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