1 /*
2 * fs/nfs/nfs4proc.c
3 *
4 * Client-side procedure declarations for NFSv4.
5 *
6 * Copyright (c) 2002 The Regents of the University of Michigan.
7 * All rights reserved.
8 *
9 * Kendrick Smith <kmsmith@umich.edu>
10 * Andy Adamson <andros@umich.edu>
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 *
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its
22 * contributors may be used to endorse or promote products derived
23 * from this software without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 */
37
38 #include <linux/mm.h>
39 #include <linux/delay.h>
40 #include <linux/errno.h>
41 #include <linux/string.h>
42 #include <linux/ratelimit.h>
43 #include <linux/printk.h>
44 #include <linux/slab.h>
45 #include <linux/sunrpc/clnt.h>
46 #include <linux/nfs.h>
47 #include <linux/nfs4.h>
48 #include <linux/nfs_fs.h>
49 #include <linux/nfs_page.h>
50 #include <linux/nfs_mount.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/module.h>
54 #include <linux/xattr.h>
55 #include <linux/utsname.h>
56 #include <linux/freezer.h>
57
58 #include "nfs4_fs.h"
59 #include "delegation.h"
60 #include "internal.h"
61 #include "iostat.h"
62 #include "callback.h"
63 #include "pnfs.h"
64 #include "netns.h"
65 #include "nfs4idmap.h"
66 #include "nfs4session.h"
67 #include "fscache.h"
68
69 #include "nfs4trace.h"
70
71 #define NFSDBG_FACILITY NFSDBG_PROC
72
73 #define NFS4_POLL_RETRY_MIN (HZ/10)
74 #define NFS4_POLL_RETRY_MAX (15*HZ)
75
76 /* file attributes which can be mapped to nfs attributes */
77 #define NFS4_VALID_ATTRS (ATTR_MODE \
78 | ATTR_UID \
79 | ATTR_GID \
80 | ATTR_SIZE \
81 | ATTR_ATIME \
82 | ATTR_MTIME \
83 | ATTR_CTIME \
84 | ATTR_ATIME_SET \
85 | ATTR_MTIME_SET)
86
87 struct nfs4_opendata;
88 static int _nfs4_proc_open(struct nfs4_opendata *data);
89 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
90 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
91 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
92 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
93 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
94 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
95 struct nfs_fattr *fattr, struct iattr *sattr,
96 struct nfs_open_context *ctx, struct nfs4_label *ilabel,
97 struct nfs4_label *olabel);
98 #ifdef CONFIG_NFS_V4_1
99 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
100 struct rpc_cred *);
101 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
102 struct rpc_cred *, bool);
103 #endif
104
105 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
106 static inline struct nfs4_label *
107 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
108 struct iattr *sattr, struct nfs4_label *label)
109 {
110 int err;
111
112 if (label == NULL)
113 return NULL;
114
115 if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
116 return NULL;
117
118 err = security_dentry_init_security(dentry, sattr->ia_mode,
119 &dentry->d_name, (void **)&label->label, &label->len);
120 if (err == 0)
121 return label;
122
123 return NULL;
124 }
125 static inline void
126 nfs4_label_release_security(struct nfs4_label *label)
127 {
128 if (label)
129 security_release_secctx(label->label, label->len);
130 }
131 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
132 {
133 if (label)
134 return server->attr_bitmask;
135
136 return server->attr_bitmask_nl;
137 }
138 #else
139 static inline struct nfs4_label *
140 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
141 struct iattr *sattr, struct nfs4_label *l)
142 { return NULL; }
143 static inline void
144 nfs4_label_release_security(struct nfs4_label *label)
145 { return; }
146 static inline u32 *
147 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
148 { return server->attr_bitmask; }
149 #endif
150
151 /* Prevent leaks of NFSv4 errors into userland */
152 static int nfs4_map_errors(int err)
153 {
154 if (err >= -1000)
155 return err;
156 switch (err) {
157 case -NFS4ERR_RESOURCE:
158 case -NFS4ERR_LAYOUTTRYLATER:
159 case -NFS4ERR_RECALLCONFLICT:
160 return -EREMOTEIO;
161 case -NFS4ERR_WRONGSEC:
162 case -NFS4ERR_WRONG_CRED:
163 return -EPERM;
164 case -NFS4ERR_BADOWNER:
165 case -NFS4ERR_BADNAME:
166 return -EINVAL;
167 case -NFS4ERR_SHARE_DENIED:
168 return -EACCES;
169 case -NFS4ERR_MINOR_VERS_MISMATCH:
170 return -EPROTONOSUPPORT;
171 case -NFS4ERR_FILE_OPEN:
172 return -EBUSY;
173 default:
174 dprintk("%s could not handle NFSv4 error %d\n",
175 __func__, -err);
176 break;
177 }
178 return -EIO;
179 }
180
181 /*
182 * This is our standard bitmap for GETATTR requests.
183 */
184 const u32 nfs4_fattr_bitmap[3] = {
185 FATTR4_WORD0_TYPE
186 | FATTR4_WORD0_CHANGE
187 | FATTR4_WORD0_SIZE
188 | FATTR4_WORD0_FSID
189 | FATTR4_WORD0_FILEID,
190 FATTR4_WORD1_MODE
191 | FATTR4_WORD1_NUMLINKS
192 | FATTR4_WORD1_OWNER
193 | FATTR4_WORD1_OWNER_GROUP
194 | FATTR4_WORD1_RAWDEV
195 | FATTR4_WORD1_SPACE_USED
196 | FATTR4_WORD1_TIME_ACCESS
197 | FATTR4_WORD1_TIME_METADATA
198 | FATTR4_WORD1_TIME_MODIFY
199 | FATTR4_WORD1_MOUNTED_ON_FILEID,
200 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
201 FATTR4_WORD2_SECURITY_LABEL
202 #endif
203 };
204
205 static const u32 nfs4_pnfs_open_bitmap[3] = {
206 FATTR4_WORD0_TYPE
207 | FATTR4_WORD0_CHANGE
208 | FATTR4_WORD0_SIZE
209 | FATTR4_WORD0_FSID
210 | FATTR4_WORD0_FILEID,
211 FATTR4_WORD1_MODE
212 | FATTR4_WORD1_NUMLINKS
213 | FATTR4_WORD1_OWNER
214 | FATTR4_WORD1_OWNER_GROUP
215 | FATTR4_WORD1_RAWDEV
216 | FATTR4_WORD1_SPACE_USED
217 | FATTR4_WORD1_TIME_ACCESS
218 | FATTR4_WORD1_TIME_METADATA
219 | FATTR4_WORD1_TIME_MODIFY,
220 FATTR4_WORD2_MDSTHRESHOLD
221 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
222 | FATTR4_WORD2_SECURITY_LABEL
223 #endif
224 };
225
226 static const u32 nfs4_open_noattr_bitmap[3] = {
227 FATTR4_WORD0_TYPE
228 | FATTR4_WORD0_FILEID,
229 };
230
231 const u32 nfs4_statfs_bitmap[3] = {
232 FATTR4_WORD0_FILES_AVAIL
233 | FATTR4_WORD0_FILES_FREE
234 | FATTR4_WORD0_FILES_TOTAL,
235 FATTR4_WORD1_SPACE_AVAIL
236 | FATTR4_WORD1_SPACE_FREE
237 | FATTR4_WORD1_SPACE_TOTAL
238 };
239
240 const u32 nfs4_pathconf_bitmap[3] = {
241 FATTR4_WORD0_MAXLINK
242 | FATTR4_WORD0_MAXNAME,
243 0
244 };
245
246 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
247 | FATTR4_WORD0_MAXREAD
248 | FATTR4_WORD0_MAXWRITE
249 | FATTR4_WORD0_LEASE_TIME,
250 FATTR4_WORD1_TIME_DELTA
251 | FATTR4_WORD1_FS_LAYOUT_TYPES,
252 FATTR4_WORD2_LAYOUT_BLKSIZE
253 | FATTR4_WORD2_CLONE_BLKSIZE
254 };
255
256 const u32 nfs4_fs_locations_bitmap[3] = {
257 FATTR4_WORD0_TYPE
258 | FATTR4_WORD0_CHANGE
259 | FATTR4_WORD0_SIZE
260 | FATTR4_WORD0_FSID
261 | FATTR4_WORD0_FILEID
262 | FATTR4_WORD0_FS_LOCATIONS,
263 FATTR4_WORD1_MODE
264 | FATTR4_WORD1_NUMLINKS
265 | FATTR4_WORD1_OWNER
266 | FATTR4_WORD1_OWNER_GROUP
267 | FATTR4_WORD1_RAWDEV
268 | FATTR4_WORD1_SPACE_USED
269 | FATTR4_WORD1_TIME_ACCESS
270 | FATTR4_WORD1_TIME_METADATA
271 | FATTR4_WORD1_TIME_MODIFY
272 | FATTR4_WORD1_MOUNTED_ON_FILEID,
273 };
274
275 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
276 struct nfs4_readdir_arg *readdir)
277 {
278 __be32 *start, *p;
279
280 if (cookie > 2) {
281 readdir->cookie = cookie;
282 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
283 return;
284 }
285
286 readdir->cookie = 0;
287 memset(&readdir->verifier, 0, sizeof(readdir->verifier));
288 if (cookie == 2)
289 return;
290
291 /*
292 * NFSv4 servers do not return entries for '.' and '..'
293 * Therefore, we fake these entries here. We let '.'
294 * have cookie 0 and '..' have cookie 1. Note that
295 * when talking to the server, we always send cookie 0
296 * instead of 1 or 2.
297 */
298 start = p = kmap_atomic(*readdir->pages);
299
300 if (cookie == 0) {
301 *p++ = xdr_one; /* next */
302 *p++ = xdr_zero; /* cookie, first word */
303 *p++ = xdr_one; /* cookie, second word */
304 *p++ = xdr_one; /* entry len */
305 memcpy(p, ".\0\0\0", 4); /* entry */
306 p++;
307 *p++ = xdr_one; /* bitmap length */
308 *p++ = htonl(FATTR4_WORD0_FILEID); /* bitmap */
309 *p++ = htonl(8); /* attribute buffer length */
310 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
311 }
312
313 *p++ = xdr_one; /* next */
314 *p++ = xdr_zero; /* cookie, first word */
315 *p++ = xdr_two; /* cookie, second word */
316 *p++ = xdr_two; /* entry len */
317 memcpy(p, "..\0\0", 4); /* entry */
318 p++;
319 *p++ = xdr_one; /* bitmap length */
320 *p++ = htonl(FATTR4_WORD0_FILEID); /* bitmap */
321 *p++ = htonl(8); /* attribute buffer length */
322 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
323
324 readdir->pgbase = (char *)p - (char *)start;
325 readdir->count -= readdir->pgbase;
326 kunmap_atomic(start);
327 }
328
329 static void nfs4_test_and_free_stateid(struct nfs_server *server,
330 nfs4_stateid *stateid,
331 struct rpc_cred *cred)
332 {
333 const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
334
335 ops->test_and_free_expired(server, stateid, cred);
336 }
337
338 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
339 nfs4_stateid *stateid,
340 struct rpc_cred *cred)
341 {
342 stateid->type = NFS4_REVOKED_STATEID_TYPE;
343 nfs4_test_and_free_stateid(server, stateid, cred);
344 }
345
346 static void nfs4_free_revoked_stateid(struct nfs_server *server,
347 const nfs4_stateid *stateid,
348 struct rpc_cred *cred)
349 {
350 nfs4_stateid tmp;
351
352 nfs4_stateid_copy(&tmp, stateid);
353 __nfs4_free_revoked_stateid(server, &tmp, cred);
354 }
355
356 static long nfs4_update_delay(long *timeout)
357 {
358 long ret;
359 if (!timeout)
360 return NFS4_POLL_RETRY_MAX;
361 if (*timeout <= 0)
362 *timeout = NFS4_POLL_RETRY_MIN;
363 if (*timeout > NFS4_POLL_RETRY_MAX)
364 *timeout = NFS4_POLL_RETRY_MAX;
365 ret = *timeout;
366 *timeout <<= 1;
367 return ret;
368 }
369
370 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
371 {
372 int res = 0;
373
374 might_sleep();
375
376 freezable_schedule_timeout_killable_unsafe(
377 nfs4_update_delay(timeout));
378 if (fatal_signal_pending(current))
379 res = -ERESTARTSYS;
380 return res;
381 }
382
383 /* This is the error handling routine for processes that are allowed
384 * to sleep.
385 */
386 static int nfs4_do_handle_exception(struct nfs_server *server,
387 int errorcode, struct nfs4_exception *exception)
388 {
389 struct nfs_client *clp = server->nfs_client;
390 struct nfs4_state *state = exception->state;
391 const nfs4_stateid *stateid = exception->stateid;
392 struct inode *inode = exception->inode;
393 int ret = errorcode;
394
395 exception->delay = 0;
396 exception->recovering = 0;
397 exception->retry = 0;
398
399 if (stateid == NULL && state != NULL)
400 stateid = &state->stateid;
401
402 switch(errorcode) {
403 case 0:
404 return 0;
405 case -NFS4ERR_DELEG_REVOKED:
406 case -NFS4ERR_ADMIN_REVOKED:
407 case -NFS4ERR_EXPIRED:
408 case -NFS4ERR_BAD_STATEID:
409 if (inode != NULL && stateid != NULL) {
410 nfs_inode_find_state_and_recover(inode,
411 stateid);
412 goto wait_on_recovery;
413 }
414 case -NFS4ERR_OPENMODE:
415 if (inode) {
416 int err;
417
418 err = nfs_async_inode_return_delegation(inode,
419 stateid);
420 if (err == 0)
421 goto wait_on_recovery;
422 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
423 exception->retry = 1;
424 break;
425 }
426 }
427 if (state == NULL)
428 break;
429 ret = nfs4_schedule_stateid_recovery(server, state);
430 if (ret < 0)
431 break;
432 goto wait_on_recovery;
433 case -NFS4ERR_STALE_STATEID:
434 case -NFS4ERR_STALE_CLIENTID:
435 nfs4_schedule_lease_recovery(clp);
436 goto wait_on_recovery;
437 case -NFS4ERR_MOVED:
438 ret = nfs4_schedule_migration_recovery(server);
439 if (ret < 0)
440 break;
441 goto wait_on_recovery;
442 case -NFS4ERR_LEASE_MOVED:
443 nfs4_schedule_lease_moved_recovery(clp);
444 goto wait_on_recovery;
445 #if defined(CONFIG_NFS_V4_1)
446 case -NFS4ERR_BADSESSION:
447 case -NFS4ERR_BADSLOT:
448 case -NFS4ERR_BAD_HIGH_SLOT:
449 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
450 case -NFS4ERR_DEADSESSION:
451 case -NFS4ERR_SEQ_FALSE_RETRY:
452 case -NFS4ERR_SEQ_MISORDERED:
453 dprintk("%s ERROR: %d Reset session\n", __func__,
454 errorcode);
455 nfs4_schedule_session_recovery(clp->cl_session, errorcode);
456 goto wait_on_recovery;
457 #endif /* defined(CONFIG_NFS_V4_1) */
458 case -NFS4ERR_FILE_OPEN:
459 if (exception->timeout > HZ) {
460 /* We have retried a decent amount, time to
461 * fail
462 */
463 ret = -EBUSY;
464 break;
465 }
466 case -NFS4ERR_DELAY:
467 nfs_inc_server_stats(server, NFSIOS_DELAY);
468 case -NFS4ERR_GRACE:
469 case -NFS4ERR_LAYOUTTRYLATER:
470 case -NFS4ERR_RECALLCONFLICT:
471 exception->delay = 1;
472 return 0;
473
474 case -NFS4ERR_RETRY_UNCACHED_REP:
475 case -NFS4ERR_OLD_STATEID:
476 exception->retry = 1;
477 break;
478 case -NFS4ERR_BADOWNER:
479 /* The following works around a Linux server bug! */
480 case -NFS4ERR_BADNAME:
481 if (server->caps & NFS_CAP_UIDGID_NOMAP) {
482 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
483 exception->retry = 1;
484 printk(KERN_WARNING "NFS: v4 server %s "
485 "does not accept raw "
486 "uid/gids. "
487 "Reenabling the idmapper.\n",
488 server->nfs_client->cl_hostname);
489 }
490 }
491 /* We failed to handle the error */
492 return nfs4_map_errors(ret);
493 wait_on_recovery:
494 exception->recovering = 1;
495 return 0;
496 }
497
498 /* This is the error handling routine for processes that are allowed
499 * to sleep.
500 */
501 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
502 {
503 struct nfs_client *clp = server->nfs_client;
504 int ret;
505
506 ret = nfs4_do_handle_exception(server, errorcode, exception);
507 if (exception->delay) {
508 ret = nfs4_delay(server->client, &exception->timeout);
509 goto out_retry;
510 }
511 if (exception->recovering) {
512 ret = nfs4_wait_clnt_recover(clp);
513 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
514 return -EIO;
515 goto out_retry;
516 }
517 return ret;
518 out_retry:
519 if (ret == 0)
520 exception->retry = 1;
521 return ret;
522 }
523
524 static int
525 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
526 int errorcode, struct nfs4_exception *exception)
527 {
528 struct nfs_client *clp = server->nfs_client;
529 int ret;
530
531 ret = nfs4_do_handle_exception(server, errorcode, exception);
532 if (exception->delay) {
533 rpc_delay(task, nfs4_update_delay(&exception->timeout));
534 goto out_retry;
535 }
536 if (exception->recovering) {
537 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
538 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
539 rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
540 goto out_retry;
541 }
542 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
543 ret = -EIO;
544 return ret;
545 out_retry:
546 if (ret == 0)
547 exception->retry = 1;
548 return ret;
549 }
550
551 static int
552 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
553 struct nfs4_state *state, long *timeout)
554 {
555 struct nfs4_exception exception = {
556 .state = state,
557 };
558
559 if (task->tk_status >= 0)
560 return 0;
561 if (timeout)
562 exception.timeout = *timeout;
563 task->tk_status = nfs4_async_handle_exception(task, server,
564 task->tk_status,
565 &exception);
566 if (exception.delay && timeout)
567 *timeout = exception.timeout;
568 if (exception.retry)
569 return -EAGAIN;
570 return 0;
571 }
572
573 /*
574 * Return 'true' if 'clp' is using an rpc_client that is integrity protected
575 * or 'false' otherwise.
576 */
577 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
578 {
579 rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
580 return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
581 }
582
583 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
584 {
585 spin_lock(&clp->cl_lock);
586 if (time_before(clp->cl_last_renewal,timestamp))
587 clp->cl_last_renewal = timestamp;
588 spin_unlock(&clp->cl_lock);
589 }
590
591 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
592 {
593 struct nfs_client *clp = server->nfs_client;
594
595 if (!nfs4_has_session(clp))
596 do_renew_lease(clp, timestamp);
597 }
598
599 struct nfs4_call_sync_data {
600 const struct nfs_server *seq_server;
601 struct nfs4_sequence_args *seq_args;
602 struct nfs4_sequence_res *seq_res;
603 };
604
605 void nfs4_init_sequence(struct nfs4_sequence_args *args,
606 struct nfs4_sequence_res *res, int cache_reply)
607 {
608 args->sa_slot = NULL;
609 args->sa_cache_this = cache_reply;
610 args->sa_privileged = 0;
611
612 res->sr_slot = NULL;
613 }
614
615 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
616 {
617 args->sa_privileged = 1;
618 }
619
620 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
621 {
622 struct nfs4_slot *slot = res->sr_slot;
623 struct nfs4_slot_table *tbl;
624
625 tbl = slot->table;
626 spin_lock(&tbl->slot_tbl_lock);
627 if (!nfs41_wake_and_assign_slot(tbl, slot))
628 nfs4_free_slot(tbl, slot);
629 spin_unlock(&tbl->slot_tbl_lock);
630
631 res->sr_slot = NULL;
632 }
633
634 static int nfs40_sequence_done(struct rpc_task *task,
635 struct nfs4_sequence_res *res)
636 {
637 if (res->sr_slot != NULL)
638 nfs40_sequence_free_slot(res);
639 return 1;
640 }
641
642 #if defined(CONFIG_NFS_V4_1)
643
644 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
645 {
646 struct nfs4_session *session;
647 struct nfs4_slot_table *tbl;
648 struct nfs4_slot *slot = res->sr_slot;
649 bool send_new_highest_used_slotid = false;
650
651 tbl = slot->table;
652 session = tbl->session;
653
654 /* Bump the slot sequence number */
655 if (slot->seq_done)
656 slot->seq_nr++;
657 slot->seq_done = 0;
658
659 spin_lock(&tbl->slot_tbl_lock);
660 /* Be nice to the server: try to ensure that the last transmitted
661 * value for highest_user_slotid <= target_highest_slotid
662 */
663 if (tbl->highest_used_slotid > tbl->target_highest_slotid)
664 send_new_highest_used_slotid = true;
665
666 if (nfs41_wake_and_assign_slot(tbl, slot)) {
667 send_new_highest_used_slotid = false;
668 goto out_unlock;
669 }
670 nfs4_free_slot(tbl, slot);
671
672 if (tbl->highest_used_slotid != NFS4_NO_SLOT)
673 send_new_highest_used_slotid = false;
674 out_unlock:
675 spin_unlock(&tbl->slot_tbl_lock);
676 res->sr_slot = NULL;
677 if (send_new_highest_used_slotid)
678 nfs41_notify_server(session->clp);
679 if (waitqueue_active(&tbl->slot_waitq))
680 wake_up_all(&tbl->slot_waitq);
681 }
682
683 static int nfs41_sequence_process(struct rpc_task *task,
684 struct nfs4_sequence_res *res)
685 {
686 struct nfs4_session *session;
687 struct nfs4_slot *slot = res->sr_slot;
688 struct nfs_client *clp;
689 bool interrupted = false;
690 int ret = 1;
691
692 if (slot == NULL)
693 goto out_noaction;
694 /* don't increment the sequence number if the task wasn't sent */
695 if (!RPC_WAS_SENT(task))
696 goto out;
697
698 session = slot->table->session;
699
700 if (slot->interrupted) {
701 slot->interrupted = 0;
702 interrupted = true;
703 }
704
705 trace_nfs4_sequence_done(session, res);
706 /* Check the SEQUENCE operation status */
707 switch (res->sr_status) {
708 case 0:
709 /* If previous op on slot was interrupted and we reused
710 * the seq# and got a reply from the cache, then retry
711 */
712 if (task->tk_status == -EREMOTEIO && interrupted) {
713 ++slot->seq_nr;
714 goto retry_nowait;
715 }
716 /* Update the slot's sequence and clientid lease timer */
717 slot->seq_done = 1;
718 clp = session->clp;
719 do_renew_lease(clp, res->sr_timestamp);
720 /* Check sequence flags */
721 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
722 !!slot->privileged);
723 nfs41_update_target_slotid(slot->table, slot, res);
724 break;
725 case 1:
726 /*
727 * sr_status remains 1 if an RPC level error occurred.
728 * The server may or may not have processed the sequence
729 * operation..
730 * Mark the slot as having hosted an interrupted RPC call.
731 */
732 slot->interrupted = 1;
733 goto out;
734 case -NFS4ERR_DELAY:
735 /* The server detected a resend of the RPC call and
736 * returned NFS4ERR_DELAY as per Section 2.10.6.2
737 * of RFC5661.
