2 * linux/include/linux/sunrpc/svc.h
4 * RPC server declarations.
6 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
14 #include <linux/sunrpc/types.h>
15 #include <linux/sunrpc/xdr.h>
16 #include <linux/sunrpc/auth.h>
17 #include <linux/sunrpc/svcauth.h>
18 #include <linux/wait.h>
22 * This is the RPC server thread function prototype
24 typedef void (*svc_thread_fn
)(struct svc_rqst
*);
28 * RPC service thread pool.
30 * Pool of threads and temporary sockets. Generally there is only
31 * a single one of these per RPC service, but on NUMA machines those
32 * services that can benefit from it (i.e. nfs but not lockd) will
33 * have one pool per NUMA node. This optimisation reduces cross-
34 * node traffic on multi-node NUMA NFS servers.
37 unsigned int sp_id
; /* pool id; also node id on NUMA */
38 spinlock_t sp_lock
; /* protects all fields */
39 struct list_head sp_threads
; /* idle server threads */
40 struct list_head sp_sockets
; /* pending sockets */
41 unsigned int sp_nrthreads
; /* # of threads in pool */
42 struct list_head sp_all_threads
; /* all server threads */
43 } ____cacheline_aligned_in_smp
;
48 * An RPC service is a ``daemon,'' possibly multithreaded, which
49 * receives and processes incoming RPC messages.
50 * It has one or more transport sockets associated with it, and maintains
51 * a list of idle threads waiting for input.
53 * We currently do not support more than one RPC program per daemon.
56 struct svc_program
* sv_program
; /* RPC program */
57 struct svc_stat
* sv_stats
; /* RPC statistics */
59 unsigned int sv_nrthreads
; /* # of server threads */
60 unsigned int sv_max_payload
; /* datagram payload size */
61 unsigned int sv_max_mesg
; /* max_payload + 1 page for overheads */
62 unsigned int sv_xdrsize
; /* XDR buffer size */
64 struct list_head sv_permsocks
; /* all permanent sockets */
65 struct list_head sv_tempsocks
; /* all temporary sockets */
66 int sv_tmpcnt
; /* count of temporary sockets */
67 struct timer_list sv_temptimer
; /* timer for aging temporary sockets */
69 char * sv_name
; /* service name */
71 unsigned int sv_nrpools
; /* number of thread pools */
72 struct svc_pool
* sv_pools
; /* array of thread pools */
74 void (*sv_shutdown
)(struct svc_serv
*serv
);
75 /* Callback to use when last thread
79 struct module
* sv_module
; /* optional module to count when
81 svc_thread_fn sv_function
; /* main function for threads */
82 int sv_kill_signal
; /* signal to kill threads */
86 * We use sv_nrthreads as a reference count. svc_destroy() drops
87 * this refcount, so we need to bump it up around operations that
88 * change the number of threads. Horrible, but there it is.
89 * Should be called with the BKL held.
91 static inline void svc_get(struct svc_serv
*serv
)
97 * Maximum payload size supported by a kernel RPC server.
98 * This is use to determine the max number of pages nfsd is
99 * willing to return in a single READ operation.
101 * These happen to all be powers of 2, which is not strictly
102 * necessary but helps enforce the real limitation, which is
103 * that they should be multiples of PAGE_CACHE_SIZE.
105 * For UDP transports, a block plus NFS,RPC, and UDP headers
106 * has to fit into the IP datagram limit of 64K. The largest
107 * feasible number for all known page sizes is probably 48K,
108 * but we choose 32K here. This is the same as the historical
109 * Linux limit; someone who cares more about NFS/UDP performance
110 * can test a larger number.
112 * For TCP transports we have more freedom. A size of 1MB is
113 * chosen to match the client limit. Other OSes are known to
114 * have larger limits, but those numbers are probably beyond
115 * the point of diminishing returns.
117 #define RPCSVC_MAXPAYLOAD (1*1024*1024u)
118 #define RPCSVC_MAXPAYLOAD_TCP RPCSVC_MAXPAYLOAD
119 #define RPCSVC_MAXPAYLOAD_UDP (32*1024u)
121 extern u32
svc_max_payload(const struct svc_rqst
*rqstp
);
124 * RPC Requsts and replies are stored in one or more pages.
125 * We maintain an array of pages for each server thread.
126 * Requests are copied into these pages as they arrive. Remaining
127 * pages are available to write the reply into.
