Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
1da177e4 LT |
3 | * Central processing for nfsd. |
4 | * | |
5 | * Authors: Olaf Kirch (okir@monad.swb.de) | |
6 | * | |
7 | * Copyright (C) 1995, 1996, 1997 Olaf Kirch <okir@monad.swb.de> | |
8 | */ | |
9 | ||
3f07c014 | 10 | #include <linux/sched/signal.h> |
83144186 | 11 | #include <linux/freezer.h> |
143cb494 | 12 | #include <linux/module.h> |
1da177e4 | 13 | #include <linux/fs_struct.h> |
c3d06f9c | 14 | #include <linux/swap.h> |
1da177e4 | 15 | |
1da177e4 | 16 | #include <linux/sunrpc/stats.h> |
1da177e4 | 17 | #include <linux/sunrpc/svcsock.h> |
36684996 | 18 | #include <linux/sunrpc/svc_xprt.h> |
1da177e4 | 19 | #include <linux/lockd/bind.h> |
a257cdd0 | 20 | #include <linux/nfsacl.h> |
ed2d8aed | 21 | #include <linux/seq_file.h> |
36684996 SM |
22 | #include <linux/inetdevice.h> |
23 | #include <net/addrconf.h> | |
24 | #include <net/ipv6.h> | |
fc5d00b0 | 25 | #include <net/net_namespace.h> |
9a74af21 BH |
26 | #include "nfsd.h" |
27 | #include "cache.h" | |
0a3adade | 28 | #include "vfs.h" |
2c2fe290 | 29 | #include "netns.h" |
65294c1f | 30 | #include "filecache.h" |
1da177e4 LT |
31 | |
32 | #define NFSDDBG_FACILITY NFSDDBG_SVC | |
33 | ||
ce0887ac OK |
34 | bool inter_copy_offload_enable; |
35 | EXPORT_SYMBOL_GPL(inter_copy_offload_enable); | |
36 | module_param(inter_copy_offload_enable, bool, 0644); | |
37 | MODULE_PARM_DESC(inter_copy_offload_enable, | |
38 | "Enable inter server to server copy offload. Default: false"); | |
39 | ||
1da177e4 | 40 | extern struct svc_program nfsd_program; |
9867d76c | 41 | static int nfsd(void *vrqstp); |
029be5d0 TM |
42 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) |
43 | static int nfsd_acl_rpcbind_set(struct net *, | |
44 | const struct svc_program *, | |
45 | u32, int, | |
46 | unsigned short, | |
47 | unsigned short); | |
e333f3bb TM |
48 | static __be32 nfsd_acl_init_request(struct svc_rqst *, |
49 | const struct svc_program *, | |
50 | struct svc_process_info *); | |
029be5d0 TM |
51 | #endif |
52 | static int nfsd_rpcbind_set(struct net *, | |
53 | const struct svc_program *, | |
54 | u32, int, | |
55 | unsigned short, | |
56 | unsigned short); | |
e333f3bb TM |
57 | static __be32 nfsd_init_request(struct svc_rqst *, |
58 | const struct svc_program *, | |
59 | struct svc_process_info *); | |
1da177e4 | 60 | |
bedbdd8b | 61 | /* |
9dd9845f | 62 | * nfsd_mutex protects nn->nfsd_serv -- both the pointer itself and the members |
bedbdd8b NB |
63 | * of the svc_serv struct. In particular, ->sv_nrthreads but also to some |
64 | * extent ->sv_temp_socks and ->sv_permsocks. It also protects nfsdstats.th_cnt | |
65 | * | |
9dd9845f | 66 | * If (out side the lock) nn->nfsd_serv is non-NULL, then it must point to a |
bedbdd8b NB |
67 | * properly initialised 'struct svc_serv' with ->sv_nrthreads > 0. That number |
68 | * of nfsd threads must exist and each must listed in ->sp_all_threads in each | |
69 | * entry of ->sv_pools[]. | |
70 | * | |
71 | * Transitions of the thread count between zero and non-zero are of particular | |
72 | * interest since the svc_serv needs to be created and initialized at that | |
73 | * point, or freed. | |
3dd98a3b JL |
74 | * |
75 | * Finally, the nfsd_mutex also protects some of the global variables that are | |
76 | * accessed when nfsd starts and that are settable via the write_* routines in | |
77 | * nfsctl.c. In particular: | |
78 | * | |
79 | * user_recovery_dirname | |
80 | * user_lease_time | |
81 | * nfsd_versions | |
bedbdd8b NB |
82 | */ |
83 | DEFINE_MUTEX(nfsd_mutex); | |
bedbdd8b | 84 | |
4bd9b0f4 AA |
85 | /* |
86 | * nfsd_drc_lock protects nfsd_drc_max_pages and nfsd_drc_pages_used. | |
87 | * nfsd_drc_max_pages limits the total amount of memory available for | |
88 | * version 4.1 DRC caches. | |
89 | * nfsd_drc_pages_used tracks the current version 4.1 DRC memory usage. | |
90 | */ | |
91 | spinlock_t nfsd_drc_lock; | |
697ce9be ZY |
92 | unsigned long nfsd_drc_max_mem; |
93 | unsigned long nfsd_drc_mem_used; | |
4bd9b0f4 | 94 | |
3fb803a9 AG |
95 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) |
96 | static struct svc_stat nfsd_acl_svcstats; | |
e9679189 | 97 | static const struct svc_version *nfsd_acl_version[] = { |
3fb803a9 AG |
98 | [2] = &nfsd_acl_version2, |
99 | [3] = &nfsd_acl_version3, | |
100 | }; | |
101 | ||
102 | #define NFSD_ACL_MINVERS 2 | |
e8c96f8c | 103 | #define NFSD_ACL_NRVERS ARRAY_SIZE(nfsd_acl_version) |
3fb803a9 AG |
104 | |
105 | static struct svc_program nfsd_acl_program = { | |
106 | .pg_prog = NFS_ACL_PROGRAM, | |
107 | .pg_nvers = NFSD_ACL_NRVERS, | |
7c149057 | 108 | .pg_vers = nfsd_acl_version, |
1a8eff6d | 109 | .pg_name = "nfsacl", |
3fb803a9 AG |
110 | .pg_class = "nfsd", |
111 | .pg_stats = &nfsd_acl_svcstats, | |
112 | .pg_authenticate = &svc_set_client, | |
e333f3bb | 113 | .pg_init_request = nfsd_acl_init_request, |
029be5d0 | 114 | .pg_rpcbind_set = nfsd_acl_rpcbind_set, |
3fb803a9 AG |
115 | }; |
116 | ||
117 | static struct svc_stat nfsd_acl_svcstats = { | |
118 | .program = &nfsd_acl_program, | |
119 | }; | |
120 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ | |
121 | ||
e9679189 | 122 | static const struct svc_version *nfsd_version[] = { |
70c3b76c N |
123 | [2] = &nfsd_version2, |
