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 <[email protected]>
  10  *  Andy Adamson   <[email protected]>
  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, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  97 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  98                             struct nfs_fattr *fattr, struct iattr *sattr,
  99                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
 100                             struct nfs4_label *olabel);
 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, (void **)&label->label, &label->len);
 131         if (err == 0)
 132                 return label;
 133
 134         return NULL;
 135 }
 136 static inline void
 137 nfs4_label_release_security(struct nfs4_label *label)
 138 {
 139         if (label)
 140                 security_release_secctx(label->label, label->len);
 141 }
 142 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 143 {
 144         if (label)
 145                 return server->attr_bitmask;
 146
 147         return server->attr_bitmask_nl;
 148 }
 149 #else
 150 static inline struct nfs4_label *
 151 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 152         struct iattr *sattr, struct nfs4_label *l)
 153 { return NULL; }
 154 static inline void
 155 nfs4_label_release_security(struct nfs4_label *label)
 156 { return; }
 157 static inline u32 *
 158 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 159 { return server->attr_bitmask; }
 160 #endif
 161
 162 /* Prevent leaks of NFSv4 errors into userland */
 163 static int nfs4_map_errors(int err)
 164 {
 165         if (err >= -1000)
 166                 return err;
 167         switch (err) {
 168         case -NFS4ERR_RESOURCE:
 169         case -NFS4ERR_LAYOUTTRYLATER:
 170         case -NFS4ERR_RECALLCONFLICT:
 171                 return -EREMOTEIO;
 172         case -NFS4ERR_WRONGSEC:
 173         case -NFS4ERR_WRONG_CRED:
 174                 return -EPERM;
 175         case -NFS4ERR_BADOWNER:
 176         case -NFS4ERR_BADNAME:
 177                 return -EINVAL;
 178         case -NFS4ERR_SHARE_DENIED:
 179                 return -EACCES;
 180         case -NFS4ERR_MINOR_VERS_MISMATCH:
 181                 return -EPROTONOSUPPORT;
 182         case -NFS4ERR_FILE_OPEN:
 183                 return -EBUSY;
 184         case -NFS4ERR_NOT_SAME:
 185                 return -ENOTSYNC;
 186         default:
 187                 dprintk("%s could not handle NFSv4 error %d\n",
 188                                 __func__, -err);
 189                 break;
 190         }
 191         return -EIO;
 192 }
 193
 194 /*
 195  * This is our standard bitmap for GETATTR requests.
 196  */
 197 const u32 nfs4_fattr_bitmap[3] = {
 198         FATTR4_WORD0_TYPE
 199         | FATTR4_WORD0_CHANGE
 200         | FATTR4_WORD0_SIZE
 201         | FATTR4_WORD0_FSID
 202         | FATTR4_WORD0_FILEID,
 203         FATTR4_WORD1_MODE
 204         | FATTR4_WORD1_NUMLINKS
 205         | FATTR4_WORD1_OWNER
 206         | FATTR4_WORD1_OWNER_GROUP
 207         | FATTR4_WORD1_RAWDEV
 208         | FATTR4_WORD1_SPACE_USED
 209         | FATTR4_WORD1_TIME_ACCESS
 210         | FATTR4_WORD1_TIME_METADATA
 211         | FATTR4_WORD1_TIME_MODIFY
 212         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 213 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 214         FATTR4_WORD2_SECURITY_LABEL
 215 #endif
 216 };
 217
 218 static const u32 nfs4_pnfs_open_bitmap[3] = {
 219         FATTR4_WORD0_TYPE
 220         | FATTR4_WORD0_CHANGE
 221         | FATTR4_WORD0_SIZE
 222         | FATTR4_WORD0_FSID
 223         | FATTR4_WORD0_FILEID,
 224         FATTR4_WORD1_MODE
 225         | FATTR4_WORD1_NUMLINKS
 226         | FATTR4_WORD1_OWNER
 227         | FATTR4_WORD1_OWNER_GROUP
 228         | FATTR4_WORD1_RAWDEV
 229         | FATTR4_WORD1_SPACE_USED
 230         | FATTR4_WORD1_TIME_ACCESS
 231         | FATTR4_WORD1_TIME_METADATA
 232         | FATTR4_WORD1_TIME_MODIFY,
 233         FATTR4_WORD2_MDSTHRESHOLD
 234 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 235         | FATTR4_WORD2_SECURITY_LABEL
 236 #endif
 237 };
 238
 239 static const u32 nfs4_open_noattr_bitmap[3] = {
 240         FATTR4_WORD0_TYPE
 241         | FATTR4_WORD0_FILEID,
 242 };
 243
 244 const u32 nfs4_statfs_bitmap[3] = {
 245         FATTR4_WORD0_FILES_AVAIL
 246         | FATTR4_WORD0_FILES_FREE
 247         | FATTR4_WORD0_FILES_TOTAL,
 248         FATTR4_WORD1_SPACE_AVAIL
 249         | FATTR4_WORD1_SPACE_FREE
 250         | FATTR4_WORD1_SPACE_TOTAL
 251 };
 252
 253 const u32 nfs4_pathconf_bitmap[3] = {
 254         FATTR4_WORD0_MAXLINK
 255         | FATTR4_WORD0_MAXNAME,
 256         0
 257 };
 258
 259 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 260                         | FATTR4_WORD0_MAXREAD
 261                         | FATTR4_WORD0_MAXWRITE
 262                         | FATTR4_WORD0_LEASE_TIME,
 263                         FATTR4_WORD1_TIME_DELTA
 264                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 265                         FATTR4_WORD2_LAYOUT_BLKSIZE
 266                         | FATTR4_WORD2_CLONE_BLKSIZE
 267                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
 268                         | FATTR4_WORD2_XATTR_SUPPORT
 269 };
 270
 271 const u32 nfs4_fs_locations_bitmap[3] = {
 272         FATTR4_WORD0_CHANGE
 273         | FATTR4_WORD0_SIZE
 274         | FATTR4_WORD0_FSID
 275         | FATTR4_WORD0_FILEID
 276         | FATTR4_WORD0_FS_LOCATIONS,
 277         FATTR4_WORD1_OWNER
 278         | FATTR4_WORD1_OWNER_GROUP
 279         | FATTR4_WORD1_RAWDEV
 280         | FATTR4_WORD1_SPACE_USED
 281         | FATTR4_WORD1_TIME_ACCESS
 282         | FATTR4_WORD1_TIME_METADATA
 283         | FATTR4_WORD1_TIME_MODIFY
 284         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 285 };
 286
 287 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 288                                     struct inode *inode, unsigned long flags)
 289 {
 290         unsigned long cache_validity;
 291
 292         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 293         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 294                 return;
 295
 296         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
 297
 298         /* Remove the attributes over which we have full control */
 299         dst[1] &= ~FATTR4_WORD1_RAWDEV;
 300         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 301                 dst[0] &= ~FATTR4_WORD0_SIZE;
 302
 303         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 304                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 305
 306         if (!(cache_validity & NFS_INO_INVALID_MODE))
 307                 dst[1] &= ~FATTR4_WORD1_MODE;
 308         if (!(cache_validity & NFS_INO_INVALID_OTHER))
 309                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
 310 }
 311
 312 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 313                 struct nfs4_readdir_arg *readdir)
 314 {
 315         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 316         __be32 *start, *p;
 317
 318         if (cookie > 2) {
 319                 readdir->cookie = cookie;
 320                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 321                 return;
 322         }
 323
 324         readdir->cookie = 0;
 325         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 326         if (cookie == 2)
 327                 return;
 328
 329         /*
 330          * NFSv4 servers do not return entries for '.' and '..'
 331          * Therefore, we fake these entries here.  We let '.'
 332          * have cookie 0 and '..' have cookie 1.  Note that
 333          * when talking to the server, we always send cookie 0
 334          * instead of 1 or 2.
 335          */
 336         start = p = kmap_atomic(*readdir->pages);
 337
 338         if (cookie == 0) {
 339                 *p++ = xdr_one;                                  /* next */
 340                 *p++ = xdr_zero;                   /* cookie, first word */
 341                 *p++ = xdr_one;                   /* cookie, second word */
 342                 *p++ = xdr_one;                             /* entry len */
 343                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 344                 p++;
 345                 *p++ = xdr_one;                         /* bitmap length */
 346                 *p++ = htonl(attrs);                           /* bitmap */
 347                 *p++ = htonl(12);             /* attribute buffer length */
 348                 *p++ = htonl(NF4DIR);
 349                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 350         }
 351
 352         *p++ = xdr_one;                                  /* next */
 353         *p++ = xdr_zero;                   /* cookie, first word */
 354         *p++ = xdr_two;                   /* cookie, second word */
 355         *p++ = xdr_two;                             /* entry len */
 356         memcpy(p, "..\0\0", 4);                         /* entry */
 357         p++;
 358         *p++ = xdr_one;                         /* bitmap length */
 359         *p++ = htonl(attrs);                           /* bitmap */
 360         *p++ = htonl(12);             /* attribute buffer length */
 361         *p++ = htonl(NF4DIR);
 362         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 363
 364         readdir->pgbase = (char *)p - (char *)start;
 365         readdir->count -= readdir->pgbase;
 366         kunmap_atomic(start);
 367 }
 368
 369 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 370                 nfs4_stateid *stateid,
 371                 const struct cred *cred)
 372 {
 373         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 374
 375         ops->test_and_free_expired(server, stateid, cred);
 376 }
 377
 378 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 383         nfs4_test_and_free_stateid(server, stateid, cred);
 384 }
 385
 386 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 387                 const nfs4_stateid *stateid,
 388                 const struct cred *cred)
 389 {
 390         nfs4_stateid tmp;
 391
 392         nfs4_stateid_copy(&tmp, stateid);
 393         __nfs4_free_revoked_stateid(server, &tmp, cred);
 394 }
 395
 396 static long nfs4_update_delay(long *timeout)
 397 {
 398         long ret;
 399         if (!timeout)
 400                 return NFS4_POLL_RETRY_MAX;
 401         if (*timeout <= 0)
 402                 *timeout = NFS4_POLL_RETRY_MIN;
 403         if (*timeout > NFS4_POLL_RETRY_MAX)
 404                 *timeout = NFS4_POLL_RETRY_MAX;
 405         ret = *timeout;
 406         *timeout <<= 1;
 407         return ret;
 408 }
 409
 410 static int nfs4_delay_killable(long *timeout)
 411 {
 412         might_sleep();
 413
 414         freezable_schedule_timeout_killable_unsafe(
 415                 nfs4_update_delay(timeout));
 416         if (!__fatal_signal_pending(current))
 417                 return 0;
 418         return -EINTR;
 419 }
 420
 421 static int nfs4_delay_interruptible(long *timeout)
 422 {
 423         might_sleep();
 424
 425         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 426         if (!signal_pending(current))
 427                 return 0;
 428         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 429 }
 430
 431 static int nfs4_delay(long *timeout, bool interruptible)
 432 {
 433         if (interruptible)
 434                 return nfs4_delay_interruptible(timeout);
 435         return nfs4_delay_killable(timeout);
 436 }
 437
 438 static const nfs4_stateid *
 439 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 440 {
 441         if (!stateid)
 442                 return NULL;
 443         switch (stateid->type) {
 444         case NFS4_OPEN_STATEID_TYPE:
 445         case NFS4_LOCK_STATEID_TYPE:
 446         case NFS4_DELEGATION_STATEID_TYPE:
 447                 return stateid;
 448         default:
 449                 break;
 450         }
 451         return NULL;
 452 }
 453
 454 /* This is the error handling routine for processes that are allowed
 455  * to sleep.
 456  */
 457 static int nfs4_do_handle_exception(struct nfs_server *server,
 458                 int errorcode, struct nfs4_exception *exception)
 459 {
 460         struct nfs_client *clp = server->nfs_client;
 461         struct nfs4_state *state = exception->state;
 462         const nfs4_stateid *stateid;
 463         struct inode *inode = exception->inode;
 464         int ret = errorcode;
 465
 466         exception->delay = 0;
 467         exception->recovering = 0;
 468         exception->retry = 0;
 469
 470         stateid = nfs4_recoverable_stateid(exception->stateid);
 471         if (stateid == NULL && state != NULL)
 472                 stateid = nfs4_recoverable_stateid(&state->stateid);
 473
 474         switch(errorcode) {
 475                 case 0:
 476                         return 0;
 477                 case -NFS4ERR_BADHANDLE:
 478                 case -ESTALE:
 479                         if (inode != NULL && S_ISREG(inode->i_mode))
 480                                 pnfs_destroy_layout(NFS_I(inode));
 481                         break;
 482                 case -NFS4ERR_DELEG_REVOKED:
 483                 case -NFS4ERR_ADMIN_REVOKED:
 484                 case -NFS4ERR_EXPIRED:
 485                 case -NFS4ERR_BAD_STATEID:
 486                 case -NFS4ERR_PARTNER_NO_AUTH:
 487                         if (inode != NULL && stateid != NULL) {
 488                                 nfs_inode_find_state_and_recover(inode,
 489                                                 stateid);
 490                                 goto wait_on_recovery;
 491                         }
 492                         fallthrough;
 493                 case -NFS4ERR_OPENMODE:
 494                         if (inode) {
 495                                 int err;
 496
 497                                 err = nfs_async_inode_return_delegation(inode,
 498                                                 stateid);
 499                                 if (err == 0)
 500                                         goto wait_on_recovery;
 501                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 502                                         exception->retry = 1;
 503                                         break;
 504                                 }
 505                         }
 506                         if (state == NULL)
 507                                 break;
 508                         ret = nfs4_schedule_stateid_recovery(server, state);
 509                         if (ret < 0)
 510                                 break;
 511                         goto wait_on_recovery;
 512                 case -NFS4ERR_STALE_STATEID:
 513                 case -NFS4ERR_STALE_CLIENTID:
 514                         nfs4_schedule_lease_recovery(clp);
 515                         goto wait_on_recovery;
 516                 case -NFS4ERR_MOVED:
 517                         ret = nfs4_schedule_migration_recovery(server);
 518                         if (ret < 0)
 519                                 break;
 520                         goto wait_on_recovery;
 521                 case -NFS4ERR_LEASE_MOVED:
 522                         nfs4_schedule_lease_moved_recovery(clp);
 523                         goto wait_on_recovery;
 524 #if defined(CONFIG_NFS_V4_1)
 525                 case -NFS4ERR_BADSESSION:
 526                 case -NFS4ERR_BADSLOT:
 527                 case -NFS4ERR_BAD_HIGH_SLOT:
 528                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 529                 case -NFS4ERR_DEADSESSION:
 530                 case -NFS4ERR_SEQ_FALSE_RETRY:
 531                 case -NFS4ERR_SEQ_MISORDERED:
 532                         /* Handled in nfs41_sequence_process() */
 533                         goto wait_on_recovery;
 534 #endif /* defined(CONFIG_NFS_V4_1) */
 535                 case -NFS4ERR_FILE_OPEN:
 536                         if (exception->timeout > HZ) {
 537                                 /* We have retried a decent amount, time to
 538                                  * fail
 539                                  */
 540                                 ret = -EBUSY;
 541                                 break;
 542                         }
 543                         fallthrough;
 544                 case -NFS4ERR_DELAY:
 545                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 546                         fallthrough;
 547                 case -NFS4ERR_GRACE:
 548                 case -NFS4ERR_LAYOUTTRYLATER:
 549                 case -NFS4ERR_RECALLCONFLICT:
 550                         exception->delay = 1;
 551                         return 0;
 552
 553                 case -NFS4ERR_RETRY_UNCACHED_REP:
 554                 case -NFS4ERR_OLD_STATEID:
 555                         exception->retry = 1;
 556                         break;
 557                 case -NFS4ERR_BADOWNER:
 558                         /* The following works around a Linux server bug! */
 559                 case -NFS4ERR_BADNAME:
 560                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 561                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 562                                 exception->retry = 1;
 563                                 printk(KERN_WARNING "NFS: v4 server %s "
 564                                                 "does not accept raw "
 565                                                 "uid/gids. "
 566                                                 "Reenabling the idmapper.\n",
 567                                                 server->nfs_client->cl_hostname);
 568                         }
 569         }
 570         /* We failed to handle the error */
 571         return nfs4_map_errors(ret);
 572 wait_on_recovery:
 573         exception->recovering = 1;
 574         return 0;
 575 }
 576
 577 /* This is the error handling routine for processes that are allowed
 578  * to sleep.
 579  */
 580 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 581 {
 582         struct nfs_client *clp = server->nfs_client;
 583         int ret;
 584
 585         ret = nfs4_do_handle_exception(server, errorcode, exception);
 586         if (exception->delay) {
 587                 ret = nfs4_delay(&exception->timeout,
 588                                 exception->interruptible);
 589                 goto out_retry;
 590         }
 591         if (exception->recovering) {
 592                 if (exception->task_is_privileged)
 593                         return -EDEADLOCK;
 594                 ret = nfs4_wait_clnt_recover(clp);
 595                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 596                         return -EIO;
 597                 goto out_retry;
 598         }
 599         return ret;
 600 out_retry:
 601         if (ret == 0)
 602                 exception->retry = 1;
 603         return ret;
 604 }
 605
 606 static int
 607 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 608                 int errorcode, struct nfs4_exception *exception)
 609 {
 610         struct nfs_client *clp = server->nfs_client;
 611         int ret;
 612
 613         ret = nfs4_do_handle_exception(server, errorcode, exception);
 614         if (exception->delay) {
 615                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 616                 goto out_retry;
 617         }
 618         if (exception->recovering) {
 619                 if (exception->task_is_privileged)
 620                         return -EDEADLOCK;
 621                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 622                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 623                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 624                 goto out_retry;
 625         }
 626         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 627                 ret = -EIO;
 628         return ret;
 629 out_retry:
 630         if (ret == 0) {
 631                 exception->retry = 1;
 632                 /*
 633                  * For NFS4ERR_MOVED, the client transport will need to
 634                  * be recomputed after migration recovery has completed.
 635                  */
 636                 if (errorcode == -NFS4ERR_MOVED)
 637                         rpc_task_release_transport(task);
 638         }
 639         return ret;
 640 }
 641
 642 int
 643 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 644                         struct nfs4_state *state, long *timeout)
 645 {
 646         struct nfs4_exception exception = {
 647                 .state = state,
 648         };
 649
 650         if (task->tk_status >= 0)
 651                 return 0;
 652         if (timeout)
 653                 exception.timeout = *timeout;
 654         task->tk_status = nfs4_async_handle_exception(task, server,
 655                         task->tk_status,
 656                         &exception);
 657         if (exception.delay && timeout)
 658                 *timeout = exception.timeout;
 659         if (exception.retry)
 660                 return -EAGAIN;
 661         return 0;
 662 }
 663
 664 /*
 665  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 666  * or 'false' otherwise.
 667  */
 668 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 669 {
 670         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 671         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 672 }
 673
 674 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 675 {
 676         spin_lock(&clp->cl_lock);
 677         if (time_before(clp->cl_last_renewal,timestamp))
 678                 clp->cl_last_renewal = timestamp;
 679         spin_unlock(&clp->cl_lock);
 680 }
 681
 682 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 683 {
 684         struct nfs_client *clp = server->nfs_client;
 685
 686         if (!nfs4_has_session(clp))
 687                 do_renew_lease(clp, timestamp);
 688 }
 689
 690 struct nfs4_call_sync_data {
 691         const struct nfs_server *seq_server;
 692         struct nfs4_sequence_args *seq_args;
 693         struct nfs4_sequence_res *seq_res;
 694 };
 695
 696 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 697                         struct nfs4_sequence_res *res, int cache_reply,
 698                         int privileged)
 699 {
 700         args->sa_slot = NULL;
 701         args->sa_cache_this = cache_reply;
 702         args->sa_privileged = privileged;
 703
 704         res->sr_slot = NULL;
 705 }
 706
 707 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 708 {
 709         struct nfs4_slot *slot = res->sr_slot;
 710         struct nfs4_slot_table *tbl;
 711
 712         tbl = slot->table;
 713         spin_lock(&tbl->slot_tbl_lock);
 714         if (!nfs41_wake_and_assign_slot(tbl, slot))
 715                 nfs4_free_slot(tbl, slot);
 716         spin_unlock(&tbl->slot_tbl_lock);
 717
 718         res->sr_slot = NULL;
 719 }
 720
 721 static int nfs40_sequence_done(struct rpc_task *task,
 722                                struct nfs4_sequence_res *res)
 723 {
 724         if (res->sr_slot != NULL)
 725                 nfs40_sequence_free_slot(res);
 726         return 1;
 727 }
 728
 729 #if defined(CONFIG_NFS_V4_1)
 730
 731 static void nfs41_release_slot(struct nfs4_slot *slot)
 732 {
 733         struct nfs4_session *session;
 734         struct nfs4_slot_table *tbl;
 735         bool send_new_highest_used_slotid = false;
 736
 737         if (!slot)
 738                 return;
 739         tbl = slot->table;
 740         session = tbl->session;
 741
 742         /* Bump the slot sequence number */
 743         if (slot->seq_done)
 744                 slot->seq_nr++;
 745         slot->seq_done = 0;
 746
 747         spin_lock(&tbl->slot_tbl_lock);
 748         /* Be nice to the server: try to ensure that the last transmitted
 749          * value for highest_user_slotid <= target_highest_slotid
 750          */
 751         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 752                 send_new_highest_used_slotid = true;
 753
 754         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 755                 send_new_highest_used_slotid = false;
 756                 goto out_unlock;
 757         }
 758         nfs4_free_slot(tbl, slot);
 759
 760         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 761                 send_new_highest_used_slotid = false;
 762 out_unlock:
 763         spin_unlock(&tbl->slot_tbl_lock);
 764         if (send_new_highest_used_slotid)
 765                 nfs41_notify_server(session->clp);
 766         if (waitqueue_active(&tbl->slot_waitq))
 767                 wake_up_all(&tbl->slot_waitq);
 768 }
 769
 770 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 771 {
 772         nfs41_release_slot(res->sr_slot);
 773         res->sr_slot = NULL;
 774 }
 775
 776 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 777                 u32 seqnr)
 778 {
 779         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 780                 slot->seq_nr_highest_sent = seqnr;
 781 }
 782 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 783                 u32 seqnr)
 784 {
 785         slot->seq_nr_highest_sent = seqnr;
 786         slot->seq_nr_last_acked = seqnr;
 787 }
 788
 789 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 790                                 struct nfs4_slot *slot)
 791 {
 792         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 793         if (!IS_ERR(task))
 794                 rpc_put_task_async(task);
 795 }
 796
 797 static int nfs41_sequence_process(struct rpc_task *task,
 798                 struct nfs4_sequence_res *res)
 799 {
 800         struct nfs4_session *session;
 801         struct nfs4_slot *slot = res->sr_slot;
 802         struct nfs_client *clp;
 803         int status;
 804         int ret = 1;
 805
 806         if (slot == NULL)
 807                 goto out_noaction;
 808         /* don't increment the sequence number if the task wasn't sent */
 809         if (!RPC_WAS_SENT(task) || slot->seq_done)
 810                 goto out;
 811
 812         session = slot->table->session;
 813         clp = session->clp;
 814
 815         trace_nfs4_sequence_done(session, res);
 816
 817         status = res->sr_status;
 818         if (task->tk_status == -NFS4ERR_DEADSESSION)
 819                 status = -NFS4ERR_DEADSESSION;
 820
 821         /* Check the SEQUENCE operation status */
 822         switch (status) {
 823         case 0:
 824                 /* Mark this sequence number as having been acked */
 825                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 826                 /* Update the slot's sequence and clientid lease timer */
 827                 slot->seq_done = 1;
 828                 do_renew_lease(clp, res->sr_timestamp);
 829                 /* Check sequence flags */
 830                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 831                                 !!slot->privileged);
 832                 nfs41_update_target_slotid(slot->table, slot, res);
 833                 break;
 834         case 1:
 835                 /*
 836                  * sr_status remains 1 if an RPC level error occurred.
 837                  * The server may or may not have processed the sequence
 838                  * operation..
 839                  */
 840                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 841                 slot->seq_done = 1;
 842                 goto out;
 843         case -NFS4ERR_DELAY:
 844                 /* The server detected a resend of the RPC call and
 845                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 846                  * of RFC5661.
 847                  */
 848                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 849                         __func__,
 850                         slot->slot_nr,
 851                         slot->seq_nr);
 852                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 853                 goto out_retry;
 854         case -NFS4ERR_RETRY_UNCACHED_REP:
 855         case -NFS4ERR_SEQ_FALSE_RETRY:
 856                 /*
 857                  * The server thinks we tried to replay a request.
 858                  * Retry the call after bumping the sequence ID.
 859                  */
 860                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 861                 goto retry_new_seq;
 862         case -NFS4ERR_BADSLOT:
 863                 /*
 864                  * The slot id we used was probably retired. Try again
 865                  * using a different slot id.
 866                  */
 867                 if (slot->slot_nr < slot->table->target_highest_slotid)
 868                         goto session_recover;
 869                 goto retry_nowait;
 870         case -NFS4ERR_SEQ_MISORDERED:
 871                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 872                 /*
 873                  * Were one or more calls using this slot interrupted?
 874                  * If the server never received the request, then our
 875                  * transmitted slot sequence number may be too high. However,
 876                  * if the server did receive the request then it might
 877                  * accidentally give us a reply with a mismatched operation.
 878                  * We can sort this out by sending a lone sequence operation
 879                  * to the server on the same slot.
 880                  */
 881                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 882                         slot->seq_nr--;
 883                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 884                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 885                                 res->sr_slot = NULL;
 886                         }
 887                         goto retry_nowait;
 888                 }
 889                 /*
 890                  * RFC5661:
 891                  * A retry might be sent while the original request is
 892                  * still in progress on the replier. The replier SHOULD
 893                  * deal with the issue by returning NFS4ERR_DELAY as the
 894                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 895                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 896                  *
 897                  * Restart the search after a delay.
 898                  */
 899                 slot->seq_nr = slot->seq_nr_highest_sent;
 900                 goto out_retry;
 901         case -NFS4ERR_BADSESSION:
 902         case -NFS4ERR_DEADSESSION:
 903         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 904                 goto session_recover;
 905         default:
 906                 /* Just update the slot sequence no. */
 907                 slot->seq_done = 1;
 908         }
 909 out:
 910         /* The session may be reset by one of the error handlers. */
 911         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 912 out_noaction:
 913         return ret;
 914 session_recover:
 915         nfs4_schedule_session_recovery(session, status);
 916         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 917         nfs41_sequence_free_slot(res);
 918         goto out;
 919 retry_new_seq:
 920         ++slot->seq_nr;
 921 retry_nowait:
 922         if (rpc_restart_call_prepare(task)) {
 923                 nfs41_sequence_free_slot(res);
 924                 task->tk_status = 0;
 925                 ret = 0;
 926         }
 927         goto out;
 928 out_retry:
 929         if (!rpc_restart_call(task))
 930                 goto out;
 931         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 932         return 0;
 933 }
 934
 935 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 936 {
 937         if (!nfs41_sequence_process(task, res))
 938                 return 0;
 939         if (res->sr_slot != NULL)
 940                 nfs41_sequence_free_slot(res);
 941         return 1;
 942
 943 }
 944 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 945
 946 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 947 {
 948         if (res->sr_slot == NULL)
 949                 return 1;
 950         if (res->sr_slot->table->session != NULL)
 951                 return nfs41_sequence_process(task, res);
 952         return nfs40_sequence_done(task, res);
 953 }
 954
 955 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 956 {
 957         if (res->sr_slot != NULL) {
 958                 if (res->sr_slot->table->session != NULL)
 959                         nfs41_sequence_free_slot(res);
 960                 else
 961                         nfs40_sequence_free_slot(res);
 962         }
 963 }
 964
 965 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 966 {
 967         if (res->sr_slot == NULL)
 968                 return 1;
 969         if (!res->sr_slot->table->session)
 970                 return nfs40_sequence_done(task, res);
 971         return nfs41_sequence_done(task, res);
 972 }
 973 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 974
 975 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 976 {
 977         struct nfs4_call_sync_data *data = calldata;
 978
 979         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 980
 981         nfs4_setup_sequence(data->seq_server->nfs_client,
 982                             data->seq_args, data->seq_res, task);
 983 }
 984
 985 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 986 {
 987         struct nfs4_call_sync_data *data = calldata;
 988
 989         nfs41_sequence_done(task, data->seq_res);
 990 }
 991
 992 static const struct rpc_call_ops nfs41_call_sync_ops = {
 993         .rpc_call_prepare = nfs41_call_sync_prepare,
 994         .rpc_call_done = nfs41_call_sync_done,
 995 };
 996
 997 #else   /* !CONFIG_NFS_V4_1 */
 998
 999 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1000 {
1001         return nfs40_sequence_done(task, res);
1002 }
1003
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006         if (res->sr_slot != NULL)
1007                 nfs40_sequence_free_slot(res);
1008 }
1009
1010 int nfs4_sequence_done(struct rpc_task *task,
1011                        struct nfs4_sequence_res *res)
1012 {
1013         return nfs40_sequence_done(task, res);
1014 }
1015 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1016
1017 #endif  /* !CONFIG_NFS_V4_1 */
1018
1019 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1020 {
1021         res->sr_timestamp = jiffies;
1022         res->sr_status_flags = 0;
1023         res->sr_status = 1;
1024 }
1025
1026 static
1027 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1028                 struct nfs4_sequence_res *res,
1029                 struct nfs4_slot *slot)
1030 {
1031         if (!slot)
1032                 return;
1033         slot->privileged = args->sa_privileged ? 1 : 0;
1034         args->sa_slot = slot;
1035
1036         res->sr_slot = slot;
1037 }
1038
1039 int nfs4_setup_sequence(struct nfs_client *client,
1040                         struct nfs4_sequence_args *args,
1041                         struct nfs4_sequence_res *res,
1042                         struct rpc_task *task)
1043 {
1044         struct nfs4_session *session = nfs4_get_session(client);
1045         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1046         struct nfs4_slot *slot;
1047
1048         /* slot already allocated? */
1049         if (res->sr_slot != NULL)
1050                 goto out_start;
1051
1052         if (session)
1053                 tbl = &session->fc_slot_table;
1054
1055         spin_lock(&tbl->slot_tbl_lock);
1056         /* The state manager will wait until the slot table is empty */
1057         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1058                 goto out_sleep;
1059
1060         slot = nfs4_alloc_slot(tbl);
1061         if (IS_ERR(slot)) {
1062                 if (slot == ERR_PTR(-ENOMEM))
1063                         goto out_sleep_timeout;
1064                 goto out_sleep;
1065         }
1066         spin_unlock(&tbl->slot_tbl_lock);
1067
1068         nfs4_sequence_attach_slot(args, res, slot);
1069
1070         trace_nfs4_setup_sequence(session, args);
1071 out_start:
1072         nfs41_sequence_res_init(res);
1073         rpc_call_start(task);
1074         return 0;
1075 out_sleep_timeout:
1076         /* Try again in 1/4 second */
1077         if (args->sa_privileged)
1078                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1079                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1080         else
1081                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1082                                 NULL, jiffies + (HZ >> 2));
1083         spin_unlock(&tbl->slot_tbl_lock);
1084         return -EAGAIN;
1085 out_sleep:
1086         if (args->sa_privileged)
1087                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1088                                 RPC_PRIORITY_PRIVILEGED);
1089         else
1090                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1091         spin_unlock(&tbl->slot_tbl_lock);
1092         return -EAGAIN;
1093 }
1094 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1095
1096 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1097 {
1098         struct nfs4_call_sync_data *data = calldata;
1099         nfs4_setup_sequence(data->seq_server->nfs_client,
1100                                 data->seq_args, data->seq_res, task);
1101 }
1102
1103 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1104 {
1105         struct nfs4_call_sync_data *data = calldata;
1106         nfs4_sequence_done(task, data->seq_res);
1107 }
1108
1109 static const struct rpc_call_ops nfs40_call_sync_ops = {
1110         .rpc_call_prepare = nfs40_call_sync_prepare,
1111         .rpc_call_done = nfs40_call_sync_done,
1112 };
1113
1114 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1115 {
1116         int ret;
1117         struct rpc_task *task;
1118
1119         task = rpc_run_task(task_setup);
1120         if (IS_ERR(task))
1121                 return PTR_ERR(task);
1122
1123         ret = task->tk_status;
1124         rpc_put_task(task);
1125         return ret;
1126 }
1127
1128 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1129                              struct nfs_server *server,
1130                              struct rpc_message *msg,
1131                              struct nfs4_sequence_args *args,
1132                              struct nfs4_sequence_res *res,
1133                              unsigned short task_flags)
1134 {
1135         struct nfs_client *clp = server->nfs_client;
1136         struct nfs4_call_sync_data data = {
1137                 .seq_server = server,
1138                 .seq_args = args,
1139                 .seq_res = res,
1140         };
1141         struct rpc_task_setup task_setup = {
1142                 .rpc_client = clnt,
1143                 .rpc_message = msg,
1144                 .callback_ops = clp->cl_mvops->call_sync_ops,
1145                 .callback_data = &data,
1146                 .flags = task_flags,
1147         };
1148
1149         return nfs4_call_sync_custom(&task_setup);
1150 }
1151
1152 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1153                                    struct nfs_server *server,
1154                                    struct rpc_message *msg,
1155                                    struct nfs4_sequence_args *args,
1156                                    struct nfs4_sequence_res *res)
1157 {
1158         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1159 }
1160
1161
1162 int nfs4_call_sync(struct rpc_clnt *clnt,
1163                    struct nfs_server *server,
1164                    struct rpc_message *msg,
1165                    struct nfs4_sequence_args *args,
1166                    struct nfs4_sequence_res *res,
1167                    int cache_reply)
1168 {
1169         nfs4_init_sequence(args, res, cache_reply, 0);
1170         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1171 }
1172
1173 static void
1174 nfs4_inc_nlink_locked(struct inode *inode)
1175 {
1176         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1177                                              NFS_INO_INVALID_CTIME |
1178                                              NFS_INO_INVALID_NLINK);
1179         inc_nlink(inode);
1180 }
1181
1182 static void
1183 nfs4_inc_nlink(struct inode *inode)
1184 {
1185         spin_lock(&inode->i_lock);
1186         nfs4_inc_nlink_locked(inode);
1187         spin_unlock(&inode->i_lock);
1188 }
1189
1190 static void
1191 nfs4_dec_nlink_locked(struct inode *inode)
1192 {
1193         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1194                                              NFS_INO_INVALID_CTIME |
1195                                              NFS_INO_INVALID_NLINK);
1196         drop_nlink(inode);
1197 }
1198
1199 static void
1200 nfs4_update_changeattr_locked(struct inode *inode,
1201                 struct nfs4_change_info *cinfo,
1202                 unsigned long timestamp, unsigned long cache_validity)
1203 {
1204         struct nfs_inode *nfsi = NFS_I(inode);
1205         u64 change_attr = inode_peek_iversion_raw(inode);
1206
1207         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1208
1209         switch (NFS_SERVER(inode)->change_attr_type) {
1210         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1211                 break;
1212         case NFS4_CHANGE_TYPE_IS_TIME_METADATA:
1213                 if ((s64)(change_attr - cinfo->after) > 0)
1214                         goto out;
1215                 break;
1216         default:
1217                 if ((s64)(change_attr - cinfo->after) >= 0)
1218                         goto out;
1219         }
1220
1221         if (cinfo->atomic && cinfo->before == change_attr) {
1222                 nfsi->attrtimeo_timestamp = jiffies;
1223         } else {
1224                 if (S_ISDIR(inode->i_mode)) {
1225                         cache_validity |= NFS_INO_INVALID_DATA;
1226                         nfs_force_lookup_revalidate(inode);
1227                 } else {
1228                         if (!NFS_PROTO(inode)->have_delegation(inode,
1229                                                                FMODE_READ))
1230                                 cache_validity |= NFS_INO_REVAL_PAGECACHE;
1231                 }
1232
1233                 if (cinfo->before != change_attr)
1234                         cache_validity |= NFS_INO_INVALID_ACCESS |
1235                                           NFS_INO_INVALID_ACL |
1236                                           NFS_INO_INVALID_XATTR;
1237         }
1238         inode_set_iversion_raw(inode, cinfo->after);
1239         nfsi->read_cache_jiffies = timestamp;
1240         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1241         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1242 out:
1243         nfs_set_cache_invalid(inode, cache_validity);
1244 }
1245
1246 void
1247 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1248                 unsigned long timestamp, unsigned long cache_validity)
1249 {
1250         spin_lock(&dir->i_lock);
1251         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1252         spin_unlock(&dir->i_lock);
1253 }
1254
1255 struct nfs4_open_createattrs {
1256         struct nfs4_label *label;
1257         struct iattr *sattr;
1258         const __u32 verf[2];
1259 };
1260
1261 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1262                 int err, struct nfs4_exception *exception)
1263 {
1264         if (err != -EINVAL)
1265                 return false;
1266         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1267                 return false;
1268         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1269         exception->retry = 1;
1270         return true;
1271 }
1272
1273 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1274 {
1275          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1276 }
1277
1278 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1279 {
1280         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1281
1282         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1283 }
1284
1285 static u32
1286 nfs4_map_atomic_open_share(struct nfs_server *server,
1287                 fmode_t fmode, int openflags)
1288 {
1289         u32 res = 0;
1290
1291         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1292         case FMODE_READ:
1293                 res = NFS4_SHARE_ACCESS_READ;
1294                 break;
1295         case FMODE_WRITE:
1296                 res = NFS4_SHARE_ACCESS_WRITE;
1297                 break;
1298         case FMODE_READ|FMODE_WRITE:
1299                 res = NFS4_SHARE_ACCESS_BOTH;
1300         }
1301         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1302                 goto out;
1303         /* Want no delegation if we're using O_DIRECT */
1304         if (openflags & O_DIRECT)
1305                 res |= NFS4_SHARE_WANT_NO_DELEG;
1306 out:
1307         return res;
1308 }
1309
1310 static enum open_claim_type4
1311 nfs4_map_atomic_open_claim(struct nfs_server *server,
1312                 enum open_claim_type4 claim)
1313 {
1314         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1315                 return claim;
1316         switch (claim) {
1317         default:
1318                 return claim;
1319         case NFS4_OPEN_CLAIM_FH:
1320                 return NFS4_OPEN_CLAIM_NULL;
1321         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1322                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1323         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1324                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1325         }
1326 }
1327
1328 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1329 {
1330         p->o_res.f_attr = &p->f_attr;
1331         p->o_res.f_label = p->f_label;
1332         p->o_res.seqid = p->o_arg.seqid;
1333         p->c_res.seqid = p->c_arg.seqid;
1334         p->o_res.server = p->o_arg.server;
1335         p->o_res.access_request = p->o_arg.access;
1336         nfs_fattr_init(&p->f_attr);
1337         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1338 }
1339
1340 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1341                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1342                 const struct nfs4_open_createattrs *c,
1343                 enum open_claim_type4 claim,
1344                 gfp_t gfp_mask)
1345 {
1346         struct dentry *parent = dget_parent(dentry);
1347         struct inode *dir = d_inode(parent);
1348         struct nfs_server *server = NFS_SERVER(dir);
1349         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1350         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1351         struct nfs4_opendata *p;
1352
1353         p = kzalloc(sizeof(*p), gfp_mask);
1354         if (p == NULL)
1355                 goto err;
1356
1357         p->f_label = nfs4_label_alloc(server, gfp_mask);
1358         if (IS_ERR(p->f_label))
1359                 goto err_free_p;
1360
1361         p->a_label = nfs4_label_alloc(server, gfp_mask);
1362         if (IS_ERR(p->a_label))
1363                 goto err_free_f;
1364
1365         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1366         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1367         if (IS_ERR(p->o_arg.seqid))
1368                 goto err_free_label;
1369         nfs_sb_active(dentry->d_sb);
1370         p->dentry = dget(dentry);
1371         p->dir = parent;
1372         p->owner = sp;
1373         atomic_inc(&sp->so_count);
1374         p->o_arg.open_flags = flags;
1375         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1376         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1377         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1378                         fmode, flags);
1379         if (flags & O_CREAT) {
1380                 p->o_arg.umask = current_umask();
1381                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1382                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1383                         p->o_arg.u.attrs = &p->attrs;
1384                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1385
1386                         memcpy(p->o_arg.u.verifier.data, c->verf,
1387                                         sizeof(p->o_arg.u.verifier.data));
1388                 }
1389         }
1390         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1391          * will return permission denied for all bits until close */
1392         if (!(flags & O_EXCL)) {
1393                 /* ask server to check for all possible rights as results
1394                  * are cached */
1395                 switch (p->o_arg.claim) {
1396                 default:
1397                         break;
1398                 case NFS4_OPEN_CLAIM_NULL:
1399                 case NFS4_OPEN_CLAIM_FH:
1400                         p->o_arg.access = NFS4_ACCESS_READ |
1401                                 NFS4_ACCESS_MODIFY |
1402                                 NFS4_ACCESS_EXTEND |
1403                                 NFS4_ACCESS_EXECUTE;
1404 #ifdef CONFIG_NFS_V4_2
1405                         if (server->caps & NFS_CAP_XATTR)
1406                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1407                                     NFS4_ACCESS_XAWRITE |
1408                                     NFS4_ACCESS_XALIST;
1409 #endif
1410                 }
1411         }
1412         p->o_arg.clientid = server->nfs_client->cl_clientid;
1413         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1414         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1415         p->o_arg.name = &dentry->d_name;
1416         p->o_arg.server = server;
1417         p->o_arg.bitmask = nfs4_bitmask(server, label);
1418         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1419         switch (p->o_arg.claim) {
1420         case NFS4_OPEN_CLAIM_NULL:
1421         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1422         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1423                 p->o_arg.fh = NFS_FH(dir);
1424                 break;
1425         case NFS4_OPEN_CLAIM_PREVIOUS:
1426         case NFS4_OPEN_CLAIM_FH:
1427         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1428         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1429                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1430         }
1431         p->c_arg.fh = &p->o_res.fh;
1432         p->c_arg.stateid = &p->o_res.stateid;
1433         p->c_arg.seqid = p->o_arg.seqid;
1434         nfs4_init_opendata_res(p);
1435         kref_init(&p->kref);
1436         return p;
1437
1438 err_free_label:
1439         nfs4_label_free(p->a_label);
1440 err_free_f:
1441         nfs4_label_free(p->f_label);
1442 err_free_p:
1443         kfree(p);
1444 err:
1445         dput(parent);
1446         return NULL;
1447 }
1448
1449 static void nfs4_opendata_free(struct kref *kref)
1450 {
1451         struct nfs4_opendata *p = container_of(kref,
1452                         struct nfs4_opendata, kref);
1453         struct super_block *sb = p->dentry->d_sb;
1454
1455         nfs4_lgopen_release(p->lgp);
1456         nfs_free_seqid(p->o_arg.seqid);
1457         nfs4_sequence_free_slot(&p->o_res.seq_res);
1458         if (p->state != NULL)
1459                 nfs4_put_open_state(p->state);
1460         nfs4_put_state_owner(p->owner);
1461
1462         nfs4_label_free(p->a_label);
1463         nfs4_label_free(p->f_label);
1464
1465         dput(p->dir);
1466         dput(p->dentry);
1467         nfs_sb_deactive(sb);
1468         nfs_fattr_free_names(&p->f_attr);
1469         kfree(p->f_attr.mdsthreshold);
1470         kfree(p);
1471 }
1472
1473 static void nfs4_opendata_put(struct nfs4_opendata *p)
1474 {
1475         if (p != NULL)
1476                 kref_put(&p->kref, nfs4_opendata_free);
1477 }
1478
1479 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1480                 fmode_t fmode)
1481 {
1482         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1483         case FMODE_READ|FMODE_WRITE:
1484                 return state->n_rdwr != 0;
1485         case FMODE_WRITE:
1486                 return state->n_wronly != 0;
1487         case FMODE_READ:
1488                 return state->n_rdonly != 0;
1489         }
1490         WARN_ON_ONCE(1);
1491         return false;
1492 }
1493
1494 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1495                 int open_mode, enum open_claim_type4 claim)
1496 {
1497         int ret = 0;
1498
1499         if (open_mode & (O_EXCL|O_TRUNC))
1500                 goto out;
1501         switch (claim) {
1502         case NFS4_OPEN_CLAIM_NULL:
1503         case NFS4_OPEN_CLAIM_FH:
1504                 goto out;
1505         default:
1506                 break;
1507         }
1508         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1509                 case FMODE_READ:
1510                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1511                                 && state->n_rdonly != 0;
1512                         break;
1513                 case FMODE_WRITE:
1514                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1515                                 && state->n_wronly != 0;
1516                         break;
1517                 case FMODE_READ|FMODE_WRITE:
1518                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1519                                 && state->n_rdwr != 0;
1520         }
1521 out:
1522         return ret;
1523 }
1524
1525 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1526                 enum open_claim_type4 claim)
1527 {
1528         if (delegation == NULL)
1529                 return 0;
1530         if ((delegation->type & fmode) != fmode)
1531                 return 0;
1532         switch (claim) {
1533         case NFS4_OPEN_CLAIM_NULL:
1534         case NFS4_OPEN_CLAIM_FH:
1535                 break;
1536         case NFS4_OPEN_CLAIM_PREVIOUS:
1537                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1538                         break;
1539                 fallthrough;
1540         default:
1541                 return 0;
1542         }
1543         nfs_mark_delegation_referenced(delegation);
1544         return 1;
1545 }
1546
1547 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1548 {
1549         switch (fmode) {
1550                 case FMODE_WRITE:
1551                         state->n_wronly++;
1552                         break;
1553                 case FMODE_READ:
1554                         state->n_rdonly++;
1555                         break;
1556                 case FMODE_READ|FMODE_WRITE:
1557                         state->n_rdwr++;
1558         }
1559         nfs4_state_set_mode_locked(state, state->state | fmode);
1560 }
1561
1562 #ifdef CONFIG_NFS_V4_1
1563 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1564 {
1565         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1566                 return true;
1567         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1568                 return true;
1569         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1570                 return true;
1571         return false;
1572 }
1573 #endif /* CONFIG_NFS_V4_1 */
1574
1575 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1576 {
1577         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1578                 wake_up_all(&state->waitq);
1579 }
1580
1581 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1582 {
1583         struct nfs_client *clp = state->owner->so_server->nfs_client;
1584         bool need_recover = false;
1585
1586         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1587                 need_recover = true;
1588         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1589                 need_recover = true;
1590         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1591                 need_recover = true;
1592         if (need_recover)
1593                 nfs4_state_mark_reclaim_nograce(clp, state);
1594 }
1595
1596 /*
1597  * Check for whether or not the caller may update the open stateid
1598  * to the value passed in by stateid.
