~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/nfs/nfs4proc.c

Version: ~ [ linux-5.15-rc1 ] ~ [ linux-5.14.5 ] ~ [ linux-5.13.18 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.66 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.147 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.206 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.246 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.282 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.283 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/file.h>
 42 #include <linux/string.h>
 43 #include <linux/ratelimit.h>
 44 #include <linux/printk.h>
 45 #include <linux/slab.h>
 46 #include <linux/sunrpc/clnt.h>
 47 #include <linux/nfs.h>
 48 #include <linux/nfs4.h>
 49 #include <linux/nfs_fs.h>
 50 #include <linux/nfs_page.h>
 51 #include <linux/nfs_mount.h>
 52 #include <linux/namei.h>
 53 #include <linux/mount.h>
 54 #include <linux/module.h>
 55 #include <linux/xattr.h>
 56 #include <linux/utsname.h>
 57 #include <linux/freezer.h>
 58 
 59 #include "nfs4_fs.h"
 60 #include "delegation.h"
 61 #include "internal.h"
 62 #include "iostat.h"
 63 #include "callback.h"
 64 #include "pnfs.h"
 65 #include "netns.h"
 66 #include "nfs4idmap.h"
 67 #include "nfs4session.h"
 68 #include "fscache.h"
 69 
 70 #include "nfs4trace.h"
 71 
 72 #define NFSDBG_FACILITY         NFSDBG_PROC
 73 
 74 #define NFS4_POLL_RETRY_MIN     (HZ/10)
 75 #define NFS4_POLL_RETRY_MAX     (15*HZ)
 76 
 77 struct nfs4_opendata;
 78 static int _nfs4_proc_open(struct nfs4_opendata *data);
 79 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
 80 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
 81 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
 82 static int nfs4_proc_getattr(struct nfs_server *, struct nfs_fh *, struct nfs_fattr *, struct nfs4_label *label);
 83 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label);
 84 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
 85                             struct nfs_fattr *fattr, struct iattr *sattr,
 86                             struct nfs4_state *state, struct nfs4_label *ilabel,
 87                             struct nfs4_label *olabel);
 88 #ifdef CONFIG_NFS_V4_1
 89 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 90                 struct rpc_cred *);
 91 static int nfs41_free_stateid(struct nfs_server *, nfs4_stateid *,
 92                 struct rpc_cred *);
 93 #endif
 94 
 95 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 96 static inline struct nfs4_label *
 97 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 98         struct iattr *sattr, struct nfs4_label *label)
 99 {
100         int err;
101 
102         if (label == NULL)
103                 return NULL;
104 
105         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
106                 return NULL;
107 
108         err = security_dentry_init_security(dentry, sattr->ia_mode,
109                                 &dentry->d_name, (void **)&label->label, &label->len);
110         if (err == 0)
111                 return label;
112 
113         return NULL;
114 }
115 static inline void
116 nfs4_label_release_security(struct nfs4_label *label)
117 {
118         if (label)
119                 security_release_secctx(label->label, label->len);
120 }
121 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
122 {
123         if (label)
124                 return server->attr_bitmask;
125 
126         return server->attr_bitmask_nl;
127 }
128 #else
129 static inline struct nfs4_label *
130 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
131         struct iattr *sattr, struct nfs4_label *l)
132 { return NULL; }
133 static inline void
134 nfs4_label_release_security(struct nfs4_label *label)
135 { return; }
136 static inline u32 *
137 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
138 { return server->attr_bitmask; }
139 #endif
140 
141 /* Prevent leaks of NFSv4 errors into userland */
142 static int nfs4_map_errors(int err)
143 {
144         if (err >= -1000)
145                 return err;
146         switch (err) {
147         case -NFS4ERR_RESOURCE:
148         case -NFS4ERR_LAYOUTTRYLATER:
149         case -NFS4ERR_RECALLCONFLICT:
150                 return -EREMOTEIO;
151         case -NFS4ERR_WRONGSEC:
152         case -NFS4ERR_WRONG_CRED:
153                 return -EPERM;
154         case -NFS4ERR_BADOWNER:
155         case -NFS4ERR_BADNAME:
156                 return -EINVAL;
157         case -NFS4ERR_SHARE_DENIED:
158                 return -EACCES;
159         case -NFS4ERR_MINOR_VERS_MISMATCH:
160                 return -EPROTONOSUPPORT;
161         case -NFS4ERR_FILE_OPEN:
162                 return -EBUSY;
163         default:
164                 dprintk("%s could not handle NFSv4 error %d\n",
165                                 __func__, -err);
166                 break;
167         }
168         return -EIO;
169 }
170 
171 /*
172  * This is our standard bitmap for GETATTR requests.
173  */
174 const u32 nfs4_fattr_bitmap[3] = {
175         FATTR4_WORD0_TYPE
176         | FATTR4_WORD0_CHANGE
177         | FATTR4_WORD0_SIZE
178         | FATTR4_WORD0_FSID
179         | FATTR4_WORD0_FILEID,
180         FATTR4_WORD1_MODE
181         | FATTR4_WORD1_NUMLINKS
182         | FATTR4_WORD1_OWNER
183         | FATTR4_WORD1_OWNER_GROUP
184         | FATTR4_WORD1_RAWDEV
185         | FATTR4_WORD1_SPACE_USED
186         | FATTR4_WORD1_TIME_ACCESS
187         | FATTR4_WORD1_TIME_METADATA
188         | FATTR4_WORD1_TIME_MODIFY
189         | FATTR4_WORD1_MOUNTED_ON_FILEID,
190 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
191         FATTR4_WORD2_SECURITY_LABEL
192 #endif
193 };
194 
195 static const u32 nfs4_pnfs_open_bitmap[3] = {
196         FATTR4_WORD0_TYPE
197         | FATTR4_WORD0_CHANGE
198         | FATTR4_WORD0_SIZE
199         | FATTR4_WORD0_FSID
200         | FATTR4_WORD0_FILEID,
201         FATTR4_WORD1_MODE
202         | FATTR4_WORD1_NUMLINKS
203         | FATTR4_WORD1_OWNER
204         | FATTR4_WORD1_OWNER_GROUP
205         | FATTR4_WORD1_RAWDEV
206         | FATTR4_WORD1_SPACE_USED
207         | FATTR4_WORD1_TIME_ACCESS
208         | FATTR4_WORD1_TIME_METADATA
209         | FATTR4_WORD1_TIME_MODIFY,
210         FATTR4_WORD2_MDSTHRESHOLD
211 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
212         | FATTR4_WORD2_SECURITY_LABEL
213 #endif
214 };
215 
216 static const u32 nfs4_open_noattr_bitmap[3] = {
217         FATTR4_WORD0_TYPE
218         | FATTR4_WORD0_CHANGE
219         | FATTR4_WORD0_FILEID,
220 };
221 
222 const u32 nfs4_statfs_bitmap[3] = {
223         FATTR4_WORD0_FILES_AVAIL
224         | FATTR4_WORD0_FILES_FREE
225         | FATTR4_WORD0_FILES_TOTAL,
226         FATTR4_WORD1_SPACE_AVAIL
227         | FATTR4_WORD1_SPACE_FREE
228         | FATTR4_WORD1_SPACE_TOTAL
229 };
230 
231 const u32 nfs4_pathconf_bitmap[3] = {
232         FATTR4_WORD0_MAXLINK
233         | FATTR4_WORD0_MAXNAME,
234         0
235 };
236 
237 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
238                         | FATTR4_WORD0_MAXREAD
239                         | FATTR4_WORD0_MAXWRITE
240                         | FATTR4_WORD0_LEASE_TIME,
241                         FATTR4_WORD1_TIME_DELTA
242                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
243                         FATTR4_WORD2_LAYOUT_BLKSIZE
244                         | FATTR4_WORD2_CLONE_BLKSIZE
245 };
246 
247 const u32 nfs4_fs_locations_bitmap[3] = {
248         FATTR4_WORD0_TYPE
249         | FATTR4_WORD0_CHANGE
250         | FATTR4_WORD0_SIZE
251         | FATTR4_WORD0_FSID
252         | FATTR4_WORD0_FILEID
253         | FATTR4_WORD0_FS_LOCATIONS,
254         FATTR4_WORD1_MODE
255         | FATTR4_WORD1_NUMLINKS
256         | FATTR4_WORD1_OWNER
257         | FATTR4_WORD1_OWNER_GROUP
258         | FATTR4_WORD1_RAWDEV
259         | FATTR4_WORD1_SPACE_USED
260         | FATTR4_WORD1_TIME_ACCESS
261         | FATTR4_WORD1_TIME_METADATA
262         | FATTR4_WORD1_TIME_MODIFY
263         | FATTR4_WORD1_MOUNTED_ON_FILEID,
264 };
265 
266 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
267                 struct nfs4_readdir_arg *readdir)
268 {
269         __be32 *start, *p;
270 
271         if (cookie > 2) {
272                 readdir->cookie = cookie;
273                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
274                 return;
275         }
276 
277         readdir->cookie = 0;
278         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
279         if (cookie == 2)
280                 return;
281         
282         /*
283          * NFSv4 servers do not return entries for '.' and '..'
284          * Therefore, we fake these entries here.  We let '.'
285          * have cookie 0 and '..' have cookie 1.  Note that
286          * when talking to the server, we always send cookie 0
287          * instead of 1 or 2.
288          */
289         start = p = kmap_atomic(*readdir->pages);
290         
291         if (cookie == 0) {
292                 *p++ = xdr_one;                                  /* next */
293                 *p++ = xdr_zero;                   /* cookie, first word */
294                 *p++ = xdr_one;                   /* cookie, second word */
295                 *p++ = xdr_one;                             /* entry len */
296                 memcpy(p, ".\0\0\0", 4);                        /* entry */
297                 p++;
298                 *p++ = xdr_one;                         /* bitmap length */
299                 *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
300                 *p++ = htonl(8);              /* attribute buffer length */
301                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
302         }
303         
304         *p++ = xdr_one;                                  /* next */
305         *p++ = xdr_zero;                   /* cookie, first word */
306         *p++ = xdr_two;                   /* cookie, second word */
307         *p++ = xdr_two;                             /* entry len */
308         memcpy(p, "..\0\0", 4);                         /* entry */
309         p++;
310         *p++ = xdr_one;                         /* bitmap length */
311         *p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
312         *p++ = htonl(8);              /* attribute buffer length */
313         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
314 
315         readdir->pgbase = (char *)p - (char *)start;
316         readdir->count -= readdir->pgbase;
317         kunmap_atomic(start);
318 }
319 
320 static long nfs4_update_delay(long *timeout)
321 {
322         long ret;
323         if (!timeout)
324                 return NFS4_POLL_RETRY_MAX;
325         if (*timeout <= 0)
326                 *timeout = NFS4_POLL_RETRY_MIN;
327         if (*timeout > NFS4_POLL_RETRY_MAX)
328                 *timeout = NFS4_POLL_RETRY_MAX;
329         ret = *timeout;
330         *timeout <<= 1;
331         return ret;
332 }
333 
334 static int nfs4_delay(struct rpc_clnt *clnt, long *timeout)
335 {
336         int res = 0;
337 
338         might_sleep();
339 
340         freezable_schedule_timeout_killable_unsafe(
341                 nfs4_update_delay(timeout));
342         if (fatal_signal_pending(current))
343                 res = -ERESTARTSYS;
344         return res;
345 }
346 
347 /* This is the error handling routine for processes that are allowed
348  * to sleep.
349  */
350 static int nfs4_do_handle_exception(struct nfs_server *server,
351                 int errorcode, struct nfs4_exception *exception)
352 {
353         struct nfs_client *clp = server->nfs_client;
354         struct nfs4_state *state = exception->state;
355         struct inode *inode = exception->inode;
356         int ret = errorcode;
357 
358         exception->delay = 0;
359         exception->recovering = 0;
360         exception->retry = 0;
361         switch(errorcode) {
362                 case 0:
363                         return 0;
364                 case -NFS4ERR_OPENMODE:
365                 case -NFS4ERR_DELEG_REVOKED:
366                 case -NFS4ERR_ADMIN_REVOKED:
367                 case -NFS4ERR_BAD_STATEID:
368                         if (inode && nfs_async_inode_return_delegation(inode,
369                                                 NULL) == 0)
370                                 goto wait_on_recovery;
371                         if (state == NULL)
372                                 break;
373                         ret = nfs4_schedule_stateid_recovery(server, state);
374                         if (ret < 0)
375                                 break;
376                         goto wait_on_recovery;
377                 case -NFS4ERR_EXPIRED:
378                         if (state != NULL) {
379                                 ret = nfs4_schedule_stateid_recovery(server, state);
380                                 if (ret < 0)
381                                         break;
382                         }
383                 case -NFS4ERR_STALE_STATEID:
384                 case -NFS4ERR_STALE_CLIENTID:
385                         nfs4_schedule_lease_recovery(clp);
386                         goto wait_on_recovery;
387                 case -NFS4ERR_MOVED:
388                         ret = nfs4_schedule_migration_recovery(server);
389                         if (ret < 0)
390                                 break;
391                         goto wait_on_recovery;
392                 case -NFS4ERR_LEASE_MOVED:
393                         nfs4_schedule_lease_moved_recovery(clp);
394                         goto wait_on_recovery;
395 #if defined(CONFIG_NFS_V4_1)
396                 case -NFS4ERR_BADSESSION:
397                 case -NFS4ERR_BADSLOT:
398                 case -NFS4ERR_BAD_HIGH_SLOT:
399                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
400                 case -NFS4ERR_DEADSESSION:
401                 case -NFS4ERR_SEQ_FALSE_RETRY:
402                 case -NFS4ERR_SEQ_MISORDERED:
403                         dprintk("%s ERROR: %d Reset session\n", __func__,
404                                 errorcode);
405                         nfs4_schedule_session_recovery(clp->cl_session, errorcode);
406                         goto wait_on_recovery;
407 #endif /* defined(CONFIG_NFS_V4_1) */
408                 case -NFS4ERR_FILE_OPEN:
409                         if (exception->timeout > HZ) {
410                                 /* We have retried a decent amount, time to
411                                  * fail
412                                  */
413                                 ret = -EBUSY;
414                                 break;
415                         }
416                 case -NFS4ERR_DELAY:
417                         nfs_inc_server_stats(server, NFSIOS_DELAY);
418                 case -NFS4ERR_GRACE:
419                         exception->delay = 1;
420                         return 0;
421 
422                 case -NFS4ERR_RETRY_UNCACHED_REP:
423                 case -NFS4ERR_OLD_STATEID:
424                         exception->retry = 1;
425                         break;
426                 case -NFS4ERR_BADOWNER:
427                         /* The following works around a Linux server bug! */
428                 case -NFS4ERR_BADNAME:
429                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
430                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
431                                 exception->retry = 1;
432                                 printk(KERN_WARNING "NFS: v4 server %s "
433                                                 "does not accept raw "
434                                                 "uid/gids. "
435                                                 "Reenabling the idmapper.\n",
436                                                 server->nfs_client->cl_hostname);
437                         }
438         }
439         /* We failed to handle the error */
440         return nfs4_map_errors(ret);
441 wait_on_recovery:
442         exception->recovering = 1;
443         return 0;
444 }
445 
446 /* This is the error handling routine for processes that are allowed
447  * to sleep.
448  */
449 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
450 {
451         struct nfs_client *clp = server->nfs_client;
452         int ret;
453 
454         ret = nfs4_do_handle_exception(server, errorcode, exception);
455         if (exception->delay) {
456                 ret = nfs4_delay(server->client, &exception->timeout);
457                 goto out_retry;
458         }
459         if (exception->recovering) {
460                 ret = nfs4_wait_clnt_recover(clp);
461                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
462                         return -EIO;
463                 goto out_retry;
464         }
465         return ret;
466 out_retry:
467         if (ret == 0)
468                 exception->retry = 1;
469         return ret;
470 }
471 
472 static int
473 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
474                 int errorcode, struct nfs4_exception *exception)
475 {
476         struct nfs_client *clp = server->nfs_client;
477         int ret;
478 
479         ret = nfs4_do_handle_exception(server, errorcode, exception);
480         if (exception->delay) {
481                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
482                 goto out_retry;
483         }
484         if (exception->recovering) {
485                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
486                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
487                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
488                 goto out_retry;
489         }
490         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
491                 ret = -EIO;
492         return ret;
493 out_retry:
494         if (ret == 0)
495                 exception->retry = 1;
496         return ret;
497 }
498 
499 static int
500 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
501                         struct nfs4_state *state, long *timeout)
502 {
503         struct nfs4_exception exception = {
504                 .state = state,
505         };
506 
507         if (task->tk_status >= 0)
508                 return 0;
509         if (timeout)
510                 exception.timeout = *timeout;
511         task->tk_status = nfs4_async_handle_exception(task, server,
512                         task->tk_status,
513                         &exception);
514         if (exception.delay && timeout)
515                 *timeout = exception.timeout;
516         if (exception.retry)
517                 return -EAGAIN;
518         return 0;
519 }
520 
521 /*
522  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
523  * or 'false' otherwise.
524  */
525 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
526 {
527         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
528 
529         if (flavor == RPC_AUTH_GSS_KRB5I ||
530             flavor == RPC_AUTH_GSS_KRB5P)
531                 return true;
532 
533         return false;
534 }
535 
536 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
537 {
538         spin_lock(&clp->cl_lock);
539         if (time_before(clp->cl_last_renewal,timestamp))
540                 clp->cl_last_renewal = timestamp;
541         spin_unlock(&clp->cl_lock);
542 }
543 
544 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
545 {
546         struct nfs_client *clp = server->nfs_client;
547 
548         if (!nfs4_has_session(clp))
549                 do_renew_lease(clp, timestamp);
550 }
551 
552 struct nfs4_call_sync_data {
553         const struct nfs_server *seq_server;
554         struct nfs4_sequence_args *seq_args;
555         struct nfs4_sequence_res *seq_res;
556 };
557 
558 void nfs4_init_sequence(struct nfs4_sequence_args *args,
559                         struct nfs4_sequence_res *res, int cache_reply)
560 {
561         args->sa_slot = NULL;
562         args->sa_cache_this = cache_reply;
563         args->sa_privileged = 0;
564 
565         res->sr_slot = NULL;
566 }
567 
568 static void nfs4_set_sequence_privileged(struct nfs4_sequence_args *args)
569 {
570         args->sa_privileged = 1;
571 }
572 
573 int nfs40_setup_sequence(struct nfs4_slot_table *tbl,
574                          struct nfs4_sequence_args *args,
575                          struct nfs4_sequence_res *res,
576                          struct rpc_task *task)
577 {
578         struct nfs4_slot *slot;
579 
580         /* slot already allocated? */
581         if (res->sr_slot != NULL)
582                 goto out_start;
583 
584         spin_lock(&tbl->slot_tbl_lock);
585         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
586                 goto out_sleep;
587 
588         slot = nfs4_alloc_slot(tbl);
589         if (IS_ERR(slot)) {
590                 if (slot == ERR_PTR(-ENOMEM))
591                         task->tk_timeout = HZ >> 2;
592                 goto out_sleep;
593         }
594         spin_unlock(&tbl->slot_tbl_lock);
595 
596         args->sa_slot = slot;
597         res->sr_slot = slot;
598 
599 out_start:
600         rpc_call_start(task);
601         return 0;
602 
603 out_sleep:
604         if (args->sa_privileged)
605                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
606                                 NULL, RPC_PRIORITY_PRIVILEGED);
607         else
608                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
609         spin_unlock(&tbl->slot_tbl_lock);
610         return -EAGAIN;
611 }
612 EXPORT_SYMBOL_GPL(nfs40_setup_sequence);
613 
614 static int nfs40_sequence_done(struct rpc_task *task,
615                                struct nfs4_sequence_res *res)
616 {
617         struct nfs4_slot *slot = res->sr_slot;
618         struct nfs4_slot_table *tbl;
619 
620         if (slot == NULL)
621                 goto out;
622 
623         tbl = slot->table;
624         spin_lock(&tbl->slot_tbl_lock);
625         if (!nfs41_wake_and_assign_slot(tbl, slot))
626                 nfs4_free_slot(tbl, slot);
627         spin_unlock(&tbl->slot_tbl_lock);
628 
629         res->sr_slot = NULL;
630 out:
631         return 1;
632 }
633 
634 #if defined(CONFIG_NFS_V4_1)
635 
636 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
637 {
638         struct nfs4_session *session;
639         struct nfs4_slot_table *tbl;
640         struct nfs4_slot *slot = res->sr_slot;
641         bool send_new_highest_used_slotid = false;
642 
643         tbl = slot->table;
644         session = tbl->session;
645 
646         spin_lock(&tbl->slot_tbl_lock);
647         /* Be nice to the server: try to ensure that the last transmitted
648          * value for highest_user_slotid <= target_highest_slotid
649          */
650         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
651                 send_new_highest_used_slotid = true;
652 
653         if (nfs41_wake_and_assign_slot(tbl, slot)) {
654                 send_new_highest_used_slotid = false;
655                 goto out_unlock;
656         }
657         nfs4_free_slot(tbl, slot);
658 
659         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
660                 send_new_highest_used_slotid = false;
661 out_unlock:
662         spin_unlock(&tbl->slot_tbl_lock);
663         res->sr_slot = NULL;
664         if (send_new_highest_used_slotid)
665                 nfs41_notify_server(session->clp);
666 }
667 
668 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
669 {
670         struct nfs4_session *session;
671         struct nfs4_slot *slot = res->sr_slot;
672         struct nfs_client *clp;
673         bool interrupted = false;
674         int ret = 1;
675 
676         if (slot == NULL)
677                 goto out_noaction;
678         /* don't increment the sequence number if the task wasn't sent */
679         if (!RPC_WAS_SENT(task))
680                 goto out;
681 
682         session = slot->table->session;
683 
684         if (slot->interrupted) {
685                 slot->interrupted = 0;
686                 interrupted = true;
687         }
688 
689         trace_nfs4_sequence_done(session, res);
690         /* Check the SEQUENCE operation status */
691         switch (res->sr_status) {
692         case 0:
693                 /* Update the slot's sequence and clientid lease timer */
694                 ++slot->seq_nr;
695                 clp = session->clp;
696                 do_renew_lease(clp, res->sr_timestamp);
697                 /* Check sequence flags */
698                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags);
699                 nfs41_update_target_slotid(slot->table, slot, res);
700                 break;
701         case 1:
702                 /*
703                  * sr_status remains 1 if an RPC level error occurred.
704                  * The server may or may not have processed the sequence
705                  * operation..
706                  * Mark the slot as having hosted an interrupted RPC call.
707                  */
708                 slot->interrupted = 1;
709                 goto out;
710         case -NFS4ERR_DELAY:
711                 /* The server detected a resend of the RPC call and
712                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
713                  * of RFC5661.
