fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.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 "sysfs.h"
  67 #include "nfs4idmap.h"
  68 #include "nfs4session.h"
  69 #include "fscache.h"
  70 #include "nfs42.h"
  71
  72 #include "nfs4trace.h"
  73
  74 #define NFSDBG_FACILITY         NFSDBG_PROC
  75
  76 #define NFS4_BITMASK_SZ         3
  77
  78 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  79 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  80
  81 /* file attributes which can be mapped to nfs attributes */
  82 #define NFS4_VALID_ATTRS (ATTR_MODE \
  83         | ATTR_UID \
  84         | ATTR_GID \
  85         | ATTR_SIZE \
  86         | ATTR_ATIME \
  87         | ATTR_MTIME \
  88         | ATTR_CTIME \
  89         | ATTR_ATIME_SET \
  90         | ATTR_MTIME_SET)
  91
  92 struct nfs4_opendata;
  93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
  97                               struct nfs_fattr *fattr, struct inode *inode);
  98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  99                             struct nfs_fattr *fattr, struct iattr *sattr,
 100                             struct nfs_open_context *ctx, struct nfs4_label *ilabel);
 101 #ifdef CONFIG_NFS_V4_1
 102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 103                 const struct cred *cred,
 104                 struct nfs4_slot *slot,
 105                 bool is_privileged);
 106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 107                 const struct cred *);
 108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 109                 const struct cred *, bool);
 110 #endif
 111 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ],
 112                              const __u32 *src, struct inode *inode,
 113                              struct nfs_server *server,
 114                              struct nfs4_label *label);
 115
 116 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 117 static inline struct nfs4_label *
 118 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 119         struct iattr *sattr, struct nfs4_label *label)
 120 {
 121         int err;
 122
 123         if (label == NULL)
 124                 return NULL;
 125
 126         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 127                 return NULL;
 128
 129         err = security_dentry_init_security(dentry, sattr->ia_mode,
 130                                 &dentry->d_name, NULL,
 131                                 (void **)&label->label, &label->len);
 132         if (err == 0)
 133                 return label;
 134
 135         return NULL;
 136 }
 137 static inline void
 138 nfs4_label_release_security(struct nfs4_label *label)
 139 {
 140         if (label)
 141                 security_release_secctx(label->label, label->len);
 142 }
 143 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 144 {
 145         if (label)
 146                 return server->attr_bitmask;
 147
 148         return server->attr_bitmask_nl;
 149 }
 150 #else
 151 static inline struct nfs4_label *
 152 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 153         struct iattr *sattr, struct nfs4_label *l)
 154 { return NULL; }
 155 static inline void
 156 nfs4_label_release_security(struct nfs4_label *label)
 157 { return; }
 158 static inline u32 *
 159 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 160 { return server->attr_bitmask; }
 161 #endif
 162
 163 /* Prevent leaks of NFSv4 errors into userland */
 164 static int nfs4_map_errors(int err)
 165 {
 166         if (err >= -1000)
 167                 return err;
 168         switch (err) {
 169         case -NFS4ERR_RESOURCE:
 170         case -NFS4ERR_LAYOUTTRYLATER:
 171         case -NFS4ERR_RECALLCONFLICT:
 172                 return -EREMOTEIO;
 173         case -NFS4ERR_WRONGSEC:
 174         case -NFS4ERR_WRONG_CRED:
 175                 return -EPERM;
 176         case -NFS4ERR_BADOWNER:
 177         case -NFS4ERR_BADNAME:
 178                 return -EINVAL;
 179         case -NFS4ERR_SHARE_DENIED:
 180                 return -EACCES;
 181         case -NFS4ERR_MINOR_VERS_MISMATCH:
 182                 return -EPROTONOSUPPORT;
 183         case -NFS4ERR_FILE_OPEN:
 184                 return -EBUSY;
 185         case -NFS4ERR_NOT_SAME:
 186                 return -ENOTSYNC;
 187         default:
 188                 dprintk("%s could not handle NFSv4 error %d\n",
 189                                 __func__, -err);
 190                 break;
 191         }
 192         return -EIO;
 193 }
 194
 195 /*
 196  * This is our standard bitmap for GETATTR requests.
 197  */
 198 const u32 nfs4_fattr_bitmap[3] = {
 199         FATTR4_WORD0_TYPE
 200         | FATTR4_WORD0_CHANGE
 201         | FATTR4_WORD0_SIZE
 202         | FATTR4_WORD0_FSID
 203         | FATTR4_WORD0_FILEID,
 204         FATTR4_WORD1_MODE
 205         | FATTR4_WORD1_NUMLINKS
 206         | FATTR4_WORD1_OWNER
 207         | FATTR4_WORD1_OWNER_GROUP
 208         | FATTR4_WORD1_RAWDEV
 209         | FATTR4_WORD1_SPACE_USED
 210         | FATTR4_WORD1_TIME_ACCESS
 211         | FATTR4_WORD1_TIME_METADATA
 212         | FATTR4_WORD1_TIME_MODIFY
 213         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 214 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 215         FATTR4_WORD2_SECURITY_LABEL
 216 #endif
 217 };
 218
 219 static const u32 nfs4_pnfs_open_bitmap[3] = {
 220         FATTR4_WORD0_TYPE
 221         | FATTR4_WORD0_CHANGE
 222         | FATTR4_WORD0_SIZE
 223         | FATTR4_WORD0_FSID
 224         | FATTR4_WORD0_FILEID,
 225         FATTR4_WORD1_MODE
 226         | FATTR4_WORD1_NUMLINKS
 227         | FATTR4_WORD1_OWNER
 228         | FATTR4_WORD1_OWNER_GROUP
 229         | FATTR4_WORD1_RAWDEV
 230         | FATTR4_WORD1_SPACE_USED
 231         | FATTR4_WORD1_TIME_ACCESS
 232         | FATTR4_WORD1_TIME_METADATA
 233         | FATTR4_WORD1_TIME_MODIFY,
 234         FATTR4_WORD2_MDSTHRESHOLD
 235 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 236         | FATTR4_WORD2_SECURITY_LABEL
 237 #endif
 238 };
 239
 240 static const u32 nfs4_open_noattr_bitmap[3] = {
 241         FATTR4_WORD0_TYPE
 242         | FATTR4_WORD0_FILEID,
 243 };
 244
 245 const u32 nfs4_statfs_bitmap[3] = {
 246         FATTR4_WORD0_FILES_AVAIL
 247         | FATTR4_WORD0_FILES_FREE
 248         | FATTR4_WORD0_FILES_TOTAL,
 249         FATTR4_WORD1_SPACE_AVAIL
 250         | FATTR4_WORD1_SPACE_FREE
 251         | FATTR4_WORD1_SPACE_TOTAL
 252 };
 253
 254 const u32 nfs4_pathconf_bitmap[3] = {
 255         FATTR4_WORD0_MAXLINK
 256         | FATTR4_WORD0_MAXNAME,
 257         0
 258 };
 259
 260 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 261                         | FATTR4_WORD0_MAXREAD
 262                         | FATTR4_WORD0_MAXWRITE
 263                         | FATTR4_WORD0_LEASE_TIME,
 264                         FATTR4_WORD1_TIME_DELTA
 265                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 266                         FATTR4_WORD2_LAYOUT_BLKSIZE
 267                         | FATTR4_WORD2_CLONE_BLKSIZE
 268                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
 269                         | FATTR4_WORD2_XATTR_SUPPORT
 270 };
 271
 272 const u32 nfs4_fs_locations_bitmap[3] = {
 273         FATTR4_WORD0_CHANGE
 274         | FATTR4_WORD0_SIZE
 275         | FATTR4_WORD0_FSID
 276         | FATTR4_WORD0_FILEID
 277         | FATTR4_WORD0_FS_LOCATIONS,
 278         FATTR4_WORD1_OWNER
 279         | FATTR4_WORD1_OWNER_GROUP
 280         | FATTR4_WORD1_RAWDEV
 281         | FATTR4_WORD1_SPACE_USED
 282         | FATTR4_WORD1_TIME_ACCESS
 283         | FATTR4_WORD1_TIME_METADATA
 284         | FATTR4_WORD1_TIME_MODIFY
 285         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 286 };
 287
 288 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 289                                     struct inode *inode, unsigned long flags)
 290 {
 291         unsigned long cache_validity;
 292
 293         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 294         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 295                 return;
 296
 297         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
 298
 299         /* Remove the attributes over which we have full control */
 300         dst[1] &= ~FATTR4_WORD1_RAWDEV;
 301         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 302                 dst[0] &= ~FATTR4_WORD0_SIZE;
 303
 304         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 305                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 306
 307         if (!(cache_validity & NFS_INO_INVALID_MODE))
 308                 dst[1] &= ~FATTR4_WORD1_MODE;
 309         if (!(cache_validity & NFS_INO_INVALID_OTHER))
 310                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
 311 }
 312
 313 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 314                 struct nfs4_readdir_arg *readdir)
 315 {
 316         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 317         __be32 *start, *p;
 318
 319         if (cookie > 2) {
 320                 readdir->cookie = cookie;
 321                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 322                 return;
 323         }
 324
 325         readdir->cookie = 0;
 326         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 327         if (cookie == 2)
 328                 return;
 329
 330         /*
 331          * NFSv4 servers do not return entries for '.' and '..'
 332          * Therefore, we fake these entries here.  We let '.'
 333          * have cookie 0 and '..' have cookie 1.  Note that
 334          * when talking to the server, we always send cookie 0
 335          * instead of 1 or 2.
 336          */
 337         start = p = kmap_atomic(*readdir->pages);
 338
 339         if (cookie == 0) {
 340                 *p++ = xdr_one;                                  /* next */
 341                 *p++ = xdr_zero;                   /* cookie, first word */
 342                 *p++ = xdr_one;                   /* cookie, second word */
 343                 *p++ = xdr_one;                             /* entry len */
 344                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 345                 p++;
 346                 *p++ = xdr_one;                         /* bitmap length */
 347                 *p++ = htonl(attrs);                           /* bitmap */
 348                 *p++ = htonl(12);             /* attribute buffer length */
 349                 *p++ = htonl(NF4DIR);
 350                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 351         }
 352
 353         *p++ = xdr_one;                                  /* next */
 354         *p++ = xdr_zero;                   /* cookie, first word */
 355         *p++ = xdr_two;                   /* cookie, second word */
 356         *p++ = xdr_two;                             /* entry len */
 357         memcpy(p, "..\0\0", 4);                         /* entry */
 358         p++;
 359         *p++ = xdr_one;                         /* bitmap length */
 360         *p++ = htonl(attrs);                           /* bitmap */
 361         *p++ = htonl(12);             /* attribute buffer length */
 362         *p++ = htonl(NF4DIR);
 363         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 364
 365         readdir->pgbase = (char *)p - (char *)start;
 366         readdir->count -= readdir->pgbase;
 367         kunmap_atomic(start);
 368 }
 369
 370 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 371                 nfs4_stateid *stateid,
 372                 const struct cred *cred)
 373 {
 374         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 375
 376         ops->test_and_free_expired(server, stateid, cred);
 377 }
 378
 379 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 380                 nfs4_stateid *stateid,
 381                 const struct cred *cred)
 382 {
 383         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 384         nfs4_test_and_free_stateid(server, stateid, cred);
 385 }
 386
 387 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 388                 const nfs4_stateid *stateid,
 389                 const struct cred *cred)
 390 {
 391         nfs4_stateid tmp;
 392
 393         nfs4_stateid_copy(&tmp, stateid);
 394         __nfs4_free_revoked_stateid(server, &tmp, cred);
 395 }
 396
 397 static long nfs4_update_delay(long *timeout)
 398 {
 399         long ret;
 400         if (!timeout)
 401                 return NFS4_POLL_RETRY_MAX;
 402         if (*timeout <= 0)
 403                 *timeout = NFS4_POLL_RETRY_MIN;
 404         if (*timeout > NFS4_POLL_RETRY_MAX)
 405                 *timeout = NFS4_POLL_RETRY_MAX;
 406         ret = *timeout;
 407         *timeout <<= 1;
 408         return ret;
 409 }
 410
 411 static int nfs4_delay_killable(long *timeout)
 412 {
 413         might_sleep();
 414
 415         freezable_schedule_timeout_killable_unsafe(
 416                 nfs4_update_delay(timeout));
 417         if (!__fatal_signal_pending(current))
 418                 return 0;
 419         return -EINTR;
 420 }
 421
 422 static int nfs4_delay_interruptible(long *timeout)
 423 {
 424         might_sleep();
 425
 426         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 427         if (!signal_pending(current))
 428                 return 0;
 429         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 430 }
 431
 432 static int nfs4_delay(long *timeout, bool interruptible)
 433 {
 434         if (interruptible)
 435                 return nfs4_delay_interruptible(timeout);
 436         return nfs4_delay_killable(timeout);
 437 }
 438
 439 static const nfs4_stateid *
 440 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 441 {
 442         if (!stateid)
 443                 return NULL;
 444         switch (stateid->type) {
 445         case NFS4_OPEN_STATEID_TYPE:
 446         case NFS4_LOCK_STATEID_TYPE:
 447         case NFS4_DELEGATION_STATEID_TYPE:
 448                 return stateid;
 449         default:
 450                 break;
 451         }
 452         return NULL;
 453 }
 454
 455 /* This is the error handling routine for processes that are allowed
 456  * to sleep.
 457  */
 458 static int nfs4_do_handle_exception(struct nfs_server *server,
 459                 int errorcode, struct nfs4_exception *exception)
 460 {
 461         struct nfs_client *clp = server->nfs_client;
 462         struct nfs4_state *state = exception->state;
 463         const nfs4_stateid *stateid;
 464         struct inode *inode = exception->inode;
 465         int ret = errorcode;
 466
 467         exception->delay = 0;
 468         exception->recovering = 0;
 469         exception->retry = 0;
 470
 471         stateid = nfs4_recoverable_stateid(exception->stateid);
 472         if (stateid == NULL && state != NULL)
 473                 stateid = nfs4_recoverable_stateid(&state->stateid);
 474
 475         switch(errorcode) {
 476                 case 0:
 477                         return 0;
 478                 case -NFS4ERR_BADHANDLE:
 479                 case -ESTALE:
 480                         if (inode != NULL && S_ISREG(inode->i_mode))
 481                                 pnfs_destroy_layout(NFS_I(inode));
 482                         break;
 483                 case -NFS4ERR_DELEG_REVOKED:
 484                 case -NFS4ERR_ADMIN_REVOKED:
 485                 case -NFS4ERR_EXPIRED:
 486                 case -NFS4ERR_BAD_STATEID:
 487                 case -NFS4ERR_PARTNER_NO_AUTH:
 488                         if (inode != NULL && stateid != NULL) {
 489                                 nfs_inode_find_state_and_recover(inode,
 490                                                 stateid);
 491                                 goto wait_on_recovery;
 492                         }
 493                         fallthrough;
 494                 case -NFS4ERR_OPENMODE:
 495                         if (inode) {
 496                                 int err;
 497
 498                                 err = nfs_async_inode_return_delegation(inode,
 499                                                 stateid);
 500                                 if (err == 0)
 501                                         goto wait_on_recovery;
 502                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 503                                         exception->retry = 1;
 504                                         break;
 505                                 }
 506                         }
 507                         if (state == NULL)
 508                                 break;
 509                         ret = nfs4_schedule_stateid_recovery(server, state);
 510                         if (ret < 0)
 511                                 break;
 512                         goto wait_on_recovery;
 513                 case -NFS4ERR_STALE_STATEID:
 514                 case -NFS4ERR_STALE_CLIENTID:
 515                         nfs4_schedule_lease_recovery(clp);
 516                         goto wait_on_recovery;
 517                 case -NFS4ERR_MOVED:
 518                         ret = nfs4_schedule_migration_recovery(server);
 519                         if (ret < 0)
 520                                 break;
 521                         goto wait_on_recovery;
 522                 case -NFS4ERR_LEASE_MOVED:
 523                         nfs4_schedule_lease_moved_recovery(clp);
 524                         goto wait_on_recovery;
 525 #if defined(CONFIG_NFS_V4_1)
 526                 case -NFS4ERR_BADSESSION:
 527                 case -NFS4ERR_BADSLOT:
 528                 case -NFS4ERR_BAD_HIGH_SLOT:
 529                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 530                 case -NFS4ERR_DEADSESSION:
 531                 case -NFS4ERR_SEQ_FALSE_RETRY:
 532                 case -NFS4ERR_SEQ_MISORDERED:
 533                         /* Handled in nfs41_sequence_process() */
 534                         goto wait_on_recovery;
 535 #endif /* defined(CONFIG_NFS_V4_1) */
 536                 case -NFS4ERR_FILE_OPEN:
 537                         if (exception->timeout > HZ) {
 538                                 /* We have retried a decent amount, time to
 539                                  * fail
 540                                  */
 541                                 ret = -EBUSY;
 542                                 break;
 543                         }
 544                         fallthrough;
 545                 case -NFS4ERR_DELAY:
 546                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 547                         fallthrough;
 548                 case -NFS4ERR_GRACE:
 549                 case -NFS4ERR_LAYOUTTRYLATER:
 550                 case -NFS4ERR_RECALLCONFLICT:
 551                         exception->delay = 1;
 552                         return 0;
 553
 554                 case -NFS4ERR_RETRY_UNCACHED_REP:
 555                 case -NFS4ERR_OLD_STATEID:
 556                         exception->retry = 1;
 557                         break;
 558                 case -NFS4ERR_BADOWNER:
 559                         /* The following works around a Linux server bug! */
 560                 case -NFS4ERR_BADNAME:
 561                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 562                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 563                                 exception->retry = 1;
 564                                 printk(KERN_WARNING "NFS: v4 server %s "
 565                                                 "does not accept raw "
 566                                                 "uid/gids. "
 567                                                 "Reenabling the idmapper.\n",
 568                                                 server->nfs_client->cl_hostname);
 569                         }
 570         }
 571         /* We failed to handle the error */
 572         return nfs4_map_errors(ret);
 573 wait_on_recovery:
 574         exception->recovering = 1;
 575         return 0;
 576 }
 577
 578 /* This is the error handling routine for processes that are allowed
 579  * to sleep.
 580  */
 581 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 582 {
 583         struct nfs_client *clp = server->nfs_client;
 584         int ret;
 585
 586         ret = nfs4_do_handle_exception(server, errorcode, exception);
 587         if (exception->delay) {
 588                 ret = nfs4_delay(&exception->timeout,
 589                                 exception->interruptible);
 590                 goto out_retry;
 591         }
 592         if (exception->recovering) {
 593                 if (exception->task_is_privileged)
 594                         return -EDEADLOCK;
 595                 ret = nfs4_wait_clnt_recover(clp);
 596                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 597                         return -EIO;
 598                 goto out_retry;
 599         }
 600         return ret;
 601 out_retry:
 602         if (ret == 0)
 603                 exception->retry = 1;
 604         return ret;
 605 }
 606
 607 static int
 608 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 609                 int errorcode, struct nfs4_exception *exception)
 610 {
 611         struct nfs_client *clp = server->nfs_client;
 612         int ret;
 613
 614         ret = nfs4_do_handle_exception(server, errorcode, exception);
 615         if (exception->delay) {
 616                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 617                 goto out_retry;
 618         }
 619         if (exception->recovering) {
 620                 if (exception->task_is_privileged)
 621                         return -EDEADLOCK;
 622                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 623                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 624                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 625                 goto out_retry;
 626         }
 627         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 628                 ret = -EIO;
 629         return ret;
 630 out_retry:
 631         if (ret == 0) {
 632                 exception->retry = 1;
 633                 /*
 634                  * For NFS4ERR_MOVED, the client transport will need to
 635                  * be recomputed after migration recovery has completed.
 636                  */
 637                 if (errorcode == -NFS4ERR_MOVED)
 638                         rpc_task_release_transport(task);
 639         }
 640         return ret;
 641 }
 642
 643 int
 644 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 645                         struct nfs4_state *state, long *timeout)
 646 {
 647         struct nfs4_exception exception = {
 648                 .state = state,
 649         };
 650
 651         if (task->tk_status >= 0)
 652                 return 0;
 653         if (timeout)
 654                 exception.timeout = *timeout;
 655         task->tk_status = nfs4_async_handle_exception(task, server,
 656                         task->tk_status,
 657                         &exception);
 658         if (exception.delay && timeout)
 659                 *timeout = exception.timeout;
 660         if (exception.retry)
 661                 return -EAGAIN;
 662         return 0;
 663 }
 664
 665 /*
 666  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 667  * or 'false' otherwise.
 668  */
 669 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 670 {
 671         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 672         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 673 }
 674
 675 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 676 {
 677         spin_lock(&clp->cl_lock);
 678         if (time_before(clp->cl_last_renewal,timestamp))
 679                 clp->cl_last_renewal = timestamp;
 680         spin_unlock(&clp->cl_lock);
 681 }
 682
 683 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 684 {
 685         struct nfs_client *clp = server->nfs_client;
 686
 687         if (!nfs4_has_session(clp))
 688                 do_renew_lease(clp, timestamp);
 689 }
 690
 691 struct nfs4_call_sync_data {
 692         const struct nfs_server *seq_server;
 693         struct nfs4_sequence_args *seq_args;
 694         struct nfs4_sequence_res *seq_res;
 695 };
 696
 697 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 698                         struct nfs4_sequence_res *res, int cache_reply,
 699                         int privileged)
 700 {
 701         args->sa_slot = NULL;
 702         args->sa_cache_this = cache_reply;
 703         args->sa_privileged = privileged;
 704
 705         res->sr_slot = NULL;
 706 }
 707
 708 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 709 {
 710         struct nfs4_slot *slot = res->sr_slot;
 711         struct nfs4_slot_table *tbl;
 712
 713         tbl = slot->table;
 714         spin_lock(&tbl->slot_tbl_lock);
 715         if (!nfs41_wake_and_assign_slot(tbl, slot))
 716                 nfs4_free_slot(tbl, slot);
 717         spin_unlock(&tbl->slot_tbl_lock);
 718
 719         res->sr_slot = NULL;
 720 }
 721
 722 static int nfs40_sequence_done(struct rpc_task *task,
 723                                struct nfs4_sequence_res *res)
 724 {
 725         if (res->sr_slot != NULL)
 726                 nfs40_sequence_free_slot(res);
 727         return 1;
 728 }
 729
 730 #if defined(CONFIG_NFS_V4_1)
 731
 732 static void nfs41_release_slot(struct nfs4_slot *slot)
 733 {
 734         struct nfs4_session *session;
 735         struct nfs4_slot_table *tbl;
 736         bool send_new_highest_used_slotid = false;
 737
 738         if (!slot)
 739                 return;
 740         tbl = slot->table;
 741         session = tbl->session;
 742
 743         /* Bump the slot sequence number */
 744         if (slot->seq_done)
 745                 slot->seq_nr++;
 746         slot->seq_done = 0;
 747
 748         spin_lock(&tbl->slot_tbl_lock);
 749         /* Be nice to the server: try to ensure that the last transmitted
 750          * value for highest_user_slotid <= target_highest_slotid
 751          */
 752         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 753                 send_new_highest_used_slotid = true;
 754
 755         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 756                 send_new_highest_used_slotid = false;
 757                 goto out_unlock;
 758         }
 759         nfs4_free_slot(tbl, slot);
 760
 761         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 762                 send_new_highest_used_slotid = false;
 763 out_unlock:
 764         spin_unlock(&tbl->slot_tbl_lock);
 765         if (send_new_highest_used_slotid)
 766                 nfs41_notify_server(session->clp);
 767         if (waitqueue_active(&tbl->slot_waitq))
 768                 wake_up_all(&tbl->slot_waitq);
 769 }
 770
 771 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 772 {
 773         nfs41_release_slot(res->sr_slot);
 774         res->sr_slot = NULL;
 775 }
 776
 777 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 778                 u32 seqnr)
 779 {
 780         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 781                 slot->seq_nr_highest_sent = seqnr;
 782 }
 783 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 784                 u32 seqnr)
 785 {
 786         slot->seq_nr_highest_sent = seqnr;
 787         slot->seq_nr_last_acked = seqnr;
 788 }
 789
 790 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 791                                 struct nfs4_slot *slot)
 792 {
 793         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 794         if (!IS_ERR(task))
 795                 rpc_put_task_async(task);
 796 }
 797
 798 static int nfs41_sequence_process(struct rpc_task *task,
 799                 struct nfs4_sequence_res *res)
 800 {
 801         struct nfs4_session *session;
 802         struct nfs4_slot *slot = res->sr_slot;
 803         struct nfs_client *clp;
 804         int status;
 805         int ret = 1;
 806
 807         if (slot == NULL)
 808                 goto out_noaction;
 809         /* don't increment the sequence number if the task wasn't sent */
 810         if (!RPC_WAS_SENT(task) || slot->seq_done)
 811                 goto out;
 812
 813         session = slot->table->session;
 814         clp = session->clp;
 815
 816         trace_nfs4_sequence_done(session, res);
 817
 818         status = res->sr_status;
 819         if (task->tk_status == -NFS4ERR_DEADSESSION)
 820                 status = -NFS4ERR_DEADSESSION;
 821
 822         /* Check the SEQUENCE operation status */
 823         switch (status) {
 824         case 0:
 825                 /* Mark this sequence number as having been acked */
 826                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 827                 /* Update the slot's sequence and clientid lease timer */
 828                 slot->seq_done = 1;
 829                 do_renew_lease(clp, res->sr_timestamp);
 830                 /* Check sequence flags */
 831                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 832                                 !!slot->privileged);
 833                 nfs41_update_target_slotid(slot->table, slot, res);
 834                 break;
 835         case 1:
 836                 /*
 837                  * sr_status remains 1 if an RPC level error occurred.
 838                  * The server may or may not have processed the sequence
 839                  * operation..
 840                  */
 841                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 842                 slot->seq_done = 1;
 843                 goto out;
 844         case -NFS4ERR_DELAY:
 845                 /* The server detected a resend of the RPC call and
 846                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 847                  * of RFC5661.
 848                  */
 849                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 850                         __func__,
 851                         slot->slot_nr,
 852                         slot->seq_nr);
 853                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 854                 goto out_retry;
 855         case -NFS4ERR_RETRY_UNCACHED_REP:
 856         case -NFS4ERR_SEQ_FALSE_RETRY:
 857                 /*
 858                  * The server thinks we tried to replay a request.
 859                  * Retry the call after bumping the sequence ID.
 860                  */
 861                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 862                 goto retry_new_seq;
 863         case -NFS4ERR_BADSLOT:
 864                 /*
 865                  * The slot id we used was probably retired. Try again
 866                  * using a different slot id.
 867                  */
 868                 if (slot->slot_nr < slot->table->target_highest_slotid)
 869                         goto session_recover;
 870                 goto retry_nowait;
 871         case -NFS4ERR_SEQ_MISORDERED:
 872                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 873                 /*
 874                  * Were one or more calls using this slot interrupted?
 875                  * If the server never received the request, then our
 876                  * transmitted slot sequence number may be too high. However,
 877                  * if the server did receive the request then it might
 878                  * accidentally give us a reply with a mismatched operation.
 879                  * We can sort this out by sending a lone sequence operation
 880                  * to the server on the same slot.
 881                  */
 882                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 883                         slot->seq_nr--;
 884                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 885                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 886                                 res->sr_slot = NULL;
 887                         }
 888                         goto retry_nowait;
 889                 }
 890                 /*
 891                  * RFC5661:
 892                  * A retry might be sent while the original request is
 893                  * still in progress on the replier. The replier SHOULD
 894                  * deal with the issue by returning NFS4ERR_DELAY as the
 895                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 896                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 897                  *
 898                  * Restart the search after a delay.
 899                  */
 900                 slot->seq_nr = slot->seq_nr_highest_sent;
 901                 goto out_retry;
 902         case -NFS4ERR_BADSESSION:
 903         case -NFS4ERR_DEADSESSION:
 904         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 905                 goto session_recover;
 906         default:
 907                 /* Just update the slot sequence no. */
 908                 slot->seq_done = 1;
 909         }
 910 out:
 911         /* The session may be reset by one of the error handlers. */
 912         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 913 out_noaction:
 914         return ret;
 915 session_recover:
 916         nfs4_schedule_session_recovery(session, status);
 917         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 918         nfs41_sequence_free_slot(res);
 919         goto out;
 920 retry_new_seq:
 921         ++slot->seq_nr;
 922 retry_nowait:
 923         if (rpc_restart_call_prepare(task)) {
 924                 nfs41_sequence_free_slot(res);
 925                 task->tk_status = 0;
 926                 ret = 0;
 927         }
 928         goto out;
 929 out_retry:
 930         if (!rpc_restart_call(task))
 931                 goto out;
 932         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 933         return 0;
 934 }
 935
 936 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 937 {
 938         if (!nfs41_sequence_process(task, res))
 939                 return 0;
 940         if (res->sr_slot != NULL)
 941                 nfs41_sequence_free_slot(res);
 942         return 1;
 943
 944 }
 945 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 946
 947 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 948 {
 949         if (res->sr_slot == NULL)
 950                 return 1;
 951         if (res->sr_slot->table->session != NULL)
 952                 return nfs41_sequence_process(task, res);
 953         return nfs40_sequence_done(task, res);
 954 }
 955
 956 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 957 {
 958         if (res->sr_slot != NULL) {
 959                 if (res->sr_slot->table->session != NULL)
 960                         nfs41_sequence_free_slot(res);
 961                 else
 962                         nfs40_sequence_free_slot(res);
 963         }
 964 }
 965
 966 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 967 {
 968         if (res->sr_slot == NULL)
 969                 return 1;
 970         if (!res->sr_slot->table->session)
 971                 return nfs40_sequence_done(task, res);
 972         return nfs41_sequence_done(task, res);
 973 }
 974 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 975
 976 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 977 {
 978         struct nfs4_call_sync_data *data = calldata;
 979
 980         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 981
 982         nfs4_setup_sequence(data->seq_server->nfs_client,
 983                             data->seq_args, data->seq_res, task);
 984 }
 985
 986 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 987 {
 988         struct nfs4_call_sync_data *data = calldata;
 989
 990         nfs41_sequence_done(task, data->seq_res);
 991 }
 992
 993 static const struct rpc_call_ops nfs41_call_sync_ops = {
 994         .rpc_call_prepare = nfs41_call_sync_prepare,
 995         .rpc_call_done = nfs41_call_sync_done,
 996 };
 997
 998 #else   /* !CONFIG_NFS_V4_1 */
 999
1000 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1001 {
1002         return nfs40_sequence_done(task, res);
1003 }
1004
1005 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1006 {
1007         if (res->sr_slot != NULL)
1008                 nfs40_sequence_free_slot(res);
1009 }
1010
1011 int nfs4_sequence_done(struct rpc_task *task,
1012                        struct nfs4_sequence_res *res)
1013 {
1014         return nfs40_sequence_done(task, res);
1015 }
1016 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1017
1018 #endif  /* !CONFIG_NFS_V4_1 */
1019
1020 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1021 {
1022         res->sr_timestamp = jiffies;
1023         res->sr_status_flags = 0;
1024         res->sr_status = 1;
1025 }
1026
1027 static
1028 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1029                 struct nfs4_sequence_res *res,
1030                 struct nfs4_slot *slot)
1031 {
1032         if (!slot)
1033                 return;
1034         slot->privileged = args->sa_privileged ? 1 : 0;
1035         args->sa_slot = slot;
1036
1037         res->sr_slot = slot;
1038 }
1039
1040 int nfs4_setup_sequence(struct nfs_client *client,
1041                         struct nfs4_sequence_args *args,
1042                         struct nfs4_sequence_res *res,
1043                         struct rpc_task *task)
1044 {
1045         struct nfs4_session *session = nfs4_get_session(client);
1046         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1047         struct nfs4_slot *slot;
1048
1049         /* slot already allocated? */
1050         if (res->sr_slot != NULL)
1051                 goto out_start;
1052
1053         if (session)
1054                 tbl = &session->fc_slot_table;
1055
1056         spin_lock(&tbl->slot_tbl_lock);
1057         /* The state manager will wait until the slot table is empty */
1058         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1059                 goto out_sleep;
1060
1061         slot = nfs4_alloc_slot(tbl);
1062         if (IS_ERR(slot)) {
1063                 if (slot == ERR_PTR(-ENOMEM))
1064                         goto out_sleep_timeout;
1065                 goto out_sleep;
1066         }
1067         spin_unlock(&tbl->slot_tbl_lock);
1068
1069         nfs4_sequence_attach_slot(args, res, slot);
1070
1071         trace_nfs4_setup_sequence(session, args);
1072 out_start:
1073         nfs41_sequence_res_init(res);
1074         rpc_call_start(task);
1075         return 0;
1076 out_sleep_timeout:
1077         /* Try again in 1/4 second */
1078         if (args->sa_privileged)
1079                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1080                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1081         else
1082                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1083                                 NULL, jiffies + (HZ >> 2));
1084         spin_unlock(&tbl->slot_tbl_lock);
1085         return -EAGAIN;
1086 out_sleep:
1087         if (args->sa_privileged)
1088                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1089                                 RPC_PRIORITY_PRIVILEGED);
1090         else
1091                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1092         spin_unlock(&tbl->slot_tbl_lock);
1093         return -EAGAIN;
1094 }
1095 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1096
1097 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1098 {
1099         struct nfs4_call_sync_data *data = calldata;
1100         nfs4_setup_sequence(data->seq_server->nfs_client,
1101                                 data->seq_args, data->seq_res, task);
1102 }
1103
1104 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1105 {
1106         struct nfs4_call_sync_data *data = calldata;
1107         nfs4_sequence_done(task, data->seq_res);
1108 }
1109
1110 static const struct rpc_call_ops nfs40_call_sync_ops = {
1111         .rpc_call_prepare = nfs40_call_sync_prepare,
1112         .rpc_call_done = nfs40_call_sync_done,
1113 };
1114
1115 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1116 {
1117         int ret;
1118         struct rpc_task *task;
1119
1120         task = rpc_run_task(task_setup);
1121         if (IS_ERR(task))
1122                 return PTR_ERR(task);
1123
1124         ret = task->tk_status;
1125         rpc_put_task(task);
1126         return ret;
1127 }
1128
1129 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1130                              struct nfs_server *server,
1131                              struct rpc_message *msg,
1132                              struct nfs4_sequence_args *args,
1133                              struct nfs4_sequence_res *res,
1134                              unsigned short task_flags)
1135 {
1136         struct nfs_client *clp = server->nfs_client;
1137         struct nfs4_call_sync_data data = {
1138                 .seq_server = server,
1139                 .seq_args = args,
1140                 .seq_res = res,
1141         };
1142         struct rpc_task_setup task_setup = {
1143                 .rpc_client = clnt,
1144                 .rpc_message = msg,
1145                 .callback_ops = clp->cl_mvops->call_sync_ops,
1146                 .callback_data = &data,
1147                 .flags = task_flags,
1148         };
1149
1150         return nfs4_call_sync_custom(&task_setup);
1151 }
1152
1153 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1154                                    struct nfs_server *server,
1155                                    struct rpc_message *msg,
1156                                    struct nfs4_sequence_args *args,
1157                                    struct nfs4_sequence_res *res)
1158 {
1159         unsigned short task_flags = 0;
1160
1161         if (server->nfs_client->cl_minorversion)
1162                 task_flags = RPC_TASK_MOVEABLE;
1163         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1164 }
1165
1166
1167 int nfs4_call_sync(struct rpc_clnt *clnt,
1168                    struct nfs_server *server,
1169                    struct rpc_message *msg,
1170                    struct nfs4_sequence_args *args,
1171                    struct nfs4_sequence_res *res,
1172                    int cache_reply)
1173 {
1174         nfs4_init_sequence(args, res, cache_reply, 0);
1175         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1176 }
1177
1178 static void
1179 nfs4_inc_nlink_locked(struct inode *inode)
1180 {
1181         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1182                                              NFS_INO_INVALID_CTIME |
1183                                              NFS_INO_INVALID_NLINK);
1184         inc_nlink(inode);
1185 }
1186
1187 static void
1188 nfs4_inc_nlink(struct inode *inode)
1189 {
1190         spin_lock(&inode->i_lock);
1191         nfs4_inc_nlink_locked(inode);
1192         spin_unlock(&inode->i_lock);
1193 }
1194
1195 static void
1196 nfs4_dec_nlink_locked(struct inode *inode)
1197 {
1198         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1199                                              NFS_INO_INVALID_CTIME |
1200                                              NFS_INO_INVALID_NLINK);
1201         drop_nlink(inode);
1202 }
1203
1204 static void
1205 nfs4_update_changeattr_locked(struct inode *inode,
1206                 struct nfs4_change_info *cinfo,
1207                 unsigned long timestamp, unsigned long cache_validity)
1208 {
1209         struct nfs_inode *nfsi = NFS_I(inode);
1210         u64 change_attr = inode_peek_iversion_raw(inode);
1211
1212         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1213         if (S_ISDIR(inode->i_mode))
1214                 cache_validity |= NFS_INO_INVALID_DATA;
1215
1216         switch (NFS_SERVER(inode)->change_attr_type) {
1217         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1218                 if (cinfo->after == change_attr)
1219                         goto out;
1220                 break;
1221         default:
1222                 if ((s64)(change_attr - cinfo->after) >= 0)
1223                         goto out;
1224         }
1225
1226         inode_set_iversion_raw(inode, cinfo->after);
1227         if (!cinfo->atomic || cinfo->before != change_attr) {
1228                 if (S_ISDIR(inode->i_mode))
1229                         nfs_force_lookup_revalidate(inode);
1230
1231                 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1232                         cache_validity |=
1233                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1234                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1235                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1236                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR |
1237                                 NFS_INO_REVAL_PAGECACHE;
1238                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1239         }
1240         nfsi->attrtimeo_timestamp = jiffies;
1241         nfsi->read_cache_jiffies = timestamp;
1242         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1243         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1244 out:
1245         nfs_set_cache_invalid(inode, cache_validity);
1246 }
1247
1248 void
1249 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1250                 unsigned long timestamp, unsigned long cache_validity)
1251 {
1252         spin_lock(&dir->i_lock);
1253         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1254         spin_unlock(&dir->i_lock);
1255 }
1256
1257 struct nfs4_open_createattrs {
1258         struct nfs4_label *label;
1259         struct iattr *sattr;
1260         const __u32 verf[2];
1261 };
1262
1263 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1264                 int err, struct nfs4_exception *exception)
1265 {
1266         if (err != -EINVAL)
1267                 return false;
1268         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1269                 return false;
1270         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1271         exception->retry = 1;
1272         return true;
1273 }
1274
1275 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1276 {
1277          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1278 }
1279
1280 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1281 {
1282         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1283
1284         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1285 }
1286
1287 static u32
1288 nfs4_map_atomic_open_share(struct nfs_server *server,
1289                 fmode_t fmode, int openflags)
1290 {
1291         u32 res = 0;
1292
1293         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1294         case FMODE_READ:
1295                 res = NFS4_SHARE_ACCESS_READ;
1296                 break;
1297         case FMODE_WRITE:
1298                 res = NFS4_SHARE_ACCESS_WRITE;
1299                 break;
1300         case FMODE_READ|FMODE_WRITE:
1301                 res = NFS4_SHARE_ACCESS_BOTH;
1302         }
1303         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1304                 goto out;
1305         /* Want no delegation if we're using O_DIRECT */
1306         if (openflags & O_DIRECT)
1307                 res |= NFS4_SHARE_WANT_NO_DELEG;
1308 out:
1309         return res;
1310 }
1311
1312 static enum open_claim_type4
1313 nfs4_map_atomic_open_claim(struct nfs_server *server,
1314                 enum open_claim_type4 claim)
1315 {
1316         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1317                 return claim;
1318         switch (claim) {
1319         default:
1320                 return claim;
1321         case NFS4_OPEN_CLAIM_FH:
1322                 return NFS4_OPEN_CLAIM_NULL;
1323         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1324                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1325         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1326                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1327         }
1328 }
1329
1330 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1331 {
1332         p->o_res.f_attr = &p->f_attr;
1333         p->o_res.seqid = p->o_arg.seqid;
1334         p->c_res.seqid = p->c_arg.seqid;
1335         p->o_res.server = p->o_arg.server;
1336         p->o_res.access_request = p->o_arg.access;
1337         nfs_fattr_init(&p->f_attr);
1338         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1339 }
1340
1341 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1342                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1343                 const struct nfs4_open_createattrs *c,
1344                 enum open_claim_type4 claim,
1345                 gfp_t gfp_mask)
1346 {
1347         struct dentry *parent = dget_parent(dentry);
1348         struct inode *dir = d_inode(parent);
1349         struct nfs_server *server = NFS_SERVER(dir);
1350         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1351         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1352         struct nfs4_opendata *p;
1353
1354         p = kzalloc(sizeof(*p), gfp_mask);
1355         if (p == NULL)
1356                 goto err;
1357
1358         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1359         if (IS_ERR(p->f_attr.label))
1360                 goto err_free_p;
1361
1362         p->a_label = nfs4_label_alloc(server, gfp_mask);
1363         if (IS_ERR(p->a_label))
1364                 goto err_free_f;
1365
1366         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1367         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1368         if (IS_ERR(p->o_arg.seqid))
1369                 goto err_free_label;
1370         nfs_sb_active(dentry->d_sb);
1371         p->dentry = dget(dentry);
1372         p->dir = parent;
1373         p->owner = sp;
1374         atomic_inc(&sp->so_count);
1375         p->o_arg.open_flags = flags;
1376         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1377         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1378         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1379                         fmode, flags);
1380         if (flags & O_CREAT) {
1381                 p->o_arg.umask = current_umask();
1382                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1383                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1384                         p->o_arg.u.attrs = &p->attrs;
1385                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1386
1387                         memcpy(p->o_arg.u.verifier.data, c->verf,
1388                                         sizeof(p->o_arg.u.verifier.data));
1389                 }
1390         }
1391         /* ask server to check for all possible rights as results
1392          * are cached */
1393         switch (p->o_arg.claim) {
1394         default:
1395                 break;
1396         case NFS4_OPEN_CLAIM_NULL:
1397         case NFS4_OPEN_CLAIM_FH:
1398                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1399                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1400                                   NFS4_ACCESS_EXECUTE;
1401 #ifdef CONFIG_NFS_V4_2
1402                 if (!(server->caps & NFS_CAP_XATTR))
1403                         break;
1404                 p->o_arg.access |= NFS4_ACCESS_XAREAD | NFS4_ACCESS_XAWRITE |
1405                                    NFS4_ACCESS_XALIST;
1406 #endif
1407         }
1408         p->o_arg.clientid = server->nfs_client->cl_clientid;
1409         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1410         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1411         p->o_arg.name = &dentry->d_name;
1412         p->o_arg.server = server;
1413         p->o_arg.bitmask = nfs4_bitmask(server, label);
1414         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1415         switch (p->o_arg.claim) {
1416         case NFS4_OPEN_CLAIM_NULL:
1417         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1418         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1419                 p->o_arg.fh = NFS_FH(dir);
1420                 break;
1421         case NFS4_OPEN_CLAIM_PREVIOUS:
1422         case NFS4_OPEN_CLAIM_FH:
1423         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1424         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1425                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1426         }
1427         p->c_arg.fh = &p->o_res.fh;
1428         p->c_arg.stateid = &p->o_res.stateid;
1429         p->c_arg.seqid = p->o_arg.seqid;
1430         nfs4_init_opendata_res(p);
1431         kref_init(&p->kref);
1432         return p;
1433
1434 err_free_label:
1435         nfs4_label_free(p->a_label);
1436 err_free_f:
1437         nfs4_label_free(p->f_attr.label);
1438 err_free_p:
1439         kfree(p);
1440 err:
1441         dput(parent);
1442         return NULL;
1443 }
1444
1445 static void nfs4_opendata_free(struct kref *kref)
1446 {
1447         struct nfs4_opendata *p = container_of(kref,
1448                         struct nfs4_opendata, kref);
1449         struct super_block *sb = p->dentry->d_sb;
1450
1451         nfs4_lgopen_release(p->lgp);
1452         nfs_free_seqid(p->o_arg.seqid);
1453         nfs4_sequence_free_slot(&p->o_res.seq_res);
1454         if (p->state != NULL)
1455                 nfs4_put_open_state(p->state);
1456         nfs4_put_state_owner(p->owner);
1457
1458         nfs4_label_free(p->a_label);
1459         nfs4_label_free(p->f_attr.label);
1460
1461         dput(p->dir);
1462         dput(p->dentry);
1463         nfs_sb_deactive(sb);
1464         nfs_fattr_free_names(&p->f_attr);
1465         kfree(p->f_attr.mdsthreshold);
1466         kfree(p);
1467 }
1468
1469 static void nfs4_opendata_put(struct nfs4_opendata *p)
1470 {
1471         if (p != NULL)
1472                 kref_put(&p->kref, nfs4_opendata_free);
1473 }
1474
1475 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1476                 fmode_t fmode)
1477 {
1478         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1479         case FMODE_READ|FMODE_WRITE:
1480                 return state->n_rdwr != 0;
1481         case FMODE_WRITE:
1482                 return state->n_wronly != 0;
1483         case FMODE_READ:
1484                 return state->n_rdonly != 0;
1485         }
1486         WARN_ON_ONCE(1);
1487         return false;
1488 }
1489
1490 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1491                 int open_mode, enum open_claim_type4 claim)
1492 {
1493         int ret = 0;
1494
1495         if (open_mode & (O_EXCL|O_TRUNC))
1496                 goto out;
1497         switch (claim) {
1498         case NFS4_OPEN_CLAIM_NULL:
1499         case NFS4_OPEN_CLAIM_FH:
1500                 goto out;
1501         default:
1502                 break;
1503         }
1504         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1505                 case FMODE_READ:
1506                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1507                                 && state->n_rdonly != 0;
1508                         break;
1509                 case FMODE_WRITE:
1510                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1511                                 && state->n_wronly != 0;
1512                         break;
1513                 case FMODE_READ|FMODE_WRITE:
1514                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1515                                 && state->n_rdwr != 0;
1516         }
1517 out:
1518         return ret;
1519 }
1520
1521 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1522                 enum open_claim_type4 claim)
1523 {
1524         if (delegation == NULL)
1525                 return 0;
1526         if ((delegation->type & fmode) != fmode)
1527                 return 0;
1528         switch (claim) {
1529         case NFS4_OPEN_CLAIM_NULL:
1530         case NFS4_OPEN_CLAIM_FH:
1531                 break;
1532         case NFS4_OPEN_CLAIM_PREVIOUS:
1533                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1534                         break;
1535                 fallthrough;
1536         default:
1537                 return 0;
1538         }
1539         nfs_mark_delegation_referenced(delegation);
1540         return 1;
1541 }
1542
1543 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1544 {
1545         switch (fmode) {
1546                 case FMODE_WRITE:
1547                         state->n_wronly++;
1548                         break;
1549                 case FMODE_READ:
1550                         state->n_rdonly++;
1551                         break;
1552                 case FMODE_READ|FMODE_WRITE:
1553                         state->n_rdwr++;
1554         }
1555         nfs4_state_set_mode_locked(state, state->state | fmode);
1556 }
1557
1558 #ifdef CONFIG_NFS_V4_1
1559 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1560 {
1561         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1562                 return true;
1563         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1564                 return true;
1565         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1566                 return true;
1567         return false;
1568 }
1569 #endif /* CONFIG_NFS_V4_1 */
1570
1571 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1572 {
1573         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1574                 wake_up_all(&state->waitq);
1575 }
1576
1577 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1578 {
1579         struct nfs_client *clp = state->owner->so_server->nfs_client;
1580         bool need_recover = false;
1581
1582         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1583                 need_recover = true;
1584         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1585                 need_recover = true;
1586         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1587                 need_recover = true;
1588         if (need_recover)
1589                 nfs4_state_mark_reclaim_nograce(clp, state);
1590 }
1591
1592 /*
1593  * Check for whether or not the caller may update the open stateid
1594  * to the value passed in by stateid.
