7896cdSJ. Bruce Fields if (rqstp->rq_deferred)
8230f0257eaSTom Tucker len = svc_deferred_recv(rqstp);
824ca7896cdSJ. Bruce Fields else
8250f0257eaSTom Tucker len = xprt->xpt_ops->xpo_recvfrom(rqstp);
826d10f27a7SJ. Bruce Fields rqstp->rq_reserved = serv->sv_max_mesg;
827d10f27a7SJ. Bruce Fields atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
828062b829cSTrond Myklebust } else
829062b829cSTrond Myklebust svc_xprt_received(xprt);
8305089f3d9SChuck Lever
83183a712e0SJeff Layton out:
8326797fa5aSJ. Bruce Fields return len;
8336797fa5aSJ. Bruce Fields }
8346797fa5aSJ. Bruce Fields
835c743b425SNeilBrown /**
836c743b425SNeilBrown * svc_recv - Receive and process the next request on any transport
837c743b425SNeilBrown * @rqstp: an idle RPC service thread
838c743b425SNeilBrown *
839c743b425SNeilBrown * This code is carefully organised not to touch any cachelines in
840c743b425SNeilBrown * the shared svc_serv structure, only cachelines in the local
841c743b425SNeilBrown * svc_pool.
8426797fa5aSJ. Bruce Fields */
svc_recv(struct svc_rqst * rqstp)843c743b425SNeilBrown void svc_recv(struct svc_rqst *rqstp)
8446797fa5aSJ. Bruce Fields {
8456797fa5aSJ. Bruce Fields struct svc_xprt *xprt = NULL;
8466797fa5aSJ. Bruce Fields struct svc_serv *serv = rqstp->rq_server;
8477b719e2bSNeilBrown int len;
8486797fa5aSJ. Bruce Fields
8497b719e2bSNeilBrown if (!svc_alloc_arg(rqstp))
850860a0d9eSJeff Layton goto out;
8516797fa5aSJ. Bruce Fields
8526797fa5aSJ. Bruce Fields try_to_freeze();
8536797fa5aSJ. Bruce Fields cond_resched();
85439039024SNeilBrown if (kthread_should_stop())
855860a0d9eSJeff Layton goto out;
8566797fa5aSJ. Bruce Fields
857c743b425SNeilBrown xprt = svc_get_next_xprt(rqstp);
8587b719e2bSNeilBrown if (!xprt)
859860a0d9eSJeff Layton goto out;
8606797fa5aSJ. Bruce Fields
8616797fa5aSJ. Bruce Fields len = svc_handle_xprt(rqstp, xprt);
8620f0257eaSTom Tucker
8630f0257eaSTom Tucker /* No data, incomplete (TCP) read, or accept() */
8649f9d2ebeSJ. Bruce Fields if (len <= 0)
865860a0d9eSJeff Layton goto out_release;
866f8335a21SChuck Lever
8675f39d271SChuck Lever trace_svc_xdr_recvfrom(&rqstp->rq_arg);
868ca7896cdSJ. Bruce Fields
8690f0257eaSTom Tucker clear_bit(XPT_OLD, &xprt->xpt_flags);
8700f0257eaSTom Tucker
8710f0257eaSTom Tucker rqstp->rq_chandle.defer = svc_defer;
8720f0257eaSTom Tucker
8730f0257eaSTom Tucker if (serv->sv_stats)
8740f0257eaSTom Tucker serv->sv_stats->netcnt++;
875f208e950SChuck Lever percpu_counter_inc(&rqstp->rq_pool->sp_messages_arrived);
876f8335a21SChuck Lever rqstp->rq_stime = ktime_get();
877f78116d3SNeilBrown svc_process(rqstp);
8787b719e2bSNeilBrown out:
8797b719e2bSNeilBrown return;
880860a0d9eSJeff Layton out_release:
881ca7896cdSJ. Bruce Fields rqstp->rq_res.len = 0;
882ca7896cdSJ. Bruce Fields svc_xprt_release(rqstp);
8830f0257eaSTom Tucker }
88424c3767eSTrond Myklebust EXPORT_SYMBOL_GPL(svc_recv);
8850f0257eaSTom Tucker
8860f0257eaSTom Tucker /*
8870f0257eaSTom Tucker * Drop request
8880f0257eaSTom Tucker */
svc_drop(struct svc_rqst * rqstp)8890f0257eaSTom Tucker void svc_drop(struct svc_rqst *rqstp)
8900f0257eaSTom Tucker {
891104f6351STrond Myklebust trace_svc_drop(rqstp);
8920f0257eaSTom Tucker svc_xprt_release(rqstp);
8930f0257eaSTom Tucker }
89424c3767eSTrond Myklebust EXPORT_SYMBOL_GPL(svc_drop);
8950f0257eaSTom Tucker
89655fcc7d9SChuck Lever /**
89755fcc7d9SChuck Lever * svc_send - Return reply to client
89855fcc7d9SChuck Lever * @rqstp: RPC transaction context
89955fcc7d9SChuck Lever *
9000f0257eaSTom Tucker */
svc_send(struct svc_rqst * rqstp)90155fcc7d9SChuck Lever void svc_send(struct svc_rqst *rqstp)
9020f0257eaSTom Tucker {
9030f0257eaSTom Tucker struct svc_xprt *xprt;
9040f0257eaSTom Tucker struct xdr_buf *xb;
90555fcc7d9SChuck Lever int status;
9060f0257eaSTom Tucker
9070f0257eaSTom Tucker xprt = rqstp->rq_xprt;
9080f0257eaSTom Tucker if (!xprt)
90955fcc7d9SChuck Lever return;
9100f0257eaSTom Tucker
9110f0257eaSTom Tucker /* calculate over-all length */
9120f0257eaSTom Tucker xb = &rqstp->rq_res;
9130f0257eaSTom Tucker xb->len = xb->head[0].iov_len +
9140f0257eaSTom Tucker xb->page_len +
9150f0257eaSTom Tucker xb->tail[0].iov_len;
9165f39d271SChuck Lever trace_svc_xdr_sendto(rqstp->rq_xid, xb);
917aaba72cdSChuck Lever trace_svc_stats_latency(rqstp);
918ca4faf54SChuck Lever
91955fcc7d9SChuck Lever status = xprt->xpt_ops->xpo_sendto(rqstp);
920ca4faf54SChuck Lever
92155fcc7d9SChuck Lever trace_svc_send(rqstp, status);
9220f0257eaSTom Tucker svc_xprt_release(rqstp);
9230f0257eaSTom Tucker }
9240f0257eaSTom Tucker
9250f0257eaSTom Tucker /*
9260f0257eaSTom Tucker * Timer function to close old temporary transports, using
9270f0257eaSTom Tucker * a mark-and-sweep algorithm.