1599  *
1600  * Note: This function relies heavily on the server implementing
1601  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1602  * correctly.
1603  * i.e. The stateid seqids have to be initialised to 1, and
1604  * are then incremented on every state transition.
1605  */
1606 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1607                 const nfs4_stateid *stateid)
1608 {
1609         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1610                 /* The common case - we're updating to a new sequence number */
1611                 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1612                         nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1613                         return true;
1614                 }
1615         } else {
1616                 /* This is the first OPEN in this generation */
1617                 if (stateid->seqid == cpu_to_be32(1))
1618                         return true;
1619         }
1620         return false;
1621 }
1622
1623 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1624 {
1625         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1626                 return;
1627         if (state->n_wronly)
1628                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1629         if (state->n_rdonly)
1630                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1631         if (state->n_rdwr)
1632                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1633         set_bit(NFS_OPEN_STATE, &state->flags);
1634 }
1635
1636 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1637                 nfs4_stateid *stateid, fmode_t fmode)
1638 {
1639         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1640         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1641         case FMODE_WRITE:
1642                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1643                 break;
1644         case FMODE_READ:
1645                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1646                 break;
1647         case 0:
1648                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1649                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1650                 clear_bit(NFS_OPEN_STATE, &state->flags);
1651         }
1652         if (stateid == NULL)
1653                 return;
1654         /* Handle OPEN+OPEN_DOWNGRADE races */
1655         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1656             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1657                 nfs_resync_open_stateid_locked(state);
1658                 goto out;
1659         }
1660         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1661                 nfs4_stateid_copy(&state->stateid, stateid);
1662         nfs4_stateid_copy(&state->open_stateid, stateid);
1663         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1664 out:
1665         nfs_state_log_update_open_stateid(state);
1666 }
1667
1668 static void nfs_clear_open_stateid(struct nfs4_state *state,
1669         nfs4_stateid *arg_stateid,
1670         nfs4_stateid *stateid, fmode_t fmode)
1671 {
1672         write_seqlock(&state->seqlock);
1673         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1674         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1675                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1676         write_sequnlock(&state->seqlock);
1677         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1678                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1679 }
1680
1681 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1682                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1683         __must_hold(&state->owner->so_lock)
1684         __must_hold(&state->seqlock)
1685         __must_hold(RCU)
1686
1687 {
1688         DEFINE_WAIT(wait);
1689         int status = 0;
1690         for (;;) {
1691
1692                 if (nfs_stateid_is_sequential(state, stateid))
1693                         break;
1694
1695                 if (status)
1696                         break;
1697                 /* Rely on seqids for serialisation with NFSv4.0 */
1698                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1699                         break;
1700
1701                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1702                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1703                 /*
1704                  * Ensure we process the state changes in the same order
1705                  * in which the server processed them by delaying the
1706                  * update of the stateid until we are in sequence.
1707                  */
1708                 write_sequnlock(&state->seqlock);
1709                 spin_unlock(&state->owner->so_lock);
1710                 rcu_read_unlock();
1711                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1712
1713                 if (!fatal_signal_pending(current)) {
1714                         if (schedule_timeout(5*HZ) == 0)
1715                                 status = -EAGAIN;
1716                         else
1717                                 status = 0;
1718                 } else
1719                         status = -EINTR;
1720                 finish_wait(&state->waitq, &wait);
1721                 rcu_read_lock();
1722                 spin_lock(&state->owner->so_lock);
1723                 write_seqlock(&state->seqlock);
1724         }
1725
1726         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1727             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1728                 nfs4_stateid_copy(freeme, &state->open_stateid);
1729                 nfs_test_and_clear_all_open_stateid(state);
1730         }
1731
1732         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1733                 nfs4_stateid_copy(&state->stateid, stateid);
1734         nfs4_stateid_copy(&state->open_stateid, stateid);
1735         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1736         nfs_state_log_update_open_stateid(state);
1737 }
1738
1739 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1740                 const nfs4_stateid *open_stateid,
1741                 fmode_t fmode,
1742                 nfs4_stateid *freeme)
1743 {
1744         /*
1745          * Protect the call to nfs4_state_set_mode_locked and
1746          * serialise the stateid update
1747          */
1748         write_seqlock(&state->seqlock);
1749         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1750         switch (fmode) {
1751         case FMODE_READ:
1752                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1753                 break;
1754         case FMODE_WRITE:
1755                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1756                 break;
1757         case FMODE_READ|FMODE_WRITE:
1758                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1759         }
1760         set_bit(NFS_OPEN_STATE, &state->flags);
1761         write_sequnlock(&state->seqlock);
1762 }
1763
1764 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1765 {
1766         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1767         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1768         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1769         clear_bit(NFS_OPEN_STATE, &state->flags);
1770 }
1771
1772 static void nfs_state_set_delegation(struct nfs4_state *state,
1773                 const nfs4_stateid *deleg_stateid,
1774                 fmode_t fmode)
1775 {
1776         /*
1777          * Protect the call to nfs4_state_set_mode_locked and
1778          * serialise the stateid update
1779          */
1780         write_seqlock(&state->seqlock);
1781         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1782         set_bit(NFS_DELEGATED_STATE, &state->flags);
1783         write_sequnlock(&state->seqlock);
1784 }
1785
1786 static void nfs_state_clear_delegation(struct nfs4_state *state)
1787 {
1788         write_seqlock(&state->seqlock);
1789         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1790         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1791         write_sequnlock(&state->seqlock);
1792 }
1793
1794 int update_open_stateid(struct nfs4_state *state,
1795                 const nfs4_stateid *open_stateid,
1796                 const nfs4_stateid *delegation,
1797                 fmode_t fmode)
1798 {
1799         struct nfs_server *server = NFS_SERVER(state->inode);
1800         struct nfs_client *clp = server->nfs_client;
1801         struct nfs_inode *nfsi = NFS_I(state->inode);
1802         struct nfs_delegation *deleg_cur;
1803         nfs4_stateid freeme = { };
1804         int ret = 0;
1805
1806         fmode &= (FMODE_READ|FMODE_WRITE);
1807
1808         rcu_read_lock();
1809         spin_lock(&state->owner->so_lock);
1810         if (open_stateid != NULL) {
1811                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1812                 ret = 1;
1813         }
1814
1815         deleg_cur = nfs4_get_valid_delegation(state->inode);
1816         if (deleg_cur == NULL)
1817                 goto no_delegation;
1818
1819         spin_lock(&deleg_cur->lock);
1820         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1821            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1822             (deleg_cur->type & fmode) != fmode)
1823                 goto no_delegation_unlock;
1824
1825         if (delegation == NULL)
1826                 delegation = &deleg_cur->stateid;
1827         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1828                 goto no_delegation_unlock;
1829
1830         nfs_mark_delegation_referenced(deleg_cur);
1831         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1832         ret = 1;
1833 no_delegation_unlock:
1834         spin_unlock(&deleg_cur->lock);
1835 no_delegation:
1836         if (ret)
1837                 update_open_stateflags(state, fmode);
1838         spin_unlock(&state->owner->so_lock);
1839         rcu_read_unlock();
1840
1841         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1842                 nfs4_schedule_state_manager(clp);
1843         if (freeme.type != 0)
1844                 nfs4_test_and_free_stateid(server, &freeme,
1845                                 state->owner->so_cred);
1846
1847         return ret;
1848 }
1849
1850 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1851                 const nfs4_stateid *stateid)
1852 {
1853         struct nfs4_state *state = lsp->ls_state;
1854         bool ret = false;
1855
1856         spin_lock(&state->state_lock);
1857         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1858                 goto out_noupdate;
1859         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1860                 goto out_noupdate;
1861         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1862         ret = true;
1863 out_noupdate:
1864         spin_unlock(&state->state_lock);
1865         return ret;
1866 }
1867
1868 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1869 {
1870         struct nfs_delegation *delegation;
1871
1872         fmode &= FMODE_READ|FMODE_WRITE;
1873         rcu_read_lock();
1874         delegation = nfs4_get_valid_delegation(inode);
1875         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1876                 rcu_read_unlock();
1877                 return;
1878         }
1879         rcu_read_unlock();
1880         nfs4_inode_return_delegation(inode);
1881 }
1882
1883 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1884 {
1885         struct nfs4_state *state = opendata->state;
1886         struct nfs_delegation *delegation;
1887         int open_mode = opendata->o_arg.open_flags;
1888         fmode_t fmode = opendata->o_arg.fmode;
1889         enum open_claim_type4 claim = opendata->o_arg.claim;
1890         nfs4_stateid stateid;
1891         int ret = -EAGAIN;
1892
1893         for (;;) {
1894                 spin_lock(&state->owner->so_lock);
1895                 if (can_open_cached(state, fmode, open_mode, claim)) {
1896                         update_open_stateflags(state, fmode);
1897                         spin_unlock(&state->owner->so_lock);
1898                         goto out_return_state;
1899                 }
1900                 spin_unlock(&state->owner->so_lock);
1901                 rcu_read_lock();
1902                 delegation = nfs4_get_valid_delegation(state->inode);
1903                 if (!can_open_delegated(delegation, fmode, claim)) {
1904                         rcu_read_unlock();
1905                         break;
1906                 }
1907                 /* Save the delegation */
1908                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1909                 rcu_read_unlock();
1910                 nfs_release_seqid(opendata->o_arg.seqid);
1911                 if (!opendata->is_recover) {
1912                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1913                         if (ret != 0)
1914                                 goto out;
1915                 }
1916                 ret = -EAGAIN;
1917
1918                 /* Try to update the stateid using the delegation */
1919                 if (update_open_stateid(state, NULL, &stateid, fmode))
1920                         goto out_return_state;
1921         }
1922 out:
1923         return ERR_PTR(ret);
1924 out_return_state:
1925         refcount_inc(&state->count);
1926         return state;
1927 }
1928
1929 static void
1930 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1931 {
1932         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1933         struct nfs_delegation *delegation;
1934         int delegation_flags = 0;
1935
1936         rcu_read_lock();
1937         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1938         if (delegation)
1939                 delegation_flags = delegation->flags;
1940         rcu_read_unlock();
1941         switch (data->o_arg.claim) {
1942         default:
1943                 break;
1944         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1945         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1946                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1947                                    "returning a delegation for "
1948                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1949                                    clp->cl_hostname);
1950                 return;
1951         }
1952         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1953                 nfs_inode_set_delegation(state->inode,
1954                                 data->owner->so_cred,
1955                                 data->o_res.delegation_type,
1956                                 &data->o_res.delegation,
1957                                 data->o_res.pagemod_limit);
1958         else
1959                 nfs_inode_reclaim_delegation(state->inode,
1960                                 data->owner->so_cred,
1961                                 data->o_res.delegation_type,
1962                                 &data->o_res.delegation,
1963                                 data->o_res.pagemod_limit);
1964
1965         if (data->o_res.do_recall)
1966                 nfs_async_inode_return_delegation(state->inode,
1967                                                   &data->o_res.delegation);
1968 }
1969
1970 /*
1971  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1972  * and update the nfs4_state.
1973  */
1974 static struct nfs4_state *
1975 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1976 {
1977         struct inode *inode = data->state->inode;
1978         struct nfs4_state *state = data->state;
1979         int ret;
1980
1981         if (!data->rpc_done) {
1982                 if (data->rpc_status)
1983                         return ERR_PTR(data->rpc_status);
1984                 /* cached opens have already been processed */
1985                 goto update;
1986         }
1987
1988         ret = nfs_refresh_inode(inode, &data->f_attr);
1989         if (ret)
1990                 return ERR_PTR(ret);
1991
1992         if (data->o_res.delegation_type != 0)
1993                 nfs4_opendata_check_deleg(data, state);
1994 update:
1995         if (!update_open_stateid(state, &data->o_res.stateid,
1996                                 NULL, data->o_arg.fmode))
1997                 return ERR_PTR(-EAGAIN);
1998         refcount_inc(&state->count);
1999
2000         return state;
2001 }
2002
2003 static struct inode *
2004 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2005 {
2006         struct inode *inode;
2007
2008         switch (data->o_arg.claim) {
2009         case NFS4_OPEN_CLAIM_NULL:
2010         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2011         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2012                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2013                         return ERR_PTR(-EAGAIN);
2014                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2015                                 &data->f_attr, data->f_label);
2016                 break;
2017         default:
2018                 inode = d_inode(data->dentry);
2019                 ihold(inode);
2020                 nfs_refresh_inode(inode, &data->f_attr);
2021         }
2022         return inode;
2023 }
2024
2025 static struct nfs4_state *
2026 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2027 {
2028         struct nfs4_state *state;
2029         struct inode *inode;
2030
2031         inode = nfs4_opendata_get_inode(data);
2032         if (IS_ERR(inode))
2033                 return ERR_CAST(inode);
2034         if (data->state != NULL && data->state->inode == inode) {
2035                 state = data->state;
2036                 refcount_inc(&state->count);
2037         } else
2038                 state = nfs4_get_open_state(inode, data->owner);
2039         iput(inode);
2040         if (state == NULL)
2041                 state = ERR_PTR(-ENOMEM);
2042         return state;
2043 }
2044
2045 static struct nfs4_state *
2046 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2047 {
2048         struct nfs4_state *state;
2049
2050         if (!data->rpc_done) {
2051                 state = nfs4_try_open_cached(data);
2052                 trace_nfs4_cached_open(data->state);
2053                 goto out;
2054         }
2055
2056         state = nfs4_opendata_find_nfs4_state(data);
2057         if (IS_ERR(state))
2058                 goto out;
2059
2060         if (data->o_res.delegation_type != 0)
2061                 nfs4_opendata_check_deleg(data, state);
2062         if (!update_open_stateid(state, &data->o_res.stateid,
2063                                 NULL, data->o_arg.fmode)) {
2064                 nfs4_put_open_state(state);
2065                 state = ERR_PTR(-EAGAIN);
2066         }
2067 out:
2068         nfs_release_seqid(data->o_arg.seqid);
2069         return state;
2070 }
2071
2072 static struct nfs4_state *
2073 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2074 {
2075         struct nfs4_state *ret;
2076
2077         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2078                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2079         else
2080                 ret = _nfs4_opendata_to_nfs4_state(data);
2081         nfs4_sequence_free_slot(&data->o_res.seq_res);
2082         return ret;
2083 }
2084
2085 static struct nfs_open_context *
2086 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2087 {
2088         struct nfs_inode *nfsi = NFS_I(state->inode);
2089         struct nfs_open_context *ctx;
2090
2091         rcu_read_lock();
2092         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2093                 if (ctx->state != state)
2094                         continue;
2095                 if ((ctx->mode & mode) != mode)
2096                         continue;
2097                 if (!get_nfs_open_context(ctx))
2098                         continue;
2099                 rcu_read_unlock();
2100                 return ctx;
2101         }
2102         rcu_read_unlock();
2103         return ERR_PTR(-ENOENT);
2104 }
2105
2106 static struct nfs_open_context *
2107 nfs4_state_find_open_context(struct nfs4_state *state)
2108 {
2109         struct nfs_open_context *ctx;
2110
2111         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2112         if (!IS_ERR(ctx))
2113                 return ctx;
2114         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2115         if (!IS_ERR(ctx))
2116                 return ctx;
2117         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2118 }
2119
2120 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2121                 struct nfs4_state *state, enum open_claim_type4 claim)
2122 {
2123         struct nfs4_opendata *opendata;
2124
2125         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2126                         NULL, claim, GFP_NOFS);
2127         if (opendata == NULL)
2128                 return ERR_PTR(-ENOMEM);
2129         opendata->state = state;
2130         refcount_inc(&state->count);
2131         return opendata;
2132 }
2133
2134 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2135                 fmode_t fmode)
2136 {
2137         struct nfs4_state *newstate;
2138         int ret;
2139
2140         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2141                 return 0;
2142         opendata->o_arg.open_flags = 0;
2143         opendata->o_arg.fmode = fmode;
2144         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2145                         NFS_SB(opendata->dentry->d_sb),
2146                         fmode, 0);
2147         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2148         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2149         nfs4_init_opendata_res(opendata);
2150         ret = _nfs4_recover_proc_open(opendata);
2151         if (ret != 0)
2152                 return ret;
2153         newstate = nfs4_opendata_to_nfs4_state(opendata);
2154         if (IS_ERR(newstate))
2155                 return PTR_ERR(newstate);
2156         if (newstate != opendata->state)
2157                 ret = -ESTALE;
2158         nfs4_close_state(newstate, fmode);
2159         return ret;
2160 }
2161
2162 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2163 {
2164         int ret;
2165
2166         /* memory barrier prior to reading state->n_* */
2167         smp_rmb();
2168         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2169         if (ret != 0)
2170                 return ret;
2171         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2172         if (ret != 0)
2173                 return ret;
2174         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2175         if (ret != 0)
2176                 return ret;
2177         /*
2178          * We may have performed cached opens for all three recoveries.
2179          * Check if we need to update the current stateid.
2180          */
2181         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2182             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2183                 write_seqlock(&state->seqlock);
2184                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2185                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2186                 write_sequnlock(&state->seqlock);
2187         }
2188         return 0;
2189 }
2190
2191 /*
2192  * OPEN_RECLAIM:
2193  *      reclaim state on the server after a reboot.
2194  */
2195 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2196 {
2197         struct nfs_delegation *delegation;
2198         struct nfs4_opendata *opendata;
2199         fmode_t delegation_type = 0;
2200         int status;
2201
2202         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2203                         NFS4_OPEN_CLAIM_PREVIOUS);
2204         if (IS_ERR(opendata))
2205                 return PTR_ERR(opendata);
2206         rcu_read_lock();
2207         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2208         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2209                 delegation_type = delegation->type;
2210         rcu_read_unlock();
2211         opendata->o_arg.u.delegation_type = delegation_type;
2212         status = nfs4_open_recover(opendata, state);
2213         nfs4_opendata_put(opendata);
2214         return status;
2215 }
2216
2217 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2218 {
2219         struct nfs_server *server = NFS_SERVER(state->inode);
2220         struct nfs4_exception exception = { };
2221         int err;
2222         do {
2223                 err = _nfs4_do_open_reclaim(ctx, state);
2224                 trace_nfs4_open_reclaim(ctx, 0, err);
2225                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2226                         continue;
2227                 if (err != -NFS4ERR_DELAY)
2228                         break;
2229                 nfs4_handle_exception(server, err, &exception);
2230         } while (exception.retry);
2231         return err;
2232 }
2233
2234 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2235 {
2236         struct nfs_open_context *ctx;
2237         int ret;
2238
2239         ctx = nfs4_state_find_open_context(state);
2240         if (IS_ERR(ctx))
2241                 return -EAGAIN;
2242         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2243         nfs_state_clear_open_state_flags(state);
2244         ret = nfs4_do_open_reclaim(ctx, state);
2245         put_nfs_open_context(ctx);
2246         return ret;
2247 }
2248
2249 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)
2250 {
2251         switch (err) {
2252                 default:
2253                         printk(KERN_ERR "NFS: %s: unhandled error "
2254                                         "%d.\n", __func__, err);
2255                         fallthrough;
2256                 case 0:
2257                 case -ENOENT:
2258                 case -EAGAIN:
2259                 case -ESTALE:
2260                 case -ETIMEDOUT:
2261                         break;
2262                 case -NFS4ERR_BADSESSION:
2263                 case -NFS4ERR_BADSLOT:
2264                 case -NFS4ERR_BAD_HIGH_SLOT:
2265                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2266                 case -NFS4ERR_DEADSESSION:
2267                         return -EAGAIN;
2268                 case -NFS4ERR_STALE_CLIENTID:
2269                 case -NFS4ERR_STALE_STATEID:
2270                         /* Don't recall a delegation if it was lost */
2271                         nfs4_schedule_lease_recovery(server->nfs_client);
2272                         return -EAGAIN;
2273                 case -NFS4ERR_MOVED:
2274                         nfs4_schedule_migration_recovery(server);
2275                         return -EAGAIN;
2276                 case -NFS4ERR_LEASE_MOVED:
2277                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2278                         return -EAGAIN;
2279                 case -NFS4ERR_DELEG_REVOKED:
2280                 case -NFS4ERR_ADMIN_REVOKED:
2281                 case -NFS4ERR_EXPIRED:
2282                 case -NFS4ERR_BAD_STATEID:
2283                 case -NFS4ERR_OPENMODE:
2284                         nfs_inode_find_state_and_recover(state->inode,
2285                                         stateid);
2286                         nfs4_schedule_stateid_recovery(server, state);
2287                         return -EAGAIN;
2288                 case -NFS4ERR_DELAY:
2289                 case -NFS4ERR_GRACE:
2290                         ssleep(1);
2291                         return -EAGAIN;
2292                 case -ENOMEM:
2293                 case -NFS4ERR_DENIED:
2294                         if (fl) {
2295                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2296                                 if (lsp)
2297                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2298                         }
2299                         return 0;
2300         }
2301         return err;
2302 }
2303
2304 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2305                 struct nfs4_state *state, const nfs4_stateid *stateid)
2306 {
2307         struct nfs_server *server = NFS_SERVER(state->inode);
2308         struct nfs4_opendata *opendata;
2309         int err = 0;
2310
2311         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2312                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2313         if (IS_ERR(opendata))
2314                 return PTR_ERR(opendata);
2315         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2316         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2317                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2318                 if (err)
2319                         goto out;
2320         }
2321         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2322                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2323                 if (err)
2324                         goto out;
2325         }
2326         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2327                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2328                 if (err)
2329                         goto out;
2330         }
2331         nfs_state_clear_delegation(state);
2332 out:
2333         nfs4_opendata_put(opendata);
2334         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2335 }
2336
2337 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2338 {
2339         struct nfs4_opendata *data = calldata;
2340
2341         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2342                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2343 }
2344
2345 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2346 {
2347         struct nfs4_opendata *data = calldata;
2348
2349         nfs40_sequence_done(task, &data->c_res.seq_res);
2350
2351         data->rpc_status = task->tk_status;
2352         if (data->rpc_status == 0) {
2353                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2354                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2355                 renew_lease(data->o_res.server, data->timestamp);
2356                 data->rpc_done = true;
2357         }
2358 }
2359
2360 static void nfs4_open_confirm_release(void *calldata)
2361 {
2362         struct nfs4_opendata *data = calldata;
2363         struct nfs4_state *state = NULL;
2364
2365         /* If this request hasn't been cancelled, do nothing */
2366         if (!data->cancelled)
2367                 goto out_free;
2368         /* In case of error, no cleanup! */
2369         if (!data->rpc_done)
2370                 goto out_free;
2371         state = nfs4_opendata_to_nfs4_state(data);
2372         if (!IS_ERR(state))
2373                 nfs4_close_state(state, data->o_arg.fmode);
2374 out_free:
2375         nfs4_opendata_put(data);
2376 }
2377
2378 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2379         .rpc_call_prepare = nfs4_open_confirm_prepare,
2380         .rpc_call_done = nfs4_open_confirm_done,
2381         .rpc_release = nfs4_open_confirm_release,
2382 };
2383
2384 /*
2385  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2386  */
2387 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2388 {
2389         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2390         struct rpc_task *task;
2391         struct  rpc_message msg = {
2392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2393                 .rpc_argp = &data->c_arg,
2394                 .rpc_resp = &data->c_res,
2395                 .rpc_cred = data->owner->so_cred,
2396         };
2397         struct rpc_task_setup task_setup_data = {
2398                 .rpc_client = server->client,
2399                 .rpc_message = &msg,
2400                 .callback_ops = &nfs4_open_confirm_ops,
2401                 .callback_data = data,
2402                 .workqueue = nfsiod_workqueue,
2403                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2404         };
2405         int status;
2406
2407         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2408                                 data->is_recover);
2409         kref_get(&data->kref);
2410         data->rpc_done = false;
2411         data->rpc_status = 0;
2412         data->timestamp = jiffies;
2413         task = rpc_run_task(&task_setup_data);
2414         if (IS_ERR(task))
2415                 return PTR_ERR(task);
2416         status = rpc_wait_for_completion_task(task);
2417         if (status != 0) {
2418                 data->cancelled = true;
2419                 smp_wmb();
2420         } else
2421                 status = data->rpc_status;
2422         rpc_put_task(task);
2423         return status;
2424 }
2425
2426 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2427 {
2428         struct nfs4_opendata *data = calldata;
2429         struct nfs4_state_owner *sp = data->owner;
2430         struct nfs_client *clp = sp->so_server->nfs_client;
2431         enum open_claim_type4 claim = data->o_arg.claim;
2432
2433         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2434                 goto out_wait;
2435         /*
2436          * Check if we still need to send an OPEN call, or if we can use
2437          * a delegation instead.
2438          */
2439         if (data->state != NULL) {
2440                 struct nfs_delegation *delegation;
2441
2442                 if (can_open_cached(data->state, data->o_arg.fmode,
2443                                         data->o_arg.open_flags, claim))
2444                         goto out_no_action;
2445                 rcu_read_lock();
2446                 delegation = nfs4_get_valid_delegation(data->state->inode);
2447                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2448                         goto unlock_no_action;
2449                 rcu_read_unlock();
2450         }
2451         /* Update client id. */
2452         data->o_arg.clientid = clp->cl_clientid;
2453         switch (claim) {
2454         default:
2455                 break;
2456         case NFS4_OPEN_CLAIM_PREVIOUS:
2457         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2458         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2459                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2460                 fallthrough;
2461         case NFS4_OPEN_CLAIM_FH:
2462                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2463         }
2464         data->timestamp = jiffies;
2465         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2466                                 &data->o_arg.seq_args,
2467                                 &data->o_res.seq_res,
2468                                 task) != 0)
2469                 nfs_release_seqid(data->o_arg.seqid);
2470
2471         /* Set the create mode (note dependency on the session type) */
2472         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2473         if (data->o_arg.open_flags & O_EXCL) {
2474                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2475                 if (nfs4_has_persistent_session(clp))
2476                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2477                 else if (clp->cl_mvops->minor_version > 0)
2478                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2479         }
2480         return;
2481 unlock_no_action:
2482         trace_nfs4_cached_open(data->state);
2483         rcu_read_unlock();
2484 out_no_action:
2485         task->tk_action = NULL;
2486 out_wait:
2487         nfs4_sequence_done(task, &data->o_res.seq_res);
2488 }
2489
2490 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2491 {
2492         struct nfs4_opendata *data = calldata;
2493
2494         data->rpc_status = task->tk_status;
2495
2496         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2497                 return;
2498
2499         if (task->tk_status == 0) {
2500                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2501                         switch (data->o_res.f_attr->mode & S_IFMT) {
2502                         case S_IFREG:
2503                                 break;
2504                         case S_IFLNK:
2505                                 data->rpc_status = -ELOOP;
2506                                 break;
2507                         case S_IFDIR:
2508                                 data->rpc_status = -EISDIR;
2509                                 break;
2510                         default:
2511                                 data->rpc_status = -ENOTDIR;
2512                         }
2513                 }
2514                 renew_lease(data->o_res.server, data->timestamp);
2515                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2516                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2517         }
2518         data->rpc_done = true;
2519 }
2520
2521 static void nfs4_open_release(void *calldata)
2522 {
2523         struct nfs4_opendata *data = calldata;
2524         struct nfs4_state *state = NULL;
2525
2526         /* If this request hasn't been cancelled, do nothing */
2527         if (!data->cancelled)
2528                 goto out_free;
2529         /* In case of error, no cleanup! */
2530         if (data->rpc_status != 0 || !data->rpc_done)
2531                 goto out_free;
2532         /* In case we need an open_confirm, no cleanup! */
2533         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2534                 goto out_free;
2535         state = nfs4_opendata_to_nfs4_state(data);
2536         if (!IS_ERR(state))
2537                 nfs4_close_state(state, data->o_arg.fmode);
2538 out_free:
2539         nfs4_opendata_put(data);
2540 }
2541
2542 static const struct rpc_call_ops nfs4_open_ops = {
2543         .rpc_call_prepare = nfs4_open_prepare,
2544         .rpc_call_done = nfs4_open_done,
2545         .rpc_release = nfs4_open_release,
2546 };
2547
2548 static int nfs4_run_open_task(struct nfs4_opendata *data,
2549                               struct nfs_open_context *ctx)
2550 {
2551         struct inode *dir = d_inode(data->dir);
2552         struct nfs_server *server = NFS_SERVER(dir);
2553         struct nfs_openargs *o_arg = &data->o_arg;
2554         struct nfs_openres *o_res = &data->o_res;
2555         struct rpc_task *task;
2556         struct rpc_message msg = {
2557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2558                 .rpc_argp = o_arg,
2559                 .rpc_resp = o_res,
2560                 .rpc_cred = data->owner->so_cred,
2561         };
2562         struct rpc_task_setup task_setup_data = {
2563                 .rpc_client = server->client,
2564                 .rpc_message = &msg,
2565                 .callback_ops = &nfs4_open_ops,
2566                 .callback_data = data,
2567                 .workqueue = nfsiod_workqueue,
2568                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2569         };
2570         int status;
2571
2572         kref_get(&data->kref);
2573         data->rpc_done = false;
2574         data->rpc_status = 0;
2575         data->cancelled = false;
2576         data->is_recover = false;
2577         if (!ctx) {
2578                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2579                 data->is_recover = true;
2580                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2581         } else {
2582                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2583                 pnfs_lgopen_prepare(data, ctx);
2584         }
2585         task = rpc_run_task(&task_setup_data);
2586         if (IS_ERR(task))
2587                 return PTR_ERR(task);
2588         status = rpc_wait_for_completion_task(task);
2589         if (status != 0) {
2590                 data->cancelled = true;
2591                 smp_wmb();
2592         } else
2593                 status = data->rpc_status;
2594         rpc_put_task(task);
2595
2596         return status;
2597 }
2598
2599 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2600 {
2601         struct inode *dir = d_inode(data->dir);
2602         struct nfs_openres *o_res = &data->o_res;
2603         int status;
2604
2605         status = nfs4_run_open_task(data, NULL);
2606         if (status != 0 || !data->rpc_done)
2607                 return status;
2608
2609         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2610
2611         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2612                 status = _nfs4_proc_open_confirm(data);
2613
2614         return status;
2615 }
2616
2617 /*
2618  * Additional permission checks in order to distinguish between an
2619  * open for read, and an open for execute. This works around the
2620  * fact that NFSv4 OPEN treats read and execute permissions as being
2621  * the same.
2622  * Note that in the non-execute case, we want to turn off permission
2623  * checking if we just created a new file (POSIX open() semantics).