1595  *
1596  * Note: This function relies heavily on the server implementing
1597  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1598  * correctly.
1599  * i.e. The stateid seqids have to be initialised to 1, and
1600  * are then incremented on every state transition.
1601  */
1602 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1603                 const nfs4_stateid *stateid)
1604 {
1605         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1606                 /* The common case - we're updating to a new sequence number */
1607                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1608                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1609                                 return true;
1610                         return false;
1611                 }
1612                 /* The server returned a new stateid */
1613         }
1614         /* This is the first OPEN in this generation */
1615         if (stateid->seqid == cpu_to_be32(1))
1616                 return true;
1617         return false;
1618 }
1619
1620 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1621 {
1622         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1623                 return;
1624         if (state->n_wronly)
1625                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1626         if (state->n_rdonly)
1627                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1628         if (state->n_rdwr)
1629                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1630         set_bit(NFS_OPEN_STATE, &state->flags);
1631 }
1632
1633 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1634                 nfs4_stateid *stateid, fmode_t fmode)
1635 {
1636         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1637         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1638         case FMODE_WRITE:
1639                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1640                 break;
1641         case FMODE_READ:
1642                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1643                 break;
1644         case 0:
1645                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1646                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1647                 clear_bit(NFS_OPEN_STATE, &state->flags);
1648         }
1649         if (stateid == NULL)
1650                 return;
1651         /* Handle OPEN+OPEN_DOWNGRADE races */
1652         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1653             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1654                 nfs_resync_open_stateid_locked(state);
1655                 goto out;
1656         }
1657         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1658                 nfs4_stateid_copy(&state->stateid, stateid);
1659         nfs4_stateid_copy(&state->open_stateid, stateid);
1660         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1661 out:
1662         nfs_state_log_update_open_stateid(state);
1663 }
1664
1665 static void nfs_clear_open_stateid(struct nfs4_state *state,
1666         nfs4_stateid *arg_stateid,
1667         nfs4_stateid *stateid, fmode_t fmode)
1668 {
1669         write_seqlock(&state->seqlock);
1670         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1671         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1672                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1673         write_sequnlock(&state->seqlock);
1674         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1675                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1676 }
1677
1678 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1679                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1680         __must_hold(&state->owner->so_lock)
1681         __must_hold(&state->seqlock)
1682         __must_hold(RCU)
1683
1684 {
1685         DEFINE_WAIT(wait);
1686         int status = 0;
1687         for (;;) {
1688
1689                 if (nfs_stateid_is_sequential(state, stateid))
1690                         break;
1691
1692                 if (status)
1693                         break;
1694                 /* Rely on seqids for serialisation with NFSv4.0 */
1695                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1696                         break;
1697
1698                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1699                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1700                 /*
1701                  * Ensure we process the state changes in the same order
1702                  * in which the server processed them by delaying the
1703                  * update of the stateid until we are in sequence.
1704                  */
1705                 write_sequnlock(&state->seqlock);
1706                 spin_unlock(&state->owner->so_lock);
1707                 rcu_read_unlock();
1708                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1709
1710                 if (!fatal_signal_pending(current)) {
1711                         if (schedule_timeout(5*HZ) == 0)
1712                                 status = -EAGAIN;
1713                         else
1714                                 status = 0;
1715                 } else
1716                         status = -EINTR;
1717                 finish_wait(&state->waitq, &wait);
1718                 rcu_read_lock();
1719                 spin_lock(&state->owner->so_lock);
1720                 write_seqlock(&state->seqlock);
1721         }
1722
1723         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1724             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1725                 nfs4_stateid_copy(freeme, &state->open_stateid);
1726                 nfs_test_and_clear_all_open_stateid(state);
1727         }
1728
1729         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1730                 nfs4_stateid_copy(&state->stateid, stateid);
1731         nfs4_stateid_copy(&state->open_stateid, stateid);
1732         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1733         nfs_state_log_update_open_stateid(state);
1734 }
1735
1736 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1737                 const nfs4_stateid *open_stateid,
1738                 fmode_t fmode,
1739                 nfs4_stateid *freeme)
1740 {
1741         /*
1742          * Protect the call to nfs4_state_set_mode_locked and
1743          * serialise the stateid update
1744          */
1745         write_seqlock(&state->seqlock);
1746         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1747         switch (fmode) {
1748         case FMODE_READ:
1749                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1750                 break;
1751         case FMODE_WRITE:
1752                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1753                 break;
1754         case FMODE_READ|FMODE_WRITE:
1755                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1756         }
1757         set_bit(NFS_OPEN_STATE, &state->flags);
1758         write_sequnlock(&state->seqlock);
1759 }
1760
1761 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1762 {
1763         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1764         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1765         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1766         clear_bit(NFS_OPEN_STATE, &state->flags);
1767 }
1768
1769 static void nfs_state_set_delegation(struct nfs4_state *state,
1770                 const nfs4_stateid *deleg_stateid,
1771                 fmode_t fmode)
1772 {
1773         /*
1774          * Protect the call to nfs4_state_set_mode_locked and
1775          * serialise the stateid update
1776          */
1777         write_seqlock(&state->seqlock);
1778         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1779         set_bit(NFS_DELEGATED_STATE, &state->flags);
1780         write_sequnlock(&state->seqlock);
1781 }
1782
1783 static void nfs_state_clear_delegation(struct nfs4_state *state)
1784 {
1785         write_seqlock(&state->seqlock);
1786         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1787         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1788         write_sequnlock(&state->seqlock);
1789 }
1790
1791 int update_open_stateid(struct nfs4_state *state,
1792                 const nfs4_stateid *open_stateid,
1793                 const nfs4_stateid *delegation,
1794                 fmode_t fmode)
1795 {
1796         struct nfs_server *server = NFS_SERVER(state->inode);
1797         struct nfs_client *clp = server->nfs_client;
1798         struct nfs_inode *nfsi = NFS_I(state->inode);
1799         struct nfs_delegation *deleg_cur;
1800         nfs4_stateid freeme = { };
1801         int ret = 0;
1802
1803         fmode &= (FMODE_READ|FMODE_WRITE);
1804
1805         rcu_read_lock();
1806         spin_lock(&state->owner->so_lock);
1807         if (open_stateid != NULL) {
1808                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1809                 ret = 1;
1810         }
1811
1812         deleg_cur = nfs4_get_valid_delegation(state->inode);
1813         if (deleg_cur == NULL)
1814                 goto no_delegation;
1815
1816         spin_lock(&deleg_cur->lock);
1817         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1818            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1819             (deleg_cur->type & fmode) != fmode)
1820                 goto no_delegation_unlock;
1821
1822         if (delegation == NULL)
1823                 delegation = &deleg_cur->stateid;
1824         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1825                 goto no_delegation_unlock;
1826
1827         nfs_mark_delegation_referenced(deleg_cur);
1828         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1829         ret = 1;
1830 no_delegation_unlock:
1831         spin_unlock(&deleg_cur->lock);
1832 no_delegation:
1833         if (ret)
1834                 update_open_stateflags(state, fmode);
1835         spin_unlock(&state->owner->so_lock);
1836         rcu_read_unlock();
1837
1838         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1839                 nfs4_schedule_state_manager(clp);
1840         if (freeme.type != 0)
1841                 nfs4_test_and_free_stateid(server, &freeme,
1842                                 state->owner->so_cred);
1843
1844         return ret;
1845 }
1846
1847 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1848                 const nfs4_stateid *stateid)
1849 {
1850         struct nfs4_state *state = lsp->ls_state;
1851         bool ret = false;
1852
1853         spin_lock(&state->state_lock);
1854         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1855                 goto out_noupdate;
1856         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1857                 goto out_noupdate;
1858         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1859         ret = true;
1860 out_noupdate:
1861         spin_unlock(&state->state_lock);
1862         return ret;
1863 }
1864
1865 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1866 {
1867         struct nfs_delegation *delegation;
1868
1869         fmode &= FMODE_READ|FMODE_WRITE;
1870         rcu_read_lock();
1871         delegation = nfs4_get_valid_delegation(inode);
1872         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1873                 rcu_read_unlock();
1874                 return;
1875         }
1876         rcu_read_unlock();
1877         nfs4_inode_return_delegation(inode);
1878 }
1879
1880 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1881 {
1882         struct nfs4_state *state = opendata->state;
1883         struct nfs_delegation *delegation;
1884         int open_mode = opendata->o_arg.open_flags;
1885         fmode_t fmode = opendata->o_arg.fmode;
1886         enum open_claim_type4 claim = opendata->o_arg.claim;
1887         nfs4_stateid stateid;
1888         int ret = -EAGAIN;
1889
1890         for (;;) {
1891                 spin_lock(&state->owner->so_lock);
1892                 if (can_open_cached(state, fmode, open_mode, claim)) {
1893                         update_open_stateflags(state, fmode);
1894                         spin_unlock(&state->owner->so_lock);
1895                         goto out_return_state;
1896                 }
1897                 spin_unlock(&state->owner->so_lock);
1898                 rcu_read_lock();
1899                 delegation = nfs4_get_valid_delegation(state->inode);
1900                 if (!can_open_delegated(delegation, fmode, claim)) {
1901                         rcu_read_unlock();
1902                         break;
1903                 }
1904                 /* Save the delegation */
1905                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1906                 rcu_read_unlock();
1907                 nfs_release_seqid(opendata->o_arg.seqid);
1908                 if (!opendata->is_recover) {
1909                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1910                         if (ret != 0)
1911                                 goto out;
1912                 }
1913                 ret = -EAGAIN;
1914
1915                 /* Try to update the stateid using the delegation */
1916                 if (update_open_stateid(state, NULL, &stateid, fmode))
1917                         goto out_return_state;
1918         }
1919 out:
1920         return ERR_PTR(ret);
1921 out_return_state:
1922         refcount_inc(&state->count);
1923         return state;
1924 }
1925
1926 static void
1927 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1928 {
1929         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1930         struct nfs_delegation *delegation;
1931         int delegation_flags = 0;
1932
1933         rcu_read_lock();
1934         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1935         if (delegation)
1936                 delegation_flags = delegation->flags;
1937         rcu_read_unlock();
1938         switch (data->o_arg.claim) {
1939         default:
1940                 break;
1941         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1942         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1943                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1944                                    "returning a delegation for "
1945                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1946                                    clp->cl_hostname);
1947                 return;
1948         }
1949         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1950                 nfs_inode_set_delegation(state->inode,
1951                                 data->owner->so_cred,
1952                                 data->o_res.delegation_type,
1953                                 &data->o_res.delegation,
1954                                 data->o_res.pagemod_limit);
1955         else
1956                 nfs_inode_reclaim_delegation(state->inode,
1957                                 data->owner->so_cred,
1958                                 data->o_res.delegation_type,
1959                                 &data->o_res.delegation,
1960                                 data->o_res.pagemod_limit);
1961
1962         if (data->o_res.do_recall)
1963                 nfs_async_inode_return_delegation(state->inode,
1964                                                   &data->o_res.delegation);
1965 }
1966
1967 /*
1968  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1969  * and update the nfs4_state.
1970  */
1971 static struct nfs4_state *
1972 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1973 {
1974         struct inode *inode = data->state->inode;
1975         struct nfs4_state *state = data->state;
1976         int ret;
1977
1978         if (!data->rpc_done) {
1979                 if (data->rpc_status)
1980                         return ERR_PTR(data->rpc_status);
1981                 /* cached opens have already been processed */
1982                 goto update;
1983         }
1984
1985         ret = nfs_refresh_inode(inode, &data->f_attr);
1986         if (ret)
1987                 return ERR_PTR(ret);
1988
1989         if (data->o_res.delegation_type != 0)
1990                 nfs4_opendata_check_deleg(data, state);
1991 update:
1992         if (!update_open_stateid(state, &data->o_res.stateid,
1993                                 NULL, data->o_arg.fmode))
1994                 return ERR_PTR(-EAGAIN);
1995         refcount_inc(&state->count);
1996
1997         return state;
1998 }
1999
2000 static struct inode *
2001 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2002 {
2003         struct inode *inode;
2004
2005         switch (data->o_arg.claim) {
2006         case NFS4_OPEN_CLAIM_NULL:
2007         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2008         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2009                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2010                         return ERR_PTR(-EAGAIN);
2011                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2012                                 &data->f_attr);
2013                 break;
2014         default:
2015                 inode = d_inode(data->dentry);
2016                 ihold(inode);
2017                 nfs_refresh_inode(inode, &data->f_attr);
2018         }
2019         return inode;
2020 }
2021
2022 static struct nfs4_state *
2023 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2024 {
2025         struct nfs4_state *state;
2026         struct inode *inode;
2027
2028         inode = nfs4_opendata_get_inode(data);
2029         if (IS_ERR(inode))
2030                 return ERR_CAST(inode);
2031         if (data->state != NULL && data->state->inode == inode) {
2032                 state = data->state;
2033                 refcount_inc(&state->count);
2034         } else
2035                 state = nfs4_get_open_state(inode, data->owner);
2036         iput(inode);
2037         if (state == NULL)
2038                 state = ERR_PTR(-ENOMEM);
2039         return state;
2040 }
2041
2042 static struct nfs4_state *
2043 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2044 {
2045         struct nfs4_state *state;
2046
2047         if (!data->rpc_done) {
2048                 state = nfs4_try_open_cached(data);
2049                 trace_nfs4_cached_open(data->state);
2050                 goto out;
2051         }
2052
2053         state = nfs4_opendata_find_nfs4_state(data);
2054         if (IS_ERR(state))
2055                 goto out;
2056
2057         if (data->o_res.delegation_type != 0)
2058                 nfs4_opendata_check_deleg(data, state);
2059         if (!update_open_stateid(state, &data->o_res.stateid,
2060                                 NULL, data->o_arg.fmode)) {
2061                 nfs4_put_open_state(state);
2062                 state = ERR_PTR(-EAGAIN);
2063         }
2064 out:
2065         nfs_release_seqid(data->o_arg.seqid);
2066         return state;
2067 }
2068
2069 static struct nfs4_state *
2070 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2071 {
2072         struct nfs4_state *ret;
2073
2074         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2075                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2076         else
2077                 ret = _nfs4_opendata_to_nfs4_state(data);
2078         nfs4_sequence_free_slot(&data->o_res.seq_res);
2079         return ret;
2080 }
2081
2082 static struct nfs_open_context *
2083 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2084 {
2085         struct nfs_inode *nfsi = NFS_I(state->inode);
2086         struct nfs_open_context *ctx;
2087
2088         rcu_read_lock();
2089         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2090                 if (ctx->state != state)
2091                         continue;
2092                 if ((ctx->mode & mode) != mode)
2093                         continue;
2094                 if (!get_nfs_open_context(ctx))
2095                         continue;
2096                 rcu_read_unlock();
2097                 return ctx;
2098         }
2099         rcu_read_unlock();
2100         return ERR_PTR(-ENOENT);
2101 }
2102
2103 static struct nfs_open_context *
2104 nfs4_state_find_open_context(struct nfs4_state *state)
2105 {
2106         struct nfs_open_context *ctx;
2107
2108         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2109         if (!IS_ERR(ctx))
2110                 return ctx;
2111         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2112         if (!IS_ERR(ctx))
2113                 return ctx;
2114         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2115 }
2116
2117 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2118                 struct nfs4_state *state, enum open_claim_type4 claim)
2119 {
2120         struct nfs4_opendata *opendata;
2121
2122         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2123                         NULL, claim, GFP_NOFS);
2124         if (opendata == NULL)
2125                 return ERR_PTR(-ENOMEM);
2126         opendata->state = state;
2127         refcount_inc(&state->count);
2128         return opendata;
2129 }
2130
2131 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2132                 fmode_t fmode)
2133 {
2134         struct nfs4_state *newstate;
2135         int ret;
2136
2137         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2138                 return 0;
2139         opendata->o_arg.open_flags = 0;
2140         opendata->o_arg.fmode = fmode;
2141         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2142                         NFS_SB(opendata->dentry->d_sb),
2143                         fmode, 0);
2144         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2145         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2146         nfs4_init_opendata_res(opendata);
2147         ret = _nfs4_recover_proc_open(opendata);
2148         if (ret != 0)
2149                 return ret;
2150         newstate = nfs4_opendata_to_nfs4_state(opendata);
2151         if (IS_ERR(newstate))
2152                 return PTR_ERR(newstate);
2153         if (newstate != opendata->state)
2154                 ret = -ESTALE;
2155         nfs4_close_state(newstate, fmode);
2156         return ret;
2157 }
2158
2159 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2160 {
2161         int ret;
2162
2163         /* memory barrier prior to reading state->n_* */
2164         smp_rmb();
2165         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2166         if (ret != 0)
2167                 return ret;
2168         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2169         if (ret != 0)
2170                 return ret;
2171         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2172         if (ret != 0)
2173                 return ret;
2174         /*
2175          * We may have performed cached opens for all three recoveries.
2176          * Check if we need to update the current stateid.
2177          */
2178         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2179             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2180                 write_seqlock(&state->seqlock);
2181                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2182                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2183                 write_sequnlock(&state->seqlock);
2184         }
2185         return 0;
2186 }
2187
2188 /*
2189  * OPEN_RECLAIM:
2190  *      reclaim state on the server after a reboot.
2191  */
2192 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2193 {
2194         struct nfs_delegation *delegation;
2195         struct nfs4_opendata *opendata;
2196         fmode_t delegation_type = 0;
2197         int status;
2198
2199         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2200                         NFS4_OPEN_CLAIM_PREVIOUS);
2201         if (IS_ERR(opendata))
2202                 return PTR_ERR(opendata);
2203         rcu_read_lock();
2204         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2205         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2206                 delegation_type = delegation->type;
2207         rcu_read_unlock();
2208         opendata->o_arg.u.delegation_type = delegation_type;
2209         status = nfs4_open_recover(opendata, state);
2210         nfs4_opendata_put(opendata);
2211         return status;
2212 }
2213
2214 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2215 {
2216         struct nfs_server *server = NFS_SERVER(state->inode);
2217         struct nfs4_exception exception = { };
2218         int err;
2219         do {
2220                 err = _nfs4_do_open_reclaim(ctx, state);
2221                 trace_nfs4_open_reclaim(ctx, 0, err);
2222                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2223                         continue;
2224                 if (err != -NFS4ERR_DELAY)
2225                         break;
2226                 nfs4_handle_exception(server, err, &exception);
2227         } while (exception.retry);
2228         return err;
2229 }
2230
2231 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2232 {
2233         struct nfs_open_context *ctx;
2234         int ret;
2235
2236         ctx = nfs4_state_find_open_context(state);
2237         if (IS_ERR(ctx))
2238                 return -EAGAIN;
2239         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2240         nfs_state_clear_open_state_flags(state);
2241         ret = nfs4_do_open_reclaim(ctx, state);
2242         put_nfs_open_context(ctx);
2243         return ret;
2244 }
2245
2246 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2247 {
2248         switch (err) {
2249                 default:
2250                         printk(KERN_ERR "NFS: %s: unhandled error "
2251                                         "%d.\n", __func__, err);
2252                         fallthrough;
2253                 case 0:
2254                 case -ENOENT:
2255                 case -EAGAIN:
2256                 case -ESTALE:
2257                 case -ETIMEDOUT:
2258                         break;
2259                 case -NFS4ERR_BADSESSION:
2260                 case -NFS4ERR_BADSLOT:
2261                 case -NFS4ERR_BAD_HIGH_SLOT:
2262                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2263                 case -NFS4ERR_DEADSESSION:
2264                         return -EAGAIN;
2265                 case -NFS4ERR_STALE_CLIENTID:
2266                 case -NFS4ERR_STALE_STATEID:
2267                         /* Don't recall a delegation if it was lost */
2268                         nfs4_schedule_lease_recovery(server->nfs_client);
2269                         return -EAGAIN;
2270                 case -NFS4ERR_MOVED:
2271                         nfs4_schedule_migration_recovery(server);
2272                         return -EAGAIN;
2273                 case -NFS4ERR_LEASE_MOVED:
2274                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2275                         return -EAGAIN;
2276                 case -NFS4ERR_DELEG_REVOKED:
2277                 case -NFS4ERR_ADMIN_REVOKED:
2278                 case -NFS4ERR_EXPIRED:
2279                 case -NFS4ERR_BAD_STATEID:
2280                 case -NFS4ERR_OPENMODE:
2281                         nfs_inode_find_state_and_recover(state->inode,
2282                                         stateid);
2283                         nfs4_schedule_stateid_recovery(server, state);
2284                         return -EAGAIN;
2285                 case -NFS4ERR_DELAY:
2286                 case -NFS4ERR_GRACE:
2287                         ssleep(1);
2288                         return -EAGAIN;
2289                 case -ENOMEM:
2290                 case -NFS4ERR_DENIED:
2291                         if (fl) {
2292                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2293                                 if (lsp)
2294                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2295                         }
2296                         return 0;
2297         }
2298         return err;
2299 }
2300
2301 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2302                 struct nfs4_state *state, const nfs4_stateid *stateid)
2303 {
2304         struct nfs_server *server = NFS_SERVER(state->inode);
2305         struct nfs4_opendata *opendata;
2306         int err = 0;
2307
2308         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2309                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2310         if (IS_ERR(opendata))
2311                 return PTR_ERR(opendata);
2312         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2313         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2314                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2315                 if (err)
2316                         goto out;
2317         }
2318         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2319                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2320                 if (err)
2321                         goto out;
2322         }
2323         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2324                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2325                 if (err)
2326                         goto out;
2327         }
2328         nfs_state_clear_delegation(state);
2329 out:
2330         nfs4_opendata_put(opendata);
2331         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2332 }
2333
2334 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2335 {
2336         struct nfs4_opendata *data = calldata;
2337
2338         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2339                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2340 }
2341
2342 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2343 {
2344         struct nfs4_opendata *data = calldata;
2345
2346         nfs40_sequence_done(task, &data->c_res.seq_res);
2347
2348         data->rpc_status = task->tk_status;
2349         if (data->rpc_status == 0) {
2350                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2351                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2352                 renew_lease(data->o_res.server, data->timestamp);
2353                 data->rpc_done = true;
2354         }
2355 }
2356
2357 static void nfs4_open_confirm_release(void *calldata)
2358 {
2359         struct nfs4_opendata *data = calldata;
2360         struct nfs4_state *state = NULL;
2361
2362         /* If this request hasn't been cancelled, do nothing */
2363         if (!data->cancelled)
2364                 goto out_free;
2365         /* In case of error, no cleanup! */
2366         if (!data->rpc_done)
2367                 goto out_free;
2368         state = nfs4_opendata_to_nfs4_state(data);
2369         if (!IS_ERR(state))
2370                 nfs4_close_state(state, data->o_arg.fmode);
2371 out_free:
2372         nfs4_opendata_put(data);
2373 }
2374
2375 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2376         .rpc_call_prepare = nfs4_open_confirm_prepare,
2377         .rpc_call_done = nfs4_open_confirm_done,
2378         .rpc_release = nfs4_open_confirm_release,
2379 };
2380
2381 /*
2382  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2383  */
2384 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2385 {
2386         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2387         struct rpc_task *task;
2388         struct  rpc_message msg = {
2389                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2390                 .rpc_argp = &data->c_arg,
2391                 .rpc_resp = &data->c_res,
2392                 .rpc_cred = data->owner->so_cred,
2393         };
2394         struct rpc_task_setup task_setup_data = {
2395                 .rpc_client = server->client,
2396                 .rpc_message = &msg,
2397                 .callback_ops = &nfs4_open_confirm_ops,
2398                 .callback_data = data,
2399                 .workqueue = nfsiod_workqueue,
2400                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2401         };
2402         int status;
2403
2404         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2405                                 data->is_recover);
2406         kref_get(&data->kref);
2407         data->rpc_done = false;
2408         data->rpc_status = 0;
2409         data->timestamp = jiffies;
2410         task = rpc_run_task(&task_setup_data);
2411         if (IS_ERR(task))
2412                 return PTR_ERR(task);
2413         status = rpc_wait_for_completion_task(task);
2414         if (status != 0) {
2415                 data->cancelled = true;
2416                 smp_wmb();
2417         } else
2418                 status = data->rpc_status;
2419         rpc_put_task(task);
2420         return status;
2421 }
2422
2423 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2424 {
2425         struct nfs4_opendata *data = calldata;
2426         struct nfs4_state_owner *sp = data->owner;
2427         struct nfs_client *clp = sp->so_server->nfs_client;
2428         enum open_claim_type4 claim = data->o_arg.claim;
2429
2430         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2431                 goto out_wait;
2432         /*
2433          * Check if we still need to send an OPEN call, or if we can use
2434          * a delegation instead.
2435          */
2436         if (data->state != NULL) {
2437                 struct nfs_delegation *delegation;
2438
2439                 if (can_open_cached(data->state, data->o_arg.fmode,
2440                                         data->o_arg.open_flags, claim))
2441                         goto out_no_action;
2442                 rcu_read_lock();
2443                 delegation = nfs4_get_valid_delegation(data->state->inode);
2444                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2445                         goto unlock_no_action;
2446                 rcu_read_unlock();
2447         }
2448         /* Update client id. */
2449         data->o_arg.clientid = clp->cl_clientid;
2450         switch (claim) {
2451         default:
2452                 break;
2453         case NFS4_OPEN_CLAIM_PREVIOUS:
2454         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2455         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2456                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2457                 fallthrough;
2458         case NFS4_OPEN_CLAIM_FH:
2459                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2460         }
2461         data->timestamp = jiffies;
2462         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2463                                 &data->o_arg.seq_args,
2464                                 &data->o_res.seq_res,
2465                                 task) != 0)
2466                 nfs_release_seqid(data->o_arg.seqid);
2467
2468         /* Set the create mode (note dependency on the session type) */
2469         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2470         if (data->o_arg.open_flags & O_EXCL) {
2471                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2472                 if (clp->cl_mvops->minor_version == 0) {
2473                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2474                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2475                          * because ACCESS will return permission denied for
2476                          * all bits until close */
2477                         data->o_res.access_request = data->o_arg.access = 0;
2478                 } else if (nfs4_has_persistent_session(clp))
2479                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2480         }
2481         return;
2482 unlock_no_action:
2483         trace_nfs4_cached_open(data->state);
2484         rcu_read_unlock();
2485 out_no_action:
2486         task->tk_action = NULL;
2487 out_wait:
2488         nfs4_sequence_done(task, &data->o_res.seq_res);
2489 }
2490
2491 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2492 {
2493         struct nfs4_opendata *data = calldata;
2494
2495         data->rpc_status = task->tk_status;
2496
2497         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2498                 return;
2499
2500         if (task->tk_status == 0) {
2501                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2502                         switch (data->o_res.f_attr->mode & S_IFMT) {
2503                         case S_IFREG:
2504                                 break;
2505                         case S_IFLNK:
2506                                 data->rpc_status = -ELOOP;
2507                                 break;
2508                         case S_IFDIR:
2509                                 data->rpc_status = -EISDIR;
2510                                 break;
2511                         default:
2512                                 data->rpc_status = -ENOTDIR;
2513                         }
2514                 }
2515                 renew_lease(data->o_res.server, data->timestamp);
2516                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2517                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2518         }
2519         data->rpc_done = true;
2520 }
2521
2522 static void nfs4_open_release(void *calldata)
2523 {
2524         struct nfs4_opendata *data = calldata;
2525         struct nfs4_state *state = NULL;
2526
2527         /* If this request hasn't been cancelled, do nothing */
2528         if (!data->cancelled)
2529                 goto out_free;
2530         /* In case of error, no cleanup! */
2531         if (data->rpc_status != 0 || !data->rpc_done)
2532                 goto out_free;
2533         /* In case we need an open_confirm, no cleanup! */
2534         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2535                 goto out_free;
2536         state = nfs4_opendata_to_nfs4_state(data);
2537         if (!IS_ERR(state))
2538                 nfs4_close_state(state, data->o_arg.fmode);
2539 out_free:
2540         nfs4_opendata_put(data);
2541 }
2542
2543 static const struct rpc_call_ops nfs4_open_ops = {
2544         .rpc_call_prepare = nfs4_open_prepare,
2545         .rpc_call_done = nfs4_open_done,
2546         .rpc_release = nfs4_open_release,
2547 };
2548
2549 static int nfs4_run_open_task(struct nfs4_opendata *data,
2550                               struct nfs_open_context *ctx)
2551 {
2552         struct inode *dir = d_inode(data->dir);
2553         struct nfs_server *server = NFS_SERVER(dir);
2554         struct nfs_openargs *o_arg = &data->o_arg;
2555         struct nfs_openres *o_res = &data->o_res;
2556         struct rpc_task *task;
2557         struct rpc_message msg = {
2558                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2559                 .rpc_argp = o_arg,
2560                 .rpc_resp = o_res,
2561                 .rpc_cred = data->owner->so_cred,
2562         };
2563         struct rpc_task_setup task_setup_data = {
2564                 .rpc_client = server->client,
2565                 .rpc_message = &msg,
2566                 .callback_ops = &nfs4_open_ops,
2567                 .callback_data = data,
2568                 .workqueue = nfsiod_workqueue,
2569                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2570         };
2571         int status;
2572
2573         if (server->nfs_client->cl_minorversion)
2574                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2575
2576         kref_get(&data->kref);
2577         data->rpc_done = false;
2578         data->rpc_status = 0;
2579         data->cancelled = false;
2580         data->is_recover = false;
2581         if (!ctx) {
2582                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2583                 data->is_recover = true;
2584                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2585         } else {
2586                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2587                 pnfs_lgopen_prepare(data, ctx);
2588         }
2589         task = rpc_run_task(&task_setup_data);
2590         if (IS_ERR(task))
2591                 return PTR_ERR(task);
2592         status = rpc_wait_for_completion_task(task);
2593         if (status != 0) {
2594                 data->cancelled = true;
2595                 smp_wmb();
2596         } else
2597                 status = data->rpc_status;
2598         rpc_put_task(task);
2599
2600         return status;
2601 }
2602
2603 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2604 {
2605         struct inode *dir = d_inode(data->dir);
2606         struct nfs_openres *o_res = &data->o_res;
2607         int status;
2608
2609         status = nfs4_run_open_task(data, NULL);
2610         if (status != 0 || !data->rpc_done)
2611                 return status;
2612
2613         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2614
2615         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2616                 status = _nfs4_proc_open_confirm(data);
2617
2618         return status;
2619 }
2620
2621 /*
2622  * Additional permission checks in order to distinguish between an
2623  * open for read, and an open for execute. This works around the
2624  * fact that NFSv4 OPEN treats read and execute permissions as being
2625  * the same.
