1 /* SPDX-License-Identifier: GPL-2.0 */
3 * linux/include/linux/sunrpc/svc.h
5 * RPC server declarations.
7 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
15 #include <linux/in6.h>
16 #include <linux/sunrpc/types.h>
17 #include <linux/sunrpc/xdr.h>
18 #include <linux/sunrpc/auth.h>
19 #include <linux/sunrpc/svcauth.h>
20 #include <linux/wait.h>
22 #include <linux/pagevec.h>
24 /* statistics for svc_pool structures */
25 struct svc_pool_stats {
26 atomic_long_t packets;
27 unsigned long sockets_queued;
28 atomic_long_t threads_woken;
29 atomic_long_t threads_timedout;
34 * RPC service thread pool.
36 * Pool of threads and temporary sockets. Generally there is only
37 * a single one of these per RPC service, but on NUMA machines those
38 * services that can benefit from it (i.e. nfs but not lockd) will
39 * have one pool per NUMA node. This optimisation reduces cross-
40 * node traffic on multi-node NUMA NFS servers.
43 unsigned int sp_id; /* pool id; also node id on NUMA */
44 spinlock_t sp_lock; /* protects all fields */
45 struct list_head sp_sockets; /* pending sockets */
46 unsigned int sp_nrthreads; /* # of threads in pool */
47 struct list_head sp_all_threads; /* all server threads */
48 struct svc_pool_stats sp_stats; /* statistics on pool operation */
49 #define SP_TASK_PENDING (0) /* still work to do even if no
51 #define SP_CONGESTED (1)
52 unsigned long sp_flags;
53 } ____cacheline_aligned_in_smp;
58 /* Callback to use when last thread exits. */
59 void (*svo_shutdown)(struct svc_serv *, struct net *);
61 /* function for service threads to run */
62 int (*svo_function)(void *);
64 /* queue up a transport for servicing */
65 void (*svo_enqueue_xprt)(struct svc_xprt *);
67 /* optional module to count when adding threads (pooled svcs only) */
68 struct module *svo_module;
74 * An RPC service is a ``daemon,'' possibly multithreaded, which
75 * receives and processes incoming RPC messages.
76 * It has one or more transport sockets associated with it, and maintains
77 * a list of idle threads waiting for input.
79 * We currently do not support more than one RPC program per daemon.
82 struct svc_program * sv_program; /* RPC program */
83 struct svc_stat * sv_stats; /* RPC statistics */
85 struct kref sv_refcnt;
86 unsigned int sv_nrthreads; /* # of server threads */
87 unsigned int sv_maxconn; /* max connections allowed or
88 * '0' causing max to be based
89 * on number of threads. */
91 unsigned int sv_max_payload; /* datagram payload size */
92 unsigned int sv_max_mesg; /* max_payload + 1 page for overheads */
93 unsigned int sv_xdrsize; /* XDR buffer size */
94 struct list_head sv_permsocks; /* all permanent sockets */
95 struct list_head sv_tempsocks; /* all temporary sockets */
96 int sv_tmpcnt; /* count of temporary sockets */
97 struct timer_list sv_temptimer; /* timer for aging temporary sockets */
99 char * sv_name; /* service name */
101 unsigned int sv_nrpools; /* number of thread pools */
102 struct svc_pool * sv_pools; /* array of thread pools */
103 const struct svc_serv_ops *sv_ops; /* server operations */
104 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
105 struct list_head sv_cb_list; /* queue for callback requests
106 * that arrive over the same
108 spinlock_t sv_cb_lock; /* protects the svc_cb_list */
109 wait_queue_head_t sv_cb_waitq; /* sleep here if there are no
110 * entries in the svc_cb_list */
111 bool sv_bc_enabled; /* service uses backchannel */
112 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
116 * svc_get() - increment reference count on a SUNRPC serv
117 * @serv: the svc_serv to have count incremented
119 * Returns: the svc_serv that was passed in.
121 static inline struct svc_serv *svc_get(struct svc_serv *serv)
123 kref_get(&serv->sv_refcnt);
127 void svc_destroy(struct kref *);
130 * svc_put - decrement reference count on a SUNRPC serv
131 * @serv: the svc_serv to have count decremented
133 * When the reference count reaches zero, svc_destroy()
134 * is called to clean up and free the serv.