124 | #if defined(CONFIG_NFSD_V3) | |
125 | [3] = &nfsd_version3, | |
126 | #endif | |
127 | #if defined(CONFIG_NFSD_V4) | |
128 | [4] = &nfsd_version4, | |
129 | #endif | |
130 | }; | |
131 | ||
132 | #define NFSD_MINVERS 2 | |
e8c96f8c | 133 | #define NFSD_NRVERS ARRAY_SIZE(nfsd_version) |
70c3b76c N |
134 | |
135 | struct svc_program nfsd_program = { | |
3fb803a9 AG |
136 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) |
137 | .pg_next = &nfsd_acl_program, | |
138 | #endif | |
70c3b76c N |
139 | .pg_prog = NFS_PROGRAM, /* program number */ |
140 | .pg_nvers = NFSD_NRVERS, /* nr of entries in nfsd_version */ | |
e333f3bb | 141 | .pg_vers = nfsd_version, /* version table */ |
70c3b76c N |
142 | .pg_name = "nfsd", /* program name */ |
143 | .pg_class = "nfsd", /* authentication class */ | |
144 | .pg_stats = &nfsd_svcstats, /* version table */ | |
145 | .pg_authenticate = &svc_set_client, /* export authentication */ | |
e333f3bb | 146 | .pg_init_request = nfsd_init_request, |
029be5d0 | 147 | .pg_rpcbind_set = nfsd_rpcbind_set, |
70c3b76c N |
148 | }; |
149 | ||
e333f3bb TM |
150 | static bool |
151 | nfsd_support_version(int vers) | |
152 | { | |
153 | if (vers >= NFSD_MINVERS && vers < NFSD_NRVERS) | |
154 | return nfsd_version[vers] != NULL; | |
155 | return false; | |
156 | } | |
157 | ||
158 | static bool * | |
159 | nfsd_alloc_versions(void) | |
160 | { | |
161 | bool *vers = kmalloc_array(NFSD_NRVERS, sizeof(bool), GFP_KERNEL); | |
162 | unsigned i; | |
163 | ||
164 | if (vers) { | |
165 | /* All compiled versions are enabled by default */ | |
166 | for (i = 0; i < NFSD_NRVERS; i++) | |
167 | vers[i] = nfsd_support_version(i); | |
168 | } | |
169 | return vers; | |
170 | } | |
171 | ||
172 | static bool * | |
173 | nfsd_alloc_minorversions(void) | |
174 | { | |
175 | bool *vers = kmalloc_array(NFSD_SUPPORTED_MINOR_VERSION + 1, | |
176 | sizeof(bool), GFP_KERNEL); | |
177 | unsigned i; | |
8daf220a | 178 | |
e333f3bb TM |
179 | if (vers) { |
180 | /* All minor versions are enabled by default */ | |
181 | for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++) | |
182 | vers[i] = nfsd_support_version(4); | |
183 | } | |
184 | return vers; | |
185 | } | |
186 | ||
187 | void | |
188 | nfsd_netns_free_versions(struct nfsd_net *nn) | |
189 | { | |
190 | kfree(nn->nfsd_versions); | |
191 | kfree(nn->nfsd4_minorversions); | |
192 | nn->nfsd_versions = NULL; | |
193 | nn->nfsd4_minorversions = NULL; | |
194 | } | |
195 | ||
196 | static void | |
197 | nfsd_netns_init_versions(struct nfsd_net *nn) | |
198 | { | |
199 | if (!nn->nfsd_versions) { | |
200 | nn->nfsd_versions = nfsd_alloc_versions(); | |
201 | nn->nfsd4_minorversions = nfsd_alloc_minorversions(); | |
202 | if (!nn->nfsd_versions || !nn->nfsd4_minorversions) | |
203 | nfsd_netns_free_versions(nn); | |
204 | } | |
205 | } | |
206 | ||
207 | int nfsd_vers(struct nfsd_net *nn, int vers, enum vers_op change) | |
6658d3a7 N |
208 | { |
209 | if (vers < NFSD_MINVERS || vers >= NFSD_NRVERS) | |
15ddb4ae | 210 | return 0; |
6658d3a7 N |
211 | switch(change) { |
212 | case NFSD_SET: | |
e333f3bb TM |
213 | if (nn->nfsd_versions) |
214 | nn->nfsd_versions[vers] = nfsd_support_version(vers); | |
1a8eff6d | 215 | break; |
6658d3a7 | 216 | case NFSD_CLEAR: |
e333f3bb TM |
217 | nfsd_netns_init_versions(nn); |
218 | if (nn->nfsd_versions) | |
219 | nn->nfsd_versions[vers] = false; | |
6658d3a7 N |
220 | break; |
221 | case NFSD_TEST: | |
e333f3bb TM |
222 | if (nn->nfsd_versions) |
223 | return nn->nfsd_versions[vers]; | |
df561f66 | 224 | fallthrough; |
6658d3a7 | 225 | case NFSD_AVAIL: |
e333f3bb | 226 | return nfsd_support_version(vers); |
6658d3a7 N |
227 | } |
228 | return 0; | |
229 | } | |
8daf220a | 230 | |
d3635ff0 | 231 | static void |
e333f3bb | 232 | nfsd_adjust_nfsd_versions4(struct nfsd_net *nn) |
d3635ff0 TM |
233 | { |
234 | unsigned i; | |
235 | ||
236 | for (i = 0; i <= NFSD_SUPPORTED_MINOR_VERSION; i++) { | |
e333f3bb | 237 | if (nn->nfsd4_minorversions[i]) |
d3635ff0 TM |
238 | return; |
239 | } | |
e333f3bb | 240 | nfsd_vers(nn, 4, NFSD_CLEAR); |
d3635ff0 TM |
241 | } |
242 | ||
e333f3bb | 243 | int nfsd_minorversion(struct nfsd_net *nn, u32 minorversion, enum vers_op change) |
8daf220a | 244 | { |
928c6fb3 N |
245 | if (minorversion > NFSD_SUPPORTED_MINOR_VERSION && |
246 | change != NFSD_AVAIL) | |
8daf220a | 247 | return -1; |
e333f3bb | 248 | |
8daf220a BH |
249 | switch(change) { |
250 | case NFSD_SET: | |
e333f3bb TM |
251 | if (nn->nfsd4_minorversions) { |
252 | nfsd_vers(nn, 4, NFSD_SET); | |
253 | nn->nfsd4_minorversions[minorversion] = | |
254 | nfsd_vers(nn, 4, NFSD_TEST); | |
255 | } | |
8daf220a BH |
256 | break; |
257 | case NFSD_CLEAR: | |
e333f3bb TM |
258 | nfsd_netns_init_versions(nn); |
259 | if (nn->nfsd4_minorversions) { | |
260 | nn->nfsd4_minorversions[minorversion] = false; | |
261 | nfsd_adjust_nfsd_versions4(nn); | |
262 | } | |
8daf220a BH |
263 | break; |
264 | case NFSD_TEST: | |
e333f3bb TM |
265 | if (nn->nfsd4_minorversions) |
266 | return nn->nfsd4_minorversions[minorversion]; | |
267 | return nfsd_vers(nn, 4, NFSD_TEST); | |
8daf220a | 268 | case NFSD_AVAIL: |
e333f3bb TM |
269 | return minorversion <= NFSD_SUPPORTED_MINOR_VERSION && |
270 | nfsd_vers(nn, 4, NFSD_AVAIL); | |
8daf220a BH |