2626  * Note that in the non-execute case, we want to turn off permission
2627  * checking if we just created a new file (POSIX open() semantics).
2628  */
2629 static int nfs4_opendata_access(const struct cred *cred,
2630                                 struct nfs4_opendata *opendata,
2631                                 struct nfs4_state *state, fmode_t fmode,
2632                                 int openflags)
2633 {
2634         struct nfs_access_entry cache;
2635         u32 mask, flags;
2636
2637         /* access call failed or for some reason the server doesn't
2638          * support any access modes -- defer access call until later */
2639         if (opendata->o_res.access_supported == 0)
2640                 return 0;
2641
2642         mask = 0;
2643         /*
2644          * Use openflags to check for exec, because fmode won't
2645          * always have FMODE_EXEC set when file open for exec.
2646          */
2647         if (openflags & __FMODE_EXEC) {
2648                 /* ONLY check for exec rights */
2649                 if (S_ISDIR(state->inode->i_mode))
2650                         mask = NFS4_ACCESS_LOOKUP;
2651                 else
2652                         mask = NFS4_ACCESS_EXECUTE;
2653         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2654                 mask = NFS4_ACCESS_READ;
2655
2656         cache.cred = cred;
2657         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2658         nfs_access_add_cache(state->inode, &cache);
2659
2660         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2661         if ((mask & ~cache.mask & flags) == 0)
2662                 return 0;
2663
2664         return -EACCES;
2665 }
2666
2667 /*
2668  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2669  */
2670 static int _nfs4_proc_open(struct nfs4_opendata *data,
2671                            struct nfs_open_context *ctx)
2672 {
2673         struct inode *dir = d_inode(data->dir);
2674         struct nfs_server *server = NFS_SERVER(dir);
2675         struct nfs_openargs *o_arg = &data->o_arg;
2676         struct nfs_openres *o_res = &data->o_res;
2677         int status;
2678
2679         status = nfs4_run_open_task(data, ctx);
2680         if (!data->rpc_done)
2681                 return status;
2682         if (status != 0) {
2683                 if (status == -NFS4ERR_BADNAME &&
2684                                 !(o_arg->open_flags & O_CREAT))
2685                         return -ENOENT;
2686                 return status;
2687         }
2688
2689         nfs_fattr_map_and_free_names(server, &data->f_attr);
2690
2691         if (o_arg->open_flags & O_CREAT) {
2692                 if (o_arg->open_flags & O_EXCL)
2693                         data->file_created = true;
2694                 else if (o_res->cinfo.before != o_res->cinfo.after)
2695                         data->file_created = true;
2696                 if (data->file_created ||
2697                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2698                         nfs4_update_changeattr(dir, &o_res->cinfo,
2699                                         o_res->f_attr->time_start,
2700                                         NFS_INO_INVALID_DATA);
2701         }
2702         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2703                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2704         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2705                 status = _nfs4_proc_open_confirm(data);
2706                 if (status != 0)
2707                         return status;
2708         }
2709         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2710                 nfs4_sequence_free_slot(&o_res->seq_res);
2711                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2712         }
2713         return 0;
2714 }
2715
2716 /*
2717  * OPEN_EXPIRED:
2718  *      reclaim state on the server after a network partition.
2719  *      Assumes caller holds the appropriate lock
2720  */
2721 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2722 {
2723         struct nfs4_opendata *opendata;
2724         int ret;
2725
2726         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2727                         NFS4_OPEN_CLAIM_FH);
2728         if (IS_ERR(opendata))
2729                 return PTR_ERR(opendata);
2730         ret = nfs4_open_recover(opendata, state);
2731         if (ret == -ESTALE)
2732                 d_drop(ctx->dentry);
2733         nfs4_opendata_put(opendata);
2734         return ret;
2735 }
2736
2737 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2738 {
2739         struct nfs_server *server = NFS_SERVER(state->inode);
2740         struct nfs4_exception exception = { };
2741         int err;
2742
2743         do {
2744                 err = _nfs4_open_expired(ctx, state);
2745                 trace_nfs4_open_expired(ctx, 0, err);
2746                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2747                         continue;
2748                 switch (err) {
2749                 default:
2750                         goto out;
2751                 case -NFS4ERR_GRACE:
2752                 case -NFS4ERR_DELAY:
2753                         nfs4_handle_exception(server, err, &exception);
2754                         err = 0;
2755                 }
2756         } while (exception.retry);
2757 out:
2758         return err;
2759 }
2760
2761 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2762 {
2763         struct nfs_open_context *ctx;
2764         int ret;
2765
2766         ctx = nfs4_state_find_open_context(state);
2767         if (IS_ERR(ctx))
2768                 return -EAGAIN;
2769         ret = nfs4_do_open_expired(ctx, state);
2770         put_nfs_open_context(ctx);
2771         return ret;
2772 }
2773
2774 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2775                 const nfs4_stateid *stateid)
2776 {
2777         nfs_remove_bad_delegation(state->inode, stateid);
2778         nfs_state_clear_delegation(state);
2779 }
2780
2781 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2782 {
2783         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2784                 nfs_finish_clear_delegation_stateid(state, NULL);
2785 }
2786
2787 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2788 {
2789         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2790         nfs40_clear_delegation_stateid(state);
2791         nfs_state_clear_open_state_flags(state);
2792         return nfs4_open_expired(sp, state);
2793 }
2794
2795 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2796                 nfs4_stateid *stateid,
2797                 const struct cred *cred)
2798 {
2799         return -NFS4ERR_BAD_STATEID;
2800 }
2801
2802 #if defined(CONFIG_NFS_V4_1)
2803 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2804                 nfs4_stateid *stateid,
2805                 const struct cred *cred)
2806 {
2807         int status;
2808
2809         switch (stateid->type) {
2810         default:
2811                 break;
2812         case NFS4_INVALID_STATEID_TYPE:
2813         case NFS4_SPECIAL_STATEID_TYPE:
2814                 return -NFS4ERR_BAD_STATEID;
2815         case NFS4_REVOKED_STATEID_TYPE:
2816                 goto out_free;
2817         }
2818
2819         status = nfs41_test_stateid(server, stateid, cred);
2820         switch (status) {
2821         case -NFS4ERR_EXPIRED:
2822         case -NFS4ERR_ADMIN_REVOKED:
2823         case -NFS4ERR_DELEG_REVOKED:
2824                 break;
2825         default:
2826                 return status;
2827         }
2828 out_free:
2829         /* Ack the revoked state to the server */
2830         nfs41_free_stateid(server, stateid, cred, true);
2831         return -NFS4ERR_EXPIRED;
2832 }
2833
2834 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2835 {
2836         struct nfs_server *server = NFS_SERVER(state->inode);
2837         nfs4_stateid stateid;
2838         struct nfs_delegation *delegation;
2839         const struct cred *cred = NULL;
2840         int status, ret = NFS_OK;
2841
2842         /* Get the delegation credential for use by test/free_stateid */
2843         rcu_read_lock();
2844         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2845         if (delegation == NULL) {
2846                 rcu_read_unlock();
2847                 nfs_state_clear_delegation(state);
2848                 return NFS_OK;
2849         }
2850
2851         spin_lock(&delegation->lock);
2852         nfs4_stateid_copy(&stateid, &delegation->stateid);
2853
2854         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2855                                 &delegation->flags)) {
2856                 spin_unlock(&delegation->lock);
2857                 rcu_read_unlock();
2858                 return NFS_OK;
2859         }
2860
2861         if (delegation->cred)
2862                 cred = get_cred(delegation->cred);
2863         spin_unlock(&delegation->lock);
2864         rcu_read_unlock();
2865         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2866         trace_nfs4_test_delegation_stateid(state, NULL, status);
2867         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2868                 nfs_finish_clear_delegation_stateid(state, &stateid);
2869         else
2870                 ret = status;
2871
2872         put_cred(cred);
2873         return ret;
2874 }
2875
2876 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2877 {
2878         nfs4_stateid tmp;
2879
2880         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2881             nfs4_copy_delegation_stateid(state->inode, state->state,
2882                                 &tmp, NULL) &&
2883             nfs4_stateid_match_other(&state->stateid, &tmp))
2884                 nfs_state_set_delegation(state, &tmp, state->state);
2885         else
2886                 nfs_state_clear_delegation(state);
2887 }
2888
2889 /**
2890  * nfs41_check_expired_locks - possibly free a lock stateid
2891  *
2892  * @state: NFSv4 state for an inode
2893  *
2894  * Returns NFS_OK if recovery for this stateid is now finished.
2895  * Otherwise a negative NFS4ERR value is returned.
2896  */
2897 static int nfs41_check_expired_locks(struct nfs4_state *state)
2898 {
2899         int status, ret = NFS_OK;
2900         struct nfs4_lock_state *lsp, *prev = NULL;
2901         struct nfs_server *server = NFS_SERVER(state->inode);
2902
2903         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2904                 goto out;
2905
2906         spin_lock(&state->state_lock);
2907         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2908                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2909                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2910
2911                         refcount_inc(&lsp->ls_count);
2912                         spin_unlock(&state->state_lock);
2913
2914                         nfs4_put_lock_state(prev);
2915                         prev = lsp;
2916
2917                         status = nfs41_test_and_free_expired_stateid(server,
2918                                         &lsp->ls_stateid,
2919                                         cred);
2920                         trace_nfs4_test_lock_stateid(state, lsp, status);
2921                         if (status == -NFS4ERR_EXPIRED ||
2922                             status == -NFS4ERR_BAD_STATEID) {
2923                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2924                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2925                                 if (!recover_lost_locks)
2926                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2927                         } else if (status != NFS_OK) {
2928                                 ret = status;
2929                                 nfs4_put_lock_state(prev);
2930                                 goto out;
2931                         }
2932                         spin_lock(&state->state_lock);
2933                 }
2934         }
2935         spin_unlock(&state->state_lock);
2936         nfs4_put_lock_state(prev);
2937 out:
2938         return ret;
2939 }
2940
2941 /**
2942  * nfs41_check_open_stateid - possibly free an open stateid
2943  *
2944  * @state: NFSv4 state for an inode
2945  *
2946  * Returns NFS_OK if recovery for this stateid is now finished.
2947  * Otherwise a negative NFS4ERR value is returned.
2948  */
2949 static int nfs41_check_open_stateid(struct nfs4_state *state)
2950 {
2951         struct nfs_server *server = NFS_SERVER(state->inode);
2952         nfs4_stateid *stateid = &state->open_stateid;
2953         const struct cred *cred = state->owner->so_cred;
2954         int status;
2955
2956         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2957                 return -NFS4ERR_BAD_STATEID;
2958         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2959         trace_nfs4_test_open_stateid(state, NULL, status);
2960         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2961                 nfs_state_clear_open_state_flags(state);
2962                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2963                 return status;
2964         }
2965         if (nfs_open_stateid_recover_openmode(state))
2966                 return -NFS4ERR_OPENMODE;
2967         return NFS_OK;
2968 }
2969
2970 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2971 {
2972         int status;
2973
2974         status = nfs41_check_delegation_stateid(state);
2975         if (status != NFS_OK)
2976                 return status;
2977         nfs41_delegation_recover_stateid(state);
2978
2979         status = nfs41_check_expired_locks(state);
2980         if (status != NFS_OK)
2981                 return status;
2982         status = nfs41_check_open_stateid(state);
2983         if (status != NFS_OK)
2984                 status = nfs4_open_expired(sp, state);
2985         return status;
2986 }
2987 #endif
2988
2989 /*
2990  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2991  * fields corresponding to attributes that were used to store the verifier.
2992  * Make sure we clobber those fields in the later setattr call
2993  */
2994 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2995                                 struct iattr *sattr, struct nfs4_label **label)
2996 {
2997         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2998         __u32 attrset[3];
2999         unsigned ret;
3000         unsigned i;
3001
3002         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3003                 attrset[i] = opendata->o_res.attrset[i];
3004                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3005                         attrset[i] &= ~bitmask[i];
3006         }
3007
3008         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3009                 sattr->ia_valid : 0;
3010
3011         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3012                 if (sattr->ia_valid & ATTR_ATIME_SET)
3013                         ret |= ATTR_ATIME_SET;
3014                 else
3015                         ret |= ATTR_ATIME;
3016         }
3017
3018         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3019                 if (sattr->ia_valid & ATTR_MTIME_SET)
3020                         ret |= ATTR_MTIME_SET;
3021                 else
3022                         ret |= ATTR_MTIME;
3023         }
3024
3025         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3026                 *label = NULL;
3027         return ret;
3028 }
3029
3030 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3031                 int flags, struct nfs_open_context *ctx)
3032 {
3033         struct nfs4_state_owner *sp = opendata->owner;
3034         struct nfs_server *server = sp->so_server;
3035         struct dentry *dentry;
3036         struct nfs4_state *state;
3037         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3038         struct inode *dir = d_inode(opendata->dir);
3039         unsigned long dir_verifier;
3040         unsigned int seq;
3041         int ret;
3042
3043         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3044         dir_verifier = nfs_save_change_attribute(dir);
3045
3046         ret = _nfs4_proc_open(opendata, ctx);
3047         if (ret != 0)
3048                 goto out;
3049
3050         state = _nfs4_opendata_to_nfs4_state(opendata);
3051         ret = PTR_ERR(state);
3052         if (IS_ERR(state))
3053                 goto out;
3054         ctx->state = state;
3055         if (server->caps & NFS_CAP_POSIX_LOCK)
3056                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3057         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3058                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3059
3060         dentry = opendata->dentry;
3061         if (d_really_is_negative(dentry)) {
3062                 struct dentry *alias;
3063                 d_drop(dentry);
3064                 alias = d_exact_alias(dentry, state->inode);
3065                 if (!alias)
3066                         alias = d_splice_alias(igrab(state->inode), dentry);
3067                 /* d_splice_alias() can't fail here - it's a non-directory */
3068                 if (alias) {
3069                         dput(ctx->dentry);
3070                         ctx->dentry = dentry = alias;
3071                 }
3072         }
3073
3074         switch(opendata->o_arg.claim) {
3075         default:
3076                 break;
3077         case NFS4_OPEN_CLAIM_NULL:
3078         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3079         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3080                 if (!opendata->rpc_done)
3081                         break;
3082                 if (opendata->o_res.delegation_type != 0)
3083                         dir_verifier = nfs_save_change_attribute(dir);
3084                 nfs_set_verifier(dentry, dir_verifier);
3085         }
3086
3087         /* Parse layoutget results before we check for access */
3088         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3089
3090         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3091                         acc_mode, flags);
3092         if (ret != 0)
3093                 goto out;
3094
3095         if (d_inode(dentry) == state->inode) {
3096                 nfs_inode_attach_open_context(ctx);
3097                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3098                         nfs4_schedule_stateid_recovery(server, state);
3099         }
3100
3101 out:
3102         if (!opendata->cancelled)
3103                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3104         return ret;
3105 }
3106
3107 /*
3108  * Returns a referenced nfs4_state
3109  */
3110 static int _nfs4_do_open(struct inode *dir,
3111                         struct nfs_open_context *ctx,
3112                         int flags,
3113                         const struct nfs4_open_createattrs *c,
3114                         int *opened)
3115 {
3116         struct nfs4_state_owner  *sp;
3117         struct nfs4_state     *state = NULL;
3118         struct nfs_server       *server = NFS_SERVER(dir);
3119         struct nfs4_opendata *opendata;
3120         struct dentry *dentry = ctx->dentry;
3121         const struct cred *cred = ctx->cred;
3122         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3123         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3124         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3125         struct iattr *sattr = c->sattr;
3126         struct nfs4_label *label = c->label;
3127         int status;
3128
3129         /* Protect against reboot recovery conflicts */
3130         status = -ENOMEM;
3131         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3132         if (sp == NULL) {
3133                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3134                 goto out_err;
3135         }
3136         status = nfs4_client_recover_expired_lease(server->nfs_client);
3137         if (status != 0)
3138                 goto err_put_state_owner;
3139         if (d_really_is_positive(dentry))
3140                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3141         status = -ENOMEM;
3142         if (d_really_is_positive(dentry))
3143                 claim = NFS4_OPEN_CLAIM_FH;
3144         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3145                         c, claim, GFP_KERNEL);
3146         if (opendata == NULL)
3147                 goto err_put_state_owner;
3148
3149         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3150                 if (!opendata->f_attr.mdsthreshold) {
3151                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3152                         if (!opendata->f_attr.mdsthreshold)
3153                                 goto err_opendata_put;
3154                 }
3155                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3156         }
3157         if (d_really_is_positive(dentry))
3158                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3159
3160         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3161         if (status != 0)
3162                 goto err_opendata_put;
3163         state = ctx->state;
3164
3165         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3166             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3167                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3168                 /*
3169                  * send create attributes which was not set by open
3170                  * with an extra setattr.
3171                  */
3172                 if (attrs || label) {
3173                         unsigned ia_old = sattr->ia_valid;
3174
3175                         sattr->ia_valid = attrs;
3176                         nfs_fattr_init(opendata->o_res.f_attr);
3177                         status = nfs4_do_setattr(state->inode, cred,
3178                                         opendata->o_res.f_attr, sattr,
3179                                         ctx, label);
3180                         if (status == 0) {
3181                                 nfs_setattr_update_inode(state->inode, sattr,
3182                                                 opendata->o_res.f_attr);
3183                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3184                         }
3185                         sattr->ia_valid = ia_old;
3186                 }
3187         }
3188         if (opened && opendata->file_created)
3189                 *opened = 1;
3190
3191         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3192                 *ctx_th = opendata->f_attr.mdsthreshold;
3193                 opendata->f_attr.mdsthreshold = NULL;
3194         }
3195
3196         nfs4_opendata_put(opendata);
3197         nfs4_put_state_owner(sp);
3198         return 0;
3199 err_opendata_put:
3200         nfs4_opendata_put(opendata);
3201 err_put_state_owner:
3202         nfs4_put_state_owner(sp);
3203 out_err:
3204         return status;
3205 }
3206
3207
3208 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3209                                         struct nfs_open_context *ctx,
3210                                         int flags,
3211                                         struct iattr *sattr,
3212                                         struct nfs4_label *label,
3213                                         int *opened)
3214 {
3215         struct nfs_server *server = NFS_SERVER(dir);
3216         struct nfs4_exception exception = {
3217                 .interruptible = true,
3218         };
3219         struct nfs4_state *res;
3220         struct nfs4_open_createattrs c = {
3221                 .label = label,
3222                 .sattr = sattr,
3223                 .verf = {
3224                         [0] = (__u32)jiffies,
3225                         [1] = (__u32)current->pid,
3226                 },
3227         };
3228         int status;
3229
3230         do {
3231                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3232                 res = ctx->state;
3233                 trace_nfs4_open_file(ctx, flags, status);
3234                 if (status == 0)
3235                         break;
3236                 /* NOTE: BAD_SEQID means the server and client disagree about the
3237                  * book-keeping w.r.t. state-changing operations
3238                  * (OPEN/CLOSE/LOCK/LOCKU...)
3239                  * It is actually a sign of a bug on the client or on the server.
3240                  *
3241                  * If we receive a BAD_SEQID error in the particular case of
3242                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3243                  * have unhashed the old state_owner for us, and that we can
3244                  * therefore safely retry using a new one. We should still warn
3245                  * the user though...
3246                  */
3247                 if (status == -NFS4ERR_BAD_SEQID) {
3248                         pr_warn_ratelimited("NFS: v4 server %s "
3249                                         " returned a bad sequence-id error!\n",
3250                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3251                         exception.retry = 1;
3252                         continue;
3253                 }
3254                 /*
3255                  * BAD_STATEID on OPEN means that the server cancelled our
3256                  * state before it received the OPEN_CONFIRM.
3257                  * Recover by retrying the request as per the discussion
3258                  * on Page 181 of RFC3530.
3259                  */
3260                 if (status == -NFS4ERR_BAD_STATEID) {
3261                         exception.retry = 1;
3262                         continue;
3263                 }
3264                 if (status == -NFS4ERR_EXPIRED) {
3265                         nfs4_schedule_lease_recovery(server->nfs_client);
3266                         exception.retry = 1;
3267                         continue;
3268                 }
3269                 if (status == -EAGAIN) {
3270                         /* We must have found a delegation */
3271                         exception.retry = 1;
3272                         continue;
3273                 }
3274                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3275                         continue;
3276                 res = ERR_PTR(nfs4_handle_exception(server,
3277                                         status, &exception));
3278         } while (exception.retry);
3279         return res;
3280 }
3281
3282 static int _nfs4_do_setattr(struct inode *inode,
3283                             struct nfs_setattrargs *arg,
3284                             struct nfs_setattrres *res,
3285                             const struct cred *cred,
3286                             struct nfs_open_context *ctx)
3287 {
3288         struct nfs_server *server = NFS_SERVER(inode);
3289         struct rpc_message msg = {
3290                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3291                 .rpc_argp       = arg,
3292                 .rpc_resp       = res,
3293                 .rpc_cred       = cred,
3294         };
3295         const struct cred *delegation_cred = NULL;
3296         unsigned long timestamp = jiffies;
3297         bool truncate;
3298         int status;
3299
3300         nfs_fattr_init(res->fattr);
3301
3302         /* Servers should only apply open mode checks for file size changes */
3303         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3304         if (!truncate) {
3305                 nfs4_inode_make_writeable(inode);
3306                 goto zero_stateid;
3307         }
3308
3309         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3310                 /* Use that stateid */
3311         } else if (ctx != NULL && ctx->state) {
3312                 struct nfs_lock_context *l_ctx;
3313                 if (!nfs4_valid_open_stateid(ctx->state))
3314                         return -EBADF;
3315                 l_ctx = nfs_get_lock_context(ctx);
3316                 if (IS_ERR(l_ctx))
3317                         return PTR_ERR(l_ctx);
3318                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3319                                                 &arg->stateid, &delegation_cred);
3320                 nfs_put_lock_context(l_ctx);
3321                 if (status == -EIO)
3322                         return -EBADF;
3323                 else if (status == -EAGAIN)
3324                         goto zero_stateid;
3325         } else {
3326 zero_stateid:
3327                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3328         }
3329         if (delegation_cred)
3330                 msg.rpc_cred = delegation_cred;
3331
3332         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3333
3334         put_cred(delegation_cred);
3335         if (status == 0 && ctx != NULL)
3336                 renew_lease(server, timestamp);
3337         trace_nfs4_setattr(inode, &arg->stateid, status);
3338         return status;
3339 }
3340
3341 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3342                            struct nfs_fattr *fattr, struct iattr *sattr,
3343                            struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3344 {
3345         struct nfs_server *server = NFS_SERVER(inode);
3346         __u32 bitmask[NFS4_BITMASK_SZ];
3347         struct nfs4_state *state = ctx ? ctx->state : NULL;
3348         struct nfs_setattrargs  arg = {
3349                 .fh             = NFS_FH(inode),
3350                 .iap            = sattr,
3351                 .server         = server,
3352                 .bitmask = bitmask,
3353                 .label          = ilabel,
3354         };
3355         struct nfs_setattrres  res = {
3356                 .fattr          = fattr,
3357                 .server         = server,
3358         };
3359         struct nfs4_exception exception = {
3360                 .state = state,
3361                 .inode = inode,
3362                 .stateid = &arg.stateid,
3363         };
3364         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3365         int err;
3366
3367         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3368                 adjust_flags |= NFS_INO_INVALID_MODE;
3369         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3370                 adjust_flags |= NFS_INO_INVALID_OTHER;
3371
3372         do {
3373                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3374                                         inode, adjust_flags);
3375
3376                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3377                 switch (err) {
3378                 case -NFS4ERR_OPENMODE:
3379                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3380                                 pr_warn_once("NFSv4: server %s is incorrectly "
3381                                                 "applying open mode checks to "
3382                                                 "a SETATTR that is not "
3383                                                 "changing file size.\n",
3384                                                 server->nfs_client->cl_hostname);
3385                         }
3386                         if (state && !(state->state & FMODE_WRITE)) {
3387                                 err = -EBADF;
3388                                 if (sattr->ia_valid & ATTR_OPEN)
3389                                         err = -EACCES;
3390                                 goto out;
3391                         }
3392                 }
3393                 err = nfs4_handle_exception(server, err, &exception);
3394         } while (exception.retry);
3395 out:
3396         return err;
3397 }
3398
3399 static bool
3400 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3401 {
3402         if (inode == NULL || !nfs_have_layout(inode))
3403                 return false;
3404
3405         return pnfs_wait_on_layoutreturn(inode, task);
3406 }
3407
3408 /*
3409  * Update the seqid of an open stateid
3410  */
3411 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3412                 struct nfs4_state *state)
3413 {
3414         __be32 seqid_open;
3415         u32 dst_seqid;
3416         int seq;
3417
3418         for (;;) {
3419                 if (!nfs4_valid_open_stateid(state))
3420                         break;
3421                 seq = read_seqbegin(&state->seqlock);
3422                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3423                         nfs4_stateid_copy(dst, &state->open_stateid);
3424                         if (read_seqretry(&state->seqlock, seq))
3425                                 continue;
3426                         break;
3427                 }
3428                 seqid_open = state->open_stateid.seqid;
3429                 if (read_seqretry(&state->seqlock, seq))
3430                         continue;
3431
3432                 dst_seqid = be32_to_cpu(dst->seqid);
3433                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3434                         dst->seqid = seqid_open;
3435                 break;
3436         }
3437 }
3438
3439 /*
3440  * Update the seqid of an open stateid after receiving
3441  * NFS4ERR_OLD_STATEID
3442  */
3443 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3444                 struct nfs4_state *state)
3445 {
3446         __be32 seqid_open;
3447         u32 dst_seqid;
3448         bool ret;
3449         int seq, status = -EAGAIN;
3450         DEFINE_WAIT(wait);
3451
3452         for (;;) {
3453                 ret = false;
3454                 if (!nfs4_valid_open_stateid(state))
3455                         break;
3456                 seq = read_seqbegin(&state->seqlock);
3457                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3458                         if (read_seqretry(&state->seqlock, seq))
3459                                 continue;
3460                         break;
3461                 }
3462
3463                 write_seqlock(&state->seqlock);
3464                 seqid_open = state->open_stateid.seqid;
3465
3466                 dst_seqid = be32_to_cpu(dst->seqid);
3467
3468                 /* Did another OPEN bump the state's seqid?  try again: */
3469                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3470                         dst->seqid = seqid_open;
3471                         write_sequnlock(&state->seqlock);
3472                         ret = true;
3473                         break;
3474                 }
3475
3476                 /* server says we're behind but we haven't seen the update yet */
3477                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3478                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3479                 write_sequnlock(&state->seqlock);
3480                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3481
3482                 if (fatal_signal_pending(current))
3483                         status = -EINTR;
3484                 else
3485                         if (schedule_timeout(5*HZ) != 0)
3486                                 status = 0;
3487
3488                 finish_wait(&state->waitq, &wait);
3489
3490                 if (!status)
3491                         continue;
3492                 if (status == -EINTR)
3493                         break;
3494
3495                 /* we slept the whole 5 seconds, we must have lost a seqid */
3496                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3497                 ret = true;
3498                 break;
3499         }
3500
3501         return ret;
3502 }
3503
3504 struct nfs4_closedata {
3505         struct inode *inode;
3506         struct nfs4_state *state;
3507         struct nfs_closeargs arg;
3508         struct nfs_closeres res;
3509         struct {
3510                 struct nfs4_layoutreturn_args arg;
3511                 struct nfs4_layoutreturn_res res;
3512                 struct nfs4_xdr_opaque_data ld_private;
3513                 u32 roc_barrier;
3514                 bool roc;
3515         } lr;
3516         struct nfs_fattr fattr;
3517         unsigned long timestamp;
3518 };
3519
3520 static void nfs4_free_closedata(void *data)
3521 {
3522         struct nfs4_closedata *calldata = data;
3523         struct nfs4_state_owner *sp = calldata->state->owner;
3524         struct super_block *sb = calldata->state->inode->i_sb;
3525
3526         if (calldata->lr.roc)
3527                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3528                                 calldata->res.lr_ret);
3529         nfs4_put_open_state(calldata->state);
3530         nfs_free_seqid(calldata->arg.seqid);
3531         nfs4_put_state_owner(sp);
3532         nfs_sb_deactive(sb);
3533         kfree(calldata);
3534 }
3535
3536 static void nfs4_close_done(struct rpc_task *task, void *data)
3537 {
3538         struct nfs4_closedata *calldata = data;
3539         struct nfs4_state *state = calldata->state;
3540         struct nfs_server *server = NFS_SERVER(calldata->inode);
3541         nfs4_stateid *res_stateid = NULL;
3542         struct nfs4_exception exception = {
3543                 .state = state,
3544                 .inode = calldata->inode,
3545                 .stateid = &calldata->arg.stateid,
3546         };
3547
3548         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3549                 return;
3550         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3551
3552         /* Handle Layoutreturn errors */
3553         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3554                           &calldata->res.lr_ret) == -EAGAIN)
3555                 goto out_restart;
3556
3557         /* hmm. we are done with the inode, and in the process of freeing
3558          * the state_owner. we keep this around to process errors
3559          */
3560         switch (task->tk_status) {
3561                 case 0:
3562                         res_stateid = &calldata->res.stateid;
3563                         renew_lease(server, calldata->timestamp);
3564                         break;
3565                 case -NFS4ERR_ACCESS:
3566                         if (calldata->arg.bitmask != NULL) {
3567                                 calldata->arg.bitmask = NULL;
3568                                 calldata->res.fattr = NULL;
3569                                 goto out_restart;
3570
3571                         }
3572                         break;
3573                 case -NFS4ERR_OLD_STATEID:
3574                         /* Did we race with OPEN? */
3575                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3576                                                 state))
3577                                 goto out_restart;
3578                         goto out_release;
3579                 case -NFS4ERR_ADMIN_REVOKED:
3580                 case -NFS4ERR_STALE_STATEID:
3581                 case -NFS4ERR_EXPIRED:
3582                         nfs4_free_revoked_stateid(server,
3583                                         &calldata->arg.stateid,
3584                                         task->tk_msg.rpc_cred);
3585                         fallthrough;
3586                 case -NFS4ERR_BAD_STATEID:
3587                         if (calldata->arg.fmode == 0)
3588                                 break;
3589                         fallthrough;
3590                 default:
3591                         task->tk_status = nfs4_async_handle_exception(task,
3592                                         server, task->tk_status, &exception);
3593                         if (exception.retry)
3594                                 goto out_restart;
3595         }
3596         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3597                         res_stateid, calldata->arg.fmode);
3598 out_release:
3599         task->tk_status = 0;
3600         nfs_release_seqid(calldata->arg.seqid);
3601         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3602         dprintk("%s: ret = %d\n", __func__, task->tk_status);
3603         return;
3604 out_restart:
3605         task->tk_status = 0;
3606         rpc_restart_call_prepare(task);
3607         goto out_release;
3608 }
3609
3610 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3611 {
3612         struct nfs4_closedata *calldata = data;
3613         struct nfs4_state *state = calldata->state;
3614         struct inode *inode = calldata->inode;
3615         struct nfs_server *server = NFS_SERVER(inode);
3616         struct pnfs_layout_hdr *lo;
3617         bool is_rdonly, is_wronly, is_rdwr;
3618         int call_close = 0;
3619
3620         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3621                 goto out_wait;
3622
3623         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3624         spin_lock(&state->owner->so_lock);
3625         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3626         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3627         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3628         /* Calculate the change in open mode */
3629         calldata->arg.fmode = 0;
3630         if (state->n_rdwr == 0) {
3631                 if (state->n_rdonly == 0)
3632                         call_close |= is_rdonly;
3633                 else if (is_rdonly)
3634                         calldata->arg.fmode |= FMODE_READ;
3635                 if (state->n_wronly == 0)
3636                         call_close |= is_wronly;
3637                 else if (is_wronly)
3638                         calldata->arg.fmode |= FMODE_WRITE;
3639                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3640                         call_close |= is_rdwr;
3641         } else if (is_rdwr)
3642                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3643
3644         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3645         if (!nfs4_valid_open_stateid(state))
3646                 call_close = 0;
3647         spin_unlock(&state->owner->so_lock);
3648
3649         if (!call_close) {
3650                 /* Note: exit _without_ calling nfs4_close_done */
3651                 goto out_no_action;
3652         }
3653
3654         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3655                 nfs_release_seqid(calldata->arg.seqid);
3656                 goto out_wait;
3657         }
3658
3659         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3660         if (lo && !pnfs_layout_is_valid(lo)) {
3661                 calldata->arg.lr_args = NULL;
3662                 calldata->res.lr_res = NULL;
3663         }
3664
3665         if (calldata->arg.fmode == 0)
3666                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3667
3668         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3669                 /* Close-to-open cache consistency revalidation */
3670                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3671                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3672                                          server->cache_consistency_bitmask,
3673                                          inode, server, NULL);
3674                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3675                 } else
3676                         calldata->arg.bitmask = NULL;
3677         }
3678
3679         calldata->arg.share_access =
3680                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3681                                 calldata->arg.fmode, 0);
3682
3683         if (calldata->res.fattr == NULL)
3684                 calldata->arg.bitmask = NULL;
3685         else if (calldata->arg.bitmask == NULL)
3686                 calldata->res.fattr = NULL;
3687         calldata->timestamp = jiffies;
3688         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3689                                 &calldata->arg.seq_args,
3690                                 &calldata->res.seq_res,
3691                                 task) != 0)
3692                 nfs_release_seqid(calldata->arg.seqid);
3693         return;
3694 out_no_action:
3695         task->tk_action = NULL;
3696 out_wait:
3697         nfs4_sequence_done(task, &calldata->res.seq_res);
3698 }
3699
3700 static const struct rpc_call_ops nfs4_close_ops = {
3701         .rpc_call_prepare = nfs4_close_prepare,
3702         .rpc_call_done = nfs4_close_done,
3703         .rpc_release = nfs4_free_closedata,
3704 };
3705
3706 /*
3707  * It is possible for data to be read/written from a mem-mapped file
3708  * after the sys_close call (which hits the vfs layer as a flush).
3709  * This means that we can't safely call nfsv4 close on a file until
3710  * the inode is cleared. This in turn means that we are not good
3711  * NFSv4 citizens - we do not indicate to the server to update the file's
3712  * share state even when we are done with one of the three share
3713  * stateid's in the inode.
3714  *
3715  * NOTE: Caller must be holding the sp->so_owner semaphore!