2624  */
2625 static int nfs4_opendata_access(const struct cred *cred,
2626                                 struct nfs4_opendata *opendata,
2627                                 struct nfs4_state *state, fmode_t fmode,
2628                                 int openflags)
2629 {
2630         struct nfs_access_entry cache;
2631         u32 mask, flags;
2632
2633         /* access call failed or for some reason the server doesn't
2634          * support any access modes -- defer access call until later */
2635         if (opendata->o_res.access_supported == 0)
2636                 return 0;
2637
2638         mask = 0;
2639         /*
2640          * Use openflags to check for exec, because fmode won't
2641          * always have FMODE_EXEC set when file open for exec.
2642          */
2643         if (openflags & __FMODE_EXEC) {
2644                 /* ONLY check for exec rights */
2645                 if (S_ISDIR(state->inode->i_mode))
2646                         mask = NFS4_ACCESS_LOOKUP;
2647                 else
2648                         mask = NFS4_ACCESS_EXECUTE;
2649         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2650                 mask = NFS4_ACCESS_READ;
2651
2652         cache.cred = cred;
2653         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2654         nfs_access_add_cache(state->inode, &cache);
2655
2656         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2657         if ((mask & ~cache.mask & flags) == 0)
2658                 return 0;
2659
2660         return -EACCES;
2661 }
2662
2663 /*
2664  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2665  */
2666 static int _nfs4_proc_open(struct nfs4_opendata *data,
2667                            struct nfs_open_context *ctx)
2668 {
2669         struct inode *dir = d_inode(data->dir);
2670         struct nfs_server *server = NFS_SERVER(dir);
2671         struct nfs_openargs *o_arg = &data->o_arg;
2672         struct nfs_openres *o_res = &data->o_res;
2673         int status;
2674
2675         status = nfs4_run_open_task(data, ctx);
2676         if (!data->rpc_done)
2677                 return status;
2678         if (status != 0) {
2679                 if (status == -NFS4ERR_BADNAME &&
2680                                 !(o_arg->open_flags & O_CREAT))
2681                         return -ENOENT;
2682                 return status;
2683         }
2684
2685         nfs_fattr_map_and_free_names(server, &data->f_attr);
2686
2687         if (o_arg->open_flags & O_CREAT) {
2688                 if (o_arg->open_flags & O_EXCL)
2689                         data->file_created = true;
2690                 else if (o_res->cinfo.before != o_res->cinfo.after)
2691                         data->file_created = true;
2692                 if (data->file_created ||
2693                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2694                         nfs4_update_changeattr(dir, &o_res->cinfo,
2695                                         o_res->f_attr->time_start,
2696                                         NFS_INO_INVALID_DATA);
2697         }
2698         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2699                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2700         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2701                 status = _nfs4_proc_open_confirm(data);
2702                 if (status != 0)
2703                         return status;
2704         }
2705         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2706                 nfs4_sequence_free_slot(&o_res->seq_res);
2707                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2708                                 o_res->f_label, NULL);
2709         }
2710         return 0;
2711 }
2712
2713 /*
2714  * OPEN_EXPIRED:
2715  *      reclaim state on the server after a network partition.
2716  *      Assumes caller holds the appropriate lock
2717  */
2718 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2719 {
2720         struct nfs4_opendata *opendata;
2721         int ret;
2722
2723         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2724                         NFS4_OPEN_CLAIM_FH);
2725         if (IS_ERR(opendata))
2726                 return PTR_ERR(opendata);
2727         ret = nfs4_open_recover(opendata, state);
2728         if (ret == -ESTALE)
2729                 d_drop(ctx->dentry);
2730         nfs4_opendata_put(opendata);
2731         return ret;
2732 }
2733
2734 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2735 {
2736         struct nfs_server *server = NFS_SERVER(state->inode);
2737         struct nfs4_exception exception = { };
2738         int err;
2739
2740         do {
2741                 err = _nfs4_open_expired(ctx, state);
2742                 trace_nfs4_open_expired(ctx, 0, err);
2743                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2744                         continue;
2745                 switch (err) {
2746                 default:
2747                         goto out;
2748                 case -NFS4ERR_GRACE:
2749                 case -NFS4ERR_DELAY:
2750                         nfs4_handle_exception(server, err, &exception);
2751                         err = 0;
2752                 }
2753         } while (exception.retry);
2754 out:
2755         return err;
2756 }
2757
2758 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2759 {
2760         struct nfs_open_context *ctx;
2761         int ret;
2762
2763         ctx = nfs4_state_find_open_context(state);
2764         if (IS_ERR(ctx))
2765                 return -EAGAIN;
2766         ret = nfs4_do_open_expired(ctx, state);
2767         put_nfs_open_context(ctx);
2768         return ret;
2769 }
2770
2771 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2772                 const nfs4_stateid *stateid)
2773 {
2774         nfs_remove_bad_delegation(state->inode, stateid);
2775         nfs_state_clear_delegation(state);
2776 }
2777
2778 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2779 {
2780         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2781                 nfs_finish_clear_delegation_stateid(state, NULL);
2782 }
2783
2784 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2785 {
2786         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2787         nfs40_clear_delegation_stateid(state);
2788         nfs_state_clear_open_state_flags(state);
2789         return nfs4_open_expired(sp, state);
2790 }
2791
2792 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2793                 nfs4_stateid *stateid,
2794                 const struct cred *cred)
2795 {
2796         return -NFS4ERR_BAD_STATEID;
2797 }
2798
2799 #if defined(CONFIG_NFS_V4_1)
2800 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2801                 nfs4_stateid *stateid,
2802                 const struct cred *cred)
2803 {
2804         int status;
2805
2806         switch (stateid->type) {
2807         default:
2808                 break;
2809         case NFS4_INVALID_STATEID_TYPE:
2810         case NFS4_SPECIAL_STATEID_TYPE:
2811                 return -NFS4ERR_BAD_STATEID;
2812         case NFS4_REVOKED_STATEID_TYPE:
2813                 goto out_free;
2814         }
2815
2816         status = nfs41_test_stateid(server, stateid, cred);
2817         switch (status) {
2818         case -NFS4ERR_EXPIRED:
2819         case -NFS4ERR_ADMIN_REVOKED:
2820         case -NFS4ERR_DELEG_REVOKED:
2821                 break;
2822         default:
2823                 return status;
2824         }
2825 out_free:
2826         /* Ack the revoked state to the server */
2827         nfs41_free_stateid(server, stateid, cred, true);
2828         return -NFS4ERR_EXPIRED;
2829 }
2830
2831 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2832 {
2833         struct nfs_server *server = NFS_SERVER(state->inode);
2834         nfs4_stateid stateid;
2835         struct nfs_delegation *delegation;
2836         const struct cred *cred = NULL;
2837         int status, ret = NFS_OK;
2838
2839         /* Get the delegation credential for use by test/free_stateid */
2840         rcu_read_lock();
2841         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2842         if (delegation == NULL) {
2843                 rcu_read_unlock();
2844                 nfs_state_clear_delegation(state);
2845                 return NFS_OK;
2846         }
2847
2848         spin_lock(&delegation->lock);
2849         nfs4_stateid_copy(&stateid, &delegation->stateid);
2850
2851         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2852                                 &delegation->flags)) {
2853                 spin_unlock(&delegation->lock);
2854                 rcu_read_unlock();
2855                 return NFS_OK;
2856         }
2857
2858         if (delegation->cred)
2859                 cred = get_cred(delegation->cred);
2860         spin_unlock(&delegation->lock);
2861         rcu_read_unlock();
2862         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2863         trace_nfs4_test_delegation_stateid(state, NULL, status);
2864         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2865                 nfs_finish_clear_delegation_stateid(state, &stateid);
2866         else
2867                 ret = status;
2868
2869         put_cred(cred);
2870         return ret;
2871 }
2872
2873 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2874 {
2875         nfs4_stateid tmp;
2876
2877         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2878             nfs4_copy_delegation_stateid(state->inode, state->state,
2879                                 &tmp, NULL) &&
2880             nfs4_stateid_match_other(&state->stateid, &tmp))
2881                 nfs_state_set_delegation(state, &tmp, state->state);
2882         else
2883                 nfs_state_clear_delegation(state);
2884 }
2885
2886 /**
2887  * nfs41_check_expired_locks - possibly free a lock stateid
2888  *
2889  * @state: NFSv4 state for an inode
2890  *
2891  * Returns NFS_OK if recovery for this stateid is now finished.
2892  * Otherwise a negative NFS4ERR value is returned.
2893  */
2894 static int nfs41_check_expired_locks(struct nfs4_state *state)
2895 {
2896         int status, ret = NFS_OK;
2897         struct nfs4_lock_state *lsp, *prev = NULL;
2898         struct nfs_server *server = NFS_SERVER(state->inode);
2899
2900         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2901                 goto out;
2902
2903         spin_lock(&state->state_lock);
2904         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2905                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2906                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2907
2908                         refcount_inc(&lsp->ls_count);
2909                         spin_unlock(&state->state_lock);
2910
2911                         nfs4_put_lock_state(prev);
2912                         prev = lsp;
2913
2914                         status = nfs41_test_and_free_expired_stateid(server,
2915                                         &lsp->ls_stateid,
2916                                         cred);
2917                         trace_nfs4_test_lock_stateid(state, lsp, status);
2918                         if (status == -NFS4ERR_EXPIRED ||
2919                             status == -NFS4ERR_BAD_STATEID) {
2920                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2921                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2922                                 if (!recover_lost_locks)
2923                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2924                         } else if (status != NFS_OK) {
2925                                 ret = status;
2926                                 nfs4_put_lock_state(prev);
2927                                 goto out;
2928                         }
2929                         spin_lock(&state->state_lock);
2930                 }
2931         }
2932         spin_unlock(&state->state_lock);
2933         nfs4_put_lock_state(prev);
2934 out:
2935         return ret;
2936 }
2937
2938 /**
2939  * nfs41_check_open_stateid - possibly free an open stateid
2940  *
2941  * @state: NFSv4 state for an inode
2942  *
2943  * Returns NFS_OK if recovery for this stateid is now finished.
2944  * Otherwise a negative NFS4ERR value is returned.
2945  */
2946 static int nfs41_check_open_stateid(struct nfs4_state *state)
2947 {
2948         struct nfs_server *server = NFS_SERVER(state->inode);
2949         nfs4_stateid *stateid = &state->open_stateid;
2950         const struct cred *cred = state->owner->so_cred;
2951         int status;
2952
2953         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2954                 return -NFS4ERR_BAD_STATEID;
2955         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2956         trace_nfs4_test_open_stateid(state, NULL, status);
2957         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2958                 nfs_state_clear_open_state_flags(state);
2959                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2960                 return status;
2961         }
2962         if (nfs_open_stateid_recover_openmode(state))
2963                 return -NFS4ERR_OPENMODE;
2964         return NFS_OK;
2965 }
2966
2967 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2968 {
2969         int status;
2970
2971         status = nfs41_check_delegation_stateid(state);
2972         if (status != NFS_OK)
2973                 return status;
2974         nfs41_delegation_recover_stateid(state);
2975
2976         status = nfs41_check_expired_locks(state);
2977         if (status != NFS_OK)
2978                 return status;
2979         status = nfs41_check_open_stateid(state);
2980         if (status != NFS_OK)
2981                 status = nfs4_open_expired(sp, state);
2982         return status;
2983 }
2984 #endif
2985
2986 /*
2987  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2988  * fields corresponding to attributes that were used to store the verifier.
2989  * Make sure we clobber those fields in the later setattr call
2990  */
2991 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2992                                 struct iattr *sattr, struct nfs4_label **label)
2993 {
2994         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2995         __u32 attrset[3];
2996         unsigned ret;
2997         unsigned i;
2998
2999         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3000                 attrset[i] = opendata->o_res.attrset[i];
3001                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3002                         attrset[i] &= ~bitmask[i];
3003         }
3004
3005         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3006                 sattr->ia_valid : 0;
3007
3008         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3009                 if (sattr->ia_valid & ATTR_ATIME_SET)
3010                         ret |= ATTR_ATIME_SET;
3011                 else
3012                         ret |= ATTR_ATIME;
3013         }
3014
3015         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3016                 if (sattr->ia_valid & ATTR_MTIME_SET)
3017                         ret |= ATTR_MTIME_SET;
3018                 else
3019                         ret |= ATTR_MTIME;
3020         }
3021
3022         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3023                 *label = NULL;
3024         return ret;
3025 }
3026
3027 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3028                 int flags, struct nfs_open_context *ctx)
3029 {
3030         struct nfs4_state_owner *sp = opendata->owner;
3031         struct nfs_server *server = sp->so_server;
3032         struct dentry *dentry;
3033         struct nfs4_state *state;
3034         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3035         struct inode *dir = d_inode(opendata->dir);
3036         unsigned long dir_verifier;
3037         unsigned int seq;
3038         int ret;
3039
3040         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3041         dir_verifier = nfs_save_change_attribute(dir);
3042
3043         ret = _nfs4_proc_open(opendata, ctx);
3044         if (ret != 0)
3045                 goto out;
3046
3047         state = _nfs4_opendata_to_nfs4_state(opendata);
3048         ret = PTR_ERR(state);
3049         if (IS_ERR(state))
3050                 goto out;
3051         ctx->state = state;
3052         if (server->caps & NFS_CAP_POSIX_LOCK)
3053                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3054         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3055                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3056
3057         dentry = opendata->dentry;
3058         if (d_really_is_negative(dentry)) {
3059                 struct dentry *alias;
3060                 d_drop(dentry);
3061                 alias = d_exact_alias(dentry, state->inode);
3062                 if (!alias)
3063                         alias = d_splice_alias(igrab(state->inode), dentry);
3064                 /* d_splice_alias() can't fail here - it's a non-directory */
3065                 if (alias) {
3066                         dput(ctx->dentry);
3067                         ctx->dentry = dentry = alias;
3068                 }
3069         }
3070
3071         switch(opendata->o_arg.claim) {
3072         default:
3073                 break;
3074         case NFS4_OPEN_CLAIM_NULL:
3075         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3076         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3077                 if (!opendata->rpc_done)
3078                         break;
3079                 if (opendata->o_res.delegation_type != 0)
3080                         dir_verifier = nfs_save_change_attribute(dir);
3081                 nfs_set_verifier(dentry, dir_verifier);
3082         }
3083
3084         /* Parse layoutget results before we check for access */
3085         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3086
3087         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3088                         acc_mode, flags);
3089         if (ret != 0)
3090                 goto out;
3091
3092         if (d_inode(dentry) == state->inode) {
3093                 nfs_inode_attach_open_context(ctx);
3094                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3095                         nfs4_schedule_stateid_recovery(server, state);
3096         }
3097
3098 out:
3099         if (!opendata->cancelled)
3100                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3101         return ret;
3102 }
3103
3104 /*
3105  * Returns a referenced nfs4_state
3106  */
3107 static int _nfs4_do_open(struct inode *dir,
3108                         struct nfs_open_context *ctx,
3109                         int flags,
3110                         const struct nfs4_open_createattrs *c,
3111                         int *opened)
3112 {
3113         struct nfs4_state_owner  *sp;
3114         struct nfs4_state     *state = NULL;
3115         struct nfs_server       *server = NFS_SERVER(dir);
3116         struct nfs4_opendata *opendata;
3117         struct dentry *dentry = ctx->dentry;
3118         const struct cred *cred = ctx->cred;
3119         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3120         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3121         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3122         struct iattr *sattr = c->sattr;
3123         struct nfs4_label *label = c->label;
3124         struct nfs4_label *olabel = NULL;
3125         int status;
3126
3127         /* Protect against reboot recovery conflicts */
3128         status = -ENOMEM;
3129         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3130         if (sp == NULL) {
3131                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3132                 goto out_err;
3133         }
3134         status = nfs4_client_recover_expired_lease(server->nfs_client);
3135         if (status != 0)
3136                 goto err_put_state_owner;
3137         if (d_really_is_positive(dentry))
3138                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3139         status = -ENOMEM;
3140         if (d_really_is_positive(dentry))
3141                 claim = NFS4_OPEN_CLAIM_FH;
3142         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3143                         c, claim, GFP_KERNEL);
3144         if (opendata == NULL)
3145                 goto err_put_state_owner;
3146
3147         if (label) {
3148                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3149                 if (IS_ERR(olabel)) {
3150                         status = PTR_ERR(olabel);
3151                         goto err_opendata_put;
3152                 }
3153         }
3154
3155         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3156                 if (!opendata->f_attr.mdsthreshold) {
3157                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3158                         if (!opendata->f_attr.mdsthreshold)
3159                                 goto err_free_label;
3160                 }
3161                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3162         }
3163         if (d_really_is_positive(dentry))
3164                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3165
3166         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3167         if (status != 0)
3168                 goto err_free_label;
3169         state = ctx->state;
3170
3171         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3172             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3173                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3174                 /*
3175                  * send create attributes which was not set by open
3176                  * with an extra setattr.
3177                  */
3178                 if (attrs || label) {
3179                         unsigned ia_old = sattr->ia_valid;
3180
3181                         sattr->ia_valid = attrs;
3182                         nfs_fattr_init(opendata->o_res.f_attr);
3183                         status = nfs4_do_setattr(state->inode, cred,
3184                                         opendata->o_res.f_attr, sattr,
3185                                         ctx, label, olabel);
3186                         if (status == 0) {
3187                                 nfs_setattr_update_inode(state->inode, sattr,
3188                                                 opendata->o_res.f_attr);
3189                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3190                         }
3191                         sattr->ia_valid = ia_old;
3192                 }
3193         }
3194         if (opened && opendata->file_created)
3195                 *opened = 1;
3196
3197         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3198                 *ctx_th = opendata->f_attr.mdsthreshold;
3199                 opendata->f_attr.mdsthreshold = NULL;
3200         }
3201
3202         nfs4_label_free(olabel);
3203
3204         nfs4_opendata_put(opendata);
3205         nfs4_put_state_owner(sp);
3206         return 0;
3207 err_free_label:
3208         nfs4_label_free(olabel);
3209 err_opendata_put:
3210         nfs4_opendata_put(opendata);
3211 err_put_state_owner:
3212         nfs4_put_state_owner(sp);
3213 out_err:
3214         return status;
3215 }
3216
3217
3218 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3219                                         struct nfs_open_context *ctx,
3220                                         int flags,
3221                                         struct iattr *sattr,
3222                                         struct nfs4_label *label,
3223                                         int *opened)
3224 {
3225         struct nfs_server *server = NFS_SERVER(dir);
3226         struct nfs4_exception exception = {
3227                 .interruptible = true,
3228         };
3229         struct nfs4_state *res;
3230         struct nfs4_open_createattrs c = {
3231                 .label = label,
3232                 .sattr = sattr,
3233                 .verf = {
3234                         [0] = (__u32)jiffies,
3235                         [1] = (__u32)current->pid,
3236                 },
3237         };
3238         int status;
3239
3240         do {
3241                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3242                 res = ctx->state;
3243                 trace_nfs4_open_file(ctx, flags, status);
3244                 if (status == 0)
3245                         break;
3246                 /* NOTE: BAD_SEQID means the server and client disagree about the
3247                  * book-keeping w.r.t. state-changing operations
3248                  * (OPEN/CLOSE/LOCK/LOCKU...)
3249                  * It is actually a sign of a bug on the client or on the server.
3250                  *
3251                  * If we receive a BAD_SEQID error in the particular case of
3252                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3253                  * have unhashed the old state_owner for us, and that we can
3254                  * therefore safely retry using a new one. We should still warn
3255                  * the user though...
3256                  */
3257                 if (status == -NFS4ERR_BAD_SEQID) {
3258                         pr_warn_ratelimited("NFS: v4 server %s "
3259                                         " returned a bad sequence-id error!\n",
3260                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3261                         exception.retry = 1;
3262                         continue;
3263                 }
3264                 /*
3265                  * BAD_STATEID on OPEN means that the server cancelled our
3266                  * state before it received the OPEN_CONFIRM.
3267                  * Recover by retrying the request as per the discussion
3268                  * on Page 181 of RFC3530.
3269                  */
3270                 if (status == -NFS4ERR_BAD_STATEID) {
3271                         exception.retry = 1;
3272                         continue;
3273                 }
3274                 if (status == -NFS4ERR_EXPIRED) {
3275                         nfs4_schedule_lease_recovery(server->nfs_client);
3276                         exception.retry = 1;
3277                         continue;
3278                 }
3279                 if (status == -EAGAIN) {
3280                         /* We must have found a delegation */
3281                         exception.retry = 1;
3282                         continue;
3283                 }
3284                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3285                         continue;
3286                 res = ERR_PTR(nfs4_handle_exception(server,
3287                                         status, &exception));
3288         } while (exception.retry);
3289         return res;
3290 }
3291
3292 static int _nfs4_do_setattr(struct inode *inode,
3293                             struct nfs_setattrargs *arg,
3294                             struct nfs_setattrres *res,
3295                             const struct cred *cred,
3296                             struct nfs_open_context *ctx)
3297 {
3298         struct nfs_server *server = NFS_SERVER(inode);
3299         struct rpc_message msg = {
3300                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3301                 .rpc_argp       = arg,
3302                 .rpc_resp       = res,
3303                 .rpc_cred       = cred,
3304         };
3305         const struct cred *delegation_cred = NULL;
3306         unsigned long timestamp = jiffies;
3307         bool truncate;
3308         int status;
3309
3310         nfs_fattr_init(res->fattr);
3311
3312         /* Servers should only apply open mode checks for file size changes */
3313         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3314         if (!truncate) {
3315                 nfs4_inode_make_writeable(inode);
3316                 goto zero_stateid;
3317         }
3318
3319         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3320                 /* Use that stateid */
3321         } else if (ctx != NULL && ctx->state) {
3322                 struct nfs_lock_context *l_ctx;
3323                 if (!nfs4_valid_open_stateid(ctx->state))
3324                         return -EBADF;
3325                 l_ctx = nfs_get_lock_context(ctx);
3326                 if (IS_ERR(l_ctx))
3327                         return PTR_ERR(l_ctx);
3328                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3329                                                 &arg->stateid, &delegation_cred);
3330                 nfs_put_lock_context(l_ctx);
3331                 if (status == -EIO)
3332                         return -EBADF;
3333                 else if (status == -EAGAIN)
3334                         goto zero_stateid;
3335         } else {
3336 zero_stateid:
3337                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3338         }
3339         if (delegation_cred)
3340                 msg.rpc_cred = delegation_cred;
3341
3342         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3343
3344         put_cred(delegation_cred);
3345         if (status == 0 && ctx != NULL)
3346                 renew_lease(server, timestamp);
3347         trace_nfs4_setattr(inode, &arg->stateid, status);
3348         return status;
3349 }
3350
3351 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3352                            struct nfs_fattr *fattr, struct iattr *sattr,
3353                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3354                            struct nfs4_label *olabel)
3355 {
3356         struct nfs_server *server = NFS_SERVER(inode);
3357         __u32 bitmask[NFS4_BITMASK_SZ];
3358         struct nfs4_state *state = ctx ? ctx->state : NULL;
3359         struct nfs_setattrargs  arg = {
3360                 .fh             = NFS_FH(inode),
3361                 .iap            = sattr,
3362                 .server         = server,
3363                 .bitmask = bitmask,
3364                 .label          = ilabel,
3365         };
3366         struct nfs_setattrres  res = {
3367                 .fattr          = fattr,
3368                 .label          = olabel,
3369                 .server         = server,
3370         };
3371         struct nfs4_exception exception = {
3372                 .state = state,
3373                 .inode = inode,
3374                 .stateid = &arg.stateid,
3375         };
3376         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3377         int err;
3378
3379         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3380                 adjust_flags |= NFS_INO_INVALID_MODE;
3381         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3382                 adjust_flags |= NFS_INO_INVALID_OTHER;
3383
3384         do {
3385                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, olabel),
3386                                         inode, adjust_flags);
3387
3388                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3389                 switch (err) {
3390                 case -NFS4ERR_OPENMODE:
3391                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3392                                 pr_warn_once("NFSv4: server %s is incorrectly "
3393                                                 "applying open mode checks to "
3394                                                 "a SETATTR that is not "
3395                                                 "changing file size.\n",
3396                                                 server->nfs_client->cl_hostname);
3397                         }
3398                         if (state && !(state->state & FMODE_WRITE)) {
3399                                 err = -EBADF;
3400                                 if (sattr->ia_valid & ATTR_OPEN)
3401                                         err = -EACCES;
3402                                 goto out;
3403                         }
3404                 }
3405                 err = nfs4_handle_exception(server, err, &exception);
3406         } while (exception.retry);
3407 out:
3408         return err;
3409 }
3410
3411 static bool
3412 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3413 {
3414         if (inode == NULL || !nfs_have_layout(inode))
3415                 return false;
3416
3417         return pnfs_wait_on_layoutreturn(inode, task);
3418 }
3419
3420 /*
3421  * Update the seqid of an open stateid
3422  */
3423 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3424                 struct nfs4_state *state)
3425 {
3426         __be32 seqid_open;
3427         u32 dst_seqid;
3428         int seq;
3429
3430         for (;;) {
3431                 if (!nfs4_valid_open_stateid(state))
3432                         break;
3433                 seq = read_seqbegin(&state->seqlock);
3434                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3435                         nfs4_stateid_copy(dst, &state->open_stateid);
3436                         if (read_seqretry(&state->seqlock, seq))
3437                                 continue;
3438                         break;
3439                 }
3440                 seqid_open = state->open_stateid.seqid;
3441                 if (read_seqretry(&state->seqlock, seq))
3442                         continue;
3443
3444                 dst_seqid = be32_to_cpu(dst->seqid);
3445                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3446                         dst->seqid = seqid_open;
3447                 break;
3448         }
3449 }
3450
3451 /*
3452  * Update the seqid of an open stateid after receiving
3453  * NFS4ERR_OLD_STATEID
3454  */
3455 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3456                 struct nfs4_state *state)
3457 {
3458         __be32 seqid_open;
3459         u32 dst_seqid;
3460         bool ret;
3461         int seq, status = -EAGAIN;
3462         DEFINE_WAIT(wait);
3463
3464         for (;;) {
3465                 ret = false;
3466                 if (!nfs4_valid_open_stateid(state))
3467                         break;
3468                 seq = read_seqbegin(&state->seqlock);
3469                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3470                         if (read_seqretry(&state->seqlock, seq))
3471                                 continue;
3472                         break;
3473                 }
3474
3475                 write_seqlock(&state->seqlock);
3476                 seqid_open = state->open_stateid.seqid;
3477
3478                 dst_seqid = be32_to_cpu(dst->seqid);
3479
3480                 /* Did another OPEN bump the state's seqid?  try again: */
3481                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3482                         dst->seqid = seqid_open;
3483                         write_sequnlock(&state->seqlock);
3484                         ret = true;
3485                         break;
3486                 }
3487
3488                 /* server says we're behind but we haven't seen the update yet */
3489                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3490                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3491                 write_sequnlock(&state->seqlock);
3492                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3493
3494                 if (fatal_signal_pending(current))
3495                         status = -EINTR;
3496                 else
3497                         if (schedule_timeout(5*HZ) != 0)
3498                                 status = 0;
3499
3500                 finish_wait(&state->waitq, &wait);
3501
3502                 if (!status)
3503                         continue;
3504                 if (status == -EINTR)
3505                         break;
3506
3507                 /* we slept the whole 5 seconds, we must have lost a seqid */
3508                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3509                 ret = true;
3510                 break;
3511         }
3512
3513         return ret;
3514 }
3515
3516 struct nfs4_closedata {
3517         struct inode *inode;
3518         struct nfs4_state *state;
3519         struct nfs_closeargs arg;
3520         struct nfs_closeres res;
3521         struct {
3522                 struct nfs4_layoutreturn_args arg;
3523                 struct nfs4_layoutreturn_res res;
3524                 struct nfs4_xdr_opaque_data ld_private;
3525                 u32 roc_barrier;
3526                 bool roc;
3527         } lr;
3528         struct nfs_fattr fattr;
3529         unsigned long timestamp;
3530 };
3531
3532 static void nfs4_free_closedata(void *data)
3533 {
3534         struct nfs4_closedata *calldata = data;
3535         struct nfs4_state_owner *sp = calldata->state->owner;
3536         struct super_block *sb = calldata->state->inode->i_sb;
3537
3538         if (calldata->lr.roc)
3539                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3540                                 calldata->res.lr_ret);
3541         nfs4_put_open_state(calldata->state);
3542         nfs_free_seqid(calldata->arg.seqid);
3543         nfs4_put_state_owner(sp);
3544         nfs_sb_deactive(sb);
3545         kfree(calldata);
3546 }
3547
3548 static void nfs4_close_done(struct rpc_task *task, void *data)
3549 {
3550         struct nfs4_closedata *calldata = data;
3551         struct nfs4_state *state = calldata->state;
3552         struct nfs_server *server = NFS_SERVER(calldata->inode);
3553         nfs4_stateid *res_stateid = NULL;
3554         struct nfs4_exception exception = {
3555                 .state = state,
3556                 .inode = calldata->inode,
3557                 .stateid = &calldata->arg.stateid,
3558         };
3559
3560         dprintk("%s: begin!\n", __func__);
3561         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3562                 return;
3563         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3564
3565         /* Handle Layoutreturn errors */
3566         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3567                           &calldata->res.lr_ret) == -EAGAIN)
3568                 goto out_restart;
3569
3570         /* hmm. we are done with the inode, and in the process of freeing
3571          * the state_owner. we keep this around to process errors
3572          */
3573         switch (task->tk_status) {
3574                 case 0:
3575                         res_stateid = &calldata->res.stateid;
3576                         renew_lease(server, calldata->timestamp);
3577                         break;
3578                 case -NFS4ERR_ACCESS:
3579                         if (calldata->arg.bitmask != NULL) {
3580                                 calldata->arg.bitmask = NULL;
3581                                 calldata->res.fattr = NULL;
3582                                 goto out_restart;
3583
3584                         }
3585                         break;
3586                 case -NFS4ERR_OLD_STATEID:
3587                         /* Did we race with OPEN? */
3588                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3589                                                 state))
3590                                 goto out_restart;
3591                         goto out_release;
3592                 case -NFS4ERR_ADMIN_REVOKED:
3593                 case -NFS4ERR_STALE_STATEID:
3594                 case -NFS4ERR_EXPIRED:
3595                         nfs4_free_revoked_stateid(server,
3596                                         &calldata->arg.stateid,
3597                                         task->tk_msg.rpc_cred);
3598                         fallthrough;
3599                 case -NFS4ERR_BAD_STATEID:
3600                         if (calldata->arg.fmode == 0)
3601                                 break;
3602                         fallthrough;
3603                 default:
3604                         task->tk_status = nfs4_async_handle_exception(task,
3605                                         server, task->tk_status, &exception);
3606                         if (exception.retry)
3607                                 goto out_restart;
3608         }
3609         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3610                         res_stateid, calldata->arg.fmode);
3611 out_release:
3612         task->tk_status = 0;
3613         nfs_release_seqid(calldata->arg.seqid);
3614         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3615         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3616         return;
3617 out_restart:
3618         task->tk_status = 0;
3619         rpc_restart_call_prepare(task);
3620         goto out_release;
3621 }
3622
3623 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3624 {
3625         struct nfs4_closedata *calldata = data;
3626         struct nfs4_state *state = calldata->state;
3627         struct inode *inode = calldata->inode;
3628         struct nfs_server *server = NFS_SERVER(inode);
3629         struct pnfs_layout_hdr *lo;
3630         bool is_rdonly, is_wronly, is_rdwr;
3631         int call_close = 0;
3632
3633         dprintk("%s: begin!\n", __func__);
3634         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3635                 goto out_wait;
3636
3637         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3638         spin_lock(&state->owner->so_lock);
3639         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3640         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3641         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3642         /* Calculate the change in open mode */
3643         calldata->arg.fmode = 0;
3644         if (state->n_rdwr == 0) {
3645                 if (state->n_rdonly == 0)
3646                         call_close |= is_rdonly;
3647                 else if (is_rdonly)
3648                         calldata->arg.fmode |= FMODE_READ;
3649                 if (state->n_wronly == 0)
3650                         call_close |= is_wronly;
3651                 else if (is_wronly)
3652                         calldata->arg.fmode |= FMODE_WRITE;
3653                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3654                         call_close |= is_rdwr;
3655         } else if (is_rdwr)
3656                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3657
3658         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3659         if (!nfs4_valid_open_stateid(state))
3660                 call_close = 0;
3661         spin_unlock(&state->owner->so_lock);
3662
3663         if (!call_close) {
3664                 /* Note: exit _without_ calling nfs4_close_done */
3665                 goto out_no_action;
3666         }
3667
3668         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3669                 nfs_release_seqid(calldata->arg.seqid);
3670                 goto out_wait;
3671         }
3672
3673         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3674         if (lo && !pnfs_layout_is_valid(lo)) {
3675                 calldata->arg.lr_args = NULL;
3676                 calldata->res.lr_res = NULL;
3677         }
3678
3679         if (calldata->arg.fmode == 0)
3680                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3681
3682         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3683                 /* Close-to-open cache consistency revalidation */
3684                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3685                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3686                                          server->cache_consistency_bitmask,
3687                                          inode, server, NULL);
3688                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3689                 } else
3690                         calldata->arg.bitmask = NULL;
3691         }
3692
3693         calldata->arg.share_access =
3694                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3695                                 calldata->arg.fmode, 0);
3696
3697         if (calldata->res.fattr == NULL)
3698                 calldata->arg.bitmask = NULL;
3699         else if (calldata->arg.bitmask == NULL)
3700                 calldata->res.fattr = NULL;
3701         calldata->timestamp = jiffies;
3702         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3703                                 &calldata->arg.seq_args,
3704                                 &calldata->res.seq_res,
3705                                 task) != 0)
3706                 nfs_release_seqid(calldata->arg.seqid);
3707         dprintk("%s: done!\n", __func__);
3708         return;
3709 out_no_action:
3710         task->tk_action = NULL;
3711 out_wait:
3712         nfs4_sequence_done(task, &calldata->res.seq_res);
3713 }
3714
3715 static const struct rpc_call_ops nfs4_close_ops = {
3716         .rpc_call_prepare = nfs4_close_prepare,
3717         .rpc_call_done = nfs4_close_done,
3718         .rpc_release = nfs4_free_closedata,
3719 };
3720
3721 /*
3722  * It is possible for data to be read/written from a mem-mapped file
3723  * after the sys_close call (which hits the vfs layer as a flush).
3724  * This means that we can't safely call nfsv4 close on a file until
3725  * the inode is cleared. This in turn means that we are not good
3726  * NFSv4 citizens - we do not indicate to the server to update the file's
3727  * share state even when we are done with one of the three share
3728  * stateid's in the inode.
3729  *
3730  * NOTE: Caller must be holding the sp->so_owner semaphore!
3731  */
3732 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3733 {
3734         struct nfs_server *server = NFS_SERVER(state->inode);
3735         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3736         struct nfs4_closedata *calldata;
3737         struct nfs4_state_owner *sp = state->owner;
3738         struct rpc_task *task;
3739         struct rpc_message msg = {
3740                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3741                 .rpc_cred = state->owner->so_cred,
3742         };
3743         struct rpc_task_setup task_setup_data = {
3744                 .rpc_client = server->client,
3745                 .rpc_message = &msg,
3746                 .callback_ops = &nfs4_close_ops,
3747                 .workqueue = nfsiod_workqueue,
3748                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3749         };
3750         int status = -ENOMEM;
3751
3752         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3753                 &task_setup_data.rpc_client, &msg);
3754
3755         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3756         if (calldata == NULL)
3757                 goto out;
3758         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3759         calldata->inode = state->inode;
3760         calldata->state = state;
3761         calldata->arg.fh = NFS_FH(state->inode);
3762         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3763                 goto out_free_calldata;
3764         /* Serialization for the sequence id */
3765         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3766         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3767         if (IS_ERR(calldata->arg.seqid))
3768                 goto out_free_calldata;
3769         nfs_fattr_init(&calldata->fattr);
3770         calldata->arg.fmode = 0;
3771         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3772         calldata->res.fattr = &calldata->fattr;
3773         calldata->res.seqid = calldata->arg.seqid;
3774         calldata->res.server = server;
3775         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3776         calldata->lr.roc = pnfs_roc(state->inode,
3777                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3778         if (calldata->lr.roc) {
3779                 calldata->arg.lr_args = &calldata->lr.arg;
3780                 calldata->res.lr_res = &calldata->lr.res;
3781         }
3782         nfs_sb_active(calldata->inode->i_sb);
3783
3784         msg.rpc_argp = &calldata->arg;
3785         msg.rpc_resp = &calldata->res;
3786         task_setup_data.callback_data = calldata;
3787         task = rpc_run_task(&task_setup_data);
3788         if (IS_ERR(task))
3789                 return PTR_ERR(task);
3790         status = 0;
3791         if (wait)
3792                 status = rpc_wait_for_completion_task(task);
3793         rpc_put_task(task);
3794         return status;
3795 out_free_calldata:
3796         kfree(calldata);
3797 out:
3798         nfs4_put_open_state(state);
3799         nfs4_put_state_owner(sp);
3800         return status;
3801 }
3802
3803 static struct inode *
3804 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3805                 int open_flags, struct iattr *attr, int *opened)
3806 {
3807         struct nfs4_state *state;
3808         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3809
3810         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3811
3812         /* Protect against concurrent sillydeletes */
3813         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3814
3815         nfs4_label_release_security(label);
3816
3817         if (IS_ERR(state))
3818                 return ERR_CAST(state);
3819         return state->inode;
3820 }
3821
3822 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3823 {
3824         if (ctx->state == NULL)
3825                 return;
3826         if (is_sync)
3827                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3828         else
3829                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3830 }
3831
3832 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3833 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3834 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3835
3836 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3837 {
3838         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3839         struct nfs4_server_caps_arg args = {
3840                 .fhandle = fhandle,
3841                 .bitmask = bitmask,
3842         };
3843         struct nfs4_server_caps_res res = {};
3844         struct rpc_message msg = {
3845                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3846                 .rpc_argp = &args,
3847                 .rpc_resp = &res,
3848         };
3849         int status;
3850         int i;
3851
3852         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3853                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3854                      FATTR4_WORD0_LINK_SUPPORT |
3855                      FATTR4_WORD0_SYMLINK_SUPPORT |
3856                      FATTR4_WORD0_ACLSUPPORT;
3857         if (minorversion)
3858                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3859
3860         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3861         if (status == 0) {
3862                 /* Sanity check the server answers */
3863                 switch (minorversion) {
3864                 case 0:
3865                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3866                         res.attr_bitmask[2] = 0;
3867                         break;
3868                 case 1:
3869                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3870                         break;
3871                 case 2:
3872                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3873                 }
3874                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3875                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3876                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3877                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3878                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3879                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3880                         server->caps |= NFS_CAP_ACLS;
3881                 if (res.has_links != 0)
3882                         server->caps |= NFS_CAP_HARDLINKS;
3883                 if (res.has_symlinks != 0)
3884                         server->caps |= NFS_CAP_SYMLINKS;
3885 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3886                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3887                         server->caps |= NFS_CAP_SECURITY_LABEL;
3888 #endif
3889                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3890                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3891                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3892                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3893                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3894                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3895                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3896                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3897                                 NFS_ATTR_FATTR_OWNER_NAME);
3898                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3899                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3900                                 NFS_ATTR_FATTR_GROUP_NAME);
3901                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3902                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3903                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3904                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3905                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3906                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3907                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3908                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3909                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3910                                 sizeof(server->attr_bitmask));
3911                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3912
3913                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3914                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3915                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3916                 server->cache_consistency_bitmask[2] = 0;
3917
3918                 /* Avoid a regression due to buggy server */
3919                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3920                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3921                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3922                         sizeof(server->exclcreat_bitmask));
3923
3924                 server->acl_bitmask = res.acl_bitmask;
3925                 server->fh_expire_type = res.fh_expire_type;
3926         }
3927
3928         return status;
3929 }
3930
3931 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3932 {
3933         struct nfs4_exception exception = {
3934                 .interruptible = true,
3935         };
3936         int err;
3937         do {
3938                 err = nfs4_handle_exception(server,
3939                                 _nfs4_server_capabilities(server, fhandle),
3940                                 &exception);
3941         } while (exception.retry);
3942         return err;
3943 }
3944
3945 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3946                 struct nfs_fsinfo *info)
3947 {
3948         u32 bitmask[3];
3949         struct nfs4_lookup_root_arg args = {
3950                 .bitmask = bitmask,
3951         };
3952         struct nfs4_lookup_res res = {
3953                 .server = server,
3954                 .fattr = info->fattr,
3955                 .fh = fhandle,
3956         };
3957         struct rpc_message msg = {
3958                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3959                 .rpc_argp = &args,
3960                 .rpc_resp = &res,
3961         };
3962
3963         bitmask[0] = nfs4_fattr_bitmap[0];
3964         bitmask[1] = nfs4_fattr_bitmap[1];
3965         /*
3966          * Process the label in the upcoming getfattr
3967          */
3968         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3969
3970         nfs_fattr_init(info->fattr);
3971         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3972 }
3973
3974 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3975                 struct nfs_fsinfo *info)
3976 {
3977         struct nfs4_exception exception = {
3978                 .interruptible = true,
3979         };
3980         int err;
3981         do {
3982                 err = _nfs4_lookup_root(server, fhandle, info);
3983                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3984                 switch (err) {
3985                 case 0:
3986                 case -NFS4ERR_WRONGSEC:
3987                         goto out;
3988                 default:
3989                         err = nfs4_handle_exception(server, err, &exception);
3990                 }
3991         } while (exception.retry);
3992 out:
3993         return err;
3994 }
3995
3996 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3997                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3998 {
3999         struct rpc_auth_create_args auth_args = {
4000                 .pseudoflavor = flavor,
4001         };
4002         struct rpc_auth *auth;
4003
4004         auth = rpcauth_create(&auth_args, server->client);
4005         if (IS_ERR(auth))
4006                 return -EACCES;
4007         return nfs4_lookup_root(server, fhandle, info);
4008 }
4009
4010 /*
4011  * Retry pseudoroot lookup with various security flavors.  We do this when:
4012  *
4013  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4014  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4015  *
4016  * Returns zero on success, or a negative NFS4ERR value, or a
4017  * negative errno value.