271 | } |
272 | return 0; | |
273 | } | |
274 | ||
1da177e4 LT |
275 | /* |
276 | * Maximum number of nfsd processes | |
277 | */ | |
278 | #define NFSD_MAXSERVS 8192 | |
279 | ||
9dd9845f | 280 | int nfsd_nrthreads(struct net *net) |
1da177e4 | 281 | { |
c7d106c9 | 282 | int rv = 0; |
9dd9845f SK |
283 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
284 | ||
c7d106c9 | 285 | mutex_lock(&nfsd_mutex); |
9dd9845f SK |
286 | if (nn->nfsd_serv) |
287 | rv = nn->nfsd_serv->sv_nrthreads; | |
c7d106c9 NB |
288 | mutex_unlock(&nfsd_mutex); |
289 | return rv; | |
1da177e4 LT |
290 | } |
291 | ||
4df493a2 | 292 | static int nfsd_init_socks(struct net *net, const struct cred *cred) |
59db4a0c BF |
293 | { |
294 | int error; | |
9dd9845f SK |
295 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
296 | ||
297 | if (!list_empty(&nn->nfsd_serv->sv_permsocks)) | |
59db4a0c BF |
298 | return 0; |
299 | ||
9dd9845f | 300 | error = svc_create_xprt(nn->nfsd_serv, "udp", net, PF_INET, NFS_PORT, |
4df493a2 | 301 | SVC_SOCK_DEFAULTS, cred); |
59db4a0c BF |
302 | if (error < 0) |
303 | return error; | |
304 | ||
9dd9845f | 305 | error = svc_create_xprt(nn->nfsd_serv, "tcp", net, PF_INET, NFS_PORT, |
4df493a2 | 306 | SVC_SOCK_DEFAULTS, cred); |
59db4a0c BF |
307 | if (error < 0) |
308 | return error; | |
309 | ||
310 | return 0; | |
311 | } | |
312 | ||
4539f149 | 313 | static int nfsd_users = 0; |
4ad9a344 | 314 | |
bda9cac1 SK |
315 | static int nfsd_startup_generic(int nrservs) |
316 | { | |
317 | int ret; | |
318 | ||
4539f149 | 319 | if (nfsd_users++) |
bda9cac1 SK |
320 | return 0; |
321 | ||
65294c1f JL |
322 | ret = nfsd_file_cache_init(); |
323 | if (ret) | |
324 | goto dec_users; | |
d9499a95 | 325 | |
bda9cac1 SK |
326 | ret = nfs4_state_start(); |
327 | if (ret) | |
501cb184 | 328 | goto out_file_cache; |
bda9cac1 SK |
329 | return 0; |
330 | ||
65294c1f JL |
331 | out_file_cache: |
332 | nfsd_file_cache_shutdown(); | |
d9499a95 KM |
333 | dec_users: |
334 | nfsd_users--; | |
bda9cac1 SK |
335 | return ret; |
336 | } | |
337 | ||
338 | static void nfsd_shutdown_generic(void) | |
339 | { | |
4539f149 SK |
340 | if (--nfsd_users) |
341 | return; | |
342 | ||
bda9cac1 | 343 | nfs4_state_shutdown(); |
65294c1f | 344 | nfsd_file_cache_shutdown(); |
bda9cac1 SK |
345 | } |
346 | ||
e333f3bb | 347 | static bool nfsd_needs_lockd(struct nfsd_net *nn) |
8ef66714 | 348 | { |
e333f3bb | 349 | return nfsd_vers(nn, 2, NFSD_TEST) || nfsd_vers(nn, 3, NFSD_TEST); |
8ef66714 KM |
350 | } |
351 | ||
27c438f5 TM |
352 | void nfsd_copy_boot_verifier(__be32 verf[2], struct nfsd_net *nn) |
353 | { | |
354 | int seq = 0; | |
355 | ||
356 | do { | |
357 | read_seqbegin_or_lock(&nn->boot_lock, &seq); | |
358 | /* | |
359 | * This is opaque to client, so no need to byte-swap. Use | |
360 | * __force to keep sparse happy. y2038 time_t overflow is | |
361 | * irrelevant in this usage | |
362 | */ | |
363 | verf[0] = (__force __be32)nn->nfssvc_boot.tv_sec; | |
364 | verf[1] = (__force __be32)nn->nfssvc_boot.tv_nsec; | |
365 | } while (need_seqretry(&nn->boot_lock, seq)); | |
366 | done_seqretry(&nn->boot_lock, seq); | |
367 | } | |
368 | ||
65643f4c | 369 | static void nfsd_reset_boot_verifier_locked(struct nfsd_net *nn) |
27c438f5 TM |
370 | { |
371 | ktime_get_real_ts64(&nn->nfssvc_boot); | |
372 | } | |
373 | ||
374 | void nfsd_reset_boot_verifier(struct nfsd_net *nn) | |
375 | { | |
376 | write_seqlock(&nn->boot_lock); | |
377 | nfsd_reset_boot_verifier_locked(nn); | |
378 | write_sequnlock(&nn->boot_lock); | |
379 | } | |
380 | ||
4df493a2 | 381 | static int nfsd_startup_net(int nrservs, struct net *net, const struct cred *cred) |
6ff50b3d | 382 | { |
2c2fe290 | 383 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
6ff50b3d SK |
384 | int ret; |
385 | ||
2c2fe290 SK |
386 | if (nn->nfsd_net_up) |
387 | return 0; | |
388 | ||
903d9bf0 | 389 | ret = nfsd_startup_generic(nrservs); |
6ff50b3d SK |
390 | if (ret) |
391 | return ret; | |
4df493a2 | 392 | ret = nfsd_init_socks(net, cred); |
903d9bf0 SK |
393 | if (ret) |
394 | goto out_socks; | |
8ef66714 | 395 | |
e333f3bb | 396 | if (nfsd_needs_lockd(nn) && !nn->lockd_up) { |
40373b12 | 397 | ret = lockd_up(net, cred); |
8ef66714 KM |
398 | if (ret) |
399 | goto out_socks; | |
e44b4bf2 | 400 | nn->lockd_up = true; |
8ef66714 KM |
401 | } |
402 | ||
9542e6a6 | 403 | ret = nfsd_file_cache_start_net(net); |
6ff50b3d SK |
404 | if (ret) |
405 | goto out_lockd; | |
9542e6a6 TM |
406 | ret = nfs4_state_start_net(net); |
407 | if (ret) | |
408 | goto out_filecache; | |
6ff50b3d | 409 | |
2c2fe290 | 410 | nn->nfsd_net_up = true; |
6ff50b3d SK |
411 | return 0; |
412 | ||
9542e6a6 TM |
413 | out_filecache: |
414 | nfsd_file_cache_shutdown_net(net); | |
6ff50b3d | 415 | out_lockd: |
8ef66714 KM |
416 | if (nn->lockd_up) { |
417 | lockd_down(net); | |
e44b4bf2 | 418 | nn->lockd_up = false; |
8ef66714 | 419 | } |
903d9bf0 | 420 | out_socks: |
bda9cac1 | 421 | nfsd_shutdown_generic(); |
4ad9a344 JL |
422 | return ret; |
423 | } | |
424 | ||
6ff50b3d SK |
425 | static void nfsd_shutdown_net(struct net *net) |
426 | { | |
2c2fe290 SK |
427 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
428 | ||
9542e6a6 | 429 | nfsd_file_cache_shutdown_net(net); |
6ff50b3d | 430 | nfs4_state_shutdown_net(net); |