3716  */
3717 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3718 {
3719         struct nfs_server *server = NFS_SERVER(state->inode);
3720         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3721         struct nfs4_closedata *calldata;
3722         struct nfs4_state_owner *sp = state->owner;
3723         struct rpc_task *task;
3724         struct rpc_message msg = {
3725                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3726                 .rpc_cred = state->owner->so_cred,
3727         };
3728         struct rpc_task_setup task_setup_data = {
3729                 .rpc_client = server->client,
3730                 .rpc_message = &msg,
3731                 .callback_ops = &nfs4_close_ops,
3732                 .workqueue = nfsiod_workqueue,
3733                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3734         };
3735         int status = -ENOMEM;
3736
3737         if (server->nfs_client->cl_minorversion)
3738                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3739
3740         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3741                 &task_setup_data.rpc_client, &msg);
3742
3743         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3744         if (calldata == NULL)
3745                 goto out;
3746         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3747         calldata->inode = state->inode;
3748         calldata->state = state;
3749         calldata->arg.fh = NFS_FH(state->inode);
3750         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3751                 goto out_free_calldata;
3752         /* Serialization for the sequence id */
3753         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3754         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3755         if (IS_ERR(calldata->arg.seqid))
3756                 goto out_free_calldata;
3757         nfs_fattr_init(&calldata->fattr);
3758         calldata->arg.fmode = 0;
3759         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3760         calldata->res.fattr = &calldata->fattr;
3761         calldata->res.seqid = calldata->arg.seqid;
3762         calldata->res.server = server;
3763         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3764         calldata->lr.roc = pnfs_roc(state->inode,
3765                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3766         if (calldata->lr.roc) {
3767                 calldata->arg.lr_args = &calldata->lr.arg;
3768                 calldata->res.lr_res = &calldata->lr.res;
3769         }
3770         nfs_sb_active(calldata->inode->i_sb);
3771
3772         msg.rpc_argp = &calldata->arg;
3773         msg.rpc_resp = &calldata->res;
3774         task_setup_data.callback_data = calldata;
3775         task = rpc_run_task(&task_setup_data);
3776         if (IS_ERR(task))
3777                 return PTR_ERR(task);
3778         status = 0;
3779         if (wait)
3780                 status = rpc_wait_for_completion_task(task);
3781         rpc_put_task(task);
3782         return status;
3783 out_free_calldata:
3784         kfree(calldata);
3785 out:
3786         nfs4_put_open_state(state);
3787         nfs4_put_state_owner(sp);
3788         return status;
3789 }
3790
3791 static struct inode *
3792 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3793                 int open_flags, struct iattr *attr, int *opened)
3794 {
3795         struct nfs4_state *state;
3796         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3797
3798         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3799
3800         /* Protect against concurrent sillydeletes */
3801         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3802
3803         nfs4_label_release_security(label);
3804
3805         if (IS_ERR(state))
3806                 return ERR_CAST(state);
3807         return state->inode;
3808 }
3809
3810 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3811 {
3812         if (ctx->state == NULL)
3813                 return;
3814         if (is_sync)
3815                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3816         else
3817                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3818 }
3819
3820 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3821 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3822 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3823
3824 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3825 {
3826         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3827         struct nfs4_server_caps_arg args = {
3828                 .fhandle = fhandle,
3829                 .bitmask = bitmask,
3830         };
3831         struct nfs4_server_caps_res res = {};
3832         struct rpc_message msg = {
3833                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3834                 .rpc_argp = &args,
3835                 .rpc_resp = &res,
3836         };
3837         int status;
3838         int i;
3839
3840         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3841                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3842                      FATTR4_WORD0_LINK_SUPPORT |
3843                      FATTR4_WORD0_SYMLINK_SUPPORT |
3844                      FATTR4_WORD0_ACLSUPPORT;
3845         if (minorversion)
3846                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3847
3848         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3849         if (status == 0) {
3850                 /* Sanity check the server answers */
3851                 switch (minorversion) {
3852                 case 0:
3853                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3854                         res.attr_bitmask[2] = 0;
3855                         break;
3856                 case 1:
3857                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3858                         break;
3859                 case 2:
3860                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3861                 }
3862                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3863                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3864                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3865                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3866                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3867                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3868                         server->caps |= NFS_CAP_ACLS;
3869                 if (res.has_links != 0)
3870                         server->caps |= NFS_CAP_HARDLINKS;
3871                 if (res.has_symlinks != 0)
3872                         server->caps |= NFS_CAP_SYMLINKS;
3873 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3874                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3875                         server->caps |= NFS_CAP_SECURITY_LABEL;
3876 #endif
3877                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3878                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3879                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3880                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3881                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3882                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3883                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3884                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3885                                 NFS_ATTR_FATTR_OWNER_NAME);
3886                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3887                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3888                                 NFS_ATTR_FATTR_GROUP_NAME);
3889                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3890                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3891                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3892                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3893                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3894                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3895                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3896                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3897                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3898                                 sizeof(server->attr_bitmask));
3899                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3900
3901                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3902                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3903                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3904                 server->cache_consistency_bitmask[2] = 0;
3905
3906                 /* Avoid a regression due to buggy server */
3907                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3908                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3909                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3910                         sizeof(server->exclcreat_bitmask));
3911
3912                 server->acl_bitmask = res.acl_bitmask;
3913                 server->fh_expire_type = res.fh_expire_type;
3914         }
3915
3916         return status;
3917 }
3918
3919 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3920 {
3921         struct nfs4_exception exception = {
3922                 .interruptible = true,
3923         };
3924         int err;
3925
3926         nfs4_server_set_init_caps(server);
3927         do {
3928                 err = nfs4_handle_exception(server,
3929                                 _nfs4_server_capabilities(server, fhandle),
3930                                 &exception);
3931         } while (exception.retry);
3932         return err;
3933 }
3934
3935 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3936                 struct nfs_fsinfo *info)
3937 {
3938         u32 bitmask[3];
3939         struct nfs4_lookup_root_arg args = {
3940                 .bitmask = bitmask,
3941         };
3942         struct nfs4_lookup_res res = {
3943                 .server = server,
3944                 .fattr = info->fattr,
3945                 .fh = fhandle,
3946         };
3947         struct rpc_message msg = {
3948                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3949                 .rpc_argp = &args,
3950                 .rpc_resp = &res,
3951         };
3952
3953         bitmask[0] = nfs4_fattr_bitmap[0];
3954         bitmask[1] = nfs4_fattr_bitmap[1];
3955         /*
3956          * Process the label in the upcoming getfattr
3957          */
3958         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3959
3960         nfs_fattr_init(info->fattr);
3961         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3962 }
3963
3964 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3965                 struct nfs_fsinfo *info)
3966 {
3967         struct nfs4_exception exception = {
3968                 .interruptible = true,
3969         };
3970         int err;
3971         do {
3972                 err = _nfs4_lookup_root(server, fhandle, info);
3973                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3974                 switch (err) {
3975                 case 0:
3976                 case -NFS4ERR_WRONGSEC:
3977                         goto out;
3978                 default:
3979                         err = nfs4_handle_exception(server, err, &exception);
3980                 }
3981         } while (exception.retry);
3982 out:
3983         return err;
3984 }
3985
3986 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3987                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3988 {
3989         struct rpc_auth_create_args auth_args = {
3990                 .pseudoflavor = flavor,
3991         };
3992         struct rpc_auth *auth;
3993
3994         auth = rpcauth_create(&auth_args, server->client);
3995         if (IS_ERR(auth))
3996                 return -EACCES;
3997         return nfs4_lookup_root(server, fhandle, info);
3998 }
3999
4000 /*
4001  * Retry pseudoroot lookup with various security flavors.  We do this when:
4002  *
4003  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4004  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4005  *
4006  * Returns zero on success, or a negative NFS4ERR value, or a
4007  * negative errno value.
4008  */
4009 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4010                               struct nfs_fsinfo *info)
4011 {
4012         /* Per 3530bis 15.33.5 */
4013         static const rpc_authflavor_t flav_array[] = {
4014                 RPC_AUTH_GSS_KRB5P,
4015                 RPC_AUTH_GSS_KRB5I,
4016                 RPC_AUTH_GSS_KRB5,
4017                 RPC_AUTH_UNIX,                  /* courtesy */
4018                 RPC_AUTH_NULL,
4019         };
4020         int status = -EPERM;
4021         size_t i;
4022
4023         if (server->auth_info.flavor_len > 0) {
4024                 /* try each flavor specified by user */
4025                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4026                         status = nfs4_lookup_root_sec(server, fhandle, info,
4027                                                 server->auth_info.flavors[i]);
4028                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4029                                 continue;
4030                         break;
4031                 }
4032         } else {
4033                 /* no flavors specified by user, try default list */
4034                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4035                         status = nfs4_lookup_root_sec(server, fhandle, info,
4036                                                       flav_array[i]);
4037                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4038                                 continue;
4039                         break;
4040                 }
4041         }
4042
4043         /*
4044          * -EACCES could mean that the user doesn't have correct permissions
4045          * to access the mount.  It could also mean that we tried to mount
4046          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4047          * existing mount programs don't handle -EACCES very well so it should
4048          * be mapped to -EPERM instead.
4049          */
4050         if (status == -EACCES)
4051                 status = -EPERM;
4052         return status;
4053 }
4054
4055 /**
4056  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4057  * @server: initialized nfs_server handle
4058  * @fhandle: we fill in the pseudo-fs root file handle
4059  * @info: we fill in an FSINFO struct
4060  * @auth_probe: probe the auth flavours
4061  *
4062  * Returns zero on success, or a negative errno.
4063  */
4064 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4065                          struct nfs_fsinfo *info,
4066                          bool auth_probe)
4067 {
4068         int status = 0;
4069
4070         if (!auth_probe)
4071                 status = nfs4_lookup_root(server, fhandle, info);
4072
4073         if (auth_probe || status == NFS4ERR_WRONGSEC)
4074                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4075                                 fhandle, info);
4076
4077         if (status == 0)
4078                 status = nfs4_server_capabilities(server, fhandle);
4079         if (status == 0)
4080                 status = nfs4_do_fsinfo(server, fhandle, info);
4081
4082         return nfs4_map_errors(status);
4083 }
4084
4085 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4086                               struct nfs_fsinfo *info)
4087 {
4088         int error;
4089         struct nfs_fattr *fattr = info->fattr;
4090
4091         error = nfs4_server_capabilities(server, mntfh);
4092         if (error < 0) {
4093                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4094                 return error;
4095         }
4096
4097         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4098         if (error < 0) {
4099                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4100                 goto out;
4101         }
4102
4103         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4104             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4105                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4106
4107 out:
4108         return error;
4109 }
4110
4111 /*
4112  * Get locations and (maybe) other attributes of a referral.
4113  * Note that we'll actually follow the referral later when
4114  * we detect fsid mismatch in inode revalidation
4115  */
4116 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4117                              const struct qstr *name, struct nfs_fattr *fattr,
4118                              struct nfs_fh *fhandle)
4119 {
4120         int status = -ENOMEM;
4121         struct page *page = NULL;
4122         struct nfs4_fs_locations *locations = NULL;
4123
4124         page = alloc_page(GFP_KERNEL);
4125         if (page == NULL)
4126                 goto out;
4127         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4128         if (locations == NULL)
4129                 goto out;
4130
4131         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4132         if (status != 0)
4133                 goto out;
4134
4135         /*
4136          * If the fsid didn't change, this is a migration event, not a
4137          * referral.  Cause us to drop into the exception handler, which
4138          * will kick off migration recovery.
4139          */
4140         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4141                 dprintk("%s: server did not return a different fsid for"
4142                         " a referral at %s\n", __func__, name->name);
4143                 status = -NFS4ERR_MOVED;
4144                 goto out;
4145         }
4146         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4147         nfs_fixup_referral_attributes(&locations->fattr);
4148
4149         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4150         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4151         memset(fhandle, 0, sizeof(struct nfs_fh));
4152 out:
4153         if (page)
4154                 __free_page(page);
4155         kfree(locations);
4156         return status;
4157 }
4158
4159 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4160                                 struct nfs_fattr *fattr, struct inode *inode)
4161 {
4162         __u32 bitmask[NFS4_BITMASK_SZ];
4163         struct nfs4_getattr_arg args = {
4164                 .fh = fhandle,
4165                 .bitmask = bitmask,
4166         };
4167         struct nfs4_getattr_res res = {
4168                 .fattr = fattr,
4169                 .server = server,
4170         };
4171         struct rpc_message msg = {
4172                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4173                 .rpc_argp = &args,
4174                 .rpc_resp = &res,
4175         };
4176         unsigned short task_flags = 0;
4177
4178         if (nfs4_has_session(server->nfs_client))
4179                 task_flags = RPC_TASK_MOVEABLE;
4180
4181         /* Is this is an attribute revalidation, subject to softreval? */
4182         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4183                 task_flags |= RPC_TASK_TIMEOUT;
4184
4185         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4186         nfs_fattr_init(fattr);
4187         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4188         return nfs4_do_call_sync(server->client, server, &msg,
4189                         &args.seq_args, &res.seq_res, task_flags);
4190 }
4191
4192 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4193                                 struct nfs_fattr *fattr, struct inode *inode)
4194 {
4195         struct nfs4_exception exception = {
4196                 .interruptible = true,
4197         };
4198         int err;
4199         do {
4200                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4201                 trace_nfs4_getattr(server, fhandle, fattr, err);
4202                 err = nfs4_handle_exception(server, err,
4203                                 &exception);
4204         } while (exception.retry);
4205         return err;
4206 }
4207
4208 /*
4209  * The file is not closed if it is opened due to the a request to change
4210  * the size of the file. The open call will not be needed once the
4211  * VFS layer lookup-intents are implemented.
4212  *
4213  * Close is called when the inode is destroyed.
4214  * If we haven't opened the file for O_WRONLY, we
4215  * need to in the size_change case to obtain a stateid.
4216  *
4217  * Got race?
4218  * Because OPEN is always done by name in nfsv4, it is
4219  * possible that we opened a different file by the same
4220  * name.  We can recognize this race condition, but we
4221  * can't do anything about it besides returning an error.
4222  *
4223  * This will be fixed with VFS changes (lookup-intent).
4224  */
4225 static int
4226 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4227                   struct iattr *sattr)
4228 {
4229         struct inode *inode = d_inode(dentry);
4230         const struct cred *cred = NULL;
4231         struct nfs_open_context *ctx = NULL;
4232         int status;
4233
4234         if (pnfs_ld_layoutret_on_setattr(inode) &&
4235             sattr->ia_valid & ATTR_SIZE &&
4236             sattr->ia_size < i_size_read(inode))
4237                 pnfs_commit_and_return_layout(inode);
4238
4239         nfs_fattr_init(fattr);
4240
4241         /* Deal with open(O_TRUNC) */
4242         if (sattr->ia_valid & ATTR_OPEN)
4243                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4244
4245         /* Optimization: if the end result is no change, don't RPC */
4246         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4247                 return 0;
4248
4249         /* Search for an existing open(O_WRITE) file */
4250         if (sattr->ia_valid & ATTR_FILE) {
4251
4252                 ctx = nfs_file_open_context(sattr->ia_file);
4253                 if (ctx)
4254                         cred = ctx->cred;
4255         }
4256
4257         /* Return any delegations if we're going to change ACLs */
4258         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4259                 nfs4_inode_make_writeable(inode);
4260
4261         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4262         if (status == 0) {
4263                 nfs_setattr_update_inode(inode, sattr, fattr);
4264                 nfs_setsecurity(inode, fattr);
4265         }
4266         return status;
4267 }
4268
4269 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4270                 struct dentry *dentry, struct nfs_fh *fhandle,
4271                 struct nfs_fattr *fattr)
4272 {
4273         struct nfs_server *server = NFS_SERVER(dir);
4274         int                    status;
4275         struct nfs4_lookup_arg args = {
4276                 .bitmask = server->attr_bitmask,
4277                 .dir_fh = NFS_FH(dir),
4278                 .name = &dentry->d_name,
4279         };
4280         struct nfs4_lookup_res res = {
4281                 .server = server,
4282                 .fattr = fattr,
4283                 .fh = fhandle,
4284         };
4285         struct rpc_message msg = {
4286                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4287                 .rpc_argp = &args,
4288                 .rpc_resp = &res,
4289         };
4290         unsigned short task_flags = 0;
4291
4292         if (server->nfs_client->cl_minorversion)
4293                 task_flags = RPC_TASK_MOVEABLE;
4294
4295         /* Is this is an attribute revalidation, subject to softreval? */
4296         if (nfs_lookup_is_soft_revalidate(dentry))
4297                 task_flags |= RPC_TASK_TIMEOUT;
4298
4299         args.bitmask = nfs4_bitmask(server, fattr->label);
4300
4301         nfs_fattr_init(fattr);
4302
4303         dprintk("NFS call  lookup %pd2\n", dentry);
4304         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4305         status = nfs4_do_call_sync(clnt, server, &msg,
4306                         &args.seq_args, &res.seq_res, task_flags);
4307         dprintk("NFS reply lookup: %d\n", status);
4308         return status;
4309 }
4310
4311 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4312 {
4313         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4314                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4315         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4316         fattr->nlink = 2;
4317 }
4318
4319 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4320                                    struct dentry *dentry, struct nfs_fh *fhandle,
4321                                    struct nfs_fattr *fattr)
4322 {
4323         struct nfs4_exception exception = {
4324                 .interruptible = true,
4325         };
4326         struct rpc_clnt *client = *clnt;
4327         const struct qstr *name = &dentry->d_name;
4328         int err;
4329         do {
4330                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4331                 trace_nfs4_lookup(dir, name, err);
4332                 switch (err) {
4333                 case -NFS4ERR_BADNAME:
4334                         err = -ENOENT;
4335                         goto out;
4336                 case -NFS4ERR_MOVED:
4337                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4338                         if (err == -NFS4ERR_MOVED)
4339                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4340                         goto out;
4341                 case -NFS4ERR_WRONGSEC:
4342                         err = -EPERM;
4343                         if (client != *clnt)
4344                                 goto out;
4345                         client = nfs4_negotiate_security(client, dir, name);
4346                         if (IS_ERR(client))
4347                                 return PTR_ERR(client);
4348
4349                         exception.retry = 1;
4350                         break;
4351                 default:
4352                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4353                 }
4354         } while (exception.retry);
4355
4356 out:
4357         if (err == 0)
4358                 *clnt = client;
4359         else if (client != *clnt)
4360                 rpc_shutdown_client(client);
4361
4362         return err;
4363 }
4364
4365 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4366                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4367 {
4368         int status;
4369         struct rpc_clnt *client = NFS_CLIENT(dir);
4370
4371         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4372         if (client != NFS_CLIENT(dir)) {
4373                 rpc_shutdown_client(client);
4374                 nfs_fixup_secinfo_attributes(fattr);
4375         }
4376         return status;
4377 }
4378
4379 struct rpc_clnt *
4380 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4381                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4382 {
4383         struct rpc_clnt *client = NFS_CLIENT(dir);
4384         int status;
4385
4386         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4387         if (status < 0)
4388                 return ERR_PTR(status);
4389         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4390 }
4391
4392 static int _nfs4_proc_lookupp(struct inode *inode,
4393                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4394 {
4395         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4396         struct nfs_server *server = NFS_SERVER(inode);
4397         int                    status;
4398         struct nfs4_lookupp_arg args = {
4399                 .bitmask = server->attr_bitmask,
4400                 .fh = NFS_FH(inode),
4401         };
4402         struct nfs4_lookupp_res res = {
4403                 .server = server,
4404                 .fattr = fattr,
4405                 .fh = fhandle,
4406         };
4407         struct rpc_message msg = {
4408                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4409                 .rpc_argp = &args,
4410                 .rpc_resp = &res,
4411         };
4412         unsigned short task_flags = 0;
4413
4414         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4415                 task_flags |= RPC_TASK_TIMEOUT;
4416
4417         args.bitmask = nfs4_bitmask(server, fattr->label);
4418
4419         nfs_fattr_init(fattr);
4420
4421         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4422         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4423                                 &res.seq_res, task_flags);
4424         dprintk("NFS reply lookupp: %d\n", status);
4425         return status;
4426 }
4427
4428 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4429                              struct nfs_fattr *fattr)
4430 {
4431         struct nfs4_exception exception = {
4432                 .interruptible = true,
4433         };
4434         int err;
4435         do {
4436                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4437                 trace_nfs4_lookupp(inode, err);
4438                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4439                                 &exception);
4440         } while (exception.retry);
4441         return err;
4442 }
4443
4444 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4445 {
4446         struct nfs_server *server = NFS_SERVER(inode);
4447         struct nfs4_accessargs args = {
4448                 .fh = NFS_FH(inode),
4449                 .access = entry->mask,
4450         };
4451         struct nfs4_accessres res = {
4452                 .server = server,
4453         };
4454         struct rpc_message msg = {
4455                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4456                 .rpc_argp = &args,
4457                 .rpc_resp = &res,
4458                 .rpc_cred = entry->cred,
4459         };
4460         int status = 0;
4461
4462         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4463                 res.fattr = nfs_alloc_fattr();
4464                 if (res.fattr == NULL)
4465                         return -ENOMEM;
4466                 args.bitmask = server->cache_consistency_bitmask;
4467         }
4468         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4469         if (!status) {
4470                 nfs_access_set_mask(entry, res.access);
4471                 if (res.fattr)
4472                         nfs_refresh_inode(inode, res.fattr);
4473         }
4474         nfs_free_fattr(res.fattr);
4475         return status;
4476 }
4477
4478 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4479 {
4480         struct nfs4_exception exception = {
4481                 .interruptible = true,
4482         };
4483         int err;
4484         do {
4485                 err = _nfs4_proc_access(inode, entry);
4486                 trace_nfs4_access(inode, err);
4487                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4488                                 &exception);
4489         } while (exception.retry);
4490         return err;
4491 }
4492
4493 /*
4494  * TODO: For the time being, we don't try to get any attributes
4495  * along with any of the zero-copy operations READ, READDIR,
4496  * READLINK, WRITE.
4497  *
4498  * In the case of the first three, we want to put the GETATTR
4499  * after the read-type operation -- this is because it is hard
4500  * to predict the length of a GETATTR response in v4, and thus
4501  * align the READ data correctly.  This means that the GETATTR
4502  * may end up partially falling into the page cache, and we should
4503  * shift it into the 'tail' of the xdr_buf before processing.
4504  * To do this efficiently, we need to know the total length
4505  * of data received, which doesn't seem to be available outside
4506  * of the RPC layer.
4507  *
4508  * In the case of WRITE, we also want to put the GETATTR after
4509  * the operation -- in this case because we want to make sure
4510  * we get the post-operation mtime and size.
4511  *
4512  * Both of these changes to the XDR layer would in fact be quite
4513  * minor, but I decided to leave them for a subsequent patch.
4514  */
4515 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4516                 unsigned int pgbase, unsigned int pglen)
4517 {
4518         struct nfs4_readlink args = {
4519                 .fh       = NFS_FH(inode),
4520                 .pgbase   = pgbase,
4521                 .pglen    = pglen,
4522                 .pages    = &page,
4523         };
4524         struct nfs4_readlink_res res;
4525         struct rpc_message msg = {
4526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4527                 .rpc_argp = &args,
4528                 .rpc_resp = &res,
4529         };
4530
4531         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4532 }
4533
4534 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4535                 unsigned int pgbase, unsigned int pglen)
4536 {
4537         struct nfs4_exception exception = {
4538                 .interruptible = true,
4539         };
4540         int err;
4541         do {
4542                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4543                 trace_nfs4_readlink(inode, err);
4544                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4545                                 &exception);
4546         } while (exception.retry);
4547         return err;
4548 }
4549
4550 /*
4551  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4552  */
4553 static int
4554 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4555                  int flags)
4556 {
4557         struct nfs_server *server = NFS_SERVER(dir);
4558         struct nfs4_label l, *ilabel = NULL;
4559         struct nfs_open_context *ctx;
4560         struct nfs4_state *state;
4561         int status = 0;
4562
4563         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4564         if (IS_ERR(ctx))
4565                 return PTR_ERR(ctx);
4566
4567         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4568
4569         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4570                 sattr->ia_mode &= ~current_umask();
4571         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4572         if (IS_ERR(state)) {
4573                 status = PTR_ERR(state);
4574                 goto out;
4575         }
4576 out:
4577         nfs4_label_release_security(ilabel);
4578         put_nfs_open_context(ctx);
4579         return status;
4580 }
4581
4582 static int
4583 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4584 {
4585         struct nfs_server *server = NFS_SERVER(dir);
4586         struct nfs_removeargs args = {
4587                 .fh = NFS_FH(dir),
4588                 .name = *name,
4589         };
4590         struct nfs_removeres res = {
4591                 .server = server,
4592         };
4593         struct rpc_message msg = {
4594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4595                 .rpc_argp = &args,
4596                 .rpc_resp = &res,
4597         };
4598         unsigned long timestamp = jiffies;
4599         int status;
4600
4601         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4602         if (status == 0) {
4603                 spin_lock(&dir->i_lock);
4604                 /* Removing a directory decrements nlink in the parent */
4605                 if (ftype == NF4DIR && dir->i_nlink > 2)
4606                         nfs4_dec_nlink_locked(dir);
4607                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4608                                               NFS_INO_INVALID_DATA);
4609                 spin_unlock(&dir->i_lock);
4610         }
4611         return status;
4612 }
4613
4614 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4615 {
4616         struct nfs4_exception exception = {
4617                 .interruptible = true,
4618         };
4619         struct inode *inode = d_inode(dentry);
4620         int err;
4621
4622         if (inode) {
4623                 if (inode->i_nlink == 1)
4624                         nfs4_inode_return_delegation(inode);
4625                 else
4626                         nfs4_inode_make_writeable(inode);
4627         }
4628         do {
4629                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4630                 trace_nfs4_remove(dir, &dentry->d_name, err);
4631                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4632                                 &exception);
4633         } while (exception.retry);
4634         return err;
4635 }
4636
4637 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4638 {
4639         struct nfs4_exception exception = {
4640                 .interruptible = true,
4641         };
4642         int err;
4643
4644         do {
4645                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4646                 trace_nfs4_remove(dir, name, err);
4647                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4648                                 &exception);
4649         } while (exception.retry);
4650         return err;
4651 }
4652
4653 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4654                 struct dentry *dentry,
4655                 struct inode *inode)
4656 {
4657         struct nfs_removeargs *args = msg->rpc_argp;
4658         struct nfs_removeres *res = msg->rpc_resp;
4659
4660         res->server = NFS_SB(dentry->d_sb);
4661         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4662         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4663
4664         nfs_fattr_init(res->dir_attr);
4665
4666         if (inode)
4667                 nfs4_inode_return_delegation(inode);
4668 }
4669
4670 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4671 {
4672         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4673                         &data->args.seq_args,
4674                         &data->res.seq_res,
4675                         task);
4676 }
4677
4678 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4679 {
4680         struct nfs_unlinkdata *data = task->tk_calldata;
4681         struct nfs_removeres *res = &data->res;
4682
4683         if (!nfs4_sequence_done(task, &res->seq_res))
4684                 return 0;
4685         if (nfs4_async_handle_error(task, res->server, NULL,
4686                                     &data->timeout) == -EAGAIN)
4687                 return 0;
4688         if (task->tk_status == 0)
4689                 nfs4_update_changeattr(dir, &res->cinfo,
4690                                 res->dir_attr->time_start,
4691                                 NFS_INO_INVALID_DATA);
4692         return 1;
4693 }
4694
4695 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4696                 struct dentry *old_dentry,
4697                 struct dentry *new_dentry)
4698 {
4699         struct nfs_renameargs *arg = msg->rpc_argp;
4700         struct nfs_renameres *res = msg->rpc_resp;
4701         struct inode *old_inode = d_inode(old_dentry);
4702         struct inode *new_inode = d_inode(new_dentry);
4703
4704         if (old_inode)
4705                 nfs4_inode_make_writeable(old_inode);
4706         if (new_inode)
4707                 nfs4_inode_return_delegation(new_inode);
4708         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4709         res->server = NFS_SB(old_dentry->d_sb);
4710         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4711 }
4712
4713 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4714 {
4715         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4716                         &data->args.seq_args,
4717                         &data->res.seq_res,
4718                         task);
4719 }
4720
4721 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4722                                  struct inode *new_dir)
4723 {
4724         struct nfs_renamedata *data = task->tk_calldata;
4725         struct nfs_renameres *res = &data->res;
4726
4727         if (!nfs4_sequence_done(task, &res->seq_res))
4728                 return 0;
4729         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4730                 return 0;
4731
4732         if (task->tk_status == 0) {
4733                 if (new_dir != old_dir) {
4734                         /* Note: If we moved a directory, nlink will change */
4735                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4736                                         res->old_fattr->time_start,
4737                                         NFS_INO_INVALID_NLINK |
4738                                             NFS_INO_INVALID_DATA);
4739                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4740                                         res->new_fattr->time_start,
4741                                         NFS_INO_INVALID_NLINK |
4742                                             NFS_INO_INVALID_DATA);
4743                 } else
4744                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4745                                         res->old_fattr->time_start,
4746                                         NFS_INO_INVALID_DATA);
4747         }
4748         return 1;
4749 }
4750
4751 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4752 {
4753         struct nfs_server *server = NFS_SERVER(inode);
4754         __u32 bitmask[NFS4_BITMASK_SZ];
4755         struct nfs4_link_arg arg = {
4756                 .fh     = NFS_FH(inode),
4757                 .dir_fh = NFS_FH(dir),
4758                 .name   = name,
4759                 .bitmask = bitmask,
4760         };
4761         struct nfs4_link_res res = {
4762                 .server = server,
4763         };
4764         struct rpc_message msg = {
4765                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4766                 .rpc_argp = &arg,
4767                 .rpc_resp = &res,
4768         };
4769         int status = -ENOMEM;
4770
4771         res.fattr = nfs_alloc_fattr_with_label(server);
4772         if (res.fattr == NULL)
4773                 goto out;
4774
4775         nfs4_inode_make_writeable(inode);
4776         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4777                                 NFS_INO_INVALID_CHANGE);
4778         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4779         if (!status) {
4780                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4781                                        NFS_INO_INVALID_DATA);
4782                 nfs4_inc_nlink(inode);
4783                 status = nfs_post_op_update_inode(inode, res.fattr);
4784                 if (!status)
4785                         nfs_setsecurity(inode, res.fattr);
4786         }
4787
4788 out:
4789         nfs_free_fattr(res.fattr);
4790         return status;
4791 }
4792
4793 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4794 {
4795         struct nfs4_exception exception = {
4796                 .interruptible = true,
4797         };
4798         int err;
4799         do {
4800                 err = nfs4_handle_exception(NFS_SERVER(inode),
4801                                 _nfs4_proc_link(inode, dir, name),
4802                                 &exception);
4803         } while (exception.retry);
4804         return err;
4805 }
4806
4807 struct nfs4_createdata {
4808         struct rpc_message msg;
4809         struct nfs4_create_arg arg;
4810         struct nfs4_create_res res;
4811         struct nfs_fh fh;
4812         struct nfs_fattr fattr;
4813 };
4814
4815 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4816                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4817 {
4818         struct nfs4_createdata *data;
4819
4820         data = kzalloc(sizeof(*data), GFP_KERNEL);
4821         if (data != NULL) {
4822                 struct nfs_server *server = NFS_SERVER(dir);
4823
4824                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4825                 if (IS_ERR(data->fattr.label))
4826                         goto out_free;
4827
4828                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4829                 data->msg.rpc_argp = &data->arg;
4830                 data->msg.rpc_resp = &data->res;
4831                 data->arg.dir_fh = NFS_FH(dir);
4832                 data->arg.server = server;
4833                 data->arg.name = name;
4834                 data->arg.attrs = sattr;
4835                 data->arg.ftype = ftype;
4836                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4837                 data->arg.umask = current_umask();
4838                 data->res.server = server;
4839                 data->res.fh = &data->fh;
4840                 data->res.fattr = &data->fattr;
4841                 nfs_fattr_init(data->res.fattr);
4842         }
4843         return data;
4844 out_free:
4845         kfree(data);
4846         return NULL;
4847 }
4848
4849 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4850 {
4851         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4852                                     &data->arg.seq_args, &data->res.seq_res, 1);
4853         if (status == 0) {
4854                 spin_lock(&dir->i_lock);
4855                 /* Creating a directory bumps nlink in the parent */
4856                 if (data->arg.ftype == NF4DIR)
4857                         nfs4_inc_nlink_locked(dir);
4858                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4859                                               data->res.fattr->time_start,
4860                                               NFS_INO_INVALID_DATA);
4861                 spin_unlock(&dir->i_lock);
4862                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4863         }
4864         return status;
4865 }
4866
4867 static void nfs4_free_createdata(struct nfs4_createdata *data)
4868 {
4869         nfs4_label_free(data->fattr.label);
4870         kfree(data);
4871 }
4872
4873 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4874                 struct page *page, unsigned int len, struct iattr *sattr,
4875                 struct nfs4_label *label)
4876 {
4877         struct nfs4_createdata *data;
4878         int status = -ENAMETOOLONG;
4879
4880         if (len > NFS4_MAXPATHLEN)
4881                 goto out;
4882
4883         status = -ENOMEM;
4884         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4885         if (data == NULL)
4886                 goto out;
4887
4888         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4889         data->arg.u.symlink.pages = &page;
4890         data->arg.u.symlink.len = len;
4891         data->arg.label = label;
4892
4893         status = nfs4_do_create(dir, dentry, data);
4894
4895         nfs4_free_createdata(data);
4896 out:
4897         return status;
4898 }
4899
4900 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4901                 struct page *page, unsigned int len, struct iattr *sattr)
4902 {
4903         struct nfs4_exception exception = {
4904                 .interruptible = true,
4905         };
4906         struct nfs4_label l, *label = NULL;
4907         int err;
4908
4909         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4910
4911         do {
4912                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4913                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4914                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4915                                 &exception);
4916         } while (exception.retry);
4917
4918         nfs4_label_release_security(label);
4919         return err;
4920 }
4921
4922 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4923                 struct iattr *sattr, struct nfs4_label *label)
4924 {
4925         struct nfs4_createdata *data;
4926         int status = -ENOMEM;
4927
4928         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4929         if (data == NULL)
4930                 goto out;
4931
4932         data->arg.label = label;
4933         status = nfs4_do_create(dir, dentry, data);
4934
4935         nfs4_free_createdata(data);
4936 out:
4937         return status;
4938 }
4939
4940 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4941                 struct iattr *sattr)
4942 {
4943         struct nfs_server *server = NFS_SERVER(dir);
4944         struct nfs4_exception exception = {
4945                 .interruptible = true,
4946         };
4947         struct nfs4_label l, *label = NULL;
4948         int err;
4949
4950         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4951
4952         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4953                 sattr->ia_mode &= ~current_umask();
4954         do {
4955                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4956                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4957                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4958                                 &exception);
4959         } while (exception.retry);
4960         nfs4_label_release_security(label);
4961
4962         return err;
4963 }
4964
4965 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
4966                               struct nfs_readdir_res *nr_res)
4967 {
4968         struct inode            *dir = d_inode(nr_arg->dentry);
4969         struct nfs_server       *server = NFS_SERVER(dir);
4970         struct nfs4_readdir_arg args = {
4971                 .fh = NFS_FH(dir),
4972                 .pages = nr_arg->pages,
4973                 .pgbase = 0,
4974                 .count = nr_arg->page_len,
4975                 .plus = nr_arg->plus,
4976         };
4977         struct nfs4_readdir_res res;
4978         struct rpc_message msg = {
4979                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4980                 .rpc_argp = &args,
4981                 .rpc_resp = &res,
4982                 .rpc_cred = nr_arg->cred,
4983         };
4984         int                     status;
4985
4986         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
4987                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
4988         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
4989                 args.bitmask = server->attr_bitmask_nl;
4990         else
4991                 args.bitmask = server->attr_bitmask;
4992
4993         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
4994         res.pgbase = args.pgbase;
4995         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
4996                         &res.seq_res, 0);
4997         if (status >= 0) {
4998                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
4999                 status += args.pgbase;
5000         }
5001
5002         nfs_invalidate_atime(dir);
5003
5004         dprintk("%s: returns %d\n", __func__, status);
5005         return status;
5006 }
5007
5008 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5009                              struct nfs_readdir_res *res)
5010 {
5011         struct nfs4_exception exception = {
5012                 .interruptible = true,
5013         };
5014         int err;
5015         do {
5016                 err = _nfs4_proc_readdir(arg, res);
5017                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5018                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5019                                             err, &exception);
5020         } while (exception.retry);
5021         return err;
5022 }
5023
5024 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5025                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5026 {
5027         struct nfs4_createdata *data;
5028         int mode = sattr->ia_mode;
5029         int status = -ENOMEM;
5030
5031         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5032         if (data == NULL)
5033                 goto out;
5034
5035         if (S_ISFIFO(mode))
5036                 data->arg.ftype = NF4FIFO;
5037         else if (S_ISBLK(mode)) {
5038                 data->arg.ftype = NF4BLK;
5039                 data->arg.u.device.specdata1 = MAJOR(rdev);
5040                 data->arg.u.device.specdata2 = MINOR(rdev);
5041         }
5042         else if (S_ISCHR(mode)) {
5043                 data->arg.ftype = NF4CHR;
5044                 data->arg.u.device.specdata1 = MAJOR(rdev);
5045                 data->arg.u.device.specdata2 = MINOR(rdev);
5046         } else if (!S_ISSOCK(mode)) {
5047                 status = -EINVAL;
5048                 goto out_free;
5049         }
5050
5051         data->arg.label = label;
5052         status = nfs4_do_create(dir, dentry, data);
5053 out_free:
5054         nfs4_free_createdata(data);
5055 out:
5056         return status;
5057 }
5058
5059 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5060                 struct iattr *sattr, dev_t rdev)
5061 {
5062         struct nfs_server *server = NFS_SERVER(dir);
5063         struct nfs4_exception exception = {
5064                 .interruptible = true,
5065         };
5066         struct nfs4_label l, *label = NULL;
5067         int err;
5068
5069         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5070
5071         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5072                 sattr->ia_mode &= ~current_umask();
5073         do {
5074                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5075                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5076                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5077                                 &exception);
5078         } while (exception.retry);
5079
5080         nfs4_label_release_security(label);
5081
5082         return err;
5083 }
5084
5085 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5086                  struct nfs_fsstat *fsstat)
5087 {
5088         struct nfs4_statfs_arg args = {
5089                 .fh = fhandle,
5090                 .bitmask = server->attr_bitmask,
5091         };
5092         struct nfs4_statfs_res res = {
5093                 .fsstat = fsstat,
5094         };
5095         struct rpc_message msg = {
5096                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5097                 .rpc_argp = &args,
5098                 .rpc_resp = &res,
5099         };
5100
5101         nfs_fattr_init(fsstat->fattr);
5102         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5103 }
5104
5105 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5106 {
5107         struct nfs4_exception exception = {
5108                 .interruptible = true,
5109         };
5110         int err;
5111         do {
5112                 err = nfs4_handle_exception(server,
5113                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5114                                 &exception);
5115         } while (exception.retry);
5116         return err;
5117 }
5118
5119 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5120                 struct nfs_fsinfo *fsinfo)
5121 {
5122         struct nfs4_fsinfo_arg args = {
5123                 .fh = fhandle,
5124                 .bitmask = server->attr_bitmask,
5125         };
5126         struct nfs4_fsinfo_res res = {
5127                 .fsinfo = fsinfo,
5128         };
5129         struct rpc_message msg = {
5130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5131                 .rpc_argp = &args,
5132                 .rpc_resp = &res,
5133         };
5134
5135         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5136 }
5137
5138 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5139 {
5140         struct nfs4_exception exception = {
5141                 .interruptible = true,
5142         };
5143         int err;
5144
5145         do {
5146                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5147                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5148                 if (err == 0) {
5149                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5150                         break;
5151                 }
5152                 err = nfs4_handle_exception(server, err, &exception);
5153         } while (exception.retry);
5154         return err;
5155 }
5156
5157 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5158 {
5159         int error;
5160
5161         nfs_fattr_init(fsinfo->fattr);
5162         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5163         if (error == 0) {
5164                 /* block layout checks this! */
5165                 server->pnfs_blksize = fsinfo->blksize;
5166                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5167         }
5168
5169         return error;
5170 }
5171
5172 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5173                 struct nfs_pathconf *pathconf)
5174 {
5175         struct nfs4_pathconf_arg args = {
5176                 .fh = fhandle,
5177                 .bitmask = server->attr_bitmask,
5178         };
5179         struct nfs4_pathconf_res res = {
5180                 .pathconf = pathconf,
5181         };
5182         struct rpc_message msg = {
5183                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5184                 .rpc_argp = &args,
5185                 .rpc_resp = &res,
5186         };
5187
5188         /* None of the pathconf attributes are mandatory to implement */
5189         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5190                 memset(pathconf, 0, sizeof(*pathconf));
5191                 return 0;
5192         }
5193
5194         nfs_fattr_init(pathconf->fattr);
5195         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5196 }
5197
5198 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5199                 struct nfs_pathconf *pathconf)
5200 {
5201         struct nfs4_exception exception = {
5202                 .interruptible = true,
5203         };
5204         int err;
5205
5206         do {
5207                 err = nfs4_handle_exception(server,
5208                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5209                                 &exception);
5210         } while (exception.retry);
5211         return err;
5212 }
5213
5214 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5215                 const struct nfs_open_context *ctx,
5216                 const struct nfs_lock_context *l_ctx,
5217                 fmode_t fmode)
5218 {
5219         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5220 }
5221 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5222
5223 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5224                 const struct nfs_open_context *ctx,
5225                 const struct nfs_lock_context *l_ctx,
5226                 fmode_t fmode)
5227 {
5228         nfs4_stateid _current_stateid;
5229
5230         /* If the current stateid represents a lost lock, then exit */
5231         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5232                 return true;
5233         return nfs4_stateid_match(stateid, &_current_stateid);
5234 }
5235
5236 static bool nfs4_error_stateid_expired(int err)
5237 {
5238         switch (err) {
5239         case -NFS4ERR_DELEG_REVOKED:
5240         case -NFS4ERR_ADMIN_REVOKED:
5241         case -NFS4ERR_BAD_STATEID:
5242         case -NFS4ERR_STALE_STATEID:
5243         case -NFS4ERR_OLD_STATEID:
5244         case -NFS4ERR_OPENMODE:
5245         case -NFS4ERR_EXPIRED:
5246                 return true;
5247         }
5248         return false;
5249 }
5250
5251 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5252 {
5253         struct nfs_server *server = NFS_SERVER(hdr->inode);
5254
5255         trace_nfs4_read(hdr, task->tk_status);
5256         if (task->tk_status < 0) {
5257                 struct nfs4_exception exception = {
5258                         .inode = hdr->inode,
5259                         .state = hdr->args.context->state,
5260                         .stateid = &hdr->args.stateid,
5261                 };
5262                 task->tk_status = nfs4_async_handle_exception(task,
5263                                 server, task->tk_status, &exception);
5264                 if (exception.retry) {
5265                         rpc_restart_call_prepare(task);
5266                         return -EAGAIN;
5267                 }
5268         }
5269
5270         if (task->tk_status > 0)
5271                 renew_lease(server, hdr->timestamp);
5272         return 0;
5273 }
5274
5275 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5276                 struct nfs_pgio_args *args)
5277 {
5278
5279         if (!nfs4_error_stateid_expired(task->tk_status) ||
5280                 nfs4_stateid_is_current(&args->stateid,
5281                                 args->context,
5282                                 args->lock_context,
5283                                 FMODE_READ))
5284                 return false;
5285         rpc_restart_call_prepare(task);
5286         return true;
5287 }
5288
5289 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5290                                          struct nfs_pgio_header *hdr)
5291 {
5292         struct nfs_server *server = NFS_SERVER(hdr->inode);
5293         struct rpc_message *msg = &task->tk_msg;
5294
5295         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5296             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5297                 server->caps &= ~NFS_CAP_READ_PLUS;
5298                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5299                 rpc_restart_call_prepare(task);
5300                 return true;
5301         }
5302         return false;
5303 }
5304
5305 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5306 {
5307         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5308                 return -EAGAIN;
5309         if (nfs4_read_stateid_changed(task, &hdr->args))
5310                 return -EAGAIN;
5311         if (nfs4_read_plus_not_supported(task, hdr))
5312                 return -EAGAIN;
5313         if (task->tk_status > 0)
5314                 nfs_invalidate_atime(hdr->inode);
5315         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5316                                     nfs4_read_done_cb(task, hdr);
5317 }
5318
5319 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5320 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5321                                     struct rpc_message *msg)
5322 {
5323         /* Note: We don't use READ_PLUS with pNFS yet */
5324         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5325                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5326 }
5327 #else
5328 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5329                                     struct rpc_message *msg)
5330 {
5331 }
5332 #endif /* CONFIG_NFS_V4_2 */
5333
5334 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5335                                  struct rpc_message *msg)
5336 {
5337         hdr->timestamp   = jiffies;
5338         if (!hdr->pgio_done_cb)
5339                 hdr->pgio_done_cb = nfs4_read_done_cb;
5340         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5341         nfs42_read_plus_support(hdr, msg);
5342         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5343 }
5344
5345 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5346                                       struct nfs_pgio_header *hdr)
5347 {
5348         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5349                         &hdr->args.seq_args,
5350                         &hdr->res.seq_res,
5351                         task))
5352                 return 0;
5353         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5354                                 hdr->args.lock_context,
5355                                 hdr->rw_mode) == -EIO)
5356                 return -EIO;
5357         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5358                 return -EIO;
5359         return 0;
5360 }
5361
5362 static int nfs4_write_done_cb(struct rpc_task *task,
5363                               struct nfs_pgio_header *hdr)
5364 {
5365         struct inode *inode = hdr->inode;
5366
5367         trace_nfs4_write(hdr, task->tk_status);
5368         if (task->tk_status < 0) {
5369                 struct nfs4_exception exception = {
5370                         .inode = hdr->inode,
5371                         .state = hdr->args.context->state,
5372                         .stateid = &hdr->args.stateid,
5373                 };
5374                 task->tk_status = nfs4_async_handle_exception(task,
5375                                 NFS_SERVER(inode), task->tk_status,
5376                                 &exception);
5377                 if (exception.retry) {
5378                         rpc_restart_call_prepare(task);
5379                         return -EAGAIN;
5380                 }
5381         }
5382         if (task->tk_status >= 0) {
5383                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5384                 nfs_writeback_update_inode(hdr);
5385         }
5386         return 0;
5387 }
5388
5389 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5390                 struct nfs_pgio_args *args)
5391 {
5392
5393         if (!nfs4_error_stateid_expired(task->tk_status) ||
5394                 nfs4_stateid_is_current(&args->stateid,
5395                                 args->context,
5396                                 args->lock_context,
5397                                 FMODE_WRITE))
5398                 return false;
5399         rpc_restart_call_prepare(task);
5400         return true;
5401 }
5402
5403 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5404 {
5405         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5406                 return -EAGAIN;
5407         if (nfs4_write_stateid_changed(task, &hdr->args))
5408                 return -EAGAIN;
5409         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5410                 nfs4_write_done_cb(task, hdr);
5411 }
5412
5413 static
5414 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5415 {
5416         /* Don't request attributes for pNFS or O_DIRECT writes */
5417         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5418                 return false;
5419         /* Otherwise, request attributes if and only if we don't hold
5420          * a delegation
5421          */
5422         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5423 }
5424
5425 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src,
5426                              struct inode *inode, struct nfs_server *server,
5427                              struct nfs4_label *label)
5428 {
5429         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5430         unsigned int i;
5431
5432         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5433
5434         if (cache_validity & NFS_INO_INVALID_CHANGE)
5435                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5436         if (cache_validity & NFS_INO_INVALID_ATIME)
5437                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5438         if (cache_validity & NFS_INO_INVALID_MODE)
5439                 bitmask[1] |= FATTR4_WORD1_MODE;
5440         if (cache_validity & NFS_INO_INVALID_OTHER)
5441                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5442         if (cache_validity & NFS_INO_INVALID_NLINK)
5443                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5444         if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL)
5445                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5446         if (cache_validity & NFS_INO_INVALID_CTIME)
5447                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5448         if (cache_validity & NFS_INO_INVALID_MTIME)
5449                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5450         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5451                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5452
5453         if (cache_validity & NFS_INO_INVALID_SIZE)
5454                 bitmask[0] |= FATTR4_WORD0_SIZE;
5455
5456         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5457                 bitmask[i] &= server->attr_bitmask[i];
5458 }
5459
5460 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5461                                   struct rpc_message *msg,
5462                                   struct rpc_clnt **clnt)
5463 {
5464         struct nfs_server *server = NFS_SERVER(hdr->inode);
5465
5466         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5467                 hdr->args.bitmask = NULL;
5468                 hdr->res.fattr = NULL;
5469         } else {
5470                 nfs4_bitmask_set(hdr->args.bitmask_store,
5471                                  server->cache_consistency_bitmask,
5472                                  hdr->inode, server, NULL);
5473                 hdr->args.bitmask = hdr->args.bitmask_store;
5474         }
5475
5476         if (!hdr->pgio_done_cb)
5477                 hdr->pgio_done_cb = nfs4_write_done_cb;
5478         hdr->res.server = server;
5479         hdr->timestamp   = jiffies;
5480
5481         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5482         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5483         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5484 }
5485
5486 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5487 {
5488         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5489                         &data->args.seq_args,
5490                         &data->res.seq_res,
5491                         task);
5492 }
5493
5494 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5495 {
5496         struct inode *inode = data->inode;
5497
5498         trace_nfs4_commit(data, task->tk_status);
5499         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5500                                     NULL, NULL) == -EAGAIN) {
5501                 rpc_restart_call_prepare(task);
5502                 return -EAGAIN;
5503         }
5504         return 0;
5505 }
5506
5507 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5508 {
5509         if (!nfs4_sequence_done(task, &data->res.seq_res))
5510                 return -EAGAIN;
5511         return data->commit_done_cb(task, data);
5512 }
5513
5514 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5515                                    struct rpc_clnt **clnt)
5516 {
5517         struct nfs_server *server = NFS_SERVER(data->inode);
5518
5519         if (data->commit_done_cb == NULL)
5520                 data->commit_done_cb = nfs4_commit_done_cb;
5521         data->res.server = server;
5522         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5523         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5524         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5525 }
5526
5527 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5528                                 struct nfs_commitres *res)
5529 {
5530         struct inode *dst_inode = file_inode(dst);
5531         struct nfs_server *server = NFS_SERVER(dst_inode);
5532         struct rpc_message msg = {
5533                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5534                 .rpc_argp = args,
5535                 .rpc_resp = res,
5536         };
5537
5538         args->fh = NFS_FH(dst_inode);
5539         return nfs4_call_sync(server->client, server, &msg,
5540                         &args->seq_args, &res->seq_res, 1);
5541 }
5542
5543 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5544 {
5545         struct nfs_commitargs args = {
5546                 .offset = offset,
5547                 .count = count,
5548         };
5549         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5550         struct nfs4_exception exception = { };
5551         int status;
5552
5553         do {
5554                 status = _nfs4_proc_commit(dst, &args, res);
5555                 status = nfs4_handle_exception(dst_server, status, &exception);
5556         } while (exception.retry);
5557
5558         return status;
5559 }
5560
5561 struct nfs4_renewdata {
5562         struct nfs_client       *client;
5563         unsigned long           timestamp;
5564 };
5565
5566 /*
5567  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5568  * standalone procedure for queueing an asynchronous RENEW.