714                  */
715                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
716                         __func__,
717                         slot->slot_nr,
718                         slot->seq_nr);
719                 goto out_retry;
720         case -NFS4ERR_BADSLOT:
721                 /*
722                  * The slot id we used was probably retired. Try again
723                  * using a different slot id.
724                  */
725                 goto retry_nowait;
726         case -NFS4ERR_SEQ_MISORDERED:
727                 /*
728                  * Was the last operation on this sequence interrupted?
729                  * If so, retry after bumping the sequence number.
730                  */
731                 if (interrupted) {
732                         ++slot->seq_nr;
733                         goto retry_nowait;
734                 }
735                 /*
736                  * Could this slot have been previously retired?
737                  * If so, then the server may be expecting seq_nr = 1!
738                  */
739                 if (slot->seq_nr != 1) {
740                         slot->seq_nr = 1;
741                         goto retry_nowait;
742                 }
743                 break;
744         case -NFS4ERR_SEQ_FALSE_RETRY:
745                 ++slot->seq_nr;
746                 goto retry_nowait;
747         default:
748                 /* Just update the slot sequence no. */
749                 ++slot->seq_nr;
750         }
751 out:
752         /* The session may be reset by one of the error handlers. */
753         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
754         nfs41_sequence_free_slot(res);
755 out_noaction:
756         return ret;
757 retry_nowait:
758         if (rpc_restart_call_prepare(task)) {
759                 task->tk_status = 0;
760                 ret = 0;
761         }
762         goto out;
763 out_retry:
764         if (!rpc_restart_call(task))
765                 goto out;
766         rpc_delay(task, NFS4_POLL_RETRY_MAX);
767         return 0;
768 }
769 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
770 
771 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
772 {
773         if (res->sr_slot == NULL)
774                 return 1;
775         if (!res->sr_slot->table->session)
776                 return nfs40_sequence_done(task, res);
777         return nfs41_sequence_done(task, res);
778 }
779 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
780 
781 int nfs41_setup_sequence(struct nfs4_session *session,
782                                 struct nfs4_sequence_args *args,
783                                 struct nfs4_sequence_res *res,
784                                 struct rpc_task *task)
785 {
786         struct nfs4_slot *slot;
787         struct nfs4_slot_table *tbl;
788 
789         dprintk("--> %s\n", __func__);
790         /* slot already allocated? */
791         if (res->sr_slot != NULL)
792                 goto out_success;
793 
794         tbl = &session->fc_slot_table;
795 
796         task->tk_timeout = 0;
797 
798         spin_lock(&tbl->slot_tbl_lock);
799         if (test_bit(NFS4_SLOT_TBL_DRAINING, &tbl->slot_tbl_state) &&
800             !args->sa_privileged) {
801                 /* The state manager will wait until the slot table is empty */
802                 dprintk("%s session is draining\n", __func__);
803                 goto out_sleep;
804         }
805 
806         slot = nfs4_alloc_slot(tbl);
807         if (IS_ERR(slot)) {
808                 /* If out of memory, try again in 1/4 second */
809                 if (slot == ERR_PTR(-ENOMEM))
810                         task->tk_timeout = HZ >> 2;
811                 dprintk("<-- %s: no free slots\n", __func__);
812                 goto out_sleep;
813         }
814         spin_unlock(&tbl->slot_tbl_lock);
815 
816         args->sa_slot = slot;
817 
818         dprintk("<-- %s slotid=%u seqid=%u\n", __func__,
819                         slot->slot_nr, slot->seq_nr);
820 
821         res->sr_slot = slot;
822         res->sr_timestamp = jiffies;
823         res->sr_status_flags = 0;
824         /*
825          * sr_status is only set in decode_sequence, and so will remain
826          * set to 1 if an rpc level failure occurs.
827          */
828         res->sr_status = 1;
829         trace_nfs4_setup_sequence(session, args);
830 out_success:
831         rpc_call_start(task);
832         return 0;
833 out_sleep:
834         /* Privileged tasks are queued with top priority */
835         if (args->sa_privileged)
836                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
837                                 NULL, RPC_PRIORITY_PRIVILEGED);
838         else
839                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
840         spin_unlock(&tbl->slot_tbl_lock);
841         return -EAGAIN;
842 }
843 EXPORT_SYMBOL_GPL(nfs41_setup_sequence);
844 
845 static int nfs4_setup_sequence(const struct nfs_server *server,
846                                struct nfs4_sequence_args *args,
847                                struct nfs4_sequence_res *res,
848                                struct rpc_task *task)
849 {
850         struct nfs4_session *session = nfs4_get_session(server);
851         int ret = 0;
852 
853         if (!session)
854                 return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
855                                             args, res, task);
856 
857         dprintk("--> %s clp %p session %p sr_slot %u\n",
858                 __func__, session->clp, session, res->sr_slot ?
859                         res->sr_slot->slot_nr : NFS4_NO_SLOT);
860 
861         ret = nfs41_setup_sequence(session, args, res, task);
862 
863         dprintk("<-- %s status=%d\n", __func__, ret);
864         return ret;
865 }
866 
867 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
868 {
869         struct nfs4_call_sync_data *data = calldata;
870         struct nfs4_session *session = nfs4_get_session(data->seq_server);
871 
872         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
873 
874         nfs41_setup_sequence(session, data->seq_args, data->seq_res, task);
875 }
876 
877 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
878 {
879         struct nfs4_call_sync_data *data = calldata;
880 
881         nfs41_sequence_done(task, data->seq_res);
882 }
883 
884 static const struct rpc_call_ops nfs41_call_sync_ops = {
885         .rpc_call_prepare = nfs41_call_sync_prepare,
886         .rpc_call_done = nfs41_call_sync_done,
887 };
888 
889 #else   /* !CONFIG_NFS_V4_1 */
890 
891 static int nfs4_setup_sequence(const struct nfs_server *server,
892                                struct nfs4_sequence_args *args,
893                                struct nfs4_sequence_res *res,
894                                struct rpc_task *task)
895 {
896         return nfs40_setup_sequence(server->nfs_client->cl_slot_tbl,
897                                     args, res, task);
898 }
899 
900 int nfs4_sequence_done(struct rpc_task *task,
901                        struct nfs4_sequence_res *res)
902 {
903         return nfs40_sequence_done(task, res);
904 }
905 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
906 
907 #endif  /* !CONFIG_NFS_V4_1 */
908 
909 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
910 {
911         struct nfs4_call_sync_data *data = calldata;
912         nfs4_setup_sequence(data->seq_server,
913                                 data->seq_args, data->seq_res, task);
914 }
915 
916 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
917 {
918         struct nfs4_call_sync_data *data = calldata;
919         nfs4_sequence_done(task, data->seq_res);
920 }
921 
922 static const struct rpc_call_ops nfs40_call_sync_ops = {
923         .rpc_call_prepare = nfs40_call_sync_prepare,
924         .rpc_call_done = nfs40_call_sync_done,
925 };
926 
927 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
928                                    struct nfs_server *server,
929                                    struct rpc_message *msg,
930                                    struct nfs4_sequence_args *args,
931                                    struct nfs4_sequence_res *res)
932 {
933         int ret;
934         struct rpc_task *task;
935         struct nfs_client *clp = server->nfs_client;
936         struct nfs4_call_sync_data data = {
937                 .seq_server = server,
938                 .seq_args = args,
939                 .seq_res = res,
940         };
941         struct rpc_task_setup task_setup = {
942                 .rpc_client = clnt,
943                 .rpc_message = msg,
944                 .callback_ops = clp->cl_mvops->call_sync_ops,
945                 .callback_data = &data
946         };
947 
948         task = rpc_run_task(&task_setup);
949         if (IS_ERR(task))
950                 ret = PTR_ERR(task);
951         else {
952                 ret = task->tk_status;
953                 rpc_put_task(task);
954         }
955         return ret;
956 }
957 
958 int nfs4_call_sync(struct rpc_clnt *clnt,
959                    struct nfs_server *server,
960                    struct rpc_message *msg,
961                    struct nfs4_sequence_args *args,
962                    struct nfs4_sequence_res *res,
963                    int cache_reply)
964 {
965         nfs4_init_sequence(args, res, cache_reply);
966         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
967 }
968 
969 static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo)
970 {
971         struct nfs_inode *nfsi = NFS_I(dir);
972 
973         spin_lock(&dir->i_lock);
974         nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
975         if (!cinfo->atomic || cinfo->before != dir->i_version)
976                 nfs_force_lookup_revalidate(dir);
977         dir->i_version = cinfo->after;
978         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
979         nfs_fscache_invalidate(dir);
980         spin_unlock(&dir->i_lock);
981 }
982 
983 struct nfs4_opendata {
984         struct kref kref;
985         struct nfs_openargs o_arg;
986         struct nfs_openres o_res;
987         struct nfs_open_confirmargs c_arg;
988         struct nfs_open_confirmres c_res;
989         struct nfs4_string owner_name;
990         struct nfs4_string group_name;
991         struct nfs4_label *a_label;
992         struct nfs_fattr f_attr;
993         struct nfs4_label *f_label;
994         struct dentry *dir;
995         struct dentry *dentry;
996         struct nfs4_state_owner *owner;
997         struct nfs4_state *state;
998         struct iattr attrs;
999         unsigned long timestamp;
1000         unsigned int rpc_done : 1;
1001         unsigned int file_created : 1;
1002         unsigned int is_recover : 1;
1003         int rpc_status;
1004         int cancelled;
1005 };
1006 
1007 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1008                 int err, struct nfs4_exception *exception)
1009 {
1010         if (err != -EINVAL)
1011                 return false;
1012         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1013                 return false;
1014         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1015         exception->retry = 1;
1016         return true;
1017 }
1018 
1019 static u32
1020 nfs4_map_atomic_open_share(struct nfs_server *server,
1021                 fmode_t fmode, int openflags)
1022 {
1023         u32 res = 0;
1024 
1025         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1026         case FMODE_READ:
1027                 res = NFS4_SHARE_ACCESS_READ;
1028                 break;
1029         case FMODE_WRITE:
1030                 res = NFS4_SHARE_ACCESS_WRITE;
1031                 break;
1032         case FMODE_READ|FMODE_WRITE:
1033                 res = NFS4_SHARE_ACCESS_BOTH;
1034         }
1035         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1036                 goto out;
1037         /* Want no delegation if we're using O_DIRECT */
1038         if (openflags & O_DIRECT)
1039                 res |= NFS4_SHARE_WANT_NO_DELEG;
1040 out:
1041         return res;
1042 }
1043 
1044 static enum open_claim_type4
1045 nfs4_map_atomic_open_claim(struct nfs_server *server,
1046                 enum open_claim_type4 claim)
1047 {
1048         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1049                 return claim;
1050         switch (claim) {
1051         default:
1052                 return claim;
1053         case NFS4_OPEN_CLAIM_FH:
1054                 return NFS4_OPEN_CLAIM_NULL;
1055         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1056                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1057         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1058                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1059         }
1060 }
1061 
1062 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1063 {
1064         p->o_res.f_attr = &p->f_attr;
1065         p->o_res.f_label = p->f_label;
1066         p->o_res.seqid = p->o_arg.seqid;
1067         p->c_res.seqid = p->c_arg.seqid;
1068         p->o_res.server = p->o_arg.server;
1069         p->o_res.access_request = p->o_arg.access;
1070         nfs_fattr_init(&p->f_attr);
1071         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1072 }
1073 
1074 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1075                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1076                 const struct iattr *attrs,
1077                 struct nfs4_label *label,
1078                 enum open_claim_type4 claim,
1079                 gfp_t gfp_mask)
1080 {
1081         struct dentry *parent = dget_parent(dentry);
1082         struct inode *dir = d_inode(parent);
1083         struct nfs_server *server = NFS_SERVER(dir);
1084         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1085         struct nfs4_opendata *p;
1086 
1087         p = kzalloc(sizeof(*p), gfp_mask);
1088         if (p == NULL)
1089                 goto err;
1090 
1091         p->f_label = nfs4_label_alloc(server, gfp_mask);
1092         if (IS_ERR(p->f_label))
1093                 goto err_free_p;
1094 
1095         p->a_label = nfs4_label_alloc(server, gfp_mask);
1096         if (IS_ERR(p->a_label))
1097                 goto err_free_f;
1098 
1099         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1100         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1101         if (IS_ERR(p->o_arg.seqid))
1102                 goto err_free_label;
1103         nfs_sb_active(dentry->d_sb);
1104         p->dentry = dget(dentry);
1105         p->dir = parent;
1106         p->owner = sp;
1107         atomic_inc(&sp->so_count);
1108         p->o_arg.open_flags = flags;
1109         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1110         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1111                         fmode, flags);
1112         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1113          * will return permission denied for all bits until close */
1114         if (!(flags & O_EXCL)) {
1115                 /* ask server to check for all possible rights as results
1116                  * are cached */
1117                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1118                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_EXECUTE;
1119         }
1120         p->o_arg.clientid = server->nfs_client->cl_clientid;
1121         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1122         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1123         p->o_arg.name = &dentry->d_name;
1124         p->o_arg.server = server;
1125         p->o_arg.bitmask = nfs4_bitmask(server, label);
1126         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1127         p->o_arg.label = nfs4_label_copy(p->a_label, label);
1128         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1129         switch (p->o_arg.claim) {
1130         case NFS4_OPEN_CLAIM_NULL:
1131         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1132         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1133                 p->o_arg.fh = NFS_FH(dir);
1134                 break;
1135         case NFS4_OPEN_CLAIM_PREVIOUS:
1136         case NFS4_OPEN_CLAIM_FH:
1137         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1138         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1139                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1140         }
1141         if (attrs != NULL && attrs->ia_valid != 0) {
1142                 __u32 verf[2];
1143 
1144                 p->o_arg.u.attrs = &p->attrs;
1145                 memcpy(&p->attrs, attrs, sizeof(p->attrs));
1146 
1147                 verf[0] = jiffies;
1148                 verf[1] = current->pid;
1149                 memcpy(p->o_arg.u.verifier.data, verf,
1150                                 sizeof(p->o_arg.u.verifier.data));
1151         }
1152         p->c_arg.fh = &p->o_res.fh;
1153         p->c_arg.stateid = &p->o_res.stateid;
1154         p->c_arg.seqid = p->o_arg.seqid;
1155         nfs4_init_opendata_res(p);
1156         kref_init(&p->kref);
1157         return p;
1158 
1159 err_free_label:
1160         nfs4_label_free(p->a_label);
1161 err_free_f:
1162         nfs4_label_free(p->f_label);
1163 err_free_p:
1164         kfree(p);
1165 err:
1166         dput(parent);
1167         return NULL;
1168 }
1169 
1170 static void nfs4_opendata_free(struct kref *kref)
1171 {
1172         struct nfs4_opendata *p = container_of(kref,
1173                         struct nfs4_opendata, kref);
1174         struct super_block *sb = p->dentry->d_sb;
1175 
1176         nfs_free_seqid(p->o_arg.seqid);
1177         if (p->state != NULL)
1178                 nfs4_put_open_state(p->state);
1179         nfs4_put_state_owner(p->owner);
1180 
1181         nfs4_label_free(p->a_label);
1182         nfs4_label_free(p->f_label);
1183 
1184         dput(p->dir);
1185         dput(p->dentry);
1186         nfs_sb_deactive(sb);
1187         nfs_fattr_free_names(&p->f_attr);
1188         kfree(p->f_attr.mdsthreshold);
1189         kfree(p);
1190 }
1191 
1192 static void nfs4_opendata_put(struct nfs4_opendata *p)
1193 {
1194         if (p != NULL)
1195                 kref_put(&p->kref, nfs4_opendata_free);
1196 }
1197 
1198 static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
1199 {
1200         int ret;
1201 
1202         ret = rpc_wait_for_completion_task(task);
1203         return ret;
1204 }
1205 
1206 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1207                 fmode_t fmode)
1208 {
1209         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1210         case FMODE_READ|FMODE_WRITE:
1211                 return state->n_rdwr != 0;
1212         case FMODE_WRITE:
1213                 return state->n_wronly != 0;
1214         case FMODE_READ:
1215                 return state->n_rdonly != 0;
1216         }
1217         WARN_ON_ONCE(1);
1218         return false;
1219 }
1220 
1221 static int can_open_cached(struct nfs4_state *state, fmode_t mode, int open_mode)
1222 {
1223         int ret = 0;
1224 
1225         if (open_mode & (O_EXCL|O_TRUNC))
1226                 goto out;
1227         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1228                 case FMODE_READ:
1229                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1230                                 && state->n_rdonly != 0;
1231                         break;
1232                 case FMODE_WRITE:
1233                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1234                                 && state->n_wronly != 0;
1235                         break;
1236                 case FMODE_READ|FMODE_WRITE:
1237                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1238                                 && state->n_rdwr != 0;
1239         }
1240 out:
1241         return ret;
1242 }
1243 
1244 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1245                 enum open_claim_type4 claim)
1246 {
1247         if (delegation == NULL)
1248                 return 0;
1249         if ((delegation->type & fmode) != fmode)
1250                 return 0;
1251         if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags))
1252                 return 0;
1253         switch (claim) {
1254         case NFS4_OPEN_CLAIM_NULL:
1255         case NFS4_OPEN_CLAIM_FH:
1256                 break;
1257         case NFS4_OPEN_CLAIM_PREVIOUS:
1258                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1259                         break;
1260         default:
1261                 return 0;
1262         }
1263         nfs_mark_delegation_referenced(delegation);
1264         return 1;
1265 }
1266 
1267 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1268 {
1269         switch (fmode) {
1270                 case FMODE_WRITE:
1271                         state->n_wronly++;
1272                         break;
1273                 case FMODE_READ:
1274                         state->n_rdonly++;
1275                         break;
1276                 case FMODE_READ|FMODE_WRITE:
1277                         state->n_rdwr++;
1278         }
1279         nfs4_state_set_mode_locked(state, state->state | fmode);
1280 }
1281 
1282 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1283 {
1284         struct nfs_client *clp = state->owner->so_server->nfs_client;
1285         bool need_recover = false;
1286 
1287         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1288                 need_recover = true;
1289         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1290                 need_recover = true;
1291         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1292                 need_recover = true;
1293         if (need_recover)
1294                 nfs4_state_mark_reclaim_nograce(clp, state);
1295 }
1296 
1297 static bool nfs_need_update_open_stateid(struct nfs4_state *state,
1298                 nfs4_stateid *stateid)
1299 {
1300         if (test_and_set_bit(NFS_OPEN_STATE, &state->flags) == 0)
1301                 return true;
1302         if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1303                 nfs_test_and_clear_all_open_stateid(state);
1304                 return true;
1305         }
1306         if (nfs4_stateid_is_newer(stateid, &state->open_stateid))
1307                 return true;
1308         return false;
1309 }
1310 
1311 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1312 {
1313         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1314                 return;
1315         if (state->n_wronly)
1316                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1317         if (state->n_rdonly)
1318                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1319         if (state->n_rdwr)
1320                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1321         set_bit(NFS_OPEN_STATE, &state->flags);
1322 }
1323 
1324 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1325                 nfs4_stateid *arg_stateid,
1326                 nfs4_stateid *stateid, fmode_t fmode)
1327 {
1328         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1329         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1330         case FMODE_WRITE:
1331                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1332                 break;
1333         case FMODE_READ:
1334                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1335                 break;
1336         case 0:
1337                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1338                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1339                 clear_bit(NFS_OPEN_STATE, &state->flags);
1340         }
1341         if (stateid == NULL)
1342                 return;
1343         /* Handle races with OPEN */
1344         if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
1345             (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1346             !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
1347                 nfs_resync_open_stateid_locked(state);
1348                 return;
1349         }
1350         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1351                 nfs4_stateid_copy(&state->stateid, stateid);
1352         nfs4_stateid_copy(&state->open_stateid, stateid);
1353 }
1354 
1355 static void nfs_clear_open_stateid(struct nfs4_state *state,
1356         nfs4_stateid *arg_stateid,
1357         nfs4_stateid *stateid, fmode_t fmode)
1358 {
1359         write_seqlock(&state->seqlock);
1360         nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
1361         write_sequnlock(&state->seqlock);
1362         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1363                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1364 }
1365 
1366 static void nfs_set_open_stateid_locked(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode)
1367 {
1368         switch (fmode) {
1369                 case FMODE_READ:
1370                         set_bit(NFS_O_RDONLY_STATE, &state->flags);
1371                         break;
1372                 case FMODE_WRITE:
1373                         set_bit(NFS_O_WRONLY_STATE, &state->flags);
1374                         break;
1375                 case FMODE_READ|FMODE_WRITE:
1376                         set_bit(NFS_O_RDWR_STATE, &state->flags);
1377         }
1378         if (!nfs_need_update_open_stateid(state, stateid))
1379                 return;
1380         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1381                 nfs4_stateid_copy(&state->stateid, stateid);
1382         nfs4_stateid_copy(&state->open_stateid, stateid);
1383 }
1384 
1385 static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, const nfs4_stateid *deleg_stateid, fmode_t fmode)
1386 {
1387         /*
1388          * Protect the call to nfs4_state_set_mode_locked and
1389          * serialise the stateid update
1390          */
1391         spin_lock(&state->owner->so_lock);
1392         write_seqlock(&state->seqlock);
1393         if (deleg_stateid != NULL) {
1394                 nfs4_stateid_copy(&state->stateid, deleg_stateid);
1395                 set_bit(NFS_DELEGATED_STATE, &state->flags);
1396         }
1397         if (open_stateid != NULL)
1398                 nfs_set_open_stateid_locked(state, open_stateid, fmode);
1399         write_sequnlock(&state->seqlock);
1400         update_open_stateflags(state, fmode);
1401         spin_unlock(&state->owner->so_lock);
1402 }
1403 
1404 static int update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_stateid, nfs4_stateid *delegation, fmode_t fmode)
1405 {
1406         struct nfs_inode *nfsi = NFS_I(state->inode);
1407         struct nfs_delegation *deleg_cur;
1408         int ret = 0;
1409 
1410         fmode &= (FMODE_READ|FMODE_WRITE);
1411 
1412         rcu_read_lock();
1413         deleg_cur = rcu_dereference(nfsi->delegation);
1414         if (deleg_cur == NULL)
1415                 goto no_delegation;
1416 
1417         spin_lock(&deleg_cur->lock);
1418         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1419            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1420             (deleg_cur->type & fmode) != fmode)
1421                 goto no_delegation_unlock;
1422 
1423         if (delegation == NULL)
1424                 delegation = &deleg_cur->stateid;
1425         else if (!nfs4_stateid_match(&deleg_cur->stateid, delegation))
1426                 goto no_delegation_unlock;
1427 
1428         nfs_mark_delegation_referenced(deleg_cur);
1429         __update_open_stateid(state, open_stateid, &deleg_cur->stateid, fmode);
1430         ret = 1;
1431 no_delegation_unlock:
1432         spin_unlock(&deleg_cur->lock);
1433 no_delegation:
1434         rcu_read_unlock();
1435 
1436         if (!ret && open_stateid != NULL) {
1437                 __update_open_stateid(state, open_stateid, NULL, fmode);
1438                 ret = 1;
1439         }
1440         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1441                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1442 
1443         return ret;
1444 }
1445 
1446 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1447                 const nfs4_stateid *stateid)
1448 {
1449         struct nfs4_state *state = lsp->ls_state;
1450         bool ret = false;
1451 
1452         spin_lock(&state->state_lock);
1453         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1454                 goto out_noupdate;
1455         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1456                 goto out_noupdate;
1457         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1458         ret = true;
1459 out_noupdate:
1460         spin_unlock(&state->state_lock);
1461         return ret;
1462 }
1463 
1464 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1465 {
1466         struct nfs_delegation *delegation;
1467 
1468         rcu_read_lock();
1469         delegation = rcu_dereference(NFS_I(inode)->delegation);
1470         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1471                 rcu_read_unlock();
1472                 return;
1473         }
1474         rcu_read_unlock();
1475         nfs4_inode_return_delegation(inode);
1476 }
1477 
1478 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1479 {
1480         struct nfs4_state *state = opendata->state;
1481         struct nfs_inode *nfsi = NFS_I(state->inode);
1482         struct nfs_delegation *delegation;
1483         int open_mode = opendata->o_arg.open_flags;
1484         fmode_t fmode = opendata->o_arg.fmode;
1485         enum open_claim_type4 claim = opendata->o_arg.claim;
1486         nfs4_stateid stateid;
1487         int ret = -EAGAIN;
1488 
1489         for (;;) {
1490                 spin_lock(&state->owner->so_lock);
1491                 if (can_open_cached(state, fmode, open_mode)) {
1492                         update_open_stateflags(state, fmode);
1493                         spin_unlock(&state->owner->so_lock);
1494                         goto out_return_state;
1495                 }
1496                 spin_unlock(&state->owner->so_lock);
1497                 rcu_read_lock();
1498                 delegation = rcu_dereference(nfsi->delegation);
1499                 if (!can_open_delegated(delegation, fmode, claim)) {
1500                         rcu_read_unlock();
1501                         break;
1502                 }
1503                 /* Save the delegation */
1504                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1505                 rcu_read_unlock();
1506                 nfs_release_seqid(opendata->o_arg.seqid);
1507                 if (!opendata->is_recover) {
1508                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1509                         if (ret != 0)
1510                                 goto out;
1511                 }
1512                 ret = -EAGAIN;
1513 
1514                 /* Try to update the stateid using the delegation */
1515                 if (update_open_stateid(state, NULL, &stateid, fmode))
1516                         goto out_return_state;
1517         }
1518 out:
1519         return ERR_PTR(ret);
1520 out_return_state:
1521         atomic_inc(&state->count);
1522         return state;
1523 }
1524 
1525 static void
1526 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1527 {
1528         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1529         struct nfs_delegation *delegation;
1530         int delegation_flags = 0;
1531 
1532         rcu_read_lock();
1533         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1534         if (delegation)
1535                 delegation_flags = delegation->flags;
1536         rcu_read_unlock();
1537         switch (data->o_arg.claim) {
1538         default:
1539                 break;
1540         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1541         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1542                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1543                                    "returning a delegation for "
1544                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1545                                    clp->cl_hostname);
1546                 return;
1547         }
1548         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1549                 nfs_inode_set_delegation(state->inode,
1550                                          data->owner->so_cred,
1551                                          &data->o_res);
1552         else
1553                 nfs_inode_reclaim_delegation(state->inode,
1554                                              data->owner->so_cred,
1555                                              &data->o_res);
1556 }
1557 
1558 /*
1559  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1560  * and update the nfs4_state.