738 */
739 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
740 __func__,
741 slot->slot_nr,
742 slot->seq_nr);
743 goto out_retry;
744 case -NFS4ERR_BADSLOT:
745 /*
746 * The slot id we used was probably retired. Try again
747 * using a different slot id.
748 */
749 goto retry_nowait;
750 case -NFS4ERR_SEQ_MISORDERED:
751 /*
752 * Was the last operation on this sequence interrupted?
753 * If so, retry after bumping the sequence number.
754 */
755 if (interrupted) {
756 ++slot->seq_nr;
757 goto retry_nowait;
758 }
759 /*
760 * Could this slot have been previously retired?
761 * If so, then the server may be expecting seq_nr = 1!
762 */
763 if (slot->seq_nr != 1) {
764 slot->seq_nr = 1;
765 goto retry_nowait;
766 }
767 break;
768 case -NFS4ERR_SEQ_FALSE_RETRY:
769 ++slot->seq_nr;
770 goto retry_nowait;
771 default:
772 /* Just update the slot sequence no. */
773 slot->seq_done = 1;
774 }
775 out:
776 /* The session may be reset by one of the error handlers. */
777 dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
778 out_noaction:
779 return ret;
780 retry_nowait:
781 if (rpc_restart_call_prepare(task)) {
782 nfs41_sequence_free_slot(res);
783 task->tk_status = 0;
784 ret = 0;
785 }
786 goto out;
787 out_retry:
788 if (!rpc_restart_call(task))
789 goto out;
790 rpc_delay(task, NFS4_POLL_RETRY_MAX);
791 return 0;
792 }
793
794 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
795 {
796 if (!nfs41_sequence_process(task, res))
797 return 0;
798 if (res->sr_slot != NULL)
799 nfs41_sequence_free_slot(res);
800 return 1;
801
802 }
803 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
804
805 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
806 {
807 if (res->sr_slot == NULL)
808 return 1;
809 if (res->sr_slot->table->session != NULL)
810 return nfs41_sequence_process(task, res);
811 return nfs40_sequence_done(task, res);
812 }
813
814 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
815 {
816 if (res->sr_slot != NULL) {
817 if (res->sr_slot->table->session != NULL)
818 nfs41_sequence_free_slot(res);
819 else
820 nfs40_sequence_free_slot(res);
821 }
822 }
823
824 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
825 {
826 if (res->sr_slot == NULL)
827 return 1;
828 if (!res->sr_slot->table->session)
829 return nfs40_sequence_done(task, res);
830 return nfs41_sequence_done(task, res);
831 }
832 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
833
834 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
835 {
836 struct nfs4_call_sync_data *data = calldata;
837
838 dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
839
840 nfs4_setup_sequence(data->seq_server->nfs_client,
841 data->seq_args, data->seq_res, task);
842 }
843
844 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
845 {
846 struct nfs4_call_sync_data *data = calldata;
847
848 nfs41_sequence_done(task, data->seq_res);
849 }
850
851 static const struct rpc_call_ops nfs41_call_sync_ops = {
852 .rpc_call_prepare = nfs41_call_sync_prepare,
853 .rpc_call_done = nfs41_call_sync_done,
854 };
855
856 #else /* !CONFIG_NFS_V4_1 */
857
858 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
859 {
860 return nfs40_sequence_done(task, res);
861 }
862
863 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
864 {
865 if (res->sr_slot != NULL)
866 nfs40_sequence_free_slot(res);
867 }
868
869 int nfs4_sequence_done(struct rpc_task *task,
870 struct nfs4_sequence_res *res)
871 {
872 return nfs40_sequence_done(task, res);
873 }
874 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
875
876 #endif /* !CONFIG_NFS_V4_1 */
877
878 int nfs4_setup_sequence(const struct nfs_client *client,
879 struct nfs4_sequence_args *args,
880 struct nfs4_sequence_res *res,
881 struct rpc_task *task)
882 {
883 struct nfs4_session *session = nfs4_get_session(client);
884 struct nfs4_slot_table *tbl = client->cl_slot_tbl;
885 struct nfs4_slot *slot;
886
887 /* slot already allocated? */
888 if (res->sr_slot != NULL)
889 goto out_start;
890
891 if (session) {
892 tbl = &session->fc_slot_table;
893 task->tk_timeout = 0;
894 }
895
896 spin_lock(&tbl->slot_tbl_lock);
897 /* The state manager will wait until the slot table is empty */
898 if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
899 goto out_sleep;
900
901 slot = nfs4_alloc_slot(tbl);
902 if (IS_ERR(slot)) {
903 /* Try again in 1/4 second */
904 if (slot == ERR_PTR(-ENOMEM))
905 task->tk_timeout = HZ >> 2;
906 goto out_sleep;
907 }
908 spin_unlock(&tbl->slot_tbl_lock);
909
910 slot->privileged = args->sa_privileged ? 1 : 0;
911 args->sa_slot = slot;
912
913 res->sr_slot = slot;
914 if (session) {
915 res->sr_timestamp = jiffies;
916 res->sr_status_flags = 0;
917 res->sr_status = 1;
918 }
919
920 trace_nfs4_setup_sequence(session, args);
921 out_start:
922 rpc_call_start(task);
923 return 0;
924
925 out_sleep:
926 if (args->sa_privileged)
927 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
928 NULL, RPC_PRIORITY_PRIVILEGED);
929 else
930 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
931 spin_unlock(&tbl->slot_tbl_lock);
932 return -EAGAIN;
933 }
934 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
935
936 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
937 {
938 struct nfs4_call_sync_data *data = calldata;
939 nfs4_setup_sequence(data->seq_server->nfs_client,
940 data->seq_args, data->seq_res, task);
941 }
942
943 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
944 {
945 struct nfs4_call_sync_data *data = calldata;
946 nfs4_sequence_done(task, data->seq_res);
947 }
948
949 static const struct rpc_call_ops nfs40_call_sync_ops = {
950 .rpc_call_prepare = nfs40_call_sync_prepare,
951 .rpc_call_done = nfs40_call_sync_done,
952 };
953
954 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
955 struct nfs_server *server,
956 struct rpc_message *msg,
957 struct nfs4_sequence_args *args,
958 struct nfs4_sequence_res *res)
959 {
960 int ret;
961 struct rpc_task *task;
962 struct nfs_client *clp = server->nfs_client;
963 struct nfs4_call_sync_data data = {
964 .seq_server = server,
965 .seq_args = args,
966 .seq_res = res,
967 };
968 struct rpc_task_setup task_setup = {
969 .rpc_client = clnt,
970 .rpc_message = msg,
971 .callback_ops = clp->cl_mvops->call_sync_ops,
972 .callback_data = &data
973 };
974
975 task = rpc_run_task(&task_setup);
976 if (IS_ERR(task))
977 ret = PTR_ERR(task);
978 else {
979 ret = task->tk_status;
980 rpc_put_task(task);
981 }
982 return ret;
983 }
984
985 int nfs4_call_sync(struct rpc_clnt *clnt,
986 struct nfs_server *server,
987 struct rpc_message *msg,
988 struct nfs4_sequence_args *args,
989 struct nfs4_sequence_res *res,
990 int cache_reply)
991 {
992 nfs4_init_sequence(args, res, cache_reply);
993 return nfs4_call_sync_sequence(clnt, server, msg, args, res);
994 }
995
996 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
997 unsigned long timestamp)
998 {
999 struct nfs_inode *nfsi = NFS_I(dir);
1000
1001 spin_lock(&dir->i_lock);
1002 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1003 if (cinfo->atomic && cinfo->before == dir->i_version) {
1004 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1005 nfsi->attrtimeo_timestamp = jiffies;
1006 } else {
1007 nfs_force_lookup_revalidate(dir);
1008 if (cinfo->before != dir->i_version)
1009 nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1010 NFS_INO_INVALID_ACL;
1011 }
1012 dir->i_version = cinfo->after;
1013 nfsi->read_cache_jiffies = timestamp;
1014 nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1015 nfs_fscache_invalidate(dir);
1016 spin_unlock(&dir->i_lock);
1017 }
1018
1019 struct nfs4_opendata {
1020 struct kref kref;
1021 struct nfs_openargs o_arg;
1022 struct nfs_openres o_res;
1023 struct nfs_open_confirmargs c_arg;
1024 struct nfs_open_confirmres c_res;
1025 struct nfs4_string owner_name;
1026 struct nfs4_string group_name;
1027 struct nfs4_label *a_label;
1028 struct nfs_fattr f_attr;
1029 struct nfs4_label *f_label;
1030 struct dentry *dir;
1031 struct dentry *dentry;
1032 struct nfs4_state_owner *owner;
1033 struct nfs4_state *state;
1034 struct iattr attrs;
1035 unsigned long timestamp;
1036 unsigned int rpc_done : 1;
1037 unsigned int file_created : 1;
1038 unsigned int is_recover : 1;
1039 int rpc_status;
1040 int cancelled;
1041 };
1042
1043 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1044 int err, struct nfs4_exception *exception)
1045 {
1046 if (err != -EINVAL)
1047 return false;
1048 if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1049 return false;
1050 server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1051 exception->retry = 1;
1052 return true;
1053 }
1054
1055 static u32
1056 nfs4_map_atomic_open_share(struct nfs_server *server,
1057 fmode_t fmode, int openflags)
1058 {
1059 u32 res = 0;
1060
1061 switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1062 case FMODE_READ:
1063 res = NFS4_SHARE_ACCESS_READ;
1064 break;
1065 case FMODE_WRITE:
1066 res = NFS4_SHARE_ACCESS_WRITE;
1067 break;
1068 case FMODE_READ|FMODE_WRITE:
1069 res = NFS4_SHARE_ACCESS_BOTH;
1070 }
1071 if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1072 goto out;
1073 /* Want no delegation if we're using O_DIRECT */
1074 if (openflags & O_DIRECT)
1075 res |= NFS4_SHARE_WANT_NO_DELEG;
1076 out:
1077 return res;
1078 }
1079
1080 static enum open_claim_type4
1081 nfs4_map_atomic_open_claim(struct nfs_server *server,
1082 enum open_claim_type4 claim)
1083 {
1084 if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1085 return claim;
1086 switch (claim) {
1087 default:
1088 return claim;
1089 case NFS4_OPEN_CLAIM_FH:
1090 return NFS4_OPEN_CLAIM_NULL;
1091 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1092 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1093 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1094 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1095 }
1096 }
1097
1098 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1099 {
1100 p->o_res.f_attr = &p->f_attr;
1101 p->o_res.f_label = p->f_label;
1102 p->o_res.seqid = p->o_arg.seqid;
1103 p->c_res.seqid = p->c_arg.seqid;
1104 p->o_res.server = p->o_arg.server;
1105 p->o_res.access_request = p->o_arg.access;
1106 nfs_fattr_init(&p->f_attr);
1107 nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1108 }
1109
1110 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1111 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1112 const struct iattr *attrs,
1113 struct nfs4_label *label,
1114 enum open_claim_type4 claim,
1115 gfp_t gfp_mask)
1116 {
1117 struct dentry *parent = dget_parent(dentry);
1118 struct inode *dir = d_inode(parent);
1119 struct nfs_server *server = NFS_SERVER(dir);
1120 struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1121 struct nfs4_opendata *p;
1122
1123 p = kzalloc(sizeof(*p), gfp_mask);
1124 if (p == NULL)
1125 goto err;
1126
1127 p->f_label = nfs4_label_alloc(server, gfp_mask);
1128 if (IS_ERR(p->f_label))
1129 goto err_free_p;
1130
1131 p->a_label = nfs4_label_alloc(server, gfp_mask);
1132 if (IS_ERR(p->a_label))
1133 goto err_free_f;
1134
1135 alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1136 p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1137 if (IS_ERR(p->o_arg.seqid))
1138 goto err_free_label;
1139 nfs_sb_active(dentry->d_sb);
1140 p->dentry = dget(dentry);
1141 p->dir = parent;
1142 p->owner = sp;
1143 atomic_inc(&sp->so_count);
1144 p->o_arg.open_flags = flags;
1145 p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1146 p->o_arg.umask = current_umask();
1147 p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1148 p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1149 fmode, flags);
1150 /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1151 * will return permission denied for all bits until close */
1152 if (!(flags & O_EXCL)) {
1153 /* ask server to check for all possible rights as results
1154 * are cached */
1155 switch (p->o_arg.claim) {
1156 default:
1157 break;
1158 case NFS4_OPEN_CLAIM_NULL:
1159 case NFS4_OPEN_CLAIM_FH:
1160 p->o_arg.access = NFS4_ACCESS_READ |
1161 NFS4_ACCESS_MODIFY |
1162 NFS4_ACCESS_EXTEND |
1163 NFS4_ACCESS_EXECUTE;
1164 }
1165 }
1166 p->o_arg.clientid = server->nfs_client->cl_clientid;
1167 p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1168 p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1169 p->o_arg.name = &dentry->d_name;
1170 p->o_arg.server = server;
1171 p->o_arg.bitmask = nfs4_bitmask(server, label);
1172 p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1173 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1174 switch (p->o_arg.claim) {
1175 case NFS4_OPEN_CLAIM_NULL:
1176 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1177 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1178 p->o_arg.fh = NFS_FH(dir);
1179 break;
1180 case NFS4_OPEN_CLAIM_PREVIOUS:
1181 case NFS4_OPEN_CLAIM_FH:
1182 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1183 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1184 p->o_arg.fh = NFS_FH(d_inode(dentry));
1185 }
1186 if (attrs != NULL && attrs->ia_valid != 0) {
1187 __u32 verf[2];
1188
1189 p->o_arg.u.attrs = &p->attrs;
1190 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1191
1192 verf[0] = jiffies;
1193 verf[1] = current->pid;
1194 memcpy(p->o_arg.u.verifier.data, verf,
1195 sizeof(p->o_arg.u.verifier.data));
1196 }
1197 p->c_arg.fh = &p->o_res.fh;
1198 p->c_arg.stateid = &p->o_res.stateid;
1199 p->c_arg.seqid = p->o_arg.seqid;
1200 nfs4_init_opendata_res(p);
1201 kref_init(&p->kref);
1202 return p;
1203
1204 err_free_label:
1205 nfs4_label_free(p->a_label);
1206 err_free_f:
1207 nfs4_label_free(p->f_label);
1208 err_free_p:
1209 kfree(p);
1210 err:
1211 dput(parent);
1212 return NULL;
1213 }
1214
1215 static void nfs4_opendata_free(struct kref *kref)
1216 {
1217 struct nfs4_opendata *p = container_of(kref,
1218 struct nfs4_opendata, kref);
1219 struct super_block *sb = p->dentry->d_sb;
1220
1221 nfs_free_seqid(p->o_arg.seqid);
1222 nfs4_sequence_free_slot(&p->o_res.seq_res);
1223 if (p->state != NULL)
1224 nfs4_put_open_state(p->state);
1225 nfs4_put_state_owner(p->owner);
1226
1227 nfs4_label_free(p->a_label);
1228 nfs4_label_free(p->f_label);
1229
1230 dput(p->dir);
1231 dput(p->dentry);
1232 nfs_sb_deactive(sb);
1233 nfs_fattr_free_names(&p->f_attr);
1234 kfree(p->f_attr.mdsthreshold);
1235 kfree(p);
1236 }
1237
1238 static void nfs4_opendata_put(struct nfs4_opendata *p)
1239 {
1240 if (p != NULL)
1241 kref_put(&p->kref, nfs4_opendata_free);
1242 }
1243
1244 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1245 fmode_t fmode)
1246 {
1247 switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1248 case FMODE_READ|FMODE_WRITE:
1249 return state->n_rdwr != 0;
1250 case FMODE_WRITE:
1251 return state->n_wronly != 0;
1252 case FMODE_READ:
1253 return state->n_rdonly != 0;
1254 }
1255 WARN_ON_ONCE(1);
1256 return false;
1257 }
1258
1259 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1260 {
1261 int ret = 0;
1262
1263 if (open_mode & (O_EXCL|O_TRUNC))
1264 goto out;
1265 switch (mode & (FMODE_READ|FMODE_WRITE)) {
1266 case FMODE_READ:
1267 ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1268 && state->n_rdonly != 0;
1269 break;
1270 case FMODE_WRITE:
1271 ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1272 && state->n_wronly != 0;
1273 break;
1274 case FMODE_READ|FMODE_WRITE:
1275 ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1276 && state->n_rdwr != 0;
1277 }
1278 out:
1279 return ret;
1280 }
1281
1282 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1283 enum open_claim_type4 claim)
1284 {
1285 if (delegation == NULL)
1286 return 0;
1287 if ((delegation->type & fmode) != fmode)
1288 return 0;
1289 if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1290 return 0;
1291 switch (claim) {
1292 case NFS4_OPEN_CLAIM_NULL:
1293 case NFS4_OPEN_CLAIM_FH:
1294 break;
1295 case NFS4_OPEN_CLAIM_PREVIOUS:
1296 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1297 break;
1298 default:
1299 return 0;
1300 }
1301 nfs_mark_delegation_referenced(delegation);
1302 return 1;
1303 }
1304
1305 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1306 {
1307 switch (fmode) {
1308 case FMODE_WRITE:
1309 state->n_wronly++;
1310 break;
1311 case FMODE_READ:
1312 state->n_rdonly++;
1313 break;
1314 case FMODE_READ|FMODE_WRITE:
1315 state->n_rdwr++;
1316 }
1317 nfs4_state_set_mode_locked(state, state->state | fmode);
1318 }
1319
1320 #ifdef CONFIG_NFS_V4_1
1321 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1322 {
1323 if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1324 return true;
1325 if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1326 return true;
1327 if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1328 return true;
1329 return false;
1330 }
1331 #endif /* CONFIG_NFS_V4_1 */
1332
1333 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1334 {
1335 struct nfs_client *clp = state->owner->so_server->nfs_client;
1336 bool need_recover = false;
1337
1338 if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1339 need_recover = true;
1340 if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1341 need_recover = true;
1342 if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1343 need_recover = true;
1344 if (need_recover)
1345 nfs4_state_mark_reclaim_nograce(clp, state);
1346 }
1347
1348 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1349 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1350 {
1351 if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1352 return true;
1353 if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1354 nfs4_stateid_copy(freeme, &state->open_stateid);
1355 nfs_test_and_clear_all_open_stateid(state);
1356 return true;
1357 }
1358 if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1359 return true;
1360 return false;
1361 }
1362
1363 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1364 {
1365 if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1366 return;
1367 if (state->n_wronly)
1368 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1369 if (state->n_rdonly)
1370 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1371 if (state->n_rdwr)
1372 set_bit(NFS_O_RDWR_STATE, &state->flags);
1373 set_bit(NFS_OPEN_STATE, &state->flags);
1374 }
1375
1376 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1377 nfs4_stateid *stateid, fmode_t fmode)
1378 {
1379 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1380 switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1381 case FMODE_WRITE:
1382 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1383 break;
1384 case FMODE_READ:
1385 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1386 break;
1387 case 0:
1388 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1389 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1390 clear_bit(NFS_OPEN_STATE, &state->flags);
1391 }
1392 if (stateid == NULL)
1393 return;
1394 /* Handle OPEN+OPEN_DOWNGRADE races */
1395 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1396 !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1397 nfs_resync_open_stateid_locked(state);
1398 return;
1399 }
1400 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1401 nfs4_stateid_copy(&state->stateid, stateid);
1402 nfs4_stateid_copy(&state->open_stateid, stateid);
1403 }
1404
1405 static void nfs_clear_open_stateid(struct nfs4_state *state,
1406 nfs4_stateid *arg_stateid,
1407 nfs4_stateid *stateid, fmode_t fmode)
1408 {
1409 write_seqlock(&state->seqlock);
1410 /* Ignore, if the CLOSE argment doesn't match the current stateid */
1411 if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1412 nfs_clear_open_stateid_locked(state, stateid, fmode);
1413 write_sequnlock(&state->seqlock);
1414 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1415 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1416 }
1417
1418 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1419 const nfs4_stateid *stateid, fmode_t fmode,
1420 nfs4_stateid *freeme)
1421 {
1422 switch (fmode) {
1423 case FMODE_READ:
1424 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1425 break;
1426 case FMODE_WRITE:
1427 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1428 break;
1429 case FMODE_READ|FMODE_WRITE:
1430 set_bit(NFS_O_RDWR_STATE, &state->flags);
1431 }
1432 if (!nfs_need_update_open_stateid(state, stateid, freeme))
1433 return;
1434 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1435 nfs4_stateid_copy(&state->stateid, stateid);
1436 nfs4_stateid_copy(&state->open_stateid, stateid);
1437 }
1438
1439 static void __update_open_stateid(struct nfs4_state *state,
1440 const nfs4_stateid *open_stateid,
1441 const nfs4_stateid *deleg_stateid,
1442 fmode_t fmode,
1443 nfs4_stateid *freeme)
1444 {
1445 /*
1446 * Protect the call to nfs4_state_set_mode_locked and
1447 * serialise the stateid update
1448 */
1449 spin_lock(&state->owner->so_lock);
1450 write_seqlock(&state->seqlock);
1451 if (deleg_stateid != NULL) {
1452 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1453 set_bit(NFS_DELEGATED_STATE, &state->flags);
1454 }
1455 if (open_stateid != NULL)
1456 nfs_set_open_stateid_locked(state, open_stateid, fmode, freeme);
1457 write_sequnlock(&state->seqlock);
1458 update_open_stateflags(state, fmode);
1459 spin_unlock(&state->owner->so_lock);
1460 }
1461
1462 static int update_open_stateid(struct nfs4_state *state,
1463 const nfs4_stateid *open_stateid,
1464 const nfs4_stateid *delegation,
1465 fmode_t fmode)
1466 {
1467 struct nfs_server *server = NFS_SERVER(state->inode);
1468 struct nfs_client *clp = server->nfs_client;
1469 struct nfs_inode *nfsi = NFS_I(state->inode);
1470 struct nfs_delegation *deleg_cur;
1471 nfs4_stateid freeme = { };
1472 int ret = 0;
1473
1474 fmode &= (FMODE_READ|FMODE_WRITE);
1475
1476 rcu_read_lock();
1477 deleg_cur = rcu_dereference(nfsi->delegation);
1478 if (deleg_cur == NULL)
1479 goto no_delegation;
1480
1481 spin_lock(&deleg_cur->lock);
1482 if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1483 test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1484 (deleg_cur->type & fmode) != fmode)
1485 goto no_delegation_unlock;
1486
1487 if (delegation == NULL)
1488 delegation = &deleg_cur->stateid;
1489 else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1490 goto no_delegation_unlock;
1491
1492 nfs_mark_delegation_referenced(deleg_cur);
1493 __update_open_stateid(state, open_stateid, &deleg_cur->stateid,
1494 fmode, &freeme);
1495 ret = 1;
1496 no_delegation_unlock:
1497 spin_unlock(&deleg_cur->lock);
1498 no_delegation:
1499 rcu_read_unlock();
1500
1501 if (!ret && open_stateid != NULL) {
1502 __update_open_stateid(state, open_stateid, NULL, fmode, &freeme);
1503 ret = 1;
1504 }
1505 if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1506 nfs4_schedule_state_manager(clp);
1507 if (freeme.type != 0)
1508 nfs4_test_and_free_stateid(server, &freeme,
1509 state->owner->so_cred);
1510
1511 return ret;
1512 }
1513
1514 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1515 const nfs4_stateid *stateid)
1516 {
1517 struct nfs4_state *state = lsp->ls_state;
1518 bool ret = false;
1519
1520 spin_lock(&state->state_lock);
1521 if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1522 goto out_noupdate;
1523 if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1524 goto out_noupdate;
1525 nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1526 ret = true;
1527 out_noupdate:
1528 spin_unlock(&state->state_lock);
1529 return ret;
1530 }
1531
1532 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1533 {
1534 struct nfs_delegation *delegation;
1535
1536 rcu_read_lock();
1537 delegation = rcu_dereference(NFS_I(inode)->delegation);
1538 if (delegation == NULL || (delegation->type & fmode) == fmode) {
1539 rcu_read_unlock();
1540 return;
1541 }
1542 rcu_read_unlock();
1543 nfs4_inode_return_delegation(inode);
1544 }
1545
1546 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1547 {
1548 struct nfs4_state *state = opendata->state;
1549 struct nfs_inode *nfsi = NFS_I(state->inode);
1550 struct nfs_delegation *delegation;
1551 int open_mode = opendata->o_arg.open_flags;
1552 fmode_t fmode = opendata->o_arg.fmode;
1553 enum open_claim_type4 claim = opendata->o_arg.claim;
1554 nfs4_stateid stateid;
1555 int ret = -EAGAIN;
1556
1557 for (;;) {
1558 spin_lock(&state->owner->so_lock);
1559 if (can_open_cached(state, fmode, open_mode)) {
1560 update_open_stateflags(state, fmode);
1561 spin_unlock(&state->owner->so_lock);
1562 goto out_return_state;
1563 }
1564 spin_unlock(&state->owner->so_lock);
1565 rcu_read_lock();
1566 delegation = rcu_dereference(nfsi->delegation);
1567 if (!can_open_delegated(delegation, fmode, claim)) {
1568 rcu_read_unlock();
1569 break;
1570 }
1571 /* Save the delegation */
1572 nfs4_stateid_copy(&stateid, &delegation->stateid);
1573 rcu_read_unlock();
1574 nfs_release_seqid(opendata->o_arg.seqid);
1575 if (!opendata->is_recover) {
1576 ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1577 if (ret != 0)
1578 goto out;
1579 }
1580 ret = -EAGAIN;
1581
1582 /* Try to update the stateid using the delegation */
1583 if (update_open_stateid(state, NULL, &stateid, fmode))
1584 goto out_return_state;
1585 }
1586 out:
1587 return ERR_PTR(ret);
1588 out_return_state:
1589 atomic_inc(&state->count);
1590 return state;
1591 }
1592
1593 static void
1594 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1595 {
1596 struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1597 struct nfs_delegation *delegation;
1598 int delegation_flags = 0;
1599
1600 rcu_read_lock();
1601 delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1602 if (delegation)
1603 delegation_flags = delegation->flags;
1604 rcu_read_unlock();
1605 switch (data->o_arg.claim) {
1606 default:
1607 break;
1608 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1609 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1610 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1611 "returning a delegation for "
1612 "OPEN(CLAIM_DELEGATE_CUR)\n",
1613 clp->cl_hostname);
1614 return;
1615 }
1616 if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1617 nfs_inode_set_delegation(state->inode,
1618 data->owner->so_cred,
1619 &data->o_res);
1620 else
1621 nfs_inode_reclaim_delegation(state->inode,
1622 data->owner->so_cred,
1623 &data->o_res);
1624 }
1625
1626 /*
1627 * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1628 * and update the nfs4_state.
