fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "sysfs.h"
  67 #include "nfs4idmap.h"
  68 #include "nfs4session.h"
  69 #include "fscache.h"
  70 #include "nfs42.h"
  71
  72 #include "nfs4trace.h"
  73
  74 #define NFSDBG_FACILITY         NFSDBG_PROC
  75
  76 #define NFS4_BITMASK_SZ         3
  77
  78 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  79 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  80
  81 /* file attributes which can be mapped to nfs attributes */
  82 #define NFS4_VALID_ATTRS (ATTR_MODE \
  83         | ATTR_UID \
  84         | ATTR_GID \
  85         | ATTR_SIZE \
  86         | ATTR_ATIME \
  87         | ATTR_MTIME \
  88         | ATTR_CTIME \
  89         | ATTR_ATIME_SET \
  90         | ATTR_MTIME_SET)
  91
  92 struct nfs4_opendata;
  93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  97 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  98                             struct nfs_fattr *fattr, struct iattr *sattr,
  99                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
 100                             struct nfs4_label *olabel);
 101 #ifdef CONFIG_NFS_V4_1
 102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 103                 const struct cred *cred,
 104                 struct nfs4_slot *slot,
 105                 bool is_privileged);
 106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 107                 const struct cred *);
 108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 109                 const struct cred *, bool);
 110 #endif
 111 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ],
 112                              const __u32 *src, struct inode *inode,
 113                              struct nfs_server *server,
 114                              struct nfs4_label *label);
 115
 116 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 117 static inline struct nfs4_label *
 118 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 119         struct iattr *sattr, struct nfs4_label *label)
 120 {
 121         int err;
 122
 123         if (label == NULL)
 124                 return NULL;
 125
 126         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 127                 return NULL;
 128
 129         err = security_dentry_init_security(dentry, sattr->ia_mode,
 130                                 &dentry->d_name, (void **)&label->label, &label->len);
 131         if (err == 0)
 132                 return label;
 133
 134         return NULL;
 135 }
 136 static inline void
 137 nfs4_label_release_security(struct nfs4_label *label)
 138 {
 139         if (label)
 140                 security_release_secctx(label->label, label->len);
 141 }
 142 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 143 {
 144         if (label)
 145                 return server->attr_bitmask;
 146
 147         return server->attr_bitmask_nl;
 148 }
 149 #else
 150 static inline struct nfs4_label *
 151 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 152         struct iattr *sattr, struct nfs4_label *l)
 153 { return NULL; }
 154 static inline void
 155 nfs4_label_release_security(struct nfs4_label *label)
 156 { return; }
 157 static inline u32 *
 158 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 159 { return server->attr_bitmask; }
 160 #endif
 161
 162 /* Prevent leaks of NFSv4 errors into userland */
 163 static int nfs4_map_errors(int err)
 164 {
 165         if (err >= -1000)
 166                 return err;
 167         switch (err) {
 168         case -NFS4ERR_RESOURCE:
 169         case -NFS4ERR_LAYOUTTRYLATER:
 170         case -NFS4ERR_RECALLCONFLICT:
 171                 return -EREMOTEIO;
 172         case -NFS4ERR_WRONGSEC:
 173         case -NFS4ERR_WRONG_CRED:
 174                 return -EPERM;
 175         case -NFS4ERR_BADOWNER:
 176         case -NFS4ERR_BADNAME:
 177                 return -EINVAL;
 178         case -NFS4ERR_SHARE_DENIED:
 179                 return -EACCES;
 180         case -NFS4ERR_MINOR_VERS_MISMATCH:
 181                 return -EPROTONOSUPPORT;
 182         case -NFS4ERR_FILE_OPEN:
 183                 return -EBUSY;
 184         case -NFS4ERR_NOT_SAME:
 185                 return -ENOTSYNC;
 186         default:
 187                 dprintk("%s could not handle NFSv4 error %d\n",
 188                                 __func__, -err);
 189                 break;
 190         }
 191         return -EIO;
 192 }
 193
 194 /*
 195  * This is our standard bitmap for GETATTR requests.
