Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/include/linux/sunrpc/svc.h | |
3 | * | |
4 | * RPC server declarations. | |
5 | * | |
6 | * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de> | |
7 | */ | |
8 | ||
9 | ||
10 | #ifndef SUNRPC_SVC_H | |
11 | #define SUNRPC_SVC_H | |
12 | ||
13 | #include <linux/in.h> | |
73df0dba | 14 | #include <linux/in6.h> |
1da177e4 LT |
15 | #include <linux/sunrpc/types.h> |
16 | #include <linux/sunrpc/xdr.h> | |
7adae489 | 17 | #include <linux/sunrpc/auth.h> |
1da177e4 LT |
18 | #include <linux/sunrpc/svcauth.h> |
19 | #include <linux/wait.h> | |
20 | #include <linux/mm.h> | |
21 | ||
a7455442 GB |
22 | /* |
23 | * This is the RPC server thread function prototype | |
24 | */ | |
9867d76c | 25 | typedef int (*svc_thread_fn)(void *); |
3262c816 GB |
26 | |
27 | /* | |
28 | * | |
29 | * RPC service thread pool. | |
30 | * | |
31 | * Pool of threads and temporary sockets. Generally there is only | |
32 | * a single one of these per RPC service, but on NUMA machines those | |
33 | * services that can benefit from it (i.e. nfs but not lockd) will | |
34 | * have one pool per NUMA node. This optimisation reduces cross- | |
35 | * node traffic on multi-node NUMA NFS servers. | |
36 | */ | |
37 | struct svc_pool { | |
38 | unsigned int sp_id; /* pool id; also node id on NUMA */ | |
39 | spinlock_t sp_lock; /* protects all fields */ | |
40 | struct list_head sp_threads; /* idle server threads */ | |
41 | struct list_head sp_sockets; /* pending sockets */ | |
42 | unsigned int sp_nrthreads; /* # of threads in pool */ | |
a7455442 | 43 | struct list_head sp_all_threads; /* all server threads */ |
3262c816 GB |
44 | } ____cacheline_aligned_in_smp; |
45 | ||
1da177e4 LT |
46 | /* |
47 | * RPC service. | |
48 | * | |
49 | * An RPC service is a ``daemon,'' possibly multithreaded, which | |
50 | * receives and processes incoming RPC messages. | |
51 | * It has one or more transport sockets associated with it, and maintains | |
52 | * a list of idle threads waiting for input. | |
53 | * | |
54 | * We currently do not support more than one RPC program per daemon. | |
55 | */ | |
56 | struct svc_serv { | |
1da177e4 LT |
57 | struct svc_program * sv_program; /* RPC program */ |
58 | struct svc_stat * sv_stats; /* RPC statistics */ | |
59 | spinlock_t sv_lock; | |
60 | unsigned int sv_nrthreads; /* # of server threads */ | |
c6b0a9f8 N |
61 | unsigned int sv_max_payload; /* datagram payload size */ |
62 | unsigned int sv_max_mesg; /* max_payload + 1 page for overheads */ | |
1da177e4 LT |
63 | unsigned int sv_xdrsize; /* XDR buffer size */ |
64 | ||
65 | struct list_head sv_permsocks; /* all permanent sockets */ | |
66 | struct list_head sv_tempsocks; /* all temporary sockets */ | |
67 | int sv_tmpcnt; /* count of temporary sockets */ | |
36bdfc8b | 68 | struct timer_list sv_temptimer; /* timer for aging temporary sockets */ |
e851db5b | 69 | sa_family_t sv_family; /* listener's address family */ |
1da177e4 LT |
70 | |
71 | char * sv_name; /* service name */ | |
bc591ccf | 72 | |
3262c816 GB |
73 | unsigned int sv_nrpools; /* number of thread pools */ |
74 | struct svc_pool * sv_pools; /* array of thread pools */ | |
75 | ||
bc591ccf N |
76 | void (*sv_shutdown)(struct svc_serv *serv); |
77 | /* Callback to use when last thread | |
78 | * exits. | |
79 | */ | |
a7455442 GB |
80 | |
81 | struct module * sv_module; /* optional module to count when | |
82 | * adding threads */ | |
83 | svc_thread_fn sv_function; /* main function for threads */ | |
1da177e4 LT |
84 | }; |
85 | ||
9a24ab57 GB |
86 | /* |
87 | * We use sv_nrthreads as a reference count. svc_destroy() drops | |