1629 */
1630 static struct nfs4_state *
1631 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1632 {
1633 struct inode *inode = data->state->inode;
1634 struct nfs4_state *state = data->state;
1635 int ret;
1636
1637 if (!data->rpc_done) {
1638 if (data->rpc_status)
1639 return ERR_PTR(data->rpc_status);
1640 /* cached opens have already been processed */
1641 goto update;
1642 }
1643
1644 ret = nfs_refresh_inode(inode, &data->f_attr);
1645 if (ret)
1646 return ERR_PTR(ret);
1647
1648 if (data->o_res.delegation_type != 0)
1649 nfs4_opendata_check_deleg(data, state);
1650 update:
1651 update_open_stateid(state, &data->o_res.stateid, NULL,
1652 data->o_arg.fmode);
1653 atomic_inc(&state->count);
1654
1655 return state;
1656 }
1657
1658 static struct nfs4_state *
1659 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1660 {
1661 struct inode *inode;
1662 struct nfs4_state *state = NULL;
1663 int ret;
1664
1665 if (!data->rpc_done) {
1666 state = nfs4_try_open_cached(data);
1667 trace_nfs4_cached_open(data->state);
1668 goto out;
1669 }
1670
1671 ret = -EAGAIN;
1672 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1673 goto err;
1674 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1675 ret = PTR_ERR(inode);
1676 if (IS_ERR(inode))
1677 goto err;
1678 ret = -ENOMEM;
1679 state = nfs4_get_open_state(inode, data->owner);
1680 if (state == NULL)
1681 goto err_put_inode;
1682 if (data->o_res.delegation_type != 0)
1683 nfs4_opendata_check_deleg(data, state);
1684 update_open_stateid(state, &data->o_res.stateid, NULL,
1685 data->o_arg.fmode);
1686 iput(inode);
1687 out:
1688 nfs_release_seqid(data->o_arg.seqid);
1689 return state;
1690 err_put_inode:
1691 iput(inode);
1692 err:
1693 return ERR_PTR(ret);
1694 }
1695
1696 static struct nfs4_state *
1697 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1698 {
1699 struct nfs4_state *ret;
1700
1701 if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1702 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
1703 else
1704 ret = _nfs4_opendata_to_nfs4_state(data);
1705 nfs4_sequence_free_slot(&data->o_res.seq_res);
1706 return ret;
1707 }
1708
1709 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1710 {
1711 struct nfs_inode *nfsi = NFS_I(state->inode);
1712 struct nfs_open_context *ctx;
1713
1714 spin_lock(&state->inode->i_lock);
1715 list_for_each_entry(ctx, &nfsi->open_files, list) {
1716 if (ctx->state != state)
1717 continue;
1718 get_nfs_open_context(ctx);
1719 spin_unlock(&state->inode->i_lock);
1720 return ctx;
1721 }
1722 spin_unlock(&state->inode->i_lock);
1723 return ERR_PTR(-ENOENT);
1724 }
1725
1726 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1727 struct nfs4_state *state, enum open_claim_type4 claim)
1728 {
1729 struct nfs4_opendata *opendata;
1730
1731 opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1732 NULL, NULL, claim, GFP_NOFS);
1733 if (opendata == NULL)
1734 return ERR_PTR(-ENOMEM);
1735 opendata->state = state;
1736 atomic_inc(&state->count);
1737 return opendata;
1738 }
1739
1740 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1741 fmode_t fmode)
1742 {
1743 struct nfs4_state *newstate;
1744 int ret;
1745
1746 if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1747 return 0;
1748 opendata->o_arg.open_flags = 0;
1749 opendata->o_arg.fmode = fmode;
1750 opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1751 NFS_SB(opendata->dentry->d_sb),
1752 fmode, 0);
1753 memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1754 memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1755 nfs4_init_opendata_res(opendata);
1756 ret = _nfs4_recover_proc_open(opendata);
1757 if (ret != 0)
1758 return ret;
1759 newstate = nfs4_opendata_to_nfs4_state(opendata);
1760 if (IS_ERR(newstate))
1761 return PTR_ERR(newstate);
1762 if (newstate != opendata->state)
1763 ret = -ESTALE;
1764 nfs4_close_state(newstate, fmode);
1765 return ret;
1766 }
1767
1768 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1769 {
1770 int ret;
1771
1772 /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1773 clear_bit(NFS_O_RDWR_STATE, &state->flags);
1774 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1775 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1776 /* memory barrier prior to reading state->n_* */
1777 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1778 clear_bit(NFS_OPEN_STATE, &state->flags);
1779 smp_rmb();
1780 ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1781 if (ret != 0)
1782 return ret;
1783 ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1784 if (ret != 0)
1785 return ret;
1786 ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1787 if (ret != 0)
1788 return ret;
1789 /*
1790 * We may have performed cached opens for all three recoveries.
1791 * Check if we need to update the current stateid.
1792 */
1793 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1794 !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1795 write_seqlock(&state->seqlock);
1796 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1797 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1798 write_sequnlock(&state->seqlock);
1799 }
1800 return 0;
1801 }
1802
1803 /*
1804 * OPEN_RECLAIM:
1805 * reclaim state on the server after a reboot.
1806 */
1807 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1808 {
1809 struct nfs_delegation *delegation;
1810 struct nfs4_opendata *opendata;
1811 fmode_t delegation_type = 0;
1812 int status;
1813
1814 opendata = nfs4_open_recoverdata_alloc(ctx, state,
1815 NFS4_OPEN_CLAIM_PREVIOUS);
1816 if (IS_ERR(opendata))
1817 return PTR_ERR(opendata);
1818 rcu_read_lock();
1819 delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1820 if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1821 delegation_type = delegation->type;
1822 rcu_read_unlock();
1823 opendata->o_arg.u.delegation_type = delegation_type;
1824 status = nfs4_open_recover(opendata, state);
1825 nfs4_opendata_put(opendata);
1826 return status;
1827 }
1828
1829 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1830 {
1831 struct nfs_server *server = NFS_SERVER(state->inode);
1832 struct nfs4_exception exception = { };
1833 int err;
1834 do {
1835 err = _nfs4_do_open_reclaim(ctx, state);
1836 trace_nfs4_open_reclaim(ctx, 0, err);
1837 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1838 continue;
1839 if (err != -NFS4ERR_DELAY)
1840 break;
1841 nfs4_handle_exception(server, err, &exception);
1842 } while (exception.retry);
1843 return err;
1844 }
1845
1846 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1847 {
1848 struct nfs_open_context *ctx;
1849 int ret;
1850
1851 ctx = nfs4_state_find_open_context(state);
1852 if (IS_ERR(ctx))
1853 return -EAGAIN;
1854 ret = nfs4_do_open_reclaim(ctx, state);
1855 put_nfs_open_context(ctx);
1856 return ret;
1857 }
1858
1859 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1860 {
1861 switch (err) {
1862 default:
1863 printk(KERN_ERR "NFS: %s: unhandled error "
1864 "%d.\n", __func__, err);
1865 case 0:
1866 case -ENOENT:
1867 case -EAGAIN:
1868 case -ESTALE:
1869 break;
1870 case -NFS4ERR_BADSESSION:
1871 case -NFS4ERR_BADSLOT:
1872 case -NFS4ERR_BAD_HIGH_SLOT:
1873 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1874 case -NFS4ERR_DEADSESSION:
1875 set_bit(NFS_DELEGATED_STATE, &state->flags);
1876 nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1877 return -EAGAIN;
1878 case -NFS4ERR_STALE_CLIENTID:
1879 case -NFS4ERR_STALE_STATEID:
1880 set_bit(NFS_DELEGATED_STATE, &state->flags);
1881 /* Don't recall a delegation if it was lost */
1882 nfs4_schedule_lease_recovery(server->nfs_client);
1883 return -EAGAIN;
1884 case -NFS4ERR_MOVED:
1885 nfs4_schedule_migration_recovery(server);
1886 return -EAGAIN;
1887 case -NFS4ERR_LEASE_MOVED:
1888 nfs4_schedule_lease_moved_recovery(server->nfs_client);
1889 return -EAGAIN;
1890 case -NFS4ERR_DELEG_REVOKED:
1891 case -NFS4ERR_ADMIN_REVOKED:
1892 case -NFS4ERR_EXPIRED:
1893 case -NFS4ERR_BAD_STATEID:
1894 case -NFS4ERR_OPENMODE:
1895 nfs_inode_find_state_and_recover(state->inode,
1896 stateid);
1897 nfs4_schedule_stateid_recovery(server, state);
1898 return -EAGAIN;
1899 case -NFS4ERR_DELAY:
1900 case -NFS4ERR_GRACE:
1901 set_bit(NFS_DELEGATED_STATE, &state->flags);
1902 ssleep(1);
1903 return -EAGAIN;
1904 case -ENOMEM:
1905 case -NFS4ERR_DENIED:
1906 /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1907 return 0;
1908 }
1909 return err;
1910 }
1911
1912 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1913 struct nfs4_state *state, const nfs4_stateid *stateid,
1914 fmode_t type)
1915 {
1916 struct nfs_server *server = NFS_SERVER(state->inode);
1917 struct nfs4_opendata *opendata;
1918 int err = 0;
1919
1920 opendata = nfs4_open_recoverdata_alloc(ctx, state,
1921 NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1922 if (IS_ERR(opendata))
1923 return PTR_ERR(opendata);
1924 nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1925 write_seqlock(&state->seqlock);
1926 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1927 write_sequnlock(&state->seqlock);
1928 clear_bit(NFS_DELEGATED_STATE, &state->flags);
1929 switch (type & (FMODE_READ|FMODE_WRITE)) {
1930 case FMODE_READ|FMODE_WRITE:
1931 case FMODE_WRITE:
1932 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1933 if (err)
1934 break;
1935 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1936 if (err)
1937 break;
1938 case FMODE_READ:
1939 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1940 }
1941 nfs4_opendata_put(opendata);
1942 return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1943 }
1944
1945 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1946 {
1947 struct nfs4_opendata *data = calldata;
1948
1949 nfs4_setup_sequence(data->o_arg.server->nfs_client,
1950 &data->c_arg.seq_args, &data->c_res.seq_res, task);
1951 }
1952
1953 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1954 {
1955 struct nfs4_opendata *data = calldata;
1956
1957 nfs40_sequence_done(task, &data->c_res.seq_res);
1958
1959 data->rpc_status = task->tk_status;
1960 if (data->rpc_status == 0) {
1961 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1962 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1963 renew_lease(data->o_res.server, data->timestamp);
1964 data->rpc_done = 1;
1965 }
1966 }
1967
1968 static void nfs4_open_confirm_release(void *calldata)
1969 {
1970 struct nfs4_opendata *data = calldata;
1971 struct nfs4_state *state = NULL;
1972
1973 /* If this request hasn't been cancelled, do nothing */
1974 if (data->cancelled == 0)
1975 goto out_free;
1976 /* In case of error, no cleanup! */
1977 if (!data->rpc_done)
1978 goto out_free;
1979 state = nfs4_opendata_to_nfs4_state(data);
1980 if (!IS_ERR(state))
1981 nfs4_close_state(state, data->o_arg.fmode);
1982 out_free:
1983 nfs4_opendata_put(data);
1984 }
1985
1986 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1987 .rpc_call_prepare = nfs4_open_confirm_prepare,
1988 .rpc_call_done = nfs4_open_confirm_done,
1989 .rpc_release = nfs4_open_confirm_release,
1990 };
1991
1992 /*
1993 * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1994 */
1995 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1996 {
1997 struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1998 struct rpc_task *task;
1999 struct rpc_message msg = {
2000 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2001 .rpc_argp = &data->c_arg,
2002 .rpc_resp = &data->c_res,
2003 .rpc_cred = data->owner->so_cred,
2004 };
2005 struct rpc_task_setup task_setup_data = {
2006 .rpc_client = server->client,
2007 .rpc_message = &msg,
2008 .callback_ops = &nfs4_open_confirm_ops,
2009 .callback_data = data,
2010 .workqueue = nfsiod_workqueue,
2011 .flags = RPC_TASK_ASYNC,
2012 };
2013 int status;
2014
2015 nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
2016 kref_get(&data->kref);
2017 data->rpc_done = 0;
2018 data->rpc_status = 0;
2019 data->timestamp = jiffies;
2020 if (data->is_recover)
2021 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
2022 task = rpc_run_task(&task_setup_data);
2023 if (IS_ERR(task))
2024 return PTR_ERR(task);
2025 status = rpc_wait_for_completion_task(task);
2026 if (status != 0) {
2027 data->cancelled = 1;
2028 smp_wmb();
2029 } else
2030 status = data->rpc_status;
2031 rpc_put_task(task);
2032 return status;
2033 }
2034
2035 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2036 {
2037 struct nfs4_opendata *data = calldata;
2038 struct nfs4_state_owner *sp = data->owner;
2039 struct nfs_client *clp = sp->so_server->nfs_client;
2040 enum open_claim_type4 claim = data->o_arg.claim;
2041
2042 if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2043 goto out_wait;
2044 /*
2045 * Check if we still need to send an OPEN call, or if we can use
2046 * a delegation instead.
2047 */
2048 if (data->state != NULL) {
2049 struct nfs_delegation *delegation;
2050
2051 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
2052 goto out_no_action;
2053 rcu_read_lock();
2054 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
2055 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2056 goto unlock_no_action;
2057 rcu_read_unlock();
2058 }
2059 /* Update client id. */
2060 data->o_arg.clientid = clp->cl_clientid;
2061 switch (claim) {
2062 default:
2063 break;
2064 case NFS4_OPEN_CLAIM_PREVIOUS:
2065 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2066 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2067 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2068 case NFS4_OPEN_CLAIM_FH:
2069 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2070 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2071 }
2072 data->timestamp = jiffies;
2073 if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2074 &data->o_arg.seq_args,
2075 &data->o_res.seq_res,
2076 task) != 0)
2077 nfs_release_seqid(data->o_arg.seqid);
2078
2079 /* Set the create mode (note dependency on the session type) */
2080 data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2081 if (data->o_arg.open_flags & O_EXCL) {
2082 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2083 if (nfs4_has_persistent_session(clp))
2084 data->o_arg.createmode = NFS4_CREATE_GUARDED;
2085 else if (clp->cl_mvops->minor_version > 0)
2086 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2087 }
2088 return;
2089 unlock_no_action:
2090 trace_nfs4_cached_open(data->state);
2091 rcu_read_unlock();
2092 out_no_action:
2093 task->tk_action = NULL;
2094 out_wait:
2095 nfs4_sequence_done(task, &data->o_res.seq_res);
2096 }
2097
2098 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2099 {
2100 struct nfs4_opendata *data = calldata;
2101
2102 data->rpc_status = task->tk_status;
2103
2104 if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2105 return;
2106
2107 if (task->tk_status == 0) {
2108 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2109 switch (data->o_res.f_attr->mode & S_IFMT) {
2110 case S_IFREG:
2111 break;
2112 case S_IFLNK:
2113 data->rpc_status = -ELOOP;
2114 break;
2115 case S_IFDIR:
2116 data->rpc_status = -EISDIR;
2117 break;
2118 default:
2119 data->rpc_status = -ENOTDIR;
2120 }
2121 }
2122 renew_lease(data->o_res.server, data->timestamp);
2123 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2124 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2125 }
2126 data->rpc_done = 1;
2127 }
2128
2129 static void nfs4_open_release(void *calldata)
2130 {
2131 struct nfs4_opendata *data = calldata;
2132 struct nfs4_state *state = NULL;
2133
2134 /* If this request hasn't been cancelled, do nothing */
2135 if (data->cancelled == 0)
2136 goto out_free;
2137 /* In case of error, no cleanup! */
2138 if (data->rpc_status != 0 || !data->rpc_done)
2139 goto out_free;
2140 /* In case we need an open_confirm, no cleanup! */
2141 if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2142 goto out_free;
2143 state = nfs4_opendata_to_nfs4_state(data);
2144 if (!IS_ERR(state))
2145 nfs4_close_state(state, data->o_arg.fmode);
2146 out_free:
2147 nfs4_opendata_put(data);
2148 }
2149
2150 static const struct rpc_call_ops nfs4_open_ops = {
2151 .rpc_call_prepare = nfs4_open_prepare,
2152 .rpc_call_done = nfs4_open_done,
2153 .rpc_release = nfs4_open_release,
2154 };
2155
2156 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2157 {
2158 struct inode *dir = d_inode(data->dir);
2159 struct nfs_server *server = NFS_SERVER(dir);
2160 struct nfs_openargs *o_arg = &data->o_arg;
2161 struct nfs_openres *o_res = &data->o_res;
2162 struct rpc_task *task;
2163 struct rpc_message msg = {
2164 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2165 .rpc_argp = o_arg,
2166 .rpc_resp = o_res,
2167 .rpc_cred = data->owner->so_cred,
2168 };
2169 struct rpc_task_setup task_setup_data = {
2170 .rpc_client = server->client,
2171 .rpc_message = &msg,
2172 .callback_ops = &nfs4_open_ops,
2173 .callback_data = data,
2174 .workqueue = nfsiod_workqueue,
2175 .flags = RPC_TASK_ASYNC,
2176 };
2177 int status;
2178
2179 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2180 kref_get(&data->kref);
2181 data->rpc_done = 0;
2182 data->rpc_status = 0;
2183 data->cancelled = 0;
2184 data->is_recover = 0;
2185 if (isrecover) {
2186 nfs4_set_sequence_privileged(&o_arg->seq_args);
2187 data->is_recover = 1;
2188 }
2189 task = rpc_run_task(&task_setup_data);
2190 if (IS_ERR(task))
2191 return PTR_ERR(task);
2192 status = rpc_wait_for_completion_task(task);
2193 if (status != 0) {
2194 data->cancelled = 1;
2195 smp_wmb();
2196 } else
2197 status = data->rpc_status;
2198 rpc_put_task(task);
2199
2200 return status;
2201 }
2202
2203 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2204 {
2205 struct inode *dir = d_inode(data->dir);
2206 struct nfs_openres *o_res = &data->o_res;
2207 int status;
2208
2209 status = nfs4_run_open_task(data, 1);
2210 if (status != 0 || !data->rpc_done)
2211 return status;
2212
2213 nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2214
2215 if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2216 status = _nfs4_proc_open_confirm(data);
2217
2218 return status;
2219 }
2220
2221 /*
2222 * Additional permission checks in order to distinguish between an
2223 * open for read, and an open for execute. This works around the
2224 * fact that NFSv4 OPEN treats read and execute permissions as being
2225 * the same.
2226 * Note that in the non-execute case, we want to turn off permission
2227 * checking if we just created a new file (POSIX open() semantics).
2228 */
2229 static int nfs4_opendata_access(struct rpc_cred *cred,
2230 struct nfs4_opendata *opendata,
2231 struct nfs4_state *state, fmode_t fmode,
2232 int openflags)
2233 {
2234 struct nfs_access_entry cache;
2235 u32 mask;
2236
2237 /* access call failed or for some reason the server doesn't
2238 * support any access modes -- defer access call until later */
2239 if (opendata->o_res.access_supported == 0)
2240 return 0;
2241
2242 mask = 0;
2243 /*
2244 * Use openflags to check for exec, because fmode won't
2245 * always have FMODE_EXEC set when file open for exec.