5569  */
5570 static void nfs4_renew_release(void *calldata)
5571 {
5572         struct nfs4_renewdata *data = calldata;
5573         struct nfs_client *clp = data->client;
5574
5575         if (refcount_read(&clp->cl_count) > 1)
5576                 nfs4_schedule_state_renewal(clp);
5577         nfs_put_client(clp);
5578         kfree(data);
5579 }
5580
5581 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5582 {
5583         struct nfs4_renewdata *data = calldata;
5584         struct nfs_client *clp = data->client;
5585         unsigned long timestamp = data->timestamp;
5586
5587         trace_nfs4_renew_async(clp, task->tk_status);
5588         switch (task->tk_status) {
5589         case 0:
5590                 break;
5591         case -NFS4ERR_LEASE_MOVED:
5592                 nfs4_schedule_lease_moved_recovery(clp);
5593                 break;
5594         default:
5595                 /* Unless we're shutting down, schedule state recovery! */
5596                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5597                         return;
5598                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5599                         nfs4_schedule_lease_recovery(clp);
5600                         return;
5601                 }
5602                 nfs4_schedule_path_down_recovery(clp);
5603         }
5604         do_renew_lease(clp, timestamp);
5605 }
5606
5607 static const struct rpc_call_ops nfs4_renew_ops = {
5608         .rpc_call_done = nfs4_renew_done,
5609         .rpc_release = nfs4_renew_release,
5610 };
5611
5612 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5613 {
5614         struct rpc_message msg = {
5615                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5616                 .rpc_argp       = clp,
5617                 .rpc_cred       = cred,
5618         };
5619         struct nfs4_renewdata *data;
5620
5621         if (renew_flags == 0)
5622                 return 0;
5623         if (!refcount_inc_not_zero(&clp->cl_count))
5624                 return -EIO;
5625         data = kmalloc(sizeof(*data), GFP_NOFS);
5626         if (data == NULL) {
5627                 nfs_put_client(clp);
5628                 return -ENOMEM;
5629         }
5630         data->client = clp;
5631         data->timestamp = jiffies;
5632         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5633                         &nfs4_renew_ops, data);
5634 }
5635
5636 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5637 {
5638         struct rpc_message msg = {
5639                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5640                 .rpc_argp       = clp,
5641                 .rpc_cred       = cred,
5642         };
5643         unsigned long now = jiffies;
5644         int status;
5645
5646         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5647         if (status < 0)
5648                 return status;
5649         do_renew_lease(clp, now);
5650         return 0;
5651 }
5652
5653 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5654 {
5655         return server->caps & NFS_CAP_ACLS;
5656 }
5657
5658 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5659  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5660  * the stack.
5661  */
5662 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5663
5664 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5665                 struct page **pages)
5666 {
5667         struct page *newpage, **spages;
5668         int rc = 0;
5669         size_t len;
5670         spages = pages;
5671
5672         do {
5673                 len = min_t(size_t, PAGE_SIZE, buflen);
5674                 newpage = alloc_page(GFP_KERNEL);
5675
5676                 if (newpage == NULL)
5677                         goto unwind;
5678                 memcpy(page_address(newpage), buf, len);
5679                 buf += len;
5680                 buflen -= len;
5681                 *pages++ = newpage;
5682                 rc++;
5683         } while (buflen != 0);
5684
5685         return rc;
5686
5687 unwind:
5688         for(; rc > 0; rc--)
5689                 __free_page(spages[rc-1]);
5690         return -ENOMEM;
5691 }
5692
5693 struct nfs4_cached_acl {
5694         int cached;
5695         size_t len;
5696         char data[];
5697 };
5698
5699 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5700 {
5701         struct nfs_inode *nfsi = NFS_I(inode);
5702
5703         spin_lock(&inode->i_lock);
5704         kfree(nfsi->nfs4_acl);
5705         nfsi->nfs4_acl = acl;
5706         spin_unlock(&inode->i_lock);
5707 }
5708
5709 static void nfs4_zap_acl_attr(struct inode *inode)
5710 {
5711         nfs4_set_cached_acl(inode, NULL);
5712 }
5713
5714 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5715 {
5716         struct nfs_inode *nfsi = NFS_I(inode);
5717         struct nfs4_cached_acl *acl;
5718         int ret = -ENOENT;
5719
5720         spin_lock(&inode->i_lock);
5721         acl = nfsi->nfs4_acl;
5722         if (acl == NULL)
5723                 goto out;
5724         if (buf == NULL) /* user is just asking for length */
5725                 goto out_len;
5726         if (acl->cached == 0)
5727                 goto out;
5728         ret = -ERANGE; /* see getxattr(2) man page */
5729         if (acl->len > buflen)
5730                 goto out;
5731         memcpy(buf, acl->data, acl->len);
5732 out_len:
5733         ret = acl->len;
5734 out:
5735         spin_unlock(&inode->i_lock);
5736         return ret;
5737 }
5738
5739 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5740 {
5741         struct nfs4_cached_acl *acl;
5742         size_t buflen = sizeof(*acl) + acl_len;
5743
5744         if (buflen <= PAGE_SIZE) {
5745                 acl = kmalloc(buflen, GFP_KERNEL);
5746                 if (acl == NULL)
5747                         goto out;
5748                 acl->cached = 1;
5749                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5750         } else {
5751                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5752                 if (acl == NULL)
5753                         goto out;
5754                 acl->cached = 0;
5755         }
5756         acl->len = acl_len;
5757 out:
5758         nfs4_set_cached_acl(inode, acl);
5759 }
5760
5761 /*
5762  * The getxattr API returns the required buffer length when called with a
5763  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5764  * the required buf.  On a NULL buf, we send a page of data to the server
5765  * guessing that the ACL request can be serviced by a page. If so, we cache
5766  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5767  * the cache. If not so, we throw away the page, and cache the required
5768  * length. The next getxattr call will then produce another round trip to
5769  * the server, this time with the input buf of the required size.
5770  */
5771 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5772 {
5773         struct page **pages;
5774         struct nfs_getaclargs args = {
5775                 .fh = NFS_FH(inode),
5776                 .acl_len = buflen,
5777         };
5778         struct nfs_getaclres res = {
5779                 .acl_len = buflen,
5780         };
5781         struct rpc_message msg = {
5782                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5783                 .rpc_argp = &args,
5784                 .rpc_resp = &res,
5785         };
5786         unsigned int npages;
5787         int ret = -ENOMEM, i;
5788         struct nfs_server *server = NFS_SERVER(inode);
5789
5790         if (buflen == 0)
5791                 buflen = server->rsize;
5792
5793         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5794         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5795         if (!pages)
5796                 return -ENOMEM;
5797
5798         args.acl_pages = pages;
5799
5800         for (i = 0; i < npages; i++) {
5801                 pages[i] = alloc_page(GFP_KERNEL);
5802                 if (!pages[i])
5803                         goto out_free;
5804         }
5805
5806         /* for decoding across pages */
5807         res.acl_scratch = alloc_page(GFP_KERNEL);
5808         if (!res.acl_scratch)
5809                 goto out_free;
5810
5811         args.acl_len = npages * PAGE_SIZE;
5812
5813         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5814                 __func__, buf, buflen, npages, args.acl_len);
5815         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5816                              &msg, &args.seq_args, &res.seq_res, 0);
5817         if (ret)
5818                 goto out_free;
5819
5820         /* Handle the case where the passed-in buffer is too short */
5821         if (res.acl_flags & NFS4_ACL_TRUNC) {
5822                 /* Did the user only issue a request for the acl length? */
5823                 if (buf == NULL)
5824                         goto out_ok;
5825                 ret = -ERANGE;
5826                 goto out_free;
5827         }
5828         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5829         if (buf) {
5830                 if (res.acl_len > buflen) {
5831                         ret = -ERANGE;
5832                         goto out_free;
5833                 }
5834                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5835         }
5836 out_ok:
5837         ret = res.acl_len;
5838 out_free:
5839         for (i = 0; i < npages; i++)
5840                 if (pages[i])
5841                         __free_page(pages[i]);
5842         if (res.acl_scratch)
5843                 __free_page(res.acl_scratch);
5844         kfree(pages);
5845         return ret;
5846 }
5847
5848 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5849 {
5850         struct nfs4_exception exception = {
5851                 .interruptible = true,
5852         };
5853         ssize_t ret;
5854         do {
5855                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5856                 trace_nfs4_get_acl(inode, ret);
5857                 if (ret >= 0)
5858                         break;
5859                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5860         } while (exception.retry);
5861         return ret;
5862 }
5863
5864 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5865 {
5866         struct nfs_server *server = NFS_SERVER(inode);
5867         int ret;
5868
5869         if (!nfs4_server_supports_acls(server))
5870                 return -EOPNOTSUPP;
5871         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
5872         if (ret < 0)
5873                 return ret;
5874         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5875                 nfs_zap_acl_cache(inode);
5876         ret = nfs4_read_cached_acl(inode, buf, buflen);
5877         if (ret != -ENOENT)
5878                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5879                  * but no cached acl data, just the acl length */
5880                 return ret;
5881         return nfs4_get_acl_uncached(inode, buf, buflen);
5882 }
5883
5884 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5885 {
5886         struct nfs_server *server = NFS_SERVER(inode);
5887         struct page *pages[NFS4ACL_MAXPAGES];
5888         struct nfs_setaclargs arg = {
5889                 .fh             = NFS_FH(inode),
5890                 .acl_pages      = pages,
5891                 .acl_len        = buflen,
5892         };
5893         struct nfs_setaclres res;
5894         struct rpc_message msg = {
5895                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5896                 .rpc_argp       = &arg,
5897                 .rpc_resp       = &res,
5898         };
5899         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5900         int ret, i;
5901
5902         /* You can't remove system.nfs4_acl: */
5903         if (buflen == 0)
5904                 return -EINVAL;
5905         if (!nfs4_server_supports_acls(server))
5906                 return -EOPNOTSUPP;
5907         if (npages > ARRAY_SIZE(pages))
5908                 return -ERANGE;
5909         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5910         if (i < 0)
5911                 return i;
5912         nfs4_inode_make_writeable(inode);
5913         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5914
5915         /*
5916          * Free each page after tx, so the only ref left is
5917          * held by the network stack
5918          */
5919         for (; i > 0; i--)
5920                 put_page(pages[i-1]);
5921
5922         /*
5923          * Acl update can result in inode attribute update.
5924          * so mark the attribute cache invalid.
5925          */
5926         spin_lock(&inode->i_lock);
5927         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
5928                                              NFS_INO_INVALID_CTIME |
5929                                              NFS_INO_REVAL_FORCED);
5930         spin_unlock(&inode->i_lock);
5931         nfs_access_zap_cache(inode);
5932         nfs_zap_acl_cache(inode);
5933         return ret;
5934 }
5935
5936 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5937 {
5938         struct nfs4_exception exception = { };
5939         int err;
5940         do {
5941                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5942                 trace_nfs4_set_acl(inode, err);
5943                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
5944                         /*
5945                          * no need to retry since the kernel
5946                          * isn't involved in encoding the ACEs.
5947                          */
5948                         err = -EINVAL;
5949                         break;
5950                 }
5951                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5952                                 &exception);
5953         } while (exception.retry);
5954         return err;
5955 }
5956
5957 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5958 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5959                                         size_t buflen)
5960 {
5961         struct nfs_server *server = NFS_SERVER(inode);
5962         struct nfs4_label label = {0, 0, buflen, buf};
5963
5964         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5965         struct nfs_fattr fattr = {
5966                 .label = &label,
5967         };
5968         struct nfs4_getattr_arg arg = {
5969                 .fh             = NFS_FH(inode),
5970                 .bitmask        = bitmask,
5971         };
5972         struct nfs4_getattr_res res = {
5973                 .fattr          = &fattr,
5974                 .server         = server,
5975         };
5976         struct rpc_message msg = {
5977                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5978                 .rpc_argp       = &arg,
5979                 .rpc_resp       = &res,
5980         };
5981         int ret;
5982
5983         nfs_fattr_init(&fattr);
5984
5985         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5986         if (ret)
5987                 return ret;
5988         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5989                 return -ENOENT;
5990         return label.len;
5991 }
5992
5993 static int nfs4_get_security_label(struct inode *inode, void *buf,
5994                                         size_t buflen)
5995 {
5996         struct nfs4_exception exception = {
5997                 .interruptible = true,
5998         };
5999         int err;
6000
6001         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6002                 return -EOPNOTSUPP;
6003
6004         do {
6005                 err = _nfs4_get_security_label(inode, buf, buflen);
6006                 trace_nfs4_get_security_label(inode, err);
6007                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6008                                 &exception);
6009         } while (exception.retry);
6010         return err;
6011 }
6012
6013 static int _nfs4_do_set_security_label(struct inode *inode,
6014                 struct nfs4_label *ilabel,
6015                 struct nfs_fattr *fattr)
6016 {
6017
6018         struct iattr sattr = {0};
6019         struct nfs_server *server = NFS_SERVER(inode);
6020         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6021         struct nfs_setattrargs arg = {
6022                 .fh             = NFS_FH(inode),
6023                 .iap            = &sattr,
6024                 .server         = server,
6025                 .bitmask        = bitmask,
6026                 .label          = ilabel,
6027         };
6028         struct nfs_setattrres res = {
6029                 .fattr          = fattr,
6030                 .server         = server,
6031         };
6032         struct rpc_message msg = {
6033                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6034                 .rpc_argp       = &arg,
6035                 .rpc_resp       = &res,
6036         };
6037         int status;
6038
6039         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6040
6041         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6042         if (status)
6043                 dprintk("%s failed: %d\n", __func__, status);
6044
6045         return status;
6046 }
6047
6048 static int nfs4_do_set_security_label(struct inode *inode,
6049                 struct nfs4_label *ilabel,
6050                 struct nfs_fattr *fattr)
6051 {
6052         struct nfs4_exception exception = { };
6053         int err;
6054
6055         do {
6056                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6057                 trace_nfs4_set_security_label(inode, err);
6058                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6059                                 &exception);
6060         } while (exception.retry);
6061         return err;
6062 }
6063
6064 static int
6065 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6066 {
6067         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6068         struct nfs_fattr *fattr;
6069         int status;
6070
6071         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6072                 return -EOPNOTSUPP;
6073
6074         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6075         if (fattr == NULL)
6076                 return -ENOMEM;
6077
6078         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6079         if (status == 0)
6080                 nfs_setsecurity(inode, fattr);
6081
6082         return status;
6083 }
6084 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6085
6086
6087 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6088                                     nfs4_verifier *bootverf)
6089 {
6090         __be32 verf[2];
6091
6092         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6093                 /* An impossible timestamp guarantees this value
6094                  * will never match a generated boot time. */
6095                 verf[0] = cpu_to_be32(U32_MAX);
6096                 verf[1] = cpu_to_be32(U32_MAX);
6097         } else {
6098                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6099                 u64 ns = ktime_to_ns(nn->boot_time);
6100
6101                 verf[0] = cpu_to_be32(ns >> 32);
6102                 verf[1] = cpu_to_be32(ns);
6103         }
6104         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6105 }
6106
6107 static size_t
6108 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6109 {
6110         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6111         struct nfs_netns_client *nn_clp = nn->nfs_client;
6112         const char *id;
6113
6114         buf[0] = '\0';
6115
6116         if (nn_clp) {
6117                 rcu_read_lock();
6118                 id = rcu_dereference(nn_clp->identifier);
6119                 if (id)
6120                         strscpy(buf, id, buflen);
6121                 rcu_read_unlock();
6122         }
6123
6124         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6125                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6126
6127         return strlen(buf);
6128 }
6129
6130 static int
6131 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6132 {
6133         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6134         size_t buflen;
6135         size_t len;
6136         char *str;
6137
6138         if (clp->cl_owner_id != NULL)
6139                 return 0;
6140
6141         rcu_read_lock();
6142         len = 14 +
6143                 strlen(clp->cl_rpcclient->cl_nodename) +
6144                 1 +
6145                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6146                 1;
6147         rcu_read_unlock();
6148
6149         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6150         if (buflen)
6151                 len += buflen + 1;
6152
6153         if (len > NFS4_OPAQUE_LIMIT + 1)
6154                 return -EINVAL;
6155
6156         /*
6157          * Since this string is allocated at mount time, and held until the
6158          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6159          * about a memory-reclaim deadlock.
6160          */
6161         str = kmalloc(len, GFP_KERNEL);
6162         if (!str)
6163                 return -ENOMEM;
6164
6165         rcu_read_lock();
6166         if (buflen)
6167                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6168                           clp->cl_rpcclient->cl_nodename, buf,
6169                           rpc_peeraddr2str(clp->cl_rpcclient,
6170                                            RPC_DISPLAY_ADDR));
6171         else
6172                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6173                           clp->cl_rpcclient->cl_nodename,
6174                           rpc_peeraddr2str(clp->cl_rpcclient,
6175                                            RPC_DISPLAY_ADDR));
6176         rcu_read_unlock();
6177
6178         clp->cl_owner_id = str;
6179         return 0;
6180 }
6181
6182 static int
6183 nfs4_init_uniform_client_string(struct nfs_client *clp)
6184 {
6185         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6186         size_t buflen;
6187         size_t len;
6188         char *str;
6189
6190         if (clp->cl_owner_id != NULL)
6191                 return 0;
6192
6193         len = 10 + 10 + 1 + 10 + 1 +
6194                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6195
6196         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6197         if (buflen)
6198                 len += buflen + 1;
6199
6200         if (len > NFS4_OPAQUE_LIMIT + 1)
6201                 return -EINVAL;
6202
6203         /*
6204          * Since this string is allocated at mount time, and held until the
6205          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6206          * about a memory-reclaim deadlock.
6207          */
6208         str = kmalloc(len, GFP_KERNEL);
6209         if (!str)
6210                 return -ENOMEM;
6211
6212         if (buflen)
6213                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6214                           clp->rpc_ops->version, clp->cl_minorversion,
6215                           buf, clp->cl_rpcclient->cl_nodename);
6216         else
6217                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6218                           clp->rpc_ops->version, clp->cl_minorversion,
6219                           clp->cl_rpcclient->cl_nodename);
6220         clp->cl_owner_id = str;
6221         return 0;
6222 }
6223
6224 /*
6225  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6226  * services.  Advertise one based on the address family of the
6227  * clientaddr.
6228  */
6229 static unsigned int
6230 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6231 {
6232         if (strchr(clp->cl_ipaddr, ':') != NULL)
6233                 return scnprintf(buf, len, "tcp6");
6234         else
6235                 return scnprintf(buf, len, "tcp");
6236 }
6237
6238 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6239 {
6240         struct nfs4_setclientid *sc = calldata;
6241
6242         if (task->tk_status == 0)
6243                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6244 }
6245
6246 static const struct rpc_call_ops nfs4_setclientid_ops = {
6247         .rpc_call_done = nfs4_setclientid_done,
6248 };
6249
6250 /**
6251  * nfs4_proc_setclientid - Negotiate client ID
6252  * @clp: state data structure
6253  * @program: RPC program for NFSv4 callback service
6254  * @port: IP port number for NFS4 callback service
6255  * @cred: credential to use for this call
6256  * @res: where to place the result
6257  *
6258  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6259  */
6260 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6261                 unsigned short port, const struct cred *cred,
6262                 struct nfs4_setclientid_res *res)
6263 {
6264         nfs4_verifier sc_verifier;
6265         struct nfs4_setclientid setclientid = {
6266                 .sc_verifier = &sc_verifier,
6267                 .sc_prog = program,
6268                 .sc_clnt = clp,
6269         };
6270         struct rpc_message msg = {
6271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6272                 .rpc_argp = &setclientid,
6273                 .rpc_resp = res,
6274                 .rpc_cred = cred,
6275         };
6276         struct rpc_task_setup task_setup_data = {
6277                 .rpc_client = clp->cl_rpcclient,
6278                 .rpc_message = &msg,
6279                 .callback_ops = &nfs4_setclientid_ops,
6280                 .callback_data = &setclientid,
6281                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6282         };
6283         unsigned long now = jiffies;
6284         int status;
6285
6286         /* nfs_client_id4 */
6287         nfs4_init_boot_verifier(clp, &sc_verifier);
6288
6289         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6290                 status = nfs4_init_uniform_client_string(clp);
6291         else
6292                 status = nfs4_init_nonuniform_client_string(clp);
6293
6294         if (status)
6295                 goto out;
6296
6297         /* cb_client4 */
6298         setclientid.sc_netid_len =
6299                                 nfs4_init_callback_netid(clp,
6300                                                 setclientid.sc_netid,
6301                                                 sizeof(setclientid.sc_netid));
6302         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6303                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6304                                 clp->cl_ipaddr, port >> 8, port & 255);
6305
6306         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6307                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6308                 clp->cl_owner_id);
6309
6310         status = nfs4_call_sync_custom(&task_setup_data);
6311         if (setclientid.sc_cred) {
6312                 kfree(clp->cl_acceptor);
6313                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6314                 put_rpccred(setclientid.sc_cred);
6315         }
6316
6317         if (status == 0)
6318                 do_renew_lease(clp, now);
6319 out:
6320         trace_nfs4_setclientid(clp, status);
6321         dprintk("NFS reply setclientid: %d\n", status);
6322         return status;
6323 }
6324
6325 /**
6326  * nfs4_proc_setclientid_confirm - Confirm client ID
6327  * @clp: state data structure
6328  * @arg: result of a previous SETCLIENTID
6329  * @cred: credential to use for this call
6330  *
6331  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6332  */
6333 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6334                 struct nfs4_setclientid_res *arg,
6335                 const struct cred *cred)
6336 {
6337         struct rpc_message msg = {
6338                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6339                 .rpc_argp = arg,
6340                 .rpc_cred = cred,
6341         };
6342         int status;
6343
6344         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6345                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6346                 clp->cl_clientid);
6347         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6348                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6349         trace_nfs4_setclientid_confirm(clp, status);
6350         dprintk("NFS reply setclientid_confirm: %d\n", status);
6351         return status;
6352 }
6353
6354 struct nfs4_delegreturndata {
6355         struct nfs4_delegreturnargs args;
6356         struct nfs4_delegreturnres res;
6357         struct nfs_fh fh;
6358         nfs4_stateid stateid;
6359         unsigned long timestamp;
6360         struct {
6361                 struct nfs4_layoutreturn_args arg;
6362                 struct nfs4_layoutreturn_res res;
6363                 struct nfs4_xdr_opaque_data ld_private;
6364                 u32 roc_barrier;
6365                 bool roc;
6366         } lr;
6367         struct nfs_fattr fattr;
6368         int rpc_status;
6369         struct inode *inode;
6370 };
6371
6372 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6373 {
6374         struct nfs4_delegreturndata *data = calldata;
6375         struct nfs4_exception exception = {
6376                 .inode = data->inode,
6377                 .stateid = &data->stateid,
6378                 .task_is_privileged = data->args.seq_args.sa_privileged,
6379         };
6380
6381         if (!nfs4_sequence_done(task, &data->res.seq_res))
6382                 return;
6383
6384         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6385
6386         /* Handle Layoutreturn errors */
6387         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6388                           &data->res.lr_ret) == -EAGAIN)
6389                 goto out_restart;
6390
6391         switch (task->tk_status) {
6392         case 0:
6393                 renew_lease(data->res.server, data->timestamp);
6394                 break;
6395         case -NFS4ERR_ADMIN_REVOKED:
6396         case -NFS4ERR_DELEG_REVOKED:
6397         case -NFS4ERR_EXPIRED:
6398                 nfs4_free_revoked_stateid(data->res.server,
6399                                 data->args.stateid,
6400                                 task->tk_msg.rpc_cred);
6401                 fallthrough;
6402         case -NFS4ERR_BAD_STATEID:
6403         case -NFS4ERR_STALE_STATEID:
6404         case -ETIMEDOUT:
6405                 task->tk_status = 0;
6406                 break;
6407         case -NFS4ERR_OLD_STATEID:
6408                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6409                         nfs4_stateid_seqid_inc(&data->stateid);
6410                 if (data->args.bitmask) {
6411                         data->args.bitmask = NULL;
6412                         data->res.fattr = NULL;
6413                 }
6414                 goto out_restart;
6415         case -NFS4ERR_ACCESS:
6416                 if (data->args.bitmask) {
6417                         data->args.bitmask = NULL;
6418                         data->res.fattr = NULL;
6419                         goto out_restart;
6420                 }
6421                 fallthrough;
6422         default:
6423                 task->tk_status = nfs4_async_handle_exception(task,
6424                                 data->res.server, task->tk_status,
6425                                 &exception);
6426                 if (exception.retry)
6427                         goto out_restart;
6428         }
6429         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6430         data->rpc_status = task->tk_status;
6431         return;
6432 out_restart:
6433         task->tk_status = 0;
6434         rpc_restart_call_prepare(task);
6435 }
6436
6437 static void nfs4_delegreturn_release(void *calldata)
6438 {
6439         struct nfs4_delegreturndata *data = calldata;
6440         struct inode *inode = data->inode;
6441
6442         if (data->lr.roc)
6443                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6444                                  data->res.lr_ret);
6445         if (inode) {
6446                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6447                 nfs_iput_and_deactive(inode);
6448         }
6449         kfree(calldata);
6450 }
6451
6452 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6453 {
6454         struct nfs4_delegreturndata *d_data;
6455         struct pnfs_layout_hdr *lo;
6456
6457         d_data = (struct nfs4_delegreturndata *)data;
6458
6459         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6460                 nfs4_sequence_done(task, &d_data->res.seq_res);
6461                 return;
6462         }
6463
6464         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6465         if (lo && !pnfs_layout_is_valid(lo)) {
6466                 d_data->args.lr_args = NULL;
6467                 d_data->res.lr_res = NULL;
6468         }
6469
6470         nfs4_setup_sequence(d_data->res.server->nfs_client,
6471                         &d_data->args.seq_args,
6472                         &d_data->res.seq_res,
6473                         task);
6474 }
6475
6476 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6477         .rpc_call_prepare = nfs4_delegreturn_prepare,
6478         .rpc_call_done = nfs4_delegreturn_done,
6479         .rpc_release = nfs4_delegreturn_release,
6480 };
6481
6482 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6483 {
6484         struct nfs4_delegreturndata *data;
6485         struct nfs_server *server = NFS_SERVER(inode);
6486         struct rpc_task *task;
6487         struct rpc_message msg = {
6488                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6489                 .rpc_cred = cred,
6490         };
6491         struct rpc_task_setup task_setup_data = {
6492                 .rpc_client = server->client,
6493                 .rpc_message = &msg,
6494                 .callback_ops = &nfs4_delegreturn_ops,
6495                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
6496         };
6497         int status = 0;
6498
6499         data = kzalloc(sizeof(*data), GFP_NOFS);
6500         if (data == NULL)
6501                 return -ENOMEM;
6502
6503         nfs4_state_protect(server->nfs_client,
6504                         NFS_SP4_MACH_CRED_CLEANUP,
6505                         &task_setup_data.rpc_client, &msg);
6506
6507         data->args.fhandle = &data->fh;
6508         data->args.stateid = &data->stateid;
6509         nfs4_bitmask_set(data->args.bitmask_store,
6510                          server->cache_consistency_bitmask, inode, server,
6511                          NULL);
6512         data->args.bitmask = data->args.bitmask_store;
6513         nfs_copy_fh(&data->fh, NFS_FH(inode));
6514         nfs4_stateid_copy(&data->stateid, stateid);
6515         data->res.fattr = &data->fattr;
6516         data->res.server = server;
6517         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6518         data->lr.arg.ld_private = &data->lr.ld_private;
6519         nfs_fattr_init(data->res.fattr);
6520         data->timestamp = jiffies;
6521         data->rpc_status = 0;
6522         data->inode = nfs_igrab_and_active(inode);
6523         if (data->inode || issync) {
6524                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6525                                         cred);
6526                 if (data->lr.roc) {
6527                         data->args.lr_args = &data->lr.arg;
6528                         data->res.lr_res = &data->lr.res;
6529                 }
6530         }
6531
6532         if (!data->inode)
6533                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6534                                    1);
6535         else
6536                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6537                                    0);
6538         task_setup_data.callback_data = data;
6539         msg.rpc_argp = &data->args;
6540         msg.rpc_resp = &data->res;
6541         task = rpc_run_task(&task_setup_data);
6542         if (IS_ERR(task))
6543                 return PTR_ERR(task);
6544         if (!issync)
6545                 goto out;
6546         status = rpc_wait_for_completion_task(task);
6547         if (status != 0)
6548                 goto out;
6549         status = data->rpc_status;
6550 out:
6551         rpc_put_task(task);
6552         return status;
6553 }
6554
6555 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6556 {
6557         struct nfs_server *server = NFS_SERVER(inode);
6558         struct nfs4_exception exception = { };
6559         int err;
6560         do {
6561                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6562                 trace_nfs4_delegreturn(inode, stateid, err);
6563                 switch (err) {
6564                         case -NFS4ERR_STALE_STATEID:
6565                         case -NFS4ERR_EXPIRED:
6566                         case 0:
6567                                 return 0;
6568                 }
6569                 err = nfs4_handle_exception(server, err, &exception);
6570         } while (exception.retry);
6571         return err;
6572 }
6573
6574 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6575 {
6576         struct inode *inode = state->inode;
6577         struct nfs_server *server = NFS_SERVER(inode);
6578         struct nfs_client *clp = server->nfs_client;
6579         struct nfs_lockt_args arg = {
6580                 .fh = NFS_FH(inode),
6581                 .fl = request,
6582         };
6583         struct nfs_lockt_res res = {
6584                 .denied = request,
6585         };
6586         struct rpc_message msg = {
6587                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6588                 .rpc_argp       = &arg,
6589                 .rpc_resp       = &res,
6590                 .rpc_cred       = state->owner->so_cred,
6591         };
6592         struct nfs4_lock_state *lsp;
6593         int status;
6594
6595         arg.lock_owner.clientid = clp->cl_clientid;
6596         status = nfs4_set_lock_state(state, request);
6597         if (status != 0)
6598                 goto out;
6599         lsp = request->fl_u.nfs4_fl.owner;
6600         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6601         arg.lock_owner.s_dev = server->s_dev;
6602         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6603         switch (status) {
6604                 case 0:
6605                         request->fl_type = F_UNLCK;
6606                         break;
6607                 case -NFS4ERR_DENIED:
6608                         status = 0;
6609         }
6610         request->fl_ops->fl_release_private(request);
6611         request->fl_ops = NULL;
6612 out:
6613         return status;
6614 }
6615
6616 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6617 {
6618         struct nfs4_exception exception = {
6619                 .interruptible = true,
6620         };
6621         int err;
6622
6623         do {
6624                 err = _nfs4_proc_getlk(state, cmd, request);
6625                 trace_nfs4_get_lock(request, state, cmd, err);
6626                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6627                                 &exception);
6628         } while (exception.retry);
6629         return err;
6630 }
6631
6632 /*
6633  * Update the seqid of a lock stateid after receiving
6634  * NFS4ERR_OLD_STATEID
6635  */
6636 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6637                 struct nfs4_lock_state *lsp)
6638 {
6639         struct nfs4_state *state = lsp->ls_state;
6640         bool ret = false;
6641
6642         spin_lock(&state->state_lock);
6643         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6644                 goto out;
6645         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6646                 nfs4_stateid_seqid_inc(dst);
6647         else
6648                 dst->seqid = lsp->ls_stateid.seqid;
6649         ret = true;
6650 out:
6651         spin_unlock(&state->state_lock);
6652         return ret;
6653 }
6654
6655 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6656                 struct nfs4_lock_state *lsp)
6657 {
6658         struct nfs4_state *state = lsp->ls_state;
6659         bool ret;
6660
6661         spin_lock(&state->state_lock);
6662         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6663         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6664         spin_unlock(&state->state_lock);
6665         return ret;
6666 }
6667
6668 struct nfs4_unlockdata {
6669         struct nfs_locku_args arg;
6670         struct nfs_locku_res res;
6671         struct nfs4_lock_state *lsp;
6672         struct nfs_open_context *ctx;
6673         struct nfs_lock_context *l_ctx;
6674         struct file_lock fl;
6675         struct nfs_server *server;
6676         unsigned long timestamp;
6677 };
6678
6679 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6680                 struct nfs_open_context *ctx,
6681                 struct nfs4_lock_state *lsp,
6682                 struct nfs_seqid *seqid)
6683 {
6684         struct nfs4_unlockdata *p;
6685         struct nfs4_state *state = lsp->ls_state;
6686         struct inode *inode = state->inode;
6687
6688         p = kzalloc(sizeof(*p), GFP_NOFS);
6689         if (p == NULL)
6690                 return NULL;
6691         p->arg.fh = NFS_FH(inode);
6692         p->arg.fl = &p->fl;
6693         p->arg.seqid = seqid;
6694         p->res.seqid = seqid;
6695         p->lsp = lsp;
6696         /* Ensure we don't close file until we're done freeing locks! */
6697         p->ctx = get_nfs_open_context(ctx);
6698         p->l_ctx = nfs_get_lock_context(ctx);
6699         locks_init_lock(&p->fl);
6700         locks_copy_lock(&p->fl, fl);
6701         p->server = NFS_SERVER(inode);
6702         spin_lock(&state->state_lock);
6703         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6704         spin_unlock(&state->state_lock);
6705         return p;
6706 }
6707
6708 static void nfs4_locku_release_calldata(void *data)
6709 {
6710         struct nfs4_unlockdata *calldata = data;
6711         nfs_free_seqid(calldata->arg.seqid);
6712         nfs4_put_lock_state(calldata->lsp);
6713         nfs_put_lock_context(calldata->l_ctx);
6714         put_nfs_open_context(calldata->ctx);
6715         kfree(calldata);
6716 }
6717
6718 static void nfs4_locku_done(struct rpc_task *task, void *data)
6719 {
6720         struct nfs4_unlockdata *calldata = data;
6721         struct nfs4_exception exception = {
6722                 .inode = calldata->lsp->ls_state->inode,
6723                 .stateid = &calldata->arg.stateid,
6724         };
6725
6726         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6727                 return;
6728         switch (task->tk_status) {
6729                 case 0:
6730                         renew_lease(calldata->server, calldata->timestamp);
6731                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6732                         if (nfs4_update_lock_stateid(calldata->lsp,
6733                                         &calldata->res.stateid))
6734                                 break;
6735                         fallthrough;
6736                 case -NFS4ERR_ADMIN_REVOKED:
6737                 case -NFS4ERR_EXPIRED:
6738                         nfs4_free_revoked_stateid(calldata->server,
6739                                         &calldata->arg.stateid,
6740                                         task->tk_msg.rpc_cred);
6741                         fallthrough;
6742                 case -NFS4ERR_BAD_STATEID:
6743                 case -NFS4ERR_STALE_STATEID:
6744                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6745                                                 calldata->lsp))
6746                                 rpc_restart_call_prepare(task);
6747                         break;
6748                 case -NFS4ERR_OLD_STATEID:
6749                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6750                                                 calldata->lsp))
6751                                 rpc_restart_call_prepare(task);
6752                         break;
6753                 default:
6754                         task->tk_status = nfs4_async_handle_exception(task,
6755                                         calldata->server, task->tk_status,
6756                                         &exception);
6757                         if (exception.retry)
6758                                 rpc_restart_call_prepare(task);
6759         }
6760         nfs_release_seqid(calldata->arg.seqid);
6761 }
6762
6763 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6764 {
6765         struct nfs4_unlockdata *calldata = data;
6766
6767         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6768                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6769                 return;
6770
6771         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6772                 goto out_wait;
6773         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6774                 /* Note: exit _without_ running nfs4_locku_done */
6775                 goto out_no_action;
6776         }
6777         calldata->timestamp = jiffies;
6778         if (nfs4_setup_sequence(calldata->server->nfs_client,
6779                                 &calldata->arg.seq_args,
6780                                 &calldata->res.seq_res,
6781                                 task) != 0)
6782                 nfs_release_seqid(calldata->arg.seqid);
6783         return;
6784 out_no_action:
6785         task->tk_action = NULL;
6786 out_wait:
6787         nfs4_sequence_done(task, &calldata->res.seq_res);
6788 }
6789
6790 static const struct rpc_call_ops nfs4_locku_ops = {
6791         .rpc_call_prepare = nfs4_locku_prepare,
6792         .rpc_call_done = nfs4_locku_done,
6793         .rpc_release = nfs4_locku_release_calldata,
6794 };
6795
6796 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6797                 struct nfs_open_context *ctx,
6798                 struct nfs4_lock_state *lsp,
6799                 struct nfs_seqid *seqid)
6800 {
6801         struct nfs4_unlockdata *data;
6802         struct rpc_message msg = {
6803                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6804                 .rpc_cred = ctx->cred,
6805         };
6806         struct rpc_task_setup task_setup_data = {
6807                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6808                 .rpc_message = &msg,
6809                 .callback_ops = &nfs4_locku_ops,
6810                 .workqueue = nfsiod_workqueue,
6811                 .flags = RPC_TASK_ASYNC,
6812         };
6813         struct nfs_client *client =
6814                 NFS_SERVER(lsp->ls_state->inode)->nfs_client;
6815
6816         if (client->cl_minorversion)
6817                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6818
6819         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6820                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6821
6822         /* Ensure this is an unlock - when canceling a lock, the
6823          * canceled lock is passed in, and it won't be an unlock.