1561  */
1562 static struct nfs4_state *
1563 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1564 {
1565         struct inode *inode = data->state->inode;
1566         struct nfs4_state *state = data->state;
1567         int ret;
1568 
1569         if (!data->rpc_done) {
1570                 if (data->rpc_status) {
1571                         ret = data->rpc_status;
1572                         goto err;
1573                 }
1574                 /* cached opens have already been processed */
1575                 goto update;
1576         }
1577 
1578         ret = nfs_refresh_inode(inode, &data->f_attr);
1579         if (ret)
1580                 goto err;
1581 
1582         if (data->o_res.delegation_type != 0)
1583                 nfs4_opendata_check_deleg(data, state);
1584 update:
1585         update_open_stateid(state, &data->o_res.stateid, NULL,
1586                             data->o_arg.fmode);
1587         atomic_inc(&state->count);
1588 
1589         return state;
1590 err:
1591         return ERR_PTR(ret);
1592 
1593 }
1594 
1595 static struct nfs4_state *
1596 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1597 {
1598         struct inode *inode;
1599         struct nfs4_state *state = NULL;
1600         int ret;
1601 
1602         if (!data->rpc_done) {
1603                 state = nfs4_try_open_cached(data);
1604                 trace_nfs4_cached_open(data->state);
1605                 goto out;
1606         }
1607 
1608         ret = -EAGAIN;
1609         if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1610                 goto err;
1611         inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh, &data->f_attr, data->f_label);
1612         ret = PTR_ERR(inode);
1613         if (IS_ERR(inode))
1614                 goto err;
1615         ret = -ENOMEM;
1616         state = nfs4_get_open_state(inode, data->owner);
1617         if (state == NULL)
1618                 goto err_put_inode;
1619         if (data->o_res.delegation_type != 0)
1620                 nfs4_opendata_check_deleg(data, state);
1621         update_open_stateid(state, &data->o_res.stateid, NULL,
1622                         data->o_arg.fmode);
1623         iput(inode);
1624 out:
1625         nfs_release_seqid(data->o_arg.seqid);
1626         return state;
1627 err_put_inode:
1628         iput(inode);
1629 err:
1630         return ERR_PTR(ret);
1631 }
1632 
1633 static struct nfs4_state *
1634 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
1635 {
1636         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
1637                 return _nfs4_opendata_reclaim_to_nfs4_state(data);
1638         return _nfs4_opendata_to_nfs4_state(data);
1639 }
1640 
1641 static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
1642 {
1643         struct nfs_inode *nfsi = NFS_I(state->inode);
1644         struct nfs_open_context *ctx;
1645 
1646         spin_lock(&state->inode->i_lock);
1647         list_for_each_entry(ctx, &nfsi->open_files, list) {
1648                 if (ctx->state != state)
1649                         continue;
1650                 get_nfs_open_context(ctx);
1651                 spin_unlock(&state->inode->i_lock);
1652                 return ctx;
1653         }
1654         spin_unlock(&state->inode->i_lock);
1655         return ERR_PTR(-ENOENT);
1656 }
1657 
1658 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
1659                 struct nfs4_state *state, enum open_claim_type4 claim)
1660 {
1661         struct nfs4_opendata *opendata;
1662 
1663         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
1664                         NULL, NULL, claim, GFP_NOFS);
1665         if (opendata == NULL)
1666                 return ERR_PTR(-ENOMEM);
1667         opendata->state = state;
1668         atomic_inc(&state->count);
1669         return opendata;
1670 }
1671 
1672 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
1673                 fmode_t fmode)
1674 {
1675         struct nfs4_state *newstate;
1676         int ret;
1677 
1678         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
1679                 return 0;
1680         opendata->o_arg.open_flags = 0;
1681         opendata->o_arg.fmode = fmode;
1682         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
1683                         NFS_SB(opendata->dentry->d_sb),
1684                         fmode, 0);
1685         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
1686         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
1687         nfs4_init_opendata_res(opendata);
1688         ret = _nfs4_recover_proc_open(opendata);
1689         if (ret != 0)
1690                 return ret; 
1691         newstate = nfs4_opendata_to_nfs4_state(opendata);
1692         if (IS_ERR(newstate))
1693                 return PTR_ERR(newstate);
1694         if (newstate != opendata->state)
1695                 ret = -ESTALE;
1696         nfs4_close_state(newstate, fmode);
1697         return ret;
1698 }
1699 
1700 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
1701 {
1702         int ret;
1703 
1704         /* Don't trigger recovery in nfs_test_and_clear_all_open_stateid */
1705         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1706         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1707         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1708         /* memory barrier prior to reading state->n_* */
1709         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1710         clear_bit(NFS_OPEN_STATE, &state->flags);
1711         smp_rmb();
1712         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1713         if (ret != 0)
1714                 return ret;
1715         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1716         if (ret != 0)
1717                 return ret;
1718         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
1719         if (ret != 0)
1720                 return ret;
1721         /*
1722          * We may have performed cached opens for all three recoveries.
1723          * Check if we need to update the current stateid.
1724          */
1725         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
1726             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
1727                 write_seqlock(&state->seqlock);
1728                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1729                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1730                 write_sequnlock(&state->seqlock);
1731         }
1732         return 0;
1733 }
1734 
1735 /*
1736  * OPEN_RECLAIM:
1737  *      reclaim state on the server after a reboot.
1738  */
1739 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1740 {
1741         struct nfs_delegation *delegation;
1742         struct nfs4_opendata *opendata;
1743         fmode_t delegation_type = 0;
1744         int status;
1745 
1746         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1747                         NFS4_OPEN_CLAIM_PREVIOUS);
1748         if (IS_ERR(opendata))
1749                 return PTR_ERR(opendata);
1750         rcu_read_lock();
1751         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1752         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
1753                 delegation_type = delegation->type;
1754         rcu_read_unlock();
1755         opendata->o_arg.u.delegation_type = delegation_type;
1756         status = nfs4_open_recover(opendata, state);
1757         nfs4_opendata_put(opendata);
1758         return status;
1759 }
1760 
1761 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
1762 {
1763         struct nfs_server *server = NFS_SERVER(state->inode);
1764         struct nfs4_exception exception = { };
1765         int err;
1766         do {
1767                 err = _nfs4_do_open_reclaim(ctx, state);
1768                 trace_nfs4_open_reclaim(ctx, 0, err);
1769                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
1770                         continue;
1771                 if (err != -NFS4ERR_DELAY)
1772                         break;
1773                 nfs4_handle_exception(server, err, &exception);
1774         } while (exception.retry);
1775         return err;
1776 }
1777 
1778 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
1779 {
1780         struct nfs_open_context *ctx;
1781         int ret;
1782 
1783         ctx = nfs4_state_find_open_context(state);
1784         if (IS_ERR(ctx))
1785                 return -EAGAIN;
1786         ret = nfs4_do_open_reclaim(ctx, state);
1787         put_nfs_open_context(ctx);
1788         return ret;
1789 }
1790 
1791 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
1792 {
1793         switch (err) {
1794                 default:
1795                         printk(KERN_ERR "NFS: %s: unhandled error "
1796                                         "%d.\n", __func__, err);
1797                 case 0:
1798                 case -ENOENT:
1799                 case -EAGAIN:
1800                 case -ESTALE:
1801                         break;
1802                 case -NFS4ERR_BADSESSION:
1803                 case -NFS4ERR_BADSLOT:
1804                 case -NFS4ERR_BAD_HIGH_SLOT:
1805                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
1806                 case -NFS4ERR_DEADSESSION:
1807                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1808                         nfs4_schedule_session_recovery(server->nfs_client->cl_session, err);
1809                         return -EAGAIN;
1810                 case -NFS4ERR_STALE_CLIENTID:
1811                 case -NFS4ERR_STALE_STATEID:
1812                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1813                 case -NFS4ERR_EXPIRED:
1814                         /* Don't recall a delegation if it was lost */
1815                         nfs4_schedule_lease_recovery(server->nfs_client);
1816                         return -EAGAIN;
1817                 case -NFS4ERR_MOVED:
1818                         nfs4_schedule_migration_recovery(server);
1819                         return -EAGAIN;
1820                 case -NFS4ERR_LEASE_MOVED:
1821                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
1822                         return -EAGAIN;
1823                 case -NFS4ERR_DELEG_REVOKED:
1824                 case -NFS4ERR_ADMIN_REVOKED:
1825                 case -NFS4ERR_BAD_STATEID:
1826                 case -NFS4ERR_OPENMODE:
1827                         nfs_inode_find_state_and_recover(state->inode,
1828                                         stateid);
1829                         nfs4_schedule_stateid_recovery(server, state);
1830                         return -EAGAIN;
1831                 case -NFS4ERR_DELAY:
1832                 case -NFS4ERR_GRACE:
1833                         set_bit(NFS_DELEGATED_STATE, &state->flags);
1834                         ssleep(1);
1835                         return -EAGAIN;
1836                 case -ENOMEM:
1837                 case -NFS4ERR_DENIED:
1838                         /* kill_proc(fl->fl_pid, SIGLOST, 1); */
1839                         return 0;
1840         }
1841         return err;
1842 }
1843 
1844 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
1845                 struct nfs4_state *state, const nfs4_stateid *stateid,
1846                 fmode_t type)
1847 {
1848         struct nfs_server *server = NFS_SERVER(state->inode);
1849         struct nfs4_opendata *opendata;
1850         int err = 0;
1851 
1852         opendata = nfs4_open_recoverdata_alloc(ctx, state,
1853                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
1854         if (IS_ERR(opendata))
1855                 return PTR_ERR(opendata);
1856         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
1857         write_seqlock(&state->seqlock);
1858         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1859         write_sequnlock(&state->seqlock);
1860         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1861         switch (type & (FMODE_READ|FMODE_WRITE)) {
1862         case FMODE_READ|FMODE_WRITE:
1863         case FMODE_WRITE:
1864                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
1865                 if (err)
1866                         break;
1867                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
1868                 if (err)
1869                         break;
1870         case FMODE_READ:
1871                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
1872         }
1873         nfs4_opendata_put(opendata);
1874         return nfs4_handle_delegation_recall_error(server, state, stateid, err);
1875 }
1876 
1877 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
1878 {
1879         struct nfs4_opendata *data = calldata;
1880 
1881         nfs40_setup_sequence(data->o_arg.server->nfs_client->cl_slot_tbl,
1882                              &data->c_arg.seq_args, &data->c_res.seq_res, task);
1883 }
1884 
1885 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
1886 {
1887         struct nfs4_opendata *data = calldata;
1888 
1889         nfs40_sequence_done(task, &data->c_res.seq_res);
1890 
1891         data->rpc_status = task->tk_status;
1892         if (data->rpc_status == 0) {
1893                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
1894                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
1895                 renew_lease(data->o_res.server, data->timestamp);
1896                 data->rpc_done = 1;
1897         }
1898 }
1899 
1900 static void nfs4_open_confirm_release(void *calldata)
1901 {
1902         struct nfs4_opendata *data = calldata;
1903         struct nfs4_state *state = NULL;
1904 
1905         /* If this request hasn't been cancelled, do nothing */
1906         if (data->cancelled == 0)
1907                 goto out_free;
1908         /* In case of error, no cleanup! */
1909         if (!data->rpc_done)
1910                 goto out_free;
1911         state = nfs4_opendata_to_nfs4_state(data);
1912         if (!IS_ERR(state))
1913                 nfs4_close_state(state, data->o_arg.fmode);
1914 out_free:
1915         nfs4_opendata_put(data);
1916 }
1917 
1918 static const struct rpc_call_ops nfs4_open_confirm_ops = {
1919         .rpc_call_prepare = nfs4_open_confirm_prepare,
1920         .rpc_call_done = nfs4_open_confirm_done,
1921         .rpc_release = nfs4_open_confirm_release,
1922 };
1923 
1924 /*
1925  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
1926  */
1927 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
1928 {
1929         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
1930         struct rpc_task *task;
1931         struct  rpc_message msg = {
1932                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
1933                 .rpc_argp = &data->c_arg,
1934                 .rpc_resp = &data->c_res,
1935                 .rpc_cred = data->owner->so_cred,
1936         };
1937         struct rpc_task_setup task_setup_data = {
1938                 .rpc_client = server->client,
1939                 .rpc_message = &msg,
1940                 .callback_ops = &nfs4_open_confirm_ops,
1941                 .callback_data = data,
1942                 .workqueue = nfsiod_workqueue,
1943                 .flags = RPC_TASK_ASYNC,
1944         };
1945         int status;
1946 
1947         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1);
1948         kref_get(&data->kref);
1949         data->rpc_done = 0;
1950         data->rpc_status = 0;
1951         data->timestamp = jiffies;
1952         if (data->is_recover)
1953                 nfs4_set_sequence_privileged(&data->c_arg.seq_args);
1954         task = rpc_run_task(&task_setup_data);
1955         if (IS_ERR(task))
1956                 return PTR_ERR(task);
1957         status = nfs4_wait_for_completion_rpc_task(task);
1958         if (status != 0) {
1959                 data->cancelled = 1;
1960                 smp_wmb();
1961         } else
1962                 status = data->rpc_status;
1963         rpc_put_task(task);
1964         return status;
1965 }
1966 
1967 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
1968 {
1969         struct nfs4_opendata *data = calldata;
1970         struct nfs4_state_owner *sp = data->owner;
1971         struct nfs_client *clp = sp->so_server->nfs_client;
1972         enum open_claim_type4 claim = data->o_arg.claim;
1973 
1974         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
1975                 goto out_wait;
1976         /*
1977          * Check if we still need to send an OPEN call, or if we can use
1978          * a delegation instead.
1979          */
1980         if (data->state != NULL) {
1981                 struct nfs_delegation *delegation;
1982 
1983                 if (can_open_cached(data->state, data->o_arg.fmode, data->o_arg.open_flags))
1984                         goto out_no_action;
1985                 rcu_read_lock();
1986                 delegation = rcu_dereference(NFS_I(data->state->inode)->delegation);
1987                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
1988                         goto unlock_no_action;
1989                 rcu_read_unlock();
1990         }
1991         /* Update client id. */
1992         data->o_arg.clientid = clp->cl_clientid;
1993         switch (claim) {
1994         default:
1995                 break;
1996         case NFS4_OPEN_CLAIM_PREVIOUS:
1997         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1998         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1999                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2000         case NFS4_OPEN_CLAIM_FH:
2001                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2002                 nfs_copy_fh(&data->o_res.fh, data->o_arg.fh);
2003         }
2004         data->timestamp = jiffies;
2005         if (nfs4_setup_sequence(data->o_arg.server,
2006                                 &data->o_arg.seq_args,
2007                                 &data->o_res.seq_res,
2008                                 task) != 0)
2009                 nfs_release_seqid(data->o_arg.seqid);
2010 
2011         /* Set the create mode (note dependency on the session type) */
2012         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2013         if (data->o_arg.open_flags & O_EXCL) {
2014                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2015                 if (nfs4_has_persistent_session(clp))
2016                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2017                 else if (clp->cl_mvops->minor_version > 0)
2018                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2019         }
2020         return;
2021 unlock_no_action:
2022         trace_nfs4_cached_open(data->state);
2023         rcu_read_unlock();
2024 out_no_action:
2025         task->tk_action = NULL;
2026 out_wait:
2027         nfs4_sequence_done(task, &data->o_res.seq_res);
2028 }
2029 
2030 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2031 {
2032         struct nfs4_opendata *data = calldata;
2033 
2034         data->rpc_status = task->tk_status;
2035 
2036         if (!nfs4_sequence_done(task, &data->o_res.seq_res))
2037                 return;
2038 
2039         if (task->tk_status == 0) {
2040                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2041                         switch (data->o_res.f_attr->mode & S_IFMT) {
2042                         case S_IFREG:
2043                                 break;
2044                         case S_IFLNK:
2045                                 data->rpc_status = -ELOOP;
2046                                 break;
2047                         case S_IFDIR:
2048                                 data->rpc_status = -EISDIR;
2049                                 break;
2050                         default:
2051                                 data->rpc_status = -ENOTDIR;
2052                         }
2053                 }
2054                 renew_lease(data->o_res.server, data->timestamp);
2055                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2056                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2057         }
2058         data->rpc_done = 1;
2059 }
2060 
2061 static void nfs4_open_release(void *calldata)
2062 {
2063         struct nfs4_opendata *data = calldata;
2064         struct nfs4_state *state = NULL;
2065 
2066         /* If this request hasn't been cancelled, do nothing */
2067         if (data->cancelled == 0)
2068                 goto out_free;
2069         /* In case of error, no cleanup! */
2070         if (data->rpc_status != 0 || !data->rpc_done)
2071                 goto out_free;
2072         /* In case we need an open_confirm, no cleanup! */
2073         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2074                 goto out_free;
2075         state = nfs4_opendata_to_nfs4_state(data);
2076         if (!IS_ERR(state))
2077                 nfs4_close_state(state, data->o_arg.fmode);
2078 out_free:
2079         nfs4_opendata_put(data);
2080 }
2081 
2082 static const struct rpc_call_ops nfs4_open_ops = {
2083         .rpc_call_prepare = nfs4_open_prepare,
2084         .rpc_call_done = nfs4_open_done,
2085         .rpc_release = nfs4_open_release,
2086 };
2087 
2088 static int nfs4_run_open_task(struct nfs4_opendata *data, int isrecover)
2089 {
2090         struct inode *dir = d_inode(data->dir);
2091         struct nfs_server *server = NFS_SERVER(dir);
2092         struct nfs_openargs *o_arg = &data->o_arg;
2093         struct nfs_openres *o_res = &data->o_res;
2094         struct rpc_task *task;
2095         struct rpc_message msg = {
2096                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2097                 .rpc_argp = o_arg,
2098                 .rpc_resp = o_res,
2099                 .rpc_cred = data->owner->so_cred,
2100         };
2101         struct rpc_task_setup task_setup_data = {
2102                 .rpc_client = server->client,
2103                 .rpc_message = &msg,
2104                 .callback_ops = &nfs4_open_ops,
2105                 .callback_data = data,
2106                 .workqueue = nfsiod_workqueue,
2107                 .flags = RPC_TASK_ASYNC,
2108         };
2109         int status;
2110 
2111         nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1);
2112         kref_get(&data->kref);
2113         data->rpc_done = 0;
2114         data->rpc_status = 0;
2115         data->cancelled = 0;
2116         data->is_recover = 0;
2117         if (isrecover) {
2118                 nfs4_set_sequence_privileged(&o_arg->seq_args);
2119                 data->is_recover = 1;
2120         }
2121         task = rpc_run_task(&task_setup_data);
2122         if (IS_ERR(task))
2123                 return PTR_ERR(task);
2124         status = nfs4_wait_for_completion_rpc_task(task);
2125         if (status != 0) {
2126                 data->cancelled = 1;
2127                 smp_wmb();
2128         } else
2129                 status = data->rpc_status;
2130         rpc_put_task(task);
2131 
2132         return status;
2133 }
2134 
2135 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2136 {
2137         struct inode *dir = d_inode(data->dir);
2138         struct nfs_openres *o_res = &data->o_res;
2139         int status;
2140 
2141         status = nfs4_run_open_task(data, 1);
2142         if (status != 0 || !data->rpc_done)
2143                 return status;
2144 
2145         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2146 
2147         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2148                 status = _nfs4_proc_open_confirm(data);
2149                 if (status != 0)
2150                         return status;
2151         }
2152 
2153         return status;
2154 }
2155 
2156 /*
2157  * Additional permission checks in order to distinguish between an
2158  * open for read, and an open for execute. This works around the
2159  * fact that NFSv4 OPEN treats read and execute permissions as being
2160  * the same.