2246 */
2247 if (openflags & __FMODE_EXEC) {
2248 /* ONLY check for exec rights */
2249 mask = MAY_EXEC;
2250 } else if ((fmode & FMODE_READ) && !opendata->file_created)
2251 mask = MAY_READ;
2252
2253 cache.cred = cred;
2254 cache.jiffies = jiffies;
2255 nfs_access_set_mask(&cache, opendata->o_res.access_result);
2256 nfs_access_add_cache(state->inode, &cache);
2257
2258 if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2259 return 0;
2260
2261 return -EACCES;
2262 }
2263
2264 /*
2265 * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2266 */
2267 static int _nfs4_proc_open(struct nfs4_opendata *data)
2268 {
2269 struct inode *dir = d_inode(data->dir);
2270 struct nfs_server *server = NFS_SERVER(dir);
2271 struct nfs_openargs *o_arg = &data->o_arg;
2272 struct nfs_openres *o_res = &data->o_res;
2273 int status;
2274
2275 status = nfs4_run_open_task(data, 0);
2276 if (!data->rpc_done)
2277 return status;
2278 if (status != 0) {
2279 if (status == -NFS4ERR_BADNAME &&
2280 !(o_arg->open_flags & O_CREAT))
2281 return -ENOENT;
2282 return status;
2283 }
2284
2285 nfs_fattr_map_and_free_names(server, &data->f_attr);
2286
2287 if (o_arg->open_flags & O_CREAT) {
2288 if (o_arg->open_flags & O_EXCL)
2289 data->file_created = 1;
2290 else if (o_res->cinfo.before != o_res->cinfo.after)
2291 data->file_created = 1;
2292 if (data->file_created || dir->i_version != o_res->cinfo.after)
2293 update_changeattr(dir, &o_res->cinfo,
2294 o_res->f_attr->time_start);
2295 }
2296 if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2297 server->caps &= ~NFS_CAP_POSIX_LOCK;
2298 if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2299 status = _nfs4_proc_open_confirm(data);
2300 if (status != 0)
2301 return status;
2302 }
2303 if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2304 nfs4_sequence_free_slot(&o_res->seq_res);
2305 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2306 }
2307 return 0;
2308 }
2309
2310 /*
2311 * OPEN_EXPIRED:
2312 * reclaim state on the server after a network partition.
2313 * Assumes caller holds the appropriate lock
2314 */
2315 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2316 {
2317 struct nfs4_opendata *opendata;
2318 int ret;
2319
2320 opendata = nfs4_open_recoverdata_alloc(ctx, state,
2321 NFS4_OPEN_CLAIM_FH);
2322 if (IS_ERR(opendata))
2323 return PTR_ERR(opendata);
2324 ret = nfs4_open_recover(opendata, state);
2325 if (ret == -ESTALE)
2326 d_drop(ctx->dentry);
2327 nfs4_opendata_put(opendata);
2328 return ret;
2329 }
2330
2331 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2332 {
2333 struct nfs_server *server = NFS_SERVER(state->inode);
2334 struct nfs4_exception exception = { };
2335 int err;
2336
2337 do {
2338 err = _nfs4_open_expired(ctx, state);
2339 trace_nfs4_open_expired(ctx, 0, err);
2340 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2341 continue;
2342 switch (err) {
2343 default:
2344 goto out;
2345 case -NFS4ERR_GRACE:
2346 case -NFS4ERR_DELAY:
2347 nfs4_handle_exception(server, err, &exception);
2348 err = 0;
2349 }
2350 } while (exception.retry);
2351 out:
2352 return err;
2353 }
2354
2355 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2356 {
2357 struct nfs_open_context *ctx;
2358 int ret;
2359
2360 ctx = nfs4_state_find_open_context(state);
2361 if (IS_ERR(ctx))
2362 return -EAGAIN;
2363 ret = nfs4_do_open_expired(ctx, state);
2364 put_nfs_open_context(ctx);
2365 return ret;
2366 }
2367
2368 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2369 const nfs4_stateid *stateid)
2370 {
2371 nfs_remove_bad_delegation(state->inode, stateid);
2372 write_seqlock(&state->seqlock);
2373 nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2374 write_sequnlock(&state->seqlock);
2375 clear_bit(NFS_DELEGATED_STATE, &state->flags);
2376 }
2377
2378 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2379 {
2380 if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2381 nfs_finish_clear_delegation_stateid(state, NULL);
2382 }
2383
2384 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2385 {
2386 /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2387 nfs40_clear_delegation_stateid(state);
2388 return nfs4_open_expired(sp, state);
2389 }
2390
2391 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2392 nfs4_stateid *stateid,
2393 struct rpc_cred *cred)
2394 {
2395 return -NFS4ERR_BAD_STATEID;
2396 }
2397
2398 #if defined(CONFIG_NFS_V4_1)
2399 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2400 nfs4_stateid *stateid,
2401 struct rpc_cred *cred)
2402 {
2403 int status;
2404
2405 switch (stateid->type) {
2406 default:
2407 break;
2408 case NFS4_INVALID_STATEID_TYPE:
2409 case NFS4_SPECIAL_STATEID_TYPE:
2410 return -NFS4ERR_BAD_STATEID;
2411 case NFS4_REVOKED_STATEID_TYPE:
2412 goto out_free;
2413 }
2414
2415 status = nfs41_test_stateid(server, stateid, cred);
2416 switch (status) {
2417 case -NFS4ERR_EXPIRED:
2418 case -NFS4ERR_ADMIN_REVOKED:
2419 case -NFS4ERR_DELEG_REVOKED:
2420 break;
2421 default:
2422 return status;
2423 }
2424 out_free:
2425 /* Ack the revoked state to the server */
2426 nfs41_free_stateid(server, stateid, cred, true);
2427 return -NFS4ERR_EXPIRED;
2428 }
2429
2430 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2431 {
2432 struct nfs_server *server = NFS_SERVER(state->inode);
2433 nfs4_stateid stateid;
2434 struct nfs_delegation *delegation;
2435 struct rpc_cred *cred;
2436 int status;
2437
2438 /* Get the delegation credential for use by test/free_stateid */
2439 rcu_read_lock();
2440 delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2441 if (delegation == NULL) {
2442 rcu_read_unlock();
2443 return;
2444 }
2445
2446 nfs4_stateid_copy(&stateid, &delegation->stateid);
2447 if (test_bit(NFS_DELEGATION_REVOKED, &delegation->flags) ||
2448 !test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2449 &delegation->flags)) {
2450 rcu_read_unlock();
2451 nfs_finish_clear_delegation_stateid(state, &stateid);
2452 return;
2453 }
2454
2455 cred = get_rpccred(delegation->cred);
2456 rcu_read_unlock();
2457 status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2458 trace_nfs4_test_delegation_stateid(state, NULL, status);
2459 if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2460 nfs_finish_clear_delegation_stateid(state, &stateid);
2461
2462 put_rpccred(cred);
2463 }
2464
2465 /**
2466 * nfs41_check_expired_locks - possibly free a lock stateid
2467 *
2468 * @state: NFSv4 state for an inode
2469 *
2470 * Returns NFS_OK if recovery for this stateid is now finished.
2471 * Otherwise a negative NFS4ERR value is returned.
2472 */
2473 static int nfs41_check_expired_locks(struct nfs4_state *state)
2474 {
2475 int status, ret = NFS_OK;
2476 struct nfs4_lock_state *lsp, *prev = NULL;
2477 struct nfs_server *server = NFS_SERVER(state->inode);
2478
2479 if (!test_bit(LK_STATE_IN_USE, &state->flags))
2480 goto out;
2481
2482 spin_lock(&state->state_lock);
2483 list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2484 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2485 struct rpc_cred *cred = lsp->ls_state->owner->so_cred;
2486
2487 atomic_inc(&lsp->ls_count);
2488 spin_unlock(&state->state_lock);
2489
2490 nfs4_put_lock_state(prev);
2491 prev = lsp;
2492
2493 status = nfs41_test_and_free_expired_stateid(server,
2494 &lsp->ls_stateid,
2495 cred);
2496 trace_nfs4_test_lock_stateid(state, lsp, status);
2497 if (status == -NFS4ERR_EXPIRED ||
2498 status == -NFS4ERR_BAD_STATEID) {
2499 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2500 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2501 if (!recover_lost_locks)
2502 set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2503 } else if (status != NFS_OK) {
2504 ret = status;
2505 nfs4_put_lock_state(prev);
2506 goto out;
2507 }
2508 spin_lock(&state->state_lock);
2509 }
2510 }
2511 spin_unlock(&state->state_lock);
2512 nfs4_put_lock_state(prev);
2513 out:
2514 return ret;
2515 }
2516
2517 /**
2518 * nfs41_check_open_stateid - possibly free an open stateid
2519 *
2520 * @state: NFSv4 state for an inode
2521 *
2522 * Returns NFS_OK if recovery for this stateid is now finished.
2523 * Otherwise a negative NFS4ERR value is returned.
2524 */
2525 static int nfs41_check_open_stateid(struct nfs4_state *state)
2526 {
2527 struct nfs_server *server = NFS_SERVER(state->inode);
2528 nfs4_stateid *stateid = &state->open_stateid;
2529 struct rpc_cred *cred = state->owner->so_cred;
2530 int status;
2531
2532 if (test_bit(NFS_OPEN_STATE, &state->flags) == 0) {
2533 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0) {
2534 if (nfs4_have_delegation(state->inode, state->state))
2535 return NFS_OK;
2536 return -NFS4ERR_OPENMODE;
2537 }
2538 return -NFS4ERR_BAD_STATEID;
2539 }
2540 status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2541 trace_nfs4_test_open_stateid(state, NULL, status);
2542 if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2543 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2544 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2545 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2546 clear_bit(NFS_OPEN_STATE, &state->flags);
2547 stateid->type = NFS4_INVALID_STATEID_TYPE;
2548 }
2549 if (status != NFS_OK)
2550 return status;
2551 if (nfs_open_stateid_recover_openmode(state))
2552 return -NFS4ERR_OPENMODE;
2553 return NFS_OK;
2554 }
2555
2556 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2557 {
2558 int status;
2559
2560 nfs41_check_delegation_stateid(state);
2561 status = nfs41_check_expired_locks(state);
2562 if (status != NFS_OK)
2563 return status;
2564 status = nfs41_check_open_stateid(state);
2565 if (status != NFS_OK)
2566 status = nfs4_open_expired(sp, state);
2567 return status;
2568 }
2569 #endif
2570
2571 /*
2572 * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2573 * fields corresponding to attributes that were used to store the verifier.
2574 * Make sure we clobber those fields in the later setattr call
2575 */
2576 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2577 struct iattr *sattr, struct nfs4_label **label)
2578 {
2579 const u32 *attrset = opendata->o_res.attrset;
2580
2581 if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2582 !(sattr->ia_valid & ATTR_ATIME_SET))
2583 sattr->ia_valid |= ATTR_ATIME;
2584
2585 if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2586 !(sattr->ia_valid & ATTR_MTIME_SET))
2587 sattr->ia_valid |= ATTR_MTIME;
2588
2589 /* Except MODE, it seems harmless of setting twice. */
2590 if (opendata->o_arg.createmode != NFS4_CREATE_EXCLUSIVE &&
2591 (attrset[1] & FATTR4_WORD1_MODE ||
2592 attrset[2] & FATTR4_WORD2_MODE_UMASK))
2593 sattr->ia_valid &= ~ATTR_MODE;
2594
2595 if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2596 *label = NULL;
2597 }
2598
2599 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2600 fmode_t fmode,
2601 int flags,
2602 struct nfs_open_context *ctx)
2603 {
2604 struct nfs4_state_owner *sp = opendata->owner;
2605 struct nfs_server *server = sp->so_server;
2606 struct dentry *dentry;
2607 struct nfs4_state *state;
2608 unsigned int seq;
2609 int ret;
2610
2611 seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2612
2613 ret = _nfs4_proc_open(opendata);
2614 if (ret != 0)
2615 goto out;
2616
2617 state = nfs4_opendata_to_nfs4_state(opendata);
2618 ret = PTR_ERR(state);
2619 if (IS_ERR(state))
2620 goto out;
2621 ctx->state = state;
2622 if (server->caps & NFS_CAP_POSIX_LOCK)
2623 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2624 if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
2625 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
2626
2627 dentry = opendata->dentry;
2628 if (d_really_is_negative(dentry)) {
2629 struct dentry *alias;
2630 d_drop(dentry);
2631 alias = d_exact_alias(dentry, state->inode);
2632 if (!alias)
2633 alias = d_splice_alias(igrab(state->inode), dentry);
2634 /* d_splice_alias() can't fail here - it's a non-directory */
2635 if (alias) {
2636 dput(ctx->dentry);
2637 ctx->dentry = dentry = alias;
2638 }
2639 nfs_set_verifier(dentry,
2640 nfs_save_change_attribute(d_inode(opendata->dir)));
2641 }
2642
2643 ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2644 if (ret != 0)
2645 goto out;
2646
2647 if (d_inode(dentry) == state->inode) {
2648 nfs_inode_attach_open_context(ctx);
2649 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2650 nfs4_schedule_stateid_recovery(server, state);
2651 }
2652 out:
2653 return ret;
2654 }
2655
2656 /*
2657 * Returns a referenced nfs4_state
2658 */
2659 static int _nfs4_do_open(struct inode *dir,
2660 struct nfs_open_context *ctx,
2661 int flags,
2662 struct iattr *sattr,
2663 struct nfs4_label *label,
2664 int *opened)
2665 {
2666 struct nfs4_state_owner *sp;
2667 struct nfs4_state *state = NULL;
2668 struct nfs_server *server = NFS_SERVER(dir);
2669 struct nfs4_opendata *opendata;
2670 struct dentry *dentry = ctx->dentry;
2671 struct rpc_cred *cred = ctx->cred;
2672 struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2673 fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2674 enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2675 struct nfs4_label *olabel = NULL;
2676 int status;
2677
2678 /* Protect against reboot recovery conflicts */
2679 status = -ENOMEM;
2680 sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2681 if (sp == NULL) {
2682 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2683 goto out_err;
2684 }
2685 status = nfs4_client_recover_expired_lease(server->nfs_client);
2686 if (status != 0)
2687 goto err_put_state_owner;
2688 if (d_really_is_positive(dentry))
2689 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2690 status = -ENOMEM;
2691 if (d_really_is_positive(dentry))
2692 claim = NFS4_OPEN_CLAIM_FH;
2693 opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2694 label, claim, GFP_KERNEL);
2695 if (opendata == NULL)
2696 goto err_put_state_owner;
2697
2698 if (label) {
2699 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2700 if (IS_ERR(olabel)) {
2701 status = PTR_ERR(olabel);
2702 goto err_opendata_put;
2703 }
2704 }
2705
2706 if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2707 if (!opendata->f_attr.mdsthreshold) {
2708 opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2709 if (!opendata->f_attr.mdsthreshold)
2710 goto err_free_label;
2711 }
2712 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2713 }
2714 if (d_really_is_positive(dentry))
2715 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2716
2717 status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2718 if (status != 0)
2719 goto err_free_label;
2720 state = ctx->state;
2721
2722 if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2723 (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2724 nfs4_exclusive_attrset(opendata, sattr, &label);
2725 /*
2726 * send create attributes which was not set by open
2727 * with an extra setattr.
2728 */
2729 if (sattr->ia_valid & NFS4_VALID_ATTRS) {
2730 nfs_fattr_init(opendata->o_res.f_attr);
2731 status = nfs4_do_setattr(state->inode, cred,
2732 opendata->o_res.f_attr, sattr,
2733 ctx, label, olabel);
2734 if (status == 0) {
2735 nfs_setattr_update_inode(state->inode, sattr,
2736 opendata->o_res.f_attr);
2737 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2738 }
2739 }
2740 }
2741 if (opened && opendata->file_created)
2742 *opened |= FILE_CREATED;
2743
2744 if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2745 *ctx_th = opendata->f_attr.mdsthreshold;
2746 opendata->f_attr.mdsthreshold = NULL;
2747 }
2748
2749 nfs4_label_free(olabel);
2750
2751 nfs4_opendata_put(opendata);
2752 nfs4_put_state_owner(sp);
2753 return 0;
2754 err_free_label:
2755 nfs4_label_free(olabel);
2756 err_opendata_put:
2757 nfs4_opendata_put(opendata);
2758 err_put_state_owner:
2759 nfs4_put_state_owner(sp);
2760 out_err:
2761 return status;
2762 }
2763
2764
2765 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2766 struct nfs_open_context *ctx,
2767 int flags,
2768 struct iattr *sattr,
2769 struct nfs4_label *label,
2770 int *opened)
2771 {
2772 struct nfs_server *server = NFS_SERVER(dir);
2773 struct nfs4_exception exception = { };
2774 struct nfs4_state *res;
2775 int status;
2776
2777 do {
2778 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2779 res = ctx->state;
2780 trace_nfs4_open_file(ctx, flags, status);
2781 if (status == 0)
2782 break;
2783 /* NOTE: BAD_SEQID means the server and client disagree about the
2784 * book-keeping w.r.t. state-changing operations
2785 * (OPEN/CLOSE/LOCK/LOCKU...)
2786 * It is actually a sign of a bug on the client or on the server.
2787 *
2788 * If we receive a BAD_SEQID error in the particular case of
2789 * doing an OPEN, we assume that nfs_increment_open_seqid() will
2790 * have unhashed the old state_owner for us, and that we can
2791 * therefore safely retry using a new one. We should still warn
2792 * the user though...
2793 */
2794 if (status == -NFS4ERR_BAD_SEQID) {
2795 pr_warn_ratelimited("NFS: v4 server %s "
2796 " returned a bad sequence-id error!\n",
2797 NFS_SERVER(dir)->nfs_client->cl_hostname);
2798 exception.retry = 1;
2799 continue;
2800 }
2801 /*
2802 * BAD_STATEID on OPEN means that the server cancelled our
2803 * state before it received the OPEN_CONFIRM.
2804 * Recover by retrying the request as per the discussion
2805 * on Page 181 of RFC3530.
2806 */
2807 if (status == -NFS4ERR_BAD_STATEID) {
2808 exception.retry = 1;
2809 continue;
2810 }
2811 if (status == -EAGAIN) {
2812 /* We must have found a delegation */
2813 exception.retry = 1;
2814 continue;
2815 }
2816 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2817 continue;
2818 res = ERR_PTR(nfs4_handle_exception(server,
2819 status, &exception));
2820 } while (exception.retry);
2821 return res;
2822 }
2823
2824 static int _nfs4_do_setattr(struct inode *inode,
2825 struct nfs_setattrargs *arg,
2826 struct nfs_setattrres *res,
2827 struct rpc_cred *cred,
2828 struct nfs_open_context *ctx)
2829 {
2830 struct nfs_server *server = NFS_SERVER(inode);
2831 struct rpc_message msg = {
2832 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2833 .rpc_argp = arg,
2834 .rpc_resp = res,
2835 .rpc_cred = cred,
2836 };
2837 struct rpc_cred *delegation_cred = NULL;
2838 unsigned long timestamp = jiffies;
2839 fmode_t fmode;
2840 bool truncate;
2841 int status;
2842
2843 nfs_fattr_init(res->fattr);
2844
2845 /* Servers should only apply open mode checks for file size changes */
2846 truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
2847 fmode = truncate ? FMODE_WRITE : FMODE_READ;
2848
2849 if (nfs4_copy_delegation_stateid(inode, fmode, &arg->stateid, &delegation_cred)) {
2850 /* Use that stateid */
2851 } else if (truncate && ctx != NULL) {
2852 struct nfs_lock_context *l_ctx;
2853 if (!nfs4_valid_open_stateid(ctx->state))
2854 return -EBADF;
2855 l_ctx = nfs_get_lock_context(ctx);
2856 if (IS_ERR(l_ctx))
2857 return PTR_ERR(l_ctx);
2858 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
2859 &arg->stateid, &delegation_cred);
2860 nfs_put_lock_context(l_ctx);
2861 if (status == -EIO)
2862 return -EBADF;
2863 } else
2864 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
2865 if (delegation_cred)
2866 msg.rpc_cred = delegation_cred;
2867
2868 status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
2869
2870 put_rpccred(delegation_cred);
2871 if (status == 0 && ctx != NULL)
2872 renew_lease(server, timestamp);
2873 trace_nfs4_setattr(inode, &arg->stateid, status);
2874 return status;
2875 }
2876
2877 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2878 struct nfs_fattr *fattr, struct iattr *sattr,
2879 struct nfs_open_context *ctx, struct nfs4_label *ilabel,
2880 struct nfs4_label *olabel)
2881 {
2882 struct nfs_server *server = NFS_SERVER(inode);
2883 struct nfs4_state *state = ctx ? ctx->state : NULL;
2884 struct nfs_setattrargs arg = {
2885 .fh = NFS_FH(inode),
2886 .iap = sattr,
2887 .server = server,
2888 .bitmask = server->attr_bitmask,
2889 .label = ilabel,
2890 };
2891 struct nfs_setattrres res = {
2892 .fattr = fattr,
2893 .label = olabel,
2894 .server = server,
2895 };
2896 struct nfs4_exception exception = {
2897 .state = state,
2898 .inode = inode,
2899 .stateid = &arg.stateid,
2900 };
2901 int err;
2902
2903 arg.bitmask = nfs4_bitmask(server, ilabel);
2904 if (ilabel)
2905 arg.bitmask = nfs4_bitmask(server, olabel);
2906
2907 do {
2908 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
2909 switch (err) {
2910 case -NFS4ERR_OPENMODE:
2911 if (!(sattr->ia_valid & ATTR_SIZE)) {
2912 pr_warn_once("NFSv4: server %s is incorrectly "
2913 "applying open mode checks to "
2914 "a SETATTR that is not "
2915 "changing file size.\n",
2916 server->nfs_client->cl_hostname);
2917 }
2918 if (state && !(state->state & FMODE_WRITE)) {
2919 err = -EBADF;
2920 if (sattr->ia_valid & ATTR_OPEN)
2921 err = -EACCES;
2922 goto out;
2923 }
2924 }
2925 err = nfs4_handle_exception(server, err, &exception);
2926 } while (exception.retry);
2927 out:
2928 return err;
2929 }
2930
2931 static bool
2932 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2933 {
2934 if (inode == NULL || !nfs_have_layout(inode))
2935 return false;
2936
2937 return pnfs_wait_on_layoutreturn(inode, task);
2938 }
2939
2940 struct nfs4_closedata {
2941 struct inode *inode;
2942 struct nfs4_state *state;
2943 struct nfs_closeargs arg;
2944 struct nfs_closeres res;
2945 struct {
2946 struct nfs4_layoutreturn_args arg;
2947 struct nfs4_layoutreturn_res res;
2948 struct nfs4_xdr_opaque_data ld_private;
2949 u32 roc_barrier;
2950 bool roc;
2951 } lr;
2952 struct nfs_fattr fattr;
2953 unsigned long timestamp;
2954 };
2955
2956 static void nfs4_free_closedata(void *data)
2957 {
2958 struct nfs4_closedata *calldata = data;
2959 struct nfs4_state_owner *sp = calldata->state->owner;
2960 struct super_block *sb = calldata->state->inode->i_sb;
2961
2962 if (calldata->lr.roc)
2963 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
2964 calldata->res.lr_ret);
2965 nfs4_put_open_state(calldata->state);
2966 nfs_free_seqid(calldata->arg.seqid);
2967 nfs4_put_state_owner(sp);
2968 nfs_sb_deactive(sb);
2969 kfree(calldata);
2970 }
2971
2972 static void nfs4_close_done(struct rpc_task *task, void *data)
2973 {
2974 struct nfs4_closedata *calldata = data;
2975 struct nfs4_state *state = calldata->state;
2976 struct nfs_server *server = NFS_SERVER(calldata->inode);
2977 nfs4_stateid *res_stateid = NULL;
2978
2979 dprintk("%s: begin!\n", __func__);
2980 if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2981 return;
2982 trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2983
2984 /* Handle Layoutreturn errors */
2985 if (calldata->arg.lr_args && task->tk_status != 0) {
2986 switch (calldata->res.lr_ret) {
2987 default:
2988 calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
2989 break;
2990 case 0:
2991 calldata->arg.lr_args = NULL;
2992 calldata->res.lr_res = NULL;
2993 break;
2994 case -NFS4ERR_ADMIN_REVOKED:
2995 case -NFS4ERR_DELEG_REVOKED:
2996 case -NFS4ERR_EXPIRED:
2997 case -NFS4ERR_BAD_STATEID:
2998 case -NFS4ERR_OLD_STATEID:
2999 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
3000 case -NFS4ERR_WRONG_CRED:
3001 calldata->arg.lr_args = NULL;
3002 calldata->res.lr_res = NULL;
3003 calldata->res.lr_ret = 0;
3004 rpc_restart_call_prepare(task);
3005 return;
3006 }
3007 }
3008
3009 /* hmm. we are done with the inode, and in the process of freeing
3010 * the state_owner. we keep this around to process errors
3011 */
3012 switch (task->tk_status) {
3013 case 0:
3014 res_stateid = &calldata->res.stateid;
3015 renew_lease(server, calldata->timestamp);
3016 break;
3017 case -NFS4ERR_ACCESS:
3018 if (calldata->arg.bitmask != NULL) {
3019 calldata->arg.bitmask = NULL;
3020 calldata->res.fattr = NULL;
3021 task->tk_status = 0;
3022 rpc_restart_call_prepare(task);
3023 goto out_release;
3024
3025 }
3026 break;
3027 case -NFS4ERR_ADMIN_REVOKED:
3028 case -NFS4ERR_STALE_STATEID:
3029 case -NFS4ERR_EXPIRED:
3030 nfs4_free_revoked_stateid(server,
3031 &calldata->arg.stateid,
3032 task->tk_msg.rpc_cred);
3033 case -NFS4ERR_OLD_STATEID:
3034 case -NFS4ERR_BAD_STATEID:
3035 if (!nfs4_stateid_match(&calldata->arg.stateid,
3036 &state->open_stateid)) {
3037 rpc_restart_call_prepare(task);
3038 goto out_release;
3039 }
3040 if (calldata->arg.fmode == 0)
3041 break;
3042 default:
3043 if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
3044 rpc_restart_call_prepare(task);
3045 goto out_release;
3046 }
3047 }
3048 nfs_clear_open_stateid(state, &calldata->arg.stateid,
3049 res_stateid, calldata->arg.fmode);
3050 out_release:
3051 nfs_release_seqid(calldata->arg.seqid);
3052 nfs_refresh_inode(calldata->inode, &calldata->fattr);
3053 dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3054 }
3055
3056 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3057 {
3058 struct nfs4_closedata *calldata = data;
3059 struct nfs4_state *state = calldata->state;
3060 struct inode *inode = calldata->inode;
3061 bool is_rdonly, is_wronly, is_rdwr;
3062 int call_close = 0;
3063
3064 dprintk("%s: begin!\n", __func__);
3065 if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3066 goto out_wait;
3067
3068 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3069 spin_lock(&state->owner->so_lock);
3070 is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3071 is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3072 is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3073 nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
3074 /* Calculate the change in open mode */
3075 calldata->arg.fmode = 0;
3076 if (state->n_rdwr == 0) {
3077 if (state->n_rdonly == 0)
3078 call_close |= is_rdonly;
3079 else if (is_rdonly)
3080 calldata->arg.fmode |= FMODE_READ;
3081 if (state->n_wronly == 0)
3082 call_close |= is_wronly;
3083 else if (is_wronly)
3084 calldata->arg.fmode |= FMODE_WRITE;
3085 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3086 call_close |= is_rdwr;
3087 } else if (is_rdwr)
3088 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3089
3090 if (!nfs4_valid_open_stateid(state) ||
3091 test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3092 call_close = 0;
3093 spin_unlock(&state->owner->so_lock);
3094
3095 if (!call_close) {
3096 /* Note: exit _without_ calling nfs4_close_done */
3097 goto out_no_action;
3098 }
3099
3100 if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3101 nfs_release_seqid(calldata->arg.seqid);
3102 goto out_wait;
3103 }
3104
3105 if (calldata->arg.fmode == 0)
3106 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3107
3108 if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3109 /* Close-to-open cache consistency revalidation */
3110 if (!nfs4_have_delegation(inode, FMODE_READ))
3111 calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3112 else
3113 calldata->arg.bitmask = NULL;
3114 }
3115
3116 calldata->arg.share_access =
3117 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3118 calldata->arg.fmode, 0);
3119
3120 if (calldata->res.fattr == NULL)
3121 calldata->arg.bitmask = NULL;
3122 else if (calldata->arg.bitmask == NULL)
3123 calldata->res.fattr = NULL;
3124 calldata->timestamp = jiffies;
3125 if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3126 &calldata->arg.seq_args,
3127 &calldata->res.seq_res,
3128 task) != 0)
3129 nfs_release_seqid(calldata->arg.seqid);
3130 dprintk("%s: done!\n", __func__);
3131 return;
3132 out_no_action:
3133 task->tk_action = NULL;
3134 out_wait:
3135 nfs4_sequence_done(task, &calldata->res.seq_res);
3136 }
3137
3138 static const struct rpc_call_ops nfs4_close_ops = {
3139 .rpc_call_prepare = nfs4_close_prepare,
3140 .rpc_call_done = nfs4_close_done,
3141 .rpc_release = nfs4_free_closedata,
3142 };
3143
3144 /*
3145 * It is possible for data to be read/written from a mem-mapped file
3146 * after the sys_close call (which hits the vfs layer as a flush).