88 | * this refcount, so we need to bump it up around operations that | |
89 | * change the number of threads. Horrible, but there it is. | |
90 | * Should be called with the BKL held. | |
91 | */ | |
92 | static inline void svc_get(struct svc_serv *serv) | |
93 | { | |
94 | serv->sv_nrthreads++; | |
95 | } | |
96 | ||
1da177e4 LT |
97 | /* |
98 | * Maximum payload size supported by a kernel RPC server. | |
99 | * This is use to determine the max number of pages nfsd is | |
100 | * willing to return in a single READ operation. | |
7adae489 GB |
101 | * |
102 | * These happen to all be powers of 2, which is not strictly | |
103 | * necessary but helps enforce the real limitation, which is | |
104 | * that they should be multiples of PAGE_CACHE_SIZE. | |
105 | * | |
106 | * For UDP transports, a block plus NFS,RPC, and UDP headers | |
107 | * has to fit into the IP datagram limit of 64K. The largest | |
108 | * feasible number for all known page sizes is probably 48K, | |
109 | * but we choose 32K here. This is the same as the historical | |
110 | * Linux limit; someone who cares more about NFS/UDP performance | |
111 | * can test a larger number. | |
112 | * | |
113 | * For TCP transports we have more freedom. A size of 1MB is | |
114 | * chosen to match the client limit. Other OSes are known to | |
115 | * have larger limits, but those numbers are probably beyond | |
116 | * the point of diminishing returns. | |
1da177e4 | 117 | */ |
7adae489 GB |
118 | #define RPCSVC_MAXPAYLOAD (1*1024*1024u) |
119 | #define RPCSVC_MAXPAYLOAD_TCP RPCSVC_MAXPAYLOAD | |
120 | #define RPCSVC_MAXPAYLOAD_UDP (32*1024u) | |
121 | ||
122 | extern u32 svc_max_payload(const struct svc_rqst *rqstp); | |
1da177e4 LT |
123 | |
124 | /* | |
125 | * RPC Requsts and replies are stored in one or more pages. | |
126 | * We maintain an array of pages for each server thread. | |
127 | * Requests are copied into these pages as they arrive. Remaining | |
128 | * pages are available to write the reply into. | |
129 | * | |
130 | * Pages are sent using ->sendpage so each server thread needs to | |
131 | * allocate more to replace those used in sending. To help keep track | |
132 | * of these pages we have a receive list where all pages initialy live, | |
133 | * and a send list where pages are moved to when there are to be part | |
134 | * of a reply. | |
135 | * | |
136 | * We use xdr_buf for holding responses as it fits well with NFS | |
137 | * read responses (that have a header, and some data pages, and possibly | |
138 | * a tail) and means we can share some client side routines. | |
139 | * | |
140 | * The xdr_buf.head kvec always points to the first page in the rq_*pages | |
141 | * list. The xdr_buf.pages pointer points to the second page on that | |
142 | * list. xdr_buf.tail points to the end of the first page. | |
143 | * This assumes that the non-page part of an rpc reply will fit | |
144 | * in a page - NFSd ensures this. lockd also has no trouble. | |
145 | * | |
146 | * Each request/reply pair can have at most one "payload", plus two pages, | |
147 | * one for the request, and one for the reply. | |
250f3915 N |
148 | * We using ->sendfile to return read data, we might need one extra page |
149 | * if the request is not page-aligned. So add another '1'. | |
1da177e4 | 150 | */ |
250f3915 N |
151 | #define RPCSVC_MAXPAGES ((RPCSVC_MAXPAYLOAD+PAGE_SIZE-1)/PAGE_SIZE \ |
152 | + 2 + 1) | |
1da177e4 | 153 | |
76994313 | 154 | static inline u32 svc_getnl(struct kvec *iov) |
1da177e4 | 155 | { |
76994313 | 156 | __be32 val, *vp; |
1da177e4 LT |
157 | vp = iov->iov_base; |
158 | val = *vp++; | |
159 | iov->iov_base = (void*)vp; | |
76994313 AD |