4018  */
4019 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4020                               struct nfs_fsinfo *info)
4021 {
4022         /* Per 3530bis 15.33.5 */
4023         static const rpc_authflavor_t flav_array[] = {
4024                 RPC_AUTH_GSS_KRB5P,
4025                 RPC_AUTH_GSS_KRB5I,
4026                 RPC_AUTH_GSS_KRB5,
4027                 RPC_AUTH_UNIX,                  /* courtesy */
4028                 RPC_AUTH_NULL,
4029         };
4030         int status = -EPERM;
4031         size_t i;
4032
4033         if (server->auth_info.flavor_len > 0) {
4034                 /* try each flavor specified by user */
4035                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4036                         status = nfs4_lookup_root_sec(server, fhandle, info,
4037                                                 server->auth_info.flavors[i]);
4038                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4039                                 continue;
4040                         break;
4041                 }
4042         } else {
4043                 /* no flavors specified by user, try default list */
4044                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4045                         status = nfs4_lookup_root_sec(server, fhandle, info,
4046                                                       flav_array[i]);
4047                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4048                                 continue;
4049                         break;
4050                 }
4051         }
4052
4053         /*
4054          * -EACCES could mean that the user doesn't have correct permissions
4055          * to access the mount.  It could also mean that we tried to mount
4056          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4057          * existing mount programs don't handle -EACCES very well so it should
4058          * be mapped to -EPERM instead.
4059          */
4060         if (status == -EACCES)
4061                 status = -EPERM;
4062         return status;
4063 }
4064
4065 /**
4066  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4067  * @server: initialized nfs_server handle
4068  * @fhandle: we fill in the pseudo-fs root file handle
4069  * @info: we fill in an FSINFO struct
4070  * @auth_probe: probe the auth flavours
4071  *
4072  * Returns zero on success, or a negative errno.
4073  */
4074 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4075                          struct nfs_fsinfo *info,
4076                          bool auth_probe)
4077 {
4078         int status = 0;
4079
4080         if (!auth_probe)
4081                 status = nfs4_lookup_root(server, fhandle, info);
4082
4083         if (auth_probe || status == NFS4ERR_WRONGSEC)
4084                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4085                                 fhandle, info);
4086
4087         if (status == 0)
4088                 status = nfs4_server_capabilities(server, fhandle);
4089         if (status == 0)
4090                 status = nfs4_do_fsinfo(server, fhandle, info);
4091
4092         return nfs4_map_errors(status);
4093 }
4094
4095 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4096                               struct nfs_fsinfo *info)
4097 {
4098         int error;
4099         struct nfs_fattr *fattr = info->fattr;
4100         struct nfs4_label *label = fattr->label;
4101
4102         error = nfs4_server_capabilities(server, mntfh);
4103         if (error < 0) {
4104                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4105                 return error;
4106         }
4107
4108         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4109         if (error < 0) {
4110                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4111                 goto out;
4112         }
4113
4114         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4115             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4116                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4117
4118 out:
4119         return error;
4120 }
4121
4122 /*
4123  * Get locations and (maybe) other attributes of a referral.
4124  * Note that we'll actually follow the referral later when
4125  * we detect fsid mismatch in inode revalidation
4126  */
4127 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4128                              const struct qstr *name, struct nfs_fattr *fattr,
4129                              struct nfs_fh *fhandle)
4130 {
4131         int status = -ENOMEM;
4132         struct page *page = NULL;
4133         struct nfs4_fs_locations *locations = NULL;
4134
4135         page = alloc_page(GFP_KERNEL);
4136         if (page == NULL)
4137                 goto out;
4138         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4139         if (locations == NULL)
4140                 goto out;
4141
4142         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4143         if (status != 0)
4144                 goto out;
4145
4146         /*
4147          * If the fsid didn't change, this is a migration event, not a
4148          * referral.  Cause us to drop into the exception handler, which
4149          * will kick off migration recovery.
4150          */
4151         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4152                 dprintk("%s: server did not return a different fsid for"
4153                         " a referral at %s\n", __func__, name->name);
4154                 status = -NFS4ERR_MOVED;
4155                 goto out;
4156         }
4157         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4158         nfs_fixup_referral_attributes(&locations->fattr);
4159
4160         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4161         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4162         memset(fhandle, 0, sizeof(struct nfs_fh));
4163 out:
4164         if (page)
4165                 __free_page(page);
4166         kfree(locations);
4167         return status;
4168 }
4169
4170 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4171                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4172                                 struct inode *inode)
4173 {
4174         __u32 bitmask[NFS4_BITMASK_SZ];
4175         struct nfs4_getattr_arg args = {
4176                 .fh = fhandle,
4177                 .bitmask = bitmask,
4178         };
4179         struct nfs4_getattr_res res = {
4180                 .fattr = fattr,
4181                 .label = label,
4182                 .server = server,
4183         };
4184         struct rpc_message msg = {
4185                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4186                 .rpc_argp = &args,
4187                 .rpc_resp = &res,
4188         };
4189         unsigned short task_flags = 0;
4190
4191         /* Is this is an attribute revalidation, subject to softreval? */
4192         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4193                 task_flags |= RPC_TASK_TIMEOUT;
4194
4195         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode, 0);
4196         nfs_fattr_init(fattr);
4197         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4198         return nfs4_do_call_sync(server->client, server, &msg,
4199                         &args.seq_args, &res.seq_res, task_flags);
4200 }
4201
4202 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4203                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4204                                 struct inode *inode)
4205 {
4206         struct nfs4_exception exception = {
4207                 .interruptible = true,
4208         };
4209         int err;
4210         do {
4211                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4212                 trace_nfs4_getattr(server, fhandle, fattr, err);
4213                 err = nfs4_handle_exception(server, err,
4214                                 &exception);
4215         } while (exception.retry);
4216         return err;
4217 }
4218
4219 /*
4220  * The file is not closed if it is opened due to the a request to change
4221  * the size of the file. The open call will not be needed once the
4222  * VFS layer lookup-intents are implemented.
4223  *
4224  * Close is called when the inode is destroyed.
4225  * If we haven't opened the file for O_WRONLY, we
4226  * need to in the size_change case to obtain a stateid.
4227  *
4228  * Got race?
4229  * Because OPEN is always done by name in nfsv4, it is
4230  * possible that we opened a different file by the same
4231  * name.  We can recognize this race condition, but we
4232  * can't do anything about it besides returning an error.
4233  *
4234  * This will be fixed with VFS changes (lookup-intent).
4235  */
4236 static int
4237 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4238                   struct iattr *sattr)
4239 {
4240         struct inode *inode = d_inode(dentry);
4241         const struct cred *cred = NULL;
4242         struct nfs_open_context *ctx = NULL;
4243         struct nfs4_label *label = NULL;
4244         int status;
4245
4246         if (pnfs_ld_layoutret_on_setattr(inode) &&
4247             sattr->ia_valid & ATTR_SIZE &&
4248             sattr->ia_size < i_size_read(inode))
4249                 pnfs_commit_and_return_layout(inode);
4250
4251         nfs_fattr_init(fattr);
4252
4253         /* Deal with open(O_TRUNC) */
4254         if (sattr->ia_valid & ATTR_OPEN)
4255                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4256
4257         /* Optimization: if the end result is no change, don't RPC */
4258         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4259                 return 0;
4260
4261         /* Search for an existing open(O_WRITE) file */
4262         if (sattr->ia_valid & ATTR_FILE) {
4263
4264                 ctx = nfs_file_open_context(sattr->ia_file);
4265                 if (ctx)
4266                         cred = ctx->cred;
4267         }
4268
4269         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4270         if (IS_ERR(label))
4271                 return PTR_ERR(label);
4272
4273         /* Return any delegations if we're going to change ACLs */
4274         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4275                 nfs4_inode_make_writeable(inode);
4276
4277         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4278         if (status == 0) {
4279                 nfs_setattr_update_inode(inode, sattr, fattr);
4280                 nfs_setsecurity(inode, fattr, label);
4281         }
4282         nfs4_label_free(label);
4283         return status;
4284 }
4285
4286 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4287                 struct dentry *dentry, struct nfs_fh *fhandle,
4288                 struct nfs_fattr *fattr, struct nfs4_label *label)
4289 {
4290         struct nfs_server *server = NFS_SERVER(dir);
4291         int                    status;
4292         struct nfs4_lookup_arg args = {
4293                 .bitmask = server->attr_bitmask,
4294                 .dir_fh = NFS_FH(dir),
4295                 .name = &dentry->d_name,
4296         };
4297         struct nfs4_lookup_res res = {
4298                 .server = server,
4299                 .fattr = fattr,
4300                 .label = label,
4301                 .fh = fhandle,
4302         };
4303         struct rpc_message msg = {
4304                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4305                 .rpc_argp = &args,
4306                 .rpc_resp = &res,
4307         };
4308         unsigned short task_flags = 0;
4309
4310         /* Is this is an attribute revalidation, subject to softreval? */
4311         if (nfs_lookup_is_soft_revalidate(dentry))
4312                 task_flags |= RPC_TASK_TIMEOUT;
4313
4314         args.bitmask = nfs4_bitmask(server, label);
4315
4316         nfs_fattr_init(fattr);
4317
4318         dprintk("NFS call  lookup %pd2\n", dentry);
4319         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4320         status = nfs4_do_call_sync(clnt, server, &msg,
4321                         &args.seq_args, &res.seq_res, task_flags);
4322         dprintk("NFS reply lookup: %d\n", status);
4323         return status;
4324 }
4325
4326 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4327 {
4328         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4329                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4330         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4331         fattr->nlink = 2;
4332 }
4333
4334 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4335                                    struct dentry *dentry, struct nfs_fh *fhandle,
4336                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4337 {
4338         struct nfs4_exception exception = {
4339                 .interruptible = true,
4340         };
4341         struct rpc_clnt *client = *clnt;
4342         const struct qstr *name = &dentry->d_name;
4343         int err;
4344         do {
4345                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4346                 trace_nfs4_lookup(dir, name, err);
4347                 switch (err) {
4348                 case -NFS4ERR_BADNAME:
4349                         err = -ENOENT;
4350                         goto out;
4351                 case -NFS4ERR_MOVED:
4352                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4353                         if (err == -NFS4ERR_MOVED)
4354                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4355                         goto out;
4356                 case -NFS4ERR_WRONGSEC:
4357                         err = -EPERM;
4358                         if (client != *clnt)
4359                                 goto out;
4360                         client = nfs4_negotiate_security(client, dir, name);
4361                         if (IS_ERR(client))
4362                                 return PTR_ERR(client);
4363
4364                         exception.retry = 1;
4365                         break;
4366                 default:
4367                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4368                 }
4369         } while (exception.retry);
4370
4371 out:
4372         if (err == 0)
4373                 *clnt = client;
4374         else if (client != *clnt)
4375                 rpc_shutdown_client(client);
4376
4377         return err;
4378 }
4379
4380 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4381                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4382                             struct nfs4_label *label)
4383 {
4384         int status;
4385         struct rpc_clnt *client = NFS_CLIENT(dir);
4386
4387         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4388         if (client != NFS_CLIENT(dir)) {
4389                 rpc_shutdown_client(client);
4390                 nfs_fixup_secinfo_attributes(fattr);
4391         }
4392         return status;
4393 }
4394
4395 struct rpc_clnt *
4396 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4397                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4398 {
4399         struct rpc_clnt *client = NFS_CLIENT(dir);
4400         int status;
4401
4402         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4403         if (status < 0)
4404                 return ERR_PTR(status);
4405         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4406 }
4407
4408 static int _nfs4_proc_lookupp(struct inode *inode,
4409                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4410                 struct nfs4_label *label)
4411 {
4412         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4413         struct nfs_server *server = NFS_SERVER(inode);
4414         int                    status;
4415         struct nfs4_lookupp_arg args = {
4416                 .bitmask = server->attr_bitmask,
4417                 .fh = NFS_FH(inode),
4418         };
4419         struct nfs4_lookupp_res res = {
4420                 .server = server,
4421                 .fattr = fattr,
4422                 .label = label,
4423                 .fh = fhandle,
4424         };
4425         struct rpc_message msg = {
4426                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4427                 .rpc_argp = &args,
4428                 .rpc_resp = &res,
4429         };
4430         unsigned short task_flags = 0;
4431
4432         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4433                 task_flags |= RPC_TASK_TIMEOUT;
4434
4435         args.bitmask = nfs4_bitmask(server, label);
4436
4437         nfs_fattr_init(fattr);
4438
4439         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4440         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4441                                 &res.seq_res, task_flags);
4442         dprintk("NFS reply lookupp: %d\n", status);
4443         return status;
4444 }
4445
4446 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4447                              struct nfs_fattr *fattr, struct nfs4_label *label)
4448 {
4449         struct nfs4_exception exception = {
4450                 .interruptible = true,
4451         };
4452         int err;
4453         do {
4454                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4455                 trace_nfs4_lookupp(inode, err);
4456                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4457                                 &exception);
4458         } while (exception.retry);
4459         return err;
4460 }
4461
4462 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4463 {
4464         struct nfs_server *server = NFS_SERVER(inode);
4465         struct nfs4_accessargs args = {
4466                 .fh = NFS_FH(inode),
4467                 .access = entry->mask,
4468         };
4469         struct nfs4_accessres res = {
4470                 .server = server,
4471         };
4472         struct rpc_message msg = {
4473                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4474                 .rpc_argp = &args,
4475                 .rpc_resp = &res,
4476                 .rpc_cred = entry->cred,
4477         };
4478         int status = 0;
4479
4480         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4481                 res.fattr = nfs_alloc_fattr();
4482                 if (res.fattr == NULL)
4483                         return -ENOMEM;
4484                 args.bitmask = server->cache_consistency_bitmask;
4485         }
4486         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4487         if (!status) {
4488                 nfs_access_set_mask(entry, res.access);
4489                 if (res.fattr)
4490                         nfs_refresh_inode(inode, res.fattr);
4491         }
4492         nfs_free_fattr(res.fattr);
4493         return status;
4494 }
4495
4496 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4497 {
4498         struct nfs4_exception exception = {
4499                 .interruptible = true,
4500         };
4501         int err;
4502         do {
4503                 err = _nfs4_proc_access(inode, entry);
4504                 trace_nfs4_access(inode, err);
4505                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4506                                 &exception);
4507         } while (exception.retry);
4508         return err;
4509 }
4510
4511 /*
4512  * TODO: For the time being, we don't try to get any attributes
4513  * along with any of the zero-copy operations READ, READDIR,
4514  * READLINK, WRITE.
4515  *
4516  * In the case of the first three, we want to put the GETATTR
4517  * after the read-type operation -- this is because it is hard
4518  * to predict the length of a GETATTR response in v4, and thus
4519  * align the READ data correctly.  This means that the GETATTR
4520  * may end up partially falling into the page cache, and we should
4521  * shift it into the 'tail' of the xdr_buf before processing.
4522  * To do this efficiently, we need to know the total length
4523  * of data received, which doesn't seem to be available outside
4524  * of the RPC layer.
4525  *
4526  * In the case of WRITE, we also want to put the GETATTR after
4527  * the operation -- in this case because we want to make sure
4528  * we get the post-operation mtime and size.
4529  *
4530  * Both of these changes to the XDR layer would in fact be quite
4531  * minor, but I decided to leave them for a subsequent patch.
4532  */
4533 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4534                 unsigned int pgbase, unsigned int pglen)
4535 {
4536         struct nfs4_readlink args = {
4537                 .fh       = NFS_FH(inode),
4538                 .pgbase   = pgbase,
4539                 .pglen    = pglen,
4540                 .pages    = &page,
4541         };
4542         struct nfs4_readlink_res res;
4543         struct rpc_message msg = {
4544                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4545                 .rpc_argp = &args,
4546                 .rpc_resp = &res,
4547         };
4548
4549         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4550 }
4551
4552 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4553                 unsigned int pgbase, unsigned int pglen)
4554 {
4555         struct nfs4_exception exception = {
4556                 .interruptible = true,
4557         };
4558         int err;
4559         do {
4560                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4561                 trace_nfs4_readlink(inode, err);
4562                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4563                                 &exception);
4564         } while (exception.retry);
4565         return err;
4566 }
4567
4568 /*
4569  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4570  */
4571 static int
4572 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4573                  int flags)
4574 {
4575         struct nfs_server *server = NFS_SERVER(dir);
4576         struct nfs4_label l, *ilabel = NULL;
4577         struct nfs_open_context *ctx;
4578         struct nfs4_state *state;
4579         int status = 0;
4580
4581         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4582         if (IS_ERR(ctx))
4583                 return PTR_ERR(ctx);
4584
4585         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4586
4587         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4588                 sattr->ia_mode &= ~current_umask();
4589         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4590         if (IS_ERR(state)) {
4591                 status = PTR_ERR(state);
4592                 goto out;
4593         }
4594 out:
4595         nfs4_label_release_security(ilabel);
4596         put_nfs_open_context(ctx);
4597         return status;
4598 }
4599
4600 static int
4601 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4602 {
4603         struct nfs_server *server = NFS_SERVER(dir);
4604         struct nfs_removeargs args = {
4605                 .fh = NFS_FH(dir),
4606                 .name = *name,
4607         };
4608         struct nfs_removeres res = {
4609                 .server = server,
4610         };
4611         struct rpc_message msg = {
4612                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4613                 .rpc_argp = &args,
4614                 .rpc_resp = &res,
4615         };
4616         unsigned long timestamp = jiffies;
4617         int status;
4618
4619         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4620         if (status == 0) {
4621                 spin_lock(&dir->i_lock);
4622                 /* Removing a directory decrements nlink in the parent */
4623                 if (ftype == NF4DIR && dir->i_nlink > 2)
4624                         nfs4_dec_nlink_locked(dir);
4625                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4626                                               NFS_INO_INVALID_DATA);
4627                 spin_unlock(&dir->i_lock);
4628         }
4629         return status;
4630 }
4631
4632 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4633 {
4634         struct nfs4_exception exception = {
4635                 .interruptible = true,
4636         };
4637         struct inode *inode = d_inode(dentry);
4638         int err;
4639
4640         if (inode) {
4641                 if (inode->i_nlink == 1)
4642                         nfs4_inode_return_delegation(inode);
4643                 else
4644                         nfs4_inode_make_writeable(inode);
4645         }
4646         do {
4647                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4648                 trace_nfs4_remove(dir, &dentry->d_name, err);
4649                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4650                                 &exception);
4651         } while (exception.retry);
4652         return err;
4653 }
4654
4655 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4656 {
4657         struct nfs4_exception exception = {
4658                 .interruptible = true,
4659         };
4660         int err;
4661
4662         do {
4663                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4664                 trace_nfs4_remove(dir, name, err);
4665                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4666                                 &exception);
4667         } while (exception.retry);
4668         return err;
4669 }
4670
4671 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4672                 struct dentry *dentry,
4673                 struct inode *inode)
4674 {
4675         struct nfs_removeargs *args = msg->rpc_argp;
4676         struct nfs_removeres *res = msg->rpc_resp;
4677
4678         res->server = NFS_SB(dentry->d_sb);
4679         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4680         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4681
4682         nfs_fattr_init(res->dir_attr);
4683
4684         if (inode)
4685                 nfs4_inode_return_delegation(inode);
4686 }
4687
4688 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4689 {
4690         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4691                         &data->args.seq_args,
4692                         &data->res.seq_res,
4693                         task);
4694 }
4695
4696 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4697 {
4698         struct nfs_unlinkdata *data = task->tk_calldata;
4699         struct nfs_removeres *res = &data->res;
4700
4701         if (!nfs4_sequence_done(task, &res->seq_res))
4702                 return 0;
4703         if (nfs4_async_handle_error(task, res->server, NULL,
4704                                     &data->timeout) == -EAGAIN)
4705                 return 0;
4706         if (task->tk_status == 0)
4707                 nfs4_update_changeattr(dir, &res->cinfo,
4708                                 res->dir_attr->time_start,
4709                                 NFS_INO_INVALID_DATA);
4710         return 1;
4711 }
4712
4713 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4714                 struct dentry *old_dentry,
4715                 struct dentry *new_dentry)
4716 {
4717         struct nfs_renameargs *arg = msg->rpc_argp;
4718         struct nfs_renameres *res = msg->rpc_resp;
4719         struct inode *old_inode = d_inode(old_dentry);
4720         struct inode *new_inode = d_inode(new_dentry);
4721
4722         if (old_inode)
4723                 nfs4_inode_make_writeable(old_inode);
4724         if (new_inode)
4725                 nfs4_inode_return_delegation(new_inode);
4726         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4727         res->server = NFS_SB(old_dentry->d_sb);
4728         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4729 }
4730
4731 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4732 {
4733         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4734                         &data->args.seq_args,
4735                         &data->res.seq_res,
4736                         task);
4737 }
4738
4739 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4740                                  struct inode *new_dir)
4741 {
4742         struct nfs_renamedata *data = task->tk_calldata;
4743         struct nfs_renameres *res = &data->res;
4744
4745         if (!nfs4_sequence_done(task, &res->seq_res))
4746                 return 0;
4747         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4748                 return 0;
4749
4750         if (task->tk_status == 0) {
4751                 if (new_dir != old_dir) {
4752                         /* Note: If we moved a directory, nlink will change */
4753                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4754                                         res->old_fattr->time_start,
4755                                         NFS_INO_INVALID_NLINK |
4756                                             NFS_INO_INVALID_DATA);
4757                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4758                                         res->new_fattr->time_start,
4759                                         NFS_INO_INVALID_NLINK |
4760                                             NFS_INO_INVALID_DATA);
4761                 } else
4762                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4763                                         res->old_fattr->time_start,
4764                                         NFS_INO_INVALID_DATA);
4765         }
4766         return 1;
4767 }
4768
4769 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4770 {
4771         struct nfs_server *server = NFS_SERVER(inode);
4772         __u32 bitmask[NFS4_BITMASK_SZ];
4773         struct nfs4_link_arg arg = {
4774                 .fh     = NFS_FH(inode),
4775                 .dir_fh = NFS_FH(dir),
4776                 .name   = name,
4777                 .bitmask = bitmask,
4778         };
4779         struct nfs4_link_res res = {
4780                 .server = server,
4781                 .label = NULL,
4782         };
4783         struct rpc_message msg = {
4784                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4785                 .rpc_argp = &arg,
4786                 .rpc_resp = &res,
4787         };
4788         int status = -ENOMEM;
4789
4790         res.fattr = nfs_alloc_fattr();
4791         if (res.fattr == NULL)
4792                 goto out;
4793
4794         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4795         if (IS_ERR(res.label)) {
4796                 status = PTR_ERR(res.label);
4797                 goto out;
4798         }
4799
4800         nfs4_inode_make_writeable(inode);
4801         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.label), inode,
4802                                 NFS_INO_INVALID_CHANGE);
4803         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4804         if (!status) {
4805                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4806                                        NFS_INO_INVALID_DATA);
4807                 nfs4_inc_nlink(inode);
4808                 status = nfs_post_op_update_inode(inode, res.fattr);
4809                 if (!status)
4810                         nfs_setsecurity(inode, res.fattr, res.label);
4811         }
4812
4813
4814         nfs4_label_free(res.label);
4815
4816 out:
4817         nfs_free_fattr(res.fattr);
4818         return status;
4819 }
4820
4821 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4822 {
4823         struct nfs4_exception exception = {
4824                 .interruptible = true,
4825         };
4826         int err;
4827         do {
4828                 err = nfs4_handle_exception(NFS_SERVER(inode),
4829                                 _nfs4_proc_link(inode, dir, name),
4830                                 &exception);
4831         } while (exception.retry);
4832         return err;
4833 }
4834
4835 struct nfs4_createdata {
4836         struct rpc_message msg;
4837         struct nfs4_create_arg arg;
4838         struct nfs4_create_res res;
4839         struct nfs_fh fh;
4840         struct nfs_fattr fattr;
4841         struct nfs4_label *label;
4842 };
4843
4844 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4845                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4846 {
4847         struct nfs4_createdata *data;
4848
4849         data = kzalloc(sizeof(*data), GFP_KERNEL);
4850         if (data != NULL) {
4851                 struct nfs_server *server = NFS_SERVER(dir);
4852
4853                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4854                 if (IS_ERR(data->label))
4855                         goto out_free;
4856
4857                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4858                 data->msg.rpc_argp = &data->arg;
4859                 data->msg.rpc_resp = &data->res;
4860                 data->arg.dir_fh = NFS_FH(dir);
4861                 data->arg.server = server;
4862                 data->arg.name = name;
4863                 data->arg.attrs = sattr;
4864                 data->arg.ftype = ftype;
4865                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4866                 data->arg.umask = current_umask();
4867                 data->res.server = server;
4868                 data->res.fh = &data->fh;
4869                 data->res.fattr = &data->fattr;
4870                 data->res.label = data->label;
4871                 nfs_fattr_init(data->res.fattr);
4872         }
4873         return data;
4874 out_free:
4875         kfree(data);
4876         return NULL;
4877 }
4878
4879 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4880 {
4881         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4882                                     &data->arg.seq_args, &data->res.seq_res, 1);
4883         if (status == 0) {
4884                 spin_lock(&dir->i_lock);
4885                 /* Creating a directory bumps nlink in the parent */
4886                 if (data->arg.ftype == NF4DIR)
4887                         nfs4_inc_nlink_locked(dir);
4888                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4889                                               data->res.fattr->time_start,
4890                                               NFS_INO_INVALID_DATA);
4891                 spin_unlock(&dir->i_lock);
4892                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4893         }
4894         return status;
4895 }
4896
4897 static void nfs4_free_createdata(struct nfs4_createdata *data)
4898 {
4899         nfs4_label_free(data->label);
4900         kfree(data);
4901 }
4902
4903 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4904                 struct page *page, unsigned int len, struct iattr *sattr,
4905                 struct nfs4_label *label)
4906 {
4907         struct nfs4_createdata *data;
4908         int status = -ENAMETOOLONG;
4909
4910         if (len > NFS4_MAXPATHLEN)
4911                 goto out;
4912
4913         status = -ENOMEM;
4914         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4915         if (data == NULL)
4916                 goto out;
4917
4918         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4919         data->arg.u.symlink.pages = &page;
4920         data->arg.u.symlink.len = len;
4921         data->arg.label = label;
4922
4923         status = nfs4_do_create(dir, dentry, data);
4924
4925         nfs4_free_createdata(data);
4926 out:
4927         return status;
4928 }
4929
4930 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4931                 struct page *page, unsigned int len, struct iattr *sattr)
4932 {
4933         struct nfs4_exception exception = {
4934                 .interruptible = true,
4935         };
4936         struct nfs4_label l, *label = NULL;
4937         int err;
4938
4939         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4940
4941         do {
4942                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4943                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4944                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4945                                 &exception);
4946         } while (exception.retry);
4947
4948         nfs4_label_release_security(label);
4949         return err;
4950 }
4951
4952 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4953                 struct iattr *sattr, struct nfs4_label *label)
4954 {
4955         struct nfs4_createdata *data;
4956         int status = -ENOMEM;
4957
4958         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4959         if (data == NULL)
4960                 goto out;
4961
4962         data->arg.label = label;
4963         status = nfs4_do_create(dir, dentry, data);
4964
4965         nfs4_free_createdata(data);
4966 out:
4967         return status;
4968 }
4969
4970 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4971                 struct iattr *sattr)
4972 {
4973         struct nfs_server *server = NFS_SERVER(dir);
4974         struct nfs4_exception exception = {
4975                 .interruptible = true,
4976         };
4977         struct nfs4_label l, *label = NULL;
4978         int err;
4979
4980         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4981
4982         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4983                 sattr->ia_mode &= ~current_umask();
4984         do {
4985                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4986                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4987                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4988                                 &exception);
4989         } while (exception.retry);
4990         nfs4_label_release_security(label);
4991
4992         return err;
4993 }
4994
4995 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
4996                               struct nfs_readdir_res *nr_res)
4997 {
4998         struct inode            *dir = d_inode(nr_arg->dentry);
4999         struct nfs_server       *server = NFS_SERVER(dir);
5000         struct nfs4_readdir_arg args = {
5001                 .fh = NFS_FH(dir),
5002                 .pages = nr_arg->pages,
5003                 .pgbase = 0,
5004                 .count = nr_arg->page_len,
5005                 .plus = nr_arg->plus,
5006         };
5007         struct nfs4_readdir_res res;
5008         struct rpc_message msg = {
5009                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5010                 .rpc_argp = &args,
5011                 .rpc_resp = &res,
5012                 .rpc_cred = nr_arg->cred,
5013         };
5014         int                     status;
5015
5016         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5017                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5018         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5019                 args.bitmask = server->attr_bitmask_nl;
5020         else
5021                 args.bitmask = server->attr_bitmask;
5022
5023         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5024         res.pgbase = args.pgbase;
5025         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5026                         &res.seq_res, 0);
5027         if (status >= 0) {
5028                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5029                 status += args.pgbase;
5030         }
5031
5032         nfs_invalidate_atime(dir);
5033
5034         dprintk("%s: returns %d\n", __func__, status);
5035         return status;
5036 }
5037
5038 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5039                              struct nfs_readdir_res *res)
5040 {
5041         struct nfs4_exception exception = {
5042                 .interruptible = true,
5043         };
5044         int err;
5045         do {
5046                 err = _nfs4_proc_readdir(arg, res);
5047                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5048                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5049                                             err, &exception);
5050         } while (exception.retry);
5051         return err;
5052 }
5053
5054 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5055                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5056 {
5057         struct nfs4_createdata *data;
5058         int mode = sattr->ia_mode;
5059         int status = -ENOMEM;
5060
5061         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5062         if (data == NULL)
5063                 goto out;
5064
5065         if (S_ISFIFO(mode))
5066                 data->arg.ftype = NF4FIFO;
5067         else if (S_ISBLK(mode)) {
5068                 data->arg.ftype = NF4BLK;
5069                 data->arg.u.device.specdata1 = MAJOR(rdev);
5070                 data->arg.u.device.specdata2 = MINOR(rdev);
5071         }
5072         else if (S_ISCHR(mode)) {
5073                 data->arg.ftype = NF4CHR;
5074                 data->arg.u.device.specdata1 = MAJOR(rdev);
5075                 data->arg.u.device.specdata2 = MINOR(rdev);
5076         } else if (!S_ISSOCK(mode)) {
5077                 status = -EINVAL;
5078                 goto out_free;
5079         }
5080
5081         data->arg.label = label;
5082         status = nfs4_do_create(dir, dentry, data);
5083 out_free:
5084         nfs4_free_createdata(data);
5085 out:
5086         return status;
5087 }
5088
5089 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5090                 struct iattr *sattr, dev_t rdev)
5091 {
5092         struct nfs_server *server = NFS_SERVER(dir);
5093         struct nfs4_exception exception = {
5094                 .interruptible = true,
5095         };
5096         struct nfs4_label l, *label = NULL;
5097         int err;
5098
5099         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5100
5101         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5102                 sattr->ia_mode &= ~current_umask();
5103         do {
5104                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5105                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5106                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5107                                 &exception);
5108         } while (exception.retry);
5109
5110         nfs4_label_release_security(label);
5111
5112         return err;
5113 }
5114
5115 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5116                  struct nfs_fsstat *fsstat)
5117 {
5118         struct nfs4_statfs_arg args = {
5119                 .fh = fhandle,
5120                 .bitmask = server->attr_bitmask,
5121         };
5122         struct nfs4_statfs_res res = {
5123                 .fsstat = fsstat,
5124         };
5125         struct rpc_message msg = {
5126                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5127                 .rpc_argp = &args,
5128                 .rpc_resp = &res,
5129         };
5130
5131         nfs_fattr_init(fsstat->fattr);
5132         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5133 }
5134
5135 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5136 {
5137         struct nfs4_exception exception = {
5138                 .interruptible = true,
5139         };
5140         int err;
5141         do {
5142                 err = nfs4_handle_exception(server,
5143                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5144                                 &exception);
5145         } while (exception.retry);
5146         return err;
5147 }
5148
5149 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5150                 struct nfs_fsinfo *fsinfo)
5151 {
5152         struct nfs4_fsinfo_arg args = {
5153                 .fh = fhandle,
5154                 .bitmask = server->attr_bitmask,
5155         };
5156         struct nfs4_fsinfo_res res = {
5157                 .fsinfo = fsinfo,
5158         };
5159         struct rpc_message msg = {
5160                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5161                 .rpc_argp = &args,
5162                 .rpc_resp = &res,
5163         };
5164
5165         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5166 }
5167
5168 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5169 {
5170         struct nfs4_exception exception = {
5171                 .interruptible = true,
5172         };
5173         int err;
5174
5175         do {
5176                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5177                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5178                 if (err == 0) {
5179                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5180                         break;
5181                 }
5182                 err = nfs4_handle_exception(server, err, &exception);
5183         } while (exception.retry);
5184         return err;
5185 }
5186
5187 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5188 {
5189         int error;
5190
5191         nfs_fattr_init(fsinfo->fattr);
5192         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5193         if (error == 0) {
5194                 /* block layout checks this! */
5195                 server->pnfs_blksize = fsinfo->blksize;
5196                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5197         }
5198
5199         return error;
5200 }
5201
5202 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5203                 struct nfs_pathconf *pathconf)
5204 {
5205         struct nfs4_pathconf_arg args = {
5206                 .fh = fhandle,
5207                 .bitmask = server->attr_bitmask,
5208         };
5209         struct nfs4_pathconf_res res = {
5210                 .pathconf = pathconf,
5211         };
5212         struct rpc_message msg = {
5213                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5214                 .rpc_argp = &args,
5215                 .rpc_resp = &res,
5216         };
5217
5218         /* None of the pathconf attributes are mandatory to implement */
5219         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5220                 memset(pathconf, 0, sizeof(*pathconf));
5221                 return 0;
5222         }
5223
5224         nfs_fattr_init(pathconf->fattr);
5225         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5226 }
5227
5228 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5229                 struct nfs_pathconf *pathconf)
5230 {
5231         struct nfs4_exception exception = {
5232                 .interruptible = true,
5233         };
5234         int err;
5235
5236         do {
5237                 err = nfs4_handle_exception(server,
5238                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5239                                 &exception);
5240         } while (exception.retry);
5241         return err;
5242 }
5243
5244 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5245                 const struct nfs_open_context *ctx,
5246                 const struct nfs_lock_context *l_ctx,
5247                 fmode_t fmode)
5248 {
5249         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5250 }
5251 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5252
5253 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5254                 const struct nfs_open_context *ctx,
5255                 const struct nfs_lock_context *l_ctx,
5256                 fmode_t fmode)
5257 {
5258         nfs4_stateid _current_stateid;
5259
5260         /* If the current stateid represents a lost lock, then exit */
5261         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5262                 return true;
5263         return nfs4_stateid_match(stateid, &_current_stateid);
5264 }
5265
5266 static bool nfs4_error_stateid_expired(int err)
5267 {
5268         switch (err) {
5269         case -NFS4ERR_DELEG_REVOKED:
5270         case -NFS4ERR_ADMIN_REVOKED:
5271         case -NFS4ERR_BAD_STATEID:
5272         case -NFS4ERR_STALE_STATEID:
5273         case -NFS4ERR_OLD_STATEID:
5274         case -NFS4ERR_OPENMODE:
5275         case -NFS4ERR_EXPIRED:
5276                 return true;
5277         }
5278         return false;
5279 }
5280
5281 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5282 {
5283         struct nfs_server *server = NFS_SERVER(hdr->inode);
5284
5285         trace_nfs4_read(hdr, task->tk_status);
5286         if (task->tk_status < 0) {
5287                 struct nfs4_exception exception = {
5288                         .inode = hdr->inode,
5289                         .state = hdr->args.context->state,
5290                         .stateid = &hdr->args.stateid,
5291                 };
5292                 task->tk_status = nfs4_async_handle_exception(task,
5293                                 server, task->tk_status, &exception);
5294                 if (exception.retry) {
5295                         rpc_restart_call_prepare(task);
5296                         return -EAGAIN;
5297                 }
5298         }
5299
5300         if (task->tk_status > 0)
5301                 renew_lease(server, hdr->timestamp);
5302         return 0;
5303 }
5304
5305 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5306                 struct nfs_pgio_args *args)
5307 {
5308
5309         if (!nfs4_error_stateid_expired(task->tk_status) ||
5310                 nfs4_stateid_is_current(&args->stateid,
5311                                 args->context,
5312                                 args->lock_context,
5313                                 FMODE_READ))
5314                 return false;
5315         rpc_restart_call_prepare(task);
5316         return true;
5317 }
5318
5319 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5320                                          struct nfs_pgio_header *hdr)
5321 {
5322         struct nfs_server *server = NFS_SERVER(hdr->inode);
5323         struct rpc_message *msg = &task->tk_msg;
5324
5325         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5326             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5327                 server->caps &= ~NFS_CAP_READ_PLUS;
5328                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5329                 rpc_restart_call_prepare(task);
5330                 return true;
5331         }
5332         return false;
5333 }
5334
5335 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5336 {
5337         dprintk("--> %s\n", __func__);
5338
5339         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5340                 return -EAGAIN;
5341         if (nfs4_read_stateid_changed(task, &hdr->args))
5342                 return -EAGAIN;
5343         if (nfs4_read_plus_not_supported(task, hdr))
5344                 return -EAGAIN;
5345         if (task->tk_status > 0)
5346                 nfs_invalidate_atime(hdr->inode);
5347         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5348                                     nfs4_read_done_cb(task, hdr);
5349 }
5350
5351 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5352 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5353                                     struct rpc_message *msg)
5354 {
5355         /* Note: We don't use READ_PLUS with pNFS yet */
5356         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5357                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5358 }
5359 #else
5360 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5361                                     struct rpc_message *msg)
5362 {
5363 }
5364 #endif /* CONFIG_NFS_V4_2 */
5365
5366 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5367                                  struct rpc_message *msg)
5368 {
5369         hdr->timestamp   = jiffies;
5370         if (!hdr->pgio_done_cb)
5371                 hdr->pgio_done_cb = nfs4_read_done_cb;
5372         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5373         nfs42_read_plus_support(hdr, msg);
5374         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5375 }
5376
5377 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5378                                       struct nfs_pgio_header *hdr)
5379 {
5380         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5381                         &hdr->args.seq_args,
5382                         &hdr->res.seq_res,
5383                         task))
5384                 return 0;
5385         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5386                                 hdr->args.lock_context,
5387                                 hdr->rw_mode) == -EIO)
5388                 return -EIO;
5389         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5390                 return -EIO;
5391         return 0;
5392 }
5393
5394 static int nfs4_write_done_cb(struct rpc_task *task,
5395                               struct nfs_pgio_header *hdr)
5396 {
5397         struct inode *inode = hdr->inode;
5398
5399         trace_nfs4_write(hdr, task->tk_status);
5400         if (task->tk_status < 0) {
5401                 struct nfs4_exception exception = {
5402                         .inode = hdr->inode,
5403                         .state = hdr->args.context->state,
5404                         .stateid = &hdr->args.stateid,
5405                 };
5406                 task->tk_status = nfs4_async_handle_exception(task,
5407                                 NFS_SERVER(inode), task->tk_status,
5408                                 &exception);
5409                 if (exception.retry) {
5410                         rpc_restart_call_prepare(task);
5411                         return -EAGAIN;
5412                 }
5413         }
5414         if (task->tk_status >= 0) {
5415                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5416                 nfs_writeback_update_inode(hdr);
5417         }
5418         return 0;
5419 }
5420
5421 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5422                 struct nfs_pgio_args *args)
5423 {
5424
5425         if (!nfs4_error_stateid_expired(task->tk_status) ||
5426                 nfs4_stateid_is_current(&args->stateid,
5427                                 args->context,
5428                                 args->lock_context,
5429                                 FMODE_WRITE))
5430                 return false;
5431         rpc_restart_call_prepare(task);
5432         return true;
5433 }
5434
5435 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5436 {
5437         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5438                 return -EAGAIN;
5439         if (nfs4_write_stateid_changed(task, &hdr->args))
5440                 return -EAGAIN;
5441         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5442                 nfs4_write_done_cb(task, hdr);
5443 }
5444
5445 static
5446 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5447 {
5448         /* Don't request attributes for pNFS or O_DIRECT writes */
5449         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5450                 return false;
5451         /* Otherwise, request attributes if and only if we don't hold
5452          * a delegation
5453          */
5454         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5455 }
5456
5457 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src,
5458                              struct inode *inode, struct nfs_server *server,
5459                              struct nfs4_label *label)
5460 {
5461         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5462         unsigned int i;
5463
5464         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5465
5466         if (cache_validity & NFS_INO_INVALID_CHANGE)
5467                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5468         if (cache_validity & NFS_INO_INVALID_ATIME)
5469                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5470         if (cache_validity & NFS_INO_INVALID_MODE)
5471                 bitmask[1] |= FATTR4_WORD1_MODE;
5472         if (cache_validity & NFS_INO_INVALID_OTHER)
5473                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5474         if (cache_validity & NFS_INO_INVALID_NLINK)
5475                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5476         if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL)
5477                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5478         if (cache_validity & NFS_INO_INVALID_CTIME)
5479                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5480         if (cache_validity & NFS_INO_INVALID_MTIME)
5481                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5482         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5483                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5484
5485         if (cache_validity & NFS_INO_INVALID_SIZE)
5486                 bitmask[0] |= FATTR4_WORD0_SIZE;
5487
5488         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5489                 bitmask[i] &= server->attr_bitmask[i];
5490 }
5491
5492 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5493                                   struct rpc_message *msg,
5494                                   struct rpc_clnt **clnt)
5495 {
5496         struct nfs_server *server = NFS_SERVER(hdr->inode);
5497
5498         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5499                 hdr->args.bitmask = NULL;
5500                 hdr->res.fattr = NULL;
5501         } else {
5502                 nfs4_bitmask_set(hdr->args.bitmask_store,
5503                                  server->cache_consistency_bitmask,
5504                                  hdr->inode, server, NULL);
5505                 hdr->args.bitmask = hdr->args.bitmask_store;
5506         }
5507
5508         if (!hdr->pgio_done_cb)
5509                 hdr->pgio_done_cb = nfs4_write_done_cb;
5510         hdr->res.server = server;
5511         hdr->timestamp   = jiffies;
5512
5513         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5514         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5515         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5516 }
5517
5518 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5519 {
5520         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5521                         &data->args.seq_args,
5522                         &data->res.seq_res,
5523                         task);
5524 }
5525
5526 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5527 {
5528         struct inode *inode = data->inode;
5529
5530         trace_nfs4_commit(data, task->tk_status);
5531         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5532                                     NULL, NULL) == -EAGAIN) {
5533                 rpc_restart_call_prepare(task);
5534                 return -EAGAIN;
5535         }
5536         return 0;
5537 }
5538
5539 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5540 {
5541         if (!nfs4_sequence_done(task, &data->res.seq_res))
5542                 return -EAGAIN;
5543         return data->commit_done_cb(task, data);
5544 }
5545
5546 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5547                                    struct rpc_clnt **clnt)
5548 {
5549         struct nfs_server *server = NFS_SERVER(data->inode);
5550
5551         if (data->commit_done_cb == NULL)
5552                 data->commit_done_cb = nfs4_commit_done_cb;
5553         data->res.server = server;
5554         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5555         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5556         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5557 }
5558
5559 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5560                                 struct nfs_commitres *res)
5561 {
5562         struct inode *dst_inode = file_inode(dst);
5563         struct nfs_server *server = NFS_SERVER(dst_inode);
5564         struct rpc_message msg = {
5565                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5566                 .rpc_argp = args,
5567                 .rpc_resp = res,
5568         };
5569
5570         args->fh = NFS_FH(dst_inode);
5571         return nfs4_call_sync(server->client, server, &msg,
5572                         &args->seq_args, &res->seq_res, 1);
5573 }
5574
5575 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5576 {
5577         struct nfs_commitargs args = {
5578                 .offset = offset,
5579                 .count = count,
5580         };
5581         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5582         struct nfs4_exception exception = { };
5583         int status;
5584
5585         do {
5586                 status = _nfs4_proc_commit(dst, &args, res);
5587                 status = nfs4_handle_exception(dst_server, status, &exception);
5588         } while (exception.retry);
5589
5590         return status;
5591 }
5592
5593 struct nfs4_renewdata {
5594         struct nfs_client       *client;
5595         unsigned long           timestamp;
5596 };
5597
5598 /*
5599  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5600  * standalone procedure for queueing an asynchronous RENEW.