 196  */
 197 const u32 nfs4_fattr_bitmap[3] = {
 198         FATTR4_WORD0_TYPE
 199         | FATTR4_WORD0_CHANGE
 200         | FATTR4_WORD0_SIZE
 201         | FATTR4_WORD0_FSID
 202         | FATTR4_WORD0_FILEID,
 203         FATTR4_WORD1_MODE
 204         | FATTR4_WORD1_NUMLINKS
 205         | FATTR4_WORD1_OWNER
 206         | FATTR4_WORD1_OWNER_GROUP
 207         | FATTR4_WORD1_RAWDEV
 208         | FATTR4_WORD1_SPACE_USED
 209         | FATTR4_WORD1_TIME_ACCESS
 210         | FATTR4_WORD1_TIME_METADATA
 211         | FATTR4_WORD1_TIME_MODIFY
 212         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 213 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 214         FATTR4_WORD2_SECURITY_LABEL
 215 #endif
 216 };
 217
 218 static const u32 nfs4_pnfs_open_bitmap[3] = {
 219         FATTR4_WORD0_TYPE
 220         | FATTR4_WORD0_CHANGE
 221         | FATTR4_WORD0_SIZE
 222         | FATTR4_WORD0_FSID
 223         | FATTR4_WORD0_FILEID,
 224         FATTR4_WORD1_MODE
 225         | FATTR4_WORD1_NUMLINKS
 226         | FATTR4_WORD1_OWNER
 227         | FATTR4_WORD1_OWNER_GROUP
 228         | FATTR4_WORD1_RAWDEV
 229         | FATTR4_WORD1_SPACE_USED
 230         | FATTR4_WORD1_TIME_ACCESS
 231         | FATTR4_WORD1_TIME_METADATA
 232         | FATTR4_WORD1_TIME_MODIFY,
 233         FATTR4_WORD2_MDSTHRESHOLD
 234 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 235         | FATTR4_WORD2_SECURITY_LABEL
 236 #endif
 237 };
 238
 239 static const u32 nfs4_open_noattr_bitmap[3] = {
 240         FATTR4_WORD0_TYPE
 241         | FATTR4_WORD0_FILEID,
 242 };
 243
 244 const u32 nfs4_statfs_bitmap[3] = {
 245         FATTR4_WORD0_FILES_AVAIL
 246         | FATTR4_WORD0_FILES_FREE
 247         | FATTR4_WORD0_FILES_TOTAL,
 248         FATTR4_WORD1_SPACE_AVAIL
 249         | FATTR4_WORD1_SPACE_FREE
 250         | FATTR4_WORD1_SPACE_TOTAL
 251 };
 252
 253 const u32 nfs4_pathconf_bitmap[3] = {
 254         FATTR4_WORD0_MAXLINK
 255         | FATTR4_WORD0_MAXNAME,
 256         0
 257 };
 258
 259 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 260                         | FATTR4_WORD0_MAXREAD
 261                         | FATTR4_WORD0_MAXWRITE
 262                         | FATTR4_WORD0_LEASE_TIME,
 263                         FATTR4_WORD1_TIME_DELTA
 264                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 265                         FATTR4_WORD2_LAYOUT_BLKSIZE
 266                         | FATTR4_WORD2_CLONE_BLKSIZE
 267                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
 268                         | FATTR4_WORD2_XATTR_SUPPORT
 269 };
 270
 271 const u32 nfs4_fs_locations_bitmap[3] = {
 272         FATTR4_WORD0_CHANGE
 273         | FATTR4_WORD0_SIZE
 274         | FATTR4_WORD0_FSID
 275         | FATTR4_WORD0_FILEID
 276         | FATTR4_WORD0_FS_LOCATIONS,
 277         FATTR4_WORD1_OWNER
 278         | FATTR4_WORD1_OWNER_GROUP
 279         | FATTR4_WORD1_RAWDEV
 280         | FATTR4_WORD1_SPACE_USED
 281         | FATTR4_WORD1_TIME_ACCESS
 282         | FATTR4_WORD1_TIME_METADATA
 283         | FATTR4_WORD1_TIME_MODIFY
 284         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 285 };
 286
 287 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 288                                     struct inode *inode, unsigned long flags)
 289 {
 290         unsigned long cache_validity;
 291
 292         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 293         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 294                 return;
 295
 296         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
 297
 298         /* Remove the attributes over which we have full control */
 299         dst[1] &= ~FATTR4_WORD1_RAWDEV;
 300         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 301                 dst[0] &= ~FATTR4_WORD0_SIZE;
 302
 303         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 304                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 305
 306         if (!(cache_validity & NFS_INO_INVALID_MODE))
 307                 dst[1] &= ~FATTR4_WORD1_MODE;
 308         if (!(cache_validity & NFS_INO_INVALID_OTHER))
 309                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
 310 }
 311
 312 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 313                 struct nfs4_readdir_arg *readdir)
 314 {
 315         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 316         __be32 *start, *p;
 317
 318         if (cookie > 2) {
 319                 readdir->cookie = cookie;
 320                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 321                 return;
 322         }
 323
 324         readdir->cookie = 0;
 325         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 326         if (cookie == 2)
 327                 return;
 328
 329         /*
 330          * NFSv4 servers do not return entries for '.' and '..'