2161  * Note that in the non-execute case, we want to turn off permission
2162  * checking if we just created a new file (POSIX open() semantics).
2163  */
2164 static int nfs4_opendata_access(struct rpc_cred *cred,
2165                                 struct nfs4_opendata *opendata,
2166                                 struct nfs4_state *state, fmode_t fmode,
2167                                 int openflags)
2168 {
2169         struct nfs_access_entry cache;
2170         u32 mask;
2171 
2172         /* access call failed or for some reason the server doesn't
2173          * support any access modes -- defer access call until later */
2174         if (opendata->o_res.access_supported == 0)
2175                 return 0;
2176 
2177         mask = 0;
2178         /*
2179          * Use openflags to check for exec, because fmode won't
2180          * always have FMODE_EXEC set when file open for exec.
2181          */
2182         if (openflags & __FMODE_EXEC) {
2183                 /* ONLY check for exec rights */
2184                 mask = MAY_EXEC;
2185         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2186                 mask = MAY_READ;
2187 
2188         cache.cred = cred;
2189         cache.jiffies = jiffies;
2190         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2191         nfs_access_add_cache(state->inode, &cache);
2192 
2193         if ((mask & ~cache.mask & (MAY_READ | MAY_EXEC)) == 0)
2194                 return 0;
2195 
2196         /* even though OPEN succeeded, access is denied. Close the file */
2197         nfs4_close_state(state, fmode);
2198         return -EACCES;
2199 }
2200 
2201 /*
2202  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2203  */
2204 static int _nfs4_proc_open(struct nfs4_opendata *data)
2205 {
2206         struct inode *dir = d_inode(data->dir);
2207         struct nfs_server *server = NFS_SERVER(dir);
2208         struct nfs_openargs *o_arg = &data->o_arg;
2209         struct nfs_openres *o_res = &data->o_res;
2210         int status;
2211 
2212         status = nfs4_run_open_task(data, 0);
2213         if (!data->rpc_done)
2214                 return status;
2215         if (status != 0) {
2216                 if (status == -NFS4ERR_BADNAME &&
2217                                 !(o_arg->open_flags & O_CREAT))
2218                         return -ENOENT;
2219                 return status;
2220         }
2221 
2222         nfs_fattr_map_and_free_names(server, &data->f_attr);
2223 
2224         if (o_arg->open_flags & O_CREAT) {
2225                 update_changeattr(dir, &o_res->cinfo);
2226                 if (o_arg->open_flags & O_EXCL)
2227                         data->file_created = 1;
2228                 else if (o_res->cinfo.before != o_res->cinfo.after)
2229                         data->file_created = 1;
2230         }
2231         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2232                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2233         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2234                 status = _nfs4_proc_open_confirm(data);
2235                 if (status != 0)
2236                         return status;
2237         }
2238         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
2239                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, o_res->f_label);
2240         return 0;
2241 }
2242 
2243 static int nfs4_recover_expired_lease(struct nfs_server *server)
2244 {
2245         return nfs4_client_recover_expired_lease(server->nfs_client);
2246 }
2247 
2248 /*
2249  * OPEN_EXPIRED:
2250  *      reclaim state on the server after a network partition.
2251  *      Assumes caller holds the appropriate lock
2252  */
2253 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2254 {
2255         struct nfs4_opendata *opendata;
2256         int ret;
2257 
2258         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2259                         NFS4_OPEN_CLAIM_FH);
2260         if (IS_ERR(opendata))
2261                 return PTR_ERR(opendata);
2262         ret = nfs4_open_recover(opendata, state);
2263         if (ret == -ESTALE)
2264                 d_drop(ctx->dentry);
2265         nfs4_opendata_put(opendata);
2266         return ret;
2267 }
2268 
2269 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2270 {
2271         struct nfs_server *server = NFS_SERVER(state->inode);
2272         struct nfs4_exception exception = { };
2273         int err;
2274 
2275         do {
2276                 err = _nfs4_open_expired(ctx, state);
2277                 trace_nfs4_open_expired(ctx, 0, err);
2278                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2279                         continue;
2280                 switch (err) {
2281                 default:
2282                         goto out;
2283                 case -NFS4ERR_GRACE:
2284                 case -NFS4ERR_DELAY:
2285                         nfs4_handle_exception(server, err, &exception);
2286                         err = 0;
2287                 }
2288         } while (exception.retry);
2289 out:
2290         return err;
2291 }
2292 
2293 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2294 {
2295         struct nfs_open_context *ctx;
2296         int ret;
2297 
2298         ctx = nfs4_state_find_open_context(state);
2299         if (IS_ERR(ctx))
2300                 return -EAGAIN;
2301         ret = nfs4_do_open_expired(ctx, state);
2302         put_nfs_open_context(ctx);
2303         return ret;
2304 }
2305 
2306 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state)
2307 {
2308         nfs_remove_bad_delegation(state->inode);
2309         write_seqlock(&state->seqlock);
2310         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2311         write_sequnlock(&state->seqlock);
2312         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2313 }
2314 
2315 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2316 {
2317         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2318                 nfs_finish_clear_delegation_stateid(state);
2319 }
2320 
2321 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2322 {
2323         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2324         nfs40_clear_delegation_stateid(state);
2325         return nfs4_open_expired(sp, state);
2326 }
2327 
2328 #if defined(CONFIG_NFS_V4_1)
2329 static void nfs41_check_delegation_stateid(struct nfs4_state *state)
2330 {
2331         struct nfs_server *server = NFS_SERVER(state->inode);
2332         nfs4_stateid stateid;
2333         struct nfs_delegation *delegation;
2334         struct rpc_cred *cred;
2335         int status;
2336 
2337         /* Get the delegation credential for use by test/free_stateid */
2338         rcu_read_lock();
2339         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2340         if (delegation == NULL) {
2341                 rcu_read_unlock();
2342                 return;
2343         }
2344 
2345         nfs4_stateid_copy(&stateid, &delegation->stateid);
2346         cred = get_rpccred(delegation->cred);
2347         rcu_read_unlock();
2348         status = nfs41_test_stateid(server, &stateid, cred);
2349         trace_nfs4_test_delegation_stateid(state, NULL, status);
2350 
2351         if (status != NFS_OK) {
2352                 /* Free the stateid unless the server explicitly
2353                  * informs us the stateid is unrecognized. */
2354                 if (status != -NFS4ERR_BAD_STATEID)
2355                         nfs41_free_stateid(server, &stateid, cred);
2356                 nfs_finish_clear_delegation_stateid(state);
2357         }
2358 
2359         put_rpccred(cred);
2360 }
2361 
2362 /**
2363  * nfs41_check_open_stateid - possibly free an open stateid
2364  *
2365  * @state: NFSv4 state for an inode
2366  *
2367  * Returns NFS_OK if recovery for this stateid is now finished.
2368  * Otherwise a negative NFS4ERR value is returned.
2369  */
2370 static int nfs41_check_open_stateid(struct nfs4_state *state)
2371 {
2372         struct nfs_server *server = NFS_SERVER(state->inode);
2373         nfs4_stateid *stateid = &state->open_stateid;
2374         struct rpc_cred *cred = state->owner->so_cred;
2375         int status;
2376 
2377         /* If a state reset has been done, test_stateid is unneeded */
2378         if ((test_bit(NFS_O_RDONLY_STATE, &state->flags) == 0) &&
2379             (test_bit(NFS_O_WRONLY_STATE, &state->flags) == 0) &&
2380             (test_bit(NFS_O_RDWR_STATE, &state->flags) == 0))
2381                 return -NFS4ERR_BAD_STATEID;
2382 
2383         status = nfs41_test_stateid(server, stateid, cred);
2384         trace_nfs4_test_open_stateid(state, NULL, status);
2385         if (status != NFS_OK) {
2386                 /* Free the stateid unless the server explicitly
2387                  * informs us the stateid is unrecognized. */
2388                 if (status != -NFS4ERR_BAD_STATEID)
2389                         nfs41_free_stateid(server, stateid, cred);
2390 
2391                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
2392                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
2393                 clear_bit(NFS_O_RDWR_STATE, &state->flags);
2394                 clear_bit(NFS_OPEN_STATE, &state->flags);
2395         }
2396         return status;
2397 }
2398 
2399 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2400 {
2401         int status;
2402 
2403         nfs41_check_delegation_stateid(state);
2404         status = nfs41_check_open_stateid(state);
2405         if (status != NFS_OK)
2406                 status = nfs4_open_expired(sp, state);
2407         return status;
2408 }
2409 #endif
2410 
2411 /*
2412  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2413  * fields corresponding to attributes that were used to store the verifier.
2414  * Make sure we clobber those fields in the later setattr call
2415  */
2416 static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2417                                 struct iattr *sattr, struct nfs4_label **label)
2418 {
2419         const u32 *attrset = opendata->o_res.attrset;
2420 
2421         if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) &&
2422             !(sattr->ia_valid & ATTR_ATIME_SET))
2423                 sattr->ia_valid |= ATTR_ATIME;
2424 
2425         if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) &&
2426             !(sattr->ia_valid & ATTR_MTIME_SET))
2427                 sattr->ia_valid |= ATTR_MTIME;
2428 
2429         /* Except MODE, it seems harmless of setting twice. */
2430         if ((attrset[1] & FATTR4_WORD1_MODE))
2431                 sattr->ia_valid &= ~ATTR_MODE;
2432 
2433         if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL)
2434                 *label = NULL;
2435 }
2436 
2437 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
2438                 fmode_t fmode,
2439                 int flags,
2440                 struct nfs_open_context *ctx)
2441 {
2442         struct nfs4_state_owner *sp = opendata->owner;
2443         struct nfs_server *server = sp->so_server;
2444         struct dentry *dentry;
2445         struct nfs4_state *state;
2446         unsigned int seq;
2447         int ret;
2448 
2449         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
2450 
2451         ret = _nfs4_proc_open(opendata);
2452         if (ret != 0)
2453                 goto out;
2454 
2455         state = nfs4_opendata_to_nfs4_state(opendata);
2456         ret = PTR_ERR(state);
2457         if (IS_ERR(state))
2458                 goto out;
2459         if (server->caps & NFS_CAP_POSIX_LOCK)
2460                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
2461 
2462         dentry = opendata->dentry;
2463         if (d_really_is_negative(dentry)) {
2464                 /* FIXME: Is this d_drop() ever needed? */
2465                 d_drop(dentry);
2466                 dentry = d_add_unique(dentry, igrab(state->inode));
2467                 if (dentry == NULL) {
2468                         dentry = opendata->dentry;
2469                 } else {
2470                         dput(ctx->dentry);
2471                         ctx->dentry = dentry;
2472                 }
2473                 nfs_set_verifier(dentry,
2474                                 nfs_save_change_attribute(d_inode(opendata->dir)));
2475         }
2476 
2477         ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags);
2478         if (ret != 0)
2479                 goto out;
2480 
2481         ctx->state = state;
2482         if (d_inode(dentry) == state->inode) {
2483                 nfs_inode_attach_open_context(ctx);
2484                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
2485                         nfs4_schedule_stateid_recovery(server, state);
2486         }
2487 out:
2488         return ret;
2489 }
2490 
2491 /*
2492  * Returns a referenced nfs4_state
2493  */
2494 static int _nfs4_do_open(struct inode *dir,
2495                         struct nfs_open_context *ctx,
2496                         int flags,
2497                         struct iattr *sattr,
2498                         struct nfs4_label *label,
2499                         int *opened)
2500 {
2501         struct nfs4_state_owner  *sp;
2502         struct nfs4_state     *state = NULL;
2503         struct nfs_server       *server = NFS_SERVER(dir);
2504         struct nfs4_opendata *opendata;
2505         struct dentry *dentry = ctx->dentry;
2506         struct rpc_cred *cred = ctx->cred;
2507         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
2508         fmode_t fmode = ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
2509         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
2510         struct nfs4_label *olabel = NULL;
2511         int status;
2512 
2513         /* Protect against reboot recovery conflicts */
2514         status = -ENOMEM;
2515         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
2516         if (sp == NULL) {
2517                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
2518                 goto out_err;
2519         }
2520         status = nfs4_recover_expired_lease(server);
2521         if (status != 0)
2522                 goto err_put_state_owner;
2523         if (d_really_is_positive(dentry))
2524                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
2525         status = -ENOMEM;
2526         if (d_really_is_positive(dentry))
2527                 claim = NFS4_OPEN_CLAIM_FH;
2528         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags, sattr,
2529                         label, claim, GFP_KERNEL);
2530         if (opendata == NULL)
2531                 goto err_put_state_owner;
2532 
2533         if (label) {
2534                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
2535                 if (IS_ERR(olabel)) {
2536                         status = PTR_ERR(olabel);
2537                         goto err_opendata_put;
2538                 }
2539         }
2540 
2541         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
2542                 if (!opendata->f_attr.mdsthreshold) {
2543                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
2544                         if (!opendata->f_attr.mdsthreshold)
2545                                 goto err_free_label;
2546                 }
2547                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
2548         }
2549         if (d_really_is_positive(dentry))
2550                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
2551 
2552         status = _nfs4_open_and_get_state(opendata, fmode, flags, ctx);
2553         if (status != 0)
2554                 goto err_free_label;
2555         state = ctx->state;
2556 
2557         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
2558             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
2559                 nfs4_exclusive_attrset(opendata, sattr, &label);
2560 
2561                 nfs_fattr_init(opendata->o_res.f_attr);
2562                 status = nfs4_do_setattr(state->inode, cred,
2563                                 opendata->o_res.f_attr, sattr,
2564                                 state, label, olabel);
2565                 if (status == 0) {
2566                         nfs_setattr_update_inode(state->inode, sattr,
2567                                         opendata->o_res.f_attr);
2568                         nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
2569                 }
2570         }
2571         if (opened && opendata->file_created)
2572                 *opened |= FILE_CREATED;
2573 
2574         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
2575                 *ctx_th = opendata->f_attr.mdsthreshold;
2576                 opendata->f_attr.mdsthreshold = NULL;
2577         }
2578 
2579         nfs4_label_free(olabel);
2580 
2581         nfs4_opendata_put(opendata);
2582         nfs4_put_state_owner(sp);
2583         return 0;
2584 err_free_label:
2585         nfs4_label_free(olabel);
2586 err_opendata_put:
2587         nfs4_opendata_put(opendata);
2588 err_put_state_owner:
2589         nfs4_put_state_owner(sp);
2590 out_err:
2591         return status;
2592 }
2593 
2594 
2595 static struct nfs4_state *nfs4_do_open(struct inode *dir,
2596                                         struct nfs_open_context *ctx,
2597                                         int flags,
2598                                         struct iattr *sattr,
2599                                         struct nfs4_label *label,
2600                                         int *opened)
2601 {
2602         struct nfs_server *server = NFS_SERVER(dir);
2603         struct nfs4_exception exception = { };
2604         struct nfs4_state *res;
2605         int status;
2606 
2607         do {
2608                 status = _nfs4_do_open(dir, ctx, flags, sattr, label, opened);
2609                 res = ctx->state;
2610                 trace_nfs4_open_file(ctx, flags, status);
2611                 if (status == 0)
2612                         break;
2613                 /* NOTE: BAD_SEQID means the server and client disagree about the
2614                  * book-keeping w.r.t. state-changing operations
2615                  * (OPEN/CLOSE/LOCK/LOCKU...)
2616                  * It is actually a sign of a bug on the client or on the server.
2617                  *
2618                  * If we receive a BAD_SEQID error in the particular case of
2619                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
2620                  * have unhashed the old state_owner for us, and that we can
2621                  * therefore safely retry using a new one. We should still warn
2622                  * the user though...
2623                  */
2624                 if (status == -NFS4ERR_BAD_SEQID) {
2625                         pr_warn_ratelimited("NFS: v4 server %s "
2626                                         " returned a bad sequence-id error!\n",
2627                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
2628                         exception.retry = 1;
2629                         continue;
2630                 }
2631                 /*
2632                  * BAD_STATEID on OPEN means that the server cancelled our
2633                  * state before it received the OPEN_CONFIRM.
2634                  * Recover by retrying the request as per the discussion
2635                  * on Page 181 of RFC3530.
2636                  */
2637                 if (status == -NFS4ERR_BAD_STATEID) {
2638                         exception.retry = 1;
2639                         continue;
2640                 }
2641                 if (status == -EAGAIN) {
2642                         /* We must have found a delegation */
2643                         exception.retry = 1;
2644                         continue;
2645                 }
2646                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
2647                         continue;
2648                 res = ERR_PTR(nfs4_handle_exception(server,
2649                                         status, &exception));
2650         } while (exception.retry);
2651         return res;
2652 }
2653 
2654 static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2655                             struct nfs_fattr *fattr, struct iattr *sattr,
2656                             struct nfs4_state *state, struct nfs4_label *ilabel,
2657                             struct nfs4_label *olabel)
2658 {
2659         struct nfs_server *server = NFS_SERVER(inode);
2660         struct nfs_setattrargs  arg = {
2661                 .fh             = NFS_FH(inode),
2662                 .iap            = sattr,
2663                 .server         = server,
2664                 .bitmask = server->attr_bitmask,
2665                 .label          = ilabel,
2666         };
2667         struct nfs_setattrres  res = {
2668                 .fattr          = fattr,
2669                 .label          = olabel,
2670                 .server         = server,
2671         };
2672         struct rpc_message msg = {
2673                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
2674                 .rpc_argp       = &arg,
2675                 .rpc_resp       = &res,
2676                 .rpc_cred       = cred,
2677         };
2678         unsigned long timestamp = jiffies;
2679         fmode_t fmode;
2680         bool truncate;
2681         int status;
2682 
2683         arg.bitmask = nfs4_bitmask(server, ilabel);
2684         if (ilabel)
2685                 arg.bitmask = nfs4_bitmask(server, olabel);
2686 
2687         nfs_fattr_init(fattr);
2688 
2689         /* Servers should only apply open mode checks for file size changes */
2690         truncate = (sattr->ia_valid & ATTR_SIZE) ? true : false;
2691         fmode = truncate ? FMODE_WRITE : FMODE_READ;
2692 
2693         if (nfs4_copy_delegation_stateid(&arg.stateid, inode, fmode)) {
2694                 /* Use that stateid */
2695         } else if (truncate && state != NULL) {
2696                 struct nfs_lockowner lockowner = {
2697                         .l_owner = current->files,
2698                         .l_pid = current->tgid,
2699                 };
2700                 if (!nfs4_valid_open_stateid(state))
2701                         return -EBADF;
2702                 if (nfs4_select_rw_stateid(&arg.stateid, state, FMODE_WRITE,
2703                                 &lockowner) == -EIO)
2704                         return -EBADF;
2705         } else
2706                 nfs4_stateid_copy(&arg.stateid, &zero_stateid);
2707 
2708         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
2709         if (status == 0 && state != NULL)
2710                 renew_lease(server, timestamp);
2711         trace_nfs4_setattr(inode, &arg.stateid, status);
2712         return status;
2713 }
2714 
2715 static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
2716                            struct nfs_fattr *fattr, struct iattr *sattr,
2717                            struct nfs4_state *state, struct nfs4_label *ilabel,
2718                            struct nfs4_label *olabel)
2719 {
2720         struct nfs_server *server = NFS_SERVER(inode);
2721         struct nfs4_exception exception = {
2722                 .state = state,
2723                 .inode = inode,
2724         };
2725         int err;
2726         do {
2727                 err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
2728                 switch (err) {
2729                 case -NFS4ERR_OPENMODE:
2730                         if (!(sattr->ia_valid & ATTR_SIZE)) {
2731                                 pr_warn_once("NFSv4: server %s is incorrectly "
2732                                                 "applying open mode checks to "
2733                                                 "a SETATTR that is not "
2734                                                 "changing file size.\n",
2735                                                 server->nfs_client->cl_hostname);
2736                         }
2737                         if (state && !(state->state & FMODE_WRITE)) {
2738                                 err = -EBADF;
2739                                 if (sattr->ia_valid & ATTR_OPEN)
2740                                         err = -EACCES;
2741                                 goto out;
2742                         }
2743                 }
2744                 err = nfs4_handle_exception(server, err, &exception);
2745         } while (exception.retry);
2746 out:
2747         return err;
2748 }
2749 
2750 static bool
2751 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
2752 {
2753         if (inode == NULL || !nfs_have_layout(inode))
2754                 return false;
2755 
2756         return pnfs_wait_on_layoutreturn(inode, task);
2757 }
2758 
2759 struct nfs4_closedata {
2760         struct inode *inode;
2761         struct nfs4_state *state;
2762         struct nfs_closeargs arg;
2763         struct nfs_closeres res;
2764         struct nfs_fattr fattr;
2765         unsigned long timestamp;
2766         bool roc;
2767         u32 roc_barrier;
2768 };
2769 
2770 static void nfs4_free_closedata(void *data)
2771 {
2772         struct nfs4_closedata *calldata = data;
2773         struct nfs4_state_owner *sp = calldata->state->owner;
2774         struct super_block *sb = calldata->state->inode->i_sb;
2775 
2776         if (calldata->roc)
2777                 pnfs_roc_release(calldata->state->inode);
2778         nfs4_put_open_state(calldata->state);
2779         nfs_free_seqid(calldata->arg.seqid);
2780         nfs4_put_state_owner(sp);
2781         nfs_sb_deactive(sb);
2782         kfree(calldata);
2783 }
2784 
2785 static void nfs4_close_done(struct rpc_task *task, void *data)
2786 {
2787         struct nfs4_closedata *calldata = data;
2788         struct nfs4_state *state = calldata->state;
2789         struct nfs_server *server = NFS_SERVER(calldata->inode);
2790         nfs4_stateid *res_stateid = NULL;
2791 
2792         dprintk("%s: begin!\n", __func__);
2793         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
2794                 return;
2795         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
2796         /* hmm. we are done with the inode, and in the process of freeing
2797          * the state_owner. we keep this around to process errors
2798          */
2799         switch (task->tk_status) {
2800                 case 0:
2801                         res_stateid = &calldata->res.stateid;
2802                         if (calldata->roc)
2803                                 pnfs_roc_set_barrier(state->inode,
2804                                                      calldata->roc_barrier);
2805                         renew_lease(server, calldata->timestamp);
2806                         break;
2807                 case -NFS4ERR_ADMIN_REVOKED:
2808                 case -NFS4ERR_STALE_STATEID:
2809                 case -NFS4ERR_OLD_STATEID:
2810                 case -NFS4ERR_BAD_STATEID:
2811                 case -NFS4ERR_EXPIRED:
2812                         if (!nfs4_stateid_match(&calldata->arg.stateid,
2813                                                 &state->open_stateid)) {
2814                                 rpc_restart_call_prepare(task);
2815                                 goto out_release;
2816                         }
2817                         if (calldata->arg.fmode == 0)
2818                                 break;
2819                 default:
2820                         if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN) {
2821                                 rpc_restart_call_prepare(task);
2822                                 goto out_release;
2823                         }
2824         }
2825         nfs_clear_open_stateid(state, &calldata->arg.stateid,
2826                         res_stateid, calldata->arg.fmode);
2827 out_release:
2828         nfs_release_seqid(calldata->arg.seqid);
2829         nfs_refresh_inode(calldata->inode, calldata->res.fattr);
2830         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
2831 }
2832 
2833 static void nfs4_close_prepare(struct rpc_task *task, void *data)
2834 {
2835         struct nfs4_closedata *calldata = data;
2836         struct nfs4_state *state = calldata->state;
2837         struct inode *inode = calldata->inode;
2838         bool is_rdonly, is_wronly, is_rdwr;
2839         int call_close = 0;
2840 
2841         dprintk("%s: begin!\n", __func__);
2842         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
2843                 goto out_wait;
2844 
2845         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
2846         spin_lock(&state->owner->so_lock);
2847         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
2848         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
2849         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
2850         nfs4_stateid_copy(&calldata->arg.stateid, &state->open_stateid);
2851         /* Calculate the change in open mode */
2852         calldata->arg.fmode = 0;
2853         if (state->n_rdwr == 0) {
2854                 if (state->n_rdonly == 0)
2855                         call_close |= is_rdonly;
2856                 else if (is_rdonly)
2857                         calldata->arg.fmode |= FMODE_READ;
2858                 if (state->n_wronly == 0)
2859                         call_close |= is_wronly;
2860                 else if (is_wronly)
2861                         calldata->arg.fmode |= FMODE_WRITE;
2862         } else if (is_rdwr)
2863                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
2864 
2865         if (calldata->arg.fmode == 0)
2866                 call_close |= is_rdwr;
2867 
2868         if (!nfs4_valid_open_stateid(state))
2869                 call_close = 0;
2870         spin_unlock(&state->owner->so_lock);
2871 
2872         if (!call_close) {
2873                 /* Note: exit _without_ calling nfs4_close_done */
2874                 goto out_no_action;
2875         }
2876 
2877         if (nfs4_wait_on_layoutreturn(inode, task)) {
2878                 nfs_release_seqid(calldata->arg.seqid);
2879                 goto out_wait;
2880         }
2881 
2882         if (calldata->arg.fmode == 0)
2883                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
2884         if (calldata->roc)
2885                 pnfs_roc_get_barrier(inode, &calldata->roc_barrier);
2886 
2887         calldata->arg.share_access =
2888                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
2889                                 calldata->arg.fmode, 0);
2890 
2891         nfs_fattr_init(calldata->res.fattr);
2892         calldata->timestamp = jiffies;
2893         if (nfs4_setup_sequence(NFS_SERVER(inode),
2894                                 &calldata->arg.seq_args,
2895                                 &calldata->res.seq_res,
2896                                 task) != 0)
2897                 nfs_release_seqid(calldata->arg.seqid);
2898         dprintk("%s: done!\n", __func__);
2899         return;
2900 out_no_action:
2901         task->tk_action = NULL;
2902 out_wait:
2903         nfs4_sequence_done(task, &calldata->res.seq_res);
2904 }
2905 
2906 static const struct rpc_call_ops nfs4_close_ops = {
2907         .rpc_call_prepare = nfs4_close_prepare,
2908         .rpc_call_done = nfs4_close_done,
2909         .rpc_release = nfs4_free_closedata,
2910 };
2911 
2912 static bool nfs4_roc(struct inode *inode)
2913 {
2914         if (!nfs_have_layout(inode))
2915                 return false;
2916         return pnfs_roc(inode);
2917 }
2918 
2919 /* 
2920  * It is possible for data to be read/written from a mem-mapped file 
2921  * after the sys_close call (which hits the vfs layer as a flush).