5601  */
5602 static void nfs4_renew_release(void *calldata)
5603 {
5604         struct nfs4_renewdata *data = calldata;
5605         struct nfs_client *clp = data->client;
5606
5607         if (refcount_read(&clp->cl_count) > 1)
5608                 nfs4_schedule_state_renewal(clp);
5609         nfs_put_client(clp);
5610         kfree(data);
5611 }
5612
5613 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5614 {
5615         struct nfs4_renewdata *data = calldata;
5616         struct nfs_client *clp = data->client;
5617         unsigned long timestamp = data->timestamp;
5618
5619         trace_nfs4_renew_async(clp, task->tk_status);
5620         switch (task->tk_status) {
5621         case 0:
5622                 break;
5623         case -NFS4ERR_LEASE_MOVED:
5624                 nfs4_schedule_lease_moved_recovery(clp);
5625                 break;
5626         default:
5627                 /* Unless we're shutting down, schedule state recovery! */
5628                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5629                         return;
5630                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5631                         nfs4_schedule_lease_recovery(clp);
5632                         return;
5633                 }
5634                 nfs4_schedule_path_down_recovery(clp);
5635         }
5636         do_renew_lease(clp, timestamp);
5637 }
5638
5639 static const struct rpc_call_ops nfs4_renew_ops = {
5640         .rpc_call_done = nfs4_renew_done,
5641         .rpc_release = nfs4_renew_release,
5642 };
5643
5644 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5645 {
5646         struct rpc_message msg = {
5647                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5648                 .rpc_argp       = clp,
5649                 .rpc_cred       = cred,
5650         };
5651         struct nfs4_renewdata *data;
5652
5653         if (renew_flags == 0)
5654                 return 0;
5655         if (!refcount_inc_not_zero(&clp->cl_count))
5656                 return -EIO;
5657         data = kmalloc(sizeof(*data), GFP_NOFS);
5658         if (data == NULL) {
5659                 nfs_put_client(clp);
5660                 return -ENOMEM;
5661         }
5662         data->client = clp;
5663         data->timestamp = jiffies;
5664         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5665                         &nfs4_renew_ops, data);
5666 }
5667
5668 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5669 {
5670         struct rpc_message msg = {
5671                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5672                 .rpc_argp       = clp,
5673                 .rpc_cred       = cred,
5674         };
5675         unsigned long now = jiffies;
5676         int status;
5677
5678         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5679         if (status < 0)
5680                 return status;
5681         do_renew_lease(clp, now);
5682         return 0;
5683 }
5684
5685 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5686 {
5687         return server->caps & NFS_CAP_ACLS;
5688 }
5689
5690 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5691  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5692  * the stack.
5693  */
5694 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5695
5696 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5697                 struct page **pages)
5698 {
5699         struct page *newpage, **spages;
5700         int rc = 0;
5701         size_t len;
5702         spages = pages;
5703
5704         do {
5705                 len = min_t(size_t, PAGE_SIZE, buflen);
5706                 newpage = alloc_page(GFP_KERNEL);
5707
5708                 if (newpage == NULL)
5709                         goto unwind;
5710                 memcpy(page_address(newpage), buf, len);
5711                 buf += len;
5712                 buflen -= len;
5713                 *pages++ = newpage;
5714                 rc++;
5715         } while (buflen != 0);
5716
5717         return rc;
5718
5719 unwind:
5720         for(; rc > 0; rc--)
5721                 __free_page(spages[rc-1]);
5722         return -ENOMEM;
5723 }
5724
5725 struct nfs4_cached_acl {
5726         int cached;
5727         size_t len;
5728         char data[];
5729 };
5730
5731 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5732 {
5733         struct nfs_inode *nfsi = NFS_I(inode);
5734
5735         spin_lock(&inode->i_lock);
5736         kfree(nfsi->nfs4_acl);
5737         nfsi->nfs4_acl = acl;
5738         spin_unlock(&inode->i_lock);
5739 }
5740
5741 static void nfs4_zap_acl_attr(struct inode *inode)
5742 {
5743         nfs4_set_cached_acl(inode, NULL);
5744 }
5745
5746 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5747 {
5748         struct nfs_inode *nfsi = NFS_I(inode);
5749         struct nfs4_cached_acl *acl;
5750         int ret = -ENOENT;
5751
5752         spin_lock(&inode->i_lock);
5753         acl = nfsi->nfs4_acl;
5754         if (acl == NULL)
5755                 goto out;
5756         if (buf == NULL) /* user is just asking for length */
5757                 goto out_len;
5758         if (acl->cached == 0)
5759                 goto out;
5760         ret = -ERANGE; /* see getxattr(2) man page */
5761         if (acl->len > buflen)
5762                 goto out;
5763         memcpy(buf, acl->data, acl->len);
5764 out_len:
5765         ret = acl->len;
5766 out:
5767         spin_unlock(&inode->i_lock);
5768         return ret;
5769 }
5770
5771 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5772 {
5773         struct nfs4_cached_acl *acl;
5774         size_t buflen = sizeof(*acl) + acl_len;
5775
5776         if (buflen <= PAGE_SIZE) {
5777                 acl = kmalloc(buflen, GFP_KERNEL);
5778                 if (acl == NULL)
5779                         goto out;
5780                 acl->cached = 1;
5781                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5782         } else {
5783                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5784                 if (acl == NULL)
5785                         goto out;
5786                 acl->cached = 0;
5787         }
5788         acl->len = acl_len;
5789 out:
5790         nfs4_set_cached_acl(inode, acl);
5791 }
5792
5793 /*
5794  * The getxattr API returns the required buffer length when called with a
5795  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5796  * the required buf.  On a NULL buf, we send a page of data to the server
5797  * guessing that the ACL request can be serviced by a page. If so, we cache
5798  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5799  * the cache. If not so, we throw away the page, and cache the required
5800  * length. The next getxattr call will then produce another round trip to
5801  * the server, this time with the input buf of the required size.
5802  */
5803 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5804 {
5805         struct page **pages;
5806         struct nfs_getaclargs args = {
5807                 .fh = NFS_FH(inode),
5808                 .acl_len = buflen,
5809         };
5810         struct nfs_getaclres res = {
5811                 .acl_len = buflen,
5812         };
5813         struct rpc_message msg = {
5814                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5815                 .rpc_argp = &args,
5816                 .rpc_resp = &res,
5817         };
5818         unsigned int npages;
5819         int ret = -ENOMEM, i;
5820         struct nfs_server *server = NFS_SERVER(inode);
5821
5822         if (buflen == 0)
5823                 buflen = server->rsize;
5824
5825         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5826         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5827         if (!pages)
5828                 return -ENOMEM;
5829
5830         args.acl_pages = pages;
5831
5832         for (i = 0; i < npages; i++) {
5833                 pages[i] = alloc_page(GFP_KERNEL);
5834                 if (!pages[i])
5835                         goto out_free;
5836         }
5837
5838         /* for decoding across pages */
5839         res.acl_scratch = alloc_page(GFP_KERNEL);
5840         if (!res.acl_scratch)
5841                 goto out_free;
5842
5843         args.acl_len = npages * PAGE_SIZE;
5844
5845         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5846                 __func__, buf, buflen, npages, args.acl_len);
5847         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5848                              &msg, &args.seq_args, &res.seq_res, 0);
5849         if (ret)
5850                 goto out_free;
5851
5852         /* Handle the case where the passed-in buffer is too short */
5853         if (res.acl_flags & NFS4_ACL_TRUNC) {
5854                 /* Did the user only issue a request for the acl length? */
5855                 if (buf == NULL)
5856                         goto out_ok;
5857                 ret = -ERANGE;
5858                 goto out_free;
5859         }
5860         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5861         if (buf) {
5862                 if (res.acl_len > buflen) {
5863                         ret = -ERANGE;
5864                         goto out_free;
5865                 }
5866                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5867         }
5868 out_ok:
5869         ret = res.acl_len;
5870 out_free:
5871         for (i = 0; i < npages; i++)
5872                 if (pages[i])
5873                         __free_page(pages[i]);
5874         if (res.acl_scratch)
5875                 __free_page(res.acl_scratch);
5876         kfree(pages);
5877         return ret;
5878 }
5879
5880 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5881 {
5882         struct nfs4_exception exception = {
5883                 .interruptible = true,
5884         };
5885         ssize_t ret;
5886         do {
5887                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5888                 trace_nfs4_get_acl(inode, ret);
5889                 if (ret >= 0)
5890                         break;
5891                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5892         } while (exception.retry);
5893         return ret;
5894 }
5895
5896 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5897 {
5898         struct nfs_server *server = NFS_SERVER(inode);
5899         int ret;
5900
5901         if (!nfs4_server_supports_acls(server))
5902                 return -EOPNOTSUPP;
5903         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
5904         if (ret < 0)
5905                 return ret;
5906         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5907                 nfs_zap_acl_cache(inode);
5908         ret = nfs4_read_cached_acl(inode, buf, buflen);
5909         if (ret != -ENOENT)
5910                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5911                  * but no cached acl data, just the acl length */
5912                 return ret;
5913         return nfs4_get_acl_uncached(inode, buf, buflen);
5914 }
5915
5916 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5917 {
5918         struct nfs_server *server = NFS_SERVER(inode);
5919         struct page *pages[NFS4ACL_MAXPAGES];
5920         struct nfs_setaclargs arg = {
5921                 .fh             = NFS_FH(inode),
5922                 .acl_pages      = pages,
5923                 .acl_len        = buflen,
5924         };
5925         struct nfs_setaclres res;
5926         struct rpc_message msg = {
5927                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5928                 .rpc_argp       = &arg,
5929                 .rpc_resp       = &res,
5930         };
5931         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5932         int ret, i;
5933
5934         /* You can't remove system.nfs4_acl: */
5935         if (buflen == 0)
5936                 return -EINVAL;
5937         if (!nfs4_server_supports_acls(server))
5938                 return -EOPNOTSUPP;
5939         if (npages > ARRAY_SIZE(pages))
5940                 return -ERANGE;
5941         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5942         if (i < 0)
5943                 return i;
5944         nfs4_inode_make_writeable(inode);
5945         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5946
5947         /*
5948          * Free each page after tx, so the only ref left is
5949          * held by the network stack
5950          */
5951         for (; i > 0; i--)
5952                 put_page(pages[i-1]);
5953
5954         /*
5955          * Acl update can result in inode attribute update.
5956          * so mark the attribute cache invalid.
5957          */
5958         spin_lock(&inode->i_lock);
5959         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
5960                                              NFS_INO_INVALID_CTIME |
5961                                              NFS_INO_REVAL_FORCED);
5962         spin_unlock(&inode->i_lock);
5963         nfs_access_zap_cache(inode);
5964         nfs_zap_acl_cache(inode);
5965         return ret;
5966 }
5967
5968 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5969 {
5970         struct nfs4_exception exception = { };
5971         int err;
5972         do {
5973                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5974                 trace_nfs4_set_acl(inode, err);
5975                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
5976                         /*
5977                          * no need to retry since the kernel
5978                          * isn't involved in encoding the ACEs.
5979                          */
5980                         err = -EINVAL;
5981                         break;
5982                 }
5983                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5984                                 &exception);
5985         } while (exception.retry);
5986         return err;
5987 }
5988
5989 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5990 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5991                                         size_t buflen)
5992 {
5993         struct nfs_server *server = NFS_SERVER(inode);
5994         struct nfs_fattr fattr;
5995         struct nfs4_label label = {0, 0, buflen, buf};
5996
5997         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5998         struct nfs4_getattr_arg arg = {
5999                 .fh             = NFS_FH(inode),
6000                 .bitmask        = bitmask,
6001         };
6002         struct nfs4_getattr_res res = {
6003                 .fattr          = &fattr,
6004                 .label          = &label,
6005                 .server         = server,
6006         };
6007         struct rpc_message msg = {
6008                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6009                 .rpc_argp       = &arg,
6010                 .rpc_resp       = &res,
6011         };
6012         int ret;
6013
6014         nfs_fattr_init(&fattr);
6015
6016         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6017         if (ret)
6018                 return ret;
6019         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6020                 return -ENOENT;
6021         return label.len;
6022 }
6023
6024 static int nfs4_get_security_label(struct inode *inode, void *buf,
6025                                         size_t buflen)
6026 {
6027         struct nfs4_exception exception = {
6028                 .interruptible = true,
6029         };
6030         int err;
6031
6032         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6033                 return -EOPNOTSUPP;
6034
6035         do {
6036                 err = _nfs4_get_security_label(inode, buf, buflen);
6037                 trace_nfs4_get_security_label(inode, err);
6038                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6039                                 &exception);
6040         } while (exception.retry);
6041         return err;
6042 }
6043
6044 static int _nfs4_do_set_security_label(struct inode *inode,
6045                 struct nfs4_label *ilabel,
6046                 struct nfs_fattr *fattr,
6047                 struct nfs4_label *olabel)
6048 {
6049
6050         struct iattr sattr = {0};
6051         struct nfs_server *server = NFS_SERVER(inode);
6052         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6053         struct nfs_setattrargs arg = {
6054                 .fh             = NFS_FH(inode),
6055                 .iap            = &sattr,
6056                 .server         = server,
6057                 .bitmask        = bitmask,
6058                 .label          = ilabel,
6059         };
6060         struct nfs_setattrres res = {
6061                 .fattr          = fattr,
6062                 .label          = olabel,
6063                 .server         = server,
6064         };
6065         struct rpc_message msg = {
6066                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6067                 .rpc_argp       = &arg,
6068                 .rpc_resp       = &res,
6069         };
6070         int status;
6071
6072         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6073
6074         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6075         if (status)
6076                 dprintk("%s failed: %d\n", __func__, status);
6077
6078         return status;
6079 }
6080
6081 static int nfs4_do_set_security_label(struct inode *inode,
6082                 struct nfs4_label *ilabel,
6083                 struct nfs_fattr *fattr,
6084                 struct nfs4_label *olabel)
6085 {
6086         struct nfs4_exception exception = { };
6087         int err;
6088
6089         do {
6090                 err = _nfs4_do_set_security_label(inode, ilabel,
6091                                 fattr, olabel);
6092                 trace_nfs4_set_security_label(inode, err);
6093                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6094                                 &exception);
6095         } while (exception.retry);
6096         return err;
6097 }
6098
6099 static int
6100 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6101 {
6102         struct nfs4_label ilabel, *olabel = NULL;
6103         struct nfs_fattr fattr;
6104         int status;
6105
6106         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6107                 return -EOPNOTSUPP;
6108
6109         nfs_fattr_init(&fattr);
6110
6111         ilabel.pi = 0;
6112         ilabel.lfs = 0;
6113         ilabel.label = (char *)buf;
6114         ilabel.len = buflen;
6115
6116         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6117         if (IS_ERR(olabel)) {
6118                 status = -PTR_ERR(olabel);
6119                 goto out;
6120         }
6121
6122         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6123         if (status == 0)
6124                 nfs_setsecurity(inode, &fattr, olabel);
6125
6126         nfs4_label_free(olabel);
6127 out:
6128         return status;
6129 }
6130 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6131
6132
6133 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6134                                     nfs4_verifier *bootverf)
6135 {
6136         __be32 verf[2];
6137
6138         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6139                 /* An impossible timestamp guarantees this value
6140                  * will never match a generated boot time. */
6141                 verf[0] = cpu_to_be32(U32_MAX);
6142                 verf[1] = cpu_to_be32(U32_MAX);
6143         } else {
6144                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6145                 u64 ns = ktime_to_ns(nn->boot_time);
6146
6147                 verf[0] = cpu_to_be32(ns >> 32);
6148                 verf[1] = cpu_to_be32(ns);
6149         }
6150         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6151 }
6152
6153 static size_t
6154 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6155 {
6156         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6157         struct nfs_netns_client *nn_clp = nn->nfs_client;
6158         const char *id;
6159
6160         buf[0] = '\0';
6161
6162         if (nn_clp) {
6163                 rcu_read_lock();
6164                 id = rcu_dereference(nn_clp->identifier);
6165                 if (id)
6166                         strscpy(buf, id, buflen);
6167                 rcu_read_unlock();
6168         }
6169
6170         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6171                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6172
6173         return strlen(buf);
6174 }
6175
6176 static int
6177 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6178 {
6179         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6180         size_t buflen;
6181         size_t len;
6182         char *str;
6183
6184         if (clp->cl_owner_id != NULL)
6185                 return 0;
6186
6187         rcu_read_lock();
6188         len = 14 +
6189                 strlen(clp->cl_rpcclient->cl_nodename) +
6190                 1 +
6191                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6192                 1;
6193         rcu_read_unlock();
6194
6195         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6196         if (buflen)
6197                 len += buflen + 1;
6198
6199         if (len > NFS4_OPAQUE_LIMIT + 1)
6200                 return -EINVAL;
6201
6202         /*
6203          * Since this string is allocated at mount time, and held until the
6204          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6205          * about a memory-reclaim deadlock.
6206          */
6207         str = kmalloc(len, GFP_KERNEL);
6208         if (!str)
6209                 return -ENOMEM;
6210
6211         rcu_read_lock();
6212         if (buflen)
6213                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6214                           clp->cl_rpcclient->cl_nodename, buf,
6215                           rpc_peeraddr2str(clp->cl_rpcclient,
6216                                            RPC_DISPLAY_ADDR));
6217         else
6218                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6219                           clp->cl_rpcclient->cl_nodename,
6220                           rpc_peeraddr2str(clp->cl_rpcclient,
6221                                            RPC_DISPLAY_ADDR));
6222         rcu_read_unlock();
6223
6224         clp->cl_owner_id = str;
6225         return 0;
6226 }
6227
6228 static int
6229 nfs4_init_uniform_client_string(struct nfs_client *clp)
6230 {
6231         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6232         size_t buflen;
6233         size_t len;
6234         char *str;
6235
6236         if (clp->cl_owner_id != NULL)
6237                 return 0;
6238
6239         len = 10 + 10 + 1 + 10 + 1 +
6240                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6241
6242         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6243         if (buflen)
6244                 len += buflen + 1;
6245
6246         if (len > NFS4_OPAQUE_LIMIT + 1)
6247                 return -EINVAL;
6248
6249         /*
6250          * Since this string is allocated at mount time, and held until the
6251          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6252          * about a memory-reclaim deadlock.
6253          */
6254         str = kmalloc(len, GFP_KERNEL);
6255         if (!str)
6256                 return -ENOMEM;
6257
6258         if (buflen)
6259                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6260                           clp->rpc_ops->version, clp->cl_minorversion,
6261                           buf, clp->cl_rpcclient->cl_nodename);
6262         else
6263                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6264                           clp->rpc_ops->version, clp->cl_minorversion,
6265                           clp->cl_rpcclient->cl_nodename);
6266         clp->cl_owner_id = str;
6267         return 0;
6268 }
6269
6270 /*
6271  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6272  * services.  Advertise one based on the address family of the
6273  * clientaddr.
6274  */
6275 static unsigned int
6276 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6277 {
6278         if (strchr(clp->cl_ipaddr, ':') != NULL)
6279                 return scnprintf(buf, len, "tcp6");
6280         else
6281                 return scnprintf(buf, len, "tcp");
6282 }
6283
6284 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6285 {
6286         struct nfs4_setclientid *sc = calldata;
6287
6288         if (task->tk_status == 0)
6289                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6290 }
6291
6292 static const struct rpc_call_ops nfs4_setclientid_ops = {
6293         .rpc_call_done = nfs4_setclientid_done,
6294 };
6295
6296 /**
6297  * nfs4_proc_setclientid - Negotiate client ID
6298  * @clp: state data structure
6299  * @program: RPC program for NFSv4 callback service
6300  * @port: IP port number for NFS4 callback service
6301  * @cred: credential to use for this call
6302  * @res: where to place the result
6303  *
6304  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6305  */
6306 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6307                 unsigned short port, const struct cred *cred,
6308                 struct nfs4_setclientid_res *res)
6309 {
6310         nfs4_verifier sc_verifier;
6311         struct nfs4_setclientid setclientid = {
6312                 .sc_verifier = &sc_verifier,
6313                 .sc_prog = program,
6314                 .sc_clnt = clp,
6315         };
6316         struct rpc_message msg = {
6317                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6318                 .rpc_argp = &setclientid,
6319                 .rpc_resp = res,
6320                 .rpc_cred = cred,
6321         };
6322         struct rpc_task_setup task_setup_data = {
6323                 .rpc_client = clp->cl_rpcclient,
6324                 .rpc_message = &msg,
6325                 .callback_ops = &nfs4_setclientid_ops,
6326                 .callback_data = &setclientid,
6327                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6328         };
6329         unsigned long now = jiffies;
6330         int status;
6331
6332         /* nfs_client_id4 */
6333         nfs4_init_boot_verifier(clp, &sc_verifier);
6334
6335         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6336                 status = nfs4_init_uniform_client_string(clp);
6337         else
6338                 status = nfs4_init_nonuniform_client_string(clp);
6339
6340         if (status)
6341                 goto out;
6342
6343         /* cb_client4 */
6344         setclientid.sc_netid_len =
6345                                 nfs4_init_callback_netid(clp,
6346                                                 setclientid.sc_netid,
6347                                                 sizeof(setclientid.sc_netid));
6348         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6349                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6350                                 clp->cl_ipaddr, port >> 8, port & 255);
6351
6352         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6353                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6354                 clp->cl_owner_id);
6355
6356         status = nfs4_call_sync_custom(&task_setup_data);
6357         if (setclientid.sc_cred) {
6358                 kfree(clp->cl_acceptor);
6359                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6360                 put_rpccred(setclientid.sc_cred);
6361         }
6362
6363         if (status == 0)
6364                 do_renew_lease(clp, now);
6365 out:
6366         trace_nfs4_setclientid(clp, status);
6367         dprintk("NFS reply setclientid: %d\n", status);
6368         return status;
6369 }
6370
6371 /**
6372  * nfs4_proc_setclientid_confirm - Confirm client ID
6373  * @clp: state data structure
6374  * @arg: result of a previous SETCLIENTID
6375  * @cred: credential to use for this call
6376  *
6377  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6378  */
6379 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6380                 struct nfs4_setclientid_res *arg,
6381                 const struct cred *cred)
6382 {
6383         struct rpc_message msg = {
6384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6385                 .rpc_argp = arg,
6386                 .rpc_cred = cred,
6387         };
6388         int status;
6389
6390         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6391                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6392                 clp->cl_clientid);
6393         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6394                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6395         trace_nfs4_setclientid_confirm(clp, status);
6396         dprintk("NFS reply setclientid_confirm: %d\n", status);
6397         return status;
6398 }
6399
6400 struct nfs4_delegreturndata {
6401         struct nfs4_delegreturnargs args;
6402         struct nfs4_delegreturnres res;
6403         struct nfs_fh fh;
6404         nfs4_stateid stateid;
6405         unsigned long timestamp;
6406         struct {
6407                 struct nfs4_layoutreturn_args arg;
6408                 struct nfs4_layoutreturn_res res;
6409                 struct nfs4_xdr_opaque_data ld_private;
6410                 u32 roc_barrier;
6411                 bool roc;
6412         } lr;
6413         struct nfs_fattr fattr;
6414         int rpc_status;
6415         struct inode *inode;
6416 };
6417
6418 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6419 {
6420         struct nfs4_delegreturndata *data = calldata;
6421         struct nfs4_exception exception = {
6422                 .inode = data->inode,
6423                 .stateid = &data->stateid,
6424                 .task_is_privileged = data->args.seq_args.sa_privileged,
6425         };
6426
6427         if (!nfs4_sequence_done(task, &data->res.seq_res))
6428                 return;
6429
6430         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6431
6432         /* Handle Layoutreturn errors */
6433         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6434                           &data->res.lr_ret) == -EAGAIN)
6435                 goto out_restart;
6436
6437         switch (task->tk_status) {
6438         case 0:
6439                 renew_lease(data->res.server, data->timestamp);
6440                 break;
6441         case -NFS4ERR_ADMIN_REVOKED:
6442         case -NFS4ERR_DELEG_REVOKED:
6443         case -NFS4ERR_EXPIRED:
6444                 nfs4_free_revoked_stateid(data->res.server,
6445                                 data->args.stateid,
6446                                 task->tk_msg.rpc_cred);
6447                 fallthrough;
6448         case -NFS4ERR_BAD_STATEID:
6449         case -NFS4ERR_STALE_STATEID:
6450         case -ETIMEDOUT:
6451                 task->tk_status = 0;
6452                 break;
6453         case -NFS4ERR_OLD_STATEID:
6454                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6455                         nfs4_stateid_seqid_inc(&data->stateid);
6456                 if (data->args.bitmask) {
6457                         data->args.bitmask = NULL;
6458                         data->res.fattr = NULL;
6459                 }
6460                 goto out_restart;
6461         case -NFS4ERR_ACCESS:
6462                 if (data->args.bitmask) {
6463                         data->args.bitmask = NULL;
6464                         data->res.fattr = NULL;
6465                         goto out_restart;
6466                 }
6467                 fallthrough;
6468         default:
6469                 task->tk_status = nfs4_async_handle_exception(task,
6470                                 data->res.server, task->tk_status,
6471                                 &exception);
6472                 if (exception.retry)
6473                         goto out_restart;
6474         }
6475         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6476         data->rpc_status = task->tk_status;
6477         return;
6478 out_restart:
6479         task->tk_status = 0;
6480         rpc_restart_call_prepare(task);
6481 }
6482
6483 static void nfs4_delegreturn_release(void *calldata)
6484 {
6485         struct nfs4_delegreturndata *data = calldata;
6486         struct inode *inode = data->inode;
6487
6488         if (data->lr.roc)
6489                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6490                                  data->res.lr_ret);
6491         if (inode) {
6492                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6493                 nfs_iput_and_deactive(inode);
6494         }
6495         kfree(calldata);
6496 }
6497
6498 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6499 {
6500         struct nfs4_delegreturndata *d_data;
6501         struct pnfs_layout_hdr *lo;
6502
6503         d_data = (struct nfs4_delegreturndata *)data;
6504
6505         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6506                 nfs4_sequence_done(task, &d_data->res.seq_res);
6507                 return;
6508         }
6509
6510         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6511         if (lo && !pnfs_layout_is_valid(lo)) {
6512                 d_data->args.lr_args = NULL;
6513                 d_data->res.lr_res = NULL;
6514         }
6515
6516         nfs4_setup_sequence(d_data->res.server->nfs_client,
6517                         &d_data->args.seq_args,
6518                         &d_data->res.seq_res,
6519                         task);
6520 }
6521
6522 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6523         .rpc_call_prepare = nfs4_delegreturn_prepare,
6524         .rpc_call_done = nfs4_delegreturn_done,
6525         .rpc_release = nfs4_delegreturn_release,
6526 };
6527
6528 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6529 {
6530         struct nfs4_delegreturndata *data;
6531         struct nfs_server *server = NFS_SERVER(inode);
6532         struct rpc_task *task;
6533         struct rpc_message msg = {
6534                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6535                 .rpc_cred = cred,
6536         };
6537         struct rpc_task_setup task_setup_data = {
6538                 .rpc_client = server->client,
6539                 .rpc_message = &msg,
6540                 .callback_ops = &nfs4_delegreturn_ops,
6541                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6542         };
6543         int status = 0;
6544
6545         data = kzalloc(sizeof(*data), GFP_NOFS);
6546         if (data == NULL)
6547                 return -ENOMEM;
6548
6549         nfs4_state_protect(server->nfs_client,
6550                         NFS_SP4_MACH_CRED_CLEANUP,
6551                         &task_setup_data.rpc_client, &msg);
6552
6553         data->args.fhandle = &data->fh;
6554         data->args.stateid = &data->stateid;
6555         nfs4_bitmask_set(data->args.bitmask_store,
6556                          server->cache_consistency_bitmask, inode, server,
6557                          NULL);
6558         data->args.bitmask = data->args.bitmask_store;
6559         nfs_copy_fh(&data->fh, NFS_FH(inode));
6560         nfs4_stateid_copy(&data->stateid, stateid);
6561         data->res.fattr = &data->fattr;
6562         data->res.server = server;
6563         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6564         data->lr.arg.ld_private = &data->lr.ld_private;
6565         nfs_fattr_init(data->res.fattr);
6566         data->timestamp = jiffies;
6567         data->rpc_status = 0;
6568         data->inode = nfs_igrab_and_active(inode);
6569         if (data->inode || issync) {
6570                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6571                                         cred);
6572                 if (data->lr.roc) {
6573                         data->args.lr_args = &data->lr.arg;
6574                         data->res.lr_res = &data->lr.res;
6575                 }
6576         }
6577
6578         if (!data->inode)
6579                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6580                                    1);
6581         else
6582                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6583                                    0);
6584         task_setup_data.callback_data = data;
6585         msg.rpc_argp = &data->args;
6586         msg.rpc_resp = &data->res;
6587         task = rpc_run_task(&task_setup_data);
6588         if (IS_ERR(task))
6589                 return PTR_ERR(task);
6590         if (!issync)
6591                 goto out;
6592         status = rpc_wait_for_completion_task(task);
6593         if (status != 0)
6594                 goto out;
6595         status = data->rpc_status;
6596 out:
6597         rpc_put_task(task);
6598         return status;
6599 }
6600
6601 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6602 {
6603         struct nfs_server *server = NFS_SERVER(inode);
6604         struct nfs4_exception exception = { };
6605         int err;
6606         do {
6607                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6608                 trace_nfs4_delegreturn(inode, stateid, err);
6609                 switch (err) {
6610                         case -NFS4ERR_STALE_STATEID:
6611                         case -NFS4ERR_EXPIRED:
6612                         case 0:
6613                                 return 0;
6614                 }
6615                 err = nfs4_handle_exception(server, err, &exception);
6616         } while (exception.retry);
6617         return err;
6618 }
6619
6620 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6621 {
6622         struct inode *inode = state->inode;
6623         struct nfs_server *server = NFS_SERVER(inode);
6624         struct nfs_client *clp = server->nfs_client;
6625         struct nfs_lockt_args arg = {
6626                 .fh = NFS_FH(inode),
6627                 .fl = request,
6628         };
6629         struct nfs_lockt_res res = {
6630                 .denied = request,
6631         };
6632         struct rpc_message msg = {
6633                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6634                 .rpc_argp       = &arg,
6635                 .rpc_resp       = &res,
6636                 .rpc_cred       = state->owner->so_cred,
6637         };
6638         struct nfs4_lock_state *lsp;
6639         int status;
6640
6641         arg.lock_owner.clientid = clp->cl_clientid;
6642         status = nfs4_set_lock_state(state, request);
6643         if (status != 0)
6644                 goto out;
6645         lsp = request->fl_u.nfs4_fl.owner;
6646         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6647         arg.lock_owner.s_dev = server->s_dev;
6648         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6649         switch (status) {
6650                 case 0:
6651                         request->fl_type = F_UNLCK;
6652                         break;
6653                 case -NFS4ERR_DENIED:
6654                         status = 0;
6655         }
6656         request->fl_ops->fl_release_private(request);
6657         request->fl_ops = NULL;
6658 out:
6659         return status;
6660 }
6661
6662 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6663 {
6664         struct nfs4_exception exception = {
6665                 .interruptible = true,
6666         };
6667         int err;
6668
6669         do {
6670                 err = _nfs4_proc_getlk(state, cmd, request);
6671                 trace_nfs4_get_lock(request, state, cmd, err);
6672                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6673                                 &exception);
6674         } while (exception.retry);
6675         return err;
6676 }
6677
6678 /*
6679  * Update the seqid of a lock stateid after receiving
6680  * NFS4ERR_OLD_STATEID
6681  */
6682 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6683                 struct nfs4_lock_state *lsp)
6684 {
6685         struct nfs4_state *state = lsp->ls_state;
6686         bool ret = false;
6687
6688         spin_lock(&state->state_lock);
6689         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6690                 goto out;
6691         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6692                 nfs4_stateid_seqid_inc(dst);
6693         else
6694                 dst->seqid = lsp->ls_stateid.seqid;
6695         ret = true;
6696 out:
6697         spin_unlock(&state->state_lock);
6698         return ret;
6699 }
6700
6701 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6702                 struct nfs4_lock_state *lsp)
6703 {
6704         struct nfs4_state *state = lsp->ls_state;
6705         bool ret;
6706
6707         spin_lock(&state->state_lock);
6708         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6709         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6710         spin_unlock(&state->state_lock);
6711         return ret;
6712 }
6713
6714 struct nfs4_unlockdata {
6715         struct nfs_locku_args arg;
6716         struct nfs_locku_res res;
6717         struct nfs4_lock_state *lsp;
6718         struct nfs_open_context *ctx;
6719         struct nfs_lock_context *l_ctx;
6720         struct file_lock fl;
6721         struct nfs_server *server;
6722         unsigned long timestamp;
6723 };
6724
6725 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6726                 struct nfs_open_context *ctx,
6727                 struct nfs4_lock_state *lsp,
6728                 struct nfs_seqid *seqid)
6729 {
6730         struct nfs4_unlockdata *p;
6731         struct nfs4_state *state = lsp->ls_state;
6732         struct inode *inode = state->inode;
6733
6734         p = kzalloc(sizeof(*p), GFP_NOFS);
6735         if (p == NULL)
6736                 return NULL;
6737         p->arg.fh = NFS_FH(inode);
6738         p->arg.fl = &p->fl;
6739         p->arg.seqid = seqid;
6740         p->res.seqid = seqid;
6741         p->lsp = lsp;
6742         /* Ensure we don't close file until we're done freeing locks! */
6743         p->ctx = get_nfs_open_context(ctx);
6744         p->l_ctx = nfs_get_lock_context(ctx);
6745         locks_init_lock(&p->fl);
6746         locks_copy_lock(&p->fl, fl);
6747         p->server = NFS_SERVER(inode);
6748         spin_lock(&state->state_lock);
6749         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6750         spin_unlock(&state->state_lock);
6751         return p;
6752 }
6753
6754 static void nfs4_locku_release_calldata(void *data)
6755 {
6756         struct nfs4_unlockdata *calldata = data;
6757         nfs_free_seqid(calldata->arg.seqid);
6758         nfs4_put_lock_state(calldata->lsp);
6759         nfs_put_lock_context(calldata->l_ctx);
6760         put_nfs_open_context(calldata->ctx);
6761         kfree(calldata);
6762 }
6763
6764 static void nfs4_locku_done(struct rpc_task *task, void *data)
6765 {
6766         struct nfs4_unlockdata *calldata = data;
6767         struct nfs4_exception exception = {
6768                 .inode = calldata->lsp->ls_state->inode,
6769                 .stateid = &calldata->arg.stateid,
6770         };
6771
6772         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6773                 return;
6774         switch (task->tk_status) {
6775                 case 0:
6776                         renew_lease(calldata->server, calldata->timestamp);
6777                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6778                         if (nfs4_update_lock_stateid(calldata->lsp,
6779                                         &calldata->res.stateid))
6780                                 break;
6781                         fallthrough;
6782                 case -NFS4ERR_ADMIN_REVOKED:
6783                 case -NFS4ERR_EXPIRED:
6784                         nfs4_free_revoked_stateid(calldata->server,
6785                                         &calldata->arg.stateid,
6786                                         task->tk_msg.rpc_cred);
6787                         fallthrough;
6788                 case -NFS4ERR_BAD_STATEID:
6789                 case -NFS4ERR_STALE_STATEID:
6790                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6791                                                 calldata->lsp))
6792                                 rpc_restart_call_prepare(task);
6793                         break;
6794                 case -NFS4ERR_OLD_STATEID:
6795                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6796                                                 calldata->lsp))
6797                                 rpc_restart_call_prepare(task);
6798                         break;
6799                 default:
6800                         task->tk_status = nfs4_async_handle_exception(task,
6801                                         calldata->server, task->tk_status,
6802                                         &exception);
6803                         if (exception.retry)
6804                                 rpc_restart_call_prepare(task);
6805         }
6806         nfs_release_seqid(calldata->arg.seqid);
6807 }
6808
6809 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6810 {
6811         struct nfs4_unlockdata *calldata = data;
6812
6813         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6814                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6815                 return;
6816
6817         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6818                 goto out_wait;
6819         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6820                 /* Note: exit _without_ running nfs4_locku_done */
6821                 goto out_no_action;
6822         }
6823         calldata->timestamp = jiffies;
6824         if (nfs4_setup_sequence(calldata->server->nfs_client,
6825                                 &calldata->arg.seq_args,
6826                                 &calldata->res.seq_res,
6827                                 task) != 0)
6828                 nfs_release_seqid(calldata->arg.seqid);
6829         return;
6830 out_no_action:
6831         task->tk_action = NULL;
6832 out_wait:
6833         nfs4_sequence_done(task, &calldata->res.seq_res);
6834 }
6835
6836 static const struct rpc_call_ops nfs4_locku_ops = {
6837         .rpc_call_prepare = nfs4_locku_prepare,
6838         .rpc_call_done = nfs4_locku_done,
6839         .rpc_release = nfs4_locku_release_calldata,
6840 };
6841
6842 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6843                 struct nfs_open_context *ctx,
6844                 struct nfs4_lock_state *lsp,
6845                 struct nfs_seqid *seqid)
6846 {
6847         struct nfs4_unlockdata *data;
6848         struct rpc_message msg = {
6849                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6850                 .rpc_cred = ctx->cred,
6851         };
6852         struct rpc_task_setup task_setup_data = {
6853                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6854                 .rpc_message = &msg,
6855                 .callback_ops = &nfs4_locku_ops,
6856                 .workqueue = nfsiod_workqueue,
6857                 .flags = RPC_TASK_ASYNC,
6858         };
6859
6860         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6861                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6862
6863         /* Ensure this is an unlock - when canceling a lock, the
6864          * canceled lock is passed in, and it won't be an unlock.