9280f0257eaSTom Tucker */
svc_age_temp_xprts(struct timer_list * t)929ff861c4dSKees Cook static void svc_age_temp_xprts(struct timer_list *t)
9300f0257eaSTom Tucker {
931ff861c4dSKees Cook struct svc_serv *serv = from_timer(serv, t, sv_temptimer);
9320f0257eaSTom Tucker struct svc_xprt *xprt;
9330f0257eaSTom Tucker struct list_head *le, *next;
9340f0257eaSTom Tucker
9350f0257eaSTom Tucker dprintk("svc_age_temp_xprts\n");
9360f0257eaSTom Tucker
9370f0257eaSTom Tucker if (!spin_trylock_bh(&serv->sv_lock)) {
9380f0257eaSTom Tucker /* busy, try again 1 sec later */
9390f0257eaSTom Tucker dprintk("svc_age_temp_xprts: busy\n");
9400f0257eaSTom Tucker mod_timer(&serv->sv_temptimer, jiffies + HZ);
9410f0257eaSTom Tucker return;
9420f0257eaSTom Tucker }
9430f0257eaSTom Tucker
9440f0257eaSTom Tucker list_for_each_safe(le, next, &serv->sv_tempsocks) {
9450f0257eaSTom Tucker xprt = list_entry(le, struct svc_xprt, xpt_list);
9460f0257eaSTom Tucker
9470f0257eaSTom Tucker /* First time through, just mark it OLD. Second time
9480f0257eaSTom Tucker * through, close it. */
9490f0257eaSTom Tucker if (!test_and_set_bit(XPT_OLD, &xprt->xpt_flags))
9500f0257eaSTom Tucker continue;
9512c935bc5SPeter Zijlstra if (kref_read(&xprt->xpt_ref) > 1 ||
952f64f9e71SJoe Perches test_bit(XPT_BUSY, &xprt->xpt_flags))
9530f0257eaSTom Tucker continue;
954e75bafbfSJ. Bruce Fields list_del_init(le);
9550f0257eaSTom Tucker set_bit(XPT_CLOSE, &xprt->xpt_flags);
9560f0257eaSTom Tucker dprintk("queuing xprt %p for closing\n", xprt);
9570f0257eaSTom Tucker
9580f0257eaSTom Tucker /* a thread will dequeue and close it soon */
9590f0257eaSTom Tucker svc_xprt_enqueue(xprt);
9600f0257eaSTom Tucker }
961e75bafbfSJ. Bruce Fields spin_unlock_bh(&serv->sv_lock);
9620f0257eaSTom Tucker
9630f0257eaSTom Tucker mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ);
9640f0257eaSTom Tucker }
9650f0257eaSTom Tucker
966c3d4879eSScott Mayhew /* Close temporary transports whose xpt_local matches server_addr immediately
967c3d4879eSScott Mayhew * instead of waiting for them to be picked up by the timer.
968c3d4879eSScott Mayhew *
969c3d4879eSScott Mayhew * This is meant to be called from a notifier_block that runs when an ip
970c3d4879eSScott Mayhew * address is deleted.