3147 * This means that we can't safely call nfsv4 close on a file until
3148 * the inode is cleared. This in turn means that we are not good
3149 * NFSv4 citizens - we do not indicate to the server to update the file's
3150 * share state even when we are done with one of the three share
3151 * stateid's in the inode.
3152 *
3153 * NOTE: Caller must be holding the sp->so_owner semaphore!
3154 */
3155 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3156 {
3157 struct nfs_server *server = NFS_SERVER(state->inode);
3158 struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3159 struct nfs4_closedata *calldata;
3160 struct nfs4_state_owner *sp = state->owner;
3161 struct rpc_task *task;
3162 struct rpc_message msg = {
3163 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3164 .rpc_cred = state->owner->so_cred,
3165 };
3166 struct rpc_task_setup task_setup_data = {
3167 .rpc_client = server->client,
3168 .rpc_message = &msg,
3169 .callback_ops = &nfs4_close_ops,
3170 .workqueue = nfsiod_workqueue,
3171 .flags = RPC_TASK_ASYNC,
3172 };
3173 int status = -ENOMEM;
3174
3175 nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3176 &task_setup_data.rpc_client, &msg);
3177
3178 calldata = kzalloc(sizeof(*calldata), gfp_mask);
3179 if (calldata == NULL)
3180 goto out;
3181 nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
3182 calldata->inode = state->inode;
3183 calldata->state = state;
3184 calldata->arg.fh = NFS_FH(state->inode);
3185 /* Serialization for the sequence id */
3186 alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3187 calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3188 if (IS_ERR(calldata->arg.seqid))
3189 goto out_free_calldata;
3190 nfs_fattr_init(&calldata->fattr);
3191 calldata->arg.fmode = 0;
3192 calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3193 calldata->res.fattr = &calldata->fattr;
3194 calldata->res.seqid = calldata->arg.seqid;
3195 calldata->res.server = server;
3196 calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3197 calldata->lr.roc = pnfs_roc(state->inode,
3198 &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3199 if (calldata->lr.roc) {
3200 calldata->arg.lr_args = &calldata->lr.arg;
3201 calldata->res.lr_res = &calldata->lr.res;
3202 }
3203 nfs_sb_active(calldata->inode->i_sb);
3204
3205 msg.rpc_argp = &calldata->arg;
3206 msg.rpc_resp = &calldata->res;
3207 task_setup_data.callback_data = calldata;
3208 task = rpc_run_task(&task_setup_data);
3209 if (IS_ERR(task))
3210 return PTR_ERR(task);
3211 status = 0;
3212 if (wait)
3213 status = rpc_wait_for_completion_task(task);
3214 rpc_put_task(task);
3215 return status;
3216 out_free_calldata:
3217 kfree(calldata);
3218 out:
3219 nfs4_put_open_state(state);
3220 nfs4_put_state_owner(sp);
3221 return status;
3222 }
3223
3224 static struct inode *
3225 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3226 int open_flags, struct iattr *attr, int *opened)
3227 {
3228 struct nfs4_state *state;
3229 struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3230
3231 label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3232
3233 /* Protect against concurrent sillydeletes */
3234 state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3235
3236 nfs4_label_release_security(label);
3237
3238 if (IS_ERR(state))
3239 return ERR_CAST(state);
3240 return state->inode;
3241 }
3242
3243 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3244 {
3245 if (ctx->state == NULL)
3246 return;
3247 if (is_sync)
3248 nfs4_close_sync(ctx->state, ctx->mode);
3249 else
3250 nfs4_close_state(ctx->state, ctx->mode);
3251 }
3252
3253 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3254 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3255 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_MODE_UMASK - 1UL)
3256
3257 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3258 {
3259 u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3260 struct nfs4_server_caps_arg args = {
3261 .fhandle = fhandle,
3262 .bitmask = bitmask,
3263 };
3264 struct nfs4_server_caps_res res = {};
3265 struct rpc_message msg = {
3266 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3267 .rpc_argp = &args,
3268 .rpc_resp = &res,
3269 };
3270 int status;
3271
3272 bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3273 FATTR4_WORD0_FH_EXPIRE_TYPE |
3274 FATTR4_WORD0_LINK_SUPPORT |
3275 FATTR4_WORD0_SYMLINK_SUPPORT |
3276 FATTR4_WORD0_ACLSUPPORT;
3277 if (minorversion)
3278 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3279
3280 status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3281 if (status == 0) {
3282 /* Sanity check the server answers */
3283 switch (minorversion) {
3284 case 0:
3285 res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3286 res.attr_bitmask[2] = 0;
3287 break;
3288 case 1:
3289 res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3290 break;
3291 case 2:
3292 res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3293 }
3294 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3295 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3296 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3297 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3298 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3299 NFS_CAP_CTIME|NFS_CAP_MTIME|
3300 NFS_CAP_SECURITY_LABEL);
3301 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3302 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3303 server->caps |= NFS_CAP_ACLS;
3304 if (res.has_links != 0)
3305 server->caps |= NFS_CAP_HARDLINKS;
3306 if (res.has_symlinks != 0)
3307 server->caps |= NFS_CAP_SYMLINKS;
3308 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3309 server->caps |= NFS_CAP_FILEID;
3310 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3311 server->caps |= NFS_CAP_MODE;
3312 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3313 server->caps |= NFS_CAP_NLINK;
3314 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3315 server->caps |= NFS_CAP_OWNER;
3316 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3317 server->caps |= NFS_CAP_OWNER_GROUP;
3318 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3319 server->caps |= NFS_CAP_ATIME;
3320 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3321 server->caps |= NFS_CAP_CTIME;
3322 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3323 server->caps |= NFS_CAP_MTIME;
3324 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3325 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3326 server->caps |= NFS_CAP_SECURITY_LABEL;
3327 #endif
3328 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3329 sizeof(server->attr_bitmask));
3330 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3331
3332 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3333 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3334 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3335 server->cache_consistency_bitmask[2] = 0;
3336 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3337 sizeof(server->exclcreat_bitmask));
3338 server->acl_bitmask = res.acl_bitmask;
3339 server->fh_expire_type = res.fh_expire_type;
3340 }
3341
3342 return status;
3343 }
3344
3345 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3346 {
3347 struct nfs4_exception exception = { };
3348 int err;
3349 do {
3350 err = nfs4_handle_exception(server,
3351 _nfs4_server_capabilities(server, fhandle),
3352 &exception);
3353 } while (exception.retry);
3354 return err;
3355 }
3356
3357 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3358 struct nfs_fsinfo *info)
3359 {
3360 u32 bitmask[3];
3361 struct nfs4_lookup_root_arg args = {
3362 .bitmask = bitmask,
3363 };
3364 struct nfs4_lookup_res res = {
3365 .server = server,
3366 .fattr = info->fattr,
3367 .fh = fhandle,
3368 };
3369 struct rpc_message msg = {
3370 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3371 .rpc_argp = &args,
3372 .rpc_resp = &res,
3373 };
3374
3375 bitmask[0] = nfs4_fattr_bitmap[0];
3376 bitmask[1] = nfs4_fattr_bitmap[1];
3377 /*
3378 * Process the label in the upcoming getfattr
3379 */
3380 bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3381
3382 nfs_fattr_init(info->fattr);
3383 return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3384 }
3385
3386 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3387 struct nfs_fsinfo *info)
3388 {
3389 struct nfs4_exception exception = { };
3390 int err;
3391 do {
3392 err = _nfs4_lookup_root(server, fhandle, info);
3393 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3394 switch (err) {
3395 case 0:
3396 case -NFS4ERR_WRONGSEC:
3397 goto out;
3398 default:
3399 err = nfs4_handle_exception(server, err, &exception);
3400 }
3401 } while (exception.retry);
3402 out:
3403 return err;
3404 }
3405
3406 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3407 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3408 {
3409 struct rpc_auth_create_args auth_args = {
3410 .pseudoflavor = flavor,
3411 };
3412 struct rpc_auth *auth;
3413
3414 auth = rpcauth_create(&auth_args, server->client);
3415 if (IS_ERR(auth))
3416 return -EACCES;
3417 return nfs4_lookup_root(server, fhandle, info);
3418 }
3419
3420 /*
3421 * Retry pseudoroot lookup with various security flavors. We do this when:
3422 *
3423 * NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3424 * NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3425 *
3426 * Returns zero on success, or a negative NFS4ERR value, or a
3427 * negative errno value.
3428 */
3429 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3430 struct nfs_fsinfo *info)
3431 {
3432 /* Per 3530bis 15.33.5 */
3433 static const rpc_authflavor_t flav_array[] = {
3434 RPC_AUTH_GSS_KRB5P,
3435 RPC_AUTH_GSS_KRB5I,
3436 RPC_AUTH_GSS_KRB5,
3437 RPC_AUTH_UNIX, /* courtesy */
3438 RPC_AUTH_NULL,
3439 };
3440 int status = -EPERM;
3441 size_t i;
3442
3443 if (server->auth_info.flavor_len > 0) {
3444 /* try each flavor specified by user */
3445 for (i = 0; i < server->auth_info.flavor_len; i++) {
3446 status = nfs4_lookup_root_sec(server, fhandle, info,
3447 server->auth_info.flavors[i]);
3448 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3449 continue;
3450 break;
3451 }
3452 } else {
3453 /* no flavors specified by user, try default list */
3454 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3455 status = nfs4_lookup_root_sec(server, fhandle, info,
3456 flav_array[i]);
3457 if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3458 continue;
3459 break;
3460 }
3461 }
3462
3463 /*
3464 * -EACCESS could mean that the user doesn't have correct permissions
3465 * to access the mount. It could also mean that we tried to mount
3466 * with a gss auth flavor, but rpc.gssd isn't running. Either way,
3467 * existing mount programs don't handle -EACCES very well so it should
3468 * be mapped to -EPERM instead.
3469 */
3470 if (status == -EACCES)
3471 status = -EPERM;
3472 return status;
3473 }
3474
3475 /**
3476 * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3477 * @server: initialized nfs_server handle
3478 * @fhandle: we fill in the pseudo-fs root file handle
3479 * @info: we fill in an FSINFO struct
3480 * @auth_probe: probe the auth flavours
3481 *
3482 * Returns zero on success, or a negative errno.
3483 */
3484 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3485 struct nfs_fsinfo *info,
3486 bool auth_probe)
3487 {
3488 int status = 0;
3489
3490 if (!auth_probe)
3491 status = nfs4_lookup_root(server, fhandle, info);
3492
3493 if (auth_probe || status == NFS4ERR_WRONGSEC)
3494 status = server->nfs_client->cl_mvops->find_root_sec(server,
3495 fhandle, info);
3496
3497 if (status == 0)
3498 status = nfs4_server_capabilities(server, fhandle);
3499 if (status == 0)
3500 status = nfs4_do_fsinfo(server, fhandle, info);
3501
3502 return nfs4_map_errors(status);
3503 }
3504
3505 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3506 struct nfs_fsinfo *info)
3507 {
3508 int error;
3509 struct nfs_fattr *fattr = info->fattr;
3510 struct nfs4_label *label = NULL;
3511
3512 error = nfs4_server_capabilities(server, mntfh);
3513 if (error < 0) {
3514 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3515 return error;
3516 }
3517
3518 label = nfs4_label_alloc(server, GFP_KERNEL);
3519 if (IS_ERR(label))
3520 return PTR_ERR(label);
3521
3522 error = nfs4_proc_getattr(server, mntfh, fattr, label);
3523 if (error < 0) {
3524 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3525 goto err_free_label;
3526 }
3527
3528 if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3529 !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3530 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3531
3532 err_free_label:
3533 nfs4_label_free(label);
3534
3535 return error;
3536 }
3537
3538 /*
3539 * Get locations and (maybe) other attributes of a referral.
3540 * Note that we'll actually follow the referral later when
3541 * we detect fsid mismatch in inode revalidation
3542 */
3543 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3544 const struct qstr *name, struct nfs_fattr *fattr,
3545 struct nfs_fh *fhandle)
3546 {
3547 int status = -ENOMEM;
3548 struct page *page = NULL;
3549 struct nfs4_fs_locations *locations = NULL;
3550
3551 page = alloc_page(GFP_KERNEL);
3552 if (page == NULL)
3553 goto out;
3554 locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3555 if (locations == NULL)
3556 goto out;
3557
3558 status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3559 if (status != 0)
3560 goto out;
3561
3562 /*
3563 * If the fsid didn't change, this is a migration event, not a
3564 * referral. Cause us to drop into the exception handler, which
3565 * will kick off migration recovery.
3566 */
3567 if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3568 dprintk("%s: server did not return a different fsid for"
3569 " a referral at %s\n", __func__, name->name);
3570 status = -NFS4ERR_MOVED;
3571 goto out;
3572 }
3573 /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3574 nfs_fixup_referral_attributes(&locations->fattr);
3575
3576 /* replace the lookup nfs_fattr with the locations nfs_fattr */
3577 memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3578 memset(fhandle, 0, sizeof(struct nfs_fh));
3579 out:
3580 if (page)
3581 __free_page(page);
3582 kfree(locations);
3583 return status;
3584 }
3585
3586 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3587 struct nfs_fattr *fattr, struct nfs4_label *label)
3588 {
3589 struct nfs4_getattr_arg args = {
3590 .fh = fhandle,
3591 .bitmask = server->attr_bitmask,
3592 };
3593 struct nfs4_getattr_res res = {
3594 .fattr = fattr,
3595 .label = label,
3596 .server = server,
3597 };
3598 struct rpc_message msg = {
3599 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3600 .rpc_argp = &args,
3601 .rpc_resp = &res,
3602 };
3603
3604 args.bitmask = nfs4_bitmask(server, label);
3605
3606 nfs_fattr_init(fattr);
3607 return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3608 }
3609
3610 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3611 struct nfs_fattr *fattr, struct nfs4_label *label)
3612 {
3613 struct nfs4_exception exception = { };
3614 int err;
3615 do {
3616 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3617 trace_nfs4_getattr(server, fhandle, fattr, err);
3618 err = nfs4_handle_exception(server, err,
3619 &exception);
3620 } while (exception.retry);
3621 return err;
3622 }
3623
3624 /*
3625 * The file is not closed if it is opened due to the a request to change
3626 * the size of the file. The open call will not be needed once the
3627 * VFS layer lookup-intents are implemented.
3628 *
3629 * Close is called when the inode is destroyed.
3630 * If we haven't opened the file for O_WRONLY, we
3631 * need to in the size_change case to obtain a stateid.
3632 *
3633 * Got race?
3634 * Because OPEN is always done by name in nfsv4, it is
3635 * possible that we opened a different file by the same
3636 * name. We can recognize this race condition, but we
3637 * can't do anything about it besides returning an error.
3638 *
3639 * This will be fixed with VFS changes (lookup-intent).
3640 */
3641 static int
3642 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3643 struct iattr *sattr)
3644 {
3645 struct inode *inode = d_inode(dentry);
3646 struct rpc_cred *cred = NULL;
3647 struct nfs_open_context *ctx = NULL;
3648 struct nfs4_label *label = NULL;
3649 int status;
3650
3651 if (pnfs_ld_layoutret_on_setattr(inode) &&
3652 sattr->ia_valid & ATTR_SIZE &&
3653 sattr->ia_size < i_size_read(inode))
3654 pnfs_commit_and_return_layout(inode);
3655
3656 nfs_fattr_init(fattr);
3657
3658 /* Deal with open(O_TRUNC) */
3659 if (sattr->ia_valid & ATTR_OPEN)
3660 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3661
3662 /* Optimization: if the end result is no change, don't RPC */
3663 if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3664 return 0;
3665
3666 /* Search for an existing open(O_WRITE) file */
3667 if (sattr->ia_valid & ATTR_FILE) {
3668
3669 ctx = nfs_file_open_context(sattr->ia_file);
3670 if (ctx)
3671 cred = ctx->cred;
3672 }
3673
3674 label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3675 if (IS_ERR(label))
3676 return PTR_ERR(label);
3677
3678 status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
3679 if (status == 0) {
3680 nfs_setattr_update_inode(inode, sattr, fattr);
3681 nfs_setsecurity(inode, fattr, label);
3682 }
3683 nfs4_label_free(label);
3684 return status;
3685 }
3686
3687 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3688 const struct qstr *name, struct nfs_fh *fhandle,
3689 struct nfs_fattr *fattr, struct nfs4_label *label)
3690 {
3691 struct nfs_server *server = NFS_SERVER(dir);
3692 int status;
3693 struct nfs4_lookup_arg args = {
3694 .bitmask = server->attr_bitmask,
3695 .dir_fh = NFS_FH(dir),
3696 .name = name,
3697 };
3698 struct nfs4_lookup_res res = {
3699 .server = server,
3700 .fattr = fattr,
3701 .label = label,
3702 .fh = fhandle,
3703 };
3704 struct rpc_message msg = {
3705 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3706 .rpc_argp = &args,
3707 .rpc_resp = &res,
3708 };
3709
3710 args.bitmask = nfs4_bitmask(server, label);
3711
3712 nfs_fattr_init(fattr);
3713
3714 dprintk("NFS call lookup %s\n", name->name);
3715 status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3716 dprintk("NFS reply lookup: %d\n", status);
3717 return status;
3718 }
3719
3720 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3721 {
3722 fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3723 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3724 fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3725 fattr->nlink = 2;
3726 }
3727
3728 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3729 const struct qstr *name, struct nfs_fh *fhandle,
3730 struct nfs_fattr *fattr, struct nfs4_label *label)
3731 {
3732 struct nfs4_exception exception = { };
3733 struct rpc_clnt *client = *clnt;
3734 int err;
3735 do {
3736 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3737 trace_nfs4_lookup(dir, name, err);
3738 switch (err) {
3739 case -NFS4ERR_BADNAME:
3740 err = -ENOENT;
3741 goto out;
3742 case -NFS4ERR_MOVED:
3743 err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3744 if (err == -NFS4ERR_MOVED)
3745 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3746 goto out;
3747 case -NFS4ERR_WRONGSEC:
3748 err = -EPERM;
3749 if (client != *clnt)
3750 goto out;
3751 client = nfs4_negotiate_security(client, dir, name);
3752 if (IS_ERR(client))
3753 return PTR_ERR(client);
3754
3755 exception.retry = 1;
3756 break;
3757 default:
3758 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3759 }
3760 } while (exception.retry);
3761
3762 out:
3763 if (err == 0)
3764 *clnt = client;
3765 else if (client != *clnt)
3766 rpc_shutdown_client(client);
3767
3768 return err;
3769 }
3770
3771 static int nfs4_proc_lookup(struct inode *dir, const struct qstr *name,
3772 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3773 struct nfs4_label *label)
3774 {
3775 int status;
3776 struct rpc_clnt *client = NFS_CLIENT(dir);
3777
3778 status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3779 if (client != NFS_CLIENT(dir)) {
3780 rpc_shutdown_client(client);
3781 nfs_fixup_secinfo_attributes(fattr);
3782 }
3783 return status;
3784 }
3785
3786 struct rpc_clnt *
3787 nfs4_proc_lookup_mountpoint(struct inode *dir, const struct qstr *name,
3788 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3789 {
3790 struct rpc_clnt *client = NFS_CLIENT(dir);
3791 int status;
3792
3793 status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3794 if (status < 0)
3795 return ERR_PTR(status);
3796 return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3797 }
3798
3799 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3800 {
3801 struct nfs_server *server = NFS_SERVER(inode);
3802 struct nfs4_accessargs args = {
3803 .fh = NFS_FH(inode),
3804 .bitmask = server->cache_consistency_bitmask,
3805 };
3806 struct nfs4_accessres res = {
3807 .server = server,
3808 };
3809 struct rpc_message msg = {
3810 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3811 .rpc_argp = &args,
3812 .rpc_resp = &res,
3813 .rpc_cred = entry->cred,
3814 };
3815 int mode = entry->mask;
3816 int status = 0;
3817
3818 /*
3819 * Determine which access bits we want to ask for...