 331          * Therefore, we fake these entries here.  We let '.'
 332          * have cookie 0 and '..' have cookie 1.  Note that
 333          * when talking to the server, we always send cookie 0
 334          * instead of 1 or 2.
 335          */
 336         start = p = kmap_atomic(*readdir->pages);
 337
 338         if (cookie == 0) {
 339                 *p++ = xdr_one;                                  /* next */
 340                 *p++ = xdr_zero;                   /* cookie, first word */
 341                 *p++ = xdr_one;                   /* cookie, second word */
 342                 *p++ = xdr_one;                             /* entry len */
 343                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 344                 p++;
 345                 *p++ = xdr_one;                         /* bitmap length */
 346                 *p++ = htonl(attrs);                           /* bitmap */
 347                 *p++ = htonl(12);             /* attribute buffer length */
 348                 *p++ = htonl(NF4DIR);
 349                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 350         }
 351
 352         *p++ = xdr_one;                                  /* next */
 353         *p++ = xdr_zero;                   /* cookie, first word */
 354         *p++ = xdr_two;                   /* cookie, second word */
 355         *p++ = xdr_two;                             /* entry len */
 356         memcpy(p, "..\0\0", 4);                         /* entry */
 357         p++;
 358         *p++ = xdr_one;                         /* bitmap length */
 359         *p++ = htonl(attrs);                           /* bitmap */
 360         *p++ = htonl(12);             /* attribute buffer length */
 361         *p++ = htonl(NF4DIR);
 362         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 363
 364         readdir->pgbase = (char *)p - (char *)start;
 365         readdir->count -= readdir->pgbase;
 366         kunmap_atomic(start);
 367 }
 368
 369 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 370                 nfs4_stateid *stateid,
 371                 const struct cred *cred)
 372 {
 373         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 374
 375         ops->test_and_free_expired(server, stateid, cred);
 376 }
 377
 378 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 383         nfs4_test_and_free_stateid(server, stateid, cred);
 384 }
 385
 386 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 387                 const nfs4_stateid *stateid,
 388                 const struct cred *cred)
 389 {
 390         nfs4_stateid tmp;
 391
 392         nfs4_stateid_copy(&tmp, stateid);
 393         __nfs4_free_revoked_stateid(server, &tmp, cred);
 394 }
 395
 396 static long nfs4_update_delay(long *timeout)
 397 {
 398         long ret;
 399         if (!timeout)
 400                 return NFS4_POLL_RETRY_MAX;
 401         if (*timeout <= 0)
 402                 *timeout = NFS4_POLL_RETRY_MIN;
 403         if (*timeout > NFS4_POLL_RETRY_MAX)
 404                 *timeout = NFS4_POLL_RETRY_MAX;
 405         ret = *timeout;
 406         *timeout <<= 1;
 407         return ret;
 408 }
 409
 410 static int nfs4_delay_killable(long *timeout)
 411 {
 412         might_sleep();
 413
 414         freezable_schedule_timeout_killable_unsafe(
 415                 nfs4_update_delay(timeout));
 416         if (!__fatal_signal_pending(current))
 417                 return 0;
 418         return -EINTR;
 419 }
 420
 421 static int nfs4_delay_interruptible(long *timeout)
 422 {
 423         might_sleep();
 424
 425         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 426         if (!signal_pending(current))
 427                 return 0;
 428         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 429 }
 430
 431 static int nfs4_delay(long *timeout, bool interruptible)
 432 {
 433         if (interruptible)
 434                 return nfs4_delay_interruptible(timeout);
 435         return nfs4_delay_killable(timeout);
 436 }
 437
 438 static const nfs4_stateid *
 439 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 440 {
 441         if (!stateid)
 442                 return NULL;
 443         switch (stateid->type) {
 444         case NFS4_OPEN_STATEID_TYPE:
 445         case NFS4_LOCK_STATEID_TYPE:
 446         case NFS4_DELEGATION_STATEID_TYPE:
 447                 return stateid;
 448         default:
 449                 break;
 450         }
 451         return NULL;
 452 }
 453
 454 /* This is the error handling routine for processes that are allowed
 455  * to sleep.