136 static inline void svc_put(struct svc_serv *serv)
138 kref_put(&serv->sv_refcnt, svc_destroy);
142 * svc_put_not_last - decrement non-final reference count on SUNRPC serv
143 * @serv: the svc_serv to have count decremented
145 * Returns: %true is refcount was decremented.
147 * If the refcount is 1, it is not decremented and instead failure is reported.
149 static inline bool svc_put_not_last(struct svc_serv *serv)
151 return refcount_dec_not_one(&serv->sv_refcnt.refcount);
155 * Maximum payload size supported by a kernel RPC server.
156 * This is use to determine the max number of pages nfsd is
157 * willing to return in a single READ operation.
159 * These happen to all be powers of 2, which is not strictly
160 * necessary but helps enforce the real limitation, which is
161 * that they should be multiples of PAGE_SIZE.
163 * For UDP transports, a block plus NFS,RPC, and UDP headers
164 * has to fit into the IP datagram limit of 64K. The largest
165 * feasible number for all known page sizes is probably 48K,
166 * but we choose 32K here. This is the same as the historical
167 * Linux limit; someone who cares more about NFS/UDP performance
168 * can test a larger number.
170 * For TCP transports we have more freedom. A size of 1MB is
171 * chosen to match the client limit. Other OSes are known to
172 * have larger limits, but those numbers are probably beyond
173 * the point of diminishing returns.
175 #define RPCSVC_MAXPAYLOAD (1*1024*1024u)
176 #define RPCSVC_MAXPAYLOAD_TCP RPCSVC_MAXPAYLOAD
177 #define RPCSVC_MAXPAYLOAD_UDP (32*1024u)
179 extern u32 svc_max_payload(const struct svc_rqst *rqstp);
182 * RPC Requsts and replies are stored in one or more pages.
183 * We maintain an array of pages for each server thread.
184 * Requests are copied into these pages as they arrive. Remaining
185 * pages are available to write the reply into.
187 * Pages are sent using ->sendpage so each server thread needs to
188 * allocate more to replace those used in sending. To help keep track
189 * of these pages we have a receive list where all pages initialy live,
190 * and a send list where pages are moved to when there are to be part
193 * We use xdr_buf for holding responses as it fits well with NFS
194 * read responses (that have a header, and some data pages, and possibly
195 * a tail) and means we can share some client side routines.
197 * The xdr_buf.head kvec always points to the first page in the rq_*pages
198 * list. The xdr_buf.pages pointer points to the second page on that
199 * list. xdr_buf.tail points to the end of the first page.
200 * This assumes that the non-page part of an rpc reply will fit
201 * in a page - NFSd ensures this. lockd also has no trouble.
203 * Each request/reply pair can have at most one "payload", plus two pages,
204 * one for the request, and one for the reply.
205 * We using ->sendfile to return read data, we might need one extra page
206 * if the request is not page-aligned. So add another '1'.