8ef66714 KM |
431 | if (nn->lockd_up) { |
432 | lockd_down(net); | |
e44b4bf2 | 433 | nn->lockd_up = false; |
8ef66714 | 434 | } |
2c2fe290 | 435 | nn->nfsd_net_up = false; |
903d9bf0 | 436 | nfsd_shutdown_generic(); |
6ff50b3d SK |
437 | } |
438 | ||
36684996 SM |
439 | static int nfsd_inetaddr_event(struct notifier_block *this, unsigned long event, |
440 | void *ptr) | |
441 | { | |
442 | struct in_ifaddr *ifa = (struct in_ifaddr *)ptr; | |
443 | struct net_device *dev = ifa->ifa_dev->dev; | |
444 | struct net *net = dev_net(dev); | |
445 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
446 | struct sockaddr_in sin; | |
447 | ||
2317dc55 VA |
448 | if ((event != NETDEV_DOWN) || |
449 | !atomic_inc_not_zero(&nn->ntf_refcnt)) | |
36684996 SM |
450 | goto out; |
451 | ||
452 | if (nn->nfsd_serv) { | |
453 | dprintk("nfsd_inetaddr_event: removed %pI4\n", &ifa->ifa_local); | |
454 | sin.sin_family = AF_INET; | |
455 | sin.sin_addr.s_addr = ifa->ifa_local; | |
456 | svc_age_temp_xprts_now(nn->nfsd_serv, (struct sockaddr *)&sin); | |
457 | } | |
2317dc55 VA |
458 | atomic_dec(&nn->ntf_refcnt); |
459 | wake_up(&nn->ntf_wq); | |
36684996 SM |
460 | |
461 | out: | |
462 | return NOTIFY_DONE; | |
463 | } | |
464 | ||
465 | static struct notifier_block nfsd_inetaddr_notifier = { | |
466 | .notifier_call = nfsd_inetaddr_event, | |
467 | }; | |
468 | ||
469 | #if IS_ENABLED(CONFIG_IPV6) | |
470 | static int nfsd_inet6addr_event(struct notifier_block *this, | |
471 | unsigned long event, void *ptr) | |
472 | { | |
473 | struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr; | |
474 | struct net_device *dev = ifa->idev->dev; | |
475 | struct net *net = dev_net(dev); | |
476 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
477 | struct sockaddr_in6 sin6; | |
478 | ||
2317dc55 VA |
479 | if ((event != NETDEV_DOWN) || |
480 | !atomic_inc_not_zero(&nn->ntf_refcnt)) | |
36684996 SM |
481 | goto out; |
482 | ||
483 | if (nn->nfsd_serv) { | |
484 | dprintk("nfsd_inet6addr_event: removed %pI6\n", &ifa->addr); | |
485 | sin6.sin6_family = AF_INET6; | |
486 | sin6.sin6_addr = ifa->addr; | |
7b19824d SM |
487 | if (ipv6_addr_type(&sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL) |
488 | sin6.sin6_scope_id = ifa->idev->dev->ifindex; | |
36684996 SM |
489 | svc_age_temp_xprts_now(nn->nfsd_serv, (struct sockaddr *)&sin6); |
490 | } | |
2317dc55 VA |
491 | atomic_dec(&nn->ntf_refcnt); |
492 | wake_up(&nn->ntf_wq); | |
36684996 SM |
493 | out: |
494 | return NOTIFY_DONE; | |
495 | } | |
496 | ||
497 | static struct notifier_block nfsd_inet6addr_notifier = { | |
498 | .notifier_call = nfsd_inet6addr_event, | |
499 | }; | |
500 | #endif | |
501 | ||
1eca45f8 VA |
502 | /* Only used under nfsd_mutex, so this atomic may be overkill: */ |
503 | static atomic_t nfsd_notifier_refcount = ATOMIC_INIT(0); | |
504 | ||
541e864f | 505 | static void nfsd_last_thread(struct svc_serv *serv, struct net *net) |
4ad9a344 | 506 | { |
903d9bf0 SK |
507 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
508 | ||
2317dc55 | 509 | atomic_dec(&nn->ntf_refcnt); |
1eca45f8 VA |
510 | /* check if the notifier still has clients */ |
511 | if (atomic_dec_return(&nfsd_notifier_refcount) == 0) { | |
512 | unregister_inetaddr_notifier(&nfsd_inetaddr_notifier); | |
36684996 | 513 | #if IS_ENABLED(CONFIG_IPV6) |
1eca45f8 | 514 | unregister_inet6addr_notifier(&nfsd_inet6addr_notifier); |
36684996 | 515 | #endif |
1eca45f8 | 516 | } |
2317dc55 | 517 | wait_event(nn->ntf_wq, atomic_read(&nn->ntf_refcnt) == 0); |
1eca45f8 | 518 | |
4ad9a344 JL |
519 | /* |
520 | * write_ports can create the server without actually starting | |
521 | * any threads--if we get shut down before any threads are | |
522 | * started, then nfsd_last_thread will be run before any of this | |
691412b4 | 523 | * other initialization has been done except the rpcb information. |
4ad9a344 | 524 | */ |
691412b4 | 525 | svc_rpcb_cleanup(serv, net); |
903d9bf0 | 526 | if (!nn->nfsd_net_up) |
4ad9a344 | 527 | return; |
16d05870 | 528 | |
691412b4 | 529 | nfsd_shutdown_net(net); |
e096bbc6 JL |
530 | printk(KERN_WARNING "nfsd: last server has exited, flushing export " |
531 | "cache\n"); | |
b3853e0e | 532 | nfsd_export_flush(net); |
bc591ccf | 533 | } |
6658d3a7 | 534 | |
e333f3bb | 535 | void nfsd_reset_versions(struct nfsd_net *nn) |
6658d3a7 | 536 | { |
6658d3a7 N |
537 | int i; |
538 | ||
800a938f | 539 | for (i = 0; i < NFSD_NRVERS; i++) |
e333f3bb | 540 | if (nfsd_vers(nn, i, NFSD_TEST)) |
800a938f | 541 | return; |
6658d3a7 | 542 | |
800a938f N |
543 | for (i = 0; i < NFSD_NRVERS; i++) |
544 | if (i != 4) | |
e333f3bb | 545 | nfsd_vers(nn, i, NFSD_SET); |
800a938f N |
546 | else { |
547 | int minor = 0; | |
e333f3bb | 548 | while (nfsd_minorversion(nn, minor, NFSD_SET) >= 0) |
800a938f N |
549 | minor++; |
550 | } | |
6658d3a7 N |
551 | } |
552 | ||
c3d06f9c AA |
553 | /* |
554 | * Each session guarantees a negotiated per slot memory cache for replies | |
555 | * which in turn consumes memory beyond the v2/v3/v4.0 server. A dedicated | |
556 | * NFSv4.1 server might want to use more memory for a DRC than a machine | |
557 | * with mutiple services. | |
558 | * | |
559 | * Impose a hard limit on the number of pages for the DRC which varies | |