6824          */
6825         fl->fl_type = F_UNLCK;
6826         if (fl->fl_flags & FL_CLOSE)
6827                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6828
6829         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6830         if (data == NULL) {
6831                 nfs_free_seqid(seqid);
6832                 return ERR_PTR(-ENOMEM);
6833         }
6834
6835         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6836         msg.rpc_argp = &data->arg;
6837         msg.rpc_resp = &data->res;
6838         task_setup_data.callback_data = data;
6839         return rpc_run_task(&task_setup_data);
6840 }
6841
6842 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6843 {
6844         struct inode *inode = state->inode;
6845         struct nfs4_state_owner *sp = state->owner;
6846         struct nfs_inode *nfsi = NFS_I(inode);
6847         struct nfs_seqid *seqid;
6848         struct nfs4_lock_state *lsp;
6849         struct rpc_task *task;
6850         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6851         int status = 0;
6852         unsigned char fl_flags = request->fl_flags;
6853
6854         status = nfs4_set_lock_state(state, request);
6855         /* Unlock _before_ we do the RPC call */
6856         request->fl_flags |= FL_EXISTS;
6857         /* Exclude nfs_delegation_claim_locks() */
6858         mutex_lock(&sp->so_delegreturn_mutex);
6859         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6860         down_read(&nfsi->rwsem);
6861         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6862                 up_read(&nfsi->rwsem);
6863                 mutex_unlock(&sp->so_delegreturn_mutex);
6864                 goto out;
6865         }
6866         up_read(&nfsi->rwsem);
6867         mutex_unlock(&sp->so_delegreturn_mutex);
6868         if (status != 0)
6869                 goto out;
6870         /* Is this a delegated lock? */
6871         lsp = request->fl_u.nfs4_fl.owner;
6872         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6873                 goto out;
6874         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6875         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6876         status = -ENOMEM;
6877         if (IS_ERR(seqid))
6878                 goto out;
6879         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6880         status = PTR_ERR(task);
6881         if (IS_ERR(task))
6882                 goto out;
6883         status = rpc_wait_for_completion_task(task);
6884         rpc_put_task(task);
6885 out:
6886         request->fl_flags = fl_flags;
6887         trace_nfs4_unlock(request, state, F_SETLK, status);
6888         return status;
6889 }
6890
6891 struct nfs4_lockdata {
6892         struct nfs_lock_args arg;
6893         struct nfs_lock_res res;
6894         struct nfs4_lock_state *lsp;
6895         struct nfs_open_context *ctx;
6896         struct file_lock fl;
6897         unsigned long timestamp;
6898         int rpc_status;
6899         int cancelled;
6900         struct nfs_server *server;
6901 };
6902
6903 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6904                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6905                 gfp_t gfp_mask)
6906 {
6907         struct nfs4_lockdata *p;
6908         struct inode *inode = lsp->ls_state->inode;
6909         struct nfs_server *server = NFS_SERVER(inode);
6910         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6911
6912         p = kzalloc(sizeof(*p), gfp_mask);
6913         if (p == NULL)
6914                 return NULL;
6915
6916         p->arg.fh = NFS_FH(inode);
6917         p->arg.fl = &p->fl;
6918         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6919         if (IS_ERR(p->arg.open_seqid))
6920                 goto out_free;
6921         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6922         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6923         if (IS_ERR(p->arg.lock_seqid))
6924                 goto out_free_seqid;
6925         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6926         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6927         p->arg.lock_owner.s_dev = server->s_dev;
6928         p->res.lock_seqid = p->arg.lock_seqid;
6929         p->lsp = lsp;
6930         p->server = server;
6931         p->ctx = get_nfs_open_context(ctx);
6932         locks_init_lock(&p->fl);
6933         locks_copy_lock(&p->fl, fl);
6934         return p;
6935 out_free_seqid:
6936         nfs_free_seqid(p->arg.open_seqid);
6937 out_free:
6938         kfree(p);
6939         return NULL;
6940 }
6941
6942 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6943 {
6944         struct nfs4_lockdata *data = calldata;
6945         struct nfs4_state *state = data->lsp->ls_state;
6946
6947         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6948                 goto out_wait;
6949         /* Do we need to do an open_to_lock_owner? */
6950         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6951                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6952                         goto out_release_lock_seqid;
6953                 }
6954                 nfs4_stateid_copy(&data->arg.open_stateid,
6955                                 &state->open_stateid);
6956                 data->arg.new_lock_owner = 1;
6957                 data->res.open_seqid = data->arg.open_seqid;
6958         } else {
6959                 data->arg.new_lock_owner = 0;
6960                 nfs4_stateid_copy(&data->arg.lock_stateid,
6961                                 &data->lsp->ls_stateid);
6962         }
6963         if (!nfs4_valid_open_stateid(state)) {
6964                 data->rpc_status = -EBADF;
6965                 task->tk_action = NULL;
6966                 goto out_release_open_seqid;
6967         }
6968         data->timestamp = jiffies;
6969         if (nfs4_setup_sequence(data->server->nfs_client,
6970                                 &data->arg.seq_args,
6971                                 &data->res.seq_res,
6972                                 task) == 0)
6973                 return;
6974 out_release_open_seqid:
6975         nfs_release_seqid(data->arg.open_seqid);
6976 out_release_lock_seqid:
6977         nfs_release_seqid(data->arg.lock_seqid);
6978 out_wait:
6979         nfs4_sequence_done(task, &data->res.seq_res);
6980         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
6981 }
6982
6983 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6984 {
6985         struct nfs4_lockdata *data = calldata;
6986         struct nfs4_lock_state *lsp = data->lsp;
6987
6988         if (!nfs4_sequence_done(task, &data->res.seq_res))
6989                 return;
6990
6991         data->rpc_status = task->tk_status;
6992         switch (task->tk_status) {
6993         case 0:
6994                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6995                                 data->timestamp);
6996                 if (data->arg.new_lock && !data->cancelled) {
6997                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6998                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6999                                 goto out_restart;
7000                 }
7001                 if (data->arg.new_lock_owner != 0) {
7002                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7003                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7004                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7005                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7006                         goto out_restart;
7007                 break;
7008         case -NFS4ERR_BAD_STATEID:
7009         case -NFS4ERR_OLD_STATEID:
7010         case -NFS4ERR_STALE_STATEID:
7011         case -NFS4ERR_EXPIRED:
7012                 if (data->arg.new_lock_owner != 0) {
7013                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7014                                                 &lsp->ls_state->open_stateid))
7015                                 goto out_restart;
7016                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7017                                                 &lsp->ls_stateid))
7018                                 goto out_restart;
7019         }
7020 out_done:
7021         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7022         return;
7023 out_restart:
7024         if (!data->cancelled)
7025                 rpc_restart_call_prepare(task);
7026         goto out_done;
7027 }
7028
7029 static void nfs4_lock_release(void *calldata)
7030 {
7031         struct nfs4_lockdata *data = calldata;
7032
7033         nfs_free_seqid(data->arg.open_seqid);
7034         if (data->cancelled && data->rpc_status == 0) {
7035                 struct rpc_task *task;
7036                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7037                                 data->arg.lock_seqid);
7038                 if (!IS_ERR(task))
7039                         rpc_put_task_async(task);
7040                 dprintk("%s: cancelling lock!\n", __func__);
7041         } else
7042                 nfs_free_seqid(data->arg.lock_seqid);
7043         nfs4_put_lock_state(data->lsp);
7044         put_nfs_open_context(data->ctx);
7045         kfree(data);
7046 }
7047
7048 static const struct rpc_call_ops nfs4_lock_ops = {
7049         .rpc_call_prepare = nfs4_lock_prepare,
7050         .rpc_call_done = nfs4_lock_done,
7051         .rpc_release = nfs4_lock_release,
7052 };
7053
7054 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7055 {
7056         switch (error) {
7057         case -NFS4ERR_ADMIN_REVOKED:
7058         case -NFS4ERR_EXPIRED:
7059         case -NFS4ERR_BAD_STATEID:
7060                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7061                 if (new_lock_owner != 0 ||
7062                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7063                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7064                 break;
7065         case -NFS4ERR_STALE_STATEID:
7066                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7067                 nfs4_schedule_lease_recovery(server->nfs_client);
7068         }
7069 }
7070
7071 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7072 {
7073         struct nfs4_lockdata *data;
7074         struct rpc_task *task;
7075         struct rpc_message msg = {
7076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7077                 .rpc_cred = state->owner->so_cred,
7078         };
7079         struct rpc_task_setup task_setup_data = {
7080                 .rpc_client = NFS_CLIENT(state->inode),
7081                 .rpc_message = &msg,
7082                 .callback_ops = &nfs4_lock_ops,
7083                 .workqueue = nfsiod_workqueue,
7084                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7085         };
7086         int ret;
7087         struct nfs_client *client = NFS_SERVER(state->inode)->nfs_client;
7088
7089         if (client->cl_minorversion)
7090                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7091
7092         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7093                         fl->fl_u.nfs4_fl.owner,
7094                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7095         if (data == NULL)
7096                 return -ENOMEM;
7097         if (IS_SETLKW(cmd))
7098                 data->arg.block = 1;
7099         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7100                                 recovery_type > NFS_LOCK_NEW);
7101         msg.rpc_argp = &data->arg;
7102         msg.rpc_resp = &data->res;
7103         task_setup_data.callback_data = data;
7104         if (recovery_type > NFS_LOCK_NEW) {
7105                 if (recovery_type == NFS_LOCK_RECLAIM)
7106                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7107         } else
7108                 data->arg.new_lock = 1;
7109         task = rpc_run_task(&task_setup_data);
7110         if (IS_ERR(task))
7111                 return PTR_ERR(task);
7112         ret = rpc_wait_for_completion_task(task);
7113         if (ret == 0) {
7114                 ret = data->rpc_status;
7115                 if (ret)
7116                         nfs4_handle_setlk_error(data->server, data->lsp,
7117                                         data->arg.new_lock_owner, ret);
7118         } else
7119                 data->cancelled = true;
7120         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7121         rpc_put_task(task);
7122         dprintk("%s: ret = %d\n", __func__, ret);
7123         return ret;
7124 }
7125
7126 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7127 {
7128         struct nfs_server *server = NFS_SERVER(state->inode);
7129         struct nfs4_exception exception = {
7130                 .inode = state->inode,
7131         };
7132         int err;
7133
7134         do {
7135                 /* Cache the lock if possible... */
7136                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7137                         return 0;
7138                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7139                 if (err != -NFS4ERR_DELAY)
7140                         break;
7141                 nfs4_handle_exception(server, err, &exception);
7142         } while (exception.retry);
7143         return err;
7144 }
7145
7146 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7147 {
7148         struct nfs_server *server = NFS_SERVER(state->inode);
7149         struct nfs4_exception exception = {
7150                 .inode = state->inode,
7151         };
7152         int err;
7153
7154         err = nfs4_set_lock_state(state, request);
7155         if (err != 0)
7156                 return err;
7157         if (!recover_lost_locks) {
7158                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7159                 return 0;
7160         }
7161         do {
7162                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7163                         return 0;
7164                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7165                 switch (err) {
7166                 default:
7167                         goto out;
7168                 case -NFS4ERR_GRACE:
7169                 case -NFS4ERR_DELAY:
7170                         nfs4_handle_exception(server, err, &exception);
7171                         err = 0;
7172                 }
7173         } while (exception.retry);
7174 out:
7175         return err;
7176 }
7177
7178 #if defined(CONFIG_NFS_V4_1)
7179 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7180 {
7181         struct nfs4_lock_state *lsp;
7182         int status;
7183
7184         status = nfs4_set_lock_state(state, request);
7185         if (status != 0)
7186                 return status;
7187         lsp = request->fl_u.nfs4_fl.owner;
7188         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7189             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7190                 return 0;
7191         return nfs4_lock_expired(state, request);
7192 }
7193 #endif
7194
7195 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7196 {
7197         struct nfs_inode *nfsi = NFS_I(state->inode);
7198         struct nfs4_state_owner *sp = state->owner;
7199         unsigned char fl_flags = request->fl_flags;
7200         int status;
7201
7202         request->fl_flags |= FL_ACCESS;
7203         status = locks_lock_inode_wait(state->inode, request);
7204         if (status < 0)
7205                 goto out;
7206         mutex_lock(&sp->so_delegreturn_mutex);
7207         down_read(&nfsi->rwsem);
7208         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7209                 /* Yes: cache locks! */
7210                 /* ...but avoid races with delegation recall... */
7211                 request->fl_flags = fl_flags & ~FL_SLEEP;
7212                 status = locks_lock_inode_wait(state->inode, request);
7213                 up_read(&nfsi->rwsem);
7214                 mutex_unlock(&sp->so_delegreturn_mutex);
7215                 goto out;
7216         }
7217         up_read(&nfsi->rwsem);
7218         mutex_unlock(&sp->so_delegreturn_mutex);
7219         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7220 out:
7221         request->fl_flags = fl_flags;
7222         return status;
7223 }
7224
7225 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7226 {
7227         struct nfs4_exception exception = {
7228                 .state = state,
7229                 .inode = state->inode,
7230                 .interruptible = true,
7231         };
7232         int err;
7233
7234         do {
7235                 err = _nfs4_proc_setlk(state, cmd, request);
7236                 if (err == -NFS4ERR_DENIED)
7237                         err = -EAGAIN;
7238                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7239                                 err, &exception);
7240         } while (exception.retry);
7241         return err;
7242 }
7243
7244 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7245 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7246
7247 static int
7248 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7249                         struct file_lock *request)
7250 {
7251         int             status = -ERESTARTSYS;
7252         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7253
7254         while(!signalled()) {
7255                 status = nfs4_proc_setlk(state, cmd, request);
7256                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7257                         break;
7258                 freezable_schedule_timeout_interruptible(timeout);
7259                 timeout *= 2;
7260                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7261                 status = -ERESTARTSYS;
7262         }
7263         return status;
7264 }
7265
7266 #ifdef CONFIG_NFS_V4_1
7267 struct nfs4_lock_waiter {
7268         struct inode            *inode;
7269         struct nfs_lowner       owner;
7270         wait_queue_entry_t      wait;
7271 };
7272
7273 static int
7274 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7275 {
7276         struct nfs4_lock_waiter *waiter =
7277                 container_of(wait, struct nfs4_lock_waiter, wait);
7278
7279         /* NULL key means to wake up everyone */
7280         if (key) {
7281                 struct cb_notify_lock_args      *cbnl = key;
7282                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7283                                                 *wowner = &waiter->owner;
7284
7285                 /* Only wake if the callback was for the same owner. */
7286                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7287                         return 0;
7288
7289                 /* Make sure it's for the right inode */
7290                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7291                         return 0;
7292         }
7293
7294         return woken_wake_function(wait, mode, flags, key);
7295 }
7296
7297 static int
7298 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7299 {
7300         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7301         struct nfs_server *server = NFS_SERVER(state->inode);
7302         struct nfs_client *clp = server->nfs_client;
7303         wait_queue_head_t *q = &clp->cl_lock_waitq;
7304         struct nfs4_lock_waiter waiter = {
7305                 .inode = state->inode,
7306                 .owner = { .clientid = clp->cl_clientid,
7307                            .id = lsp->ls_seqid.owner_id,
7308                            .s_dev = server->s_dev },
7309         };
7310         int status;
7311
7312         /* Don't bother with waitqueue if we don't expect a callback */
7313         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7314                 return nfs4_retry_setlk_simple(state, cmd, request);
7315
7316         init_wait(&waiter.wait);
7317         waiter.wait.func = nfs4_wake_lock_waiter;
7318         add_wait_queue(q, &waiter.wait);
7319
7320         do {
7321                 status = nfs4_proc_setlk(state, cmd, request);
7322                 if (status != -EAGAIN || IS_SETLK(cmd))
7323                         break;
7324
7325                 status = -ERESTARTSYS;
7326                 freezer_do_not_count();
7327                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7328                            NFS4_LOCK_MAXTIMEOUT);
7329                 freezer_count();
7330         } while (!signalled());
7331
7332         remove_wait_queue(q, &waiter.wait);
7333
7334         return status;
7335 }
7336 #else /* !CONFIG_NFS_V4_1 */
7337 static inline int
7338 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7339 {
7340         return nfs4_retry_setlk_simple(state, cmd, request);
7341 }
7342 #endif
7343
7344 static int
7345 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7346 {
7347         struct nfs_open_context *ctx;
7348         struct nfs4_state *state;
7349         int status;
7350
7351         /* verify open state */
7352         ctx = nfs_file_open_context(filp);
7353         state = ctx->state;
7354
7355         if (IS_GETLK(cmd)) {
7356                 if (state != NULL)
7357                         return nfs4_proc_getlk(state, F_GETLK, request);
7358                 return 0;
7359         }
7360
7361         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7362                 return -EINVAL;
7363
7364         if (request->fl_type == F_UNLCK) {
7365                 if (state != NULL)
7366                         return nfs4_proc_unlck(state, cmd, request);
7367                 return 0;
7368         }
7369
7370         if (state == NULL)
7371                 return -ENOLCK;
7372
7373         if ((request->fl_flags & FL_POSIX) &&
7374             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7375                 return -ENOLCK;
7376
7377         /*
7378          * Don't rely on the VFS having checked the file open mode,
7379          * since it won't do this for flock() locks.
7380          */
7381         switch (request->fl_type) {
7382         case F_RDLCK:
7383                 if (!(filp->f_mode & FMODE_READ))
7384                         return -EBADF;
7385                 break;
7386         case F_WRLCK:
7387                 if (!(filp->f_mode & FMODE_WRITE))
7388                         return -EBADF;
7389         }
7390
7391         status = nfs4_set_lock_state(state, request);
7392         if (status != 0)
7393                 return status;
7394
7395         return nfs4_retry_setlk(state, cmd, request);
7396 }
7397
7398 static int nfs4_delete_lease(struct file *file, void **priv)
7399 {
7400         return generic_setlease(file, F_UNLCK, NULL, priv);
7401 }
7402
7403 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7404                           void **priv)
7405 {
7406         struct inode *inode = file_inode(file);
7407         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7408         int ret;
7409
7410         /* No delegation, no lease */
7411         if (!nfs4_have_delegation(inode, type))
7412                 return -EAGAIN;
7413         ret = generic_setlease(file, arg, lease, priv);
7414         if (ret || nfs4_have_delegation(inode, type))
7415                 return ret;
7416         /* We raced with a delegation return */
7417         nfs4_delete_lease(file, priv);
7418         return -EAGAIN;
7419 }
7420
7421 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7422                        void **priv)
7423 {
7424         switch (arg) {
7425         case F_RDLCK:
7426         case F_WRLCK:
7427                 return nfs4_add_lease(file, arg, lease, priv);
7428         case F_UNLCK:
7429                 return nfs4_delete_lease(file, priv);
7430         default:
7431                 return -EINVAL;
7432         }
7433 }
7434
7435 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7436 {
7437         struct nfs_server *server = NFS_SERVER(state->inode);
7438         int err;
7439
7440         err = nfs4_set_lock_state(state, fl);
7441         if (err != 0)
7442                 return err;
7443         do {
7444                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7445                 if (err != -NFS4ERR_DELAY)
7446                         break;
7447                 ssleep(1);
7448         } while (err == -NFS4ERR_DELAY);
7449         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7450 }
7451
7452 struct nfs_release_lockowner_data {
7453         struct nfs4_lock_state *lsp;
7454         struct nfs_server *server;
7455         struct nfs_release_lockowner_args args;
7456         struct nfs_release_lockowner_res res;
7457         unsigned long timestamp;
7458 };
7459
7460 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7461 {
7462         struct nfs_release_lockowner_data *data = calldata;
7463         struct nfs_server *server = data->server;
7464         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7465                            &data->res.seq_res, task);
7466         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7467         data->timestamp = jiffies;
7468 }
7469
7470 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7471 {
7472         struct nfs_release_lockowner_data *data = calldata;
7473         struct nfs_server *server = data->server;
7474
7475         nfs40_sequence_done(task, &data->res.seq_res);
7476
7477         switch (task->tk_status) {
7478         case 0:
7479                 renew_lease(server, data->timestamp);
7480                 break;
7481         case -NFS4ERR_STALE_CLIENTID:
7482         case -NFS4ERR_EXPIRED:
7483                 nfs4_schedule_lease_recovery(server->nfs_client);
7484                 break;
7485         case -NFS4ERR_LEASE_MOVED:
7486         case -NFS4ERR_DELAY:
7487                 if (nfs4_async_handle_error(task, server,
7488                                             NULL, NULL) == -EAGAIN)
7489                         rpc_restart_call_prepare(task);
7490         }
7491 }
7492
7493 static void nfs4_release_lockowner_release(void *calldata)
7494 {
7495         struct nfs_release_lockowner_data *data = calldata;
7496         nfs4_free_lock_state(data->server, data->lsp);
7497         kfree(calldata);
7498 }
7499
7500 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7501         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7502         .rpc_call_done = nfs4_release_lockowner_done,
7503         .rpc_release = nfs4_release_lockowner_release,
7504 };
7505
7506 static void
7507 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7508 {
7509         struct nfs_release_lockowner_data *data;
7510         struct rpc_message msg = {
7511                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7512         };
7513
7514         if (server->nfs_client->cl_mvops->minor_version != 0)
7515                 return;
7516
7517         data = kmalloc(sizeof(*data), GFP_NOFS);
7518         if (!data)
7519                 return;
7520         data->lsp = lsp;
7521         data->server = server;
7522         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7523         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7524         data->args.lock_owner.s_dev = server->s_dev;
7525
7526         msg.rpc_argp = &data->args;
7527         msg.rpc_resp = &data->res;
7528         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7529         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7530 }
7531
7532 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7533
7534 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7535                                    struct user_namespace *mnt_userns,
7536                                    struct dentry *unused, struct inode *inode,
7537                                    const char *key, const void *buf,
7538                                    size_t buflen, int flags)
7539 {
7540         return nfs4_proc_set_acl(inode, buf, buflen);
7541 }
7542
7543 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7544                                    struct dentry *unused, struct inode *inode,
7545                                    const char *key, void *buf, size_t buflen)
7546 {
7547         return nfs4_proc_get_acl(inode, buf, buflen);
7548 }
7549
7550 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7551 {
7552         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7553 }
7554
7555 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7556
7557 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7558                                      struct user_namespace *mnt_userns,
7559                                      struct dentry *unused, struct inode *inode,
7560                                      const char *key, const void *buf,
7561                                      size_t buflen, int flags)
7562 {
7563         if (security_ismaclabel(key))
7564                 return nfs4_set_security_label(inode, buf, buflen);
7565
7566         return -EOPNOTSUPP;
7567 }
7568
7569 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7570                                      struct dentry *unused, struct inode *inode,
7571                                      const char *key, void *buf, size_t buflen)
7572 {
7573         if (security_ismaclabel(key))
7574                 return nfs4_get_security_label(inode, buf, buflen);
7575         return -EOPNOTSUPP;
7576 }
7577
7578 static ssize_t
7579 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7580 {
7581         int len = 0;
7582
7583         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7584                 len = security_inode_listsecurity(inode, list, list_len);
7585                 if (len >= 0 && list_len && len > list_len)
7586                         return -ERANGE;
7587         }
7588         return len;
7589 }
7590
7591 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7592         .prefix = XATTR_SECURITY_PREFIX,
7593         .get    = nfs4_xattr_get_nfs4_label,
7594         .set    = nfs4_xattr_set_nfs4_label,
7595 };
7596
7597 #else
7598
7599 static ssize_t
7600 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7601 {
7602         return 0;
7603 }
7604
7605 #endif
7606
7607 #ifdef CONFIG_NFS_V4_2
7608 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7609                                     struct user_namespace *mnt_userns,
7610                                     struct dentry *unused, struct inode *inode,
7611                                     const char *key, const void *buf,
7612                                     size_t buflen, int flags)
7613 {
7614         struct nfs_access_entry cache;
7615         int ret;
7616
7617         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7618                 return -EOPNOTSUPP;
7619
7620         /*
7621          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7622          * flags right now. Handling of xattr operations use the normal
7623          * file read/write permissions.
7624          *
7625          * Just in case the server has other ideas (which RFC 8276 allows),
7626          * do a cached access check for the XA* flags to possibly avoid
7627          * doing an RPC and getting EACCES back.
7628          */
7629         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7630                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7631                         return -EACCES;
7632         }
7633
7634         if (buf == NULL) {
7635                 ret = nfs42_proc_removexattr(inode, key);
7636                 if (!ret)
7637                         nfs4_xattr_cache_remove(inode, key);
7638         } else {
7639                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7640                 if (!ret)
7641                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7642         }
7643
7644         return ret;
7645 }
7646
7647 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7648                                     struct dentry *unused, struct inode *inode,
7649                                     const char *key, void *buf, size_t buflen)
7650 {
7651         struct nfs_access_entry cache;
7652         ssize_t ret;
7653
7654         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7655                 return -EOPNOTSUPP;
7656
7657         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7658                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7659                         return -EACCES;
7660         }
7661
7662         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7663         if (ret)
7664                 return ret;
7665
7666         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7667         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7668                 return ret;
7669
7670         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7671
7672         return ret;
7673 }
7674
7675 static ssize_t
7676 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7677 {
7678         u64 cookie;
7679         bool eof;
7680         ssize_t ret, size;
7681         char *buf;
7682         size_t buflen;
7683         struct nfs_access_entry cache;
7684
7685         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7686                 return 0;
7687
7688         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7689                 if (!(cache.mask & NFS_ACCESS_XALIST))
7690                         return 0;
7691         }
7692
7693         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7694         if (ret)
7695                 return ret;
7696
7697         ret = nfs4_xattr_cache_list(inode, list, list_len);
7698         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7699                 return ret;
7700
7701         cookie = 0;
7702         eof = false;
7703         buflen = list_len ? list_len : XATTR_LIST_MAX;
7704         buf = list_len ? list : NULL;
7705         size = 0;
7706
7707         while (!eof) {
7708                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7709                     &cookie, &eof);
7710                 if (ret < 0)
7711                         return ret;
7712
7713                 if (list_len) {
7714                         buf += ret;
7715                         buflen -= ret;
7716                 }
7717                 size += ret;
7718         }
7719
7720         if (list_len)
7721                 nfs4_xattr_cache_set_list(inode, list, size);
7722
7723         return size;
7724 }
7725
7726 #else
7727
7728 static ssize_t
7729 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7730 {
7731         return 0;
7732 }
7733 #endif /* CONFIG_NFS_V4_2 */
7734
7735 /*
7736  * nfs_fhget will use either the mounted_on_fileid or the fileid
7737  */
7738 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7739 {
7740         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7741                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7742               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7743               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7744                 return;
7745
7746         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7747                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7748         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7749         fattr->nlink = 2;
7750 }
7751
7752 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7753                                    const struct qstr *name,
7754                                    struct nfs4_fs_locations *fs_locations,
7755                                    struct page *page)
7756 {
7757         struct nfs_server *server = NFS_SERVER(dir);
7758         u32 bitmask[3];
7759         struct nfs4_fs_locations_arg args = {
7760                 .dir_fh = NFS_FH(dir),
7761                 .name = name,
7762                 .page = page,
7763                 .bitmask = bitmask,
7764         };
7765         struct nfs4_fs_locations_res res = {
7766                 .fs_locations = fs_locations,
7767         };
7768         struct rpc_message msg = {
7769                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7770                 .rpc_argp = &args,
7771                 .rpc_resp = &res,
7772         };
7773         int status;
7774
7775         dprintk("%s: start\n", __func__);
7776
7777         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7778         bitmask[1] = nfs4_fattr_bitmap[1];
7779
7780         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7781          * is not supported */
7782         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7783                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7784         else
7785                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7786
7787         nfs_fattr_init(&fs_locations->fattr);
7788         fs_locations->server = server;
7789         fs_locations->nlocations = 0;
7790         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7791         dprintk("%s: returned status = %d\n", __func__, status);
7792         return status;
7793 }
7794
7795 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7796                            const struct qstr *name,
7797                            struct nfs4_fs_locations *fs_locations,
7798                            struct page *page)
7799 {
7800         struct nfs4_exception exception = {
7801                 .interruptible = true,
7802         };
7803         int err;
7804         do {
7805                 err = _nfs4_proc_fs_locations(client, dir, name,
7806                                 fs_locations, page);
7807                 trace_nfs4_get_fs_locations(dir, name, err);
7808                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7809                                 &exception);
7810         } while (exception.retry);
7811         return err;
7812 }
7813
7814 /*
7815  * This operation also signals the server that this client is
7816  * performing migration recovery.  The server can stop returning
7817  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7818  * appended to this compound to identify the client ID which is
7819  * performing recovery.
7820  */
7821 static int _nfs40_proc_get_locations(struct inode *inode,
7822                                      struct nfs4_fs_locations *locations,
7823                                      struct page *page, const struct cred *cred)
7824 {
7825         struct nfs_server *server = NFS_SERVER(inode);
7826         struct rpc_clnt *clnt = server->client;
7827         u32 bitmask[2] = {
7828                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7829         };
7830         struct nfs4_fs_locations_arg args = {
7831                 .clientid       = server->nfs_client->cl_clientid,
7832                 .fh             = NFS_FH(inode),
7833                 .page           = page,
7834                 .bitmask        = bitmask,
7835                 .migration      = 1,            /* skip LOOKUP */
7836                 .renew          = 1,            /* append RENEW */
7837         };
7838         struct nfs4_fs_locations_res res = {
7839                 .fs_locations   = locations,
7840                 .migration      = 1,
7841                 .renew          = 1,
7842         };
7843         struct rpc_message msg = {
7844                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7845                 .rpc_argp       = &args,
7846                 .rpc_resp       = &res,
7847                 .rpc_cred       = cred,
7848         };
7849         unsigned long now = jiffies;
7850         int status;
7851
7852         nfs_fattr_init(&locations->fattr);
7853         locations->server = server;
7854         locations->nlocations = 0;
7855
7856         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7857         status = nfs4_call_sync_sequence(clnt, server, &msg,
7858                                         &args.seq_args, &res.seq_res);
7859         if (status)
7860                 return status;
7861
7862         renew_lease(server, now);
7863         return 0;
7864 }
7865
7866 #ifdef CONFIG_NFS_V4_1
7867
7868 /*
7869  * This operation also signals the server that this client is
7870  * performing migration recovery.  The server can stop asserting
7871  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7872  * performing this operation is identified in the SEQUENCE
7873  * operation in this compound.
7874  *
7875  * When the client supports GETATTR(fs_locations_info), it can
7876  * be plumbed in here.