6865          */
6866         fl->fl_type = F_UNLCK;
6867         if (fl->fl_flags & FL_CLOSE)
6868                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6869
6870         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6871         if (data == NULL) {
6872                 nfs_free_seqid(seqid);
6873                 return ERR_PTR(-ENOMEM);
6874         }
6875
6876         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6877         msg.rpc_argp = &data->arg;
6878         msg.rpc_resp = &data->res;
6879         task_setup_data.callback_data = data;
6880         return rpc_run_task(&task_setup_data);
6881 }
6882
6883 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6884 {
6885         struct inode *inode = state->inode;
6886         struct nfs4_state_owner *sp = state->owner;
6887         struct nfs_inode *nfsi = NFS_I(inode);
6888         struct nfs_seqid *seqid;
6889         struct nfs4_lock_state *lsp;
6890         struct rpc_task *task;
6891         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6892         int status = 0;
6893         unsigned char fl_flags = request->fl_flags;
6894
6895         status = nfs4_set_lock_state(state, request);
6896         /* Unlock _before_ we do the RPC call */
6897         request->fl_flags |= FL_EXISTS;
6898         /* Exclude nfs_delegation_claim_locks() */
6899         mutex_lock(&sp->so_delegreturn_mutex);
6900         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6901         down_read(&nfsi->rwsem);
6902         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6903                 up_read(&nfsi->rwsem);
6904                 mutex_unlock(&sp->so_delegreturn_mutex);
6905                 goto out;
6906         }
6907         up_read(&nfsi->rwsem);
6908         mutex_unlock(&sp->so_delegreturn_mutex);
6909         if (status != 0)
6910                 goto out;
6911         /* Is this a delegated lock? */
6912         lsp = request->fl_u.nfs4_fl.owner;
6913         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6914                 goto out;
6915         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6916         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6917         status = -ENOMEM;
6918         if (IS_ERR(seqid))
6919                 goto out;
6920         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6921         status = PTR_ERR(task);
6922         if (IS_ERR(task))
6923                 goto out;
6924         status = rpc_wait_for_completion_task(task);
6925         rpc_put_task(task);
6926 out:
6927         request->fl_flags = fl_flags;
6928         trace_nfs4_unlock(request, state, F_SETLK, status);
6929         return status;
6930 }
6931
6932 struct nfs4_lockdata {
6933         struct nfs_lock_args arg;
6934         struct nfs_lock_res res;
6935         struct nfs4_lock_state *lsp;
6936         struct nfs_open_context *ctx;
6937         struct file_lock fl;
6938         unsigned long timestamp;
6939         int rpc_status;
6940         int cancelled;
6941         struct nfs_server *server;
6942 };
6943
6944 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6945                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6946                 gfp_t gfp_mask)
6947 {
6948         struct nfs4_lockdata *p;
6949         struct inode *inode = lsp->ls_state->inode;
6950         struct nfs_server *server = NFS_SERVER(inode);
6951         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6952
6953         p = kzalloc(sizeof(*p), gfp_mask);
6954         if (p == NULL)
6955                 return NULL;
6956
6957         p->arg.fh = NFS_FH(inode);
6958         p->arg.fl = &p->fl;
6959         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6960         if (IS_ERR(p->arg.open_seqid))
6961                 goto out_free;
6962         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6963         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6964         if (IS_ERR(p->arg.lock_seqid))
6965                 goto out_free_seqid;
6966         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6967         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6968         p->arg.lock_owner.s_dev = server->s_dev;
6969         p->res.lock_seqid = p->arg.lock_seqid;
6970         p->lsp = lsp;
6971         p->server = server;
6972         p->ctx = get_nfs_open_context(ctx);
6973         locks_init_lock(&p->fl);
6974         locks_copy_lock(&p->fl, fl);
6975         return p;
6976 out_free_seqid:
6977         nfs_free_seqid(p->arg.open_seqid);
6978 out_free:
6979         kfree(p);
6980         return NULL;
6981 }
6982
6983 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6984 {
6985         struct nfs4_lockdata *data = calldata;
6986         struct nfs4_state *state = data->lsp->ls_state;
6987
6988         dprintk("%s: begin!\n", __func__);
6989         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6990                 goto out_wait;
6991         /* Do we need to do an open_to_lock_owner? */
6992         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6993                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6994                         goto out_release_lock_seqid;
6995                 }
6996                 nfs4_stateid_copy(&data->arg.open_stateid,
6997                                 &state->open_stateid);
6998                 data->arg.new_lock_owner = 1;
6999                 data->res.open_seqid = data->arg.open_seqid;
7000         } else {
7001                 data->arg.new_lock_owner = 0;
7002                 nfs4_stateid_copy(&data->arg.lock_stateid,
7003                                 &data->lsp->ls_stateid);
7004         }
7005         if (!nfs4_valid_open_stateid(state)) {
7006                 data->rpc_status = -EBADF;
7007                 task->tk_action = NULL;
7008                 goto out_release_open_seqid;
7009         }
7010         data->timestamp = jiffies;
7011         if (nfs4_setup_sequence(data->server->nfs_client,
7012                                 &data->arg.seq_args,
7013                                 &data->res.seq_res,
7014                                 task) == 0)
7015                 return;
7016 out_release_open_seqid:
7017         nfs_release_seqid(data->arg.open_seqid);
7018 out_release_lock_seqid:
7019         nfs_release_seqid(data->arg.lock_seqid);
7020 out_wait:
7021         nfs4_sequence_done(task, &data->res.seq_res);
7022         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
7023 }
7024
7025 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7026 {
7027         struct nfs4_lockdata *data = calldata;
7028         struct nfs4_lock_state *lsp = data->lsp;
7029
7030         dprintk("%s: begin!\n", __func__);
7031
7032         if (!nfs4_sequence_done(task, &data->res.seq_res))
7033                 return;
7034
7035         data->rpc_status = task->tk_status;
7036         switch (task->tk_status) {
7037         case 0:
7038                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7039                                 data->timestamp);
7040                 if (data->arg.new_lock && !data->cancelled) {
7041                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7042                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7043                                 goto out_restart;
7044                 }
7045                 if (data->arg.new_lock_owner != 0) {
7046                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7047                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7048                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7049                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7050                         goto out_restart;
7051                 break;
7052         case -NFS4ERR_BAD_STATEID:
7053         case -NFS4ERR_OLD_STATEID:
7054         case -NFS4ERR_STALE_STATEID:
7055         case -NFS4ERR_EXPIRED:
7056                 if (data->arg.new_lock_owner != 0) {
7057                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7058                                                 &lsp->ls_state->open_stateid))
7059                                 goto out_restart;
7060                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7061                                                 &lsp->ls_stateid))
7062                                 goto out_restart;
7063         }
7064 out_done:
7065         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7066         return;
7067 out_restart:
7068         if (!data->cancelled)
7069                 rpc_restart_call_prepare(task);
7070         goto out_done;
7071 }
7072
7073 static void nfs4_lock_release(void *calldata)
7074 {
7075         struct nfs4_lockdata *data = calldata;
7076
7077         dprintk("%s: begin!\n", __func__);
7078         nfs_free_seqid(data->arg.open_seqid);
7079         if (data->cancelled && data->rpc_status == 0) {
7080                 struct rpc_task *task;
7081                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7082                                 data->arg.lock_seqid);
7083                 if (!IS_ERR(task))
7084                         rpc_put_task_async(task);
7085                 dprintk("%s: cancelling lock!\n", __func__);
7086         } else
7087                 nfs_free_seqid(data->arg.lock_seqid);
7088         nfs4_put_lock_state(data->lsp);
7089         put_nfs_open_context(data->ctx);
7090         kfree(data);
7091         dprintk("%s: done!\n", __func__);
7092 }
7093
7094 static const struct rpc_call_ops nfs4_lock_ops = {
7095         .rpc_call_prepare = nfs4_lock_prepare,
7096         .rpc_call_done = nfs4_lock_done,
7097         .rpc_release = nfs4_lock_release,
7098 };
7099
7100 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7101 {
7102         switch (error) {
7103         case -NFS4ERR_ADMIN_REVOKED:
7104         case -NFS4ERR_EXPIRED:
7105         case -NFS4ERR_BAD_STATEID:
7106                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7107                 if (new_lock_owner != 0 ||
7108                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7109                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7110                 break;
7111         case -NFS4ERR_STALE_STATEID:
7112                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7113                 nfs4_schedule_lease_recovery(server->nfs_client);
7114         }
7115 }
7116
7117 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7118 {
7119         struct nfs4_lockdata *data;
7120         struct rpc_task *task;
7121         struct rpc_message msg = {
7122                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7123                 .rpc_cred = state->owner->so_cred,
7124         };
7125         struct rpc_task_setup task_setup_data = {
7126                 .rpc_client = NFS_CLIENT(state->inode),
7127                 .rpc_message = &msg,
7128                 .callback_ops = &nfs4_lock_ops,
7129                 .workqueue = nfsiod_workqueue,
7130                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7131         };
7132         int ret;
7133
7134         dprintk("%s: begin!\n", __func__);
7135         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7136                         fl->fl_u.nfs4_fl.owner,
7137                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7138         if (data == NULL)
7139                 return -ENOMEM;
7140         if (IS_SETLKW(cmd))
7141                 data->arg.block = 1;
7142         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7143                                 recovery_type > NFS_LOCK_NEW);
7144         msg.rpc_argp = &data->arg;
7145         msg.rpc_resp = &data->res;
7146         task_setup_data.callback_data = data;
7147         if (recovery_type > NFS_LOCK_NEW) {
7148                 if (recovery_type == NFS_LOCK_RECLAIM)
7149                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7150         } else
7151                 data->arg.new_lock = 1;
7152         task = rpc_run_task(&task_setup_data);
7153         if (IS_ERR(task))
7154                 return PTR_ERR(task);
7155         ret = rpc_wait_for_completion_task(task);
7156         if (ret == 0) {
7157                 ret = data->rpc_status;
7158                 if (ret)
7159                         nfs4_handle_setlk_error(data->server, data->lsp,
7160                                         data->arg.new_lock_owner, ret);
7161         } else
7162                 data->cancelled = true;
7163         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7164         rpc_put_task(task);
7165         dprintk("%s: done, ret = %d!\n", __func__, ret);
7166         return ret;
7167 }
7168
7169 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7170 {
7171         struct nfs_server *server = NFS_SERVER(state->inode);
7172         struct nfs4_exception exception = {
7173                 .inode = state->inode,
7174         };
7175         int err;
7176
7177         do {
7178                 /* Cache the lock if possible... */
7179                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7180                         return 0;
7181                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7182                 if (err != -NFS4ERR_DELAY)
7183                         break;
7184                 nfs4_handle_exception(server, err, &exception);
7185         } while (exception.retry);
7186         return err;
7187 }
7188
7189 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7190 {
7191         struct nfs_server *server = NFS_SERVER(state->inode);
7192         struct nfs4_exception exception = {
7193                 .inode = state->inode,
7194         };
7195         int err;
7196
7197         err = nfs4_set_lock_state(state, request);
7198         if (err != 0)
7199                 return err;
7200         if (!recover_lost_locks) {
7201                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7202                 return 0;
7203         }
7204         do {
7205                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7206                         return 0;
7207                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7208                 switch (err) {
7209                 default:
7210                         goto out;
7211                 case -NFS4ERR_GRACE:
7212                 case -NFS4ERR_DELAY:
7213                         nfs4_handle_exception(server, err, &exception);
7214                         err = 0;
7215                 }
7216         } while (exception.retry);
7217 out:
7218         return err;
7219 }
7220
7221 #if defined(CONFIG_NFS_V4_1)
7222 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7223 {
7224         struct nfs4_lock_state *lsp;
7225         int status;
7226
7227         status = nfs4_set_lock_state(state, request);
7228         if (status != 0)
7229                 return status;
7230         lsp = request->fl_u.nfs4_fl.owner;
7231         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7232             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7233                 return 0;
7234         return nfs4_lock_expired(state, request);
7235 }
7236 #endif
7237
7238 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7239 {
7240         struct nfs_inode *nfsi = NFS_I(state->inode);
7241         struct nfs4_state_owner *sp = state->owner;
7242         unsigned char fl_flags = request->fl_flags;
7243         int status;
7244
7245         request->fl_flags |= FL_ACCESS;
7246         status = locks_lock_inode_wait(state->inode, request);
7247         if (status < 0)
7248                 goto out;
7249         mutex_lock(&sp->so_delegreturn_mutex);
7250         down_read(&nfsi->rwsem);
7251         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7252                 /* Yes: cache locks! */
7253                 /* ...but avoid races with delegation recall... */
7254                 request->fl_flags = fl_flags & ~FL_SLEEP;
7255                 status = locks_lock_inode_wait(state->inode, request);
7256                 up_read(&nfsi->rwsem);
7257                 mutex_unlock(&sp->so_delegreturn_mutex);
7258                 goto out;
7259         }
7260         up_read(&nfsi->rwsem);
7261         mutex_unlock(&sp->so_delegreturn_mutex);
7262         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7263 out:
7264         request->fl_flags = fl_flags;
7265         return status;
7266 }
7267
7268 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7269 {
7270         struct nfs4_exception exception = {
7271                 .state = state,
7272                 .inode = state->inode,
7273                 .interruptible = true,
7274         };
7275         int err;
7276
7277         do {
7278                 err = _nfs4_proc_setlk(state, cmd, request);
7279                 if (err == -NFS4ERR_DENIED)
7280                         err = -EAGAIN;
7281                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7282                                 err, &exception);
7283         } while (exception.retry);
7284         return err;
7285 }
7286
7287 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7288 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7289
7290 static int
7291 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7292                         struct file_lock *request)
7293 {
7294         int             status = -ERESTARTSYS;
7295         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7296
7297         while(!signalled()) {
7298                 status = nfs4_proc_setlk(state, cmd, request);
7299                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7300                         break;
7301                 freezable_schedule_timeout_interruptible(timeout);
7302                 timeout *= 2;
7303                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7304                 status = -ERESTARTSYS;
7305         }
7306         return status;
7307 }
7308
7309 #ifdef CONFIG_NFS_V4_1
7310 struct nfs4_lock_waiter {
7311         struct inode            *inode;
7312         struct nfs_lowner       owner;
7313         wait_queue_entry_t      wait;
7314 };
7315
7316 static int
7317 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7318 {
7319         struct nfs4_lock_waiter *waiter =
7320                 container_of(wait, struct nfs4_lock_waiter, wait);
7321
7322         /* NULL key means to wake up everyone */
7323         if (key) {
7324                 struct cb_notify_lock_args      *cbnl = key;
7325                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7326                                                 *wowner = &waiter->owner;
7327
7328                 /* Only wake if the callback was for the same owner. */
7329                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7330                         return 0;
7331
7332                 /* Make sure it's for the right inode */
7333                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7334                         return 0;
7335         }
7336
7337         return woken_wake_function(wait, mode, flags, key);
7338 }
7339
7340 static int
7341 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7342 {
7343         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7344         struct nfs_server *server = NFS_SERVER(state->inode);
7345         struct nfs_client *clp = server->nfs_client;
7346         wait_queue_head_t *q = &clp->cl_lock_waitq;
7347         struct nfs4_lock_waiter waiter = {
7348                 .inode = state->inode,
7349                 .owner = { .clientid = clp->cl_clientid,
7350                            .id = lsp->ls_seqid.owner_id,
7351                            .s_dev = server->s_dev },
7352         };
7353         int status;
7354
7355         /* Don't bother with waitqueue if we don't expect a callback */
7356         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7357                 return nfs4_retry_setlk_simple(state, cmd, request);
7358
7359         init_wait(&waiter.wait);
7360         waiter.wait.func = nfs4_wake_lock_waiter;
7361         add_wait_queue(q, &waiter.wait);
7362
7363         do {
7364                 status = nfs4_proc_setlk(state, cmd, request);
7365                 if (status != -EAGAIN || IS_SETLK(cmd))
7366                         break;
7367
7368                 status = -ERESTARTSYS;
7369                 freezer_do_not_count();
7370                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7371                            NFS4_LOCK_MAXTIMEOUT);
7372                 freezer_count();
7373         } while (!signalled());
7374
7375         remove_wait_queue(q, &waiter.wait);
7376
7377         return status;
7378 }
7379 #else /* !CONFIG_NFS_V4_1 */
7380 static inline int
7381 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7382 {
7383         return nfs4_retry_setlk_simple(state, cmd, request);
7384 }
7385 #endif
7386
7387 static int
7388 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7389 {
7390         struct nfs_open_context *ctx;
7391         struct nfs4_state *state;
7392         int status;
7393
7394         /* verify open state */
7395         ctx = nfs_file_open_context(filp);
7396         state = ctx->state;
7397
7398         if (IS_GETLK(cmd)) {
7399                 if (state != NULL)
7400                         return nfs4_proc_getlk(state, F_GETLK, request);
7401                 return 0;
7402         }
7403
7404         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7405                 return -EINVAL;
7406
7407         if (request->fl_type == F_UNLCK) {
7408                 if (state != NULL)
7409                         return nfs4_proc_unlck(state, cmd, request);
7410                 return 0;
7411         }
7412
7413         if (state == NULL)
7414                 return -ENOLCK;
7415
7416         if ((request->fl_flags & FL_POSIX) &&
7417             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7418                 return -ENOLCK;
7419
7420         /*
7421          * Don't rely on the VFS having checked the file open mode,
7422          * since it won't do this for flock() locks.
7423          */
7424         switch (request->fl_type) {
7425         case F_RDLCK:
7426                 if (!(filp->f_mode & FMODE_READ))
7427                         return -EBADF;
7428                 break;
7429         case F_WRLCK:
7430                 if (!(filp->f_mode & FMODE_WRITE))
7431                         return -EBADF;
7432         }
7433
7434         status = nfs4_set_lock_state(state, request);
7435         if (status != 0)
7436                 return status;
7437
7438         return nfs4_retry_setlk(state, cmd, request);
7439 }
7440
7441 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7442 {
7443         struct nfs_server *server = NFS_SERVER(state->inode);
7444         int err;
7445
7446         err = nfs4_set_lock_state(state, fl);
7447         if (err != 0)
7448                 return err;
7449         do {
7450                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7451                 if (err != -NFS4ERR_DELAY)
7452                         break;
7453                 ssleep(1);
7454         } while (err == -NFS4ERR_DELAY);
7455         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7456 }
7457
7458 struct nfs_release_lockowner_data {
7459         struct nfs4_lock_state *lsp;
7460         struct nfs_server *server;
7461         struct nfs_release_lockowner_args args;
7462         struct nfs_release_lockowner_res res;
7463         unsigned long timestamp;
7464 };
7465
7466 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7467 {
7468         struct nfs_release_lockowner_data *data = calldata;
7469         struct nfs_server *server = data->server;
7470         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7471                            &data->res.seq_res, task);
7472         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7473         data->timestamp = jiffies;
7474 }
7475
7476 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7477 {
7478         struct nfs_release_lockowner_data *data = calldata;
7479         struct nfs_server *server = data->server;
7480
7481         nfs40_sequence_done(task, &data->res.seq_res);
7482
7483         switch (task->tk_status) {
7484         case 0:
7485                 renew_lease(server, data->timestamp);
7486                 break;
7487         case -NFS4ERR_STALE_CLIENTID:
7488         case -NFS4ERR_EXPIRED:
7489                 nfs4_schedule_lease_recovery(server->nfs_client);
7490                 break;
7491         case -NFS4ERR_LEASE_MOVED:
7492         case -NFS4ERR_DELAY:
7493                 if (nfs4_async_handle_error(task, server,
7494                                             NULL, NULL) == -EAGAIN)
7495                         rpc_restart_call_prepare(task);
7496         }
7497 }
7498
7499 static void nfs4_release_lockowner_release(void *calldata)
7500 {
7501         struct nfs_release_lockowner_data *data = calldata;
7502         nfs4_free_lock_state(data->server, data->lsp);
7503         kfree(calldata);
7504 }
7505
7506 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7507         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7508         .rpc_call_done = nfs4_release_lockowner_done,
7509         .rpc_release = nfs4_release_lockowner_release,
7510 };
7511
7512 static void
7513 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7514 {
7515         struct nfs_release_lockowner_data *data;
7516         struct rpc_message msg = {
7517                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7518         };
7519
7520         if (server->nfs_client->cl_mvops->minor_version != 0)
7521                 return;
7522
7523         data = kmalloc(sizeof(*data), GFP_NOFS);
7524         if (!data)
7525                 return;
7526         data->lsp = lsp;
7527         data->server = server;
7528         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7529         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7530         data->args.lock_owner.s_dev = server->s_dev;
7531
7532         msg.rpc_argp = &data->args;
7533         msg.rpc_resp = &data->res;
7534         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7535         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7536 }
7537
7538 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7539
7540 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7541                                    struct user_namespace *mnt_userns,
7542                                    struct dentry *unused, struct inode *inode,
7543                                    const char *key, const void *buf,
7544                                    size_t buflen, int flags)
7545 {
7546         return nfs4_proc_set_acl(inode, buf, buflen);
7547 }
7548
7549 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7550                                    struct dentry *unused, struct inode *inode,
7551                                    const char *key, void *buf, size_t buflen)
7552 {
7553         return nfs4_proc_get_acl(inode, buf, buflen);
7554 }
7555
7556 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7557 {
7558         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7559 }
7560
7561 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7562
7563 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7564                                      struct user_namespace *mnt_userns,
7565                                      struct dentry *unused, struct inode *inode,
7566                                      const char *key, const void *buf,
7567                                      size_t buflen, int flags)
7568 {
7569         if (security_ismaclabel(key))
7570                 return nfs4_set_security_label(inode, buf, buflen);
7571
7572         return -EOPNOTSUPP;
7573 }
7574
7575 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7576                                      struct dentry *unused, struct inode *inode,
7577                                      const char *key, void *buf, size_t buflen)
7578 {
7579         if (security_ismaclabel(key))
7580                 return nfs4_get_security_label(inode, buf, buflen);
7581         return -EOPNOTSUPP;
7582 }
7583
7584 static ssize_t
7585 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7586 {
7587         int len = 0;
7588
7589         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7590                 len = security_inode_listsecurity(inode, list, list_len);
7591                 if (len >= 0 && list_len && len > list_len)
7592                         return -ERANGE;
7593         }
7594         return len;
7595 }
7596
7597 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7598         .prefix = XATTR_SECURITY_PREFIX,
7599         .get    = nfs4_xattr_get_nfs4_label,
7600         .set    = nfs4_xattr_set_nfs4_label,
7601 };
7602
7603 #else
7604
7605 static ssize_t
7606 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7607 {
7608         return 0;
7609 }
7610
7611 #endif
7612
7613 #ifdef CONFIG_NFS_V4_2
7614 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7615                                     struct user_namespace *mnt_userns,
7616                                     struct dentry *unused, struct inode *inode,
7617                                     const char *key, const void *buf,
7618                                     size_t buflen, int flags)
7619 {
7620         struct nfs_access_entry cache;
7621         int ret;
7622
7623         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7624                 return -EOPNOTSUPP;
7625
7626         /*
7627          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7628          * flags right now. Handling of xattr operations use the normal
7629          * file read/write permissions.
7630          *
7631          * Just in case the server has other ideas (which RFC 8276 allows),
7632          * do a cached access check for the XA* flags to possibly avoid
7633          * doing an RPC and getting EACCES back.
7634          */
7635         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7636                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7637                         return -EACCES;
7638         }
7639
7640         if (buf == NULL) {
7641                 ret = nfs42_proc_removexattr(inode, key);
7642                 if (!ret)
7643                         nfs4_xattr_cache_remove(inode, key);
7644         } else {
7645                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7646                 if (!ret)
7647                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7648         }
7649
7650         return ret;
7651 }
7652
7653 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7654                                     struct dentry *unused, struct inode *inode,
7655                                     const char *key, void *buf, size_t buflen)
7656 {
7657         struct nfs_access_entry cache;
7658         ssize_t ret;
7659
7660         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7661                 return -EOPNOTSUPP;
7662
7663         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7664                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7665                         return -EACCES;
7666         }
7667
7668         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7669         if (ret)
7670                 return ret;
7671
7672         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7673         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7674                 return ret;
7675
7676         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7677
7678         return ret;
7679 }
7680
7681 static ssize_t
7682 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7683 {
7684         u64 cookie;
7685         bool eof;
7686         ssize_t ret, size;
7687         char *buf;
7688         size_t buflen;
7689         struct nfs_access_entry cache;
7690
7691         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7692                 return 0;
7693
7694         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7695                 if (!(cache.mask & NFS_ACCESS_XALIST))
7696                         return 0;
7697         }
7698
7699         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7700         if (ret)
7701                 return ret;
7702
7703         ret = nfs4_xattr_cache_list(inode, list, list_len);
7704         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7705                 return ret;
7706
7707         cookie = 0;
7708         eof = false;
7709         buflen = list_len ? list_len : XATTR_LIST_MAX;
7710         buf = list_len ? list : NULL;
7711         size = 0;
7712
7713         while (!eof) {
7714                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7715                     &cookie, &eof);
7716                 if (ret < 0)
7717                         return ret;
7718
7719                 if (list_len) {
7720                         buf += ret;
7721                         buflen -= ret;
7722                 }
7723                 size += ret;
7724         }
7725
7726         if (list_len)
7727                 nfs4_xattr_cache_set_list(inode, list, size);
7728
7729         return size;
7730 }
7731
7732 #else
7733
7734 static ssize_t
7735 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7736 {
7737         return 0;
7738 }
7739 #endif /* CONFIG_NFS_V4_2 */
7740
7741 /*
7742  * nfs_fhget will use either the mounted_on_fileid or the fileid
7743  */
7744 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7745 {
7746         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7747                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7748               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7749               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7750                 return;
7751
7752         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7753                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7754         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7755         fattr->nlink = 2;
7756 }
7757
7758 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7759                                    const struct qstr *name,
7760                                    struct nfs4_fs_locations *fs_locations,
7761                                    struct page *page)
7762 {
7763         struct nfs_server *server = NFS_SERVER(dir);
7764         u32 bitmask[3];
7765         struct nfs4_fs_locations_arg args = {
7766                 .dir_fh = NFS_FH(dir),
7767                 .name = name,
7768                 .page = page,
7769                 .bitmask = bitmask,
7770         };
7771         struct nfs4_fs_locations_res res = {
7772                 .fs_locations = fs_locations,
7773         };
7774         struct rpc_message msg = {
7775                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7776                 .rpc_argp = &args,
7777                 .rpc_resp = &res,
7778         };
7779         int status;
7780
7781         dprintk("%s: start\n", __func__);
7782
7783         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7784         bitmask[1] = nfs4_fattr_bitmap[1];
7785
7786         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7787          * is not supported */
7788         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7789                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7790         else
7791                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7792
7793         nfs_fattr_init(&fs_locations->fattr);
7794         fs_locations->server = server;
7795         fs_locations->nlocations = 0;
7796         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7797         dprintk("%s: returned status = %d\n", __func__, status);
7798         return status;
7799 }
7800
7801 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7802                            const struct qstr *name,
7803                            struct nfs4_fs_locations *fs_locations,
7804                            struct page *page)
7805 {
7806         struct nfs4_exception exception = {
7807                 .interruptible = true,
7808         };
7809         int err;
7810         do {
7811                 err = _nfs4_proc_fs_locations(client, dir, name,
7812                                 fs_locations, page);
7813                 trace_nfs4_get_fs_locations(dir, name, err);
7814                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7815                                 &exception);
7816         } while (exception.retry);
7817         return err;
7818 }
7819
7820 /*
7821  * This operation also signals the server that this client is
7822  * performing migration recovery.  The server can stop returning
7823  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7824  * appended to this compound to identify the client ID which is
7825  * performing recovery.
7826  */
7827 static int _nfs40_proc_get_locations(struct inode *inode,
7828                                      struct nfs4_fs_locations *locations,
7829                                      struct page *page, const struct cred *cred)
7830 {
7831         struct nfs_server *server = NFS_SERVER(inode);
7832         struct rpc_clnt *clnt = server->client;
7833         u32 bitmask[2] = {
7834                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7835         };
7836         struct nfs4_fs_locations_arg args = {
7837                 .clientid       = server->nfs_client->cl_clientid,
7838                 .fh             = NFS_FH(inode),
7839                 .page           = page,
7840                 .bitmask        = bitmask,
7841                 .migration      = 1,            /* skip LOOKUP */
7842                 .renew          = 1,            /* append RENEW */
7843         };
7844         struct nfs4_fs_locations_res res = {
7845                 .fs_locations   = locations,
7846                 .migration      = 1,
7847                 .renew          = 1,
7848         };
7849         struct rpc_message msg = {
7850                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7851                 .rpc_argp       = &args,
7852                 .rpc_resp       = &res,
7853                 .rpc_cred       = cred,
7854         };
7855         unsigned long now = jiffies;
7856         int status;
7857
7858         nfs_fattr_init(&locations->fattr);
7859         locations->server = server;
7860         locations->nlocations = 0;
7861
7862         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7863         status = nfs4_call_sync_sequence(clnt, server, &msg,
7864                                         &args.seq_args, &res.seq_res);
7865         if (status)
7866                 return status;
7867
7868         renew_lease(server, now);
7869         return 0;
7870 }
7871
7872 #ifdef CONFIG_NFS_V4_1
7873
7874 /*
7875  * This operation also signals the server that this client is
7876  * performing migration recovery.  The server can stop asserting
7877  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7878  * performing this operation is identified in the SEQUENCE
7879  * operation in this compound.
7880  *
7881  * When the client supports GETATTR(fs_locations_info), it can
7882  * be plumbed in here.
7883  */
7884 static int _nfs41_proc_get_locations(struct inode *inode,
7885                                      struct nfs4_fs_locations *locations,
7886                                      struct page *page, const struct cred *cred)
7887 {
7888         struct nfs_server *server = NFS_SERVER(inode);
7889         struct rpc_clnt *clnt = server->client;
7890         u32 bitmask[2] = {
7891                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7892         };
7893         struct nfs4_fs_locations_arg args = {
7894                 .fh             = NFS_FH(inode),
7895                 .page           = page,
7896                 .bitmask        = bitmask,
7897                 .migration      = 1,            /* skip LOOKUP */
7898         };
7899         struct nfs4_fs_locations_res res = {
7900                 .fs_locations   = locations,
7901                 .migration      = 1,
7902         };
7903         struct rpc_message msg = {
7904                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7905                 .rpc_argp       = &args,
7906                 .rpc_resp       = &res,
7907                 .rpc_cred       = cred,
7908         };
7909         int status;
7910
7911         nfs_fattr_init(&locations->fattr);
7912         locations->server = server;
7913         locations->nlocations = 0;
7914
7915         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7916         status = nfs4_call_sync_sequence(clnt, server, &msg,
7917                                         &args.seq_args, &res.seq_res);
7918         if (status == NFS4_OK &&
7919             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7920                 status = -NFS4ERR_LEASE_MOVED;
7921         return status;
7922 }
7923
7924 #endif  /* CONFIG_NFS_V4_1 */
7925
7926 /**
7927  * nfs4_proc_get_locations - discover locations for a migrated FSID
7928  * @inode: inode on FSID that is migrating
7929  * @locations: result of query
7930  * @page: buffer
7931  * @cred: credential to use for this operation
7932  *
7933  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7934  * operation failed, or a negative errno if a local error occurred.