971c3d4879eSScott Mayhew */
svc_age_temp_xprts_now(struct svc_serv * serv,struct sockaddr * server_addr)972c3d4879eSScott Mayhew void svc_age_temp_xprts_now(struct svc_serv *serv, struct sockaddr *server_addr)
973c3d4879eSScott Mayhew {
974c3d4879eSScott Mayhew struct svc_xprt *xprt;
975c3d4879eSScott Mayhew struct list_head *le, *next;
976c3d4879eSScott Mayhew LIST_HEAD(to_be_closed);
977c3d4879eSScott Mayhew
978c3d4879eSScott Mayhew spin_lock_bh(&serv->sv_lock);
979c3d4879eSScott Mayhew list_for_each_safe(le, next, &serv->sv_tempsocks) {
980c3d4879eSScott Mayhew xprt = list_entry(le, struct svc_xprt, xpt_list);
981c3d4879eSScott Mayhew if (rpc_cmp_addr(server_addr, (struct sockaddr *)
982c3d4879eSScott Mayhew &xprt->xpt_local)) {
983c3d4879eSScott Mayhew dprintk("svc_age_temp_xprts_now: found %p\n", xprt);
984c3d4879eSScott Mayhew list_move(le, &to_be_closed);
985c3d4879eSScott Mayhew }
986c3d4879eSScott Mayhew }
987c3d4879eSScott Mayhew spin_unlock_bh(&serv->sv_lock);
988c3d4879eSScott Mayhew
989c3d4879eSScott Mayhew while (!list_empty(&to_be_closed)) {
990c3d4879eSScott Mayhew le = to_be_closed.next;
991c3d4879eSScott Mayhew list_del_init(le);
992c3d4879eSScott Mayhew xprt = list_entry(le, struct svc_xprt, xpt_list);
993546125d1SScott Mayhew set_bit(XPT_CLOSE, &xprt->xpt_flags);
994546125d1SScott Mayhew set_bit(XPT_KILL_TEMP, &xprt->xpt_flags);
995546125d1SScott Mayhew dprintk("svc_age_temp_xprts_now: queuing xprt %p for closing\n",
996546125d1SScott Mayhew xprt);
997546125d1SScott Mayhew svc_xprt_enqueue(xprt);
998c3d4879eSScott Mayhew }
999c3d4879eSScott Mayhew }
1000c3d4879eSScott Mayhew EXPORT_SYMBOL_GPL(svc_age_temp_xprts_now);
1001c3d4879eSScott Mayhew
call_xpt_users(struct svc_xprt * xprt)1002edc7a894SJ. Bruce Fields static void call_xpt_users(struct svc_xprt *xprt)
1003edc7a894SJ. Bruce Fields {
1004edc7a894SJ. Bruce Fields struct svc_xpt_user *u;
1005edc7a894SJ. Bruce Fields
1006edc7a894SJ. Bruce Fields spin_lock(&xprt->xpt_lock);
1007edc7a894SJ. Bruce Fields while (!list_empty(&xprt->xpt_users)) {
1008edc7a894SJ. Bruce Fields u = list_first_entry(&xprt->xpt_users, struct svc_xpt_user, list);
1009bb6ad557STrond Myklebust list_del_init(&u->list);
1010edc7a894SJ. Bruce Fields u->callback(u);
1011edc7a894SJ. Bruce Fields }
1012edc7a894SJ. Bruce Fields spin_unlock(&xprt->xpt_lock);
1013edc7a894SJ. Bruce Fields }
1014edc7a894SJ. Bruce Fields
10150f0257eaSTom Tucker /*
10160f0257eaSTom Tucker * Remove a dead transport
10170f0257eaSTom Tucker */
svc_delete_xprt(struct svc_xprt * xprt)10187710ec36SJ. Bruce Fields static void svc_delete_xprt(struct svc_xprt *xprt)
10190f0257eaSTom Tucker {
10200f0257eaSTom Tucker struct svc_serv *serv = xprt->xpt_server;
102122945e4aSTom Tucker struct svc_deferred_req *dr;
102222945e4aSTom Tucker
102322945e4aSTom Tucker if (test_and_set_bit(XPT_DEAD, &xprt->xpt_flags))
102411bbb0f7SChuck Lever return;
10250f0257eaSTom Tucker
102611bbb0f7SChuck Lever trace_svc_xprt_detach(xprt);
10270f0257eaSTom Tucker xprt->xpt_ops->xpo_detach(xprt);
10286221f1d9SChuck Lever if (xprt->xpt_bc_xprt)
10296221f1d9SChuck Lever xprt->xpt_bc_xprt->ops->close(xprt->xpt_bc_xprt);
10300f0257eaSTom Tucker
10310f0257eaSTom Tucker spin_lock_bh(&serv->sv_lock);
10320f0257eaSTom Tucker list_del_init(&xprt->xpt_list);
103301047298SWeston Andros Adamson WARN_ON_ONCE(!list_empty(&xprt->xpt_ready));
10340f0257eaSTom Tucker if (test_bit(XPT_TEMP, &xprt->xpt_flags))
10350f0257eaSTom Tucker serv->sv_tmpcnt--;
1036788e69e5SJ. Bruce Fields spin_unlock_bh(&serv->sv_lock);
103722945e4aSTom Tucker
1038ab1b18f7SNeil Brown while ((dr = svc_deferred_dequeue(xprt)) != NULL)
1039948f072aSNeilBrown free_deferred(xprt, dr);
104022945e4aSTom Tucker
1041edc7a894SJ. Bruce Fields call_xpt_users(xprt);
104222945e4aSTom Tucker svc_xprt_put(xprt);
10430f0257eaSTom Tucker }
10440f0257eaSTom Tucker
10454355d767SChuck Lever /**
10464355d767SChuck Lever * svc_xprt_close - Close a client connection
10474355d767SChuck Lever * @xprt: transport to disconnect
10484355d767SChuck Lever *
10494355d767SChuck Lever */
svc_xprt_close(struct svc_xprt * xprt)10504355d767SChuck Lever void svc_xprt_close(struct svc_xprt *xprt)
10510f0257eaSTom Tucker {
105211bbb0f7SChuck Lever trace_svc_xprt_close(xprt);
10530f0257eaSTom Tucker set_bit(XPT_CLOSE, &xprt->xpt_flags);
10540f0257eaSTom Tucker if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
10550f0257eaSTom Tucker /* someone else will have to effect the close */