160 | iov->iov_len -= sizeof(__be32); |
161 | return ntohl(val); | |
162 | } | |
163 | ||
164 | static inline void svc_putnl(struct kvec *iov, u32 val) | |
165 | { | |
166 | __be32 *vp = iov->iov_base + iov->iov_len; | |
167 | *vp = htonl(val); | |
168 | iov->iov_len += sizeof(__be32); | |
169 | } | |
170 | ||
171 | static inline __be32 svc_getu32(struct kvec *iov) | |
172 | { | |
173 | __be32 val, *vp; | |
174 | vp = iov->iov_base; | |
175 | val = *vp++; | |
176 | iov->iov_base = (void*)vp; | |
177 | iov->iov_len -= sizeof(__be32); | |
1da177e4 LT |
178 | return val; |
179 | } | |
180 | ||
181 | static inline void svc_ungetu32(struct kvec *iov) | |
182 | { | |
76994313 | 183 | __be32 *vp = (__be32 *)iov->iov_base; |
1da177e4 LT |
184 | iov->iov_base = (void *)(vp - 1); |
185 | iov->iov_len += sizeof(*vp); | |
186 | } | |
187 | ||
76994313 | 188 | static inline void svc_putu32(struct kvec *iov, __be32 val) |
1da177e4 | 189 | { |
76994313 | 190 | __be32 *vp = iov->iov_base + iov->iov_len; |
1da177e4 | 191 | *vp = val; |
76994313 | 192 | iov->iov_len += sizeof(__be32); |
1da177e4 LT |
193 | } |
194 | ||
73df0dba CL |
195 | union svc_addr_u { |
196 | struct in_addr addr; | |
73df0dba | 197 | struct in6_addr addr6; |
73df0dba CL |
198 | }; |
199 | ||
1da177e4 LT |
200 | /* |
201 | * The context of a single thread, including the request currently being | |
202 | * processed. | |
1da177e4 LT |
203 | */ |
204 | struct svc_rqst { | |
205 | struct list_head rq_list; /* idle list */ | |
a7455442 | 206 | struct list_head rq_all; /* all threads list */ |
57b1d3ba | 207 | struct svc_xprt * rq_xprt; /* transport ptr */ |
27459f09 CL |
208 | struct sockaddr_storage rq_addr; /* peer address */ |
209 | size_t rq_addrlen; | |
1da177e4 LT |
210 | |
211 | struct svc_serv * rq_server; /* RPC service definition */ | |
3262c816 | 212 | struct svc_pool * rq_pool; /* thread pool */ |
1da177e4 LT |
213 | struct svc_procedure * rq_procinfo; /* procedure info */ |
214 | struct auth_ops * rq_authop; /* authentication flavour */ | |
c4170583 | 215 | u32 rq_flavor; /* pseudoflavor */ |
1da177e4 | 216 | struct svc_cred rq_cred; /* auth info */ |
5148bf4e | 217 | void * rq_xprt_ctxt; /* transport specific context ptr */ |
1da177e4 LT |
218 | struct svc_deferred_req*rq_deferred; /* deferred request we are replaying */ |
219 | ||
260c1d12 | 220 | size_t rq_xprt_hlen; /* xprt header len */ |
1da177e4 LT |
221 | struct xdr_buf rq_arg; |
222 | struct xdr_buf rq_res; | |
44524359 N |
223 | struct page * rq_pages[RPCSVC_MAXPAGES]; |
224 | struct page * *rq_respages; /* points into rq_pages */ | |
225 | int rq_resused; /* number of pages used for result */ | |
1da177e4 | 226 | |
3cc03b16 N |
227 | struct kvec rq_vec[RPCSVC_MAXPAGES]; /* generally useful.. */ |
228 | ||
d8ed029d | 229 | __be32 rq_xid; /* transmission id */ |
1da177e4 LT |
230 | u32 rq_prog; /* program number */ |
231 | u32 rq_vers; /* program version */ | |
232 | u32 rq_proc; /* procedure number */ | |
233 | u32 rq_prot; /* IP protocol */ | |
234 | unsigned short | |
235 | rq_secure : 1; /* secure port */ | |
236 | ||
73df0dba CL |
237 | union svc_addr_u rq_daddr; /* dest addr of request |
238 | * - reply from here */ | |
1da177e4 LT |
239 | |
240 | void * rq_argp; /* decoded arguments */ | |
241 | void * rq_resp; /* xdr'd results */ | |
242 | void * rq_auth_data; /* flavor-specific data */ | |
243 | ||
244 | int rq_reserved; /* space on socket outq | |
245 | * reserved for this request | |
246 | */ | |
247 | ||
248 | struct cache_req rq_chandle; /* handle passed to caches for | |
249 | * request delaying | |
250 | */ | |
251 | /* Catering to nfsd */ | |
252 | struct auth_domain * rq_client; /* RPC peer info */ | |