2922  * This means that we can't safely call nfsv4 close on a file until 
2923  * the inode is cleared. This in turn means that we are not good
2924  * NFSv4 citizens - we do not indicate to the server to update the file's 
2925  * share state even when we are done with one of the three share 
2926  * stateid's in the inode.
2927  *
2928  * NOTE: Caller must be holding the sp->so_owner semaphore!
2929  */
2930 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
2931 {
2932         struct nfs_server *server = NFS_SERVER(state->inode);
2933         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
2934         struct nfs4_closedata *calldata;
2935         struct nfs4_state_owner *sp = state->owner;
2936         struct rpc_task *task;
2937         struct rpc_message msg = {
2938                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
2939                 .rpc_cred = state->owner->so_cred,
2940         };
2941         struct rpc_task_setup task_setup_data = {
2942                 .rpc_client = server->client,
2943                 .rpc_message = &msg,
2944                 .callback_ops = &nfs4_close_ops,
2945                 .workqueue = nfsiod_workqueue,
2946                 .flags = RPC_TASK_ASYNC,
2947         };
2948         int status = -ENOMEM;
2949 
2950         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
2951                 &task_setup_data.rpc_client, &msg);
2952 
2953         calldata = kzalloc(sizeof(*calldata), gfp_mask);
2954         if (calldata == NULL)
2955                 goto out;
2956         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1);
2957         calldata->inode = state->inode;
2958         calldata->state = state;
2959         calldata->arg.fh = NFS_FH(state->inode);
2960         /* Serialization for the sequence id */
2961         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
2962         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
2963         if (IS_ERR(calldata->arg.seqid))
2964                 goto out_free_calldata;
2965         calldata->arg.fmode = 0;
2966         calldata->arg.bitmask = server->cache_consistency_bitmask;
2967         calldata->res.fattr = &calldata->fattr;
2968         calldata->res.seqid = calldata->arg.seqid;
2969         calldata->res.server = server;
2970         calldata->roc = nfs4_roc(state->inode);
2971         nfs_sb_active(calldata->inode->i_sb);
2972 
2973         msg.rpc_argp = &calldata->arg;
2974         msg.rpc_resp = &calldata->res;
2975         task_setup_data.callback_data = calldata;
2976         task = rpc_run_task(&task_setup_data);
2977         if (IS_ERR(task))
2978                 return PTR_ERR(task);
2979         status = 0;
2980         if (wait)
2981                 status = rpc_wait_for_completion_task(task);
2982         rpc_put_task(task);
2983         return status;
2984 out_free_calldata:
2985         kfree(calldata);
2986 out:
2987         nfs4_put_open_state(state);
2988         nfs4_put_state_owner(sp);
2989         return status;
2990 }
2991 
2992 static struct inode *
2993 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
2994                 int open_flags, struct iattr *attr, int *opened)
2995 {
2996         struct nfs4_state *state;
2997         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
2998 
2999         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3000 
3001         /* Protect against concurrent sillydeletes */
3002         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3003 
3004         nfs4_label_release_security(label);
3005 
3006         if (IS_ERR(state))
3007                 return ERR_CAST(state);
3008         return state->inode;
3009 }
3010 
3011 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3012 {
3013         if (ctx->state == NULL)
3014                 return;
3015         if (is_sync)
3016                 nfs4_close_sync(ctx->state, ctx->mode);
3017         else
3018                 nfs4_close_state(ctx->state, ctx->mode);
3019 }
3020 
3021 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3022 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3023 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_SECURITY_LABEL - 1UL)
3024 
3025 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3026 {
3027         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3028         struct nfs4_server_caps_arg args = {
3029                 .fhandle = fhandle,
3030                 .bitmask = bitmask,
3031         };
3032         struct nfs4_server_caps_res res = {};
3033         struct rpc_message msg = {
3034                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3035                 .rpc_argp = &args,
3036                 .rpc_resp = &res,
3037         };
3038         int status;
3039 
3040         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3041                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3042                      FATTR4_WORD0_LINK_SUPPORT |
3043                      FATTR4_WORD0_SYMLINK_SUPPORT |
3044                      FATTR4_WORD0_ACLSUPPORT;
3045         if (minorversion)
3046                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3047 
3048         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3049         if (status == 0) {
3050                 /* Sanity check the server answers */
3051                 switch (minorversion) {
3052                 case 0:
3053                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3054                         res.attr_bitmask[2] = 0;
3055                         break;
3056                 case 1:
3057                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3058                         break;
3059                 case 2:
3060                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3061                 }
3062                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3063                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3064                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3065                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3066                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3067                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3068                                 NFS_CAP_SECURITY_LABEL);
3069                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3070                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3071                         server->caps |= NFS_CAP_ACLS;
3072                 if (res.has_links != 0)
3073                         server->caps |= NFS_CAP_HARDLINKS;
3074                 if (res.has_symlinks != 0)
3075                         server->caps |= NFS_CAP_SYMLINKS;
3076                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3077                         server->caps |= NFS_CAP_FILEID;
3078                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3079                         server->caps |= NFS_CAP_MODE;
3080                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3081                         server->caps |= NFS_CAP_NLINK;
3082                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3083                         server->caps |= NFS_CAP_OWNER;
3084                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3085                         server->caps |= NFS_CAP_OWNER_GROUP;
3086                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3087                         server->caps |= NFS_CAP_ATIME;
3088                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3089                         server->caps |= NFS_CAP_CTIME;
3090                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3091                         server->caps |= NFS_CAP_MTIME;
3092 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3093                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3094                         server->caps |= NFS_CAP_SECURITY_LABEL;
3095 #endif
3096                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3097                                 sizeof(server->attr_bitmask));
3098                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3099 
3100                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3101                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3102                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3103                 server->cache_consistency_bitmask[2] = 0;
3104                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3105                         sizeof(server->exclcreat_bitmask));
3106                 server->acl_bitmask = res.acl_bitmask;
3107                 server->fh_expire_type = res.fh_expire_type;
3108         }
3109 
3110         return status;
3111 }
3112 
3113 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3114 {
3115         struct nfs4_exception exception = { };
3116         int err;
3117         do {
3118                 err = nfs4_handle_exception(server,
3119                                 _nfs4_server_capabilities(server, fhandle),
3120                                 &exception);
3121         } while (exception.retry);
3122         return err;
3123 }
3124 
3125 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3126                 struct nfs_fsinfo *info)
3127 {
3128         u32 bitmask[3];
3129         struct nfs4_lookup_root_arg args = {
3130                 .bitmask = bitmask,
3131         };
3132         struct nfs4_lookup_res res = {
3133                 .server = server,
3134                 .fattr = info->fattr,
3135                 .fh = fhandle,
3136         };
3137         struct rpc_message msg = {
3138                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3139                 .rpc_argp = &args,
3140                 .rpc_resp = &res,
3141         };
3142 
3143         bitmask[0] = nfs4_fattr_bitmap[0];
3144         bitmask[1] = nfs4_fattr_bitmap[1];
3145         /*
3146          * Process the label in the upcoming getfattr
3147          */
3148         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3149 
3150         nfs_fattr_init(info->fattr);
3151         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3152 }
3153 
3154 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3155                 struct nfs_fsinfo *info)
3156 {
3157         struct nfs4_exception exception = { };
3158         int err;
3159         do {
3160                 err = _nfs4_lookup_root(server, fhandle, info);
3161                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3162                 switch (err) {
3163                 case 0:
3164                 case -NFS4ERR_WRONGSEC:
3165                         goto out;
3166                 default:
3167                         err = nfs4_handle_exception(server, err, &exception);
3168                 }
3169         } while (exception.retry);
3170 out:
3171         return err;
3172 }
3173 
3174 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3175                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3176 {
3177         struct rpc_auth_create_args auth_args = {
3178                 .pseudoflavor = flavor,
3179         };
3180         struct rpc_auth *auth;
3181         int ret;
3182 
3183         auth = rpcauth_create(&auth_args, server->client);
3184         if (IS_ERR(auth)) {
3185                 ret = -EACCES;
3186                 goto out;
3187         }
3188         ret = nfs4_lookup_root(server, fhandle, info);
3189 out:
3190         return ret;
3191 }
3192 
3193 /*
3194  * Retry pseudoroot lookup with various security flavors.  We do this when:
3195  *
3196  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3197  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3198  *
3199  * Returns zero on success, or a negative NFS4ERR value, or a
3200  * negative errno value.
3201  */
3202 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3203                               struct nfs_fsinfo *info)
3204 {
3205         /* Per 3530bis 15.33.5 */
3206         static const rpc_authflavor_t flav_array[] = {
3207                 RPC_AUTH_GSS_KRB5P,
3208                 RPC_AUTH_GSS_KRB5I,
3209                 RPC_AUTH_GSS_KRB5,
3210                 RPC_AUTH_UNIX,                  /* courtesy */
3211                 RPC_AUTH_NULL,
3212         };
3213         int status = -EPERM;
3214         size_t i;
3215 
3216         if (server->auth_info.flavor_len > 0) {
3217                 /* try each flavor specified by user */
3218                 for (i = 0; i < server->auth_info.flavor_len; i++) {
3219                         status = nfs4_lookup_root_sec(server, fhandle, info,
3220                                                 server->auth_info.flavors[i]);
3221                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3222                                 continue;
3223                         break;
3224                 }
3225         } else {
3226                 /* no flavors specified by user, try default list */
3227                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
3228                         status = nfs4_lookup_root_sec(server, fhandle, info,
3229                                                       flav_array[i]);
3230                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
3231                                 continue;
3232                         break;
3233                 }
3234         }
3235 
3236         /*
3237          * -EACCESS could mean that the user doesn't have correct permissions
3238          * to access the mount.  It could also mean that we tried to mount
3239          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
3240          * existing mount programs don't handle -EACCES very well so it should
3241          * be mapped to -EPERM instead.
3242          */
3243         if (status == -EACCES)
3244                 status = -EPERM;
3245         return status;
3246 }
3247 
3248 static int nfs4_do_find_root_sec(struct nfs_server *server,
3249                 struct nfs_fh *fhandle, struct nfs_fsinfo *info)
3250 {
3251         int mv = server->nfs_client->cl_minorversion;
3252         return nfs_v4_minor_ops[mv]->find_root_sec(server, fhandle, info);
3253 }
3254 
3255 /**
3256  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
3257  * @server: initialized nfs_server handle
3258  * @fhandle: we fill in the pseudo-fs root file handle
3259  * @info: we fill in an FSINFO struct
3260  * @auth_probe: probe the auth flavours
3261  *
3262  * Returns zero on success, or a negative errno.
3263  */
3264 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
3265                          struct nfs_fsinfo *info,
3266                          bool auth_probe)
3267 {
3268         int status = 0;
3269 
3270         if (!auth_probe)
3271                 status = nfs4_lookup_root(server, fhandle, info);
3272 
3273         if (auth_probe || status == NFS4ERR_WRONGSEC)
3274                 status = nfs4_do_find_root_sec(server, fhandle, info);
3275 
3276         if (status == 0)
3277                 status = nfs4_server_capabilities(server, fhandle);
3278         if (status == 0)
3279                 status = nfs4_do_fsinfo(server, fhandle, info);
3280 
3281         return nfs4_map_errors(status);
3282 }
3283 
3284 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
3285                               struct nfs_fsinfo *info)
3286 {
3287         int error;
3288         struct nfs_fattr *fattr = info->fattr;
3289         struct nfs4_label *label = NULL;
3290 
3291         error = nfs4_server_capabilities(server, mntfh);
3292         if (error < 0) {
3293                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
3294                 return error;
3295         }
3296 
3297         label = nfs4_label_alloc(server, GFP_KERNEL);
3298         if (IS_ERR(label))
3299                 return PTR_ERR(label);
3300 
3301         error = nfs4_proc_getattr(server, mntfh, fattr, label);
3302         if (error < 0) {
3303                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
3304                 goto err_free_label;
3305         }
3306 
3307         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
3308             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
3309                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
3310 
3311 err_free_label:
3312         nfs4_label_free(label);
3313 
3314         return error;
3315 }
3316 
3317 /*
3318  * Get locations and (maybe) other attributes of a referral.
3319  * Note that we'll actually follow the referral later when
3320  * we detect fsid mismatch in inode revalidation
3321  */
3322 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
3323                              const struct qstr *name, struct nfs_fattr *fattr,
3324                              struct nfs_fh *fhandle)
3325 {
3326         int status = -ENOMEM;
3327         struct page *page = NULL;
3328         struct nfs4_fs_locations *locations = NULL;
3329 
3330         page = alloc_page(GFP_KERNEL);
3331         if (page == NULL)
3332                 goto out;
3333         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
3334         if (locations == NULL)
3335                 goto out;
3336 
3337         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
3338         if (status != 0)
3339                 goto out;
3340 
3341         /*
3342          * If the fsid didn't change, this is a migration event, not a
3343          * referral.  Cause us to drop into the exception handler, which
3344          * will kick off migration recovery.
3345          */
3346         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
3347                 dprintk("%s: server did not return a different fsid for"
3348                         " a referral at %s\n", __func__, name->name);
3349                 status = -NFS4ERR_MOVED;
3350                 goto out;
3351         }
3352         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
3353         nfs_fixup_referral_attributes(&locations->fattr);
3354 
3355         /* replace the lookup nfs_fattr with the locations nfs_fattr */
3356         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
3357         memset(fhandle, 0, sizeof(struct nfs_fh));
3358 out:
3359         if (page)
3360                 __free_page(page);
3361         kfree(locations);
3362         return status;
3363 }
3364 
3365 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3366                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3367 {
3368         struct nfs4_getattr_arg args = {
3369                 .fh = fhandle,
3370                 .bitmask = server->attr_bitmask,
3371         };
3372         struct nfs4_getattr_res res = {
3373                 .fattr = fattr,
3374                 .label = label,
3375                 .server = server,
3376         };
3377         struct rpc_message msg = {
3378                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
3379                 .rpc_argp = &args,
3380                 .rpc_resp = &res,
3381         };
3382 
3383         args.bitmask = nfs4_bitmask(server, label);
3384 
3385         nfs_fattr_init(fattr);
3386         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3387 }
3388 
3389 static int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
3390                                 struct nfs_fattr *fattr, struct nfs4_label *label)
3391 {
3392         struct nfs4_exception exception = { };
3393         int err;
3394         do {
3395                 err = _nfs4_proc_getattr(server, fhandle, fattr, label);
3396                 trace_nfs4_getattr(server, fhandle, fattr, err);
3397                 err = nfs4_handle_exception(server, err,
3398                                 &exception);
3399         } while (exception.retry);
3400         return err;
3401 }
3402 
3403 /* 
3404  * The file is not closed if it is opened due to the a request to change
3405  * the size of the file. The open call will not be needed once the
3406  * VFS layer lookup-intents are implemented.
3407  *
3408  * Close is called when the inode is destroyed.
3409  * If we haven't opened the file for O_WRONLY, we
3410  * need to in the size_change case to obtain a stateid.
3411  *
3412  * Got race?
3413  * Because OPEN is always done by name in nfsv4, it is
3414  * possible that we opened a different file by the same
3415  * name.  We can recognize this race condition, but we
3416  * can't do anything about it besides returning an error.
3417  *
3418  * This will be fixed with VFS changes (lookup-intent).
3419  */
3420 static int
3421 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
3422                   struct iattr *sattr)
3423 {
3424         struct inode *inode = d_inode(dentry);
3425         struct rpc_cred *cred = NULL;
3426         struct nfs4_state *state = NULL;
3427         struct nfs4_label *label = NULL;
3428         int status;
3429 
3430         if (pnfs_ld_layoutret_on_setattr(inode) &&
3431             sattr->ia_valid & ATTR_SIZE &&
3432             sattr->ia_size < i_size_read(inode))
3433                 pnfs_commit_and_return_layout(inode);
3434 
3435         nfs_fattr_init(fattr);
3436         
3437         /* Deal with open(O_TRUNC) */
3438         if (sattr->ia_valid & ATTR_OPEN)
3439                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
3440 
3441         /* Optimization: if the end result is no change, don't RPC */
3442         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
3443                 return 0;
3444 
3445         /* Search for an existing open(O_WRITE) file */
3446         if (sattr->ia_valid & ATTR_FILE) {
3447                 struct nfs_open_context *ctx;
3448 
3449                 ctx = nfs_file_open_context(sattr->ia_file);
3450                 if (ctx) {
3451                         cred = ctx->cred;
3452                         state = ctx->state;
3453                 }
3454         }
3455 
3456         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
3457         if (IS_ERR(label))
3458                 return PTR_ERR(label);
3459 
3460         status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
3461         if (status == 0) {
3462                 nfs_setattr_update_inode(inode, sattr, fattr);
3463                 nfs_setsecurity(inode, fattr, label);
3464         }
3465         nfs4_label_free(label);
3466         return status;
3467 }
3468 
3469 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
3470                 const struct qstr *name, struct nfs_fh *fhandle,
3471                 struct nfs_fattr *fattr, struct nfs4_label *label)
3472 {
3473         struct nfs_server *server = NFS_SERVER(dir);
3474         int                    status;
3475         struct nfs4_lookup_arg args = {
3476                 .bitmask = server->attr_bitmask,
3477                 .dir_fh = NFS_FH(dir),
3478                 .name = name,
3479         };
3480         struct nfs4_lookup_res res = {
3481                 .server = server,
3482                 .fattr = fattr,
3483                 .label = label,
3484                 .fh = fhandle,
3485         };
3486         struct rpc_message msg = {
3487                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
3488                 .rpc_argp = &args,
3489                 .rpc_resp = &res,
3490         };
3491 
3492         args.bitmask = nfs4_bitmask(server, label);
3493 
3494         nfs_fattr_init(fattr);
3495 
3496         dprintk("NFS call  lookup %s\n", name->name);
3497         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args, &res.seq_res, 0);
3498         dprintk("NFS reply lookup: %d\n", status);
3499         return status;
3500 }
3501 
3502 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
3503 {
3504         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
3505                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
3506         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
3507         fattr->nlink = 2;
3508 }
3509 
3510 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
3511                                    struct qstr *name, struct nfs_fh *fhandle,
3512                                    struct nfs_fattr *fattr, struct nfs4_label *label)
3513 {
3514         struct nfs4_exception exception = { };
3515         struct rpc_clnt *client = *clnt;
3516         int err;
3517         do {
3518                 err = _nfs4_proc_lookup(client, dir, name, fhandle, fattr, label);
3519                 trace_nfs4_lookup(dir, name, err);
3520                 switch (err) {
3521                 case -NFS4ERR_BADNAME:
3522                         err = -ENOENT;
3523                         goto out;
3524                 case -NFS4ERR_MOVED:
3525                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
3526                         if (err == -NFS4ERR_MOVED)
3527                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3528                         goto out;
3529                 case -NFS4ERR_WRONGSEC:
3530                         err = -EPERM;
3531                         if (client != *clnt)
3532                                 goto out;
3533                         client = nfs4_negotiate_security(client, dir, name);
3534                         if (IS_ERR(client))
3535                                 return PTR_ERR(client);
3536 
3537                         exception.retry = 1;
3538                         break;
3539                 default:
3540                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
3541                 }
3542         } while (exception.retry);
3543 
3544 out:
3545         if (err == 0)
3546                 *clnt = client;
3547         else if (client != *clnt)
3548                 rpc_shutdown_client(client);
3549 
3550         return err;
3551 }
3552 
3553 static int nfs4_proc_lookup(struct inode *dir, struct qstr *name,
3554                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
3555                             struct nfs4_label *label)
3556 {
3557         int status;
3558         struct rpc_clnt *client = NFS_CLIENT(dir);
3559 
3560         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, label);
3561         if (client != NFS_CLIENT(dir)) {
3562                 rpc_shutdown_client(client);
3563                 nfs_fixup_secinfo_attributes(fattr);
3564         }
3565         return status;
3566 }
3567 
3568 struct rpc_clnt *
3569 nfs4_proc_lookup_mountpoint(struct inode *dir, struct qstr *name,
3570                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
3571 {
3572         struct rpc_clnt *client = NFS_CLIENT(dir);
3573         int status;
3574 
3575         status = nfs4_proc_lookup_common(&client, dir, name, fhandle, fattr, NULL);
3576         if (status < 0)
3577                 return ERR_PTR(status);
3578         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
3579 }
3580 
3581 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3582 {
3583         struct nfs_server *server = NFS_SERVER(inode);
3584         struct nfs4_accessargs args = {
3585                 .fh = NFS_FH(inode),
3586                 .bitmask = server->cache_consistency_bitmask,
3587         };
3588         struct nfs4_accessres res = {
3589                 .server = server,
3590         };
3591         struct rpc_message msg = {
3592                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
3593                 .rpc_argp = &args,
3594                 .rpc_resp = &res,
3595                 .rpc_cred = entry->cred,
3596         };
3597         int mode = entry->mask;
3598         int status = 0;
3599 
3600         /*
3601          * Determine which access bits we want to ask for...