10560f0257eaSTom Tucker return;
1057b1763316SJ. Bruce Fields /*
1058b1763316SJ. Bruce Fields * We expect svc_close_xprt() to work even when no threads are
1059b1763316SJ. Bruce Fields * running (e.g., while configuring the server before starting
1060b1763316SJ. Bruce Fields * any threads), so if the transport isn't busy, we delete
1061b1763316SJ. Bruce Fields * it ourself:
1062b1763316SJ. Bruce Fields */
10630f0257eaSTom Tucker svc_delete_xprt(xprt);
10640f0257eaSTom Tucker }
10654355d767SChuck Lever EXPORT_SYMBOL_GPL(svc_xprt_close);
10660f0257eaSTom Tucker
svc_close_list(struct svc_serv * serv,struct list_head * xprt_list,struct net * net)1067cc630d9fSJ. Bruce Fields static int svc_close_list(struct svc_serv *serv, struct list_head *xprt_list, struct net *net)
10680f0257eaSTom Tucker {
10690f0257eaSTom Tucker struct svc_xprt *xprt;
1070cc630d9fSJ. Bruce Fields int ret = 0;
10710f0257eaSTom Tucker
1072c7de87ffSJoe Korty spin_lock_bh(&serv->sv_lock);
1073b4f36f88SJ. Bruce Fields list_for_each_entry(xprt, xprt_list, xpt_list) {
10747b147f1fSStanislav Kinsbursky if (xprt->xpt_net != net)
10757b147f1fSStanislav Kinsbursky continue;
1076cc630d9fSJ. Bruce Fields ret++;
10770f0257eaSTom Tucker set_bit(XPT_CLOSE, &xprt->xpt_flags);
1078cc630d9fSJ. Bruce Fields svc_xprt_enqueue(xprt);
10790f0257eaSTom Tucker }
1080c7de87ffSJoe Korty spin_unlock_bh(&serv->sv_lock);
1081cc630d9fSJ. Bruce Fields return ret;
10820f0257eaSTom Tucker }
10830f0257eaSTom Tucker
svc_dequeue_net(struct svc_serv * serv,struct net * net)1084cc630d9fSJ. Bruce Fields static struct svc_xprt *svc_dequeue_net(struct svc_serv *serv, struct net *net)
10852fefb8a0SJ. Bruce Fields {
1086b4f36f88SJ. Bruce Fields struct svc_pool *pool;
1087b4f36f88SJ. Bruce Fields struct svc_xprt *xprt;
1088b4f36f88SJ. Bruce Fields struct svc_xprt *tmp;
1089b4f36f88SJ. Bruce Fields int i;
1090b4f36f88SJ. Bruce Fields
1091b4f36f88SJ. Bruce Fields for (i = 0; i < serv->sv_nrpools; i++) {
1092b4f36f88SJ. Bruce Fields pool = &serv->sv_pools[i];
1093b4f36f88SJ. Bruce Fields
1094b4f36f88SJ. Bruce Fields spin_lock_bh(&pool->sp_lock);
10956f513365SStanislav Kinsbursky list_for_each_entry_safe(xprt, tmp, &pool->sp_sockets, xpt_ready) {
10967b147f1fSStanislav Kinsbursky if (xprt->xpt_net != net)
10977b147f1fSStanislav Kinsbursky continue;
1098b4f36f88SJ. Bruce Fields list_del_init(&xprt->xpt_ready);
1099cc630d9fSJ. Bruce Fields spin_unlock_bh(&pool->sp_lock);
1100cc630d9fSJ. Bruce Fields return xprt;
1101b4f36f88SJ. Bruce Fields }
1102b4f36f88SJ. Bruce Fields spin_unlock_bh(&pool->sp_lock);
1103b4f36f88SJ. Bruce Fields }
1104cc630d9fSJ. Bruce Fields return NULL;
11056f513365SStanislav Kinsbursky }
11066f513365SStanislav Kinsbursky
svc_clean_up_xprts(struct svc_serv * serv,struct net * net)1107cc630d9fSJ. Bruce Fields static void svc_clean_up_xprts(struct svc_serv *serv, struct net *net)
11086f513365SStanislav Kinsbursky {
11096f513365SStanislav Kinsbursky struct svc_xprt *xprt;
11106f513365SStanislav Kinsbursky
1111cc630d9fSJ. Bruce Fields while ((xprt = svc_dequeue_net(serv, net))) {
1112cc630d9fSJ. Bruce Fields set_bit(XPT_CLOSE, &xprt->xpt_flags);
1113719f8bccSJ. Bruce Fields svc_delete_xprt(xprt);
11143a22bf50SStanislav Kinsbursky }
1115cc630d9fSJ. Bruce Fields }
11163a22bf50SStanislav Kinsbursky
1117c7d7ec8fSChuck Lever /**
1118c7d7ec8fSChuck Lever * svc_xprt_destroy_all - Destroy transports associated with @serv
1119c7d7ec8fSChuck Lever * @serv: RPC service to be shut down
1120c7d7ec8fSChuck Lever * @net: target network namespace
1121c7d7ec8fSChuck Lever *
1122cc630d9fSJ. Bruce Fields * Server threads may still be running (especially in the case where the
1123cc630d9fSJ. Bruce Fields * service is still running in other network namespaces).
1124cc630d9fSJ. Bruce Fields *
1125cc630d9fSJ. Bruce Fields * So we shut down sockets the same way we would on a running server, by
1126cc630d9fSJ. Bruce Fields * setting XPT_CLOSE, enqueuing, and letting a thread pick it up to do
1127cc630d9fSJ. Bruce Fields * the close. In the case there are no such other threads,
1128cc630d9fSJ. Bruce Fields * threads running, svc_clean_up_xprts() does a simple version of a
1129cc630d9fSJ. Bruce Fields * server's main event loop, and in the case where there are other
1130cc630d9fSJ. Bruce Fields * threads, we may need to wait a little while and then check again to
1131cc630d9fSJ. Bruce Fields * see if they're done.