208 #define RPCSVC_MAXPAGES ((RPCSVC_MAXPAYLOAD+PAGE_SIZE-1)/PAGE_SIZE \
211 static inline u32 svc_getnl(struct kvec *iov)
216 iov->iov_base = (void*)vp;
217 iov->iov_len -= sizeof(__be32);
221 static inline void svc_putnl(struct kvec *iov, u32 val)
223 __be32 *vp = iov->iov_base + iov->iov_len;
225 iov->iov_len += sizeof(__be32);
228 static inline __be32 svc_getu32(struct kvec *iov)
233 iov->iov_base = (void*)vp;
234 iov->iov_len -= sizeof(__be32);
238 static inline void svc_ungetu32(struct kvec *iov)
240 __be32 *vp = (__be32 *)iov->iov_base;
241 iov->iov_base = (void *)(vp - 1);
242 iov->iov_len += sizeof(*vp);
245 static inline void svc_putu32(struct kvec *iov, __be32 val)
247 __be32 *vp = iov->iov_base + iov->iov_len;
249 iov->iov_len += sizeof(__be32);
253 * The context of a single thread, including the request currently being
257 struct list_head rq_all; /* all threads list */
258 struct rcu_head rq_rcu_head; /* for RCU deferred kfree */
259 struct svc_xprt * rq_xprt; /* transport ptr */
261 struct sockaddr_storage rq_addr; /* peer address */
263 struct sockaddr_storage rq_daddr; /* dest addr of request
264 * - reply from here */
267 struct svc_serv * rq_server; /* RPC service definition */
268 struct svc_pool * rq_pool; /* thread pool */
269 const struct svc_procedure *rq_procinfo;/* procedure info */
270 struct auth_ops * rq_authop; /* authentication flavour */
271 struct svc_cred rq_cred; /* auth info */
272 void * rq_xprt_ctxt; /* transport specific context ptr */
273 struct svc_deferred_req*rq_deferred; /* deferred request we are replaying */
275 size_t rq_xprt_hlen; /* xprt header len */
276 struct xdr_buf rq_arg;
277 struct xdr_stream rq_arg_stream;
278 struct xdr_stream rq_res_stream;
279 struct page *rq_scratch_page;
280 struct xdr_buf rq_res;
281 struct page *rq_pages[RPCSVC_MAXPAGES + 1];
282 struct page * *rq_respages; /* points into rq_pages */
283 struct page * *rq_next_page; /* next reply page to use */
284 struct page * *rq_page_end; /* one past the last page */
286 struct pagevec rq_pvec;
287 struct kvec rq_vec[RPCSVC_MAXPAGES]; /* generally useful.. */
288 struct bio_vec rq_bvec[RPCSVC_MAXPAGES];
290 __be32 rq_xid; /* transmission id */
291 u32 rq_prog; /* program number */
292 u32 rq_vers; /* program version */
293 u32 rq_proc; /* procedure number */
294 u32 rq_prot; /* IP protocol */
295 int rq_cachetype; /* catering to nfsd */
296 #define RQ_SECURE (0) /* secure port */
297 #define RQ_LOCAL (1) /* local request */
298 #define RQ_USEDEFERRAL (2) /* use deferral */
299 #define RQ_DROPME (3) /* drop current reply */
300 #define RQ_SPLICE_OK (4) /* turned off in gss privacy
301 * to prevent encrypting page
303 #define RQ_VICTIM (5) /* about to be shut down */
304 #define RQ_BUSY (6) /* request is busy */
305 #define RQ_DATA (7) /* request has data */
306 unsigned long rq_flags; /* flags field */
307 ktime_t rq_qtime; /* enqueue time */
309 void * rq_argp; /* decoded arguments */
310 void * rq_resp; /* xdr'd results */
311 void * rq_auth_data; /* flavor-specific data */
312 __be32 rq_auth_stat; /* authentication status */
313 int rq_auth_slack; /* extra space xdr code
314 * should leave in head
317 int rq_reserved; /* space on socket outq
318 * reserved for this request
320 ktime_t rq_stime; /* start time */
322 struct cache_req rq_chandle; /* handle passed to caches for
325 /* Catering to nfsd */
326 struct auth_domain * rq_client; /* RPC peer info */
327 struct auth_domain * rq_gssclient; /* "gss/"-style peer info */
328 struct svc_cacherep * rq_cacherep; /* cache info */
329 struct task_struct *rq_task; /* service thread */
330 spinlock_t rq_lock; /* per-request lock */
331 struct net *rq_bc_net; /* pointer to backchannel's
334 void ** rq_lease_breaker; /* The v4 client breaking a lease */
337 #define SVC_NET(rqst) (rqst->rq_xprt ? rqst->rq_xprt->xpt_net : rqst->rq_bc_net)