3602          */
3603         if (mode & MAY_READ)
3604                 args.access |= NFS4_ACCESS_READ;
3605         if (S_ISDIR(inode->i_mode)) {
3606                 if (mode & MAY_WRITE)
3607                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE;
3608                 if (mode & MAY_EXEC)
3609                         args.access |= NFS4_ACCESS_LOOKUP;
3610         } else {
3611                 if (mode & MAY_WRITE)
3612                         args.access |= NFS4_ACCESS_MODIFY | NFS4_ACCESS_EXTEND;
3613                 if (mode & MAY_EXEC)
3614                         args.access |= NFS4_ACCESS_EXECUTE;
3615         }
3616 
3617         res.fattr = nfs_alloc_fattr();
3618         if (res.fattr == NULL)
3619                 return -ENOMEM;
3620 
3621         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3622         if (!status) {
3623                 nfs_access_set_mask(entry, res.access);
3624                 nfs_refresh_inode(inode, res.fattr);
3625         }
3626         nfs_free_fattr(res.fattr);
3627         return status;
3628 }
3629 
3630 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
3631 {
3632         struct nfs4_exception exception = { };
3633         int err;
3634         do {
3635                 err = _nfs4_proc_access(inode, entry);
3636                 trace_nfs4_access(inode, err);
3637                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3638                                 &exception);
3639         } while (exception.retry);
3640         return err;
3641 }
3642 
3643 /*
3644  * TODO: For the time being, we don't try to get any attributes
3645  * along with any of the zero-copy operations READ, READDIR,
3646  * READLINK, WRITE.
3647  *
3648  * In the case of the first three, we want to put the GETATTR
3649  * after the read-type operation -- this is because it is hard
3650  * to predict the length of a GETATTR response in v4, and thus
3651  * align the READ data correctly.  This means that the GETATTR
3652  * may end up partially falling into the page cache, and we should
3653  * shift it into the 'tail' of the xdr_buf before processing.
3654  * To do this efficiently, we need to know the total length
3655  * of data received, which doesn't seem to be available outside
3656  * of the RPC layer.
3657  *
3658  * In the case of WRITE, we also want to put the GETATTR after
3659  * the operation -- in this case because we want to make sure
3660  * we get the post-operation mtime and size.
3661  *
3662  * Both of these changes to the XDR layer would in fact be quite
3663  * minor, but I decided to leave them for a subsequent patch.
3664  */
3665 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
3666                 unsigned int pgbase, unsigned int pglen)
3667 {
3668         struct nfs4_readlink args = {
3669                 .fh       = NFS_FH(inode),
3670                 .pgbase   = pgbase,
3671                 .pglen    = pglen,
3672                 .pages    = &page,
3673         };
3674         struct nfs4_readlink_res res;
3675         struct rpc_message msg = {
3676                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
3677                 .rpc_argp = &args,
3678                 .rpc_resp = &res,
3679         };
3680 
3681         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
3682 }
3683 
3684 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
3685                 unsigned int pgbase, unsigned int pglen)
3686 {
3687         struct nfs4_exception exception = { };
3688         int err;
3689         do {
3690                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
3691                 trace_nfs4_readlink(inode, err);
3692                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
3693                                 &exception);
3694         } while (exception.retry);
3695         return err;
3696 }
3697 
3698 /*
3699  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
3700  */
3701 static int
3702 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
3703                  int flags)
3704 {
3705         struct nfs4_label l, *ilabel = NULL;
3706         struct nfs_open_context *ctx;
3707         struct nfs4_state *state;
3708         int status = 0;
3709 
3710         ctx = alloc_nfs_open_context(dentry, FMODE_READ);
3711         if (IS_ERR(ctx))
3712                 return PTR_ERR(ctx);
3713 
3714         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
3715 
3716         sattr->ia_mode &= ~current_umask();
3717         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
3718         if (IS_ERR(state)) {
3719                 status = PTR_ERR(state);
3720                 goto out;
3721         }
3722 out:
3723         nfs4_label_release_security(ilabel);
3724         put_nfs_open_context(ctx);
3725         return status;
3726 }
3727 
3728 static int _nfs4_proc_remove(struct inode *dir, struct qstr *name)
3729 {
3730         struct nfs_server *server = NFS_SERVER(dir);
3731         struct nfs_removeargs args = {
3732                 .fh = NFS_FH(dir),
3733                 .name = *name,
3734         };
3735         struct nfs_removeres res = {
3736                 .server = server,
3737         };
3738         struct rpc_message msg = {
3739                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
3740                 .rpc_argp = &args,
3741                 .rpc_resp = &res,
3742         };
3743         int status;
3744 
3745         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
3746         if (status == 0)
3747                 update_changeattr(dir, &res.cinfo);
3748         return status;
3749 }
3750 
3751 static int nfs4_proc_remove(struct inode *dir, struct qstr *name)
3752 {
3753         struct nfs4_exception exception = { };
3754         int err;
3755         do {
3756                 err = _nfs4_proc_remove(dir, name);
3757                 trace_nfs4_remove(dir, name, err);
3758                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3759                                 &exception);
3760         } while (exception.retry);
3761         return err;
3762 }
3763 
3764 static void nfs4_proc_unlink_setup(struct rpc_message *msg, struct inode *dir)
3765 {
3766         struct nfs_server *server = NFS_SERVER(dir);
3767         struct nfs_removeargs *args = msg->rpc_argp;
3768         struct nfs_removeres *res = msg->rpc_resp;
3769 
3770         res->server = server;
3771         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
3772         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1);
3773 
3774         nfs_fattr_init(res->dir_attr);
3775 }
3776 
3777 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
3778 {
3779         nfs4_setup_sequence(NFS_SERVER(data->dir),
3780                         &data->args.seq_args,
3781                         &data->res.seq_res,
3782                         task);
3783 }
3784 
3785 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
3786 {
3787         struct nfs_unlinkdata *data = task->tk_calldata;
3788         struct nfs_removeres *res = &data->res;
3789 
3790         if (!nfs4_sequence_done(task, &res->seq_res))
3791                 return 0;
3792         if (nfs4_async_handle_error(task, res->server, NULL,
3793                                     &data->timeout) == -EAGAIN)
3794                 return 0;
3795         update_changeattr(dir, &res->cinfo);
3796         return 1;
3797 }
3798 
3799 static void nfs4_proc_rename_setup(struct rpc_message *msg, struct inode *dir)
3800 {
3801         struct nfs_server *server = NFS_SERVER(dir);
3802         struct nfs_renameargs *arg = msg->rpc_argp;
3803         struct nfs_renameres *res = msg->rpc_resp;
3804 
3805         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
3806         res->server = server;
3807         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1);
3808 }
3809 
3810 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
3811 {
3812         nfs4_setup_sequence(NFS_SERVER(data->old_dir),
3813                         &data->args.seq_args,
3814                         &data->res.seq_res,
3815                         task);
3816 }
3817 
3818 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
3819                                  struct inode *new_dir)
3820 {
3821         struct nfs_renamedata *data = task->tk_calldata;
3822         struct nfs_renameres *res = &data->res;
3823 
3824         if (!nfs4_sequence_done(task, &res->seq_res))
3825                 return 0;
3826         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
3827                 return 0;
3828 
3829         update_changeattr(old_dir, &res->old_cinfo);
3830         update_changeattr(new_dir, &res->new_cinfo);
3831         return 1;
3832 }
3833 
3834 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3835 {
3836         struct nfs_server *server = NFS_SERVER(inode);
3837         struct nfs4_link_arg arg = {
3838                 .fh     = NFS_FH(inode),
3839                 .dir_fh = NFS_FH(dir),
3840                 .name   = name,
3841                 .bitmask = server->attr_bitmask,
3842         };
3843         struct nfs4_link_res res = {
3844                 .server = server,
3845                 .label = NULL,
3846         };
3847         struct rpc_message msg = {
3848                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
3849                 .rpc_argp = &arg,
3850                 .rpc_resp = &res,
3851         };
3852         int status = -ENOMEM;
3853 
3854         res.fattr = nfs_alloc_fattr();
3855         if (res.fattr == NULL)
3856                 goto out;
3857 
3858         res.label = nfs4_label_alloc(server, GFP_KERNEL);
3859         if (IS_ERR(res.label)) {
3860                 status = PTR_ERR(res.label);
3861                 goto out;
3862         }
3863         arg.bitmask = nfs4_bitmask(server, res.label);
3864 
3865         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
3866         if (!status) {
3867                 update_changeattr(dir, &res.cinfo);
3868                 status = nfs_post_op_update_inode(inode, res.fattr);
3869                 if (!status)
3870                         nfs_setsecurity(inode, res.fattr, res.label);
3871         }
3872 
3873 
3874         nfs4_label_free(res.label);
3875 
3876 out:
3877         nfs_free_fattr(res.fattr);
3878         return status;
3879 }
3880 
3881 static int nfs4_proc_link(struct inode *inode, struct inode *dir, struct qstr *name)
3882 {
3883         struct nfs4_exception exception = { };
3884         int err;
3885         do {
3886                 err = nfs4_handle_exception(NFS_SERVER(inode),
3887                                 _nfs4_proc_link(inode, dir, name),
3888                                 &exception);
3889         } while (exception.retry);
3890         return err;
3891 }
3892 
3893 struct nfs4_createdata {
3894         struct rpc_message msg;
3895         struct nfs4_create_arg arg;
3896         struct nfs4_create_res res;
3897         struct nfs_fh fh;
3898         struct nfs_fattr fattr;
3899         struct nfs4_label *label;
3900 };
3901 
3902 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
3903                 struct qstr *name, struct iattr *sattr, u32 ftype)
3904 {
3905         struct nfs4_createdata *data;
3906 
3907         data = kzalloc(sizeof(*data), GFP_KERNEL);
3908         if (data != NULL) {
3909                 struct nfs_server *server = NFS_SERVER(dir);
3910 
3911                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
3912                 if (IS_ERR(data->label))
3913                         goto out_free;
3914 
3915                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
3916                 data->msg.rpc_argp = &data->arg;
3917                 data->msg.rpc_resp = &data->res;
3918                 data->arg.dir_fh = NFS_FH(dir);
3919                 data->arg.server = server;
3920                 data->arg.name = name;
3921                 data->arg.attrs = sattr;
3922                 data->arg.ftype = ftype;
3923                 data->arg.bitmask = nfs4_bitmask(server, data->label);
3924                 data->res.server = server;
3925                 data->res.fh = &data->fh;
3926                 data->res.fattr = &data->fattr;
3927                 data->res.label = data->label;
3928                 nfs_fattr_init(data->res.fattr);
3929         }
3930         return data;
3931 out_free:
3932         kfree(data);
3933         return NULL;
3934 }
3935 
3936 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
3937 {
3938         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
3939                                     &data->arg.seq_args, &data->res.seq_res, 1);
3940         if (status == 0) {
3941                 update_changeattr(dir, &data->res.dir_cinfo);
3942                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
3943         }
3944         return status;
3945 }
3946 
3947 static void nfs4_free_createdata(struct nfs4_createdata *data)
3948 {
3949         nfs4_label_free(data->label);
3950         kfree(data);
3951 }
3952 
3953 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3954                 struct page *page, unsigned int len, struct iattr *sattr,
3955                 struct nfs4_label *label)
3956 {
3957         struct nfs4_createdata *data;
3958         int status = -ENAMETOOLONG;
3959 
3960         if (len > NFS4_MAXPATHLEN)
3961                 goto out;
3962 
3963         status = -ENOMEM;
3964         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
3965         if (data == NULL)
3966                 goto out;
3967 
3968         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
3969         data->arg.u.symlink.pages = &page;
3970         data->arg.u.symlink.len = len;
3971         data->arg.label = label;
3972         
3973         status = nfs4_do_create(dir, dentry, data);
3974 
3975         nfs4_free_createdata(data);
3976 out:
3977         return status;
3978 }
3979 
3980 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
3981                 struct page *page, unsigned int len, struct iattr *sattr)
3982 {
3983         struct nfs4_exception exception = { };
3984         struct nfs4_label l, *label = NULL;
3985         int err;
3986 
3987         label = nfs4_label_init_security(dir, dentry, sattr, &l);
3988 
3989         do {
3990                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
3991                 trace_nfs4_symlink(dir, &dentry->d_name, err);
3992                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
3993                                 &exception);
3994         } while (exception.retry);
3995 
3996         nfs4_label_release_security(label);
3997         return err;
3998 }
3999 
4000 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4001                 struct iattr *sattr, struct nfs4_label *label)
4002 {
4003         struct nfs4_createdata *data;
4004         int status = -ENOMEM;
4005 
4006         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4007         if (data == NULL)
4008                 goto out;
4009 
4010         data->arg.label = label;
4011         status = nfs4_do_create(dir, dentry, data);
4012 
4013         nfs4_free_createdata(data);
4014 out:
4015         return status;
4016 }
4017 
4018 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4019                 struct iattr *sattr)
4020 {
4021         struct nfs4_exception exception = { };
4022         struct nfs4_label l, *label = NULL;
4023         int err;
4024 
4025         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4026 
4027         sattr->ia_mode &= ~current_umask();
4028         do {
4029                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4030                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4031                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4032                                 &exception);
4033         } while (exception.retry);
4034         nfs4_label_release_security(label);
4035 
4036         return err;
4037 }
4038 
4039 static int _nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4040                 u64 cookie, struct page **pages, unsigned int count, int plus)
4041 {
4042         struct inode            *dir = d_inode(dentry);
4043         struct nfs4_readdir_arg args = {
4044                 .fh = NFS_FH(dir),
4045                 .pages = pages,
4046                 .pgbase = 0,
4047                 .count = count,
4048                 .bitmask = NFS_SERVER(d_inode(dentry))->attr_bitmask,
4049                 .plus = plus,
4050         };
4051         struct nfs4_readdir_res res;
4052         struct rpc_message msg = {
4053                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4054                 .rpc_argp = &args,
4055                 .rpc_resp = &res,
4056                 .rpc_cred = cred,
4057         };
4058         int                     status;
4059 
4060         dprintk("%s: dentry = %pd2, cookie = %Lu\n", __func__,
4061                         dentry,
4062                         (unsigned long long)cookie);
4063         nfs4_setup_readdir(cookie, NFS_I(dir)->cookieverf, dentry, &args);
4064         res.pgbase = args.pgbase;
4065         status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &msg, &args.seq_args, &res.seq_res, 0);
4066         if (status >= 0) {
4067                 memcpy(NFS_I(dir)->cookieverf, res.verifier.data, NFS4_VERIFIER_SIZE);
4068                 status += args.pgbase;
4069         }
4070 
4071         nfs_invalidate_atime(dir);
4072 
4073         dprintk("%s: returns %d\n", __func__, status);
4074         return status;
4075 }
4076 
4077 static int nfs4_proc_readdir(struct dentry *dentry, struct rpc_cred *cred,
4078                 u64 cookie, struct page **pages, unsigned int count, int plus)
4079 {
4080         struct nfs4_exception exception = { };
4081         int err;
4082         do {
4083                 err = _nfs4_proc_readdir(dentry, cred, cookie,
4084                                 pages, count, plus);
4085                 trace_nfs4_readdir(d_inode(dentry), err);
4086                 err = nfs4_handle_exception(NFS_SERVER(d_inode(dentry)), err,
4087                                 &exception);
4088         } while (exception.retry);
4089         return err;
4090 }
4091 
4092 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4093                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
4094 {
4095         struct nfs4_createdata *data;
4096         int mode = sattr->ia_mode;
4097         int status = -ENOMEM;
4098 
4099         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
4100         if (data == NULL)
4101                 goto out;
4102 
4103         if (S_ISFIFO(mode))
4104                 data->arg.ftype = NF4FIFO;
4105         else if (S_ISBLK(mode)) {
4106                 data->arg.ftype = NF4BLK;
4107                 data->arg.u.device.specdata1 = MAJOR(rdev);
4108                 data->arg.u.device.specdata2 = MINOR(rdev);
4109         }
4110         else if (S_ISCHR(mode)) {
4111                 data->arg.ftype = NF4CHR;
4112                 data->arg.u.device.specdata1 = MAJOR(rdev);
4113                 data->arg.u.device.specdata2 = MINOR(rdev);
4114         } else if (!S_ISSOCK(mode)) {
4115                 status = -EINVAL;
4116                 goto out_free;
4117         }
4118 
4119         data->arg.label = label;
4120         status = nfs4_do_create(dir, dentry, data);
4121 out_free:
4122         nfs4_free_createdata(data);
4123 out:
4124         return status;
4125 }
4126 
4127 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
4128                 struct iattr *sattr, dev_t rdev)
4129 {
4130         struct nfs4_exception exception = { };
4131         struct nfs4_label l, *label = NULL;
4132         int err;
4133 
4134         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4135 
4136         sattr->ia_mode &= ~current_umask();
4137         do {
4138                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
4139                 trace_nfs4_mknod(dir, &dentry->d_name, err);
4140                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4141                                 &exception);
4142         } while (exception.retry);
4143 
4144         nfs4_label_release_security(label);
4145 
4146         return err;
4147 }
4148 
4149 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
4150                  struct nfs_fsstat *fsstat)
4151 {
4152         struct nfs4_statfs_arg args = {
4153                 .fh = fhandle,
4154                 .bitmask = server->attr_bitmask,
4155         };
4156         struct nfs4_statfs_res res = {
4157                 .fsstat = fsstat,
4158         };
4159         struct rpc_message msg = {
4160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
4161                 .rpc_argp = &args,
4162                 .rpc_resp = &res,
4163         };
4164 
4165         nfs_fattr_init(fsstat->fattr);
4166         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4167 }
4168 
4169 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
4170 {
4171         struct nfs4_exception exception = { };
4172         int err;
4173         do {
4174                 err = nfs4_handle_exception(server,
4175                                 _nfs4_proc_statfs(server, fhandle, fsstat),
4176                                 &exception);
4177         } while (exception.retry);
4178         return err;
4179 }
4180 
4181 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
4182                 struct nfs_fsinfo *fsinfo)
4183 {
4184         struct nfs4_fsinfo_arg args = {
4185                 .fh = fhandle,
4186                 .bitmask = server->attr_bitmask,
4187         };
4188         struct nfs4_fsinfo_res res = {
4189                 .fsinfo = fsinfo,
4190         };
4191         struct rpc_message msg = {
4192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
4193                 .rpc_argp = &args,
4194                 .rpc_resp = &res,
4195         };
4196 
4197         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4198 }
4199 
4200 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4201 {
4202         struct nfs4_exception exception = { };
4203         unsigned long now = jiffies;
4204         int err;
4205 
4206         do {
4207                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
4208                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
4209                 if (err == 0) {
4210                         struct nfs_client *clp = server->nfs_client;
4211 
4212                         spin_lock(&clp->cl_lock);
4213                         clp->cl_lease_time = fsinfo->lease_time * HZ;
4214                         clp->cl_last_renewal = now;
4215                         spin_unlock(&clp->cl_lock);
4216                         break;
4217                 }
4218                 err = nfs4_handle_exception(server, err, &exception);
4219         } while (exception.retry);
4220         return err;
4221 }
4222 
4223 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
4224 {
4225         int error;
4226 
4227         nfs_fattr_init(fsinfo->fattr);
4228         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
4229         if (error == 0) {
4230                 /* block layout checks this! */
4231                 server->pnfs_blksize = fsinfo->blksize;
4232                 set_pnfs_layoutdriver(server, fhandle, fsinfo->layouttype);
4233         }
4234 
4235         return error;
4236 }
4237 
4238 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4239                 struct nfs_pathconf *pathconf)
4240 {
4241         struct nfs4_pathconf_arg args = {
4242                 .fh = fhandle,
4243                 .bitmask = server->attr_bitmask,
4244         };
4245         struct nfs4_pathconf_res res = {
4246                 .pathconf = pathconf,
4247         };
4248         struct rpc_message msg = {
4249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
4250                 .rpc_argp = &args,
4251                 .rpc_resp = &res,
4252         };
4253 
4254         /* None of the pathconf attributes are mandatory to implement */
4255         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
4256                 memset(pathconf, 0, sizeof(*pathconf));
4257                 return 0;
4258         }
4259 
4260         nfs_fattr_init(pathconf->fattr);
4261         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4262 }
4263 
4264 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
4265                 struct nfs_pathconf *pathconf)
4266 {
4267         struct nfs4_exception exception = { };
4268         int err;
4269 
4270         do {
4271                 err = nfs4_handle_exception(server,
4272                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
4273                                 &exception);
4274         } while (exception.retry);
4275         return err;
4276 }
4277 
4278 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
4279                 const struct nfs_open_context *ctx,
4280                 const struct nfs_lock_context *l_ctx,
4281                 fmode_t fmode)
4282 {
4283         const struct nfs_lockowner *lockowner = NULL;
4284 
4285         if (l_ctx != NULL)
4286                 lockowner = &l_ctx->lockowner;
4287         return nfs4_select_rw_stateid(stateid, ctx->state, fmode, lockowner);
4288 }
4289 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
4290 
4291 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
4292                 const struct nfs_open_context *ctx,
4293                 const struct nfs_lock_context *l_ctx,
4294                 fmode_t fmode)
4295 {
4296         nfs4_stateid current_stateid;
4297 
4298         /* If the current stateid represents a lost lock, then exit */
4299         if (nfs4_set_rw_stateid(&current_stateid, ctx, l_ctx, fmode) == -EIO)
4300                 return true;
4301         return nfs4_stateid_match(stateid, &current_stateid);
4302 }
4303 
4304 static bool nfs4_error_stateid_expired(int err)
4305 {
4306         switch (err) {
4307         case -NFS4ERR_DELEG_REVOKED:
4308         case -NFS4ERR_ADMIN_REVOKED:
4309         case -NFS4ERR_BAD_STATEID:
4310         case -NFS4ERR_STALE_STATEID:
4311         case -NFS4ERR_OLD_STATEID:
4312         case -NFS4ERR_OPENMODE:
4313         case -NFS4ERR_EXPIRED:
4314                 return true;
4315         }
4316         return false;
4317 }
4318 
4319 void __nfs4_read_done_cb(struct nfs_pgio_header *hdr)
4320 {
4321         nfs_invalidate_atime(hdr->inode);
4322 }
4323 
4324 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
4325 {
4326         struct nfs_server *server = NFS_SERVER(hdr->inode);
4327 
4328         trace_nfs4_read(hdr, task->tk_status);
4329         if (nfs4_async_handle_error(task, server,
4330                                     hdr->args.context->state,
4331                                     NULL) == -EAGAIN) {
4332                 rpc_restart_call_prepare(task);
4333                 return -EAGAIN;
4334         }
4335 
4336         __nfs4_read_done_cb(hdr);
4337         if (task->tk_status > 0)
4338                 renew_lease(server, hdr->timestamp);
4339         return 0;