560 | * according to the machines free pages. This is of course only a default. | |
561 | * | |
562 | * For now this is a #defined shift which could be under admin control | |
563 | * in the future. | |
564 | */ | |
565 | static void set_max_drc(void) | |
566 | { | |
44d8660d | 567 | #define NFSD_DRC_SIZE_SHIFT 7 |
0c193054 AA |
568 | nfsd_drc_max_mem = (nr_free_buffer_pages() |
569 | >> NFSD_DRC_SIZE_SHIFT) * PAGE_SIZE; | |
570 | nfsd_drc_mem_used = 0; | |
4bd9b0f4 | 571 | spin_lock_init(&nfsd_drc_lock); |
697ce9be | 572 | dprintk("%s nfsd_drc_max_mem %lu \n", __func__, nfsd_drc_max_mem); |
c3d06f9c | 573 | } |
bedbdd8b | 574 | |
87b0fc7d | 575 | static int nfsd_get_default_max_blksize(void) |
02a375f0 | 576 | { |
87b0fc7d BF |
577 | struct sysinfo i; |
578 | unsigned long long target; | |
579 | unsigned long ret; | |
bedbdd8b | 580 | |
87b0fc7d | 581 | si_meminfo(&i); |
508f9227 | 582 | target = (i.totalram - i.totalhigh) << PAGE_SHIFT; |
87b0fc7d BF |
583 | /* |
584 | * Aim for 1/4096 of memory per thread This gives 1MB on 4Gig | |
585 | * machines, but only uses 32K on 128M machines. Bottom out at | |
586 | * 8K on 32M and smaller. Of course, this is only a default. | |
587 | */ | |
588 | target >>= 12; | |
589 | ||
590 | ret = NFSSVC_MAXBLKSIZE; | |
591 | while (ret > target && ret >= 8*1024*2) | |
592 | ret /= 2; | |
593 | return ret; | |
594 | } | |
595 | ||
afea5657 | 596 | static const struct svc_serv_ops nfsd_thread_sv_ops = { |
b9e13cdf JL |
597 | .svo_shutdown = nfsd_last_thread, |
598 | .svo_function = nfsd, | |
599 | .svo_enqueue_xprt = svc_xprt_do_enqueue, | |
598e2359 | 600 | .svo_setup = svc_set_num_threads, |
b9e13cdf | 601 | .svo_module = THIS_MODULE, |
ea126e74 JL |
602 | }; |
603 | ||
44fb26c6 | 604 | bool i_am_nfsd(void) |
28df3d15 BF |
605 | { |
606 | return kthread_func(current) == nfsd; | |
607 | } | |
608 | ||
6777436b | 609 | int nfsd_create_serv(struct net *net) |
87b0fc7d | 610 | { |
9793f7c8 | 611 | int error; |
b9c0ef85 | 612 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
9793f7c8 | 613 | |
bedbdd8b | 614 | WARN_ON(!mutex_is_locked(&nfsd_mutex)); |
9dd9845f SK |
615 | if (nn->nfsd_serv) { |
616 | svc_get(nn->nfsd_serv); | |
02a375f0 N |
617 | return 0; |
618 | } | |
87b0fc7d BF |
619 | if (nfsd_max_blksize == 0) |
620 | nfsd_max_blksize = nfsd_get_default_max_blksize(); | |
e333f3bb | 621 | nfsd_reset_versions(nn); |
9dd9845f | 622 | nn->nfsd_serv = svc_create_pooled(&nfsd_program, nfsd_max_blksize, |
b9e13cdf | 623 | &nfsd_thread_sv_ops); |
9dd9845f | 624 | if (nn->nfsd_serv == NULL) |
628b3687 | 625 | return -ENOMEM; |
bedbdd8b | 626 | |
5b8db00b | 627 | nn->nfsd_serv->sv_maxconn = nn->max_connections; |
9dd9845f | 628 | error = svc_bind(nn->nfsd_serv, net); |
9793f7c8 | 629 | if (error < 0) { |
9dd9845f | 630 | svc_destroy(nn->nfsd_serv); |
9793f7c8 SK |
631 | return error; |
632 | } | |
633 | ||
628b3687 | 634 | set_max_drc(); |
1eca45f8 VA |
635 | /* check if the notifier is already set */ |
636 | if (atomic_inc_return(&nfsd_notifier_refcount) == 1) { | |
637 | register_inetaddr_notifier(&nfsd_inetaddr_notifier); | |
36684996 | 638 | #if IS_ENABLED(CONFIG_IPV6) |
1eca45f8 | 639 | register_inet6addr_notifier(&nfsd_inet6addr_notifier); |
36684996 | 640 | #endif |
1eca45f8 | 641 | } |
2317dc55 | 642 | atomic_inc(&nn->ntf_refcnt); |
27c438f5 | 643 | nfsd_reset_boot_verifier(nn); |
87b0fc7d | 644 | return 0; |
02a375f0 N |
645 | } |
646 | ||
9dd9845f | 647 | int nfsd_nrpools(struct net *net) |
eed2965a | 648 | { |
9dd9845f SK |
649 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
650 | ||
651 | if (nn->nfsd_serv == NULL) | |
eed2965a GB |
652 | return 0; |
653 | else | |
9dd9845f | 654 | return nn->nfsd_serv->sv_nrpools; |
eed2965a GB |
655 | } |
656 | ||
9dd9845f | 657 | int nfsd_get_nrthreads(int n, int *nthreads, struct net *net) |
eed2965a GB |
658 | { |
659 | int i = 0; | |
9dd9845f | 660 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
eed2965a | 661 | |
9dd9845f SK |
662 | if (nn->nfsd_serv != NULL) { |
663 | for (i = 0; i < nn->nfsd_serv->sv_nrpools && i < n; i++) | |
664 | nthreads[i] = nn->nfsd_serv->sv_pools[i].sp_nrthreads; | |
eed2965a GB |
665 | } |
666 | ||
667 | return 0; | |
668 | } | |
669 | ||
9dd9845f SK |
670 | void nfsd_destroy(struct net *net) |
671 | { | |
672 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
673 | int destroy = (nn->nfsd_serv->sv_nrthreads == 1); | |
674 | ||
675 | if (destroy) | |
676 | svc_shutdown_net(nn->nfsd_serv, net); | |
677 | svc_destroy(nn->nfsd_serv); | |
678 | if (destroy) | |
679 | nn->nfsd_serv = NULL; | |
680 | } | |
681 | ||
3938a0d5 | 682 | int nfsd_set_nrthreads(int n, int *nthreads, struct net *net) |
eed2965a GB |
683 | { |
684 | int i = 0; | |
685 | int tot = 0; | |
686 | int err = 0; | |
9dd9845f | 687 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
eed2965a | 688 | |
bedbdd8b NB |
689 | WARN_ON(!mutex_is_locked(&nfsd_mutex)); |
690 | ||
9dd9845f | 691 | if (nn->nfsd_serv == NULL || n <= 0) |
eed2965a GB |
692 | return 0; |
693 | ||
9dd9845f SK |
694 | if (n > nn->nfsd_serv->sv_nrpools) |
695 | n = nn->nfsd_serv->sv_nrpools; | |
eed2965a GB |
696 | |
697 | /* enforce a global maximum number of threads */ | |
698 | tot = 0; | |
699 | for (i = 0; i < n; i++) { | |