1132cc630d9fSJ. Bruce Fields */
svc_xprt_destroy_all(struct svc_serv * serv,struct net * net)1133c7d7ec8fSChuck Lever void svc_xprt_destroy_all(struct svc_serv *serv, struct net *net)
11343a22bf50SStanislav Kinsbursky {
1135cc630d9fSJ. Bruce Fields int delay = 0;
11366f513365SStanislav Kinsbursky
1137cc630d9fSJ. Bruce Fields while (svc_close_list(serv, &serv->sv_permsocks, net) +
1138cc630d9fSJ. Bruce Fields svc_close_list(serv, &serv->sv_tempsocks, net)) {
1139cc630d9fSJ. Bruce Fields
1140cc630d9fSJ. Bruce Fields svc_clean_up_xprts(serv, net);
1141cc630d9fSJ. Bruce Fields msleep(delay++);
1142cc630d9fSJ. Bruce Fields }
11432fefb8a0SJ. Bruce Fields }
1144c7d7ec8fSChuck Lever EXPORT_SYMBOL_GPL(svc_xprt_destroy_all);
11452fefb8a0SJ. Bruce Fields
11460f0257eaSTom Tucker /*
11470f0257eaSTom Tucker * Handle defer and revisit of requests
11480f0257eaSTom Tucker */
11490f0257eaSTom Tucker
svc_revisit(struct cache_deferred_req * dreq,int too_many)11500f0257eaSTom Tucker static void svc_revisit(struct cache_deferred_req *dreq, int too_many)
11510f0257eaSTom Tucker {
11520f0257eaSTom Tucker struct svc_deferred_req *dr =
11530f0257eaSTom Tucker container_of(dreq, struct svc_deferred_req, handle);
11540f0257eaSTom Tucker struct svc_xprt *xprt = dr->xprt;
11550f0257eaSTom Tucker
115622945e4aSTom Tucker spin_lock(&xprt->xpt_lock);
115722945e4aSTom Tucker set_bit(XPT_DEFERRED, &xprt->xpt_flags);
115822945e4aSTom Tucker if (too_many || test_bit(XPT_DEAD, &xprt->xpt_flags)) {
115922945e4aSTom Tucker spin_unlock(&xprt->xpt_lock);
11608954c5c2SChuck Lever trace_svc_defer_drop(dr);
1161948f072aSNeilBrown free_deferred(xprt, dr);
11620f0257eaSTom Tucker svc_xprt_put(xprt);
11630f0257eaSTom Tucker return;
11640f0257eaSTom Tucker }
11650f0257eaSTom Tucker dr->xprt = NULL;
11660f0257eaSTom Tucker list_add(&dr->handle.recent, &xprt->xpt_deferred);
11670f0257eaSTom Tucker spin_unlock(&xprt->xpt_lock);
11688954c5c2SChuck Lever trace_svc_defer_queue(dr);
11690f0257eaSTom Tucker svc_xprt_enqueue(xprt);
11700f0257eaSTom Tucker svc_xprt_put(xprt);
11710f0257eaSTom Tucker }
11720f0257eaSTom Tucker
1173260c1d12STom Tucker /*
1174260c1d12STom Tucker * Save the request off for later processing. The request buffer looks
1175260c1d12STom Tucker * like this:
1176260c1d12STom Tucker *
1177260c1d12STom Tucker * <xprt-header><rpc-header><rpc-pagelist><rpc-tail>
1178260c1d12STom Tucker *
1179260c1d12STom Tucker * This code can only handle requests that consist of an xprt-header
1180260c1d12STom Tucker * and rpc-header.
1181260c1d12STom Tucker */
svc_defer(struct cache_req * req)11820f0257eaSTom Tucker static struct cache_deferred_req *svc_defer(struct cache_req *req)
11830f0257eaSTom Tucker {
11840f0257eaSTom Tucker struct svc_rqst *rqstp = container_of(req, struct svc_rqst, rq_chandle);
11850f0257eaSTom Tucker struct svc_deferred_req *dr;
11860f0257eaSTom Tucker
118730660e04SJeff Layton if (rqstp->rq_arg.page_len || !test_bit(RQ_USEDEFERRAL, &rqstp->rq_flags))
11880f0257eaSTom Tucker return NULL; /* if more than a page, give up FIXME */
11890f0257eaSTom Tucker if (rqstp->rq_deferred) {
11900f0257eaSTom Tucker dr = rqstp->rq_deferred;
11910f0257eaSTom Tucker rqstp->rq_deferred = NULL;
11920f0257eaSTom Tucker } else {
1193260c1d12STom Tucker size_t skip;
1194260c1d12STom Tucker size_t size;
11950f0257eaSTom Tucker /* FIXME maybe discard if size too large */
1196260c1d12STom Tucker size = sizeof(struct svc_deferred_req) + rqstp->rq_arg.len;
11970f0257eaSTom Tucker dr = kmalloc(size, GFP_KERNEL);
11980f0257eaSTom Tucker if (dr == NULL)
11990f0257eaSTom Tucker return NULL;
12000f0257eaSTom Tucker
12010f0257eaSTom Tucker dr->handle.owner = rqstp->rq_server;
12020f0257eaSTom Tucker dr->prot = rqstp->rq_prot;
12030f0257eaSTom Tucker memcpy(&dr->addr, &rqstp->rq_addr, rqstp->rq_addrlen);
12040f0257eaSTom Tucker dr->addrlen = rqstp->rq_addrlen;
12050f0257eaSTom Tucker dr->daddr = rqstp->rq_daddr;
12060f0257eaSTom Tucker dr->argslen = rqstp->rq_arg.len >> 2;
1207260c1d12STom Tucker
1208260c1d12STom Tucker /* back up head to the start of the buffer and copy */
1209260c1d12STom Tucker skip = rqstp->rq_arg.len - rqstp->rq_arg.head[0].iov_len;