7877  */
7878 static int _nfs41_proc_get_locations(struct inode *inode,
7879                                      struct nfs4_fs_locations *locations,
7880                                      struct page *page, const struct cred *cred)
7881 {
7882         struct nfs_server *server = NFS_SERVER(inode);
7883         struct rpc_clnt *clnt = server->client;
7884         u32 bitmask[2] = {
7885                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7886         };
7887         struct nfs4_fs_locations_arg args = {
7888                 .fh             = NFS_FH(inode),
7889                 .page           = page,
7890                 .bitmask        = bitmask,
7891                 .migration      = 1,            /* skip LOOKUP */
7892         };
7893         struct nfs4_fs_locations_res res = {
7894                 .fs_locations   = locations,
7895                 .migration      = 1,
7896         };
7897         struct rpc_message msg = {
7898                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7899                 .rpc_argp       = &args,
7900                 .rpc_resp       = &res,
7901                 .rpc_cred       = cred,
7902         };
7903         int status;
7904
7905         nfs_fattr_init(&locations->fattr);
7906         locations->server = server;
7907         locations->nlocations = 0;
7908
7909         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7910         status = nfs4_call_sync_sequence(clnt, server, &msg,
7911                                         &args.seq_args, &res.seq_res);
7912         if (status == NFS4_OK &&
7913             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7914                 status = -NFS4ERR_LEASE_MOVED;
7915         return status;
7916 }
7917
7918 #endif  /* CONFIG_NFS_V4_1 */
7919
7920 /**
7921  * nfs4_proc_get_locations - discover locations for a migrated FSID
7922  * @inode: inode on FSID that is migrating
7923  * @locations: result of query
7924  * @page: buffer
7925  * @cred: credential to use for this operation
7926  *
7927  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7928  * operation failed, or a negative errno if a local error occurred.
7929  *
7930  * On success, "locations" is filled in, but if the server has
7931  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7932  * asserted.
7933  *
7934  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7935  * from this client that require migration recovery.
7936  */
7937 int nfs4_proc_get_locations(struct inode *inode,
7938                             struct nfs4_fs_locations *locations,
7939                             struct page *page, const struct cred *cred)
7940 {
7941         struct nfs_server *server = NFS_SERVER(inode);
7942         struct nfs_client *clp = server->nfs_client;
7943         const struct nfs4_mig_recovery_ops *ops =
7944                                         clp->cl_mvops->mig_recovery_ops;
7945         struct nfs4_exception exception = {
7946                 .interruptible = true,
7947         };
7948         int status;
7949
7950         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7951                 (unsigned long long)server->fsid.major,
7952                 (unsigned long long)server->fsid.minor,
7953                 clp->cl_hostname);
7954         nfs_display_fhandle(NFS_FH(inode), __func__);
7955
7956         do {
7957                 status = ops->get_locations(inode, locations, page, cred);
7958                 if (status != -NFS4ERR_DELAY)
7959                         break;
7960                 nfs4_handle_exception(server, status, &exception);
7961         } while (exception.retry);
7962         return status;
7963 }
7964
7965 /*
7966  * This operation also signals the server that this client is
7967  * performing "lease moved" recovery.  The server can stop
7968  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7969  * is appended to this compound to identify the client ID which is
7970  * performing recovery.
7971  */
7972 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7973 {
7974         struct nfs_server *server = NFS_SERVER(inode);
7975         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7976         struct rpc_clnt *clnt = server->client;
7977         struct nfs4_fsid_present_arg args = {
7978                 .fh             = NFS_FH(inode),
7979                 .clientid       = clp->cl_clientid,
7980                 .renew          = 1,            /* append RENEW */
7981         };
7982         struct nfs4_fsid_present_res res = {
7983                 .renew          = 1,
7984         };
7985         struct rpc_message msg = {
7986                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7987                 .rpc_argp       = &args,
7988                 .rpc_resp       = &res,
7989                 .rpc_cred       = cred,
7990         };
7991         unsigned long now = jiffies;
7992         int status;
7993
7994         res.fh = nfs_alloc_fhandle();
7995         if (res.fh == NULL)
7996                 return -ENOMEM;
7997
7998         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7999         status = nfs4_call_sync_sequence(clnt, server, &msg,
8000                                                 &args.seq_args, &res.seq_res);
8001         nfs_free_fhandle(res.fh);
8002         if (status)
8003                 return status;
8004
8005         do_renew_lease(clp, now);
8006         return 0;
8007 }
8008
8009 #ifdef CONFIG_NFS_V4_1
8010
8011 /*
8012  * This operation also signals the server that this client is
8013  * performing "lease moved" recovery.  The server can stop asserting
8014  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8015  * this operation is identified in the SEQUENCE operation in this
8016  * compound.
8017  */
8018 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8019 {
8020         struct nfs_server *server = NFS_SERVER(inode);
8021         struct rpc_clnt *clnt = server->client;
8022         struct nfs4_fsid_present_arg args = {
8023                 .fh             = NFS_FH(inode),
8024         };
8025         struct nfs4_fsid_present_res res = {
8026         };
8027         struct rpc_message msg = {
8028                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8029                 .rpc_argp       = &args,
8030                 .rpc_resp       = &res,
8031                 .rpc_cred       = cred,
8032         };
8033         int status;
8034
8035         res.fh = nfs_alloc_fhandle();
8036         if (res.fh == NULL)
8037                 return -ENOMEM;
8038
8039         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8040         status = nfs4_call_sync_sequence(clnt, server, &msg,
8041                                                 &args.seq_args, &res.seq_res);
8042         nfs_free_fhandle(res.fh);
8043         if (status == NFS4_OK &&
8044             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8045                 status = -NFS4ERR_LEASE_MOVED;
8046         return status;
8047 }
8048
8049 #endif  /* CONFIG_NFS_V4_1 */
8050
8051 /**
8052  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8053  * @inode: inode on FSID to check
8054  * @cred: credential to use for this operation
8055  *
8056  * Server indicates whether the FSID is present, moved, or not
8057  * recognized.  This operation is necessary to clear a LEASE_MOVED
8058  * condition for this client ID.
8059  *
8060  * Returns NFS4_OK if the FSID is present on this server,
8061  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8062  *  NFS4ERR code if some error occurred on the server, or a
8063  *  negative errno if a local failure occurred.
8064  */
8065 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8066 {
8067         struct nfs_server *server = NFS_SERVER(inode);
8068         struct nfs_client *clp = server->nfs_client;
8069         const struct nfs4_mig_recovery_ops *ops =
8070                                         clp->cl_mvops->mig_recovery_ops;
8071         struct nfs4_exception exception = {
8072                 .interruptible = true,
8073         };
8074         int status;
8075
8076         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8077                 (unsigned long long)server->fsid.major,
8078                 (unsigned long long)server->fsid.minor,
8079                 clp->cl_hostname);
8080         nfs_display_fhandle(NFS_FH(inode), __func__);
8081
8082         do {
8083                 status = ops->fsid_present(inode, cred);
8084                 if (status != -NFS4ERR_DELAY)
8085                         break;
8086                 nfs4_handle_exception(server, status, &exception);
8087         } while (exception.retry);
8088         return status;
8089 }
8090
8091 /*
8092  * If 'use_integrity' is true and the state managment nfs_client
8093  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8094  * and the machine credential as per RFC3530bis and RFC5661 Security
8095  * Considerations sections. Otherwise, just use the user cred with the
8096  * filesystem's rpc_client.
8097  */
8098 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8099 {
8100         int status;
8101         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8102         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8103         struct nfs4_secinfo_arg args = {
8104                 .dir_fh = NFS_FH(dir),
8105                 .name   = name,
8106         };
8107         struct nfs4_secinfo_res res = {
8108                 .flavors     = flavors,
8109         };
8110         struct rpc_message msg = {
8111                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8112                 .rpc_argp = &args,
8113                 .rpc_resp = &res,
8114         };
8115         struct nfs4_call_sync_data data = {
8116                 .seq_server = NFS_SERVER(dir),
8117                 .seq_args = &args.seq_args,
8118                 .seq_res = &res.seq_res,
8119         };
8120         struct rpc_task_setup task_setup = {
8121                 .rpc_client = clnt,
8122                 .rpc_message = &msg,
8123                 .callback_ops = clp->cl_mvops->call_sync_ops,
8124                 .callback_data = &data,
8125                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8126         };
8127         const struct cred *cred = NULL;
8128
8129         if (use_integrity) {
8130                 clnt = clp->cl_rpcclient;
8131                 task_setup.rpc_client = clnt;
8132
8133                 cred = nfs4_get_clid_cred(clp);
8134                 msg.rpc_cred = cred;
8135         }
8136
8137         dprintk("NFS call  secinfo %s\n", name->name);
8138
8139         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8140         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8141         status = nfs4_call_sync_custom(&task_setup);
8142
8143         dprintk("NFS reply  secinfo: %d\n", status);
8144
8145         put_cred(cred);
8146         return status;
8147 }
8148
8149 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8150                       struct nfs4_secinfo_flavors *flavors)
8151 {
8152         struct nfs4_exception exception = {
8153                 .interruptible = true,
8154         };
8155         int err;
8156         do {
8157                 err = -NFS4ERR_WRONGSEC;
8158
8159                 /* try to use integrity protection with machine cred */
8160                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8161                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8162
8163                 /*
8164                  * if unable to use integrity protection, or SECINFO with
8165                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8166                  * disallowed by spec, but exists in deployed servers) use
8167                  * the current filesystem's rpc_client and the user cred.
8168                  */
8169                 if (err == -NFS4ERR_WRONGSEC)
8170                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8171
8172                 trace_nfs4_secinfo(dir, name, err);
8173                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8174                                 &exception);
8175         } while (exception.retry);
8176         return err;
8177 }
8178
8179 #ifdef CONFIG_NFS_V4_1
8180 /*
8181  * Check the exchange flags returned by the server for invalid flags, having
8182  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8183  * DS flags set.
8184  */
8185 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8186 {
8187         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8188                 goto out_inval;
8189         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8190                 goto out_inval;
8191         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8192             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8193                 goto out_inval;
8194         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8195                 goto out_inval;
8196         return NFS_OK;
8197 out_inval:
8198         return -NFS4ERR_INVAL;
8199 }
8200
8201 static bool
8202 nfs41_same_server_scope(struct nfs41_server_scope *a,
8203                         struct nfs41_server_scope *b)
8204 {
8205         if (a->server_scope_sz != b->server_scope_sz)
8206                 return false;
8207         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8208 }
8209
8210 static void
8211 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8212 {
8213         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8214         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8215         struct nfs_client *clp = args->client;
8216
8217         switch (task->tk_status) {
8218         case -NFS4ERR_BADSESSION:
8219         case -NFS4ERR_DEADSESSION:
8220                 nfs4_schedule_session_recovery(clp->cl_session,
8221                                 task->tk_status);
8222         }
8223         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8224                         res->dir != NFS4_CDFS4_BOTH) {
8225                 rpc_task_close_connection(task);
8226                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8227                         rpc_restart_call(task);
8228         }
8229 }
8230
8231 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8232         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8233 };
8234
8235 /*
8236  * nfs4_proc_bind_one_conn_to_session()
8237  *
8238  * The 4.1 client currently uses the same TCP connection for the
8239  * fore and backchannel.
8240  */
8241 static
8242 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8243                 struct rpc_xprt *xprt,
8244                 struct nfs_client *clp,
8245                 const struct cred *cred)
8246 {
8247         int status;
8248         struct nfs41_bind_conn_to_session_args args = {
8249                 .client = clp,
8250                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8251                 .retries = 0,
8252         };
8253         struct nfs41_bind_conn_to_session_res res;
8254         struct rpc_message msg = {
8255                 .rpc_proc =
8256                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8257                 .rpc_argp = &args,
8258                 .rpc_resp = &res,
8259                 .rpc_cred = cred,
8260         };
8261         struct rpc_task_setup task_setup_data = {
8262                 .rpc_client = clnt,
8263                 .rpc_xprt = xprt,
8264                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8265                 .rpc_message = &msg,
8266                 .flags = RPC_TASK_TIMEOUT,
8267         };
8268         struct rpc_task *task;
8269
8270         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8271         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8272                 args.dir = NFS4_CDFC4_FORE;
8273
8274         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8275         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8276                 args.dir = NFS4_CDFC4_FORE;
8277
8278         task = rpc_run_task(&task_setup_data);
8279         if (!IS_ERR(task)) {
8280                 status = task->tk_status;
8281                 rpc_put_task(task);
8282         } else
8283                 status = PTR_ERR(task);
8284         trace_nfs4_bind_conn_to_session(clp, status);
8285         if (status == 0) {
8286                 if (memcmp(res.sessionid.data,
8287                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8288                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8289                         return -EIO;
8290                 }
8291                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8292                         dprintk("NFS: %s: Unexpected direction from server\n",
8293                                 __func__);
8294                         return -EIO;
8295                 }
8296                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8297                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8298                                 __func__);
8299                         return -EIO;
8300                 }
8301         }
8302
8303         return status;
8304 }
8305
8306 struct rpc_bind_conn_calldata {
8307         struct nfs_client *clp;
8308         const struct cred *cred;
8309 };
8310
8311 static int
8312 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8313                 struct rpc_xprt *xprt,
8314                 void *calldata)
8315 {
8316         struct rpc_bind_conn_calldata *p = calldata;
8317
8318         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8319 }
8320
8321 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8322 {
8323         struct rpc_bind_conn_calldata data = {
8324                 .clp = clp,
8325                 .cred = cred,
8326         };
8327         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8328                         nfs4_proc_bind_conn_to_session_callback, &data);
8329 }
8330
8331 /*
8332  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8333  * and operations we'd like to see to enable certain features in the allow map
8334  */
8335 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8336         .how = SP4_MACH_CRED,
8337         .enforce.u.words = {
8338                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8339                       1 << (OP_EXCHANGE_ID - 32) |
8340                       1 << (OP_CREATE_SESSION - 32) |
8341                       1 << (OP_DESTROY_SESSION - 32) |
8342                       1 << (OP_DESTROY_CLIENTID - 32)
8343         },
8344         .allow.u.words = {
8345                 [0] = 1 << (OP_CLOSE) |
8346                       1 << (OP_OPEN_DOWNGRADE) |
8347                       1 << (OP_LOCKU) |
8348                       1 << (OP_DELEGRETURN) |
8349                       1 << (OP_COMMIT),
8350                 [1] = 1 << (OP_SECINFO - 32) |
8351                       1 << (OP_SECINFO_NO_NAME - 32) |
8352                       1 << (OP_LAYOUTRETURN - 32) |
8353                       1 << (OP_TEST_STATEID - 32) |
8354                       1 << (OP_FREE_STATEID - 32) |
8355                       1 << (OP_WRITE - 32)
8356         }
8357 };
8358
8359 /*
8360  * Select the state protection mode for client `clp' given the server results
8361  * from exchange_id in `sp'.
8362  *
8363  * Returns 0 on success, negative errno otherwise.
8364  */
8365 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8366                                  struct nfs41_state_protection *sp)
8367 {
8368         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8369                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8370                       1 << (OP_EXCHANGE_ID - 32) |
8371                       1 << (OP_CREATE_SESSION - 32) |
8372                       1 << (OP_DESTROY_SESSION - 32) |
8373                       1 << (OP_DESTROY_CLIENTID - 32)
8374         };
8375         unsigned long flags = 0;
8376         unsigned int i;
8377         int ret = 0;
8378
8379         if (sp->how == SP4_MACH_CRED) {
8380                 /* Print state protect result */
8381                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8382                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8383                         if (test_bit(i, sp->enforce.u.longs))
8384                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8385                         if (test_bit(i, sp->allow.u.longs))
8386                                 dfprintk(MOUNT, "  allow op %d\n", i);
8387                 }
8388
8389                 /* make sure nothing is on enforce list that isn't supported */
8390                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8391                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8392                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8393                                 ret = -EINVAL;
8394                                 goto out;
8395                         }
8396                 }
8397
8398                 /*
8399                  * Minimal mode - state operations are allowed to use machine
8400                  * credential.  Note this already happens by default, so the
8401                  * client doesn't have to do anything more than the negotiation.
8402                  *
8403                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8404                  *       we're already using the machine cred for exchange_id
8405                  *       and will never use a different cred.
8406                  */
8407                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8408                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8409                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8410                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8411                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8412                         dfprintk(MOUNT, "  minimal mode enabled\n");
8413                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8414                 } else {
8415                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8416                         ret = -EINVAL;
8417                         goto out;
8418                 }
8419
8420                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8421                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8422                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8423                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8424                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8425                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8426                 }
8427
8428                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8429                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8430                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8431                 }
8432
8433                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8434                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8435                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8436                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8437                 }
8438
8439                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8440                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8441                         dfprintk(MOUNT, "  stateid mode enabled\n");
8442                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8443                 }
8444
8445                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8446                         dfprintk(MOUNT, "  write mode enabled\n");
8447                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8448                 }
8449
8450                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8451                         dfprintk(MOUNT, "  commit mode enabled\n");
8452                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8453                 }
8454         }
8455 out:
8456         clp->cl_sp4_flags = flags;
8457         return ret;
8458 }
8459
8460 struct nfs41_exchange_id_data {
8461         struct nfs41_exchange_id_res res;
8462         struct nfs41_exchange_id_args args;
8463 };
8464
8465 static void nfs4_exchange_id_release(void *data)
8466 {
8467         struct nfs41_exchange_id_data *cdata =
8468                                         (struct nfs41_exchange_id_data *)data;
8469
8470         nfs_put_client(cdata->args.client);
8471         kfree(cdata->res.impl_id);
8472         kfree(cdata->res.server_scope);
8473         kfree(cdata->res.server_owner);
8474         kfree(cdata);
8475 }
8476
8477 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8478         .rpc_release = nfs4_exchange_id_release,
8479 };
8480
8481 /*
8482  * _nfs4_proc_exchange_id()
8483  *
8484  * Wrapper for EXCHANGE_ID operation.
8485  */
8486 static struct rpc_task *
8487 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8488                         u32 sp4_how, struct rpc_xprt *xprt)
8489 {
8490         struct rpc_message msg = {
8491                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8492                 .rpc_cred = cred,
8493         };
8494         struct rpc_task_setup task_setup_data = {
8495                 .rpc_client = clp->cl_rpcclient,
8496                 .callback_ops = &nfs4_exchange_id_call_ops,
8497                 .rpc_message = &msg,
8498                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8499         };
8500         struct nfs41_exchange_id_data *calldata;
8501         int status;
8502
8503         if (!refcount_inc_not_zero(&clp->cl_count))
8504                 return ERR_PTR(-EIO);
8505
8506         status = -ENOMEM;
8507         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8508         if (!calldata)
8509                 goto out;
8510
8511         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8512
8513         status = nfs4_init_uniform_client_string(clp);
8514         if (status)
8515                 goto out_calldata;
8516
8517         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8518                                                 GFP_NOFS);
8519         status = -ENOMEM;
8520         if (unlikely(calldata->res.server_owner == NULL))
8521                 goto out_calldata;
8522
8523         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8524                                         GFP_NOFS);
8525         if (unlikely(calldata->res.server_scope == NULL))
8526                 goto out_server_owner;
8527
8528         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8529         if (unlikely(calldata->res.impl_id == NULL))
8530                 goto out_server_scope;
8531
8532         switch (sp4_how) {
8533         case SP4_NONE:
8534                 calldata->args.state_protect.how = SP4_NONE;
8535                 break;
8536
8537         case SP4_MACH_CRED:
8538                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8539                 break;
8540
8541         default:
8542                 /* unsupported! */
8543                 WARN_ON_ONCE(1);
8544                 status = -EINVAL;
8545                 goto out_impl_id;
8546         }
8547         if (xprt) {
8548                 task_setup_data.rpc_xprt = xprt;
8549                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8550                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8551                                 sizeof(calldata->args.verifier.data));
8552         }
8553         calldata->args.client = clp;
8554         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8555         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8556 #ifdef CONFIG_NFS_V4_1_MIGRATION
8557         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8558 #endif
8559         msg.rpc_argp = &calldata->args;
8560         msg.rpc_resp = &calldata->res;
8561         task_setup_data.callback_data = calldata;
8562
8563         return rpc_run_task(&task_setup_data);
8564
8565 out_impl_id:
8566         kfree(calldata->res.impl_id);
8567 out_server_scope:
8568         kfree(calldata->res.server_scope);
8569 out_server_owner:
8570         kfree(calldata->res.server_owner);
8571 out_calldata:
8572         kfree(calldata);
8573 out:
8574         nfs_put_client(clp);
8575         return ERR_PTR(status);
8576 }
8577
8578 /*
8579  * _nfs4_proc_exchange_id()
8580  *
8581  * Wrapper for EXCHANGE_ID operation.
8582  */
8583 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8584                         u32 sp4_how)
8585 {
8586         struct rpc_task *task;
8587         struct nfs41_exchange_id_args *argp;
8588         struct nfs41_exchange_id_res *resp;
8589         unsigned long now = jiffies;
8590         int status;
8591
8592         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8593         if (IS_ERR(task))
8594                 return PTR_ERR(task);
8595
8596         argp = task->tk_msg.rpc_argp;
8597         resp = task->tk_msg.rpc_resp;
8598         status = task->tk_status;
8599         if (status  != 0)
8600                 goto out;
8601
8602         status = nfs4_check_cl_exchange_flags(resp->flags,
8603                         clp->cl_mvops->minor_version);
8604         if (status  != 0)
8605                 goto out;
8606
8607         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8608         if (status != 0)
8609                 goto out;
8610
8611         do_renew_lease(clp, now);
8612
8613         clp->cl_clientid = resp->clientid;
8614         clp->cl_exchange_flags = resp->flags;
8615         clp->cl_seqid = resp->seqid;
8616         /* Client ID is not confirmed */
8617         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8618                 clear_bit(NFS4_SESSION_ESTABLISHED,
8619                           &clp->cl_session->session_state);
8620
8621         if (clp->cl_serverscope != NULL &&
8622             !nfs41_same_server_scope(clp->cl_serverscope,
8623                                 resp->server_scope)) {
8624                 dprintk("%s: server_scope mismatch detected\n",
8625                         __func__);
8626                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8627         }
8628
8629         swap(clp->cl_serverowner, resp->server_owner);
8630         swap(clp->cl_serverscope, resp->server_scope);
8631         swap(clp->cl_implid, resp->impl_id);
8632
8633         /* Save the EXCHANGE_ID verifier session trunk tests */
8634         memcpy(clp->cl_confirm.data, argp->verifier.data,
8635                sizeof(clp->cl_confirm.data));
8636 out:
8637         trace_nfs4_exchange_id(clp, status);
8638         rpc_put_task(task);
8639         return status;
8640 }
8641
8642 /*
8643  * nfs4_proc_exchange_id()
8644  *
8645  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8646  *
8647  * Since the clientid has expired, all compounds using sessions
8648  * associated with the stale clientid will be returning
8649  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8650  * be in some phase of session reset.
8651  *
8652  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8653  */
8654 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8655 {
8656         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8657         int status;
8658
8659         /* try SP4_MACH_CRED if krb5i/p */
8660         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8661             authflavor == RPC_AUTH_GSS_KRB5P) {
8662                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8663                 if (!status)
8664                         return 0;
8665         }
8666
8667         /* try SP4_NONE */
8668         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8669 }
8670
8671 /**
8672  * nfs4_test_session_trunk
8673  *
8674  * This is an add_xprt_test() test function called from
8675  * rpc_clnt_setup_test_and_add_xprt.
8676  *
8677  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8678  * and is dereferrenced in nfs4_exchange_id_release
8679  *
8680  * Upon success, add the new transport to the rpc_clnt
8681  *
8682  * @clnt: struct rpc_clnt to get new transport
8683  * @xprt: the rpc_xprt to test
8684  * @data: call data for _nfs4_proc_exchange_id.
8685  */
8686 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8687                             void *data)
8688 {
8689         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8690         struct rpc_task *task;
8691         int status;
8692
8693         u32 sp4_how;
8694
8695         dprintk("--> %s try %s\n", __func__,
8696                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8697
8698         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8699
8700         /* Test connection for session trunking. Async exchange_id call */
8701         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8702         if (IS_ERR(task))
8703                 return;
8704
8705         status = task->tk_status;
8706         if (status == 0)
8707                 status = nfs4_detect_session_trunking(adata->clp,
8708                                 task->tk_msg.rpc_resp, xprt);
8709
8710         if (status == 0)
8711                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8712
8713         rpc_put_task(task);
8714 }
8715 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8716
8717 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8718                 const struct cred *cred)
8719 {
8720         struct rpc_message msg = {
8721                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8722                 .rpc_argp = clp,
8723                 .rpc_cred = cred,
8724         };
8725         int status;
8726
8727         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8728                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8729         trace_nfs4_destroy_clientid(clp, status);
8730         if (status)
8731                 dprintk("NFS: Got error %d from the server %s on "
8732                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8733         return status;
8734 }
8735
8736 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8737                 const struct cred *cred)
8738 {
8739         unsigned int loop;
8740         int ret;
8741
8742         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8743                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8744                 switch (ret) {
8745                 case -NFS4ERR_DELAY:
8746                 case -NFS4ERR_CLIENTID_BUSY:
8747                         ssleep(1);
8748                         break;
8749                 default:
8750                         return ret;
8751                 }
8752         }
8753         return 0;
8754 }
8755
8756 int nfs4_destroy_clientid(struct nfs_client *clp)
8757 {
8758         const struct cred *cred;
8759         int ret = 0;
8760
8761         if (clp->cl_mvops->minor_version < 1)
8762                 goto out;
8763         if (clp->cl_exchange_flags == 0)
8764                 goto out;
8765         if (clp->cl_preserve_clid)
8766                 goto out;
8767         cred = nfs4_get_clid_cred(clp);
8768         ret = nfs4_proc_destroy_clientid(clp, cred);
8769         put_cred(cred);
8770         switch (ret) {
8771         case 0:
8772         case -NFS4ERR_STALE_CLIENTID:
8773                 clp->cl_exchange_flags = 0;
8774         }
8775 out:
8776         return ret;
8777 }
8778
8779 #endif /* CONFIG_NFS_V4_1 */
8780
8781 struct nfs4_get_lease_time_data {
8782         struct nfs4_get_lease_time_args *args;
8783         struct nfs4_get_lease_time_res *res;
8784         struct nfs_client *clp;
8785 };
8786
8787 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8788                                         void *calldata)
8789 {
8790         struct nfs4_get_lease_time_data *data =
8791                         (struct nfs4_get_lease_time_data *)calldata;
8792
8793         /* just setup sequence, do not trigger session recovery
8794            since we're invoked within one */
8795         nfs4_setup_sequence(data->clp,
8796                         &data->args->la_seq_args,
8797                         &data->res->lr_seq_res,
8798                         task);
8799 }
8800
8801 /*
8802  * Called from nfs4_state_manager thread for session setup, so don't recover
8803  * from sequence operation or clientid errors.
8804  */
8805 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8806 {
8807         struct nfs4_get_lease_time_data *data =
8808                         (struct nfs4_get_lease_time_data *)calldata;
8809
8810         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8811                 return;
8812         switch (task->tk_status) {
8813         case -NFS4ERR_DELAY:
8814         case -NFS4ERR_GRACE:
8815                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8816                 task->tk_status = 0;
8817                 fallthrough;
8818         case -NFS4ERR_RETRY_UNCACHED_REP:
8819                 rpc_restart_call_prepare(task);
8820                 return;
8821         }
8822 }
8823
8824 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8825         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8826         .rpc_call_done = nfs4_get_lease_time_done,
8827 };
8828
8829 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8830 {
8831         struct nfs4_get_lease_time_args args;
8832         struct nfs4_get_lease_time_res res = {
8833                 .lr_fsinfo = fsinfo,
8834         };
8835         struct nfs4_get_lease_time_data data = {
8836                 .args = &args,
8837                 .res = &res,
8838                 .clp = clp,
8839         };
8840         struct rpc_message msg = {
8841                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8842                 .rpc_argp = &args,
8843                 .rpc_resp = &res,
8844         };
8845         struct rpc_task_setup task_setup = {
8846                 .rpc_client = clp->cl_rpcclient,
8847                 .rpc_message = &msg,
8848                 .callback_ops = &nfs4_get_lease_time_ops,
8849                 .callback_data = &data,
8850                 .flags = RPC_TASK_TIMEOUT,
8851         };
8852
8853         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8854         return nfs4_call_sync_custom(&task_setup);
8855 }
8856
8857 #ifdef CONFIG_NFS_V4_1
8858
8859 /*
8860  * Initialize the values to be used by the client in CREATE_SESSION
8861  * If nfs4_init_session set the fore channel request and response sizes,
8862  * use them.
8863  *
8864  * Set the back channel max_resp_sz_cached to zero to force the client to
8865  * always set csa_cachethis to FALSE because the current implementation
8866  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8867  */
8868 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8869                                     struct rpc_clnt *clnt)
8870 {
8871         unsigned int max_rqst_sz, max_resp_sz;
8872         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8873         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8874
8875         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8876         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8877
8878         /* Fore channel attributes */
8879         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8880         args->fc_attrs.max_resp_sz = max_resp_sz;
8881         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8882         args->fc_attrs.max_reqs = max_session_slots;
8883
8884         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8885                 "max_ops=%u max_reqs=%u\n",
8886                 __func__,
8887                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8888                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8889
8890         /* Back channel attributes */
8891         args->bc_attrs.max_rqst_sz = max_bc_payload;
8892         args->bc_attrs.max_resp_sz = max_bc_payload;
8893         args->bc_attrs.max_resp_sz_cached = 0;
8894         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8895         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8896         if (args->bc_attrs.max_reqs > max_bc_slots)
8897                 args->bc_attrs.max_reqs = max_bc_slots;
8898
8899         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8900                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8901                 __func__,
8902                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8903                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8904                 args->bc_attrs.max_reqs);
8905 }
8906
8907 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8908                 struct nfs41_create_session_res *res)
8909 {
8910         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8911         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8912
8913         if (rcvd->max_resp_sz > sent->max_resp_sz)
8914                 return -EINVAL;
8915         /*
8916          * Our requested max_ops is the minimum we need; we're not
8917          * prepared to break up compounds into smaller pieces than that.
8918          * So, no point even trying to continue if the server won't
8919          * cooperate:
8920          */
8921         if (rcvd->max_ops < sent->max_ops)
8922                 return -EINVAL;
8923         if (rcvd->max_reqs == 0)
8924                 return -EINVAL;
8925         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8926                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8927         return 0;
8928 }
8929
8930 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8931                 struct nfs41_create_session_res *res)
8932 {
8933         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8934         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8935
8936         if (!(res->flags & SESSION4_BACK_CHAN))
8937                 goto out;
8938         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8939                 return -EINVAL;
8940         if (rcvd->max_resp_sz < sent->max_resp_sz)
8941                 return -EINVAL;
8942         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8943                 return -EINVAL;
8944         if (rcvd->max_ops > sent->max_ops)
8945                 return -EINVAL;
8946         if (rcvd->max_reqs > sent->max_reqs)
8947                 return -EINVAL;
8948 out:
8949         return 0;
8950 }
8951
8952 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8953                                      struct nfs41_create_session_res *res)
8954 {
8955         int ret;
8956
8957         ret = nfs4_verify_fore_channel_attrs(args, res);
8958         if (ret)
8959                 return ret;
8960         return nfs4_verify_back_channel_attrs(args, res);
8961 }
8962
8963 static void nfs4_update_session(struct nfs4_session *session,
8964                 struct nfs41_create_session_res *res)
8965 {
8966         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8967         /* Mark client id and session as being confirmed */
8968         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8969         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8970         session->flags = res->flags;
8971         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8972         if (res->flags & SESSION4_BACK_CHAN)
8973                 memcpy(&session->bc_attrs, &res->bc_attrs,
8974                                 sizeof(session->bc_attrs));
8975 }
8976
8977 static int _nfs4_proc_create_session(struct nfs_client *clp,
8978                 const struct cred *cred)
8979 {
8980         struct nfs4_session *session = clp->cl_session;
8981         struct nfs41_create_session_args args = {
8982                 .client = clp,
8983                 .clientid = clp->cl_clientid,
8984                 .seqid = clp->cl_seqid,
8985                 .cb_program = NFS4_CALLBACK,
8986         };
8987         struct nfs41_create_session_res res;
8988
8989         struct rpc_message msg = {
8990                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8991                 .rpc_argp = &args,
8992                 .rpc_resp = &res,
8993                 .rpc_cred = cred,
8994         };
8995         int status;
8996
8997         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8998         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8999
9000         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9001                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9002         trace_nfs4_create_session(clp, status);
9003
9004         switch (status) {
9005         case -NFS4ERR_STALE_CLIENTID:
9006         case -NFS4ERR_DELAY:
9007         case -ETIMEDOUT:
9008         case -EACCES:
9009         case -EAGAIN:
9010                 goto out;
9011         }
9012
9013         clp->cl_seqid++;
9014         if (!status) {
9015                 /* Verify the session's negotiated channel_attrs values */
9016                 status = nfs4_verify_channel_attrs(&args, &res);
9017                 /* Increment the clientid slot sequence id */
9018                 if (status)
9019                         goto out;
9020                 nfs4_update_session(session, &res);
9021         }
9022 out:
9023         return status;
9024 }
9025
9026 /*
9027  * Issues a CREATE_SESSION operation to the server.
9028  * It is the responsibility of the caller to verify the session is
9029  * expired before calling this routine.
9030  */
9031 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9032 {
9033         int status;
9034         unsigned *ptr;
9035         struct nfs4_session *session = clp->cl_session;
9036
9037         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9038
9039         status = _nfs4_proc_create_session(clp, cred);
9040         if (status)
9041                 goto out;
9042
9043         /* Init or reset the session slot tables */
9044         status = nfs4_setup_session_slot_tables(session);
9045         dprintk("slot table setup returned %d\n", status);
9046         if (status)
9047                 goto out;
9048
9049         ptr = (unsigned *)&session->sess_id.data[0];
9050         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9051                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9052 out:
9053         return status;
9054 }
9055
9056 /*
9057  * Issue the over-the-wire RPC DESTROY_SESSION.
9058  * The caller must serialize access to this routine.
9059  */
9060 int nfs4_proc_destroy_session(struct nfs4_session *session,
9061                 const struct cred *cred)
9062 {
9063         struct rpc_message msg = {
9064                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9065                 .rpc_argp = session,
9066                 .rpc_cred = cred,
9067         };
9068         int status = 0;
9069
9070         /* session is still being setup */
9071         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9072                 return 0;
9073
9074         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9075                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9076         trace_nfs4_destroy_session(session->clp, status);
9077
9078         if (status)
9079                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9080                         "Session has been destroyed regardless...\n", status);
9081         return status;
9082 }
9083
9084 /*
9085  * Renew the cl_session lease.
9086  */
9087 struct nfs4_sequence_data {
9088         struct nfs_client *clp;
9089         struct nfs4_sequence_args args;
9090         struct nfs4_sequence_res res;
9091 };
9092
9093 static void nfs41_sequence_release(void *data)
9094 {
9095         struct nfs4_sequence_data *calldata = data;
9096         struct nfs_client *clp = calldata->clp;
9097
9098         if (refcount_read(&clp->cl_count) > 1)
9099                 nfs4_schedule_state_renewal(clp);
9100         nfs_put_client(clp);
9101         kfree(calldata);
9102 }
9103
9104 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9105 {
9106         switch(task->tk_status) {
9107         case -NFS4ERR_DELAY:
9108                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9109                 return -EAGAIN;
9110         default:
9111                 nfs4_schedule_lease_recovery(clp);
9112         }
9113         return 0;
9114 }
9115
9116 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9117 {
9118         struct nfs4_sequence_data *calldata = data;
9119         struct nfs_client *clp = calldata->clp;
9120
9121         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9122                 return;
9123
9124         trace_nfs4_sequence(clp, task->tk_status);
9125         if (task->tk_status < 0) {
9126                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9127                 if (refcount_read(&clp->cl_count) == 1)
9128                         return;
9129
9130                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9131                         rpc_restart_call_prepare(task);
9132                         return;
9133                 }
9134         }
9135         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9136 }
9137
9138 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9139 {
9140         struct nfs4_sequence_data *calldata = data;
9141         struct nfs_client *clp = calldata->clp;
9142         struct nfs4_sequence_args *args;
9143         struct nfs4_sequence_res *res;
9144
9145         args = task->tk_msg.rpc_argp;
9146         res = task->tk_msg.rpc_resp;
9147
9148         nfs4_setup_sequence(clp, args, res, task);
9149 }
9150
9151 static const struct rpc_call_ops nfs41_sequence_ops = {
9152         .rpc_call_done = nfs41_sequence_call_done,
9153         .rpc_call_prepare = nfs41_sequence_prepare,
9154         .rpc_release = nfs41_sequence_release,
9155 };
9156
9157 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9158                 const struct cred *cred,
9159                 struct nfs4_slot *slot,
9160                 bool is_privileged)
9161 {
9162         struct nfs4_sequence_data *calldata;
9163         struct rpc_message msg = {
9164                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9165                 .rpc_cred = cred,
9166         };
9167         struct rpc_task_setup task_setup_data = {
9168                 .rpc_client = clp->cl_rpcclient,
9169                 .rpc_message = &msg,
9170                 .callback_ops = &nfs41_sequence_ops,
9171                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9172         };
9173         struct rpc_task *ret;
9174
9175         ret = ERR_PTR(-EIO);
9176         if (!refcount_inc_not_zero(&clp->cl_count))
9177                 goto out_err;
9178
9179         ret = ERR_PTR(-ENOMEM);
9180         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9181         if (calldata == NULL)
9182                 goto out_put_clp;
9183         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9184         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9185         msg.rpc_argp = &calldata->args;
9186         msg.rpc_resp = &calldata->res;
9187         calldata->clp = clp;
9188         task_setup_data.callback_data = calldata;
9189
9190         ret = rpc_run_task(&task_setup_data);
9191         if (IS_ERR(ret))
9192                 goto out_err;
9193         return ret;
9194 out_put_clp:
9195         nfs_put_client(clp);
9196 out_err:
9197         nfs41_release_slot(slot);
9198         return ret;
9199 }
9200
9201 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9202 {
9203         struct rpc_task *task;
9204         int ret = 0;
9205
9206         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9207                 return -EAGAIN;
9208         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9209         if (IS_ERR(task))
9210                 ret = PTR_ERR(task);
9211         else
9212                 rpc_put_task_async(task);
9213         dprintk("<-- %s status=%d\n", __func__, ret);
9214         return ret;
9215 }
9216
9217 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9218 {
9219         struct rpc_task *task;
9220         int ret;
9221
9222         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9223         if (IS_ERR(task)) {
9224                 ret = PTR_ERR(task);
9225                 goto out;
9226         }
9227         ret = rpc_wait_for_completion_task(task);
9228         if (!ret)
9229                 ret = task->tk_status;
9230         rpc_put_task(task);
9231 out:
9232         dprintk("<-- %s status=%d\n", __func__, ret);
9233         return ret;
9234 }
9235
9236 struct nfs4_reclaim_complete_data {
9237         struct nfs_client *clp;
9238         struct nfs41_reclaim_complete_args arg;
9239         struct nfs41_reclaim_complete_res res;
9240 };
9241
9242 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9243 {
9244         struct nfs4_reclaim_complete_data *calldata = data;
9245
9246         nfs4_setup_sequence(calldata->clp,
9247                         &calldata->arg.seq_args,
9248                         &calldata->res.seq_res,
9249                         task);
9250 }
9251
9252 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9253 {
9254         switch(task->tk_status) {
9255         case 0:
9256                 wake_up_all(&clp->cl_lock_waitq);
9257                 fallthrough;
9258         case -NFS4ERR_COMPLETE_ALREADY:
9259         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9260                 break;
9261         case -NFS4ERR_DELAY:
9262                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9263                 fallthrough;
9264         case -NFS4ERR_RETRY_UNCACHED_REP:
9265                 return -EAGAIN;
9266         case -NFS4ERR_BADSESSION:
9267         case -NFS4ERR_DEADSESSION:
9268         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9269                 break;
9270         default:
9271                 nfs4_schedule_lease_recovery(clp);
9272         }
9273         return 0;
9274 }
9275
9276 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9277 {
9278         struct nfs4_reclaim_complete_data *calldata = data;
9279         struct nfs_client *clp = calldata->clp;
9280         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9281
9282         if (!nfs41_sequence_done(task, res))
9283                 return;
9284
9285         trace_nfs4_reclaim_complete(clp, task->tk_status);
9286         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9287                 rpc_restart_call_prepare(task);
9288                 return;
9289         }
9290 }
9291
9292 static void nfs4_free_reclaim_complete_data(void *data)
9293 {
9294         struct nfs4_reclaim_complete_data *calldata = data;
9295
9296         kfree(calldata);
9297 }
9298
9299 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9300         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9301         .rpc_call_done = nfs4_reclaim_complete_done,
9302         .rpc_release = nfs4_free_reclaim_complete_data,
9303 };
9304
9305 /*
9306  * Issue a global reclaim complete.