3ab4d8b1 | 253 | struct auth_domain * rq_gssclient; /* "gss/"-style peer info */ |
1da177e4 LT |
254 | struct svc_cacherep * rq_cacherep; /* cache info */ |
255 | struct knfsd_fh * rq_reffh; /* Referrence filehandle, used to | |
256 | * determine what device number | |
257 | * to report (real or virtual) | |
258 | */ | |
cf8208d0 | 259 | int rq_splice_ok; /* turned off in gss privacy |
5c04c46a BF |
260 | * to prevent encrypting page |
261 | * cache pages */ | |
1da177e4 | 262 | wait_queue_head_t rq_wait; /* synchronization */ |
a7455442 | 263 | struct task_struct *rq_task; /* service thread */ |
1da177e4 LT |
264 | }; |
265 | ||
27459f09 CL |
266 | /* |
267 | * Rigorous type checking on sockaddr type conversions | |
268 | */ | |
5344b12d | 269 | static inline struct sockaddr_in *svc_addr_in(const struct svc_rqst *rqst) |
27459f09 CL |
270 | { |
271 | return (struct sockaddr_in *) &rqst->rq_addr; | |
272 | } | |
273 | ||
5344b12d | 274 | static inline struct sockaddr_in6 *svc_addr_in6(const struct svc_rqst *rqst) |
95756482 CL |
275 | { |
276 | return (struct sockaddr_in6 *) &rqst->rq_addr; | |
277 | } | |
278 | ||
5344b12d | 279 | static inline struct sockaddr *svc_addr(const struct svc_rqst *rqst) |
27459f09 CL |
280 | { |
281 | return (struct sockaddr *) &rqst->rq_addr; | |
282 | } | |
283 | ||
1da177e4 LT |
284 | /* |
285 | * Check buffer bounds after decoding arguments | |
286 | */ | |
287 | static inline int | |
d8ed029d | 288 | xdr_argsize_check(struct svc_rqst *rqstp, __be32 *p) |
1da177e4 LT |
289 | { |
290 | char *cp = (char *)p; | |
291 | struct kvec *vec = &rqstp->rq_arg.head[0]; | |
0ba7536d N |
292 | return cp >= (char*)vec->iov_base |
293 | && cp <= (char*)vec->iov_base + vec->iov_len; | |
1da177e4 LT |
294 | } |
295 | ||
296 | static inline int | |
d8ed029d | 297 | xdr_ressize_check(struct svc_rqst *rqstp, __be32 *p) |
1da177e4 LT |
298 | { |
299 | struct kvec *vec = &rqstp->rq_res.head[0]; | |
300 | char *cp = (char*)p; | |
301 | ||
302 | vec->iov_len = cp - (char*)vec->iov_base; | |
303 | ||
304 | return vec->iov_len <= PAGE_SIZE; | |
305 | } | |
306 | ||
44524359 | 307 | static inline void svc_free_res_pages(struct svc_rqst *rqstp) |
a257cdd0 | 308 | { |
44524359 N |
309 | while (rqstp->rq_resused) { |
310 | struct page **pp = (rqstp->rq_respages + | |
311 | --rqstp->rq_resused); | |
312 | if (*pp) { | |
313 | put_page(*pp); | |
314 | *pp = NULL; | |
1da177e4 LT |
315 | } |
316 | } | |
317 | } | |
318 | ||
1da177e4 LT |
319 | struct svc_deferred_req { |
320 | u32 prot; /* protocol (UDP or TCP) */ | |
8c7b0172 | 321 | struct svc_xprt *xprt; |
24422222 CL |
322 | struct sockaddr_storage addr; /* where reply must go */ |
323 | size_t addrlen; | |
73df0dba | 324 | union svc_addr_u daddr; /* where reply must come from */ |
1da177e4 | 325 | struct cache_deferred_req handle; |
260c1d12 | 326 | size_t xprt_hlen; |
1da177e4 | 327 | int argslen; |
d8ed029d | 328 | __be32 args[0]; |
1da177e4 LT |
329 | }; |
330 | ||
331 | /* | |
9ba02638 | 332 | * List of RPC programs on the same transport endpoint |
1da177e4 LT |
333 | */ |
334 | struct svc_program { | |
9ba02638 | 335 | struct svc_program * pg_next; /* other programs (same xprt) */ |
1da177e4 LT |
336 | u32 pg_prog; /* program number */ |
337 | unsigned int pg_lovers; /* lowest version */ | |
338 | unsigned int pg_hivers; /* lowest version */ | |
339 | unsigned int pg_nvers; /* number of versions */ | |
340 | struct svc_version ** pg_vers; /* version array */ | |
341 | char * pg_name; /* service name */ | |
342 | char * pg_class; /* class name: services sharing authentication */ | |
343 | struct svc_stat * pg_stats; /* rpc statistics */ | |