129 * Pages are sent using ->sendpage so each server thread needs to
130 * allocate more to replace those used in sending. To help keep track
131 * of these pages we have a receive list where all pages initialy live,
132 * and a send list where pages are moved to when there are to be part
135 * We use xdr_buf for holding responses as it fits well with NFS
136 * read responses (that have a header, and some data pages, and possibly
137 * a tail) and means we can share some client side routines.
139 * The xdr_buf.head kvec always points to the first page in the rq_*pages
140 * list. The xdr_buf.pages pointer points to the second page on that
141 * list. xdr_buf.tail points to the end of the first page.
142 * This assumes that the non-page part of an rpc reply will fit
143 * in a page - NFSd ensures this. lockd also has no trouble.
145 * Each request/reply pair can have at most one "payload", plus two pages,
146 * one for the request, and one for the reply.
147 * We using ->sendfile to return read data, we might need one extra page
148 * if the request is not page-aligned. So add another '1'.
150 #define RPCSVC_MAXPAGES ((RPCSVC_MAXPAYLOAD+PAGE_SIZE-1)/PAGE_SIZE \
153 static inline u32
svc_getnl(struct kvec
*iov
)
158 iov
->iov_base
= (void*)vp
;
159 iov
->iov_len
-= sizeof(__be32
);
163 static inline void svc_putnl(struct kvec
*iov
, u32 val
)
165 __be32
*vp
= iov
->iov_base
+ iov
->iov_len
;
167 iov
->iov_len
+= sizeof(__be32
);
170 static inline __be32
svc_getu32(struct kvec
*iov
)
175 iov
->iov_base
= (void*)vp
;
176 iov
->iov_len
-= sizeof(__be32
);
180 static inline void svc_ungetu32(struct kvec
*iov
)
182 __be32
*vp
= (__be32
*)iov
->iov_base
;
183 iov
->iov_base
= (void *)(vp
- 1);
184 iov
->iov_len
+= sizeof(*vp
);
187 static inline void svc_putu32(struct kvec
*iov
, __be32 val
)
189 __be32
*vp
= iov
->iov_base
+ iov
->iov_len
;
191 iov
->iov_len
+= sizeof(__be32
);
196 * The context of a single thread, including the request currently being
200 struct list_head rq_list
; /* idle list */
201 struct list_head rq_all
; /* all threads list */
202 struct svc_sock
* rq_sock
; /* socket */
203 struct sockaddr_in rq_addr
; /* peer address */
206 struct svc_serv
* rq_server
; /* RPC service definition */
207 struct svc_pool
* rq_pool
; /* thread pool */
208 struct svc_procedure
* rq_procinfo
; /* procedure info */
209 struct auth_ops
* rq_authop
; /* authentication flavour */
210 struct svc_cred rq_cred
; /* auth info */
211 struct sk_buff
* rq_skbuff
; /* fast recv inet buffer */
212 struct svc_deferred_req
*rq_deferred
; /* deferred request we are replaying */
214 struct xdr_buf rq_arg
;
215 struct xdr_buf rq_res
;
216 struct page
* rq_pages
[RPCSVC_MAXPAGES
];
217 struct page
* *rq_respages
; /* points into rq_pages */
218 int rq_resused
; /* number of pages used for result */
220 struct kvec rq_vec
[RPCSVC_MAXPAGES
]; /* generally useful.. */
222 __be32 rq_xid
; /* transmission id */
223 u32 rq_prog
; /* program number */
224 u32 rq_vers
; /* program version */
225 u32 rq_proc
; /* procedure number */
226 u32 rq_prot
; /* IP protocol */
228 rq_secure
: 1; /* secure port */
231 __be32 rq_daddr
; /* dest addr of request - reply from here */
233 void * rq_argp
; /* decoded arguments */
234 void * rq_resp
; /* xdr'd results */
235 void * rq_auth_data
; /* flavor-specific data */
237 int rq_reserved
; /* space on socket outq
238 * reserved for this request
241 struct cache_req rq_chandle
; /* handle passed to caches for
244 /* Catering to nfsd */
245 struct auth_domain
* rq_client
; /* RPC peer info */
246 struct svc_cacherep
* rq_cacherep
; /* cache info */
247 struct knfsd_fh
* rq_reffh
; /* Referrence filehandle, used to
248 * determine what device number