4340 }
4341 
4342 static bool nfs4_read_stateid_changed(struct rpc_task *task,
4343                 struct nfs_pgio_args *args)
4344 {
4345 
4346         if (!nfs4_error_stateid_expired(task->tk_status) ||
4347                 nfs4_stateid_is_current(&args->stateid,
4348                                 args->context,
4349                                 args->lock_context,
4350                                 FMODE_READ))
4351                 return false;
4352         rpc_restart_call_prepare(task);
4353         return true;
4354 }
4355 
4356 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4357 {
4358 
4359         dprintk("--> %s\n", __func__);
4360 
4361         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4362                 return -EAGAIN;
4363         if (nfs4_read_stateid_changed(task, &hdr->args))
4364                 return -EAGAIN;
4365         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4366                                     nfs4_read_done_cb(task, hdr);
4367 }
4368 
4369 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
4370                                  struct rpc_message *msg)
4371 {
4372         hdr->timestamp   = jiffies;
4373         hdr->pgio_done_cb = nfs4_read_done_cb;
4374         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
4375         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0);
4376 }
4377 
4378 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
4379                                       struct nfs_pgio_header *hdr)
4380 {
4381         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode),
4382                         &hdr->args.seq_args,
4383                         &hdr->res.seq_res,
4384                         task))
4385                 return 0;
4386         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
4387                                 hdr->args.lock_context,
4388                                 hdr->rw_ops->rw_mode) == -EIO)
4389                 return -EIO;
4390         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
4391                 return -EIO;
4392         return 0;
4393 }
4394 
4395 static int nfs4_write_done_cb(struct rpc_task *task,
4396                               struct nfs_pgio_header *hdr)
4397 {
4398         struct inode *inode = hdr->inode;
4399 
4400         trace_nfs4_write(hdr, task->tk_status);
4401         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4402                                     hdr->args.context->state,
4403                                     NULL) == -EAGAIN) {
4404                 rpc_restart_call_prepare(task);
4405                 return -EAGAIN;
4406         }
4407         if (task->tk_status >= 0) {
4408                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
4409                 nfs_writeback_update_inode(hdr);
4410         }
4411         return 0;
4412 }
4413 
4414 static bool nfs4_write_stateid_changed(struct rpc_task *task,
4415                 struct nfs_pgio_args *args)
4416 {
4417 
4418         if (!nfs4_error_stateid_expired(task->tk_status) ||
4419                 nfs4_stateid_is_current(&args->stateid,
4420                                 args->context,
4421                                 args->lock_context,
4422                                 FMODE_WRITE))
4423                 return false;
4424         rpc_restart_call_prepare(task);
4425         return true;
4426 }
4427 
4428 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
4429 {
4430         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
4431                 return -EAGAIN;
4432         if (nfs4_write_stateid_changed(task, &hdr->args))
4433                 return -EAGAIN;
4434         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
4435                 nfs4_write_done_cb(task, hdr);
4436 }
4437 
4438 static
4439 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
4440 {
4441         /* Don't request attributes for pNFS or O_DIRECT writes */
4442         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
4443                 return false;
4444         /* Otherwise, request attributes if and only if we don't hold
4445          * a delegation
4446          */
4447         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
4448 }
4449 
4450 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
4451                                   struct rpc_message *msg)
4452 {
4453         struct nfs_server *server = NFS_SERVER(hdr->inode);
4454 
4455         if (!nfs4_write_need_cache_consistency_data(hdr)) {
4456                 hdr->args.bitmask = NULL;
4457                 hdr->res.fattr = NULL;
4458         } else
4459                 hdr->args.bitmask = server->cache_consistency_bitmask;
4460 
4461         if (!hdr->pgio_done_cb)
4462                 hdr->pgio_done_cb = nfs4_write_done_cb;
4463         hdr->res.server = server;
4464         hdr->timestamp   = jiffies;
4465 
4466         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
4467         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 1);
4468 }
4469 
4470 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
4471 {
4472         nfs4_setup_sequence(NFS_SERVER(data->inode),
4473                         &data->args.seq_args,
4474                         &data->res.seq_res,
4475                         task);
4476 }
4477 
4478 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
4479 {
4480         struct inode *inode = data->inode;
4481 
4482         trace_nfs4_commit(data, task->tk_status);
4483         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
4484                                     NULL, NULL) == -EAGAIN) {
4485                 rpc_restart_call_prepare(task);
4486                 return -EAGAIN;
4487         }
4488         return 0;
4489 }
4490 
4491 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
4492 {
4493         if (!nfs4_sequence_done(task, &data->res.seq_res))
4494                 return -EAGAIN;
4495         return data->commit_done_cb(task, data);
4496 }
4497 
4498 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg)
4499 {
4500         struct nfs_server *server = NFS_SERVER(data->inode);
4501 
4502         if (data->commit_done_cb == NULL)
4503                 data->commit_done_cb = nfs4_commit_done_cb;
4504         data->res.server = server;
4505         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
4506         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
4507 }
4508 
4509 struct nfs4_renewdata {
4510         struct nfs_client       *client;
4511         unsigned long           timestamp;
4512 };
4513 
4514 /*
4515  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
4516  * standalone procedure for queueing an asynchronous RENEW.
4517  */
4518 static void nfs4_renew_release(void *calldata)
4519 {
4520         struct nfs4_renewdata *data = calldata;
4521         struct nfs_client *clp = data->client;
4522 
4523         if (atomic_read(&clp->cl_count) > 1)
4524                 nfs4_schedule_state_renewal(clp);
4525         nfs_put_client(clp);
4526         kfree(data);
4527 }
4528 
4529 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
4530 {
4531         struct nfs4_renewdata *data = calldata;
4532         struct nfs_client *clp = data->client;
4533         unsigned long timestamp = data->timestamp;
4534 
4535         trace_nfs4_renew_async(clp, task->tk_status);
4536         switch (task->tk_status) {
4537         case 0:
4538                 break;
4539         case -NFS4ERR_LEASE_MOVED:
4540                 nfs4_schedule_lease_moved_recovery(clp);
4541                 break;
4542         default:
4543                 /* Unless we're shutting down, schedule state recovery! */
4544                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
4545                         return;
4546                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
4547                         nfs4_schedule_lease_recovery(clp);
4548                         return;
4549                 }
4550                 nfs4_schedule_path_down_recovery(clp);
4551         }
4552         do_renew_lease(clp, timestamp);
4553 }
4554 
4555 static const struct rpc_call_ops nfs4_renew_ops = {
4556         .rpc_call_done = nfs4_renew_done,
4557         .rpc_release = nfs4_renew_release,
4558 };
4559 
4560 static int nfs4_proc_async_renew(struct nfs_client *clp, struct rpc_cred *cred, unsigned renew_flags)
4561 {
4562         struct rpc_message msg = {
4563                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4564                 .rpc_argp       = clp,
4565                 .rpc_cred       = cred,
4566         };
4567         struct nfs4_renewdata *data;
4568 
4569         if (renew_flags == 0)
4570                 return 0;
4571         if (!atomic_inc_not_zero(&clp->cl_count))
4572                 return -EIO;
4573         data = kmalloc(sizeof(*data), GFP_NOFS);
4574         if (data == NULL)
4575                 return -ENOMEM;
4576         data->client = clp;
4577         data->timestamp = jiffies;
4578         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
4579                         &nfs4_renew_ops, data);
4580 }
4581 
4582 static int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred)
4583 {
4584         struct rpc_message msg = {
4585                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
4586                 .rpc_argp       = clp,
4587                 .rpc_cred       = cred,
4588         };
4589         unsigned long now = jiffies;
4590         int status;
4591 
4592         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
4593         if (status < 0)
4594                 return status;
4595         do_renew_lease(clp, now);
4596         return 0;
4597 }
4598 
4599 static inline int nfs4_server_supports_acls(struct nfs_server *server)
4600 {
4601         return server->caps & NFS_CAP_ACLS;
4602 }
4603 
4604 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
4605  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
4606  * the stack.
4607  */
4608 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
4609 
4610 static int buf_to_pages_noslab(const void *buf, size_t buflen,
4611                 struct page **pages)
4612 {
4613         struct page *newpage, **spages;
4614         int rc = 0;
4615         size_t len;
4616         spages = pages;
4617 
4618         do {
4619                 len = min_t(size_t, PAGE_SIZE, buflen);
4620                 newpage = alloc_page(GFP_KERNEL);
4621 
4622                 if (newpage == NULL)
4623                         goto unwind;
4624                 memcpy(page_address(newpage), buf, len);
4625                 buf += len;
4626                 buflen -= len;
4627                 *pages++ = newpage;
4628                 rc++;
4629         } while (buflen != 0);
4630 
4631         return rc;
4632 
4633 unwind:
4634         for(; rc > 0; rc--)
4635                 __free_page(spages[rc-1]);
4636         return -ENOMEM;
4637 }
4638 
4639 struct nfs4_cached_acl {
4640         int cached;
4641         size_t len;
4642         char data[0];
4643 };
4644 
4645 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
4646 {
4647         struct nfs_inode *nfsi = NFS_I(inode);
4648 
4649         spin_lock(&inode->i_lock);
4650         kfree(nfsi->nfs4_acl);
4651         nfsi->nfs4_acl = acl;
4652         spin_unlock(&inode->i_lock);
4653 }
4654 
4655 static void nfs4_zap_acl_attr(struct inode *inode)
4656 {
4657         nfs4_set_cached_acl(inode, NULL);
4658 }
4659 
4660 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
4661 {
4662         struct nfs_inode *nfsi = NFS_I(inode);
4663         struct nfs4_cached_acl *acl;
4664         int ret = -ENOENT;
4665 
4666         spin_lock(&inode->i_lock);
4667         acl = nfsi->nfs4_acl;
4668         if (acl == NULL)
4669                 goto out;
4670         if (buf == NULL) /* user is just asking for length */
4671                 goto out_len;
4672         if (acl->cached == 0)
4673                 goto out;
4674         ret = -ERANGE; /* see getxattr(2) man page */
4675         if (acl->len > buflen)
4676                 goto out;
4677         memcpy(buf, acl->data, acl->len);
4678 out_len:
4679         ret = acl->len;
4680 out:
4681         spin_unlock(&inode->i_lock);
4682         return ret;
4683 }
4684 
4685 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
4686 {
4687         struct nfs4_cached_acl *acl;
4688         size_t buflen = sizeof(*acl) + acl_len;
4689 
4690         if (buflen <= PAGE_SIZE) {
4691                 acl = kmalloc(buflen, GFP_KERNEL);
4692                 if (acl == NULL)
4693                         goto out;
4694                 acl->cached = 1;
4695                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
4696         } else {
4697                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
4698                 if (acl == NULL)
4699                         goto out;
4700                 acl->cached = 0;
4701         }
4702         acl->len = acl_len;
4703 out:
4704         nfs4_set_cached_acl(inode, acl);
4705 }
4706 
4707 /*
4708  * The getxattr API returns the required buffer length when called with a
4709  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
4710  * the required buf.  On a NULL buf, we send a page of data to the server
4711  * guessing that the ACL request can be serviced by a page. If so, we cache
4712  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
4713  * the cache. If not so, we throw away the page, and cache the required
4714  * length. The next getxattr call will then produce another round trip to
4715  * the server, this time with the input buf of the required size.
4716  */
4717 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4718 {
4719         struct page *pages[NFS4ACL_MAXPAGES] = {NULL, };
4720         struct nfs_getaclargs args = {
4721                 .fh = NFS_FH(inode),
4722                 .acl_pages = pages,
4723                 .acl_len = buflen,
4724         };
4725         struct nfs_getaclres res = {
4726                 .acl_len = buflen,
4727         };
4728         struct rpc_message msg = {
4729                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
4730                 .rpc_argp = &args,
4731                 .rpc_resp = &res,
4732         };
4733         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4734         int ret = -ENOMEM, i;
4735 
4736         /* As long as we're doing a round trip to the server anyway,
4737          * let's be prepared for a page of acl data. */
4738         if (npages == 0)
4739                 npages = 1;
4740         if (npages > ARRAY_SIZE(pages))
4741                 return -ERANGE;
4742 
4743         for (i = 0; i < npages; i++) {
4744                 pages[i] = alloc_page(GFP_KERNEL);
4745                 if (!pages[i])
4746                         goto out_free;
4747         }
4748 
4749         /* for decoding across pages */
4750         res.acl_scratch = alloc_page(GFP_KERNEL);
4751         if (!res.acl_scratch)
4752                 goto out_free;
4753 
4754         args.acl_len = npages * PAGE_SIZE;
4755 
4756         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
4757                 __func__, buf, buflen, npages, args.acl_len);
4758         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
4759                              &msg, &args.seq_args, &res.seq_res, 0);
4760         if (ret)
4761                 goto out_free;
4762 
4763         /* Handle the case where the passed-in buffer is too short */
4764         if (res.acl_flags & NFS4_ACL_TRUNC) {
4765                 /* Did the user only issue a request for the acl length? */
4766                 if (buf == NULL)
4767                         goto out_ok;
4768                 ret = -ERANGE;
4769                 goto out_free;
4770         }
4771         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
4772         if (buf) {
4773                 if (res.acl_len > buflen) {
4774                         ret = -ERANGE;
4775                         goto out_free;
4776                 }
4777                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
4778         }
4779 out_ok:
4780         ret = res.acl_len;
4781 out_free:
4782         for (i = 0; i < npages; i++)
4783                 if (pages[i])
4784                         __free_page(pages[i]);
4785         if (res.acl_scratch)
4786                 __free_page(res.acl_scratch);
4787         return ret;
4788 }
4789 
4790 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
4791 {
4792         struct nfs4_exception exception = { };
4793         ssize_t ret;
4794         do {
4795                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
4796                 trace_nfs4_get_acl(inode, ret);
4797                 if (ret >= 0)
4798                         break;
4799                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
4800         } while (exception.retry);
4801         return ret;
4802 }
4803 
4804 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
4805 {
4806         struct nfs_server *server = NFS_SERVER(inode);
4807         int ret;
4808 
4809         if (!nfs4_server_supports_acls(server))
4810                 return -EOPNOTSUPP;
4811         ret = nfs_revalidate_inode(server, inode);
4812         if (ret < 0)
4813                 return ret;
4814         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
4815                 nfs_zap_acl_cache(inode);
4816         ret = nfs4_read_cached_acl(inode, buf, buflen);
4817         if (ret != -ENOENT)
4818                 /* -ENOENT is returned if there is no ACL or if there is an ACL
4819                  * but no cached acl data, just the acl length */
4820                 return ret;
4821         return nfs4_get_acl_uncached(inode, buf, buflen);
4822 }
4823 
4824 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4825 {
4826         struct nfs_server *server = NFS_SERVER(inode);
4827         struct page *pages[NFS4ACL_MAXPAGES];
4828         struct nfs_setaclargs arg = {
4829                 .fh             = NFS_FH(inode),
4830                 .acl_pages      = pages,
4831                 .acl_len        = buflen,
4832         };
4833         struct nfs_setaclres res;
4834         struct rpc_message msg = {
4835                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
4836                 .rpc_argp       = &arg,
4837                 .rpc_resp       = &res,
4838         };
4839         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
4840         int ret, i;
4841 
4842         if (!nfs4_server_supports_acls(server))
4843                 return -EOPNOTSUPP;
4844         if (npages > ARRAY_SIZE(pages))
4845                 return -ERANGE;
4846         i = buf_to_pages_noslab(buf, buflen, arg.acl_pages);
4847         if (i < 0)
4848                 return i;
4849         nfs4_inode_return_delegation(inode);
4850         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4851 
4852         /*
4853          * Free each page after tx, so the only ref left is
4854          * held by the network stack
4855          */
4856         for (; i > 0; i--)
4857                 put_page(pages[i-1]);
4858 
4859         /*
4860          * Acl update can result in inode attribute update.
4861          * so mark the attribute cache invalid.
4862          */
4863         spin_lock(&inode->i_lock);
4864         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATTR;
4865         spin_unlock(&inode->i_lock);
4866         nfs_access_zap_cache(inode);
4867         nfs_zap_acl_cache(inode);
4868         return ret;
4869 }
4870 
4871 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
4872 {
4873         struct nfs4_exception exception = { };
4874         int err;
4875         do {
4876                 err = __nfs4_proc_set_acl(inode, buf, buflen);
4877                 trace_nfs4_set_acl(inode, err);
4878                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4879                                 &exception);
4880         } while (exception.retry);
4881         return err;
4882 }
4883 
4884 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
4885 static int _nfs4_get_security_label(struct inode *inode, void *buf,
4886                                         size_t buflen)
4887 {
4888         struct nfs_server *server = NFS_SERVER(inode);
4889         struct nfs_fattr fattr;
4890         struct nfs4_label label = {0, 0, buflen, buf};
4891 
4892         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4893         struct nfs4_getattr_arg arg = {
4894                 .fh             = NFS_FH(inode),
4895                 .bitmask        = bitmask,
4896         };
4897         struct nfs4_getattr_res res = {
4898                 .fattr          = &fattr,
4899                 .label          = &label,
4900                 .server         = server,
4901         };
4902         struct rpc_message msg = {
4903                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4904                 .rpc_argp       = &arg,
4905                 .rpc_resp       = &res,
4906         };
4907         int ret;
4908 
4909         nfs_fattr_init(&fattr);
4910 
4911         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
4912         if (ret)
4913                 return ret;
4914         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
4915                 return -ENOENT;
4916         if (buflen < label.len)
4917                 return -ERANGE;
4918         return 0;
4919 }
4920 
4921 static int nfs4_get_security_label(struct inode *inode, void *buf,
4922                                         size_t buflen)
4923 {
4924         struct nfs4_exception exception = { };
4925         int err;
4926 
4927         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
4928                 return -EOPNOTSUPP;
4929 
4930         do {
4931                 err = _nfs4_get_security_label(inode, buf, buflen);
4932                 trace_nfs4_get_security_label(inode, err);
4933                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4934                                 &exception);
4935         } while (exception.retry);
4936         return err;
4937 }
4938 
4939 static int _nfs4_do_set_security_label(struct inode *inode,
4940                 struct nfs4_label *ilabel,
4941                 struct nfs_fattr *fattr,
4942                 struct nfs4_label *olabel)
4943 {
4944 
4945         struct iattr sattr = {0};
4946         struct nfs_server *server = NFS_SERVER(inode);
4947         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
4948         struct nfs_setattrargs arg = {
4949                 .fh             = NFS_FH(inode),
4950                 .iap            = &sattr,
4951                 .server         = server,
4952                 .bitmask        = bitmask,
4953                 .label          = ilabel,
4954         };
4955         struct nfs_setattrres res = {
4956                 .fattr          = fattr,
4957                 .label          = olabel,
4958                 .server         = server,
4959         };
4960         struct rpc_message msg = {
4961                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
4962                 .rpc_argp       = &arg,
4963                 .rpc_resp       = &res,
4964         };
4965         int status;
4966 
4967         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
4968 
4969         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4970         if (status)
4971                 dprintk("%s failed: %d\n", __func__, status);
4972 
4973         return status;
4974 }
4975 
4976 static int nfs4_do_set_security_label(struct inode *inode,
4977                 struct nfs4_label *ilabel,
4978                 struct nfs_fattr *fattr,
4979                 struct nfs4_label *olabel)
4980 {
4981         struct nfs4_exception exception = { };
4982         int err;
4983 
4984         do {
4985                 err = _nfs4_do_set_security_label(inode, ilabel,
4986                                 fattr, olabel);
4987                 trace_nfs4_set_security_label(inode, err);
4988                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4989                                 &exception);
4990         } while (exception.retry);
4991         return err;
4992 }
4993 
4994 static int
4995 nfs4_set_security_label(struct dentry *dentry, const void *buf, size_t buflen)
4996 {
4997         struct nfs4_label ilabel, *olabel = NULL;
4998         struct nfs_fattr fattr;
4999         struct rpc_cred *cred;
5000         struct inode *inode = d_inode(dentry);
5001         int status;
5002 
5003         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5004                 return -EOPNOTSUPP;
5005 
5006         nfs_fattr_init(&fattr);
5007 
5008         ilabel.pi = 0;
5009         ilabel.lfs = 0;
5010         ilabel.label = (char *)buf;
5011         ilabel.len = buflen;
5012 
5013         cred = rpc_lookup_cred();
5014         if (IS_ERR(cred))
5015                 return PTR_ERR(cred);
5016 
5017         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
5018         if (IS_ERR(olabel)) {
5019                 status = -PTR_ERR(olabel);
5020                 goto out;
5021         }
5022 
5023         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
5024         if (status == 0)
5025                 nfs_setsecurity(inode, &fattr, olabel);
5026 
5027         nfs4_label_free(olabel);
5028 out:
5029         put_rpccred(cred);
5030         return status;
5031 }
5032 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
5033 
5034 
5035 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
5036                                     nfs4_verifier *bootverf)
5037 {
5038         __be32 verf[2];
5039 
5040         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
5041                 /* An impossible timestamp guarantees this value
5042                  * will never match a generated boot time. */
5043                 verf[0] = 0;
5044                 verf[1] = cpu_to_be32(NSEC_PER_SEC + 1);
5045         } else {
5046                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
5047                 verf[0] = cpu_to_be32(nn->boot_time.tv_sec);
5048                 verf[1] = cpu_to_be32(nn->boot_time.tv_nsec);
5049         }
5050         memcpy(bootverf->data, verf, sizeof(bootverf->data));
5051 }
5052 
5053 static int
5054 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
5055 {
5056         size_t len;
5057         char *str;
5058 
5059         if (clp->cl_owner_id != NULL)
5060                 return 0;
5061 
5062         rcu_read_lock();
5063         len = 14 + strlen(clp->cl_ipaddr) + 1 +
5064                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
5065                 1 +
5066                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) +
5067                 1;
5068         rcu_read_unlock();
5069 
5070         if (len > NFS4_OPAQUE_LIMIT + 1)
5071                 return -EINVAL;
5072 
5073         /*
5074          * Since this string is allocated at mount time, and held until the
5075          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5076          * about a memory-reclaim deadlock.