3820 */
3821 if (mode & MAY_READ)
3822 args.access |= NFS4_ACCESS_READ;
3823 if (S_ISDIR(inode->i_mode)) {
3824 if (mode & MAY_WRITE)
3825 args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3826 if (mode & MAY_EXEC)
3827 args.access |= NFS4_ACCESS_LOOKUP;
3828 } else {
3829 if (mode & MAY_WRITE)
3830 args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3831 if (mode & MAY_EXEC)
3832 args.access |= NFS4_ACCESS_EXECUTE;
3833 }
3834
3835 res.fattr = nfs_alloc_fattr();
3836 if (res.fattr == NULL)
3837 return -ENOMEM;
3838
3839 status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3840 if (!status) {
3841 nfs_access_set_mask(entry, res.access);
3842 nfs_refresh_inode(inode, res.fattr);
3843 }
3844 nfs_free_fattr(res.fattr);
3845 return status;
3846 }
3847
3848 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3849 {
3850 struct nfs4_exception exception = { };
3851 int err;
3852 do {
3853 err = _nfs4_proc_access(inode, entry);
3854 trace_nfs4_access(inode, err);
3855 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3856 &exception);
3857 } while (exception.retry);
3858 return err;
3859 }
3860
3861 /*
3862 * TODO: For the time being, we don't try to get any attributes
3863 * along with any of the zero-copy operations READ, READDIR,
3864 * READLINK, WRITE.
3865 *
3866 * In the case of the first three, we want to put the GETATTR
3867 * after the read-type operation -- this is because it is hard
3868 * to predict the length of a GETATTR response in v4, and thus
3869 * align the READ data correctly. This means that the GETATTR
3870 * may end up partially falling into the page cache, and we should
3871 * shift it into the 'tail' of the xdr_buf before processing.
3872 * To do this efficiently, we need to know the total length
3873 * of data received, which doesn't seem to be available outside
3874 * of the RPC layer.
3875 *
3876 * In the case of WRITE, we also want to put the GETATTR after
3877 * the operation -- in this case because we want to make sure
3878 * we get the post-operation mtime and size.
3879 *
3880 * Both of these changes to the XDR layer would in fact be quite
3881 * minor, but I decided to leave them for a subsequent patch.
3882 */
3883 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3884 unsigned int pgbase, unsigned int pglen)
3885 {
3886 struct nfs4_readlink args = {
3887 .fh = NFS_FH(inode),
3888 .pgbase = pgbase,
3889 .pglen = pglen,
3890 .pages = &page,
3891 };
3892 struct nfs4_readlink_res res;
3893 struct rpc_message msg = {
3894 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3895 .rpc_argp = &args,
3896 .rpc_resp = &res,
3897 };
3898
3899 return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3900 }
3901
3902 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3903 unsigned int pgbase, unsigned int pglen)
3904 {
3905 struct nfs4_exception exception = { };
3906 int err;
3907 do {
3908 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3909 trace_nfs4_readlink(inode, err);
3910 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3911 &exception);
3912 } while (exception.retry);
3913 return err;
3914 }
3915
3916 /*
3917 * This is just for mknod. open(O_CREAT) will always do ->open_context().
3918 */
3919 static int
3920 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3921 int flags)
3922 {
3923 struct nfs_server *server = NFS_SERVER(dir);
3924 struct nfs4_label l, *ilabel = NULL;
3925 struct nfs_open_context *ctx;
3926 struct nfs4_state *state;
3927 int status = 0;
3928
3929 ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
3930 if (IS_ERR(ctx))
3931 return PTR_ERR(ctx);
3932
3933 ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3934
3935 if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
3936 sattr->ia_mode &= ~current_umask();
3937 state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3938 if (IS_ERR(state)) {
3939 status = PTR_ERR(state);
3940 goto out;
3941 }
3942 out:
3943 nfs4_label_release_security(ilabel);
3944 put_nfs_open_context(ctx);
3945 return status;
3946 }
3947
3948 static int _nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3949 {
3950 struct nfs_server *server = NFS_SERVER(dir);
3951 struct nfs_removeargs args = {
3952 .fh = NFS_FH(dir),
3953 .name = *name,
3954 };
3955 struct nfs_removeres res = {
3956 .server = server,
3957 };
3958 struct rpc_message msg = {
3959 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3960 .rpc_argp = &args,
3961 .rpc_resp = &res,
3962 };
3963 unsigned long timestamp = jiffies;
3964 int status;
3965
3966 status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3967 if (status == 0)
3968 update_changeattr(dir, &res.cinfo, timestamp);
3969 return status;
3970 }
3971
3972 static int nfs4_proc_remove(struct inode *dir, const struct qstr *name)
3973 {
3974 struct nfs4_exception exception = { };
3975 int err;
3976 do {
3977 err = _nfs4_proc_remove(dir, name);
3978 trace_nfs4_remove(dir, name, err);
3979 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3980 &exception);
3981 } while (exception.retry);
3982 return err;
3983 }
3984
3985 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3986 {
3987 struct nfs_server *server = NFS_SERVER(dir);
3988 struct nfs_removeargs *args = msg->rpc_argp;
3989 struct nfs_removeres *res = msg->rpc_resp;
3990
3991 res->server = server;
3992 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3993 nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3994
3995 nfs_fattr_init(res->dir_attr);
3996 }
3997
3998 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3999 {
4000 nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4001 &data->args.seq_args,
4002 &data->res.seq_res,
4003 task);
4004 }
4005
4006 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4007 {
4008 struct nfs_unlinkdata *data = task->tk_calldata;
4009 struct nfs_removeres *res = &data->res;
4010
4011 if (!nfs4_sequence_done(task, &res->seq_res))
4012 return 0;
4013 if (nfs4_async_handle_error(task, res->server, NULL,
4014 &data->timeout) == -EAGAIN)
4015 return 0;
4016 if (task->tk_status == 0)
4017 update_changeattr(dir, &res->cinfo, res->dir_attr->time_start);
4018 return 1;
4019 }
4020
4021 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
4022 {
4023 struct nfs_server *server = NFS_SERVER(dir);
4024 struct nfs_renameargs *arg = msg->rpc_argp;
4025 struct nfs_renameres *res = msg->rpc_resp;
4026
4027 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4028 res->server = server;
4029 nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
4030 }
4031
4032 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4033 {
4034 nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4035 &data->args.seq_args,
4036 &data->res.seq_res,
4037 task);
4038 }
4039
4040 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4041 struct inode *new_dir)
4042 {
4043 struct nfs_renamedata *data = task->tk_calldata;
4044 struct nfs_renameres *res = &data->res;
4045
4046 if (!nfs4_sequence_done(task, &res->seq_res))
4047 return 0;
4048 if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4049 return 0;
4050
4051 if (task->tk_status == 0) {
4052 update_changeattr(old_dir, &res->old_cinfo, res->old_fattr->time_start);
4053 if (new_dir != old_dir)
4054 update_changeattr(new_dir, &res->new_cinfo, res->new_fattr->time_start);
4055 }
4056 return 1;
4057 }
4058
4059 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4060 {
4061 struct nfs_server *server = NFS_SERVER(inode);
4062 struct nfs4_link_arg arg = {
4063 .fh = NFS_FH(inode),
4064 .dir_fh = NFS_FH(dir),
4065 .name = name,
4066 .bitmask = server->attr_bitmask,
4067 };
4068 struct nfs4_link_res res = {
4069 .server = server,
4070 .label = NULL,
4071 };
4072 struct rpc_message msg = {
4073 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4074 .rpc_argp = &arg,
4075 .rpc_resp = &res,
4076 };
4077 int status = -ENOMEM;
4078
4079 res.fattr = nfs_alloc_fattr();
4080 if (res.fattr == NULL)
4081 goto out;
4082
4083 res.label = nfs4_label_alloc(server, GFP_KERNEL);
4084 if (IS_ERR(res.label)) {
4085 status = PTR_ERR(res.label);
4086 goto out;
4087 }
4088 arg.bitmask = nfs4_bitmask(server, res.label);
4089
4090 status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4091 if (!status) {
4092 update_changeattr(dir, &res.cinfo, res.fattr->time_start);
4093 status = nfs_post_op_update_inode(inode, res.fattr);
4094 if (!status)
4095 nfs_setsecurity(inode, res.fattr, res.label);
4096 }
4097
4098
4099 nfs4_label_free(res.label);
4100
4101 out:
4102 nfs_free_fattr(res.fattr);
4103 return status;
4104 }
4105
4106 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4107 {
4108 struct nfs4_exception exception = { };
4109 int err;
4110 do {
4111 err = nfs4_handle_exception(NFS_SERVER(inode),
4112 _nfs4_proc_link(inode, dir, name),
4113 &exception);
4114 } while (exception.retry);
4115 return err;
4116 }
4117
4118 struct nfs4_createdata {
4119 struct rpc_message msg;
4120 struct nfs4_create_arg arg;
4121 struct nfs4_create_res res;
4122 struct nfs_fh fh;
4123 struct nfs_fattr fattr;
4124 struct nfs4_label *label;
4125 };
4126
4127 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4128 const struct qstr *name, struct iattr *sattr, u32 ftype)
4129 {
4130 struct nfs4_createdata *data;
4131
4132 data = kzalloc(sizeof(*data), GFP_KERNEL);
4133 if (data != NULL) {
4134 struct nfs_server *server = NFS_SERVER(dir);
4135
4136 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4137 if (IS_ERR(data->label))
4138 goto out_free;
4139
4140 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4141 data->msg.rpc_argp = &data->arg;
4142 data->msg.rpc_resp = &data->res;
4143 data->arg.dir_fh = NFS_FH(dir);
4144 data->arg.server = server;
4145 data->arg.name = name;
4146 data->arg.attrs = sattr;
4147 data->arg.ftype = ftype;
4148 data->arg.bitmask = nfs4_bitmask(server, data->label);
4149 data->arg.umask = current_umask();
4150 data->res.server = server;
4151 data->res.fh = &data->fh;
4152 data->res.fattr = &data->fattr;
4153 data->res.label = data->label;
4154 nfs_fattr_init(data->res.fattr);
4155 }
4156 return data;
4157 out_free:
4158 kfree(data);
4159 return NULL;
4160 }
4161
4162 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4163 {
4164 int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4165 &data->arg.seq_args, &data->res.seq_res, 1);
4166 if (status == 0) {
4167 update_changeattr(dir, &data->res.dir_cinfo,
4168 data->res.fattr->time_start);
4169 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4170 }
4171 return status;
4172 }
4173
4174 static void nfs4_free_createdata(struct nfs4_createdata *data)
4175 {
4176 nfs4_label_free(data->label);
4177 kfree(data);
4178 }
4179
4180 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4181 struct page *page, unsigned int len, struct iattr *sattr,
4182 struct nfs4_label *label)
4183 {
4184 struct nfs4_createdata *data;
4185 int status = -ENAMETOOLONG;
4186
4187 if (len > NFS4_MAXPATHLEN)
4188 goto out;
4189
4190 status = -ENOMEM;
4191 data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4192 if (data == NULL)
4193 goto out;
4194
4195 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4196 data->arg.u.symlink.pages = &page;
4197 data->arg.u.symlink.len = len;
4198 data->arg.label = label;
4199
4200 status = nfs4_do_create(dir, dentry, data);
4201
4202 nfs4_free_createdata(data);
4203 out:
4204 return status;
4205 }
4206
4207 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4208 struct page *page, unsigned int len, struct iattr *sattr)
4209 {
4210 struct nfs4_exception exception = { };
4211 struct nfs4_label l, *label = NULL;
4212 int err;
4213
4214 label = nfs4_label_init_security(dir, dentry, sattr, &l);
4215
4216 do {
4217 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4218 trace_nfs4_symlink(dir, &dentry->d_name, err);
4219 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4220 &exception);
4221 } while (exception.retry);
4222
4223 nfs4_label_release_security(label);
4224 return err;
4225 }
4226
4227 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4228 struct iattr *sattr, struct nfs4_label *label)
4229 {
4230 struct nfs4_createdata *data;
4231 int status = -ENOMEM;
4232
4233 data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4234 if (data == NULL)
4235 goto out;
4236
4237 data->arg.label = label;
4238 status = nfs4_do_create(dir, dentry, data);
4239
4240 nfs4_free_createdata(data);
4241 out:
4242 return status;
4243 }
4244
4245 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4246 struct iattr *sattr)
4247 {
4248 struct nfs_server *server = NFS_SERVER(dir);
4249 struct nfs4_exception exception = { };
4250 struct nfs4_label l, *label = NULL;
4251 int err;
4252
4253 label = nfs4_label_init_security(dir, dentry, sattr, &l);
4254
4255 if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4256 sattr->ia_mode &= ~current_umask();
4257 do {
4258 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4259 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4260 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4261 &exception);
4262 } while (exception.retry);
4263 nfs4_label_release_security(label);
4264
4265 return err;
4266 }
4267
4268 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4269 u64 cookie, struct page **pages, unsigned int count, int plus)
4270 {
4271 struct inode *dir = d_inode(dentry);
4272 struct nfs4_readdir_arg args = {
4273 .fh = NFS_FH(dir),
4274 .pages = pages,
4275 .pgbase = 0,
4276 .count = count,
4277 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4278 .plus = plus,
4279 };
4280 struct nfs4_readdir_res res;
4281 struct rpc_message msg = {
4282 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4283 .rpc_argp = &args,
4284 .rpc_resp = &res,
4285 .rpc_cred = cred,
4286 };
4287 int status;
4288
4289 dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4290 dentry,
4291 (unsigned long long)cookie);
4292 nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4293 res.pgbase = args.pgbase;
4294 status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4295 if (status >= 0) {
4296 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4297 status += args.pgbase;
4298 }
4299
4300 nfs_invalidate_atime(dir);
4301
4302 dprintk("%s: returns %d\n", __func__, status);
4303 return status;
4304 }
4305
4306 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4307 u64 cookie, struct page **pages, unsigned int count, int plus)
4308 {
4309 struct nfs4_exception exception = { };
4310 int err;
4311 do {
4312 err = _nfs4_proc_readdir(dentry, cred, cookie,
4313 pages, count, plus);
4314 trace_nfs4_readdir(d_inode(dentry), err);
4315 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4316 &exception);
4317 } while (exception.retry);
4318 return err;
4319 }
4320
4321 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4322 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4323 {
4324 struct nfs4_createdata *data;
4325 int mode = sattr->ia_mode;
4326 int status = -ENOMEM;
4327
4328 data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4329 if (data == NULL)
4330 goto out;
4331
4332 if (S_ISFIFO(mode))
4333 data->arg.ftype = NF4FIFO;
4334 else if (S_ISBLK(mode)) {
4335 data->arg.ftype = NF4BLK;
4336 data->arg.u.device.specdata1 = MAJOR(rdev);
4337 data->arg.u.device.specdata2 = MINOR(rdev);
4338 }
4339 else if (S_ISCHR(mode)) {
4340 data->arg.ftype = NF4CHR;
4341 data->arg.u.device.specdata1 = MAJOR(rdev);
4342 data->arg.u.device.specdata2 = MINOR(rdev);
4343 } else if (!S_ISSOCK(mode)) {
4344 status = -EINVAL;
4345 goto out_free;
4346 }
4347
4348 data->arg.label = label;
4349 status = nfs4_do_create(dir, dentry, data);
4350 out_free:
4351 nfs4_free_createdata(data);
4352 out:
4353 return status;
4354 }
4355
4356 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4357 struct iattr *sattr, dev_t rdev)
4358 {
4359 struct nfs_server *server = NFS_SERVER(dir);
4360 struct nfs4_exception exception = { };
4361 struct nfs4_label l, *label = NULL;
4362 int err;
4363
4364 label = nfs4_label_init_security(dir, dentry, sattr, &l);
4365
4366 if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4367 sattr->ia_mode &= ~current_umask();
4368 do {
4369 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4370 trace_nfs4_mknod(dir, &dentry->d_name, err);
4371 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4372 &exception);
4373 } while (exception.retry);
4374
4375 nfs4_label_release_security(label);
4376
4377 return err;
4378 }
4379
4380 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4381 struct nfs_fsstat *fsstat)
4382 {
4383 struct nfs4_statfs_arg args = {
4384 .fh = fhandle,
4385 .bitmask = server->attr_bitmask,
4386 };
4387 struct nfs4_statfs_res res = {
4388 .fsstat = fsstat,
4389 };
4390 struct rpc_message msg = {
4391 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4392 .rpc_argp = &args,
4393 .rpc_resp = &res,
4394 };
4395
4396 nfs_fattr_init(fsstat->fattr);
4397 return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4398 }
4399
4400 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4401 {
4402 struct nfs4_exception exception = { };
4403 int err;
4404 do {
4405 err = nfs4_handle_exception(server,
4406 _nfs4_proc_statfs(server, fhandle, fsstat),
4407 &exception);
4408 } while (exception.retry);
4409 return err;
4410 }
4411
4412 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4413 struct nfs_fsinfo *fsinfo)
4414 {
4415 struct nfs4_fsinfo_arg args = {
4416 .fh = fhandle,
4417 .bitmask = server->attr_bitmask,
4418 };
4419 struct nfs4_fsinfo_res res = {
4420 .fsinfo = fsinfo,
4421 };
4422 struct rpc_message msg = {
4423 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4424 .rpc_argp = &args,
4425 .rpc_resp = &res,
4426 };
4427
4428 return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4429 }
4430
4431 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4432 {
4433 struct nfs4_exception exception = { };
4434 unsigned long now = jiffies;
4435 int err;
4436
4437 do {
4438 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4439 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4440 if (err == 0) {
4441 nfs4_set_lease_period(server->nfs_client,
4442 fsinfo->lease_time * HZ,
4443 now);
4444 break;
4445 }
4446 err = nfs4_handle_exception(server, err, &exception);
4447 } while (exception.retry);
4448 return err;
4449 }
4450
4451 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4452 {
4453 int error;
4454
4455 nfs_fattr_init(fsinfo->fattr);
4456 error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4457 if (error == 0) {
4458 /* block layout checks this! */
4459 server->pnfs_blksize = fsinfo->blksize;
4460 set_pnfs_layoutdriver(server, fhandle, fsinfo);
4461 }
4462
4463 return error;
4464 }
4465
4466 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4467 struct nfs_pathconf *pathconf)
4468 {
4469 struct nfs4_pathconf_arg args = {
4470 .fh = fhandle,
4471 .bitmask = server->attr_bitmask,
4472 };
4473 struct nfs4_pathconf_res res = {
4474 .pathconf = pathconf,
4475 };
4476 struct rpc_message msg = {
4477 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4478 .rpc_argp = &args,
4479 .rpc_resp = &res,
4480 };
4481
4482 /* None of the pathconf attributes are mandatory to implement */
4483 if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4484 memset(pathconf, 0, sizeof(*pathconf));
4485 return 0;
4486 }
4487
4488 nfs_fattr_init(pathconf->fattr);
4489 return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4490 }
4491
4492 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4493 struct nfs_pathconf *pathconf)
4494 {
4495 struct nfs4_exception exception = { };
4496 int err;
4497
4498 do {
4499 err = nfs4_handle_exception(server,
4500 _nfs4_proc_pathconf(server, fhandle, pathconf),
4501 &exception);
4502 } while (exception.retry);
4503 return err;
4504 }
4505
4506 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4507 const struct nfs_open_context *ctx,
4508 const struct nfs_lock_context *l_ctx,
4509 fmode_t fmode)
4510 {
4511 return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
4512 }
4513 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4514
4515 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4516 const struct nfs_open_context *ctx,
4517 const struct nfs_lock_context *l_ctx,
4518 fmode_t fmode)
4519 {
4520 nfs4_stateid current_stateid;
4521
4522 /* If the current stateid represents a lost lock, then exit */
4523 if (nfs4_set_rw_stateid(¤t_stateid, ctx, l_ctx, fmode) == -EIO)
4524 return true;
4525 return nfs4_stateid_match(stateid, ¤t_stateid);
4526 }
4527
4528 static bool nfs4_error_stateid_expired(int err)
4529 {
4530 switch (err) {
4531 case -NFS4ERR_DELEG_REVOKED:
4532 case -NFS4ERR_ADMIN_REVOKED:
4533 case -NFS4ERR_BAD_STATEID:
4534 case -NFS4ERR_STALE_STATEID:
4535 case -NFS4ERR_OLD_STATEID:
4536 case -NFS4ERR_OPENMODE:
4537 case -NFS4ERR_EXPIRED:
4538 return true;
4539 }
4540 return false;
4541 }
4542
4543 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4544 {
4545 struct nfs_server *server = NFS_SERVER(hdr->inode);
4546
4547 trace_nfs4_read(hdr, task->tk_status);
4548 if (task->tk_status < 0) {
4549 struct nfs4_exception exception = {
4550 .inode = hdr->inode,
4551 .state = hdr->args.context->state,
4552 .stateid = &hdr->args.stateid,
4553 };
4554 task->tk_status = nfs4_async_handle_exception(task,
4555 server, task->tk_status, &exception);
4556 if (exception.retry) {
4557 rpc_restart_call_prepare(task);
4558 return -EAGAIN;
4559 }
4560 }
4561
4562 if (task->tk_status > 0)
4563 renew_lease(server, hdr->timestamp);
4564 return 0;
4565 }
4566
4567 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4568 struct nfs_pgio_args *args)
4569 {
4570
4571 if (!nfs4_error_stateid_expired(task->tk_status) ||
4572 nfs4_stateid_is_current(&args->stateid,
4573 args->context,
4574 args->lock_context,
4575 FMODE_READ))
4576 return false;
4577 rpc_restart_call_prepare(task);
4578 return true;
4579 }
4580
4581 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4582 {
4583
4584 dprintk("--> %s\n", __func__);
4585
4586 if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4587 return -EAGAIN;
4588 if (nfs4_read_stateid_changed(task, &hdr->args))
4589 return -EAGAIN;
4590 if (task->tk_status > 0)
4591 nfs_invalidate_atime(hdr->inode);
4592 return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4593 nfs4_read_done_cb(task, hdr);
4594 }
4595
4596 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4597 struct rpc_message *msg)
4598 {
4599 hdr->timestamp = jiffies;
4600 if (!hdr->pgio_done_cb)
4601 hdr->pgio_done_cb = nfs4_read_done_cb;
4602 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4603 nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4604 }
4605
4606 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4607 struct nfs_pgio_header *hdr)
4608 {
4609 if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
4610 &hdr->args.seq_args,
4611 &hdr->res.seq_res,
4612 task))
4613 return 0;
4614 if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4615 hdr->args.lock_context,
4616 hdr->rw_ops->rw_mode) == -EIO)
4617 return -EIO;
4618 if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4619 return -EIO;
4620 return 0;
4621 }
4622
4623 static int nfs4_write_done_cb(struct rpc_task *task,
4624 struct nfs_pgio_header *hdr)
4625 {
4626 struct inode *inode = hdr->inode;
4627
4628 trace_nfs4_write(hdr, task->tk_status);
4629 if (task->tk_status < 0) {
4630 struct nfs4_exception exception = {
4631 .inode = hdr->inode,
4632 .state = hdr->args.context->state,
4633 .stateid = &hdr->args.stateid,
4634 };
4635 task->tk_status = nfs4_async_handle_exception(task,
4636 NFS_SERVER(inode), task->tk_status,
4637 &exception);
4638 if (exception.retry) {
4639 rpc_restart_call_prepare(task);
4640 return -EAGAIN;
4641 }
4642 }
4643 if (task->tk_status >= 0) {
4644 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4645 nfs_writeback_update_inode(hdr);
4646 }
4647 return 0;
4648 }
4649
4650 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4651 struct nfs_pgio_args *args)
4652 {
4653
4654 if (!nfs4_error_stateid_expired(task->tk_status) ||
4655 nfs4_stateid_is_current(&args->stateid,
4656 args->context,
4657 args->lock_context,
4658 FMODE_WRITE))
4659 return false;
4660 rpc_restart_call_prepare(task);
4661 return true;
4662 }
4663
4664 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4665 {
4666 if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4667 return -EAGAIN;
4668 if (nfs4_write_stateid_changed(task, &hdr->args))
4669 return -EAGAIN;
4670 return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4671 nfs4_write_done_cb(task, hdr);
4672 }
4673
4674 static
4675 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4676 {
4677 /* Don't request attributes for pNFS or O_DIRECT writes */
4678 if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4679 return false;
4680 /* Otherwise, request attributes if and only if we don't hold
4681 * a delegation
4682 */
4683 return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4684 }
4685
4686 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4687 struct rpc_message *msg)
4688 {
4689 struct nfs_server *server = NFS_SERVER(hdr->inode);
4690
4691 if (!nfs4_write_need_cache_consistency_data(hdr)) {
4692 hdr->args.bitmask = NULL;
4693 hdr->res.fattr = NULL;
4694 } else
4695 hdr->args.bitmask = server->cache_consistency_bitmask;
4696
4697 if (!hdr->pgio_done_cb)
4698 hdr->pgio_done_cb = nfs4_write_done_cb;
4699 hdr->res.server = server;
4700 hdr->timestamp = jiffies;
4701
4702 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4703 nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4704 }
4705
4706 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4707 {
4708 nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
4709 &data->args.seq_args,
4710 &data->res.seq_res,
4711 task);
4712 }
4713
4714 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4715 {
4716 struct inode *inode = data->inode;
4717
4718 trace_nfs4_commit(data, task->tk_status);
4719 if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4720 NULL, NULL) == -EAGAIN) {
4721 rpc_restart_call_prepare(task);
4722 return -EAGAIN;
4723 }
4724 return 0;
4725 }
4726
4727 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4728 {
4729 if (!nfs4_sequence_done(task, &data->res.seq_res))
4730 return -EAGAIN;
4731 return data->commit_done_cb(task, data);
4732 }
4733
4734 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4735 {
4736 struct nfs_server *server = NFS_SERVER(data->inode);
4737
4738 if (data->commit_done_cb == NULL)
4739 data->commit_done_cb = nfs4_commit_done_cb;
4740 data->res.server = server;
4741 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4742 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4743 }
4744
4745 struct nfs4_renewdata {
4746 struct nfs_client *client;
4747 unsigned long timestamp;
4748 };
4749
4750 /*
4751 * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4752 * standalone procedure for queueing an asynchronous RENEW.