12100f0257eaSTom Tucker memcpy(dr->args, rqstp->rq_arg.head[0].iov_base - skip,
12110f0257eaSTom Tucker dr->argslen << 2);
12120f0257eaSTom Tucker }
1213948f072aSNeilBrown dr->xprt_ctxt = rqstp->rq_xprt_ctxt;
1214eb8d3a2cSNeilBrown rqstp->rq_xprt_ctxt = NULL;
12158954c5c2SChuck Lever trace_svc_defer(rqstp);
12160f0257eaSTom Tucker svc_xprt_get(rqstp->rq_xprt);
12170f0257eaSTom Tucker dr->xprt = rqstp->rq_xprt;
12187827c81fSChuck Lever set_bit(RQ_DROPME, &rqstp->rq_flags);
12190f0257eaSTom Tucker
12200f0257eaSTom Tucker dr->handle.revisit = svc_revisit;
12210f0257eaSTom Tucker return &dr->handle;
12220f0257eaSTom Tucker }
12230f0257eaSTom Tucker
12240f0257eaSTom Tucker /*
12250f0257eaSTom Tucker * recv data from a deferred request into an active one
12260f0257eaSTom Tucker */
svc_deferred_recv(struct svc_rqst * rqstp)12278954c5c2SChuck Lever static noinline int svc_deferred_recv(struct svc_rqst *rqstp)
12280f0257eaSTom Tucker {
12290f0257eaSTom Tucker struct svc_deferred_req *dr = rqstp->rq_deferred;
12300f0257eaSTom Tucker
12318954c5c2SChuck Lever trace_svc_defer_recv(dr);
12328954c5c2SChuck Lever
1233260c1d12STom Tucker /* setup iov_base past transport header */
1234983084b2SChuck Lever rqstp->rq_arg.head[0].iov_base = dr->args;
1235260c1d12STom Tucker /* The iov_len does not include the transport header bytes */
1236983084b2SChuck Lever rqstp->rq_arg.head[0].iov_len = dr->argslen << 2;
12370f0257eaSTom Tucker rqstp->rq_arg.page_len = 0;
1238260c1d12STom Tucker /* The rq_arg.len includes the transport header bytes */
12390f0257eaSTom Tucker rqstp->rq_arg.len = dr->argslen << 2;
12400f0257eaSTom Tucker rqstp->rq_prot = dr->prot;
12410f0257eaSTom Tucker memcpy(&rqstp->rq_addr, &dr->addr, dr->addrlen);
12420f0257eaSTom Tucker rqstp->rq_addrlen = dr->addrlen;
1243260c1d12STom Tucker /* Save off transport header len in case we get deferred again */
12440f0257eaSTom Tucker rqstp->rq_daddr = dr->daddr;
12450f0257eaSTom Tucker rqstp->rq_respages = rqstp->rq_pages;
1246773f91b2SChuck Lever rqstp->rq_xprt_ctxt = dr->xprt_ctxt;
1247948f072aSNeilBrown
1248948f072aSNeilBrown dr->xprt_ctxt = NULL;
124982011c80SChuck Lever svc_xprt_received(rqstp->rq_xprt);
1250983084b2SChuck Lever return dr->argslen << 2;
12510f0257eaSTom Tucker }
12520f0257eaSTom Tucker
12530f0257eaSTom Tucker
svc_deferred_dequeue(struct svc_xprt * xprt)12540f0257eaSTom Tucker static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
12550f0257eaSTom Tucker {
12560f0257eaSTom Tucker struct svc_deferred_req *dr = NULL;
12570f0257eaSTom Tucker
12580f0257eaSTom Tucker if (!test_bit(XPT_DEFERRED, &xprt->xpt_flags))
12590f0257eaSTom Tucker return NULL;
12600f0257eaSTom Tucker spin_lock(&xprt->xpt_lock);
12610f0257eaSTom Tucker if (!list_empty(&xprt->xpt_deferred)) {
12620f0257eaSTom Tucker dr = list_entry(xprt->xpt_deferred.next,
12630f0257eaSTom Tucker struct svc_deferred_req,
12640f0257eaSTom Tucker handle.recent);
12650f0257eaSTom Tucker list_del_init(&dr->handle.recent);
126662bac4afSJ. Bruce Fields } else
126762bac4afSJ. Bruce Fields clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
12680f0257eaSTom Tucker spin_unlock(&xprt->xpt_lock);
12690f0257eaSTom Tucker return dr;
12700f0257eaSTom Tucker }
12717fcb98d5STom Tucker
1272156e6209SChuck Lever /**
1273156e6209SChuck Lever * svc_find_xprt - find an RPC transport instance
1274156e6209SChuck Lever * @serv: pointer to svc_serv to search
1275156e6209SChuck Lever * @xcl_name: C string containing transport's class name
12764cb54ca2SStanislav Kinsbursky * @net: owner net pointer
1277156e6209SChuck Lever * @af: Address family of transport's local address
1278156e6209SChuck Lever * @port: transport's IP port number
1279156e6209SChuck Lever *
12807fcb98d5STom Tucker * Return the transport instance pointer for the endpoint accepting
12817fcb98d5STom Tucker * connections/peer traffic from the specified transport class,
12827fcb98d5STom Tucker * address family and port.
12837fcb98d5STom Tucker *
12847fcb98d5STom Tucker * Specifying 0 for the address family or port is effectively a
12857fcb98d5STom Tucker * wild-card, and will result in matching the first transport in the
12867fcb98d5STom Tucker * service's list that has a matching class name.