7935  *
7936  * On success, "locations" is filled in, but if the server has
7937  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7938  * asserted.
7939  *
7940  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7941  * from this client that require migration recovery.
7942  */
7943 int nfs4_proc_get_locations(struct inode *inode,
7944                             struct nfs4_fs_locations *locations,
7945                             struct page *page, const struct cred *cred)
7946 {
7947         struct nfs_server *server = NFS_SERVER(inode);
7948         struct nfs_client *clp = server->nfs_client;
7949         const struct nfs4_mig_recovery_ops *ops =
7950                                         clp->cl_mvops->mig_recovery_ops;
7951         struct nfs4_exception exception = {
7952                 .interruptible = true,
7953         };
7954         int status;
7955
7956         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7957                 (unsigned long long)server->fsid.major,
7958                 (unsigned long long)server->fsid.minor,
7959                 clp->cl_hostname);
7960         nfs_display_fhandle(NFS_FH(inode), __func__);
7961
7962         do {
7963                 status = ops->get_locations(inode, locations, page, cred);
7964                 if (status != -NFS4ERR_DELAY)
7965                         break;
7966                 nfs4_handle_exception(server, status, &exception);
7967         } while (exception.retry);
7968         return status;
7969 }
7970
7971 /*
7972  * This operation also signals the server that this client is
7973  * performing "lease moved" recovery.  The server can stop
7974  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7975  * is appended to this compound to identify the client ID which is
7976  * performing recovery.
7977  */
7978 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7979 {
7980         struct nfs_server *server = NFS_SERVER(inode);
7981         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7982         struct rpc_clnt *clnt = server->client;
7983         struct nfs4_fsid_present_arg args = {
7984                 .fh             = NFS_FH(inode),
7985                 .clientid       = clp->cl_clientid,
7986                 .renew          = 1,            /* append RENEW */
7987         };
7988         struct nfs4_fsid_present_res res = {
7989                 .renew          = 1,
7990         };
7991         struct rpc_message msg = {
7992                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7993                 .rpc_argp       = &args,
7994                 .rpc_resp       = &res,
7995                 .rpc_cred       = cred,
7996         };
7997         unsigned long now = jiffies;
7998         int status;
7999
8000         res.fh = nfs_alloc_fhandle();
8001         if (res.fh == NULL)
8002                 return -ENOMEM;
8003
8004         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8005         status = nfs4_call_sync_sequence(clnt, server, &msg,
8006                                                 &args.seq_args, &res.seq_res);
8007         nfs_free_fhandle(res.fh);
8008         if (status)
8009                 return status;
8010
8011         do_renew_lease(clp, now);
8012         return 0;
8013 }
8014
8015 #ifdef CONFIG_NFS_V4_1
8016
8017 /*
8018  * This operation also signals the server that this client is
8019  * performing "lease moved" recovery.  The server can stop asserting
8020  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8021  * this operation is identified in the SEQUENCE operation in this
8022  * compound.
8023  */
8024 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8025 {
8026         struct nfs_server *server = NFS_SERVER(inode);
8027         struct rpc_clnt *clnt = server->client;
8028         struct nfs4_fsid_present_arg args = {
8029                 .fh             = NFS_FH(inode),
8030         };
8031         struct nfs4_fsid_present_res res = {
8032         };
8033         struct rpc_message msg = {
8034                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8035                 .rpc_argp       = &args,
8036                 .rpc_resp       = &res,
8037                 .rpc_cred       = cred,
8038         };
8039         int status;
8040
8041         res.fh = nfs_alloc_fhandle();
8042         if (res.fh == NULL)
8043                 return -ENOMEM;
8044
8045         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8046         status = nfs4_call_sync_sequence(clnt, server, &msg,
8047                                                 &args.seq_args, &res.seq_res);
8048         nfs_free_fhandle(res.fh);
8049         if (status == NFS4_OK &&
8050             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8051                 status = -NFS4ERR_LEASE_MOVED;
8052         return status;
8053 }
8054
8055 #endif  /* CONFIG_NFS_V4_1 */
8056
8057 /**
8058  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8059  * @inode: inode on FSID to check
8060  * @cred: credential to use for this operation
8061  *
8062  * Server indicates whether the FSID is present, moved, or not
8063  * recognized.  This operation is necessary to clear a LEASE_MOVED
8064  * condition for this client ID.
8065  *
8066  * Returns NFS4_OK if the FSID is present on this server,
8067  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8068  *  NFS4ERR code if some error occurred on the server, or a
8069  *  negative errno if a local failure occurred.
8070  */
8071 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8072 {
8073         struct nfs_server *server = NFS_SERVER(inode);
8074         struct nfs_client *clp = server->nfs_client;
8075         const struct nfs4_mig_recovery_ops *ops =
8076                                         clp->cl_mvops->mig_recovery_ops;
8077         struct nfs4_exception exception = {
8078                 .interruptible = true,
8079         };
8080         int status;
8081
8082         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8083                 (unsigned long long)server->fsid.major,
8084                 (unsigned long long)server->fsid.minor,
8085                 clp->cl_hostname);
8086         nfs_display_fhandle(NFS_FH(inode), __func__);
8087
8088         do {
8089                 status = ops->fsid_present(inode, cred);
8090                 if (status != -NFS4ERR_DELAY)
8091                         break;
8092                 nfs4_handle_exception(server, status, &exception);
8093         } while (exception.retry);
8094         return status;
8095 }
8096
8097 /*
8098  * If 'use_integrity' is true and the state managment nfs_client
8099  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8100  * and the machine credential as per RFC3530bis and RFC5661 Security
8101  * Considerations sections. Otherwise, just use the user cred with the
8102  * filesystem's rpc_client.
8103  */
8104 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8105 {
8106         int status;
8107         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8108         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8109         struct nfs4_secinfo_arg args = {
8110                 .dir_fh = NFS_FH(dir),
8111                 .name   = name,
8112         };
8113         struct nfs4_secinfo_res res = {
8114                 .flavors     = flavors,
8115         };
8116         struct rpc_message msg = {
8117                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8118                 .rpc_argp = &args,
8119                 .rpc_resp = &res,
8120         };
8121         struct nfs4_call_sync_data data = {
8122                 .seq_server = NFS_SERVER(dir),
8123                 .seq_args = &args.seq_args,
8124                 .seq_res = &res.seq_res,
8125         };
8126         struct rpc_task_setup task_setup = {
8127                 .rpc_client = clnt,
8128                 .rpc_message = &msg,
8129                 .callback_ops = clp->cl_mvops->call_sync_ops,
8130                 .callback_data = &data,
8131                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8132         };
8133         const struct cred *cred = NULL;
8134
8135         if (use_integrity) {
8136                 clnt = clp->cl_rpcclient;
8137                 task_setup.rpc_client = clnt;
8138
8139                 cred = nfs4_get_clid_cred(clp);
8140                 msg.rpc_cred = cred;
8141         }
8142
8143         dprintk("NFS call  secinfo %s\n", name->name);
8144
8145         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8146         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8147         status = nfs4_call_sync_custom(&task_setup);
8148
8149         dprintk("NFS reply  secinfo: %d\n", status);
8150
8151         put_cred(cred);
8152         return status;
8153 }
8154
8155 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8156                       struct nfs4_secinfo_flavors *flavors)
8157 {
8158         struct nfs4_exception exception = {
8159                 .interruptible = true,
8160         };
8161         int err;
8162         do {
8163                 err = -NFS4ERR_WRONGSEC;
8164
8165                 /* try to use integrity protection with machine cred */
8166                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8167                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8168
8169                 /*
8170                  * if unable to use integrity protection, or SECINFO with
8171                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8172                  * disallowed by spec, but exists in deployed servers) use
8173                  * the current filesystem's rpc_client and the user cred.
8174                  */
8175                 if (err == -NFS4ERR_WRONGSEC)
8176                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8177
8178                 trace_nfs4_secinfo(dir, name, err);
8179                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8180                                 &exception);
8181         } while (exception.retry);
8182         return err;
8183 }
8184
8185 #ifdef CONFIG_NFS_V4_1
8186 /*
8187  * Check the exchange flags returned by the server for invalid flags, having
8188  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8189  * DS flags set.
8190  */
8191 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8192 {
8193         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8194                 goto out_inval;
8195         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8196                 goto out_inval;
8197         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8198             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8199                 goto out_inval;
8200         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8201                 goto out_inval;
8202         return NFS_OK;
8203 out_inval:
8204         return -NFS4ERR_INVAL;
8205 }
8206
8207 static bool
8208 nfs41_same_server_scope(struct nfs41_server_scope *a,
8209                         struct nfs41_server_scope *b)
8210 {
8211         if (a->server_scope_sz != b->server_scope_sz)
8212                 return false;
8213         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8214 }
8215
8216 static void
8217 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8218 {
8219         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8220         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8221         struct nfs_client *clp = args->client;
8222
8223         switch (task->tk_status) {
8224         case -NFS4ERR_BADSESSION:
8225         case -NFS4ERR_DEADSESSION:
8226                 nfs4_schedule_session_recovery(clp->cl_session,
8227                                 task->tk_status);
8228         }
8229         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8230                         res->dir != NFS4_CDFS4_BOTH) {
8231                 rpc_task_close_connection(task);
8232                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8233                         rpc_restart_call(task);
8234         }
8235 }
8236
8237 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8238         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8239 };
8240
8241 /*
8242  * nfs4_proc_bind_one_conn_to_session()
8243  *
8244  * The 4.1 client currently uses the same TCP connection for the
8245  * fore and backchannel.
8246  */
8247 static
8248 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8249                 struct rpc_xprt *xprt,
8250                 struct nfs_client *clp,
8251                 const struct cred *cred)
8252 {
8253         int status;
8254         struct nfs41_bind_conn_to_session_args args = {
8255                 .client = clp,
8256                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8257                 .retries = 0,
8258         };
8259         struct nfs41_bind_conn_to_session_res res;
8260         struct rpc_message msg = {
8261                 .rpc_proc =
8262                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8263                 .rpc_argp = &args,
8264                 .rpc_resp = &res,
8265                 .rpc_cred = cred,
8266         };
8267         struct rpc_task_setup task_setup_data = {
8268                 .rpc_client = clnt,
8269                 .rpc_xprt = xprt,
8270                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8271                 .rpc_message = &msg,
8272                 .flags = RPC_TASK_TIMEOUT,
8273         };
8274         struct rpc_task *task;
8275
8276         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8277         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8278                 args.dir = NFS4_CDFC4_FORE;
8279
8280         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8281         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8282                 args.dir = NFS4_CDFC4_FORE;
8283
8284         task = rpc_run_task(&task_setup_data);
8285         if (!IS_ERR(task)) {
8286                 status = task->tk_status;
8287                 rpc_put_task(task);
8288         } else
8289                 status = PTR_ERR(task);
8290         trace_nfs4_bind_conn_to_session(clp, status);
8291         if (status == 0) {
8292                 if (memcmp(res.sessionid.data,
8293                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8294                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8295                         return -EIO;
8296                 }
8297                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8298                         dprintk("NFS: %s: Unexpected direction from server\n",
8299                                 __func__);
8300                         return -EIO;
8301                 }
8302                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8303                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8304                                 __func__);
8305                         return -EIO;
8306                 }
8307         }
8308
8309         return status;
8310 }
8311
8312 struct rpc_bind_conn_calldata {
8313         struct nfs_client *clp;
8314         const struct cred *cred;
8315 };
8316
8317 static int
8318 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8319                 struct rpc_xprt *xprt,
8320                 void *calldata)
8321 {
8322         struct rpc_bind_conn_calldata *p = calldata;
8323
8324         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8325 }
8326
8327 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8328 {
8329         struct rpc_bind_conn_calldata data = {
8330                 .clp = clp,
8331                 .cred = cred,
8332         };
8333         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8334                         nfs4_proc_bind_conn_to_session_callback, &data);
8335 }
8336
8337 /*
8338  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8339  * and operations we'd like to see to enable certain features in the allow map
8340  */
8341 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8342         .how = SP4_MACH_CRED,
8343         .enforce.u.words = {
8344                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8345                       1 << (OP_EXCHANGE_ID - 32) |
8346                       1 << (OP_CREATE_SESSION - 32) |
8347                       1 << (OP_DESTROY_SESSION - 32) |
8348                       1 << (OP_DESTROY_CLIENTID - 32)
8349         },
8350         .allow.u.words = {
8351                 [0] = 1 << (OP_CLOSE) |
8352                       1 << (OP_OPEN_DOWNGRADE) |
8353                       1 << (OP_LOCKU) |
8354                       1 << (OP_DELEGRETURN) |
8355                       1 << (OP_COMMIT),
8356                 [1] = 1 << (OP_SECINFO - 32) |
8357                       1 << (OP_SECINFO_NO_NAME - 32) |
8358                       1 << (OP_LAYOUTRETURN - 32) |
8359                       1 << (OP_TEST_STATEID - 32) |
8360                       1 << (OP_FREE_STATEID - 32) |
8361                       1 << (OP_WRITE - 32)
8362         }
8363 };
8364
8365 /*
8366  * Select the state protection mode for client `clp' given the server results
8367  * from exchange_id in `sp'.
8368  *
8369  * Returns 0 on success, negative errno otherwise.
8370  */
8371 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8372                                  struct nfs41_state_protection *sp)
8373 {
8374         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8375                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8376                       1 << (OP_EXCHANGE_ID - 32) |
8377                       1 << (OP_CREATE_SESSION - 32) |
8378                       1 << (OP_DESTROY_SESSION - 32) |
8379                       1 << (OP_DESTROY_CLIENTID - 32)
8380         };
8381         unsigned long flags = 0;
8382         unsigned int i;
8383         int ret = 0;
8384
8385         if (sp->how == SP4_MACH_CRED) {
8386                 /* Print state protect result */
8387                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8388                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8389                         if (test_bit(i, sp->enforce.u.longs))
8390                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8391                         if (test_bit(i, sp->allow.u.longs))
8392                                 dfprintk(MOUNT, "  allow op %d\n", i);
8393                 }
8394
8395                 /* make sure nothing is on enforce list that isn't supported */
8396                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8397                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8398                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8399                                 ret = -EINVAL;
8400                                 goto out;
8401                         }
8402                 }
8403
8404                 /*
8405                  * Minimal mode - state operations are allowed to use machine
8406                  * credential.  Note this already happens by default, so the
8407                  * client doesn't have to do anything more than the negotiation.
8408                  *
8409                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8410                  *       we're already using the machine cred for exchange_id
8411                  *       and will never use a different cred.
8412                  */
8413                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8414                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8415                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8416                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8417                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8418                         dfprintk(MOUNT, "  minimal mode enabled\n");
8419                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8420                 } else {
8421                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8422                         ret = -EINVAL;
8423                         goto out;
8424                 }
8425
8426                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8427                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8428                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8429                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8430                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8431                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8432                 }
8433
8434                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8435                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8436                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8437                 }
8438
8439                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8440                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8441                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8442                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8443                 }
8444
8445                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8446                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8447                         dfprintk(MOUNT, "  stateid mode enabled\n");
8448                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8449                 }
8450
8451                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8452                         dfprintk(MOUNT, "  write mode enabled\n");
8453                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8454                 }
8455
8456                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8457                         dfprintk(MOUNT, "  commit mode enabled\n");
8458                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8459                 }
8460         }
8461 out:
8462         clp->cl_sp4_flags = flags;
8463         return ret;
8464 }
8465
8466 struct nfs41_exchange_id_data {
8467         struct nfs41_exchange_id_res res;
8468         struct nfs41_exchange_id_args args;
8469 };
8470
8471 static void nfs4_exchange_id_release(void *data)
8472 {
8473         struct nfs41_exchange_id_data *cdata =
8474                                         (struct nfs41_exchange_id_data *)data;
8475
8476         nfs_put_client(cdata->args.client);
8477         kfree(cdata->res.impl_id);
8478         kfree(cdata->res.server_scope);
8479         kfree(cdata->res.server_owner);
8480         kfree(cdata);
8481 }
8482
8483 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8484         .rpc_release = nfs4_exchange_id_release,
8485 };
8486
8487 /*
8488  * _nfs4_proc_exchange_id()
8489  *
8490  * Wrapper for EXCHANGE_ID operation.
8491  */
8492 static struct rpc_task *
8493 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8494                         u32 sp4_how, struct rpc_xprt *xprt)
8495 {
8496         struct rpc_message msg = {
8497                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8498                 .rpc_cred = cred,
8499         };
8500         struct rpc_task_setup task_setup_data = {
8501                 .rpc_client = clp->cl_rpcclient,
8502                 .callback_ops = &nfs4_exchange_id_call_ops,
8503                 .rpc_message = &msg,
8504                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8505         };
8506         struct nfs41_exchange_id_data *calldata;
8507         int status;
8508
8509         if (!refcount_inc_not_zero(&clp->cl_count))
8510                 return ERR_PTR(-EIO);
8511
8512         status = -ENOMEM;
8513         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8514         if (!calldata)
8515                 goto out;
8516
8517         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8518
8519         status = nfs4_init_uniform_client_string(clp);
8520         if (status)
8521                 goto out_calldata;
8522
8523         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8524                                                 GFP_NOFS);
8525         status = -ENOMEM;
8526         if (unlikely(calldata->res.server_owner == NULL))
8527                 goto out_calldata;
8528
8529         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8530                                         GFP_NOFS);
8531         if (unlikely(calldata->res.server_scope == NULL))
8532                 goto out_server_owner;
8533
8534         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8535         if (unlikely(calldata->res.impl_id == NULL))
8536                 goto out_server_scope;
8537
8538         switch (sp4_how) {
8539         case SP4_NONE:
8540                 calldata->args.state_protect.how = SP4_NONE;
8541                 break;
8542
8543         case SP4_MACH_CRED:
8544                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8545                 break;
8546
8547         default:
8548                 /* unsupported! */
8549                 WARN_ON_ONCE(1);
8550                 status = -EINVAL;
8551                 goto out_impl_id;
8552         }
8553         if (xprt) {
8554                 task_setup_data.rpc_xprt = xprt;
8555                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8556                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8557                                 sizeof(calldata->args.verifier.data));
8558         }
8559         calldata->args.client = clp;
8560         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8561         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8562 #ifdef CONFIG_NFS_V4_1_MIGRATION
8563         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8564 #endif
8565         msg.rpc_argp = &calldata->args;
8566         msg.rpc_resp = &calldata->res;
8567         task_setup_data.callback_data = calldata;
8568
8569         return rpc_run_task(&task_setup_data);
8570
8571 out_impl_id:
8572         kfree(calldata->res.impl_id);
8573 out_server_scope:
8574         kfree(calldata->res.server_scope);
8575 out_server_owner:
8576         kfree(calldata->res.server_owner);
8577 out_calldata:
8578         kfree(calldata);
8579 out:
8580         nfs_put_client(clp);
8581         return ERR_PTR(status);
8582 }
8583
8584 /*
8585  * _nfs4_proc_exchange_id()
8586  *
8587  * Wrapper for EXCHANGE_ID operation.
8588  */
8589 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8590                         u32 sp4_how)
8591 {
8592         struct rpc_task *task;
8593         struct nfs41_exchange_id_args *argp;
8594         struct nfs41_exchange_id_res *resp;
8595         unsigned long now = jiffies;
8596         int status;
8597
8598         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8599         if (IS_ERR(task))
8600                 return PTR_ERR(task);
8601
8602         argp = task->tk_msg.rpc_argp;
8603         resp = task->tk_msg.rpc_resp;
8604         status = task->tk_status;
8605         if (status  != 0)
8606                 goto out;
8607
8608         status = nfs4_check_cl_exchange_flags(resp->flags,
8609                         clp->cl_mvops->minor_version);
8610         if (status  != 0)
8611                 goto out;
8612
8613         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8614         if (status != 0)
8615                 goto out;
8616
8617         do_renew_lease(clp, now);
8618
8619         clp->cl_clientid = resp->clientid;
8620         clp->cl_exchange_flags = resp->flags;
8621         clp->cl_seqid = resp->seqid;
8622         /* Client ID is not confirmed */
8623         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8624                 clear_bit(NFS4_SESSION_ESTABLISHED,
8625                           &clp->cl_session->session_state);
8626
8627         if (clp->cl_serverscope != NULL &&
8628             !nfs41_same_server_scope(clp->cl_serverscope,
8629                                 resp->server_scope)) {
8630                 dprintk("%s: server_scope mismatch detected\n",
8631                         __func__);
8632                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8633         }
8634
8635         swap(clp->cl_serverowner, resp->server_owner);
8636         swap(clp->cl_serverscope, resp->server_scope);
8637         swap(clp->cl_implid, resp->impl_id);
8638
8639         /* Save the EXCHANGE_ID verifier session trunk tests */
8640         memcpy(clp->cl_confirm.data, argp->verifier.data,
8641                sizeof(clp->cl_confirm.data));
8642 out:
8643         trace_nfs4_exchange_id(clp, status);
8644         rpc_put_task(task);
8645         return status;
8646 }
8647
8648 /*
8649  * nfs4_proc_exchange_id()
8650  *
8651  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8652  *
8653  * Since the clientid has expired, all compounds using sessions
8654  * associated with the stale clientid will be returning
8655  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8656  * be in some phase of session reset.
8657  *
8658  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8659  */
8660 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8661 {
8662         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8663         int status;
8664
8665         /* try SP4_MACH_CRED if krb5i/p */
8666         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8667             authflavor == RPC_AUTH_GSS_KRB5P) {
8668                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8669                 if (!status)
8670                         return 0;
8671         }
8672
8673         /* try SP4_NONE */
8674         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8675 }
8676
8677 /**
8678  * nfs4_test_session_trunk
8679  *
8680  * This is an add_xprt_test() test function called from
8681  * rpc_clnt_setup_test_and_add_xprt.
8682  *
8683  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8684  * and is dereferrenced in nfs4_exchange_id_release
8685  *
8686  * Upon success, add the new transport to the rpc_clnt
8687  *
8688  * @clnt: struct rpc_clnt to get new transport
8689  * @xprt: the rpc_xprt to test
8690  * @data: call data for _nfs4_proc_exchange_id.
8691  */
8692 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8693                             void *data)
8694 {
8695         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8696         struct rpc_task *task;
8697         int status;
8698
8699         u32 sp4_how;
8700
8701         dprintk("--> %s try %s\n", __func__,
8702                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8703
8704         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8705
8706         /* Test connection for session trunking. Async exchange_id call */
8707         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8708         if (IS_ERR(task))
8709                 return;
8710
8711         status = task->tk_status;
8712         if (status == 0)
8713                 status = nfs4_detect_session_trunking(adata->clp,
8714                                 task->tk_msg.rpc_resp, xprt);
8715
8716         if (status == 0)
8717                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8718
8719         rpc_put_task(task);
8720 }
8721 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8722
8723 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8724                 const struct cred *cred)
8725 {
8726         struct rpc_message msg = {
8727                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8728                 .rpc_argp = clp,
8729                 .rpc_cred = cred,
8730         };
8731         int status;
8732
8733         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8734                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8735         trace_nfs4_destroy_clientid(clp, status);
8736         if (status)
8737                 dprintk("NFS: Got error %d from the server %s on "
8738                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8739         return status;
8740 }
8741
8742 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8743                 const struct cred *cred)
8744 {
8745         unsigned int loop;
8746         int ret;
8747
8748         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8749                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8750                 switch (ret) {
8751                 case -NFS4ERR_DELAY:
8752                 case -NFS4ERR_CLIENTID_BUSY:
8753                         ssleep(1);
8754                         break;
8755                 default:
8756                         return ret;
8757                 }
8758         }
8759         return 0;
8760 }
8761
8762 int nfs4_destroy_clientid(struct nfs_client *clp)
8763 {
8764         const struct cred *cred;
8765         int ret = 0;
8766
8767         if (clp->cl_mvops->minor_version < 1)
8768                 goto out;
8769         if (clp->cl_exchange_flags == 0)
8770                 goto out;
8771         if (clp->cl_preserve_clid)
8772                 goto out;
8773         cred = nfs4_get_clid_cred(clp);
8774         ret = nfs4_proc_destroy_clientid(clp, cred);
8775         put_cred(cred);
8776         switch (ret) {
8777         case 0:
8778         case -NFS4ERR_STALE_CLIENTID:
8779                 clp->cl_exchange_flags = 0;
8780         }
8781 out:
8782         return ret;
8783 }
8784
8785 #endif /* CONFIG_NFS_V4_1 */
8786
8787 struct nfs4_get_lease_time_data {
8788         struct nfs4_get_lease_time_args *args;
8789         struct nfs4_get_lease_time_res *res;
8790         struct nfs_client *clp;
8791 };
8792
8793 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8794                                         void *calldata)
8795 {
8796         struct nfs4_get_lease_time_data *data =
8797                         (struct nfs4_get_lease_time_data *)calldata;
8798
8799         dprintk("--> %s\n", __func__);
8800         /* just setup sequence, do not trigger session recovery
8801            since we're invoked within one */
8802         nfs4_setup_sequence(data->clp,
8803                         &data->args->la_seq_args,
8804                         &data->res->lr_seq_res,
8805                         task);
8806         dprintk("<-- %s\n", __func__);
8807 }
8808
8809 /*
8810  * Called from nfs4_state_manager thread for session setup, so don't recover
8811  * from sequence operation or clientid errors.
8812  */
8813 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8814 {
8815         struct nfs4_get_lease_time_data *data =
8816                         (struct nfs4_get_lease_time_data *)calldata;
8817
8818         dprintk("--> %s\n", __func__);
8819         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8820                 return;
8821         switch (task->tk_status) {
8822         case -NFS4ERR_DELAY:
8823         case -NFS4ERR_GRACE:
8824                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8825                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8826                 task->tk_status = 0;
8827                 fallthrough;
8828         case -NFS4ERR_RETRY_UNCACHED_REP:
8829                 rpc_restart_call_prepare(task);
8830                 return;
8831         }
8832         dprintk("<-- %s\n", __func__);
8833 }
8834
8835 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8836         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8837         .rpc_call_done = nfs4_get_lease_time_done,
8838 };
8839
8840 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8841 {
8842         struct nfs4_get_lease_time_args args;
8843         struct nfs4_get_lease_time_res res = {
8844                 .lr_fsinfo = fsinfo,
8845         };
8846         struct nfs4_get_lease_time_data data = {
8847                 .args = &args,
8848                 .res = &res,
8849                 .clp = clp,
8850         };
8851         struct rpc_message msg = {
8852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8853                 .rpc_argp = &args,
8854                 .rpc_resp = &res,
8855         };
8856         struct rpc_task_setup task_setup = {
8857                 .rpc_client = clp->cl_rpcclient,
8858                 .rpc_message = &msg,
8859                 .callback_ops = &nfs4_get_lease_time_ops,
8860                 .callback_data = &data,
8861                 .flags = RPC_TASK_TIMEOUT,
8862         };
8863
8864         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8865         return nfs4_call_sync_custom(&task_setup);
8866 }
8867
8868 #ifdef CONFIG_NFS_V4_1
8869
8870 /*
8871  * Initialize the values to be used by the client in CREATE_SESSION
8872  * If nfs4_init_session set the fore channel request and response sizes,
8873  * use them.
8874  *
8875  * Set the back channel max_resp_sz_cached to zero to force the client to
8876  * always set csa_cachethis to FALSE because the current implementation
8877  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8878  */
8879 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8880                                     struct rpc_clnt *clnt)
8881 {
8882         unsigned int max_rqst_sz, max_resp_sz;
8883         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8884         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8885
8886         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8887         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8888
8889         /* Fore channel attributes */
8890         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8891         args->fc_attrs.max_resp_sz = max_resp_sz;
8892         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8893         args->fc_attrs.max_reqs = max_session_slots;
8894
8895         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8896                 "max_ops=%u max_reqs=%u\n",
8897                 __func__,
8898                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8899                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8900
8901         /* Back channel attributes */
8902         args->bc_attrs.max_rqst_sz = max_bc_payload;
8903         args->bc_attrs.max_resp_sz = max_bc_payload;
8904         args->bc_attrs.max_resp_sz_cached = 0;
8905         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8906         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8907         if (args->bc_attrs.max_reqs > max_bc_slots)
8908                 args->bc_attrs.max_reqs = max_bc_slots;
8909
8910         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8911                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8912                 __func__,
8913                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8914                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8915                 args->bc_attrs.max_reqs);
8916 }
8917
8918 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8919                 struct nfs41_create_session_res *res)
8920 {
8921         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8922         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8923
8924         if (rcvd->max_resp_sz > sent->max_resp_sz)
8925                 return -EINVAL;
8926         /*
8927          * Our requested max_ops is the minimum we need; we're not
8928          * prepared to break up compounds into smaller pieces than that.
8929          * So, no point even trying to continue if the server won't
8930          * cooperate:
8931          */
8932         if (rcvd->max_ops < sent->max_ops)
8933                 return -EINVAL;
8934         if (rcvd->max_reqs == 0)
8935                 return -EINVAL;
8936         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8937                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8938         return 0;
8939 }
8940
8941 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8942                 struct nfs41_create_session_res *res)
8943 {
8944         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8945         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8946
8947         if (!(res->flags & SESSION4_BACK_CHAN))
8948                 goto out;
8949         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8950                 return -EINVAL;
8951         if (rcvd->max_resp_sz < sent->max_resp_sz)
8952                 return -EINVAL;
8953         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8954                 return -EINVAL;
8955         if (rcvd->max_ops > sent->max_ops)
8956                 return -EINVAL;
8957         if (rcvd->max_reqs > sent->max_reqs)
8958                 return -EINVAL;
8959 out:
8960         return 0;
8961 }
8962
8963 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8964                                      struct nfs41_create_session_res *res)
8965 {
8966         int ret;
8967
8968         ret = nfs4_verify_fore_channel_attrs(args, res);
8969         if (ret)
8970                 return ret;
8971         return nfs4_verify_back_channel_attrs(args, res);
8972 }
8973
8974 static void nfs4_update_session(struct nfs4_session *session,
8975                 struct nfs41_create_session_res *res)
8976 {
8977         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8978         /* Mark client id and session as being confirmed */
8979         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8980         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8981         session->flags = res->flags;
8982         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8983         if (res->flags & SESSION4_BACK_CHAN)
8984                 memcpy(&session->bc_attrs, &res->bc_attrs,
8985                                 sizeof(session->bc_attrs));
8986 }
8987
8988 static int _nfs4_proc_create_session(struct nfs_client *clp,
8989                 const struct cred *cred)
8990 {
8991         struct nfs4_session *session = clp->cl_session;
8992         struct nfs41_create_session_args args = {
8993                 .client = clp,
8994                 .clientid = clp->cl_clientid,
8995                 .seqid = clp->cl_seqid,
8996                 .cb_program = NFS4_CALLBACK,
8997         };
8998         struct nfs41_create_session_res res;
8999
9000         struct rpc_message msg = {
9001                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9002                 .rpc_argp = &args,
9003                 .rpc_resp = &res,
9004                 .rpc_cred = cred,
9005         };
9006         int status;
9007
9008         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9009         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9010
9011         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9012                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9013         trace_nfs4_create_session(clp, status);
9014
9015         switch (status) {
9016         case -NFS4ERR_STALE_CLIENTID:
9017         case -NFS4ERR_DELAY:
9018         case -ETIMEDOUT:
9019         case -EACCES:
9020         case -EAGAIN:
9021                 goto out;
9022         }
9023
9024         clp->cl_seqid++;
9025         if (!status) {
9026                 /* Verify the session's negotiated channel_attrs values */
9027                 status = nfs4_verify_channel_attrs(&args, &res);
9028                 /* Increment the clientid slot sequence id */
9029                 if (status)
9030                         goto out;
9031                 nfs4_update_session(session, &res);
9032         }
9033 out:
9034         return status;
9035 }
9036
9037 /*
9038  * Issues a CREATE_SESSION operation to the server.
9039  * It is the responsibility of the caller to verify the session is
9040  * expired before calling this routine.
9041  */
9042 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9043 {
9044         int status;
9045         unsigned *ptr;
9046         struct nfs4_session *session = clp->cl_session;
9047
9048         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9049
9050         status = _nfs4_proc_create_session(clp, cred);
9051         if (status)
9052                 goto out;
9053
9054         /* Init or reset the session slot tables */
9055         status = nfs4_setup_session_slot_tables(session);
9056         dprintk("slot table setup returned %d\n", status);
9057         if (status)
9058                 goto out;
9059
9060         ptr = (unsigned *)&session->sess_id.data[0];
9061         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9062                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9063 out:
9064         dprintk("<-- %s\n", __func__);
9065         return status;
9066 }
9067
9068 /*
9069  * Issue the over-the-wire RPC DESTROY_SESSION.
9070  * The caller must serialize access to this routine.
9071  */
9072 int nfs4_proc_destroy_session(struct nfs4_session *session,
9073                 const struct cred *cred)
9074 {
9075         struct rpc_message msg = {
9076                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9077                 .rpc_argp = session,
9078                 .rpc_cred = cred,
9079         };
9080         int status = 0;
9081
9082         dprintk("--> nfs4_proc_destroy_session\n");
9083
9084         /* session is still being setup */
9085         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9086                 return 0;
9087
9088         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9089                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9090         trace_nfs4_destroy_session(session->clp, status);
9091
9092         if (status)
9093                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9094                         "Session has been destroyed regardless...\n", status);
9095
9096         dprintk("<-- nfs4_proc_destroy_session\n");
9097         return status;
9098 }
9099
9100 /*
9101  * Renew the cl_session lease.
9102  */
9103 struct nfs4_sequence_data {
9104         struct nfs_client *clp;
9105         struct nfs4_sequence_args args;
9106         struct nfs4_sequence_res res;
9107 };
9108
9109 static void nfs41_sequence_release(void *data)
9110 {
9111         struct nfs4_sequence_data *calldata = data;
9112         struct nfs_client *clp = calldata->clp;
9113
9114         if (refcount_read(&clp->cl_count) > 1)
9115                 nfs4_schedule_state_renewal(clp);
9116         nfs_put_client(clp);
9117         kfree(calldata);
9118 }
9119
9120 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9121 {
9122         switch(task->tk_status) {
9123         case -NFS4ERR_DELAY:
9124                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9125                 return -EAGAIN;
9126         default:
9127                 nfs4_schedule_lease_recovery(clp);
9128         }
9129         return 0;
9130 }
9131
9132 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9133 {
9134         struct nfs4_sequence_data *calldata = data;
9135         struct nfs_client *clp = calldata->clp;
9136
9137         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9138                 return;
9139
9140         trace_nfs4_sequence(clp, task->tk_status);
9141         if (task->tk_status < 0) {
9142                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9143                 if (refcount_read(&clp->cl_count) == 1)
9144                         goto out;
9145
9146                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9147                         rpc_restart_call_prepare(task);
9148                         return;
9149                 }
9150         }
9151         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9152 out:
9153         dprintk("<-- %s\n", __func__);
9154 }
9155
9156 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9157 {
9158         struct nfs4_sequence_data *calldata = data;
9159         struct nfs_client *clp = calldata->clp;
9160         struct nfs4_sequence_args *args;
9161         struct nfs4_sequence_res *res;
9162
9163         args = task->tk_msg.rpc_argp;
9164         res = task->tk_msg.rpc_resp;
9165
9166         nfs4_setup_sequence(clp, args, res, task);
9167 }
9168
9169 static const struct rpc_call_ops nfs41_sequence_ops = {
9170         .rpc_call_done = nfs41_sequence_call_done,
9171         .rpc_call_prepare = nfs41_sequence_prepare,
9172         .rpc_release = nfs41_sequence_release,
9173 };
9174
9175 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9176                 const struct cred *cred,
9177                 struct nfs4_slot *slot,
9178                 bool is_privileged)
9179 {
9180         struct nfs4_sequence_data *calldata;
9181         struct rpc_message msg = {
9182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9183                 .rpc_cred = cred,
9184         };
9185         struct rpc_task_setup task_setup_data = {
9186                 .rpc_client = clp->cl_rpcclient,
9187                 .rpc_message = &msg,
9188                 .callback_ops = &nfs41_sequence_ops,
9189                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9190         };
9191         struct rpc_task *ret;
9192
9193         ret = ERR_PTR(-EIO);
9194         if (!refcount_inc_not_zero(&clp->cl_count))
9195                 goto out_err;
9196
9197         ret = ERR_PTR(-ENOMEM);
9198         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9199         if (calldata == NULL)
9200                 goto out_put_clp;
9201         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9202         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9203         msg.rpc_argp = &calldata->args;
9204         msg.rpc_resp = &calldata->res;
9205         calldata->clp = clp;
9206         task_setup_data.callback_data = calldata;
9207
9208         ret = rpc_run_task(&task_setup_data);
9209         if (IS_ERR(ret))
9210                 goto out_err;
9211         return ret;
9212 out_put_clp:
9213         nfs_put_client(clp);
9214 out_err:
9215         nfs41_release_slot(slot);
9216         return ret;
9217 }
9218
9219 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9220 {
9221         struct rpc_task *task;
9222         int ret = 0;
9223
9224         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9225                 return -EAGAIN;
9226         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9227         if (IS_ERR(task))
9228                 ret = PTR_ERR(task);
9229         else
9230                 rpc_put_task_async(task);
9231         dprintk("<-- %s status=%d\n", __func__, ret);
9232         return ret;
9233 }
9234
9235 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9236 {
9237         struct rpc_task *task;
9238         int ret;
9239
9240         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9241         if (IS_ERR(task)) {
9242                 ret = PTR_ERR(task);
9243                 goto out;
9244         }
9245         ret = rpc_wait_for_completion_task(task);
9246         if (!ret)
9247                 ret = task->tk_status;
9248         rpc_put_task(task);
9249 out:
9250         dprintk("<-- %s status=%d\n", __func__, ret);
9251         return ret;
9252 }
9253
9254 struct nfs4_reclaim_complete_data {
9255         struct nfs_client *clp;
9256         struct nfs41_reclaim_complete_args arg;
9257         struct nfs41_reclaim_complete_res res;
9258 };
9259
9260 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9261 {
9262         struct nfs4_reclaim_complete_data *calldata = data;
9263
9264         nfs4_setup_sequence(calldata->clp,
9265                         &calldata->arg.seq_args,
9266                         &calldata->res.seq_res,
9267                         task);
9268 }
9269
9270 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9271 {
9272         switch(task->tk_status) {
9273         case 0:
9274                 wake_up_all(&clp->cl_lock_waitq);
9275                 fallthrough;
9276         case -NFS4ERR_COMPLETE_ALREADY:
9277         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9278                 break;
9279         case -NFS4ERR_DELAY:
9280                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9281                 fallthrough;
9282         case -NFS4ERR_RETRY_UNCACHED_REP:
9283                 return -EAGAIN;
9284         case -NFS4ERR_BADSESSION:
9285         case -NFS4ERR_DEADSESSION:
9286         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9287                 break;
9288         default:
9289                 nfs4_schedule_lease_recovery(clp);
9290         }
9291         return 0;
9292 }
9293
9294 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9295 {
9296         struct nfs4_reclaim_complete_data *calldata = data;
9297         struct nfs_client *clp = calldata->clp;
9298         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9299
9300         dprintk("--> %s\n", __func__);
9301         if (!nfs41_sequence_done(task, res))
9302                 return;
9303
9304         trace_nfs4_reclaim_complete(clp, task->tk_status);
9305         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9306                 rpc_restart_call_prepare(task);
9307                 return;
9308         }
9309         dprintk("<-- %s\n", __func__);
9310 }
9311
9312 static void nfs4_free_reclaim_complete_data(void *data)
9313 {
9314         struct nfs4_reclaim_complete_data *calldata = data;
9315
9316         kfree(calldata);
9317 }
9318
9319 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9320         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9321         .rpc_call_done = nfs4_reclaim_complete_done,
9322         .rpc_release = nfs4_free_reclaim_complete_data,
9323 };
9324
9325 /*
9326  * Issue a global reclaim complete.