4753 */
4754 static void nfs4_renew_release(void *calldata)
4755 {
4756 struct nfs4_renewdata *data = calldata;
4757 struct nfs_client *clp = data->client;
4758
4759 if (atomic_read(&clp->cl_count) > 1)
4760 nfs4_schedule_state_renewal(clp);
4761 nfs_put_client(clp);
4762 kfree(data);
4763 }
4764
4765 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4766 {
4767 struct nfs4_renewdata *data = calldata;
4768 struct nfs_client *clp = data->client;
4769 unsigned long timestamp = data->timestamp;
4770
4771 trace_nfs4_renew_async(clp, task->tk_status);
4772 switch (task->tk_status) {
4773 case 0:
4774 break;
4775 case -NFS4ERR_LEASE_MOVED:
4776 nfs4_schedule_lease_moved_recovery(clp);
4777 break;
4778 default:
4779 /* Unless we're shutting down, schedule state recovery! */
4780 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4781 return;
4782 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4783 nfs4_schedule_lease_recovery(clp);
4784 return;
4785 }
4786 nfs4_schedule_path_down_recovery(clp);
4787 }
4788 do_renew_lease(clp, timestamp);
4789 }
4790
4791 static const struct rpc_call_ops nfs4_renew_ops = {
4792 .rpc_call_done = nfs4_renew_done,
4793 .rpc_release = nfs4_renew_release,
4794 };
4795
4796 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4797 {
4798 struct rpc_message msg = {
4799 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4800 .rpc_argp = clp,
4801 .rpc_cred = cred,
4802 };
4803 struct nfs4_renewdata *data;
4804
4805 if (renew_flags == 0)
4806 return 0;
4807 if (!atomic_inc_not_zero(&clp->cl_count))
4808 return -EIO;
4809 data = kmalloc(sizeof(*data), GFP_NOFS);
4810 if (data == NULL)
4811 return -ENOMEM;
4812 data->client = clp;
4813 data->timestamp = jiffies;
4814 return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4815 &nfs4_renew_ops, data);
4816 }
4817
4818 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4819 {
4820 struct rpc_message msg = {
4821 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4822 .rpc_argp = clp,
4823 .rpc_cred = cred,
4824 };
4825 unsigned long now = jiffies;
4826 int status;
4827
4828 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4829 if (status < 0)
4830 return status;
4831 do_renew_lease(clp, now);
4832 return 0;
4833 }
4834
4835 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4836 {
4837 return server->caps & NFS_CAP_ACLS;
4838 }
4839
4840 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4841 * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4842 * the stack.
4843 */
4844 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4845
4846 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4847 struct page **pages)
4848 {
4849 struct page *newpage, **spages;
4850 int rc = 0;
4851 size_t len;
4852 spages = pages;
4853
4854 do {
4855 len = min_t(size_t, PAGE_SIZE, buflen);
4856 newpage = alloc_page(GFP_KERNEL);
4857
4858 if (newpage == NULL)
4859 goto unwind;
4860 memcpy(page_address(newpage), buf, len);
4861 buf += len;
4862 buflen -= len;
4863 *pages++ = newpage;
4864 rc++;
4865 } while (buflen != 0);
4866
4867 return rc;
4868
4869 unwind:
4870 for(; rc > 0; rc--)
4871 __free_page(spages[rc-1]);
4872 return -ENOMEM;
4873 }
4874
4875 struct nfs4_cached_acl {
4876 int cached;
4877 size_t len;
4878 char data[0];
4879 };
4880
4881 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4882 {
4883 struct nfs_inode *nfsi = NFS_I(inode);
4884
4885 spin_lock(&inode->i_lock);
4886 kfree(nfsi->nfs4_acl);
4887 nfsi->nfs4_acl = acl;
4888 spin_unlock(&inode->i_lock);
4889 }
4890
4891 static void nfs4_zap_acl_attr(struct inode *inode)
4892 {
4893 nfs4_set_cached_acl(inode, NULL);
4894 }
4895
4896 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4897 {
4898 struct nfs_inode *nfsi = NFS_I(inode);
4899 struct nfs4_cached_acl *acl;
4900 int ret = -ENOENT;
4901
4902 spin_lock(&inode->i_lock);
4903 acl = nfsi->nfs4_acl;
4904 if (acl == NULL)
4905 goto out;
4906 if (buf == NULL) /* user is just asking for length */
4907 goto out_len;
4908 if (acl->cached == 0)
4909 goto out;
4910 ret = -ERANGE; /* see getxattr(2) man page */
4911 if (acl->len > buflen)
4912 goto out;
4913 memcpy(buf, acl->data, acl->len);
4914 out_len:
4915 ret = acl->len;
4916 out:
4917 spin_unlock(&inode->i_lock);
4918 return ret;
4919 }
4920
4921 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4922 {
4923 struct nfs4_cached_acl *acl;
4924 size_t buflen = sizeof(*acl) + acl_len;
4925
4926 if (buflen <= PAGE_SIZE) {
4927 acl = kmalloc(buflen, GFP_KERNEL);
4928 if (acl == NULL)
4929 goto out;
4930 acl->cached = 1;
4931 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4932 } else {
4933 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4934 if (acl == NULL)
4935 goto out;
4936 acl->cached = 0;
4937 }
4938 acl->len = acl_len;
4939 out:
4940 nfs4_set_cached_acl(inode, acl);
4941 }
4942
4943 /*
4944 * The getxattr API returns the required buffer length when called with a
4945 * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4946 * the required buf. On a NULL buf, we send a page of data to the server
4947 * guessing that the ACL request can be serviced by a page. If so, we cache
4948 * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4949 * the cache. If not so, we throw away the page, and cache the required
4950 * length. The next getxattr call will then produce another round trip to
4951 * the server, this time with the input buf of the required size.
4952 */
4953 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4954 {
4955 struct page *pages[NFS4ACL_MAXPAGES + 1] = {NULL, };
4956 struct nfs_getaclargs args = {
4957 .fh = NFS_FH(inode),
4958 .acl_pages = pages,
4959 .acl_len = buflen,
4960 };
4961 struct nfs_getaclres res = {
4962 .acl_len = buflen,
4963 };
4964 struct rpc_message msg = {
4965 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4966 .rpc_argp = &args,
4967 .rpc_resp = &res,
4968 };
4969 unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
4970 int ret = -ENOMEM, i;
4971
4972 if (npages > ARRAY_SIZE(pages))
4973 return -ERANGE;
4974
4975 for (i = 0; i < npages; i++) {
4976 pages[i] = alloc_page(GFP_KERNEL);
4977 if (!pages[i])
4978 goto out_free;
4979 }
4980
4981 /* for decoding across pages */
4982 res.acl_scratch = alloc_page(GFP_KERNEL);
4983 if (!res.acl_scratch)
4984 goto out_free;
4985
4986 args.acl_len = npages * PAGE_SIZE;
4987
4988 dprintk("%s buf %p buflen %zu npages %d args.acl_len %zu\n",
4989 __func__, buf, buflen, npages, args.acl_len);
4990 ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4991 &msg, &args.seq_args, &res.seq_res, 0);
4992 if (ret)
4993 goto out_free;
4994
4995 /* Handle the case where the passed-in buffer is too short */
4996 if (res.acl_flags & NFS4_ACL_TRUNC) {
4997 /* Did the user only issue a request for the acl length? */
4998 if (buf == NULL)
4999 goto out_ok;
5000 ret = -ERANGE;
5001 goto out_free;
5002 }
5003 nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5004 if (buf) {
5005 if (res.acl_len > buflen) {
5006 ret = -ERANGE;
5007 goto out_free;
5008 }
5009 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5010 }
5011 out_ok:
5012 ret = res.acl_len;
5013 out_free:
5014 for (i = 0; i < npages; i++)
5015 if (pages[i])
5016 __free_page(pages[i]);
5017 if (res.acl_scratch)
5018 __free_page(res.acl_scratch);
5019 return ret;
5020 }
5021
5022 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5023 {
5024 struct nfs4_exception exception = { };
5025 ssize_t ret;
5026 do {
5027 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5028 trace_nfs4_get_acl(inode, ret);
5029 if (ret >= 0)
5030 break;
5031 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5032 } while (exception.retry);
5033 return ret;
5034 }
5035
5036 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5037 {
5038 struct nfs_server *server = NFS_SERVER(inode);
5039 int ret;
5040
5041 if (!nfs4_server_supports_acls(server))
5042 return -EOPNOTSUPP;
5043 ret = nfs_revalidate_inode(server, inode);
5044 if (ret < 0)
5045 return ret;
5046 if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5047 nfs_zap_acl_cache(inode);
5048 ret = nfs4_read_cached_acl(inode, buf, buflen);
5049 if (ret != -ENOENT)
5050 /* -ENOENT is returned if there is no ACL or if there is an ACL
5051 * but no cached acl data, just the acl length */
5052 return ret;
5053 return nfs4_get_acl_uncached(inode, buf, buflen);
5054 }
5055
5056 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5057 {
5058 struct nfs_server *server = NFS_SERVER(inode);
5059 struct page *pages[NFS4ACL_MAXPAGES];
5060 struct nfs_setaclargs arg = {
5061 .fh = NFS_FH(inode),
5062 .acl_pages = pages,
5063 .acl_len = buflen,
5064 };
5065 struct nfs_setaclres res;
5066 struct rpc_message msg = {
5067 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5068 .rpc_argp = &arg,
5069 .rpc_resp = &res,
5070 };
5071 unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5072 int ret, i;
5073
5074 if (!nfs4_server_supports_acls(server))
5075 return -EOPNOTSUPP;
5076 if (npages > ARRAY_SIZE(pages))
5077 return -ERANGE;
5078 i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5079 if (i < 0)
5080 return i;
5081 nfs4_inode_return_delegation(inode);
5082 ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5083
5084 /*
5085 * Free each page after tx, so the only ref left is
5086 * held by the network stack
5087 */
5088 for (; i > 0; i--)
5089 put_page(pages[i-1]);
5090
5091 /*
5092 * Acl update can result in inode attribute update.
5093 * so mark the attribute cache invalid.
5094 */
5095 spin_lock(&inode->i_lock);
5096 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
5097 spin_unlock(&inode->i_lock);
5098 nfs_access_zap_cache(inode);
5099 nfs_zap_acl_cache(inode);
5100 return ret;
5101 }
5102
5103 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5104 {
5105 struct nfs4_exception exception = { };
5106 int err;
5107 do {
5108 err = __nfs4_proc_set_acl(inode, buf, buflen);
5109 trace_nfs4_set_acl(inode, err);
5110 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5111 &exception);
5112 } while (exception.retry);
5113 return err;
5114 }
5115
5116 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5117 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5118 size_t buflen)
5119 {
5120 struct nfs_server *server = NFS_SERVER(inode);
5121 struct nfs_fattr fattr;
5122 struct nfs4_label label = {0, 0, buflen, buf};
5123
5124 u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5125 struct nfs4_getattr_arg arg = {
5126 .fh = NFS_FH(inode),
5127 .bitmask = bitmask,
5128 };
5129 struct nfs4_getattr_res res = {
5130 .fattr = &fattr,
5131 .label = &label,
5132 .server = server,
5133 };
5134 struct rpc_message msg = {
5135 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5136 .rpc_argp = &arg,
5137 .rpc_resp = &res,
5138 };
5139 int ret;
5140
5141 nfs_fattr_init(&fattr);
5142
5143 ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5144 if (ret)
5145 return ret;
5146 if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5147 return -ENOENT;
5148 if (buflen < label.len)
5149 return -ERANGE;
5150 return 0;
5151 }
5152
5153 static int nfs4_get_security_label(struct inode *inode, void *buf,
5154 size_t buflen)
5155 {
5156 struct nfs4_exception exception = { };
5157 int err;
5158
5159 if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5160 return -EOPNOTSUPP;
5161
5162 do {
5163 err = _nfs4_get_security_label(inode, buf, buflen);
5164 trace_nfs4_get_security_label(inode, err);
5165 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5166 &exception);
5167 } while (exception.retry);
5168 return err;
5169 }
5170
5171 static int _nfs4_do_set_security_label(struct inode *inode,
5172 struct nfs4_label *ilabel,
5173 struct nfs_fattr *fattr,
5174 struct nfs4_label *olabel)
5175 {
5176
5177 struct iattr sattr = {0};
5178 struct nfs_server *server = NFS_SERVER(inode);
5179 const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5180 struct nfs_setattrargs arg = {
5181 .fh = NFS_FH(inode),
5182 .iap = &sattr,
5183 .server = server,
5184 .bitmask = bitmask,
5185 .label = ilabel,
5186 };
5187 struct nfs_setattrres res = {
5188 .fattr = fattr,
5189 .label = olabel,
5190 .server = server,
5191 };
5192 struct rpc_message msg = {
5193 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
5194 .rpc_argp = &arg,
5195 .rpc_resp = &res,
5196 };
5197 int status;
5198
5199 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
5200
5201 status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5202 if (status)
5203 dprintk("%s failed: %d\n", __func__, status);
5204
5205 return status;
5206 }
5207
5208 static int nfs4_do_set_security_label(struct inode *inode,
5209 struct nfs4_label *ilabel,
5210 struct nfs_fattr *fattr,
5211 struct nfs4_label *olabel)
5212 {
5213 struct nfs4_exception exception = { };
5214 int err;
5215
5216 do {
5217 err = _nfs4_do_set_security_label(inode, ilabel,
5218 fattr, olabel);
5219 trace_nfs4_set_security_label(inode, err);
5220 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5221 &exception);
5222 } while (exception.retry);
5223 return err;
5224 }
5225
5226 static int
5227 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
5228 {
5229 struct nfs4_label ilabel, *olabel = NULL;
5230 struct nfs_fattr fattr;
5231 struct rpc_cred *cred;
5232 int status;
5233
5234 if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5235 return -EOPNOTSUPP;
5236
5237 nfs_fattr_init(&fattr);
5238
5239 ilabel.pi = 0;
5240 ilabel.lfs = 0;
5241 ilabel.label = (char *)buf;
5242 ilabel.len = buflen;
5243
5244 cred = rpc_lookup_cred();
5245 if (IS_ERR(cred))
5246 return PTR_ERR(cred);
5247
5248 olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5249 if (IS_ERR(olabel)) {
5250 status = -PTR_ERR(olabel);
5251 goto out;
5252 }
5253
5254 status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5255 if (status == 0)
5256 nfs_setsecurity(inode, &fattr, olabel);
5257
5258 nfs4_label_free(olabel);
5259 out:
5260 put_rpccred(cred);
5261 return status;
5262 }
5263 #endif /* CONFIG_NFS_V4_SECURITY_LABEL */
5264
5265
5266 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5267 nfs4_verifier *bootverf)
5268 {
5269 __be32 verf[2];
5270
5271 if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5272 /* An impossible timestamp guarantees this value
5273 * will never match a generated boot time. */
5274 verf[0] = cpu_to_be32(U32_MAX);
5275 verf[1] = cpu_to_be32(U32_MAX);
5276 } else {
5277 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5278 u64 ns = ktime_to_ns(nn->boot_time);
5279
5280 verf[0] = cpu_to_be32(ns >> 32);
5281 verf[1] = cpu_to_be32(ns);
5282 }
5283 memcpy(bootverf->data, verf, sizeof(bootverf->data));
5284 }
5285
5286 static int
5287 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5288 {
5289 size_t len;
5290 char *str;
5291
5292 if (clp->cl_owner_id != NULL)
5293 return 0;
5294
5295 rcu_read_lock();
5296 len = 14 + strlen(clp->cl_ipaddr) + 1 +
5297 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5298 1 +
5299 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5300 1;
5301 rcu_read_unlock();
5302
5303 if (len > NFS4_OPAQUE_LIMIT + 1)
5304 return -EINVAL;
5305
5306 /*
5307 * Since this string is allocated at mount time, and held until the
5308 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5309 * about a memory-reclaim deadlock.
5310 */
5311 str = kmalloc(len, GFP_KERNEL);
5312 if (!str)
5313 return -ENOMEM;
5314
5315 rcu_read_lock();
5316 scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5317 clp->cl_ipaddr,
5318 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5319 rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5320 rcu_read_unlock();
5321
5322 clp->cl_owner_id = str;
5323 return 0;
5324 }
5325
5326 static int
5327 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5328 {
5329 size_t len;
5330 char *str;
5331
5332 len = 10 + 10 + 1 + 10 + 1 +
5333 strlen(nfs4_client_id_uniquifier) + 1 +
5334 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5335
5336 if (len > NFS4_OPAQUE_LIMIT + 1)
5337 return -EINVAL;
5338
5339 /*
5340 * Since this string is allocated at mount time, and held until the
5341 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5342 * about a memory-reclaim deadlock.
5343 */
5344 str = kmalloc(len, GFP_KERNEL);
5345 if (!str)
5346 return -ENOMEM;
5347
5348 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5349 clp->rpc_ops->version, clp->cl_minorversion,
5350 nfs4_client_id_uniquifier,
5351 clp->cl_rpcclient->cl_nodename);
5352 clp->cl_owner_id = str;
5353 return 0;
5354 }
5355
5356 static int
5357 nfs4_init_uniform_client_string(struct nfs_client *clp)
5358 {
5359 size_t len;
5360 char *str;
5361
5362 if (clp->cl_owner_id != NULL)
5363 return 0;
5364
5365 if (nfs4_client_id_uniquifier[0] != '\0')
5366 return nfs4_init_uniquifier_client_string(clp);
5367
5368 len = 10 + 10 + 1 + 10 + 1 +
5369 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5370
5371 if (len > NFS4_OPAQUE_LIMIT + 1)
5372 return -EINVAL;
5373
5374 /*
5375 * Since this string is allocated at mount time, and held until the
5376 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5377 * about a memory-reclaim deadlock.
5378 */
5379 str = kmalloc(len, GFP_KERNEL);
5380 if (!str)
5381 return -ENOMEM;
5382
5383 scnprintf(str, len, "Linux NFSv%u.%u %s",
5384 clp->rpc_ops->version, clp->cl_minorversion,
5385 clp->cl_rpcclient->cl_nodename);
5386 clp->cl_owner_id = str;
5387 return 0;
5388 }
5389
5390 /*
5391 * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5392 * services. Advertise one based on the address family of the
5393 * clientaddr.