12877fcb98d5STom Tucker */
svc_find_xprt(struct svc_serv * serv,const char * xcl_name,struct net * net,const sa_family_t af,const unsigned short port)1288156e6209SChuck Lever struct svc_xprt *svc_find_xprt(struct svc_serv *serv, const char *xcl_name,
12894cb54ca2SStanislav Kinsbursky struct net *net, const sa_family_t af,
12904cb54ca2SStanislav Kinsbursky const unsigned short port)
12917fcb98d5STom Tucker {
12927fcb98d5STom Tucker struct svc_xprt *xprt;
12937fcb98d5STom Tucker struct svc_xprt *found = NULL;
12947fcb98d5STom Tucker
12957fcb98d5STom Tucker /* Sanity check the args */
1296156e6209SChuck Lever if (serv == NULL || xcl_name == NULL)
12977fcb98d5STom Tucker return found;
12987fcb98d5STom Tucker
12997fcb98d5STom Tucker spin_lock_bh(&serv->sv_lock);
13007fcb98d5STom Tucker list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
13014cb54ca2SStanislav Kinsbursky if (xprt->xpt_net != net)
13024cb54ca2SStanislav Kinsbursky continue;
13037fcb98d5STom Tucker if (strcmp(xprt->xpt_class->xcl_name, xcl_name))
13047fcb98d5STom Tucker continue;
13057fcb98d5STom Tucker if (af != AF_UNSPEC && af != xprt->xpt_local.ss_family)
13067fcb98d5STom Tucker continue;
1307156e6209SChuck Lever if (port != 0 && port != svc_xprt_local_port(xprt))
13087fcb98d5STom Tucker continue;
13097fcb98d5STom Tucker found = xprt;
1310a217813fSTom Tucker svc_xprt_get(xprt);
13117fcb98d5STom Tucker break;
13127fcb98d5STom Tucker }
13137fcb98d5STom Tucker spin_unlock_bh(&serv->sv_lock);
13147fcb98d5STom Tucker return found;
13157fcb98d5STom Tucker }
13167fcb98d5STom Tucker EXPORT_SYMBOL_GPL(svc_find_xprt);
13179571af18STom Tucker
svc_one_xprt_name(const struct svc_xprt * xprt,char * pos,int remaining)1318335c54bdSChuck Lever static int svc_one_xprt_name(const struct svc_xprt *xprt,
1319335c54bdSChuck Lever char *pos, int remaining)
1320335c54bdSChuck Lever {
1321335c54bdSChuck Lever int len;
1322335c54bdSChuck Lever
1323335c54bdSChuck Lever len = snprintf(pos, remaining, "%s %u\n",
1324335c54bdSChuck Lever xprt->xpt_class->xcl_name,
1325335c54bdSChuck Lever svc_xprt_local_port(xprt));
1326335c54bdSChuck Lever if (len >= remaining)
1327335c54bdSChuck Lever return -ENAMETOOLONG;
1328335c54bdSChuck Lever return len;
1329335c54bdSChuck Lever }
1330335c54bdSChuck Lever
1331335c54bdSChuck Lever /**
1332335c54bdSChuck Lever * svc_xprt_names - format a buffer with a list of transport names
1333335c54bdSChuck Lever * @serv: pointer to an RPC service
1334335c54bdSChuck Lever * @buf: pointer to a buffer to be filled in
1335335c54bdSChuck Lever * @buflen: length of buffer to be filled in
1336335c54bdSChuck Lever *
1337335c54bdSChuck Lever * Fills in @buf with a string containing a list of transport names,
1338335c54bdSChuck Lever * each name terminated with '\n'.
1339335c54bdSChuck Lever *
1340335c54bdSChuck Lever * Returns positive length of the filled-in string on success; otherwise
1341335c54bdSChuck Lever * a negative errno value is returned if an error occurs.
13429571af18STom Tucker */
svc_xprt_names(struct svc_serv * serv,char * buf,const int buflen)1343335c54bdSChuck Lever int svc_xprt_names(struct svc_serv *serv, char *buf, const int buflen)
13449571af18STom Tucker {
13459571af18STom Tucker struct svc_xprt *xprt;
1346335c54bdSChuck Lever int len, totlen;
1347335c54bdSChuck Lever char *pos;
13489571af18STom Tucker
13499571af18STom Tucker /* Sanity check args */
13509571af18STom Tucker if (!serv)
13519571af18STom Tucker return 0;
13529571af18STom Tucker
13539571af18STom Tucker spin_lock_bh(&serv->sv_lock);
1354335c54bdSChuck Lever
1355335c54bdSChuck Lever pos = buf;
1356335c54bdSChuck Lever totlen = 0;
13579571af18STom Tucker list_for_each_entry(xprt, &serv->sv_permsocks, xpt_list) {
1358335c54bdSChuck Lever len = svc_one_xprt_name(xprt, pos, buflen - totlen);
1359335c54bdSChuck Lever if (len < 0) {
1360335c54bdSChuck Lever *buf = '\0';
1361335c54bdSChuck Lever totlen = len;
1362335c54bdSChuck Lever }
1363335c54bdSChuck Lever if (len <= 0)
13649571af18STom Tucker break;
1365335c54bdSChuck Lever
1366335c54bdSChuck Lever pos += len;
13679571af18STom Tucker totlen += len;
13689571af18STom Tucker }