9307  */
9308 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9309                 const struct cred *cred)
9310 {
9311         struct nfs4_reclaim_complete_data *calldata;
9312         struct rpc_message msg = {
9313                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9314                 .rpc_cred = cred,
9315         };
9316         struct rpc_task_setup task_setup_data = {
9317                 .rpc_client = clp->cl_rpcclient,
9318                 .rpc_message = &msg,
9319                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9320                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9321         };
9322         int status = -ENOMEM;
9323
9324         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9325         if (calldata == NULL)
9326                 goto out;
9327         calldata->clp = clp;
9328         calldata->arg.one_fs = 0;
9329
9330         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9331         msg.rpc_argp = &calldata->arg;
9332         msg.rpc_resp = &calldata->res;
9333         task_setup_data.callback_data = calldata;
9334         status = nfs4_call_sync_custom(&task_setup_data);
9335 out:
9336         dprintk("<-- %s status=%d\n", __func__, status);
9337         return status;
9338 }
9339
9340 static void
9341 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9342 {
9343         struct nfs4_layoutget *lgp = calldata;
9344         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9345
9346         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9347                                 &lgp->res.seq_res, task);
9348 }
9349
9350 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9351 {
9352         struct nfs4_layoutget *lgp = calldata;
9353
9354         nfs41_sequence_process(task, &lgp->res.seq_res);
9355 }
9356
9357 static int
9358 nfs4_layoutget_handle_exception(struct rpc_task *task,
9359                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9360 {
9361         struct inode *inode = lgp->args.inode;
9362         struct nfs_server *server = NFS_SERVER(inode);
9363         struct pnfs_layout_hdr *lo = lgp->lo;
9364         int nfs4err = task->tk_status;
9365         int err, status = 0;
9366         LIST_HEAD(head);
9367
9368         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9369
9370         nfs4_sequence_free_slot(&lgp->res.seq_res);
9371
9372         switch (nfs4err) {
9373         case 0:
9374                 goto out;
9375
9376         /*
9377          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9378          * on the file. set tk_status to -ENODATA to tell upper layer to
9379          * retry go inband.
9380          */
9381         case -NFS4ERR_LAYOUTUNAVAILABLE:
9382                 status = -ENODATA;
9383                 goto out;
9384         /*
9385          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9386          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9387          */
9388         case -NFS4ERR_BADLAYOUT:
9389                 status = -EOVERFLOW;
9390                 goto out;
9391         /*
9392          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9393          * (or clients) writing to the same RAID stripe except when
9394          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9395          *
9396          * Treat it like we would RECALLCONFLICT -- we retry for a little
9397          * while, and then eventually give up.
9398          */
9399         case -NFS4ERR_LAYOUTTRYLATER:
9400                 if (lgp->args.minlength == 0) {
9401                         status = -EOVERFLOW;
9402                         goto out;
9403                 }
9404                 status = -EBUSY;
9405                 break;
9406         case -NFS4ERR_RECALLCONFLICT:
9407                 status = -ERECALLCONFLICT;
9408                 break;
9409         case -NFS4ERR_DELEG_REVOKED:
9410         case -NFS4ERR_ADMIN_REVOKED:
9411         case -NFS4ERR_EXPIRED:
9412         case -NFS4ERR_BAD_STATEID:
9413                 exception->timeout = 0;
9414                 spin_lock(&inode->i_lock);
9415                 /* If the open stateid was bad, then recover it. */
9416                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9417                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9418                         spin_unlock(&inode->i_lock);
9419                         exception->state = lgp->args.ctx->state;
9420                         exception->stateid = &lgp->args.stateid;
9421                         break;
9422                 }
9423
9424                 /*
9425                  * Mark the bad layout state as invalid, then retry
9426                  */
9427                 pnfs_mark_layout_stateid_invalid(lo, &head);
9428                 spin_unlock(&inode->i_lock);
9429                 nfs_commit_inode(inode, 0);
9430                 pnfs_free_lseg_list(&head);
9431                 status = -EAGAIN;
9432                 goto out;
9433         }
9434
9435         err = nfs4_handle_exception(server, nfs4err, exception);
9436         if (!status) {
9437                 if (exception->retry)
9438                         status = -EAGAIN;
9439                 else
9440                         status = err;
9441         }
9442 out:
9443         return status;
9444 }
9445
9446 size_t max_response_pages(struct nfs_server *server)
9447 {
9448         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9449         return nfs_page_array_len(0, max_resp_sz);
9450 }
9451
9452 static void nfs4_layoutget_release(void *calldata)
9453 {
9454         struct nfs4_layoutget *lgp = calldata;
9455
9456         nfs4_sequence_free_slot(&lgp->res.seq_res);
9457         pnfs_layoutget_free(lgp);
9458 }
9459
9460 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9461         .rpc_call_prepare = nfs4_layoutget_prepare,
9462         .rpc_call_done = nfs4_layoutget_done,
9463         .rpc_release = nfs4_layoutget_release,
9464 };
9465
9466 struct pnfs_layout_segment *
9467 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9468 {
9469         struct inode *inode = lgp->args.inode;
9470         struct nfs_server *server = NFS_SERVER(inode);
9471         struct rpc_task *task;
9472         struct rpc_message msg = {
9473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9474                 .rpc_argp = &lgp->args,
9475                 .rpc_resp = &lgp->res,
9476                 .rpc_cred = lgp->cred,
9477         };
9478         struct rpc_task_setup task_setup_data = {
9479                 .rpc_client = server->client,
9480                 .rpc_message = &msg,
9481                 .callback_ops = &nfs4_layoutget_call_ops,
9482                 .callback_data = lgp,
9483                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9484                          RPC_TASK_MOVEABLE,
9485         };
9486         struct pnfs_layout_segment *lseg = NULL;
9487         struct nfs4_exception exception = {
9488                 .inode = inode,
9489                 .timeout = *timeout,
9490         };
9491         int status = 0;
9492
9493         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9494
9495         task = rpc_run_task(&task_setup_data);
9496
9497         status = rpc_wait_for_completion_task(task);
9498         if (status != 0)
9499                 goto out;
9500
9501         if (task->tk_status < 0) {
9502                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9503                 *timeout = exception.timeout;
9504         } else if (lgp->res.layoutp->len == 0) {
9505                 status = -EAGAIN;
9506                 *timeout = nfs4_update_delay(&exception.timeout);
9507         } else
9508                 lseg = pnfs_layout_process(lgp);
9509 out:
9510         trace_nfs4_layoutget(lgp->args.ctx,
9511                         &lgp->args.range,
9512                         &lgp->res.range,
9513                         &lgp->res.stateid,
9514                         status);
9515
9516         rpc_put_task(task);
9517         dprintk("<-- %s status=%d\n", __func__, status);
9518         if (status)
9519                 return ERR_PTR(status);
9520         return lseg;
9521 }
9522
9523 static void
9524 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9525 {
9526         struct nfs4_layoutreturn *lrp = calldata;
9527
9528         nfs4_setup_sequence(lrp->clp,
9529                         &lrp->args.seq_args,
9530                         &lrp->res.seq_res,
9531                         task);
9532         if (!pnfs_layout_is_valid(lrp->args.layout))
9533                 rpc_exit(task, 0);
9534 }
9535
9536 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9537 {
9538         struct nfs4_layoutreturn *lrp = calldata;
9539         struct nfs_server *server;
9540
9541         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9542                 return;
9543
9544         /*
9545          * Was there an RPC level error? Assume the call succeeded,
9546          * and that we need to release the layout
9547          */
9548         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9549                 lrp->res.lrs_present = 0;
9550                 return;
9551         }
9552
9553         server = NFS_SERVER(lrp->args.inode);
9554         switch (task->tk_status) {
9555         case -NFS4ERR_OLD_STATEID:
9556                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9557                                         &lrp->args.range,
9558                                         lrp->args.inode))
9559                         goto out_restart;
9560                 fallthrough;
9561         default:
9562                 task->tk_status = 0;
9563                 fallthrough;
9564         case 0:
9565                 break;
9566         case -NFS4ERR_DELAY:
9567                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9568                         break;
9569                 goto out_restart;
9570         }
9571         return;
9572 out_restart:
9573         task->tk_status = 0;
9574         nfs4_sequence_free_slot(&lrp->res.seq_res);
9575         rpc_restart_call_prepare(task);
9576 }
9577
9578 static void nfs4_layoutreturn_release(void *calldata)
9579 {
9580         struct nfs4_layoutreturn *lrp = calldata;
9581         struct pnfs_layout_hdr *lo = lrp->args.layout;
9582
9583         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9584                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9585         nfs4_sequence_free_slot(&lrp->res.seq_res);
9586         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9587                 lrp->ld_private.ops->free(&lrp->ld_private);
9588         pnfs_put_layout_hdr(lrp->args.layout);
9589         nfs_iput_and_deactive(lrp->inode);
9590         put_cred(lrp->cred);
9591         kfree(calldata);
9592 }
9593
9594 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9595         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9596         .rpc_call_done = nfs4_layoutreturn_done,
9597         .rpc_release = nfs4_layoutreturn_release,
9598 };
9599
9600 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9601 {
9602         struct rpc_task *task;
9603         struct rpc_message msg = {
9604                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9605                 .rpc_argp = &lrp->args,
9606                 .rpc_resp = &lrp->res,
9607                 .rpc_cred = lrp->cred,
9608         };
9609         struct rpc_task_setup task_setup_data = {
9610                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9611                 .rpc_message = &msg,
9612                 .callback_ops = &nfs4_layoutreturn_call_ops,
9613                 .callback_data = lrp,
9614                 .flags = RPC_TASK_MOVEABLE,
9615         };
9616         int status = 0;
9617
9618         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9619                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9620                         &task_setup_data.rpc_client, &msg);
9621
9622         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9623         if (!sync) {
9624                 if (!lrp->inode) {
9625                         nfs4_layoutreturn_release(lrp);
9626                         return -EAGAIN;
9627                 }
9628                 task_setup_data.flags |= RPC_TASK_ASYNC;
9629         }
9630         if (!lrp->inode)
9631                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9632                                    1);
9633         else
9634                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9635                                    0);
9636         task = rpc_run_task(&task_setup_data);
9637         if (IS_ERR(task))
9638                 return PTR_ERR(task);
9639         if (sync)
9640                 status = task->tk_status;
9641         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9642         dprintk("<-- %s status=%d\n", __func__, status);
9643         rpc_put_task(task);
9644         return status;
9645 }
9646
9647 static int
9648 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9649                 struct pnfs_device *pdev,
9650                 const struct cred *cred)
9651 {
9652         struct nfs4_getdeviceinfo_args args = {
9653                 .pdev = pdev,
9654                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9655                         NOTIFY_DEVICEID4_DELETE,
9656         };
9657         struct nfs4_getdeviceinfo_res res = {
9658                 .pdev = pdev,
9659         };
9660         struct rpc_message msg = {
9661                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9662                 .rpc_argp = &args,
9663                 .rpc_resp = &res,
9664                 .rpc_cred = cred,
9665         };
9666         int status;
9667
9668         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9669         if (res.notification & ~args.notify_types)
9670                 dprintk("%s: unsupported notification\n", __func__);
9671         if (res.notification != args.notify_types)
9672                 pdev->nocache = 1;
9673
9674         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9675
9676         dprintk("<-- %s status=%d\n", __func__, status);
9677
9678         return status;
9679 }
9680
9681 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9682                 struct pnfs_device *pdev,
9683                 const struct cred *cred)
9684 {
9685         struct nfs4_exception exception = { };
9686         int err;
9687
9688         do {
9689                 err = nfs4_handle_exception(server,
9690                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9691                                         &exception);
9692         } while (exception.retry);
9693         return err;
9694 }
9695 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9696
9697 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9698 {
9699         struct nfs4_layoutcommit_data *data = calldata;
9700         struct nfs_server *server = NFS_SERVER(data->args.inode);
9701
9702         nfs4_setup_sequence(server->nfs_client,
9703                         &data->args.seq_args,
9704                         &data->res.seq_res,
9705                         task);
9706 }
9707
9708 static void
9709 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9710 {
9711         struct nfs4_layoutcommit_data *data = calldata;
9712         struct nfs_server *server = NFS_SERVER(data->args.inode);
9713
9714         if (!nfs41_sequence_done(task, &data->res.seq_res))
9715                 return;
9716
9717         switch (task->tk_status) { /* Just ignore these failures */
9718         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9719         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9720         case -NFS4ERR_BADLAYOUT:     /* no layout */
9721         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9722                 task->tk_status = 0;
9723                 break;
9724         case 0:
9725                 break;
9726         default:
9727                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9728                         rpc_restart_call_prepare(task);
9729                         return;
9730                 }
9731         }
9732 }
9733
9734 static void nfs4_layoutcommit_release(void *calldata)
9735 {
9736         struct nfs4_layoutcommit_data *data = calldata;
9737
9738         pnfs_cleanup_layoutcommit(data);
9739         nfs_post_op_update_inode_force_wcc(data->args.inode,
9740                                            data->res.fattr);
9741         put_cred(data->cred);
9742         nfs_iput_and_deactive(data->inode);
9743         kfree(data);
9744 }
9745
9746 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9747         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9748         .rpc_call_done = nfs4_layoutcommit_done,
9749         .rpc_release = nfs4_layoutcommit_release,
9750 };
9751
9752 int
9753 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9754 {
9755         struct rpc_message msg = {
9756                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9757                 .rpc_argp = &data->args,
9758                 .rpc_resp = &data->res,
9759                 .rpc_cred = data->cred,
9760         };
9761         struct rpc_task_setup task_setup_data = {
9762                 .task = &data->task,
9763                 .rpc_client = NFS_CLIENT(data->args.inode),
9764                 .rpc_message = &msg,
9765                 .callback_ops = &nfs4_layoutcommit_ops,
9766                 .callback_data = data,
9767                 .flags = RPC_TASK_MOVEABLE,
9768         };
9769         struct rpc_task *task;
9770         int status = 0;
9771
9772         dprintk("NFS: initiating layoutcommit call. sync %d "
9773                 "lbw: %llu inode %lu\n", sync,
9774                 data->args.lastbytewritten,
9775                 data->args.inode->i_ino);
9776
9777         if (!sync) {
9778                 data->inode = nfs_igrab_and_active(data->args.inode);
9779                 if (data->inode == NULL) {
9780                         nfs4_layoutcommit_release(data);
9781                         return -EAGAIN;
9782                 }
9783                 task_setup_data.flags = RPC_TASK_ASYNC;
9784         }
9785         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9786         task = rpc_run_task(&task_setup_data);
9787         if (IS_ERR(task))
9788                 return PTR_ERR(task);
9789         if (sync)
9790                 status = task->tk_status;
9791         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9792         dprintk("%s: status %d\n", __func__, status);
9793         rpc_put_task(task);
9794         return status;
9795 }
9796
9797 /*
9798  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9799  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9800  */
9801 static int
9802 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9803                     struct nfs_fsinfo *info,
9804                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9805 {
9806         struct nfs41_secinfo_no_name_args args = {
9807                 .style = SECINFO_STYLE_CURRENT_FH,
9808         };
9809         struct nfs4_secinfo_res res = {
9810                 .flavors = flavors,
9811         };
9812         struct rpc_message msg = {
9813                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9814                 .rpc_argp = &args,
9815                 .rpc_resp = &res,
9816         };
9817         struct nfs4_call_sync_data data = {
9818                 .seq_server = server,
9819                 .seq_args = &args.seq_args,
9820                 .seq_res = &res.seq_res,
9821         };
9822         struct rpc_task_setup task_setup = {
9823                 .rpc_client = server->client,
9824                 .rpc_message = &msg,
9825                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9826                 .callback_data = &data,
9827                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9828         };
9829         const struct cred *cred = NULL;
9830         int status;
9831
9832         if (use_integrity) {
9833                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9834
9835                 cred = nfs4_get_clid_cred(server->nfs_client);
9836                 msg.rpc_cred = cred;
9837         }
9838
9839         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9840         status = nfs4_call_sync_custom(&task_setup);
9841         dprintk("<-- %s status=%d\n", __func__, status);
9842
9843         put_cred(cred);
9844
9845         return status;
9846 }
9847
9848 static int
9849 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9850                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9851 {
9852         struct nfs4_exception exception = {
9853                 .interruptible = true,
9854         };
9855         int err;
9856         do {
9857                 /* first try using integrity protection */
9858                 err = -NFS4ERR_WRONGSEC;
9859
9860                 /* try to use integrity protection with machine cred */
9861                 if (_nfs4_is_integrity_protected(server->nfs_client))
9862                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9863                                                           flavors, true);
9864
9865                 /*
9866                  * if unable to use integrity protection, or SECINFO with
9867                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9868                  * disallowed by spec, but exists in deployed servers) use
9869                  * the current filesystem's rpc_client and the user cred.
9870                  */
9871                 if (err == -NFS4ERR_WRONGSEC)
9872                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9873                                                           flavors, false);
9874
9875                 switch (err) {
9876                 case 0:
9877                 case -NFS4ERR_WRONGSEC:
9878                 case -ENOTSUPP:
9879                         goto out;
9880                 default:
9881                         err = nfs4_handle_exception(server, err, &exception);
9882                 }
9883         } while (exception.retry);
9884 out:
9885         return err;
9886 }
9887
9888 static int
9889 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9890                     struct nfs_fsinfo *info)
9891 {
9892         int err;
9893         struct page *page;
9894         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9895         struct nfs4_secinfo_flavors *flavors;
9896         struct nfs4_secinfo4 *secinfo;
9897         int i;
9898
9899         page = alloc_page(GFP_KERNEL);
9900         if (!page) {
9901                 err = -ENOMEM;
9902                 goto out;
9903         }
9904
9905         flavors = page_address(page);
9906         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9907
9908         /*
9909          * Fall back on "guess and check" method if
9910          * the server doesn't support SECINFO_NO_NAME
9911          */
9912         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9913                 err = nfs4_find_root_sec(server, fhandle, info);
9914                 goto out_freepage;
9915         }
9916         if (err)
9917                 goto out_freepage;
9918
9919         for (i = 0; i < flavors->num_flavors; i++) {
9920                 secinfo = &flavors->flavors[i];
9921
9922                 switch (secinfo->flavor) {
9923                 case RPC_AUTH_NULL:
9924                 case RPC_AUTH_UNIX:
9925                 case RPC_AUTH_GSS:
9926                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9927                                         &secinfo->flavor_info);
9928                         break;
9929                 default:
9930                         flavor = RPC_AUTH_MAXFLAVOR;
9931                         break;
9932                 }
9933
9934                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9935                         flavor = RPC_AUTH_MAXFLAVOR;
9936
9937                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9938                         err = nfs4_lookup_root_sec(server, fhandle,
9939                                                    info, flavor);
9940                         if (!err)
9941                                 break;
9942                 }
9943         }
9944
9945         if (flavor == RPC_AUTH_MAXFLAVOR)
9946                 err = -EPERM;
9947
9948 out_freepage:
9949         put_page(page);
9950         if (err == -EACCES)
9951                 return -EPERM;
9952 out:
9953         return err;
9954 }
9955
9956 static int _nfs41_test_stateid(struct nfs_server *server,
9957                 nfs4_stateid *stateid,
9958                 const struct cred *cred)
9959 {
9960         int status;
9961         struct nfs41_test_stateid_args args = {
9962                 .stateid = stateid,
9963         };
9964         struct nfs41_test_stateid_res res;
9965         struct rpc_message msg = {
9966                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9967                 .rpc_argp = &args,
9968                 .rpc_resp = &res,
9969                 .rpc_cred = cred,
9970         };
9971         struct rpc_clnt *rpc_client = server->client;
9972
9973         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9974                 &rpc_client, &msg);
9975
9976         dprintk("NFS call  test_stateid %p\n", stateid);
9977         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9978         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9979                         &args.seq_args, &res.seq_res);
9980         if (status != NFS_OK) {
9981                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9982                 return status;
9983         }
9984         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9985         return -res.status;
9986 }
9987
9988 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9989                 int err, struct nfs4_exception *exception)
9990 {
9991         exception->retry = 0;
9992         switch(err) {
9993         case -NFS4ERR_DELAY:
9994         case -NFS4ERR_RETRY_UNCACHED_REP:
9995                 nfs4_handle_exception(server, err, exception);
9996                 break;
9997         case -NFS4ERR_BADSESSION:
9998         case -NFS4ERR_BADSLOT:
9999         case -NFS4ERR_BAD_HIGH_SLOT:
10000         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10001         case -NFS4ERR_DEADSESSION:
10002                 nfs4_do_handle_exception(server, err, exception);
10003         }
10004 }
10005
10006 /**
10007  * nfs41_test_stateid - perform a TEST_STATEID operation
10008  *
10009  * @server: server / transport on which to perform the operation
10010  * @stateid: state ID to test
10011  * @cred: credential
10012  *
10013  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10014  * Otherwise a negative NFS4ERR value is returned if the operation
10015  * failed or the state ID is not currently valid.
10016  */
10017 static int nfs41_test_stateid(struct nfs_server *server,
10018                 nfs4_stateid *stateid,
10019                 const struct cred *cred)
10020 {
10021         struct nfs4_exception exception = {
10022                 .interruptible = true,
10023         };
10024         int err;
10025         do {
10026                 err = _nfs41_test_stateid(server, stateid, cred);
10027                 nfs4_handle_delay_or_session_error(server, err, &exception);
10028         } while (exception.retry);
10029         return err;
10030 }
10031
10032 struct nfs_free_stateid_data {
10033         struct nfs_server *server;
10034         struct nfs41_free_stateid_args args;
10035         struct nfs41_free_stateid_res res;
10036 };
10037
10038 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10039 {
10040         struct nfs_free_stateid_data *data = calldata;
10041         nfs4_setup_sequence(data->server->nfs_client,
10042                         &data->args.seq_args,
10043                         &data->res.seq_res,
10044                         task);
10045 }
10046
10047 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10048 {
10049         struct nfs_free_stateid_data *data = calldata;
10050
10051         nfs41_sequence_done(task, &data->res.seq_res);
10052
10053         switch (task->tk_status) {
10054         case -NFS4ERR_DELAY:
10055                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10056                         rpc_restart_call_prepare(task);
10057         }
10058 }
10059
10060 static void nfs41_free_stateid_release(void *calldata)
10061 {
10062         struct nfs_free_stateid_data *data = calldata;
10063         struct nfs_client *clp = data->server->nfs_client;
10064
10065         nfs_put_client(clp);
10066         kfree(calldata);
10067 }
10068
10069 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10070         .rpc_call_prepare = nfs41_free_stateid_prepare,
10071         .rpc_call_done = nfs41_free_stateid_done,
10072         .rpc_release = nfs41_free_stateid_release,
10073 };
10074
10075 /**
10076  * nfs41_free_stateid - perform a FREE_STATEID operation
10077  *
10078  * @server: server / transport on which to perform the operation
10079  * @stateid: state ID to release
10080  * @cred: credential
10081  * @privileged: set to true if this call needs to be privileged
10082  *
10083  * Note: this function is always asynchronous.
10084  */
10085 static int nfs41_free_stateid(struct nfs_server *server,
10086                 const nfs4_stateid *stateid,
10087                 const struct cred *cred,
10088                 bool privileged)
10089 {
10090         struct rpc_message msg = {
10091                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10092                 .rpc_cred = cred,
10093         };
10094         struct rpc_task_setup task_setup = {
10095                 .rpc_client = server->client,
10096                 .rpc_message = &msg,
10097                 .callback_ops = &nfs41_free_stateid_ops,
10098                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10099         };
10100         struct nfs_free_stateid_data *data;
10101         struct rpc_task *task;
10102         struct nfs_client *clp = server->nfs_client;
10103
10104         if (!refcount_inc_not_zero(&clp->cl_count))
10105                 return -EIO;
10106
10107         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10108                 &task_setup.rpc_client, &msg);
10109
10110         dprintk("NFS call  free_stateid %p\n", stateid);
10111         data = kmalloc(sizeof(*data), GFP_NOFS);
10112         if (!data)
10113                 return -ENOMEM;
10114         data->server = server;
10115         nfs4_stateid_copy(&data->args.stateid, stateid);
10116
10117         task_setup.callback_data = data;
10118
10119         msg.rpc_argp = &data->args;
10120         msg.rpc_resp = &data->res;
10121         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10122         task = rpc_run_task(&task_setup);
10123         if (IS_ERR(task))
10124                 return PTR_ERR(task);
10125         rpc_put_task(task);
10126         return 0;
10127 }
10128
10129 static void
10130 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10131 {
10132         const struct cred *cred = lsp->ls_state->owner->so_cred;
10133
10134         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10135         nfs4_free_lock_state(server, lsp);
10136 }
10137
10138 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10139                 const nfs4_stateid *s2)
10140 {
10141         if (s1->type != s2->type)
10142                 return false;
10143
10144         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10145                 return false;
10146
10147         if (s1->seqid == s2->seqid)
10148                 return true;
10149
10150         return s1->seqid == 0 || s2->seqid == 0;
10151 }
10152
10153 #endif /* CONFIG_NFS_V4_1 */
10154
10155 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10156                 const nfs4_stateid *s2)
10157 {
10158         return nfs4_stateid_match(s1, s2);
10159 }
10160
10161
10162 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10163         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10164         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10165         .recover_open   = nfs4_open_reclaim,
10166         .recover_lock   = nfs4_lock_reclaim,
10167         .establish_clid = nfs4_init_clientid,
10168         .detect_trunking = nfs40_discover_server_trunking,
10169 };
10170
10171 #if defined(CONFIG_NFS_V4_1)
10172 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10173         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10174         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10175         .recover_open   = nfs4_open_reclaim,
10176         .recover_lock   = nfs4_lock_reclaim,
10177         .establish_clid = nfs41_init_clientid,
10178         .reclaim_complete = nfs41_proc_reclaim_complete,
10179         .detect_trunking = nfs41_discover_server_trunking,
10180 };
10181 #endif /* CONFIG_NFS_V4_1 */
10182
10183 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10184         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10185         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10186         .recover_open   = nfs40_open_expired,
10187         .recover_lock   = nfs4_lock_expired,
10188         .establish_clid = nfs4_init_clientid,
10189 };
10190
10191 #if defined(CONFIG_NFS_V4_1)
10192 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10193         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10194         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10195         .recover_open   = nfs41_open_expired,
10196         .recover_lock   = nfs41_lock_expired,
10197         .establish_clid = nfs41_init_clientid,
10198 };
10199 #endif /* CONFIG_NFS_V4_1 */
10200
10201 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10202         .sched_state_renewal = nfs4_proc_async_renew,
10203         .get_state_renewal_cred = nfs4_get_renew_cred,
10204         .renew_lease = nfs4_proc_renew,
10205 };
10206
10207 #if defined(CONFIG_NFS_V4_1)
10208 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10209         .sched_state_renewal = nfs41_proc_async_sequence,
10210         .get_state_renewal_cred = nfs4_get_machine_cred,
10211         .renew_lease = nfs4_proc_sequence,
10212 };
10213 #endif
10214
10215 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10216         .get_locations = _nfs40_proc_get_locations,
10217         .fsid_present = _nfs40_proc_fsid_present,
10218 };
10219
10220 #if defined(CONFIG_NFS_V4_1)
10221 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10222         .get_locations = _nfs41_proc_get_locations,
10223         .fsid_present = _nfs41_proc_fsid_present,
10224 };
10225 #endif  /* CONFIG_NFS_V4_1 */
10226
10227 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10228         .minor_version = 0,
10229         .init_caps = NFS_CAP_READDIRPLUS
10230                 | NFS_CAP_ATOMIC_OPEN
10231                 | NFS_CAP_POSIX_LOCK,
10232         .init_client = nfs40_init_client,
10233         .shutdown_client = nfs40_shutdown_client,
10234         .match_stateid = nfs4_match_stateid,
10235         .find_root_sec = nfs4_find_root_sec,
10236         .free_lock_state = nfs4_release_lockowner,
10237         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10238         .alloc_seqid = nfs_alloc_seqid,
10239         .call_sync_ops = &nfs40_call_sync_ops,
10240         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10241         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10242         .state_renewal_ops = &nfs40_state_renewal_ops,
10243         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10244 };
10245
10246 #if defined(CONFIG_NFS_V4_1)
10247 static struct nfs_seqid *
10248 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10249 {
10250         return NULL;
10251 }
10252
10253 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10254         .minor_version = 1,
10255         .init_caps = NFS_CAP_READDIRPLUS
10256                 | NFS_CAP_ATOMIC_OPEN
10257                 | NFS_CAP_POSIX_LOCK
10258                 | NFS_CAP_STATEID_NFSV41
10259                 | NFS_CAP_ATOMIC_OPEN_V1
10260                 | NFS_CAP_LGOPEN,
10261         .init_client = nfs41_init_client,
10262         .shutdown_client = nfs41_shutdown_client,
10263         .match_stateid = nfs41_match_stateid,
10264         .find_root_sec = nfs41_find_root_sec,
10265         .free_lock_state = nfs41_free_lock_state,
10266         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10267         .alloc_seqid = nfs_alloc_no_seqid,
10268         .session_trunk = nfs4_test_session_trunk,
10269         .call_sync_ops = &nfs41_call_sync_ops,
10270         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10271         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10272         .state_renewal_ops = &nfs41_state_renewal_ops,
10273         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10274 };
10275 #endif
10276
10277 #if defined(CONFIG_NFS_V4_2)
10278 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10279         .minor_version = 2,
10280         .init_caps = NFS_CAP_READDIRPLUS
10281                 | NFS_CAP_ATOMIC_OPEN
10282                 | NFS_CAP_POSIX_LOCK
10283                 | NFS_CAP_STATEID_NFSV41
10284                 | NFS_CAP_ATOMIC_OPEN_V1
10285                 | NFS_CAP_LGOPEN
10286                 | NFS_CAP_ALLOCATE
10287                 | NFS_CAP_COPY
10288                 | NFS_CAP_OFFLOAD_CANCEL
10289                 | NFS_CAP_COPY_NOTIFY
10290                 | NFS_CAP_DEALLOCATE
10291                 | NFS_CAP_SEEK
10292                 | NFS_CAP_LAYOUTSTATS
10293                 | NFS_CAP_CLONE
10294                 | NFS_CAP_LAYOUTERROR
10295                 | NFS_CAP_READ_PLUS,
10296         .init_client = nfs41_init_client,
10297         .shutdown_client = nfs41_shutdown_client,
10298         .match_stateid = nfs41_match_stateid,
10299         .find_root_sec = nfs41_find_root_sec,
10300         .free_lock_state = nfs41_free_lock_state,
10301         .call_sync_ops = &nfs41_call_sync_ops,
10302         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10303         .alloc_seqid = nfs_alloc_no_seqid,
10304         .session_trunk = nfs4_test_session_trunk,
10305         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10306         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10307         .state_renewal_ops = &nfs41_state_renewal_ops,
10308         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10309 };
10310 #endif
10311
10312 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10313         [0] = &nfs_v4_0_minor_ops,
10314 #if defined(CONFIG_NFS_V4_1)
10315         [1] = &nfs_v4_1_minor_ops,
10316 #endif
10317 #if defined(CONFIG_NFS_V4_2)
10318         [2] = &nfs_v4_2_minor_ops,
10319 #endif
10320 };
10321
10322 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10323 {
10324         ssize_t error, error2, error3;
10325
10326         error = generic_listxattr(dentry, list, size);
10327         if (error < 0)
10328                 return error;
10329         if (list) {
10330                 list += error;
10331                 size -= error;
10332         }
10333
10334         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10335         if (error2 < 0)
10336                 return error2;
10337
10338         if (list) {
10339                 list += error2;
10340                 size -= error2;
10341         }
10342
10343         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10344         if (error3 < 0)
10345                 return error3;
10346
10347         return error + error2 + error3;
10348 }
10349
10350 static const struct inode_operations nfs4_dir_inode_operations = {
10351         .create         = nfs_create,
10352         .lookup         = nfs_lookup,
10353         .atomic_open    = nfs_atomic_open,
10354         .link           = nfs_link,
10355         .unlink         = nfs_unlink,
10356         .symlink        = nfs_symlink,
10357         .mkdir          = nfs_mkdir,
10358         .rmdir          = nfs_rmdir,
10359         .mknod          = nfs_mknod,
10360         .rename         = nfs_rename,
10361         .permission     = nfs_permission,
10362         .getattr        = nfs_getattr,
10363         .setattr        = nfs_setattr,
10364         .listxattr      = nfs4_listxattr,
10365 };
10366
10367 static const struct inode_operations nfs4_file_inode_operations = {
10368         .permission     = nfs_permission,
10369         .getattr        = nfs_getattr,
10370         .setattr        = nfs_setattr,
10371         .listxattr      = nfs4_listxattr,
10372 };
10373
10374 const struct nfs_rpc_ops nfs_v4_clientops = {
10375         .version        = 4,                    /* protocol version */
10376         .dentry_ops     = &nfs4_dentry_operations,
10377         .dir_inode_ops  = &nfs4_dir_inode_operations,
10378         .file_inode_ops = &nfs4_file_inode_operations,
10379         .file_ops       = &nfs4_file_operations,
10380         .getroot        = nfs4_proc_get_root,
10381         .submount       = nfs4_submount,
10382         .try_get_tree   = nfs4_try_get_tree,
10383         .getattr        = nfs4_proc_getattr,
10384         .setattr        = nfs4_proc_setattr,
10385         .lookup         = nfs4_proc_lookup,
10386         .lookupp        = nfs4_proc_lookupp,
10387         .access         = nfs4_proc_access,
10388         .readlink       = nfs4_proc_readlink,
10389         .create         = nfs4_proc_create,
10390         .remove         = nfs4_proc_remove,
10391         .unlink_setup   = nfs4_proc_unlink_setup,
10392         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10393         .unlink_done    = nfs4_proc_unlink_done,
10394         .rename_setup   = nfs4_proc_rename_setup,
10395         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10396         .rename_done    = nfs4_proc_rename_done,
10397         .link           = nfs4_proc_link,
10398         .symlink        = nfs4_proc_symlink,
10399         .mkdir          = nfs4_proc_mkdir,
10400         .rmdir          = nfs4_proc_rmdir,
10401         .readdir        = nfs4_proc_readdir,
10402         .mknod          = nfs4_proc_mknod,
10403         .statfs         = nfs4_proc_statfs,
10404         .fsinfo         = nfs4_proc_fsinfo,
10405         .pathconf       = nfs4_proc_pathconf,
10406         .set_capabilities = nfs4_server_capabilities,
10407         .decode_dirent  = nfs4_decode_dirent,
10408         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10409         .read_setup     = nfs4_proc_read_setup,
10410         .read_done      = nfs4_read_done,
10411         .write_setup    = nfs4_proc_write_setup,
10412         .write_done     = nfs4_write_done,
10413         .commit_setup   = nfs4_proc_commit_setup,
10414         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10415         .commit_done    = nfs4_commit_done,
10416         .lock           = nfs4_proc_lock,
10417         .clear_acl_cache = nfs4_zap_acl_attr,
10418         .close_context  = nfs4_close_context,
10419         .open_context   = nfs4_atomic_open,
10420         .have_delegation = nfs4_have_delegation,
10421         .alloc_client   = nfs4_alloc_client,
10422         .init_client    = nfs4_init_client,
10423         .free_client    = nfs4_free_client,
10424         .create_server  = nfs4_create_server,
10425         .clone_server   = nfs_clone_server,
10426 };
10427
10428 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10429         .name   = XATTR_NAME_NFSV4_ACL,
10430         .list   = nfs4_xattr_list_nfs4_acl,
10431         .get    = nfs4_xattr_get_nfs4_acl,
10432         .set    = nfs4_xattr_set_nfs4_acl,
10433 };
10434
10435 #ifdef CONFIG_NFS_V4_2
10436 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10437         .prefix = XATTR_USER_PREFIX,
10438         .get    = nfs4_xattr_get_nfs4_user,
10439         .set    = nfs4_xattr_set_nfs4_user,
10440 };
10441 #endif
10442
10443 const struct xattr_handler *nfs4_xattr_handlers[] = {
10444         &nfs4_xattr_nfs4_acl_handler,
10445 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10446         &nfs4_xattr_nfs4_label_handler,
10447 #endif
10448 #ifdef CONFIG_NFS_V4_2
10449         &nfs4_xattr_nfs4_user_handler,
10450 #endif
10451         NULL
10452 };