9327  */
9328 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9329                 const struct cred *cred)
9330 {
9331         struct nfs4_reclaim_complete_data *calldata;
9332         struct rpc_message msg = {
9333                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9334                 .rpc_cred = cred,
9335         };
9336         struct rpc_task_setup task_setup_data = {
9337                 .rpc_client = clp->cl_rpcclient,
9338                 .rpc_message = &msg,
9339                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9340                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9341         };
9342         int status = -ENOMEM;
9343
9344         dprintk("--> %s\n", __func__);
9345         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9346         if (calldata == NULL)
9347                 goto out;
9348         calldata->clp = clp;
9349         calldata->arg.one_fs = 0;
9350
9351         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9352         msg.rpc_argp = &calldata->arg;
9353         msg.rpc_resp = &calldata->res;
9354         task_setup_data.callback_data = calldata;
9355         status = nfs4_call_sync_custom(&task_setup_data);
9356 out:
9357         dprintk("<-- %s status=%d\n", __func__, status);
9358         return status;
9359 }
9360
9361 static void
9362 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9363 {
9364         struct nfs4_layoutget *lgp = calldata;
9365         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9366
9367         dprintk("--> %s\n", __func__);
9368         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9369                                 &lgp->res.seq_res, task);
9370         dprintk("<-- %s\n", __func__);
9371 }
9372
9373 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9374 {
9375         struct nfs4_layoutget *lgp = calldata;
9376
9377         dprintk("--> %s\n", __func__);
9378         nfs41_sequence_process(task, &lgp->res.seq_res);
9379         dprintk("<-- %s\n", __func__);
9380 }
9381
9382 static int
9383 nfs4_layoutget_handle_exception(struct rpc_task *task,
9384                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9385 {
9386         struct inode *inode = lgp->args.inode;
9387         struct nfs_server *server = NFS_SERVER(inode);
9388         struct pnfs_layout_hdr *lo;
9389         int nfs4err = task->tk_status;
9390         int err, status = 0;
9391         LIST_HEAD(head);
9392
9393         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9394
9395         nfs4_sequence_free_slot(&lgp->res.seq_res);
9396
9397         switch (nfs4err) {
9398         case 0:
9399                 goto out;
9400
9401         /*
9402          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9403          * on the file. set tk_status to -ENODATA to tell upper layer to
9404          * retry go inband.
9405          */
9406         case -NFS4ERR_LAYOUTUNAVAILABLE:
9407                 status = -ENODATA;
9408                 goto out;
9409         /*
9410          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9411          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9412          */
9413         case -NFS4ERR_BADLAYOUT:
9414                 status = -EOVERFLOW;
9415                 goto out;
9416         /*
9417          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9418          * (or clients) writing to the same RAID stripe except when
9419          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9420          *
9421          * Treat it like we would RECALLCONFLICT -- we retry for a little
9422          * while, and then eventually give up.
9423          */
9424         case -NFS4ERR_LAYOUTTRYLATER:
9425                 if (lgp->args.minlength == 0) {
9426                         status = -EOVERFLOW;
9427                         goto out;
9428                 }
9429                 status = -EBUSY;
9430                 break;
9431         case -NFS4ERR_RECALLCONFLICT:
9432                 status = -ERECALLCONFLICT;
9433                 break;
9434         case -NFS4ERR_DELEG_REVOKED:
9435         case -NFS4ERR_ADMIN_REVOKED:
9436         case -NFS4ERR_EXPIRED:
9437         case -NFS4ERR_BAD_STATEID:
9438                 exception->timeout = 0;
9439                 spin_lock(&inode->i_lock);
9440                 lo = NFS_I(inode)->layout;
9441                 /* If the open stateid was bad, then recover it. */
9442                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9443                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9444                         spin_unlock(&inode->i_lock);
9445                         exception->state = lgp->args.ctx->state;
9446                         exception->stateid = &lgp->args.stateid;
9447                         break;
9448                 }
9449
9450                 /*
9451                  * Mark the bad layout state as invalid, then retry
9452                  */
9453                 pnfs_mark_layout_stateid_invalid(lo, &head);
9454                 spin_unlock(&inode->i_lock);
9455                 nfs_commit_inode(inode, 0);
9456                 pnfs_free_lseg_list(&head);
9457                 status = -EAGAIN;
9458                 goto out;
9459         }
9460
9461         err = nfs4_handle_exception(server, nfs4err, exception);
9462         if (!status) {
9463                 if (exception->retry)
9464                         status = -EAGAIN;
9465                 else
9466                         status = err;
9467         }
9468 out:
9469         dprintk("<-- %s\n", __func__);
9470         return status;
9471 }
9472
9473 size_t max_response_pages(struct nfs_server *server)
9474 {
9475         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9476         return nfs_page_array_len(0, max_resp_sz);
9477 }
9478
9479 static void nfs4_layoutget_release(void *calldata)
9480 {
9481         struct nfs4_layoutget *lgp = calldata;
9482
9483         dprintk("--> %s\n", __func__);
9484         nfs4_sequence_free_slot(&lgp->res.seq_res);
9485         pnfs_layoutget_free(lgp);
9486         dprintk("<-- %s\n", __func__);
9487 }
9488
9489 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9490         .rpc_call_prepare = nfs4_layoutget_prepare,
9491         .rpc_call_done = nfs4_layoutget_done,
9492         .rpc_release = nfs4_layoutget_release,
9493 };
9494
9495 struct pnfs_layout_segment *
9496 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9497 {
9498         struct inode *inode = lgp->args.inode;
9499         struct nfs_server *server = NFS_SERVER(inode);
9500         struct rpc_task *task;
9501         struct rpc_message msg = {
9502                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9503                 .rpc_argp = &lgp->args,
9504                 .rpc_resp = &lgp->res,
9505                 .rpc_cred = lgp->cred,
9506         };
9507         struct rpc_task_setup task_setup_data = {
9508                 .rpc_client = server->client,
9509                 .rpc_message = &msg,
9510                 .callback_ops = &nfs4_layoutget_call_ops,
9511                 .callback_data = lgp,
9512                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9513         };
9514         struct pnfs_layout_segment *lseg = NULL;
9515         struct nfs4_exception exception = {
9516                 .inode = inode,
9517                 .timeout = *timeout,
9518         };
9519         int status = 0;
9520
9521         dprintk("--> %s\n", __func__);
9522
9523         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9524         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9525
9526         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9527
9528         task = rpc_run_task(&task_setup_data);
9529
9530         status = rpc_wait_for_completion_task(task);
9531         if (status != 0)
9532                 goto out;
9533
9534         if (task->tk_status < 0) {
9535                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9536                 *timeout = exception.timeout;
9537         } else if (lgp->res.layoutp->len == 0) {
9538                 status = -EAGAIN;
9539                 *timeout = nfs4_update_delay(&exception.timeout);
9540         } else
9541                 lseg = pnfs_layout_process(lgp);
9542 out:
9543         trace_nfs4_layoutget(lgp->args.ctx,
9544                         &lgp->args.range,
9545                         &lgp->res.range,
9546                         &lgp->res.stateid,
9547                         status);
9548
9549         rpc_put_task(task);
9550         dprintk("<-- %s status=%d\n", __func__, status);
9551         if (status)
9552                 return ERR_PTR(status);
9553         return lseg;
9554 }
9555
9556 static void
9557 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9558 {
9559         struct nfs4_layoutreturn *lrp = calldata;
9560
9561         dprintk("--> %s\n", __func__);
9562         nfs4_setup_sequence(lrp->clp,
9563                         &lrp->args.seq_args,
9564                         &lrp->res.seq_res,
9565                         task);
9566         if (!pnfs_layout_is_valid(lrp->args.layout))
9567                 rpc_exit(task, 0);
9568 }
9569
9570 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9571 {
9572         struct nfs4_layoutreturn *lrp = calldata;
9573         struct nfs_server *server;
9574
9575         dprintk("--> %s\n", __func__);
9576
9577         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9578                 return;
9579
9580         /*
9581          * Was there an RPC level error? Assume the call succeeded,
9582          * and that we need to release the layout
9583          */
9584         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9585                 lrp->res.lrs_present = 0;
9586                 return;
9587         }
9588
9589         server = NFS_SERVER(lrp->args.inode);
9590         switch (task->tk_status) {
9591         case -NFS4ERR_OLD_STATEID:
9592                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9593                                         &lrp->args.range,
9594                                         lrp->args.inode))
9595                         goto out_restart;
9596                 fallthrough;
9597         default:
9598                 task->tk_status = 0;
9599                 fallthrough;
9600         case 0:
9601                 break;
9602         case -NFS4ERR_DELAY:
9603                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9604                         break;
9605                 goto out_restart;
9606         }
9607         dprintk("<-- %s\n", __func__);
9608         return;
9609 out_restart:
9610         task->tk_status = 0;
9611         nfs4_sequence_free_slot(&lrp->res.seq_res);
9612         rpc_restart_call_prepare(task);
9613 }
9614
9615 static void nfs4_layoutreturn_release(void *calldata)
9616 {
9617         struct nfs4_layoutreturn *lrp = calldata;
9618         struct pnfs_layout_hdr *lo = lrp->args.layout;
9619
9620         dprintk("--> %s\n", __func__);
9621         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9622                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9623         nfs4_sequence_free_slot(&lrp->res.seq_res);
9624         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9625                 lrp->ld_private.ops->free(&lrp->ld_private);
9626         pnfs_put_layout_hdr(lrp->args.layout);
9627         nfs_iput_and_deactive(lrp->inode);
9628         put_cred(lrp->cred);
9629         kfree(calldata);
9630         dprintk("<-- %s\n", __func__);
9631 }
9632
9633 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9634         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9635         .rpc_call_done = nfs4_layoutreturn_done,
9636         .rpc_release = nfs4_layoutreturn_release,
9637 };
9638
9639 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9640 {
9641         struct rpc_task *task;
9642         struct rpc_message msg = {
9643                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9644                 .rpc_argp = &lrp->args,
9645                 .rpc_resp = &lrp->res,
9646                 .rpc_cred = lrp->cred,
9647         };
9648         struct rpc_task_setup task_setup_data = {
9649                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9650                 .rpc_message = &msg,
9651                 .callback_ops = &nfs4_layoutreturn_call_ops,
9652                 .callback_data = lrp,
9653         };
9654         int status = 0;
9655
9656         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9657                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9658                         &task_setup_data.rpc_client, &msg);
9659
9660         dprintk("--> %s\n", __func__);
9661         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9662         if (!sync) {
9663                 if (!lrp->inode) {
9664                         nfs4_layoutreturn_release(lrp);
9665                         return -EAGAIN;
9666                 }
9667                 task_setup_data.flags |= RPC_TASK_ASYNC;
9668         }
9669         if (!lrp->inode)
9670                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9671                                    1);
9672         else
9673                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9674                                    0);
9675         task = rpc_run_task(&task_setup_data);
9676         if (IS_ERR(task))
9677                 return PTR_ERR(task);
9678         if (sync)
9679                 status = task->tk_status;
9680         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9681         dprintk("<-- %s status=%d\n", __func__, status);
9682         rpc_put_task(task);
9683         return status;
9684 }
9685
9686 static int
9687 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9688                 struct pnfs_device *pdev,
9689                 const struct cred *cred)
9690 {
9691         struct nfs4_getdeviceinfo_args args = {
9692                 .pdev = pdev,
9693                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9694                         NOTIFY_DEVICEID4_DELETE,
9695         };
9696         struct nfs4_getdeviceinfo_res res = {
9697                 .pdev = pdev,
9698         };
9699         struct rpc_message msg = {
9700                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9701                 .rpc_argp = &args,
9702                 .rpc_resp = &res,
9703                 .rpc_cred = cred,
9704         };
9705         int status;
9706
9707         dprintk("--> %s\n", __func__);
9708         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9709         if (res.notification & ~args.notify_types)
9710                 dprintk("%s: unsupported notification\n", __func__);
9711         if (res.notification != args.notify_types)
9712                 pdev->nocache = 1;
9713
9714         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9715
9716         dprintk("<-- %s status=%d\n", __func__, status);
9717
9718         return status;
9719 }
9720
9721 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9722                 struct pnfs_device *pdev,
9723                 const struct cred *cred)
9724 {
9725         struct nfs4_exception exception = { };
9726         int err;
9727
9728         do {
9729                 err = nfs4_handle_exception(server,
9730                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9731                                         &exception);
9732         } while (exception.retry);
9733         return err;
9734 }
9735 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9736
9737 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9738 {
9739         struct nfs4_layoutcommit_data *data = calldata;
9740         struct nfs_server *server = NFS_SERVER(data->args.inode);
9741
9742         nfs4_setup_sequence(server->nfs_client,
9743                         &data->args.seq_args,
9744                         &data->res.seq_res,
9745                         task);
9746 }
9747
9748 static void
9749 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9750 {
9751         struct nfs4_layoutcommit_data *data = calldata;
9752         struct nfs_server *server = NFS_SERVER(data->args.inode);
9753
9754         if (!nfs41_sequence_done(task, &data->res.seq_res))
9755                 return;
9756
9757         switch (task->tk_status) { /* Just ignore these failures */
9758         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9759         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9760         case -NFS4ERR_BADLAYOUT:     /* no layout */
9761         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9762                 task->tk_status = 0;
9763                 break;
9764         case 0:
9765                 break;
9766         default:
9767                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9768                         rpc_restart_call_prepare(task);
9769                         return;
9770                 }
9771         }
9772 }
9773
9774 static void nfs4_layoutcommit_release(void *calldata)
9775 {
9776         struct nfs4_layoutcommit_data *data = calldata;
9777
9778         pnfs_cleanup_layoutcommit(data);
9779         nfs_post_op_update_inode_force_wcc(data->args.inode,
9780                                            data->res.fattr);
9781         put_cred(data->cred);
9782         nfs_iput_and_deactive(data->inode);
9783         kfree(data);
9784 }
9785
9786 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9787         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9788         .rpc_call_done = nfs4_layoutcommit_done,
9789         .rpc_release = nfs4_layoutcommit_release,
9790 };
9791
9792 int
9793 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9794 {
9795         struct rpc_message msg = {
9796                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9797                 .rpc_argp = &data->args,
9798                 .rpc_resp = &data->res,
9799                 .rpc_cred = data->cred,
9800         };
9801         struct rpc_task_setup task_setup_data = {
9802                 .task = &data->task,
9803                 .rpc_client = NFS_CLIENT(data->args.inode),
9804                 .rpc_message = &msg,
9805                 .callback_ops = &nfs4_layoutcommit_ops,
9806                 .callback_data = data,
9807         };
9808         struct rpc_task *task;
9809         int status = 0;
9810
9811         dprintk("NFS: initiating layoutcommit call. sync %d "
9812                 "lbw: %llu inode %lu\n", sync,
9813                 data->args.lastbytewritten,
9814                 data->args.inode->i_ino);
9815
9816         if (!sync) {
9817                 data->inode = nfs_igrab_and_active(data->args.inode);
9818                 if (data->inode == NULL) {
9819                         nfs4_layoutcommit_release(data);
9820                         return -EAGAIN;
9821                 }
9822                 task_setup_data.flags = RPC_TASK_ASYNC;
9823         }
9824         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9825         task = rpc_run_task(&task_setup_data);
9826         if (IS_ERR(task))
9827                 return PTR_ERR(task);
9828         if (sync)
9829                 status = task->tk_status;
9830         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9831         dprintk("%s: status %d\n", __func__, status);
9832         rpc_put_task(task);
9833         return status;
9834 }
9835
9836 /*
9837  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9838  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9839  */
9840 static int
9841 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9842                     struct nfs_fsinfo *info,
9843                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9844 {
9845         struct nfs41_secinfo_no_name_args args = {
9846                 .style = SECINFO_STYLE_CURRENT_FH,
9847         };
9848         struct nfs4_secinfo_res res = {
9849                 .flavors = flavors,
9850         };
9851         struct rpc_message msg = {
9852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9853                 .rpc_argp = &args,
9854                 .rpc_resp = &res,
9855         };
9856         struct nfs4_call_sync_data data = {
9857                 .seq_server = server,
9858                 .seq_args = &args.seq_args,
9859                 .seq_res = &res.seq_res,
9860         };
9861         struct rpc_task_setup task_setup = {
9862                 .rpc_client = server->client,
9863                 .rpc_message = &msg,
9864                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9865                 .callback_data = &data,
9866                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9867         };
9868         const struct cred *cred = NULL;
9869         int status;
9870
9871         if (use_integrity) {
9872                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9873
9874                 cred = nfs4_get_clid_cred(server->nfs_client);
9875                 msg.rpc_cred = cred;
9876         }
9877
9878         dprintk("--> %s\n", __func__);
9879         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9880         status = nfs4_call_sync_custom(&task_setup);
9881         dprintk("<-- %s status=%d\n", __func__, status);
9882
9883         put_cred(cred);
9884
9885         return status;
9886 }
9887
9888 static int
9889 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9890                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9891 {
9892         struct nfs4_exception exception = {
9893                 .interruptible = true,
9894         };
9895         int err;
9896         do {
9897                 /* first try using integrity protection */
9898                 err = -NFS4ERR_WRONGSEC;
9899
9900                 /* try to use integrity protection with machine cred */
9901                 if (_nfs4_is_integrity_protected(server->nfs_client))
9902                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9903                                                           flavors, true);
9904
9905                 /*
9906                  * if unable to use integrity protection, or SECINFO with
9907                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9908                  * disallowed by spec, but exists in deployed servers) use
9909                  * the current filesystem's rpc_client and the user cred.
9910                  */
9911                 if (err == -NFS4ERR_WRONGSEC)
9912                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9913                                                           flavors, false);
9914
9915                 switch (err) {
9916                 case 0:
9917                 case -NFS4ERR_WRONGSEC:
9918                 case -ENOTSUPP:
9919                         goto out;
9920                 default:
9921                         err = nfs4_handle_exception(server, err, &exception);
9922                 }
9923         } while (exception.retry);
9924 out:
9925         return err;
9926 }
9927
9928 static int
9929 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9930                     struct nfs_fsinfo *info)
9931 {
9932         int err;
9933         struct page *page;
9934         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9935         struct nfs4_secinfo_flavors *flavors;
9936         struct nfs4_secinfo4 *secinfo;
9937         int i;
9938
9939         page = alloc_page(GFP_KERNEL);
9940         if (!page) {
9941                 err = -ENOMEM;
9942                 goto out;
9943         }
9944
9945         flavors = page_address(page);
9946         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9947
9948         /*
9949          * Fall back on "guess and check" method if
9950          * the server doesn't support SECINFO_NO_NAME
9951          */
9952         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9953                 err = nfs4_find_root_sec(server, fhandle, info);
9954                 goto out_freepage;
9955         }
9956         if (err)
9957                 goto out_freepage;
9958
9959         for (i = 0; i < flavors->num_flavors; i++) {
9960                 secinfo = &flavors->flavors[i];
9961
9962                 switch (secinfo->flavor) {
9963                 case RPC_AUTH_NULL:
9964                 case RPC_AUTH_UNIX:
9965                 case RPC_AUTH_GSS:
9966                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9967                                         &secinfo->flavor_info);
9968                         break;
9969                 default:
9970                         flavor = RPC_AUTH_MAXFLAVOR;
9971                         break;
9972                 }
9973
9974                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9975                         flavor = RPC_AUTH_MAXFLAVOR;
9976
9977                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9978                         err = nfs4_lookup_root_sec(server, fhandle,
9979                                                    info, flavor);
9980                         if (!err)
9981                                 break;
9982                 }
9983         }
9984
9985         if (flavor == RPC_AUTH_MAXFLAVOR)
9986                 err = -EPERM;
9987
9988 out_freepage:
9989         put_page(page);
9990         if (err == -EACCES)
9991                 return -EPERM;
9992 out:
9993         return err;
9994 }
9995
9996 static int _nfs41_test_stateid(struct nfs_server *server,
9997                 nfs4_stateid *stateid,
9998                 const struct cred *cred)
9999 {
10000         int status;
10001         struct nfs41_test_stateid_args args = {
10002                 .stateid = stateid,
10003         };
10004         struct nfs41_test_stateid_res res;
10005         struct rpc_message msg = {
10006                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10007                 .rpc_argp = &args,
10008                 .rpc_resp = &res,
10009                 .rpc_cred = cred,
10010         };
10011         struct rpc_clnt *rpc_client = server->client;
10012
10013         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10014                 &rpc_client, &msg);
10015
10016         dprintk("NFS call  test_stateid %p\n", stateid);
10017         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10018         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10019                         &args.seq_args, &res.seq_res);
10020         if (status != NFS_OK) {
10021                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10022                 return status;
10023         }
10024         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10025         return -res.status;
10026 }
10027
10028 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10029                 int err, struct nfs4_exception *exception)
10030 {
10031         exception->retry = 0;
10032         switch(err) {
10033         case -NFS4ERR_DELAY:
10034         case -NFS4ERR_RETRY_UNCACHED_REP:
10035                 nfs4_handle_exception(server, err, exception);
10036                 break;
10037         case -NFS4ERR_BADSESSION:
10038         case -NFS4ERR_BADSLOT:
10039         case -NFS4ERR_BAD_HIGH_SLOT:
10040         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10041         case -NFS4ERR_DEADSESSION:
10042                 nfs4_do_handle_exception(server, err, exception);
10043         }
10044 }
10045
10046 /**
10047  * nfs41_test_stateid - perform a TEST_STATEID operation
10048  *
10049  * @server: server / transport on which to perform the operation
10050  * @stateid: state ID to test
10051  * @cred: credential
10052  *
10053  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10054  * Otherwise a negative NFS4ERR value is returned if the operation
10055  * failed or the state ID is not currently valid.
10056  */
10057 static int nfs41_test_stateid(struct nfs_server *server,
10058                 nfs4_stateid *stateid,
10059                 const struct cred *cred)
10060 {
10061         struct nfs4_exception exception = {
10062                 .interruptible = true,
10063         };
10064         int err;
10065         do {
10066                 err = _nfs41_test_stateid(server, stateid, cred);
10067                 nfs4_handle_delay_or_session_error(server, err, &exception);
10068         } while (exception.retry);
10069         return err;
10070 }
10071
10072 struct nfs_free_stateid_data {
10073         struct nfs_server *server;
10074         struct nfs41_free_stateid_args args;
10075         struct nfs41_free_stateid_res res;
10076 };
10077
10078 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10079 {
10080         struct nfs_free_stateid_data *data = calldata;
10081         nfs4_setup_sequence(data->server->nfs_client,
10082                         &data->args.seq_args,
10083                         &data->res.seq_res,
10084                         task);
10085 }
10086
10087 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10088 {
10089         struct nfs_free_stateid_data *data = calldata;
10090
10091         nfs41_sequence_done(task, &data->res.seq_res);
10092
10093         switch (task->tk_status) {
10094         case -NFS4ERR_DELAY:
10095                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10096                         rpc_restart_call_prepare(task);
10097         }
10098 }
10099
10100 static void nfs41_free_stateid_release(void *calldata)
10101 {
10102         kfree(calldata);
10103 }
10104
10105 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10106         .rpc_call_prepare = nfs41_free_stateid_prepare,
10107         .rpc_call_done = nfs41_free_stateid_done,
10108         .rpc_release = nfs41_free_stateid_release,
10109 };
10110
10111 /**
10112  * nfs41_free_stateid - perform a FREE_STATEID operation
10113  *
10114  * @server: server / transport on which to perform the operation
10115  * @stateid: state ID to release
10116  * @cred: credential
10117  * @privileged: set to true if this call needs to be privileged
10118  *
10119  * Note: this function is always asynchronous.
10120  */
10121 static int nfs41_free_stateid(struct nfs_server *server,
10122                 const nfs4_stateid *stateid,
10123                 const struct cred *cred,
10124                 bool privileged)
10125 {
10126         struct rpc_message msg = {
10127                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10128                 .rpc_cred = cred,
10129         };
10130         struct rpc_task_setup task_setup = {
10131                 .rpc_client = server->client,
10132                 .rpc_message = &msg,
10133                 .callback_ops = &nfs41_free_stateid_ops,
10134                 .flags = RPC_TASK_ASYNC,
10135         };
10136         struct nfs_free_stateid_data *data;
10137         struct rpc_task *task;
10138
10139         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10140                 &task_setup.rpc_client, &msg);
10141
10142         dprintk("NFS call  free_stateid %p\n", stateid);
10143         data = kmalloc(sizeof(*data), GFP_NOFS);
10144         if (!data)
10145                 return -ENOMEM;
10146         data->server = server;
10147         nfs4_stateid_copy(&data->args.stateid, stateid);
10148
10149         task_setup.callback_data = data;
10150
10151         msg.rpc_argp = &data->args;
10152         msg.rpc_resp = &data->res;
10153         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10154         task = rpc_run_task(&task_setup);
10155         if (IS_ERR(task))
10156                 return PTR_ERR(task);
10157         rpc_put_task(task);
10158         return 0;
10159 }
10160
10161 static void
10162 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10163 {
10164         const struct cred *cred = lsp->ls_state->owner->so_cred;
10165
10166         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10167         nfs4_free_lock_state(server, lsp);
10168 }
10169
10170 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10171                 const nfs4_stateid *s2)
10172 {
10173         if (s1->type != s2->type)
10174                 return false;
10175
10176         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10177                 return false;
10178
10179         if (s1->seqid == s2->seqid)
10180                 return true;
10181
10182         return s1->seqid == 0 || s2->seqid == 0;
10183 }
10184
10185 #endif /* CONFIG_NFS_V4_1 */
10186
10187 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10188                 const nfs4_stateid *s2)
10189 {
10190         return nfs4_stateid_match(s1, s2);
10191 }
10192
10193
10194 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10195         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10196         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10197         .recover_open   = nfs4_open_reclaim,
10198         .recover_lock   = nfs4_lock_reclaim,
10199         .establish_clid = nfs4_init_clientid,
10200         .detect_trunking = nfs40_discover_server_trunking,
10201 };
10202
10203 #if defined(CONFIG_NFS_V4_1)
10204 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10205         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10206         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10207         .recover_open   = nfs4_open_reclaim,
10208         .recover_lock   = nfs4_lock_reclaim,
10209         .establish_clid = nfs41_init_clientid,
10210         .reclaim_complete = nfs41_proc_reclaim_complete,
10211         .detect_trunking = nfs41_discover_server_trunking,
10212 };
10213 #endif /* CONFIG_NFS_V4_1 */
10214
10215 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10216         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10217         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10218         .recover_open   = nfs40_open_expired,
10219         .recover_lock   = nfs4_lock_expired,
10220         .establish_clid = nfs4_init_clientid,
10221 };
10222
10223 #if defined(CONFIG_NFS_V4_1)
10224 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10225         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10226         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10227         .recover_open   = nfs41_open_expired,
10228         .recover_lock   = nfs41_lock_expired,
10229         .establish_clid = nfs41_init_clientid,
10230 };
10231 #endif /* CONFIG_NFS_V4_1 */
10232
10233 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10234         .sched_state_renewal = nfs4_proc_async_renew,
10235         .get_state_renewal_cred = nfs4_get_renew_cred,
10236         .renew_lease = nfs4_proc_renew,
10237 };
10238
10239 #if defined(CONFIG_NFS_V4_1)
10240 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10241         .sched_state_renewal = nfs41_proc_async_sequence,
10242         .get_state_renewal_cred = nfs4_get_machine_cred,
10243         .renew_lease = nfs4_proc_sequence,
10244 };
10245 #endif
10246
10247 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10248         .get_locations = _nfs40_proc_get_locations,
10249         .fsid_present = _nfs40_proc_fsid_present,
10250 };
10251
10252 #if defined(CONFIG_NFS_V4_1)
10253 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10254         .get_locations = _nfs41_proc_get_locations,
10255         .fsid_present = _nfs41_proc_fsid_present,
10256 };
10257 #endif  /* CONFIG_NFS_V4_1 */
10258
10259 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10260         .minor_version = 0,
10261         .init_caps = NFS_CAP_READDIRPLUS
10262                 | NFS_CAP_ATOMIC_OPEN
10263                 | NFS_CAP_POSIX_LOCK,
10264         .init_client = nfs40_init_client,
10265         .shutdown_client = nfs40_shutdown_client,
10266         .match_stateid = nfs4_match_stateid,
10267         .find_root_sec = nfs4_find_root_sec,
10268         .free_lock_state = nfs4_release_lockowner,
10269         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10270         .alloc_seqid = nfs_alloc_seqid,
10271         .call_sync_ops = &nfs40_call_sync_ops,
10272         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10273         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10274         .state_renewal_ops = &nfs40_state_renewal_ops,
10275         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10276 };
10277
10278 #if defined(CONFIG_NFS_V4_1)
10279 static struct nfs_seqid *
10280 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10281 {
10282         return NULL;
10283 }
10284
10285 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10286         .minor_version = 1,
10287         .init_caps = NFS_CAP_READDIRPLUS
10288                 | NFS_CAP_ATOMIC_OPEN
10289                 | NFS_CAP_POSIX_LOCK
10290                 | NFS_CAP_STATEID_NFSV41
10291                 | NFS_CAP_ATOMIC_OPEN_V1
10292                 | NFS_CAP_LGOPEN,
10293         .init_client = nfs41_init_client,
10294         .shutdown_client = nfs41_shutdown_client,
10295         .match_stateid = nfs41_match_stateid,
10296         .find_root_sec = nfs41_find_root_sec,
10297         .free_lock_state = nfs41_free_lock_state,
10298         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10299         .alloc_seqid = nfs_alloc_no_seqid,
10300         .session_trunk = nfs4_test_session_trunk,
10301         .call_sync_ops = &nfs41_call_sync_ops,
10302         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10303         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10304         .state_renewal_ops = &nfs41_state_renewal_ops,
10305         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10306 };
10307 #endif
10308
10309 #if defined(CONFIG_NFS_V4_2)
10310 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10311         .minor_version = 2,
10312         .init_caps = NFS_CAP_READDIRPLUS
10313                 | NFS_CAP_ATOMIC_OPEN
10314                 | NFS_CAP_POSIX_LOCK
10315                 | NFS_CAP_STATEID_NFSV41
10316                 | NFS_CAP_ATOMIC_OPEN_V1
10317                 | NFS_CAP_LGOPEN
10318                 | NFS_CAP_ALLOCATE
10319                 | NFS_CAP_COPY
10320                 | NFS_CAP_OFFLOAD_CANCEL
10321                 | NFS_CAP_COPY_NOTIFY
10322                 | NFS_CAP_DEALLOCATE
10323                 | NFS_CAP_SEEK
10324                 | NFS_CAP_LAYOUTSTATS
10325                 | NFS_CAP_CLONE
10326                 | NFS_CAP_LAYOUTERROR
10327                 | NFS_CAP_READ_PLUS,
10328         .init_client = nfs41_init_client,
10329         .shutdown_client = nfs41_shutdown_client,
10330         .match_stateid = nfs41_match_stateid,
10331         .find_root_sec = nfs41_find_root_sec,
10332         .free_lock_state = nfs41_free_lock_state,
10333         .call_sync_ops = &nfs41_call_sync_ops,
10334         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10335         .alloc_seqid = nfs_alloc_no_seqid,
10336         .session_trunk = nfs4_test_session_trunk,
10337         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10338         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10339         .state_renewal_ops = &nfs41_state_renewal_ops,
10340         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10341 };
10342 #endif
10343
10344 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10345         [0] = &nfs_v4_0_minor_ops,
10346 #if defined(CONFIG_NFS_V4_1)
10347         [1] = &nfs_v4_1_minor_ops,
10348 #endif
10349 #if defined(CONFIG_NFS_V4_2)
10350         [2] = &nfs_v4_2_minor_ops,
10351 #endif
10352 };
10353
10354 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10355 {
10356         ssize_t error, error2, error3;
10357
10358         error = generic_listxattr(dentry, list, size);
10359         if (error < 0)
10360                 return error;
10361         if (list) {
10362                 list += error;
10363                 size -= error;
10364         }
10365
10366         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10367         if (error2 < 0)
10368                 return error2;
10369
10370         if (list) {
10371                 list += error2;
10372                 size -= error2;
10373         }
10374
10375         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10376         if (error3 < 0)
10377                 return error3;
10378
10379         return error + error2 + error3;
10380 }
10381
10382 static const struct inode_operations nfs4_dir_inode_operations = {
10383         .create         = nfs_create,
10384         .lookup         = nfs_lookup,
10385         .atomic_open    = nfs_atomic_open,
10386         .link           = nfs_link,
10387         .unlink         = nfs_unlink,
10388         .symlink        = nfs_symlink,
10389         .mkdir          = nfs_mkdir,
10390         .rmdir          = nfs_rmdir,
10391         .mknod          = nfs_mknod,
10392         .rename         = nfs_rename,
10393         .permission     = nfs_permission,
10394         .getattr        = nfs_getattr,
10395         .setattr        = nfs_setattr,
10396         .listxattr      = nfs4_listxattr,
10397 };
10398
10399 static const struct inode_operations nfs4_file_inode_operations = {
10400         .permission     = nfs_permission,
10401         .getattr        = nfs_getattr,
10402         .setattr        = nfs_setattr,
10403         .listxattr      = nfs4_listxattr,
10404 };
10405
10406 const struct nfs_rpc_ops nfs_v4_clientops = {
10407         .version        = 4,                    /* protocol version */
10408         .dentry_ops     = &nfs4_dentry_operations,
10409         .dir_inode_ops  = &nfs4_dir_inode_operations,
10410         .file_inode_ops = &nfs4_file_inode_operations,
10411         .file_ops       = &nfs4_file_operations,
10412         .getroot        = nfs4_proc_get_root,
10413         .submount       = nfs4_submount,
10414         .try_get_tree   = nfs4_try_get_tree,
10415         .getattr        = nfs4_proc_getattr,
10416         .setattr        = nfs4_proc_setattr,
10417         .lookup         = nfs4_proc_lookup,
10418         .lookupp        = nfs4_proc_lookupp,
10419         .access         = nfs4_proc_access,
10420         .readlink       = nfs4_proc_readlink,
10421         .create         = nfs4_proc_create,
10422         .remove         = nfs4_proc_remove,
10423         .unlink_setup   = nfs4_proc_unlink_setup,
10424         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10425         .unlink_done    = nfs4_proc_unlink_done,
10426         .rename_setup   = nfs4_proc_rename_setup,
10427         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10428         .rename_done    = nfs4_proc_rename_done,
10429         .link           = nfs4_proc_link,
10430         .symlink        = nfs4_proc_symlink,
10431         .mkdir          = nfs4_proc_mkdir,
10432         .rmdir          = nfs4_proc_rmdir,
10433         .readdir        = nfs4_proc_readdir,
10434         .mknod          = nfs4_proc_mknod,
10435         .statfs         = nfs4_proc_statfs,
10436         .fsinfo         = nfs4_proc_fsinfo,
10437         .pathconf       = nfs4_proc_pathconf,
10438         .set_capabilities = nfs4_server_capabilities,
10439         .decode_dirent  = nfs4_decode_dirent,
10440         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10441         .read_setup     = nfs4_proc_read_setup,
10442         .read_done      = nfs4_read_done,
10443         .write_setup    = nfs4_proc_write_setup,
10444         .write_done     = nfs4_write_done,
10445         .commit_setup   = nfs4_proc_commit_setup,
10446         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10447         .commit_done    = nfs4_commit_done,
10448         .lock           = nfs4_proc_lock,
10449         .clear_acl_cache = nfs4_zap_acl_attr,
10450         .close_context  = nfs4_close_context,
10451         .open_context   = nfs4_atomic_open,
10452         .have_delegation = nfs4_have_delegation,
10453         .alloc_client   = nfs4_alloc_client,
10454         .init_client    = nfs4_init_client,
10455         .free_client    = nfs4_free_client,
10456         .create_server  = nfs4_create_server,
10457         .clone_server   = nfs_clone_server,
10458 };
10459
10460 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10461         .name   = XATTR_NAME_NFSV4_ACL,
10462         .list   = nfs4_xattr_list_nfs4_acl,
10463         .get    = nfs4_xattr_get_nfs4_acl,
10464         .set    = nfs4_xattr_set_nfs4_acl,
10465 };
10466
10467 #ifdef CONFIG_NFS_V4_2
10468 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10469         .prefix = XATTR_USER_PREFIX,
10470         .get    = nfs4_xattr_get_nfs4_user,
10471         .set    = nfs4_xattr_set_nfs4_user,
10472 };
10473 #endif
10474
10475 const struct xattr_handler *nfs4_xattr_handlers[] = {
10476         &nfs4_xattr_nfs4_acl_handler,
10477 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10478         &nfs4_xattr_nfs4_label_handler,
10479 #endif
10480 #ifdef CONFIG_NFS_V4_2
10481         &nfs4_xattr_nfs4_user_handler,
10482 #endif
10483         NULL
10484 };