5077          */
5078         str = kmalloc(len, GFP_KERNEL);
5079         if (!str)
5080                 return -ENOMEM;
5081 
5082         rcu_read_lock();
5083         scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
5084                         clp->cl_ipaddr,
5085                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
5086                         rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
5087         rcu_read_unlock();
5088 
5089         clp->cl_owner_id = str;
5090         return 0;
5091 }
5092 
5093 static int
5094 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
5095 {
5096         size_t len;
5097         char *str;
5098 
5099         len = 10 + 10 + 1 + 10 + 1 +
5100                 strlen(nfs4_client_id_uniquifier) + 1 +
5101                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5102 
5103         if (len > NFS4_OPAQUE_LIMIT + 1)
5104                 return -EINVAL;
5105 
5106         /*
5107          * Since this string is allocated at mount time, and held until the
5108          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5109          * about a memory-reclaim deadlock.
5110          */
5111         str = kmalloc(len, GFP_KERNEL);
5112         if (!str)
5113                 return -ENOMEM;
5114 
5115         scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
5116                         clp->rpc_ops->version, clp->cl_minorversion,
5117                         nfs4_client_id_uniquifier,
5118                         clp->cl_rpcclient->cl_nodename);
5119         clp->cl_owner_id = str;
5120         return 0;
5121 }
5122 
5123 static int
5124 nfs4_init_uniform_client_string(struct nfs_client *clp)
5125 {
5126         size_t len;
5127         char *str;
5128 
5129         if (clp->cl_owner_id != NULL)
5130                 return 0;
5131 
5132         if (nfs4_client_id_uniquifier[0] != '\0')
5133                 return nfs4_init_uniquifier_client_string(clp);
5134 
5135         len = 10 + 10 + 1 + 10 + 1 +
5136                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
5137 
5138         if (len > NFS4_OPAQUE_LIMIT + 1)
5139                 return -EINVAL;
5140 
5141         /*
5142          * Since this string is allocated at mount time, and held until the
5143          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
5144          * about a memory-reclaim deadlock.
5145          */
5146         str = kmalloc(len, GFP_KERNEL);
5147         if (!str)
5148                 return -ENOMEM;
5149 
5150         scnprintf(str, len, "Linux NFSv%u.%u %s",
5151                         clp->rpc_ops->version, clp->cl_minorversion,
5152                         clp->cl_rpcclient->cl_nodename);
5153         clp->cl_owner_id = str;
5154         return 0;
5155 }
5156 
5157 /*
5158  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
5159  * services.  Advertise one based on the address family of the
5160  * clientaddr.
5161  */
5162 static unsigned int
5163 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
5164 {
5165         if (strchr(clp->cl_ipaddr, ':') != NULL)
5166                 return scnprintf(buf, len, "tcp6");
5167         else
5168                 return scnprintf(buf, len, "tcp");
5169 }
5170 
5171 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
5172 {
5173         struct nfs4_setclientid *sc = calldata;
5174 
5175         if (task->tk_status == 0)
5176                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
5177 }
5178 
5179 static const struct rpc_call_ops nfs4_setclientid_ops = {
5180         .rpc_call_done = nfs4_setclientid_done,
5181 };
5182 
5183 /**
5184  * nfs4_proc_setclientid - Negotiate client ID
5185  * @clp: state data structure
5186  * @program: RPC program for NFSv4 callback service
5187  * @port: IP port number for NFS4 callback service
5188  * @cred: RPC credential to use for this call
5189  * @res: where to place the result
5190  *
5191  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5192  */
5193 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
5194                 unsigned short port, struct rpc_cred *cred,
5195                 struct nfs4_setclientid_res *res)
5196 {
5197         nfs4_verifier sc_verifier;
5198         struct nfs4_setclientid setclientid = {
5199                 .sc_verifier = &sc_verifier,
5200                 .sc_prog = program,
5201                 .sc_clnt = clp,
5202         };
5203         struct rpc_message msg = {
5204                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
5205                 .rpc_argp = &setclientid,
5206                 .rpc_resp = res,
5207                 .rpc_cred = cred,
5208         };
5209         struct rpc_task *task;
5210         struct rpc_task_setup task_setup_data = {
5211                 .rpc_client = clp->cl_rpcclient,
5212                 .rpc_message = &msg,
5213                 .callback_ops = &nfs4_setclientid_ops,
5214                 .callback_data = &setclientid,
5215                 .flags = RPC_TASK_TIMEOUT,
5216         };
5217         int status;
5218 
5219         /* nfs_client_id4 */
5220         nfs4_init_boot_verifier(clp, &sc_verifier);
5221 
5222         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
5223                 status = nfs4_init_uniform_client_string(clp);
5224         else
5225                 status = nfs4_init_nonuniform_client_string(clp);
5226 
5227         if (status)
5228                 goto out;
5229 
5230         /* cb_client4 */
5231         setclientid.sc_netid_len =
5232                                 nfs4_init_callback_netid(clp,
5233                                                 setclientid.sc_netid,
5234                                                 sizeof(setclientid.sc_netid));
5235         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
5236                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
5237                                 clp->cl_ipaddr, port >> 8, port & 255);
5238 
5239         dprintk("NFS call  setclientid auth=%s, '%s'\n",
5240                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5241                 clp->cl_owner_id);
5242         task = rpc_run_task(&task_setup_data);
5243         if (IS_ERR(task)) {
5244                 status = PTR_ERR(task);
5245                 goto out;
5246         }
5247         status = task->tk_status;
5248         if (setclientid.sc_cred) {
5249                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
5250                 put_rpccred(setclientid.sc_cred);
5251         }
5252         rpc_put_task(task);
5253 out:
5254         trace_nfs4_setclientid(clp, status);
5255         dprintk("NFS reply setclientid: %d\n", status);
5256         return status;
5257 }
5258 
5259 /**
5260  * nfs4_proc_setclientid_confirm - Confirm client ID
5261  * @clp: state data structure
5262  * @res: result of a previous SETCLIENTID
5263  * @cred: RPC credential to use for this call
5264  *
5265  * Returns zero, a negative errno, or a negative NFS4ERR status code.
5266  */
5267 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
5268                 struct nfs4_setclientid_res *arg,
5269                 struct rpc_cred *cred)
5270 {
5271         struct rpc_message msg = {
5272                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
5273                 .rpc_argp = arg,
5274                 .rpc_cred = cred,
5275         };
5276         int status;
5277 
5278         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
5279                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
5280                 clp->cl_clientid);
5281         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5282         trace_nfs4_setclientid_confirm(clp, status);
5283         dprintk("NFS reply setclientid_confirm: %d\n", status);
5284         return status;
5285 }
5286 
5287 struct nfs4_delegreturndata {
5288         struct nfs4_delegreturnargs args;
5289         struct nfs4_delegreturnres res;
5290         struct nfs_fh fh;
5291         nfs4_stateid stateid;
5292         unsigned long timestamp;
5293         struct nfs_fattr fattr;
5294         int rpc_status;
5295         struct inode *inode;
5296         bool roc;
5297         u32 roc_barrier;
5298 };
5299 
5300 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
5301 {
5302         struct nfs4_delegreturndata *data = calldata;
5303 
5304         if (!nfs4_sequence_done(task, &data->res.seq_res))
5305                 return;
5306 
5307         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
5308         switch (task->tk_status) {
5309         case 0:
5310                 renew_lease(data->res.server, data->timestamp);
5311         case -NFS4ERR_ADMIN_REVOKED:
5312         case -NFS4ERR_DELEG_REVOKED:
5313         case -NFS4ERR_BAD_STATEID:
5314         case -NFS4ERR_OLD_STATEID:
5315         case -NFS4ERR_STALE_STATEID:
5316         case -NFS4ERR_EXPIRED:
5317                 task->tk_status = 0;
5318                 if (data->roc)
5319                         pnfs_roc_set_barrier(data->inode, data->roc_barrier);
5320                 break;
5321         default:
5322                 if (nfs4_async_handle_error(task, data->res.server,
5323                                             NULL, NULL) == -EAGAIN) {
5324                         rpc_restart_call_prepare(task);
5325                         return;
5326                 }
5327         }
5328         data->rpc_status = task->tk_status;
5329 }
5330 
5331 static void nfs4_delegreturn_release(void *calldata)
5332 {
5333         struct nfs4_delegreturndata *data = calldata;
5334         struct inode *inode = data->inode;
5335 
5336         if (inode) {
5337                 if (data->roc)
5338                         pnfs_roc_release(inode);
5339                 nfs_iput_and_deactive(inode);
5340         }
5341         kfree(calldata);
5342 }
5343 
5344 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
5345 {
5346         struct nfs4_delegreturndata *d_data;
5347 
5348         d_data = (struct nfs4_delegreturndata *)data;
5349 
5350         if (nfs4_wait_on_layoutreturn(d_data->inode, task))
5351                 return;
5352 
5353         if (d_data->roc)
5354                 pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier);
5355 
5356         nfs4_setup_sequence(d_data->res.server,
5357                         &d_data->args.seq_args,
5358                         &d_data->res.seq_res,
5359                         task);
5360 }
5361 
5362 static const struct rpc_call_ops nfs4_delegreturn_ops = {
5363         .rpc_call_prepare = nfs4_delegreturn_prepare,
5364         .rpc_call_done = nfs4_delegreturn_done,
5365         .rpc_release = nfs4_delegreturn_release,
5366 };
5367 
5368 static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5369 {
5370         struct nfs4_delegreturndata *data;
5371         struct nfs_server *server = NFS_SERVER(inode);
5372         struct rpc_task *task;
5373         struct rpc_message msg = {
5374                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
5375                 .rpc_cred = cred,
5376         };
5377         struct rpc_task_setup task_setup_data = {
5378                 .rpc_client = server->client,
5379                 .rpc_message = &msg,
5380                 .callback_ops = &nfs4_delegreturn_ops,
5381                 .flags = RPC_TASK_ASYNC,
5382         };
5383         int status = 0;
5384 
5385         data = kzalloc(sizeof(*data), GFP_NOFS);
5386         if (data == NULL)
5387                 return -ENOMEM;
5388         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
5389 
5390         nfs4_state_protect(server->nfs_client,
5391                         NFS_SP4_MACH_CRED_CLEANUP,
5392                         &task_setup_data.rpc_client, &msg);
5393 
5394         data->args.fhandle = &data->fh;
5395         data->args.stateid = &data->stateid;
5396         data->args.bitmask = server->cache_consistency_bitmask;
5397         nfs_copy_fh(&data->fh, NFS_FH(inode));
5398         nfs4_stateid_copy(&data->stateid, stateid);
5399         data->res.fattr = &data->fattr;
5400         data->res.server = server;
5401         nfs_fattr_init(data->res.fattr);
5402         data->timestamp = jiffies;
5403         data->rpc_status = 0;
5404         data->inode = nfs_igrab_and_active(inode);
5405         if (data->inode)
5406                 data->roc = nfs4_roc(inode);
5407 
5408         task_setup_data.callback_data = data;
5409         msg.rpc_argp = &data->args;
5410         msg.rpc_resp = &data->res;
5411         task = rpc_run_task(&task_setup_data);
5412         if (IS_ERR(task))
5413                 return PTR_ERR(task);
5414         if (!issync)
5415                 goto out;
5416         status = nfs4_wait_for_completion_rpc_task(task);
5417         if (status != 0)
5418                 goto out;
5419         status = data->rpc_status;
5420         if (status == 0)
5421                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
5422         else
5423                 nfs_refresh_inode(inode, &data->fattr);
5424 out:
5425         rpc_put_task(task);
5426         return status;
5427 }
5428 
5429 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync)
5430 {
5431         struct nfs_server *server = NFS_SERVER(inode);
5432         struct nfs4_exception exception = { };
5433         int err;
5434         do {
5435                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
5436                 trace_nfs4_delegreturn(inode, stateid, err);
5437                 switch (err) {
5438                         case -NFS4ERR_STALE_STATEID:
5439                         case -NFS4ERR_EXPIRED:
5440                         case 0:
5441                                 return 0;
5442                 }
5443                 err = nfs4_handle_exception(server, err, &exception);
5444         } while (exception.retry);
5445         return err;
5446 }
5447 
5448 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
5449 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
5450 
5451 /* 
5452  * sleep, with exponential backoff, and retry the LOCK operation. 
5453  */
5454 static unsigned long
5455 nfs4_set_lock_task_retry(unsigned long timeout)
5456 {
5457         freezable_schedule_timeout_killable_unsafe(timeout);
5458         timeout <<= 1;
5459         if (timeout > NFS4_LOCK_MAXTIMEOUT)
5460                 return NFS4_LOCK_MAXTIMEOUT;
5461         return timeout;
5462 }
5463 
5464 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5465 {
5466         struct inode *inode = state->inode;
5467         struct nfs_server *server = NFS_SERVER(inode);
5468         struct nfs_client *clp = server->nfs_client;
5469         struct nfs_lockt_args arg = {
5470                 .fh = NFS_FH(inode),
5471                 .fl = request,
5472         };
5473         struct nfs_lockt_res res = {
5474                 .denied = request,
5475         };
5476         struct rpc_message msg = {
5477                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
5478                 .rpc_argp       = &arg,
5479                 .rpc_resp       = &res,
5480                 .rpc_cred       = state->owner->so_cred,
5481         };
5482         struct nfs4_lock_state *lsp;
5483         int status;
5484 
5485         arg.lock_owner.clientid = clp->cl_clientid;
5486         status = nfs4_set_lock_state(state, request);
5487         if (status != 0)
5488                 goto out;
5489         lsp = request->fl_u.nfs4_fl.owner;
5490         arg.lock_owner.id = lsp->ls_seqid.owner_id;
5491         arg.lock_owner.s_dev = server->s_dev;
5492         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5493         switch (status) {
5494                 case 0:
5495                         request->fl_type = F_UNLCK;
5496                         break;
5497                 case -NFS4ERR_DENIED:
5498                         status = 0;
5499         }
5500         request->fl_ops->fl_release_private(request);
5501         request->fl_ops = NULL;
5502 out:
5503         return status;
5504 }
5505 
5506 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
5507 {
5508         struct nfs4_exception exception = { };
5509         int err;
5510 
5511         do {
5512                 err = _nfs4_proc_getlk(state, cmd, request);
5513                 trace_nfs4_get_lock(request, state, cmd, err);
5514                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
5515                                 &exception);
5516         } while (exception.retry);
5517         return err;
5518 }
5519 
5520 static int do_vfs_lock(struct inode *inode, struct file_lock *fl)
5521 {
5522         return locks_lock_inode_wait(inode, fl);
5523 }
5524 
5525 struct nfs4_unlockdata {
5526         struct nfs_locku_args arg;
5527         struct nfs_locku_res res;
5528         struct nfs4_lock_state *lsp;
5529         struct nfs_open_context *ctx;
5530         struct file_lock fl;
5531         struct nfs_server *server;
5532         unsigned long timestamp;
5533 };
5534 
5535 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
5536                 struct nfs_open_context *ctx,
5537                 struct nfs4_lock_state *lsp,
5538                 struct nfs_seqid *seqid)
5539 {
5540         struct nfs4_unlockdata *p;
5541         struct inode *inode = lsp->ls_state->inode;
5542 
5543         p = kzalloc(sizeof(*p), GFP_NOFS);
5544         if (p == NULL)
5545                 return NULL;
5546         p->arg.fh = NFS_FH(inode);
5547         p->arg.fl = &p->fl;
5548         p->arg.seqid = seqid;
5549         p->res.seqid = seqid;
5550         p->lsp = lsp;
5551         atomic_inc(&lsp->ls_count);
5552         /* Ensure we don't close file until we're done freeing locks! */
5553         p->ctx = get_nfs_open_context(ctx);
5554         memcpy(&p->fl, fl, sizeof(p->fl));
5555         p->server = NFS_SERVER(inode);
5556         return p;
5557 }
5558 
5559 static void nfs4_locku_release_calldata(void *data)
5560 {
5561         struct nfs4_unlockdata *calldata = data;
5562         nfs_free_seqid(calldata->arg.seqid);
5563         nfs4_put_lock_state(calldata->lsp);
5564         put_nfs_open_context(calldata->ctx);
5565         kfree(calldata);
5566 }
5567 
5568 static void nfs4_locku_done(struct rpc_task *task, void *data)
5569 {
5570         struct nfs4_unlockdata *calldata = data;
5571 
5572         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
5573                 return;
5574         switch (task->tk_status) {
5575                 case 0:
5576                         renew_lease(calldata->server, calldata->timestamp);
5577                         do_vfs_lock(calldata->lsp->ls_state->inode, &calldata->fl);
5578                         if (nfs4_update_lock_stateid(calldata->lsp,
5579                                         &calldata->res.stateid))
5580                                 break;
5581                 case -NFS4ERR_BAD_STATEID:
5582                 case -NFS4ERR_OLD_STATEID:
5583                 case -NFS4ERR_STALE_STATEID:
5584                 case -NFS4ERR_EXPIRED:
5585                         if (!nfs4_stateid_match(&calldata->arg.stateid,
5586                                                 &calldata->lsp->ls_stateid))
5587                                 rpc_restart_call_prepare(task);
5588                         break;
5589                 default:
5590                         if (nfs4_async_handle_error(task, calldata->server,
5591                                                     NULL, NULL) == -EAGAIN)
5592                                 rpc_restart_call_prepare(task);
5593         }
5594         nfs_release_seqid(calldata->arg.seqid);
5595 }
5596 
5597 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
5598 {
5599         struct nfs4_unlockdata *calldata = data;
5600 
5601         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
5602                 goto out_wait;
5603         nfs4_stateid_copy(&calldata->arg.stateid, &calldata->lsp->ls_stateid);
5604         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
5605                 /* Note: exit _without_ running nfs4_locku_done */
5606                 goto out_no_action;
5607         }
5608         calldata->timestamp = jiffies;
5609         if (nfs4_setup_sequence(calldata->server,
5610                                 &calldata->arg.seq_args,
5611                                 &calldata->res.seq_res,
5612                                 task) != 0)
5613                 nfs_release_seqid(calldata->arg.seqid);
5614         return;
5615 out_no_action:
5616         task->tk_action = NULL;
5617 out_wait:
5618         nfs4_sequence_done(task, &calldata->res.seq_res);
5619 }
5620 
5621 static const struct rpc_call_ops nfs4_locku_ops = {
5622         .rpc_call_prepare = nfs4_locku_prepare,
5623         .rpc_call_done = nfs4_locku_done,
5624         .rpc_release = nfs4_locku_release_calldata,
5625 };
5626 
5627 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
5628                 struct nfs_open_context *ctx,
5629                 struct nfs4_lock_state *lsp,
5630                 struct nfs_seqid *seqid)
5631 {
5632         struct nfs4_unlockdata *data;
5633         struct rpc_message msg = {
5634                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
5635                 .rpc_cred = ctx->cred,
5636         };
5637         struct rpc_task_setup task_setup_data = {
5638                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
5639                 .rpc_message = &msg,
5640                 .callback_ops = &nfs4_locku_ops,
5641                 .workqueue = nfsiod_workqueue,
5642                 .flags = RPC_TASK_ASYNC,
5643         };
5644 
5645         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
5646                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
5647 
5648         /* Ensure this is an unlock - when canceling a lock, the
5649          * canceled lock is passed in, and it won't be an unlock.
5650          */
5651         fl->fl_type = F_UNLCK;
5652 
5653         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
5654         if (data == NULL) {
5655                 nfs_free_seqid(seqid);
5656                 return ERR_PTR(-ENOMEM);
5657         }
5658 
5659         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1);
5660         msg.rpc_argp = &data->arg;
5661         msg.rpc_resp = &data->res;
5662         task_setup_data.callback_data = data;
5663         return rpc_run_task(&task_setup_data);
5664 }
5665 
5666 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
5667 {
5668         struct inode *inode = state->inode;
5669         struct nfs4_state_owner *sp = state->owner;
5670         struct nfs_inode *nfsi = NFS_I(inode);
5671         struct nfs_seqid *seqid;
5672         struct nfs4_lock_state *lsp;
5673         struct rpc_task *task;
5674         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
5675         int status = 0;
5676         unsigned char fl_flags = request->fl_flags;
5677 
5678         status = nfs4_set_lock_state(state, request);
5679         /* Unlock _before_ we do the RPC call */
5680         request->fl_flags |= FL_EXISTS;
5681         /* Exclude nfs_delegation_claim_locks() */
5682         mutex_lock(&sp->so_delegreturn_mutex);
5683         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
5684         down_read(&nfsi->rwsem);
5685         if (do_vfs_lock(inode, request) == -ENOENT) {
5686                 up_read(&nfsi->rwsem);
5687                 mutex_unlock(&sp->so_delegreturn_mutex);
5688                 goto out;
5689         }
5690         up_read(&nfsi->rwsem);
5691         mutex_unlock(&sp->so_delegreturn_mutex);
5692         if (status != 0)
5693                 goto out;
5694         /* Is this a delegated lock? */
5695         lsp = request->fl_u.nfs4_fl.owner;
5696         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)