5394 */
5395 static unsigned int
5396 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5397 {
5398 if (strchr(clp->cl_ipaddr, ':') != NULL)
5399 return scnprintf(buf, len, "tcp6");
5400 else
5401 return scnprintf(buf, len, "tcp");
5402 }
5403
5404 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5405 {
5406 struct nfs4_setclientid *sc = calldata;
5407
5408 if (task->tk_status == 0)
5409 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5410 }
5411
5412 static const struct rpc_call_ops nfs4_setclientid_ops = {
5413 .rpc_call_done = nfs4_setclientid_done,
5414 };
5415
5416 /**
5417 * nfs4_proc_setclientid - Negotiate client ID
5418 * @clp: state data structure
5419 * @program: RPC program for NFSv4 callback service
5420 * @port: IP port number for NFS4 callback service
5421 * @cred: RPC credential to use for this call
5422 * @res: where to place the result
5423 *
5424 * Returns zero, a negative errno, or a negative NFS4ERR status code.
5425 */
5426 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5427 unsigned short port, struct rpc_cred *cred,
5428 struct nfs4_setclientid_res *res)
5429 {
5430 nfs4_verifier sc_verifier;
5431 struct nfs4_setclientid setclientid = {
5432 .sc_verifier = &sc_verifier,
5433 .sc_prog = program,
5434 .sc_clnt = clp,
5435 };
5436 struct rpc_message msg = {
5437 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5438 .rpc_argp = &setclientid,
5439 .rpc_resp = res,
5440 .rpc_cred = cred,
5441 };
5442 struct rpc_task *task;
5443 struct rpc_task_setup task_setup_data = {
5444 .rpc_client = clp->cl_rpcclient,
5445 .rpc_message = &msg,
5446 .callback_ops = &nfs4_setclientid_ops,
5447 .callback_data = &setclientid,
5448 .flags = RPC_TASK_TIMEOUT,
5449 };
5450 int status;
5451
5452 /* nfs_client_id4 */
5453 nfs4_init_boot_verifier(clp, &sc_verifier);
5454
5455 if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5456 status = nfs4_init_uniform_client_string(clp);
5457 else
5458 status = nfs4_init_nonuniform_client_string(clp);
5459
5460 if (status)
5461 goto out;
5462
5463 /* cb_client4 */
5464 setclientid.sc_netid_len =
5465 nfs4_init_callback_netid(clp,
5466 setclientid.sc_netid,
5467 sizeof(setclientid.sc_netid));
5468 setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5469 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5470 clp->cl_ipaddr, port >> 8, port & 255);
5471
5472 dprintk("NFS call setclientid auth=%s, '%s'\n",
5473 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5474 clp->cl_owner_id);
5475 task = rpc_run_task(&task_setup_data);
5476 if (IS_ERR(task)) {
5477 status = PTR_ERR(task);
5478 goto out;
5479 }
5480 status = task->tk_status;
5481 if (setclientid.sc_cred) {
5482 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5483 put_rpccred(setclientid.sc_cred);
5484 }
5485 rpc_put_task(task);
5486 out:
5487 trace_nfs4_setclientid(clp, status);
5488 dprintk("NFS reply setclientid: %d\n", status);
5489 return status;
5490 }
5491
5492 /**
5493 * nfs4_proc_setclientid_confirm - Confirm client ID
5494 * @clp: state data structure
5495 * @res: result of a previous SETCLIENTID
5496 * @cred: RPC credential to use for this call
5497 *
5498 * Returns zero, a negative errno, or a negative NFS4ERR status code.
5499 */
5500 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5501 struct nfs4_setclientid_res *arg,
5502 struct rpc_cred *cred)
5503 {
5504 struct rpc_message msg = {
5505 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5506 .rpc_argp = arg,
5507 .rpc_cred = cred,
5508 };
5509 int status;
5510
5511 dprintk("NFS call setclientid_confirm auth=%s, (client ID %llx)\n",
5512 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5513 clp->cl_clientid);
5514 status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5515 trace_nfs4_setclientid_confirm(clp, status);
5516 dprintk("NFS reply setclientid_confirm: %d\n", status);
5517 return status;
5518 }
5519
5520 struct nfs4_delegreturndata {
5521 struct nfs4_delegreturnargs args;
5522 struct nfs4_delegreturnres res;
5523 struct nfs_fh fh;
5524 nfs4_stateid stateid;
5525 unsigned long timestamp;
5526 struct {
5527 struct nfs4_layoutreturn_args arg;
5528 struct nfs4_layoutreturn_res res;
5529 struct nfs4_xdr_opaque_data ld_private;
5530 u32 roc_barrier;
5531 bool roc;
5532 } lr;
5533 struct nfs_fattr fattr;
5534 int rpc_status;
5535 struct inode *inode;
5536 };
5537
5538 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5539 {
5540 struct nfs4_delegreturndata *data = calldata;
5541
5542 if (!nfs4_sequence_done(task, &data->res.seq_res))
5543 return;
5544
5545 trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5546
5547 /* Handle Layoutreturn errors */
5548 if (data->args.lr_args && task->tk_status != 0) {
5549 switch(data->res.lr_ret) {
5550 default:
5551 data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5552 break;
5553 case 0:
5554 data->args.lr_args = NULL;
5555 data->res.lr_res = NULL;
5556 break;
5557 case -NFS4ERR_ADMIN_REVOKED:
5558 case -NFS4ERR_DELEG_REVOKED:
5559 case -NFS4ERR_EXPIRED:
5560 case -NFS4ERR_BAD_STATEID:
5561 case -NFS4ERR_OLD_STATEID:
5562 case -NFS4ERR_UNKNOWN_LAYOUTTYPE:
5563 case -NFS4ERR_WRONG_CRED:
5564 data->args.lr_args = NULL;
5565 data->res.lr_res = NULL;
5566 data->res.lr_ret = 0;
5567 rpc_restart_call_prepare(task);
5568 return;
5569 }
5570 }
5571
5572 switch (task->tk_status) {
5573 case 0:
5574 renew_lease(data->res.server, data->timestamp);
5575 break;
5576 case -NFS4ERR_ADMIN_REVOKED:
5577 case -NFS4ERR_DELEG_REVOKED:
5578 case -NFS4ERR_EXPIRED:
5579 nfs4_free_revoked_stateid(data->res.server,
5580 data->args.stateid,
5581 task->tk_msg.rpc_cred);
5582 case -NFS4ERR_BAD_STATEID:
5583 case -NFS4ERR_OLD_STATEID:
5584 case -NFS4ERR_STALE_STATEID:
5585 task->tk_status = 0;
5586 break;
5587 case -NFS4ERR_ACCESS:
5588 if (data->args.bitmask) {
5589 data->args.bitmask = NULL;
5590 data->res.fattr = NULL;
5591 task->tk_status = 0;
5592 rpc_restart_call_prepare(task);
5593 return;
5594 }
5595 default:
5596 if (nfs4_async_handle_error(task, data->res.server,
5597 NULL, NULL) == -EAGAIN) {
5598 rpc_restart_call_prepare(task);
5599 return;
5600 }
5601 }
5602 data->rpc_status = task->tk_status;
5603 }
5604
5605 static void nfs4_delegreturn_release(void *calldata)
5606 {
5607 struct nfs4_delegreturndata *data = calldata;
5608 struct inode *inode = data->inode;
5609
5610 if (inode) {
5611 if (data->lr.roc)
5612 pnfs_roc_release(&data->lr.arg, &data->lr.res,
5613 data->res.lr_ret);
5614 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5615 nfs_iput_and_deactive(inode);
5616 }
5617 kfree(calldata);
5618 }
5619
5620 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5621 {
5622 struct nfs4_delegreturndata *d_data;
5623
5624 d_data = (struct nfs4_delegreturndata *)data;
5625
5626 if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task))
5627 return;
5628
5629 nfs4_setup_sequence(d_data->res.server->nfs_client,
5630 &d_data->args.seq_args,
5631 &d_data->res.seq_res,
5632 task);
5633 }
5634
5635 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5636 .rpc_call_prepare = nfs4_delegreturn_prepare,
5637 .rpc_call_done = nfs4_delegreturn_done,
5638 .rpc_release = nfs4_delegreturn_release,
5639 };
5640
5641 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5642 {
5643 struct nfs4_delegreturndata *data;
5644 struct nfs_server *server = NFS_SERVER(inode);
5645 struct rpc_task *task;
5646 struct rpc_message msg = {
5647 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5648 .rpc_cred = cred,
5649 };
5650 struct rpc_task_setup task_setup_data = {
5651 .rpc_client = server->client,
5652 .rpc_message = &msg,
5653 .callback_ops = &nfs4_delegreturn_ops,
5654 .flags = RPC_TASK_ASYNC,
5655 };
5656 int status = 0;
5657
5658 data = kzalloc(sizeof(*data), GFP_NOFS);
5659 if (data == NULL)
5660 return -ENOMEM;
5661 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5662
5663 nfs4_state_protect(server->nfs_client,
5664 NFS_SP4_MACH_CRED_CLEANUP,
5665 &task_setup_data.rpc_client, &msg);
5666
5667 data->args.fhandle = &data->fh;
5668 data->args.stateid = &data->stateid;
5669 data->args.bitmask = server->cache_consistency_bitmask;
5670 nfs_copy_fh(&data->fh, NFS_FH(inode));
5671 nfs4_stateid_copy(&data->stateid, stateid);
5672 data->res.fattr = &data->fattr;
5673 data->res.server = server;
5674 data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
5675 data->lr.arg.ld_private = &data->lr.ld_private;
5676 nfs_fattr_init(data->res.fattr);
5677 data->timestamp = jiffies;
5678 data->rpc_status = 0;
5679 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred);
5680 data->inode = nfs_igrab_and_active(inode);
5681 if (data->inode) {
5682 if (data->lr.roc) {
5683 data->args.lr_args = &data->lr.arg;
5684 data->res.lr_res = &data->lr.res;
5685 }
5686 } else if (data->lr.roc) {
5687 pnfs_roc_release(&data->lr.arg, &data->lr.res, 0);
5688 data->lr.roc = false;
5689 }
5690
5691 task_setup_data.callback_data = data;
5692 msg.rpc_argp = &data->args;
5693 msg.rpc_resp = &data->res;
5694 task = rpc_run_task(&task_setup_data);
5695 if (IS_ERR(task))
5696 return PTR_ERR(task);
5697 if (!issync)
5698 goto out;
5699 status = rpc_wait_for_completion_task(task);
5700 if (status != 0)
5701 goto out;
5702 status = data->rpc_status;
5703 out:
5704 rpc_put_task(task);
5705 return status;
5706 }
5707
5708 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5709 {
5710 struct nfs_server *server = NFS_SERVER(inode);
5711 struct nfs4_exception exception = { };
5712 int err;
5713 do {
5714 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5715 trace_nfs4_delegreturn(inode, stateid, err);
5716 switch (err) {
5717 case -NFS4ERR_STALE_STATEID:
5718 case -NFS4ERR_EXPIRED:
5719 case 0:
5720 return 0;
5721 }
5722 err = nfs4_handle_exception(server, err, &exception);
5723 } while (exception.retry);
5724 return err;
5725 }
5726
5727 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5728 {
5729 struct inode *inode = state->inode;
5730 struct nfs_server *server = NFS_SERVER(inode);
5731 struct nfs_client *clp = server->nfs_client;
5732 struct nfs_lockt_args arg = {
5733 .fh = NFS_FH(inode),
5734 .fl = request,
5735 };
5736 struct nfs_lockt_res res = {
5737 .denied = request,
5738 };
5739 struct rpc_message msg = {
5740 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5741 .rpc_argp = &arg,
5742 .rpc_resp = &res,
5743 .rpc_cred = state->owner->so_cred,
5744 };
5745 struct nfs4_lock_state *lsp;
5746 int status;
5747
5748 arg.lock_owner.clientid = clp->cl_clientid;
5749 status = nfs4_set_lock_state(state, request);
5750 if (status != 0)
5751 goto out;
5752 lsp = request->fl_u.nfs4_fl.owner;
5753 arg.lock_owner.id = lsp->ls_seqid.owner_id;
5754 arg.lock_owner.s_dev = server->s_dev;
5755 status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5756 switch (status) {
5757 case 0:
5758 request->fl_type = F_UNLCK;
5759 break;
5760 case -NFS4ERR_DENIED:
5761 status = 0;
5762 }
5763 request->fl_ops->fl_release_private(request);
5764 request->fl_ops = NULL;
5765 out:
5766 return status;
5767 }
5768
5769 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5770 {
5771 struct nfs4_exception exception = { };
5772 int err;
5773
5774 do {
5775 err = _nfs4_proc_getlk(state, cmd, request);
5776 trace_nfs4_get_lock(request, state, cmd, err);
5777 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5778 &exception);
5779 } while (exception.retry);
5780 return err;
5781 }
5782
5783 struct nfs4_unlockdata {
5784 struct nfs_locku_args arg;
5785 struct nfs_locku_res res;
5786 struct nfs4_lock_state *lsp;
5787 struct nfs_open_context *ctx;
5788 struct file_lock fl;
5789 struct nfs_server *server;
5790 unsigned long timestamp;
5791 };
5792
5793 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5794 struct nfs_open_context *ctx,
5795 struct nfs4_lock_state *lsp,
5796 struct nfs_seqid *seqid)
5797 {
5798 struct nfs4_unlockdata *p;
5799 struct inode *inode = lsp->ls_state->inode;
5800
5801 p = kzalloc(sizeof(*p), GFP_NOFS);
5802 if (p == NULL)
5803 return NULL;
5804 p->arg.fh = NFS_FH(inode);
5805 p->arg.fl = &p->fl;
5806 p->arg.seqid = seqid;
5807 p->res.seqid = seqid;
5808 p->lsp = lsp;
5809 atomic_inc(&lsp->ls_count);
5810 /* Ensure we don't close file until we're done freeing locks! */
5811 p->ctx = get_nfs_open_context(ctx);
5812 memcpy(&p->fl, fl, sizeof(p->fl));
5813 p->server = NFS_SERVER(inode);
5814 return p;
5815 }
5816
5817 static void nfs4_locku_release_calldata(void *data)
5818 {
5819 struct nfs4_unlockdata *calldata = data;
5820 nfs_free_seqid(calldata->arg.seqid);
5821 nfs4_put_lock_state(calldata->lsp);
5822 put_nfs_open_context(calldata->ctx);
5823 kfree(calldata);
5824 }
5825
5826 static void nfs4_locku_done(struct rpc_task *task, void *data)
5827 {
5828 struct nfs4_unlockdata *calldata = data;
5829
5830 if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5831 return;
5832 switch (task->tk_status) {
5833 case 0:
5834 renew_lease(calldata->server, calldata->timestamp);
5835 locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
5836 if (nfs4_update_lock_stateid(calldata->lsp,
5837 &calldata->res.stateid))
5838 break;
5839 case -NFS4ERR_ADMIN_REVOKED:
5840 case -NFS4ERR_EXPIRED:
5841 nfs4_free_revoked_stateid(calldata->server,
5842 &calldata->arg.stateid,
5843 task->tk_msg.rpc_cred);
5844 case -NFS4ERR_BAD_STATEID:
5845 case -NFS4ERR_OLD_STATEID:
5846 case -NFS4ERR_STALE_STATEID:
5847 if (!nfs4_stateid_match(&calldata->arg.stateid,
5848 &calldata->lsp->ls_stateid))
5849 rpc_restart_call_prepare(task);
5850 break;
5851 default:
5852 if (nfs4_async_handle_error(task, calldata->server,
5853 NULL, NULL) == -EAGAIN)
5854 rpc_restart_call_prepare(task);
5855 }
5856 nfs_release_seqid(calldata->arg.seqid);
5857 }
5858
5859 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5860 {
5861 struct nfs4_unlockdata *calldata = data;
5862
5863 if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5864 goto out_wait;
5865 nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5866 if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5867 /* Note: exit _without_ running nfs4_locku_done */
5868 goto out_no_action;
5869 }
5870 calldata->timestamp = jiffies;
5871 if (nfs4_setup_sequence(calldata->server->nfs_client,
5872 &calldata->arg.seq_args,
5873 &calldata->res.seq_res,
5874 task) != 0)
5875 nfs_release_seqid(calldata->arg.seqid);
5876 return;
5877 out_no_action:
5878 task->tk_action = NULL;
5879 out_wait:
5880 nfs4_sequence_done(task, &calldata->res.seq_res);
5881 }
5882
5883 static const struct rpc_call_ops nfs4_locku_ops = {
5884 .rpc_call_prepare = nfs4_locku_prepare,
5885 .rpc_call_done = nfs4_locku_done,
5886 .rpc_release = nfs4_locku_release_calldata,
5887 };
5888
5889 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5890 struct nfs_open_context *ctx,
5891 struct nfs4_lock_state *lsp,
5892 struct nfs_seqid *seqid)
5893 {
5894 struct nfs4_unlockdata *data;
5895 struct rpc_message msg = {
5896 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5897 .rpc_cred = ctx->cred,
5898 };
5899 struct rpc_task_setup task_setup_data = {
5900 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5901 .rpc_message = &msg,
5902 .callback_ops = &nfs4_locku_ops,
5903 .workqueue = nfsiod_workqueue,
5904 .flags = RPC_TASK_ASYNC,
5905 };
5906
5907 nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5908 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5909
5910 /* Ensure this is an unlock - when canceling a lock, the
5911 * canceled lock is passed in, and it won't be an unlock.
5912 */
5913 fl->fl_type = F_UNLCK;
5914
5915 data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5916 if (data == NULL) {
5917 nfs_free_seqid(seqid);
5918 return ERR_PTR(-ENOMEM);
5919 }
5920
5921 nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5922 msg.rpc_argp = &data->arg;
5923 msg.rpc_resp = &data->res;
5924 task_setup_data.callback_data = data;
5925 return rpc_run_task(&task_setup_data);
5926 }
5927
5928 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5929 {
5930 struct inode *inode = state->inode;
5931 struct nfs4_state_owner *sp = state->owner;
5932 struct nfs_inode *nfsi = NFS_I(inode);
5933 struct nfs_seqid *seqid;
5934 struct nfs4_lock_state *lsp;
5935 struct rpc_task *task;
5936 struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5937 int status = 0;
5938 unsigned char fl_flags = request->fl_flags;
5939
5940 status = nfs4_set_lock_state(state, request);
5941 /* Unlock _before_ we do the RPC call */
5942 request->fl_flags |= FL_EXISTS;
5943 /* Exclude nfs_delegation_claim_locks() */
5944 mutex_lock(&sp->so_delegreturn_mutex);
5945 /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5946 down_read(&nfsi->rwsem);
5947 if (locks_lock_inode_wait(inode, request) == -ENOENT) {
5948 up_read(&nfsi->rwsem);
5949 mutex_unlock(&sp->so_delegreturn_mutex);
5950 goto out;
5951 }
5952 up_read(&nfsi->rwsem);
5953 mutex_unlock(&sp->so_delegreturn_mutex);
5954 if (status != 0)
5955 goto out;
5956 /* Is this a delegated lock? */
5957 lsp = request->fl_u.nfs4_fl.owner;
5958 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
5959 goto out;
5960 alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
5961 seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
5962 status = -ENOMEM;
5963 if (IS_ERR(seqid))
5964 goto out;
5965 task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
5966 status = PTR_ERR(task);
5967 if (IS_ERR(task))
5968 goto out;
5969 status = rpc_wait_for_completion_task(task);
5970 rpc_put_task(task);
5971 out:
5972 request->fl_flags = fl_flags;
5973 trace_nfs4_unlock(request, state, F_SETLK, status);
5974 return status;
5975 }
5976
5977 struct nfs4_lockdata {
5978 struct nfs_lock_args arg;
5979 struct nfs_lock_res res;
5980 struct nfs4_lock_state *lsp;
5981 struct nfs_open_context *ctx;
5982 struct file_lock fl;
5983 unsigned long timestamp;
5984 int rpc_status;
5985 int cancelled;
5986 struct nfs_server *server;
5987 };
5988
5989 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
5990 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
5991 gfp_t gfp_mask)
5992 {
5993 struct nfs4_lockdata *p;
5994 struct inode *inode = lsp->ls_state->inode;
5995 struct nfs_server *server = NFS_SERVER(inode);
5996 struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5997
5998 p = kzalloc(sizeof(*p), gfp_mask);
5999 if (p == NULL)
6000 return NULL;
6001
6002 p->arg.fh = NFS_FH(inode);
6003 p->arg.fl = &p->fl;
6004 p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6005 if (IS_ERR(p->arg.open_seqid))
6006 goto out_free;
6007 alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6008 p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6009 if (IS_ERR(p->arg.lock_seqid))
6010 goto out_free_seqid;
6011 p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6012 p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6013 p->arg.lock_owner.s_dev = server->s_dev;
6014 p->res.lock_seqid = p->arg.lock_seqid;
6015 p->lsp = lsp;
6016 p->server = server;
6017 atomic_inc(&lsp->ls_count);
6018 p->ctx = get_nfs_open_context(ctx);
6019 memcpy(&p->fl, fl, sizeof(p->fl));
6020 return p;
6021 out_free_seqid:
6022 nfs_free_seqid(p->arg.open_seqid);
6023 out_free:
6024 kfree(p);
6025 return NULL;
6026 }
6027
6028 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6029 {
6030 struct nfs4_lockdata *data = calldata;
6031 struct nfs4_state *state = data->lsp->ls_state;
6032
6033 dprintk("%s: begin!\n", __func__);
6034 if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6035 goto out_wait;
6036 /* Do we need to do an open_to_lock_owner? */
6037 if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6038 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6039 goto out_release_lock_seqid;
6040 }
6041 nfs4_stateid_copy(&data->arg.open_stateid,
6042 &state->open_stateid);
6043 data->arg.new_lock_owner = 1;
6044 data->res.open_seqid = data->arg.open_seqid;
6045 } else {
6046 data->arg.new_lock_owner = 0;
6047 nfs4_stateid_copy(&data->arg.lock_stateid,
6048 &data->lsp->ls_stateid);
6049 }
6050 if (!nfs4_valid_open_stateid(state)) {
6051 data->rpc_status = -EBADF;
6052 task->tk_action = NULL;
6053 goto out_release_open_seqid;
6054 }
6055 data->timestamp = jiffies;
6056 if (nfs4_setup_sequence(data->server->nfs_client,
6057 &data->arg.seq_args,
6058 &data->res.seq_res,
6059 task) == 0)
6060 return;
6061 out_release_open_seqid:
6062 nfs_release_seqid(data->arg.open_seqid);
6063 out_release_lock_seqid:
6064 nfs_release_seqid(data->arg.lock_seqid);
6065 out_wait:
6066 nfs4_sequence_done(task, &data->res.seq_res);
6067 dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6068 }
6069
6070 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6071 {
6072 struct nfs4_lockdata *data = calldata;
6073 struct nfs4_lock_state *lsp = data->lsp;
6074
6075 dprintk("%s: begin!\n", __func__);
6076
6077 if (!nfs4_sequence_done(task, &data->res.seq_res))
6078 return;
6079
6080 data->rpc_status = task->tk_status;
6081 switch (task->tk_status) {
6082 case 0:
6083 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6084 data->