3c7aa15d | 700 | nthreads[i] = min(nthreads[i], NFSD_MAXSERVS); |
eed2965a GB |
701 | tot += nthreads[i]; |
702 | } | |
703 | if (tot > NFSD_MAXSERVS) { | |
704 | /* total too large: scale down requested numbers */ | |
705 | for (i = 0; i < n && tot > 0; i++) { | |
706 | int new = nthreads[i] * NFSD_MAXSERVS / tot; | |
707 | tot -= (nthreads[i] - new); | |
708 | nthreads[i] = new; | |
709 | } | |
710 | for (i = 0; i < n && tot > 0; i++) { | |
711 | nthreads[i]--; | |
712 | tot--; | |
713 | } | |
714 | } | |
715 | ||
716 | /* | |
717 | * There must always be a thread in pool 0; the admin | |
718 | * can't shut down NFS completely using pool_threads. | |
719 | */ | |
720 | if (nthreads[0] == 0) | |
721 | nthreads[0] = 1; | |
722 | ||
723 | /* apply the new numbers */ | |
9dd9845f | 724 | svc_get(nn->nfsd_serv); |
eed2965a | 725 | for (i = 0; i < n; i++) { |
598e2359 JL |
726 | err = nn->nfsd_serv->sv_ops->svo_setup(nn->nfsd_serv, |
727 | &nn->nfsd_serv->sv_pools[i], nthreads[i]); | |
eed2965a GB |
728 | if (err) |
729 | break; | |
730 | } | |
19f7e2ca | 731 | nfsd_destroy(net); |
eed2965a GB |
732 | return err; |
733 | } | |
734 | ||
ac77efbe JL |
735 | /* |
736 | * Adjust the number of threads and return the new number of threads. | |
737 | * This is also the function that starts the server if necessary, if | |
738 | * this is the first time nrservs is nonzero. | |
739 | */ | |
1da177e4 | 740 | int |
4df493a2 | 741 | nfsd_svc(int nrservs, struct net *net, const struct cred *cred) |
1da177e4 LT |
742 | { |
743 | int error; | |
774f8bbd | 744 | bool nfsd_up_before; |
9dd9845f | 745 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
bedbdd8b NB |
746 | |
747 | mutex_lock(&nfsd_mutex); | |
6658d3a7 | 748 | dprintk("nfsd: creating service\n"); |
3c7aa15d KM |
749 | |
750 | nrservs = max(nrservs, 0); | |
751 | nrservs = min(nrservs, NFSD_MAXSERVS); | |
671e1fcf | 752 | error = 0; |
3c7aa15d | 753 | |
9dd9845f | 754 | if (nrservs == 0 && nn->nfsd_serv == NULL) |
671e1fcf N |
755 | goto out; |
756 | ||
7627d7dc SM |
757 | strlcpy(nn->nfsd_name, utsname()->nodename, |
758 | sizeof(nn->nfsd_name)); | |
759 | ||
6777436b | 760 | error = nfsd_create_serv(net); |
02a375f0 | 761 | if (error) |
774f8bbd BF |
762 | goto out; |
763 | ||
903d9bf0 | 764 | nfsd_up_before = nn->nfsd_net_up; |
774f8bbd | 765 | |
4df493a2 | 766 | error = nfsd_startup_net(nrservs, net, cred); |
af4718f3 BF |
767 | if (error) |
768 | goto out_destroy; | |
598e2359 JL |
769 | error = nn->nfsd_serv->sv_ops->svo_setup(nn->nfsd_serv, |
770 | NULL, nrservs); | |
774f8bbd BF |
771 | if (error) |
772 | goto out_shutdown; | |
9dd9845f | 773 | /* We are holding a reference to nn->nfsd_serv which |
af4718f3 BF |
774 | * we don't want to count in the return value, |
775 | * so subtract 1 | |
776 | */ | |
9dd9845f | 777 | error = nn->nfsd_serv->sv_nrthreads - 1; |
4ad9a344 | 778 | out_shutdown: |
774f8bbd | 779 | if (error < 0 && !nfsd_up_before) |
541e864f | 780 | nfsd_shutdown_net(net); |
774f8bbd | 781 | out_destroy: |
19f7e2ca | 782 | nfsd_destroy(net); /* Release server */ |
4ad9a344 | 783 | out: |
bedbdd8b | 784 | mutex_unlock(&nfsd_mutex); |
1da177e4 LT |
785 | return error; |
786 | } | |
787 | ||
029be5d0 TM |
788 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) |
789 | static bool | |
790 | nfsd_support_acl_version(int vers) | |
791 | { | |
792 | if (vers >= NFSD_ACL_MINVERS && vers < NFSD_ACL_NRVERS) | |
793 | return nfsd_acl_version[vers] != NULL; | |
794 | return false; | |
795 | } | |
796 | ||
797 | static int | |
798 | nfsd_acl_rpcbind_set(struct net *net, const struct svc_program *progp, | |
799 | u32 version, int family, unsigned short proto, | |
800 | unsigned short port) | |
801 | { | |
802 | if (!nfsd_support_acl_version(version) || | |
e333f3bb | 803 | !nfsd_vers(net_generic(net, nfsd_net_id), version, NFSD_TEST)) |
029be5d0 TM |
804 | return 0; |
805 | return svc_generic_rpcbind_set(net, progp, version, family, | |
806 | proto, port); | |
807 | } | |
e333f3bb TM |
808 | |
809 | static __be32 | |
810 | nfsd_acl_init_request(struct svc_rqst *rqstp, | |
811 | const struct svc_program *progp, | |
812 | struct svc_process_info *ret) | |
813 | { | |
814 | struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id); | |
815 | int i; | |
816 | ||
817 | if (likely(nfsd_support_acl_version(rqstp->rq_vers) && | |
818 | nfsd_vers(nn, rqstp->rq_vers, NFSD_TEST))) | |
819 | return svc_generic_init_request(rqstp, progp, ret); | |
820 | ||
821 | ret->mismatch.lovers = NFSD_ACL_NRVERS; | |
822 | for (i = NFSD_ACL_MINVERS; i < NFSD_ACL_NRVERS; i++) { | |
823 | if (nfsd_support_acl_version(rqstp->rq_vers) && | |
824 | nfsd_vers(nn, i, NFSD_TEST)) { | |
825 | ret->mismatch.lovers = i; | |
826 | break; | |
827 | } | |
828 | } | |
829 | if (ret->mismatch.lovers == NFSD_ACL_NRVERS) | |
830 | return rpc_prog_unavail; | |
831 | ret->mismatch.hivers = NFSD_ACL_MINVERS; | |
832 | for (i = NFSD_ACL_NRVERS - 1; i >= NFSD_ACL_MINVERS; i--) { | |
833 | if (nfsd_support_acl_version(rqstp->rq_vers) && | |
834 | nfsd_vers(nn, i, NFSD_TEST)) { | |
835 | ret->mismatch.hivers = i; | |
836 | break; | |
837 | } | |
838 | } | |
839 | return rpc_prog_mismatch; | |
840 | } | |
029be5d0 TM |
841 | #endif |
842 | ||
843 | static int | |
844 | nfsd_rpcbind_set(struct net *net, const struct svc_program *progp, | |