344 | int (*pg_authenticate)(struct svc_rqst *); | |
345 | }; | |
346 | ||
347 | /* | |
348 | * RPC program version | |
349 | */ | |
350 | struct svc_version { | |
351 | u32 vs_vers; /* version number */ | |
352 | u32 vs_nproc; /* number of procedures */ | |
353 | struct svc_procedure * vs_proc; /* per-procedure info */ | |
354 | u32 vs_xdrsize; /* xdrsize needed for this version */ | |
355 | ||
bc5fea42 OK |
356 | unsigned int vs_hidden : 1; /* Don't register with portmapper. |
357 | * Only used for nfsacl so far. */ | |
358 | ||
1da177e4 LT |
359 | /* Override dispatch function (e.g. when caching replies). |
360 | * A return value of 0 means drop the request. | |
361 | * vs_dispatch == NULL means use default dispatcher. | |
362 | */ | |
d8ed029d | 363 | int (*vs_dispatch)(struct svc_rqst *, __be32 *); |
1da177e4 LT |
364 | }; |
365 | ||
366 | /* | |
367 | * RPC procedure info | |
368 | */ | |
7111c66e | 369 | typedef __be32 (*svc_procfunc)(struct svc_rqst *, void *argp, void *resp); |
1da177e4 LT |
370 | struct svc_procedure { |
371 | svc_procfunc pc_func; /* process the request */ | |
372 | kxdrproc_t pc_decode; /* XDR decode args */ | |
373 | kxdrproc_t pc_encode; /* XDR encode result */ | |
374 | kxdrproc_t pc_release; /* XDR free result */ | |
375 | unsigned int pc_argsize; /* argument struct size */ | |
376 | unsigned int pc_ressize; /* result struct size */ | |
377 | unsigned int pc_count; /* call count */ | |
378 | unsigned int pc_cachetype; /* cache info (NFS) */ | |
379 | unsigned int pc_xdrressize; /* maximum size of XDR reply */ | |
380 | }; | |
381 | ||
1da177e4 LT |
382 | /* |
383 | * Function prototypes. | |
384 | */ | |
e851db5b CL |
385 | struct svc_serv *svc_create(struct svc_program *, unsigned int, sa_family_t, |
386 | void (*shutdown)(struct svc_serv *)); | |
0113ab34 JL |
387 | struct svc_rqst *svc_prepare_thread(struct svc_serv *serv, |
388 | struct svc_pool *pool); | |
1da177e4 | 389 | void svc_exit_thread(struct svc_rqst *); |
a7455442 | 390 | struct svc_serv * svc_create_pooled(struct svc_program *, unsigned int, |
e851db5b CL |
391 | sa_family_t, void (*shutdown)(struct svc_serv *), |
392 | svc_thread_fn, struct module *); | |
a7455442 | 393 | int svc_set_num_threads(struct svc_serv *, struct svc_pool *, int); |
1da177e4 | 394 | void svc_destroy(struct svc_serv *); |
6fb2b47f | 395 | int svc_process(struct svc_rqst *); |
a26cfad6 CL |
396 | int svc_register(const struct svc_serv *, const unsigned short, |
397 | const unsigned short); | |
398 | ||
1da177e4 LT |
399 | void svc_wake_up(struct svc_serv *); |
400 | void svc_reserve(struct svc_rqst *rqstp, int space); | |
bfd24160 | 401 | struct svc_pool * svc_pool_for_cpu(struct svc_serv *serv, int cpu); |
ad06e4bd CL |
402 | char * svc_print_addr(struct svc_rqst *, char *, size_t); |
403 | ||
404 | #define RPC_MAX_ADDRBUFLEN (63U) | |
1da177e4 | 405 | |
cd123012 JL |
406 | /* |
407 | * When we want to reduce the size of the reserved space in the response | |
408 | * buffer, we need to take into account the size of any checksum data that | |
409 | * may be at the end of the packet. This is difficult to determine exactly | |
410 | * for all cases without actually generating the checksum, so we just use a | |
411 | * static value. | |
412 | */ | |
fbb7878c | 413 | static inline void svc_reserve_auth(struct svc_rqst *rqstp, int space) |
cd123012 | 414 | { |
fbb7878c | 415 | int added_space = 0; |
cd123012 | 416 | |
fbb7878c BF |
417 | if (rqstp->rq_authop->flavour) |
418 | added_space = RPC_MAX_AUTH_SIZE; | |
419 | svc_reserve(rqstp, space + added_space); | |
cd123012 JL |
420 | } |
421 | ||
1da177e4 | 422 | #endif /* SUNRPC_SVC_H */ |