1369335c54bdSChuck Lever
13709571af18STom Tucker spin_unlock_bh(&serv->sv_lock);
13719571af18STom Tucker return totlen;
13729571af18STom Tucker }
13739571af18STom Tucker EXPORT_SYMBOL_GPL(svc_xprt_names);
137403cf6c9fSGreg Banks
137503cf6c9fSGreg Banks
137603cf6c9fSGreg Banks /*----------------------------------------------------------------------------*/
137703cf6c9fSGreg Banks
svc_pool_stats_start(struct seq_file * m,loff_t * pos)137803cf6c9fSGreg Banks static void *svc_pool_stats_start(struct seq_file *m, loff_t *pos)
137903cf6c9fSGreg Banks {
138003cf6c9fSGreg Banks unsigned int pidx = (unsigned int)*pos;
138103cf6c9fSGreg Banks struct svc_serv *serv = m->private;
138203cf6c9fSGreg Banks
138303cf6c9fSGreg Banks dprintk("svc_pool_stats_start, *pidx=%u\n", pidx);
138403cf6c9fSGreg Banks
138503cf6c9fSGreg Banks if (!pidx)
138603cf6c9fSGreg Banks return SEQ_START_TOKEN;
138703cf6c9fSGreg Banks return (pidx > serv->sv_nrpools ? NULL : &serv->sv_pools[pidx-1]);
138803cf6c9fSGreg Banks }
138903cf6c9fSGreg Banks
svc_pool_stats_next(struct seq_file * m,void * p,loff_t * pos)139003cf6c9fSGreg Banks static void *svc_pool_stats_next(struct seq_file *m, void *p, loff_t *pos)
139103cf6c9fSGreg Banks {
139203cf6c9fSGreg Banks struct svc_pool *pool = p;
139303cf6c9fSGreg Banks struct svc_serv *serv = m->private;
139403cf6c9fSGreg Banks
139503cf6c9fSGreg Banks dprintk("svc_pool_stats_next, *pos=%llu\n", *pos);
139603cf6c9fSGreg Banks
139703cf6c9fSGreg Banks if (p == SEQ_START_TOKEN) {
139803cf6c9fSGreg Banks pool = &serv->sv_pools[0];
139903cf6c9fSGreg Banks } else {
140003cf6c9fSGreg Banks unsigned int pidx = (pool - &serv->sv_pools[0]);
140103cf6c9fSGreg Banks if (pidx < serv->sv_nrpools-1)
140203cf6c9fSGreg Banks pool = &serv->sv_pools[pidx+1];
140303cf6c9fSGreg Banks else
140403cf6c9fSGreg Banks pool = NULL;
140503cf6c9fSGreg Banks }
140603cf6c9fSGreg Banks ++*pos;
140703cf6c9fSGreg Banks return pool;
140803cf6c9fSGreg Banks }
140903cf6c9fSGreg Banks
svc_pool_stats_stop(struct seq_file * m,void * p)141003cf6c9fSGreg Banks static void svc_pool_stats_stop(struct seq_file *m, void *p)
141103cf6c9fSGreg Banks {
141203cf6c9fSGreg Banks }
141303cf6c9fSGreg Banks
svc_pool_stats_show(struct seq_file * m,void * p)141403cf6c9fSGreg Banks static int svc_pool_stats_show(struct seq_file *m, void *p)
141503cf6c9fSGreg Banks {
141603cf6c9fSGreg Banks struct svc_pool *pool = p;
141703cf6c9fSGreg Banks
141803cf6c9fSGreg Banks if (p == SEQ_START_TOKEN) {
141978c210efSJ. Bruce Fields seq_puts(m, "# pool packets-arrived sockets-enqueued threads-woken threads-timedout\n");
142003cf6c9fSGreg Banks return 0;
142103cf6c9fSGreg Banks }
142203cf6c9fSGreg Banks
1423c743b425SNeilBrown seq_printf(m, "%u %llu %llu %llu 0\n",
142403cf6c9fSGreg Banks pool->sp_id,
1425f208e950SChuck Lever percpu_counter_sum_positive(&pool->sp_messages_arrived),
1426ccf08bedSChuck Lever percpu_counter_sum_positive(&pool->sp_sockets_queued),
1427c743b425SNeilBrown percpu_counter_sum_positive(&pool->sp_threads_woken));
142803cf6c9fSGreg Banks
142903cf6c9fSGreg Banks return 0;
143003cf6c9fSGreg Banks }
143103cf6c9fSGreg Banks
143203cf6c9fSGreg Banks static const struct seq_operations svc_pool_stats_seq_ops = {
143303cf6c9fSGreg Banks .start = svc_pool_stats_start,
143403cf6c9fSGreg Banks .next = svc_pool_stats_next,
143503cf6c9fSGreg Banks .stop = svc_pool_stats_stop,
143603cf6c9fSGreg Banks .show = svc_pool_stats_show,
143703cf6c9fSGreg Banks };
143803cf6c9fSGreg Banks
svc_pool_stats_open(struct svc_serv * serv,struct file * file)143903cf6c9fSGreg Banks int svc_pool_stats_open(struct svc_serv *serv, struct file *file)
144003cf6c9fSGreg Banks {
144103cf6c9fSGreg Banks int err;
144203cf6c9fSGreg Banks
144303cf6c9fSGreg Banks err = seq_open(file, &svc_pool_stats_seq_ops);
144403cf6c9fSGreg Banks if (!err)
144503cf6c9fSGreg Banks ((struct seq_file *) file->private_data)->private = serv;
144603cf6c9fSGreg Banks return err;
144703cf6c9fSGreg Banks }
144803cf6c9fSGreg Banks EXPORT_SYMBOL(svc_pool_stats_open);
144903cf6c9fSGreg Banks
145003cf6c9fSGreg Banks /*----------------------------------------------------------------------------*/
1451