845 | u32 version, int family, unsigned short proto, | |
846 | unsigned short port) | |
847 | { | |
e333f3bb | 848 | if (!nfsd_vers(net_generic(net, nfsd_net_id), version, NFSD_TEST)) |
029be5d0 TM |
849 | return 0; |
850 | return svc_generic_rpcbind_set(net, progp, version, family, | |
851 | proto, port); | |
852 | } | |
1da177e4 | 853 | |
e333f3bb TM |
854 | static __be32 |
855 | nfsd_init_request(struct svc_rqst *rqstp, | |
856 | const struct svc_program *progp, | |
857 | struct svc_process_info *ret) | |
858 | { | |
859 | struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id); | |
860 | int i; | |
861 | ||
862 | if (likely(nfsd_vers(nn, rqstp->rq_vers, NFSD_TEST))) | |
863 | return svc_generic_init_request(rqstp, progp, ret); | |
864 | ||
865 | ret->mismatch.lovers = NFSD_NRVERS; | |
866 | for (i = NFSD_MINVERS; i < NFSD_NRVERS; i++) { | |
867 | if (nfsd_vers(nn, i, NFSD_TEST)) { | |
868 | ret->mismatch.lovers = i; | |
869 | break; | |
870 | } | |
871 | } | |
872 | if (ret->mismatch.lovers == NFSD_NRVERS) | |
873 | return rpc_prog_unavail; | |
874 | ret->mismatch.hivers = NFSD_MINVERS; | |
875 | for (i = NFSD_NRVERS - 1; i >= NFSD_MINVERS; i--) { | |
876 | if (nfsd_vers(nn, i, NFSD_TEST)) { | |
877 | ret->mismatch.hivers = i; | |
878 | break; | |
879 | } | |
880 | } | |
881 | return rpc_prog_mismatch; | |
882 | } | |
883 | ||
1da177e4 LT |
884 | /* |
885 | * This is the NFS server kernel thread | |
886 | */ | |
9867d76c JL |
887 | static int |
888 | nfsd(void *vrqstp) | |
1da177e4 | 889 | { |
9867d76c | 890 | struct svc_rqst *rqstp = (struct svc_rqst *) vrqstp; |
88c47666 SK |
891 | struct svc_xprt *perm_sock = list_entry(rqstp->rq_server->sv_permsocks.next, typeof(struct svc_xprt), xpt_list); |
892 | struct net *net = perm_sock->xpt_net; | |
5b8db00b | 893 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); |
5b444cc9 | 894 | int err; |
1da177e4 LT |
895 | |
896 | /* Lock module and set up kernel thread */ | |
bedbdd8b | 897 | mutex_lock(&nfsd_mutex); |
1da177e4 | 898 | |
9867d76c | 899 | /* At this point, the thread shares current->fs |
47057abd AG |
900 | * with the init process. We need to create files with the |
901 | * umask as defined by the client instead of init's umask. */ | |
3e93cd67 | 902 | if (unshare_fs_struct() < 0) { |
1da177e4 LT |
903 | printk("Unable to start nfsd thread: out of memory\n"); |
904 | goto out; | |
905 | } | |
3e93cd67 | 906 | |
1da177e4 LT |
907 | current->fs->umask = 0; |
908 | ||
9867d76c JL |
909 | /* |
910 | * thread is spawned with all signals set to SIG_IGN, re-enable | |
100766f8 | 911 | * the ones that will bring down the thread |
9867d76c | 912 | */ |
100766f8 JL |
913 | allow_signal(SIGKILL); |
914 | allow_signal(SIGHUP); | |
915 | allow_signal(SIGINT); | |
916 | allow_signal(SIGQUIT); | |
bedbdd8b | 917 | |
1da177e4 | 918 | nfsdstats.th_cnt++; |
bedbdd8b NB |
919 | mutex_unlock(&nfsd_mutex); |
920 | ||
83144186 | 921 | set_freezable(); |
1da177e4 LT |
922 | |
923 | /* | |
924 | * The main request loop | |
925 | */ | |
926 | for (;;) { | |
5b8db00b JL |
927 | /* Update sv_maxconn if it has changed */ |
928 | rqstp->rq_server->sv_maxconn = nn->max_connections; | |
929 | ||
1da177e4 LT |
930 | /* |
931 | * Find a socket with data available and call its | |
932 | * recvfrom routine. | |
933 | */ | |
6fb2b47f | 934 | while ((err = svc_recv(rqstp, 60*60*HZ)) == -EAGAIN) |
1da177e4 | 935 | ; |
9867d76c | 936 | if (err == -EINTR) |
1da177e4 | 937 | break; |
e0e81739 | 938 | validate_process_creds(); |
6fb2b47f | 939 | svc_process(rqstp); |
e0e81739 | 940 | validate_process_creds(); |
1da177e4 LT |
941 | } |
942 | ||
24e36663 | 943 | /* Clear signals before calling svc_exit_thread() */ |
9e416052 | 944 | flush_signals(current); |
1da177e4 | 945 | |
bedbdd8b | 946 | mutex_lock(&nfsd_mutex); |
1da177e4 LT |
947 | nfsdstats.th_cnt --; |
948 | ||
949 | out: | |
57c8b13e | 950 | rqstp->rq_server = NULL; |
786185b5 | 951 | |
1da177e4 LT |
952 | /* Release the thread */ |
953 | svc_exit_thread(rqstp); | |
954 | ||
88c47666 | 955 | nfsd_destroy(net); |
57c8b13e | 956 | |
1da177e4 | 957 | /* Release module */ |
bedbdd8b | 958 | mutex_unlock(&nfsd_mutex); |
1da177e4 | 959 | module_put_and_exit(0); |
9867d76c | 960 | return 0; |
1da177e4 LT |
961 | } |
962 | ||
32c1eb0c AA |
963 | static __be32 map_new_errors(u32 vers, __be32 nfserr) |
964 | { | |
965 | if (nfserr == nfserr_jukebox && vers == 2) | |
966 | return nfserr_dropit; | |
967 | if (nfserr == nfserr_wrongsec && vers < 4) | |
968 | return nfserr_acces; | |
969 | return nfserr; | |
970 | } | |
971 | ||
e6838a29 BF |
972 | /* |
973 | * A write procedure can have a large argument, and a read procedure can | |
974 | * have a large reply, but no NFSv2 or NFSv3 procedure has argument and | |
975 | * reply that can both be larger than a page. The xdr code has taken | |
976 | * advantage of this assumption to be a sloppy about bounds checking in | |
977 | * some cases. Pending a rewrite of the NFSv2/v3 xdr code to fix that | |
978 | * problem, we enforce these assumptions here: | |
979 | */ | |
980 | static bool nfs_request_too_big(struct svc_rqst *rqstp, | |
860bda29 | 981 | const struct svc_procedure *proc) |
e6838a29 BF |
982 | { |
983 | /* | |
984 | * The ACL code has more careful bounds-checking and is not | |