340 * Rigorous type checking on sockaddr type conversions
342 static inline struct sockaddr_in *svc_addr_in(const struct svc_rqst *rqst)
344 return (struct sockaddr_in *) &rqst->rq_addr;
347 static inline struct sockaddr_in6 *svc_addr_in6(const struct svc_rqst *rqst)
349 return (struct sockaddr_in6 *) &rqst->rq_addr;
352 static inline struct sockaddr *svc_addr(const struct svc_rqst *rqst)
354 return (struct sockaddr *) &rqst->rq_addr;
357 static inline struct sockaddr_in *svc_daddr_in(const struct svc_rqst *rqst)
359 return (struct sockaddr_in *) &rqst->rq_daddr;
362 static inline struct sockaddr_in6 *svc_daddr_in6(const struct svc_rqst *rqst)
364 return (struct sockaddr_in6 *) &rqst->rq_daddr;
367 static inline struct sockaddr *svc_daddr(const struct svc_rqst *rqst)
369 return (struct sockaddr *) &rqst->rq_daddr;
373 * Check buffer bounds after decoding arguments
376 xdr_argsize_check(struct svc_rqst *rqstp, __be32 *p)
378 char *cp = (char *)p;
379 struct kvec *vec = &rqstp->rq_arg.head[0];
380 return cp >= (char*)vec->iov_base
381 && cp <= (char*)vec->iov_base + vec->iov_len;
385 xdr_ressize_check(struct svc_rqst *rqstp, __be32 *p)
387 struct kvec *vec = &rqstp->rq_res.head[0];
390 vec->iov_len = cp - (char*)vec->iov_base;
392 return vec->iov_len <= PAGE_SIZE;
395 static inline void svc_free_res_pages(struct svc_rqst *rqstp)
397 while (rqstp->rq_next_page != rqstp->rq_respages) {
398 struct page **pp = --rqstp->rq_next_page;
406 struct svc_deferred_req {
407 u32 prot; /* protocol (UDP or TCP) */
408 struct svc_xprt *xprt;
409 struct sockaddr_storage addr; /* where reply must go */
411 struct sockaddr_storage daddr; /* where reply must come from */
413 struct cache_deferred_req handle;
419 struct svc_process_info {
421 int (*dispatch)(struct svc_rqst *, __be32 *);
430 * List of RPC programs on the same transport endpoint
433 struct svc_program * pg_next; /* other programs (same xprt) */
434 u32 pg_prog; /* program number */
435 unsigned int pg_lovers; /* lowest version */
436 unsigned int pg_hivers; /* highest version */
437 unsigned int pg_nvers; /* number of versions */
438 const struct svc_version **pg_vers; /* version array */
439 char * pg_name; /* service name */
440 char * pg_class; /* class name: services sharing authentication */
441 struct svc_stat * pg_stats; /* rpc statistics */
442 int (*pg_authenticate)(struct svc_rqst *);
443 __be32 (*pg_init_request)(struct svc_rqst *,
444 const struct svc_program *,
445 struct svc_process_info *);
446 int (*pg_rpcbind_set)(struct net *net,
447 const struct svc_program *,
448 u32 version, int family,
449 unsigned short proto,
450 unsigned short port);
454 * RPC program version
457 u32 vs_vers; /* version number */
458 u32 vs_nproc; /* number of procedures */
459 const struct svc_procedure *vs_proc; /* per-procedure info */
460 unsigned int *vs_count; /* call counts */
461 u32 vs_xdrsize; /* xdrsize needed for this version */
463 /* Don't register with rpcbind */
466 /* Don't care if the rpcbind registration fails */
469 /* Need xprt with congestion control */
470 bool vs_need_cong_ctrl;
472 /* Dispatch function */
473 int (*vs_dispatch)(struct svc_rqst *, __be32 *);
479 struct svc_procedure {
480 /* process the request: */
481 __be32 (*pc_func)(struct svc_rqst *);
482 /* XDR decode args: */
483 bool (*pc_decode)(struct svc_rqst *rqstp,
484 struct xdr_stream *xdr);
485 /* XDR encode result: */
486 bool (*pc_encode)(struct svc_rqst *rqstp,
487 struct xdr_stream *xdr);
488 /* XDR free result: */
489 void (*pc_release)(struct svc_rqst *);
490 unsigned int pc_argsize; /* argument struct size */
491 unsigned int pc_ressize; /* result struct size */
492 unsigned int pc_cachetype; /* cache info (NFS) */
493 unsigned int pc_xdrressize; /* maximum size of XDR reply */
494 const char * pc_name; /* for display */
498 * Function prototypes.