249 * to report (real or virtual)
251 int rq_sendfile_ok
; /* turned off in gss privacy
252 * to prevent encrypting page
254 wait_queue_head_t rq_wait
; /* synchronization */
255 struct task_struct
*rq_task
; /* service thread */
259 * Check buffer bounds after decoding arguments
262 xdr_argsize_check(struct svc_rqst
*rqstp
, __be32
*p
)
264 char *cp
= (char *)p
;
265 struct kvec
*vec
= &rqstp
->rq_arg
.head
[0];
266 return cp
>= (char*)vec
->iov_base
267 && cp
<= (char*)vec
->iov_base
+ vec
->iov_len
;
271 xdr_ressize_check(struct svc_rqst
*rqstp
, __be32
*p
)
273 struct kvec
*vec
= &rqstp
->rq_res
.head
[0];
276 vec
->iov_len
= cp
- (char*)vec
->iov_base
;
278 return vec
->iov_len
<= PAGE_SIZE
;
281 static inline void svc_free_res_pages(struct svc_rqst
*rqstp
)
283 while (rqstp
->rq_resused
) {
284 struct page
**pp
= (rqstp
->rq_respages
+
285 --rqstp
->rq_resused
);
293 struct svc_deferred_req
{
294 u32 prot
; /* protocol (UDP or TCP) */
295 struct sockaddr_in addr
;
296 struct svc_sock
*svsk
; /* where reply must go */
297 __be32 daddr
; /* where reply must come from */
298 struct cache_deferred_req handle
;
304 * List of RPC programs on the same transport endpoint
307 struct svc_program
* pg_next
; /* other programs (same xprt) */
308 u32 pg_prog
; /* program number */
309 unsigned int pg_lovers
; /* lowest version */
310 unsigned int pg_hivers
; /* lowest version */
311 unsigned int pg_nvers
; /* number of versions */
312 struct svc_version
** pg_vers
; /* version array */
313 char * pg_name
; /* service name */
314 char * pg_class
; /* class name: services sharing authentication */
315 struct svc_stat
* pg_stats
; /* rpc statistics */
316 int (*pg_authenticate
)(struct svc_rqst
*);
320 * RPC program version
323 u32 vs_vers
; /* version number */
324 u32 vs_nproc
; /* number of procedures */
325 struct svc_procedure
* vs_proc
; /* per-procedure info */
326 u32 vs_xdrsize
; /* xdrsize needed for this version */
328 unsigned int vs_hidden
: 1; /* Don't register with portmapper.
329 * Only used for nfsacl so far. */
331 /* Override dispatch function (e.g. when caching replies).
332 * A return value of 0 means drop the request.
333 * vs_dispatch == NULL means use default dispatcher.
335 int (*vs_dispatch
)(struct svc_rqst
*, __be32
*);
341 typedef __be32 (*svc_procfunc
)(struct svc_rqst
*, void *argp
, void *resp
);
342 struct svc_procedure
{
343 svc_procfunc pc_func
; /* process the request */
344 kxdrproc_t pc_decode
; /* XDR decode args */
345 kxdrproc_t pc_encode
; /* XDR encode result */
346 kxdrproc_t pc_release
; /* XDR free result */
347 unsigned int pc_argsize
; /* argument struct size */
348 unsigned int pc_ressize
; /* result struct size */
349 unsigned int pc_count
; /* call count */
350 unsigned int pc_cachetype
; /* cache info (NFS) */
351 unsigned int pc_xdrressize
; /* maximum size of XDR reply */
355 * Function prototypes.
357 struct svc_serv
* svc_create(struct svc_program
*, unsigned int,
358 void (*shutdown
)(struct svc_serv
*));
359 int svc_create_thread(svc_thread_fn
, struct svc_serv
*);
360 void svc_exit_thread(struct svc_rqst
*);
361 struct svc_serv
* svc_create_pooled(struct svc_program
*, unsigned int,
362 void (*shutdown
)(struct svc_serv
*),
363 svc_thread_fn
, int sig
, struct module
*);
364 int svc_set_num_threads(struct svc_serv
*, struct svc_pool
*, int);
365 void svc_destroy(struct svc_serv
*);
366 int svc_process(struct svc_rqst
*);
367 int svc_register(struct svc_serv
*, int, unsigned short);
368 void svc_wake_up(struct svc_serv
*);
369 void svc_reserve(struct svc_rqst
*rqstp
, int space
);
370 struct svc_pool
* svc_pool_for_cpu(struct svc_serv
*serv
, int cpu
);
372 #endif /* SUNRPC_SVC_H */