985 | * susceptible to this problem: | |
986 | */ | |
987 | if (rqstp->rq_prog != NFS_PROGRAM) | |
988 | return false; | |
989 | /* | |
990 | * Ditto NFSv4 (which can in theory have argument and reply both | |
991 | * more than a page): | |
992 | */ | |
993 | if (rqstp->rq_vers >= 4) | |
994 | return false; | |
995 | /* The reply will be small, we're OK: */ | |
996 | if (proc->pc_xdrressize > 0 && | |
997 | proc->pc_xdrressize < XDR_QUADLEN(PAGE_SIZE)) | |
998 | return false; | |
999 | ||
1000 | return rqstp->rq_arg.len > PAGE_SIZE; | |
1001 | } | |
1002 | ||
383c440d CL |
1003 | /** |
1004 | * nfsd_dispatch - Process an NFS or NFSACL Request | |
1005 | * @rqstp: incoming request | |
1006 | * @statp: pointer to location of accept_stat field in RPC Reply buffer | |
1007 | * | |
1008 | * This RPC dispatcher integrates the NFS server's duplicate reply cache. | |
1009 | * | |
1010 | * Return values: | |
1011 | * %0: Processing complete; do not send a Reply | |
1012 | * %1: Processing complete; send Reply in rqstp->rq_res | |
1013 | */ | |
1014 | int nfsd_dispatch(struct svc_rqst *rqstp, __be32 *statp) | |
1da177e4 | 1015 | { |
4c96cb56 CL |
1016 | const struct svc_procedure *proc = rqstp->rq_procinfo; |
1017 | struct kvec *argv = &rqstp->rq_arg.head[0]; | |
1018 | struct kvec *resv = &rqstp->rq_res.head[0]; | |
1019 | __be32 nfserr, *nfserrp; | |
1da177e4 LT |
1020 | |
1021 | dprintk("nfsd_dispatch: vers %d proc %d\n", | |
1022 | rqstp->rq_vers, rqstp->rq_proc); | |
1da177e4 | 1023 | |
85085aac CL |
1024 | if (nfs_request_too_big(rqstp, proc)) |
1025 | goto out_too_large; | |
1026 | ||
1091006c BF |
1027 | /* |
1028 | * Give the xdr decoder a chance to change this if it wants | |
1029 | * (necessary in the NFSv4.0 compound case) | |
1030 | */ | |
1031 | rqstp->rq_cachetype = proc->pc_cachetype; | |
85085aac CL |
1032 | if (!proc->pc_decode(rqstp, argv->iov_base)) |
1033 | goto out_decode_err; | |
1091006c | 1034 | |
1091006c | 1035 | switch (nfsd_cache_lookup(rqstp)) { |
84c138e7 CL |
1036 | case RC_DOIT: |
1037 | break; | |
1da177e4 | 1038 | case RC_REPLY: |
85085aac | 1039 | goto out_cached_reply; |
84c138e7 | 1040 | case RC_DROPIT: |
85085aac | 1041 | goto out_dropit; |
1da177e4 LT |
1042 | } |
1043 | ||
383c440d CL |
1044 | /* |
1045 | * Need to grab the location to store the status, as | |
1046 | * NFSv4 does some encoding while processing | |
1da177e4 | 1047 | */ |
4c96cb56 CL |
1048 | nfserrp = resv->iov_base + resv->iov_len; |
1049 | resv->iov_len += sizeof(__be32); | |
1da177e4 | 1050 | |
a6beb732 | 1051 | nfserr = proc->pc_func(rqstp); |
32c1eb0c | 1052 | nfserr = map_new_errors(rqstp->rq_vers, nfserr); |
85085aac CL |
1053 | if (nfserr == nfserr_dropit || test_bit(RQ_DROPME, &rqstp->rq_flags)) |
1054 | goto out_update_drop; | |
1da177e4 LT |
1055 | |
1056 | if (rqstp->rq_proc != 0) | |
1057 | *nfserrp++ = nfserr; | |
1058 | ||
383c440d | 1059 | /* |
1da177e4 LT |
1060 | * For NFSv2, additional info is never returned in case of an error. |
1061 | */ | |
85085aac CL |
1062 | if (!(nfserr && rqstp->rq_vers == 2)) |
1063 | if (!proc->pc_encode(rqstp, nfserrp)) | |
1064 | goto out_encode_err; | |
1da177e4 | 1065 | |
57d276d7 | 1066 | nfsd_cache_update(rqstp, rqstp->rq_cachetype, statp + 1); |
85085aac CL |
1067 | out_cached_reply: |
1068 | return 1; | |
1069 | ||
1070 | out_too_large: | |
1071 | dprintk("nfsd: NFSv%d argument too large\n", rqstp->rq_vers); | |
1072 | *statp = rpc_garbage_args; | |
1073 | return 1; | |
1074 | ||
1075 | out_decode_err: | |
1076 | dprintk("nfsd: failed to decode arguments!\n"); | |
1077 | *statp = rpc_garbage_args; | |
1078 | return 1; | |
1079 | ||
1080 | out_update_drop: | |
1081 | dprintk("nfsd: Dropping request; may be revisited later\n"); | |
1082 | nfsd_cache_update(rqstp, RC_NOCACHE, NULL); | |
1083 | out_dropit: | |
1084 | return 0; | |
1085 | ||
1086 | out_encode_err: | |
1087 | dprintk("nfsd: failed to encode result!\n"); | |
1088 | nfsd_cache_update(rqstp, RC_NOCACHE, NULL); | |
1089 | *statp = rpc_system_err; | |
1da177e4 LT |
1090 | return 1; |
1091 | } | |
03cf6c9f GB |
1092 | |
1093 | int nfsd_pool_stats_open(struct inode *inode, struct file *file) | |
1094 | { | |
ed2d8aed | 1095 | int ret; |
11f77942 | 1096 | struct nfsd_net *nn = net_generic(inode->i_sb->s_fs_info, nfsd_net_id); |
9dd9845f | 1097 | |
ed2d8aed | 1098 | mutex_lock(&nfsd_mutex); |
9dd9845f | 1099 | if (nn->nfsd_serv == NULL) { |
ed2d8aed | 1100 | mutex_unlock(&nfsd_mutex); |
03cf6c9f | 1101 | return -ENODEV; |
ed2d8aed RY |
1102 | } |
1103 | /* bump up the psudo refcount while traversing */ | |
9dd9845f SK |
1104 | svc_get(nn->nfsd_serv); |
1105 | ret = svc_pool_stats_open(nn->nfsd_serv, file); | |
ed2d8aed RY |
1106 | mutex_unlock(&nfsd_mutex); |
1107 | return ret; | |
1108 | } | |
1109 | ||
1110 | int nfsd_pool_stats_release(struct inode *inode, struct file *file) | |
1111 | { | |
1112 | int ret = seq_release(inode, file); | |
11f77942 | 1113 | struct net *net = inode->i_sb->s_fs_info; |
786185b5 | 1114 | |
ed2d8aed RY |
1115 | mutex_lock(&nfsd_mutex); |
1116 | /* this function really, really should have been called svc_put() */ | |
19f7e2ca | 1117 | nfsd_destroy(net); |
ed2d8aed RY |
1118 | mutex_unlock(&nfsd_mutex); |
1119 | return ret; | |
03cf6c9f | 1120 | } |