500 int svc_rpcb_setup(struct svc_serv *serv, struct net *net);
501 void svc_rpcb_cleanup(struct svc_serv *serv, struct net *net);
502 int svc_bind(struct svc_serv *serv, struct net *net);
503 struct svc_serv *svc_create(struct svc_program *, unsigned int,
504 const struct svc_serv_ops *);
505 struct svc_rqst *svc_rqst_alloc(struct svc_serv *serv,
506 struct svc_pool *pool, int node);
507 struct svc_rqst *svc_prepare_thread(struct svc_serv *serv,
508 struct svc_pool *pool, int node);
509 void svc_rqst_replace_page(struct svc_rqst *rqstp,
511 void svc_rqst_free(struct svc_rqst *);
512 void svc_exit_thread(struct svc_rqst *);
513 struct svc_serv * svc_create_pooled(struct svc_program *, unsigned int,
514 const struct svc_serv_ops *);
515 int svc_set_num_threads(struct svc_serv *, struct svc_pool *, int);
516 int svc_pool_stats_open(struct svc_serv *serv, struct file *file);
517 void svc_shutdown_net(struct svc_serv *, struct net *);
518 int svc_process(struct svc_rqst *);
519 int bc_svc_process(struct svc_serv *, struct rpc_rqst *,
521 int svc_register(const struct svc_serv *, struct net *, const int,
522 const unsigned short, const unsigned short);
524 void svc_wake_up(struct svc_serv *);
525 void svc_reserve(struct svc_rqst *rqstp, int space);
526 struct svc_pool * svc_pool_for_cpu(struct svc_serv *serv, int cpu);
527 char * svc_print_addr(struct svc_rqst *, char *, size_t);
528 const char * svc_proc_name(const struct svc_rqst *rqstp);
529 int svc_encode_result_payload(struct svc_rqst *rqstp,
531 unsigned int length);
532 unsigned int svc_fill_write_vector(struct svc_rqst *rqstp,
533 struct xdr_buf *payload);
534 char *svc_fill_symlink_pathname(struct svc_rqst *rqstp,
535 struct kvec *first, void *p,
537 __be32 svc_generic_init_request(struct svc_rqst *rqstp,
538 const struct svc_program *progp,
539 struct svc_process_info *procinfo);
540 int svc_generic_rpcbind_set(struct net *net,
541 const struct svc_program *progp,
542 u32 version, int family,
543 unsigned short proto,
544 unsigned short port);
545 int svc_rpcbind_set_version(struct net *net,
546 const struct svc_program *progp,
547 u32 version, int family,
548 unsigned short proto,
549 unsigned short port);
551 #define RPC_MAX_ADDRBUFLEN (63U)
554 * When we want to reduce the size of the reserved space in the response
555 * buffer, we need to take into account the size of any checksum data that
556 * may be at the end of the packet. This is difficult to determine exactly
557 * for all cases without actually generating the checksum, so we just use a
560 static inline void svc_reserve_auth(struct svc_rqst *rqstp, int space)
562 svc_reserve(rqstp, space + rqstp->rq_auth_slack);
566 * svcxdr_init_decode - Prepare an xdr_stream for svc Call decoding
567 * @rqstp: controlling server RPC transaction context
570 static inline void svcxdr_init_decode(struct svc_rqst *rqstp)
572 struct xdr_stream *xdr = &rqstp->rq_arg_stream;
573 struct kvec *argv = rqstp->rq_arg.head;
575 xdr_init_decode(xdr, &rqstp->rq_arg, argv->iov_base, NULL);
576 xdr_set_scratch_page(xdr, rqstp->rq_scratch_page);
580 * svcxdr_init_encode - Prepare an xdr_stream for svc Reply encoding
581 * @rqstp: controlling server RPC transaction context
584 static inline void svcxdr_init_encode(struct svc_rqst *rqstp)
586 struct xdr_stream *xdr = &rqstp->rq_res_stream;
587 struct xdr_buf *buf = &rqstp->rq_res;
588 struct kvec *resv = buf->head;
590 xdr_reset_scratch_buffer(xdr);
594 xdr->p = resv->iov_base + resv->iov_len;
595 xdr->end = resv->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
596 buf->len = resv->iov_len;
597 xdr->page_ptr = buf->pages - 1;
598 buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages);
599 buf->buflen -= rqstp->rq_auth_slack;
603 #endif /* SUNRPC_SVC_H */