 456  */
 457 static int nfs4_do_handle_exception(struct nfs_server *server,
 458                 int errorcode, struct nfs4_exception *exception)
 459 {
 460         struct nfs_client *clp = server->nfs_client;
 461         struct nfs4_state *state = exception->state;
 462         const nfs4_stateid *stateid;
 463         struct inode *inode = exception->inode;
 464         int ret = errorcode;
 465
 466         exception->delay = 0;
 467         exception->recovering = 0;
 468         exception->retry = 0;
 469
 470         stateid = nfs4_recoverable_stateid(exception->stateid);
 471         if (stateid == NULL && state != NULL)
 472                 stateid = nfs4_recoverable_stateid(&state->stateid);
 473
 474         switch(errorcode) {
 475                 case 0:
 476                         return 0;
 477                 case -NFS4ERR_BADHANDLE:
 478                 case -ESTALE:
 479                         if (inode != NULL && S_ISREG(inode->i_mode))
 480                                 pnfs_destroy_layout(NFS_I(inode));
 481                         break;
 482                 case -NFS4ERR_DELEG_REVOKED:
 483                 case -NFS4ERR_ADMIN_REVOKED:
 484                 case -NFS4ERR_EXPIRED:
 485                 case -NFS4ERR_BAD_STATEID:
 486                 case -NFS4ERR_PARTNER_NO_AUTH:
 487                         if (inode != NULL && stateid != NULL) {
 488                                 nfs_inode_find_state_and_recover(inode,
 489                                                 stateid);
 490                                 goto wait_on_recovery;
 491                         }
 492                         fallthrough;
 493                 case -NFS4ERR_OPENMODE:
 494                         if (inode) {
 495                                 int err;
 496
 497                                 err = nfs_async_inode_return_delegation(inode,
 498                                                 stateid);
 499                                 if (err == 0)
 500                                         goto wait_on_recovery;
 501                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 502                                         exception->retry = 1;
 503                                         break;
 504                                 }
 505                         }
 506                         if (state == NULL)
 507                                 break;
 508                         ret = nfs4_schedule_stateid_recovery(server, state);
 509                         if (ret < 0)
 510                                 break;
 511                         goto wait_on_recovery;
 512                 case -NFS4ERR_STALE_STATEID:
 513                 case -NFS4ERR_STALE_CLIENTID:
 514                         nfs4_schedule_lease_recovery(clp);
 515                         goto wait_on_recovery;
 516                 case -NFS4ERR_MOVED:
 517                         ret = nfs4_schedule_migration_recovery(server);
 518                         if (ret < 0)
 519                                 break;
 520                         goto wait_on_recovery;
 521                 case -NFS4ERR_LEASE_MOVED:
 522                         nfs4_schedule_lease_moved_recovery(clp);
 523                         goto wait_on_recovery;
 524 #if defined(CONFIG_NFS_V4_1)
 525                 case -NFS4ERR_BADSESSION:
 526                 case -NFS4ERR_BADSLOT:
 527                 case -NFS4ERR_BAD_HIGH_SLOT:
 528                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 529                 case -NFS4ERR_DEADSESSION:
 530                 case -NFS4ERR_SEQ_FALSE_RETRY:
 531                 case -NFS4ERR_SEQ_MISORDERED:
 532                         /* Handled in nfs41_sequence_process() */
 533                         goto wait_on_recovery;
 534 #endif /* defined(CONFIG_NFS_V4_1) */
 535                 case -NFS4ERR_FILE_OPEN:
 536                         if (exception->timeout > HZ) {
 537                                 /* We have retried a decent amount, time to
 538                                  * fail
 539                                  */
 540                                 ret = -EBUSY;
 541                                 break;
 542                         }
 543                         fallthrough;
 544                 case -NFS4ERR_DELAY:
 545                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 546                         fallthrough;
 547                 case -NFS4ERR_GRACE:
 548                 case -NFS4ERR_LAYOUTTRYLATER:
 549                 case -NFS4ERR_RECALLCONFLICT:
 550                         exception->delay = 1;
 551                         return 0;
 552
 553                 case -NFS4ERR_RETRY_UNCACHED_REP:
 554                 case -NFS4ERR_OLD_STATEID:
 555                         exception->retry = 1;
 556                         break;
 557                 case -NFS4ERR_BADOWNER:
 558                         /* The following works around a Linux server bug! */
 559                 case -NFS4ERR_BADNAME:
 560                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 561                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 562                                 exception->retry = 1;
 563                                 printk(KERN_WARNING "NFS: v4 server %s "
 564                                                 "does not accept raw "
 565                                                 "uid/gids. "
 566                                                 "Reenabling the idmapper.\n",
 567                                                 server->nfs_client->cl_hostname);
 568                         }
 569         }
 570         /* We failed to handle the error */
 571         return nfs4_map_errors(ret);
 572 wait_on_recovery:
 573         exception->recovering = 1;
 574         return 0;
 575 }
 576
 577 /* This is the error handling routine for processes that are allowed
 578  * to sleep.
 579  */
 580 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 581 {
 582         struct nfs_client *clp = server->nfs_client;
 583         int ret;
 584
 585         ret = nfs4_do_handle_exception(server, errorcode, exception);
 586         if (exception->delay) {
 587                 ret = nfs4_delay(&exception->timeout,
 588                                 exception->interruptible);
 589                 goto out_retry;
 590         }
 591         if (exception->recovering) {
 592                 if (exception->task_is_privileged)
 593                         return -EDEADLOCK;
 594                 ret = nfs4_wait_clnt_recover(clp);
 595                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 596                         return -EIO;
 597                 goto out_retry;
 598         }
 599         return ret;
 600 out_retry:
 601         if (ret == 0)
 602                 exception->retry = 1;
 603         return ret;
 604 }
 605
 606 static int
 607 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 608                 int errorcode, struct nfs4_exception *exception)
 609 {
 610         struct nfs_client *clp = server->nfs_client;
 611         int ret;
 612
 613         ret = nfs4_do_handle_exception(server, errorcode, exception);
 614         if (exception->delay) {
 615                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 616                 goto out_retry;
 617         }
 618         if (exception->recovering) {
 619                 if (exception->task_is_privileged)
 620                         return -EDEADLOCK;
 621                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 622                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 623                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 624                 goto out_retry;
 625         }
 626         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 627                 ret = -EIO;
 628         return ret;
 629 out_retry:
 630         if (ret == 0) {
 631                 exception->retry = 1;
 632                 /*
 633                  * For NFS4ERR_MOVED, the client transport will need to
 634                  * be recomputed after migration recovery has completed.
 635                  */
 636                 if (errorcode == -NFS4ERR_MOVED)
 637                         rpc_task_release_transport(task);
 638         }
 639         return ret;
 640 }
 641
 642 int
 643 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 644                         struct nfs4_state *state, long *timeout)
 645 {
 646         struct nfs4_exception exception = {
 647                 .state = state,
 648         };
 649
 650         if (task->tk_status >= 0)
 651                 return 0;
 652         if (timeout)
 653                 exception.timeout = *timeout;
 654         task->tk_status = nfs4_async_handle_exception(task, server,
 655                         task->tk_status,
 656                         &exception);
 657         if (exception.delay && timeout)
 658                 *timeout = exception.timeout;
 659         if (exception.retry)
 660                 return -EAGAIN;
 661         return 0;
 662 }
 663
 664 /*
 665  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 666  * or 'false' otherwise.
 667  */
 668 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 669 {
 670         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 671         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 672 }
 673
 674 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 675 {
 676         spin_lock(&clp->cl_lock);
 677         if (time_before(clp->cl_last_renewal,timestamp))
 678                 clp->cl_last_renewal = timestamp;
 679         spin_unlock(&clp->cl_lock);
 680 }
 681
 682 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 683 {
 684         struct nfs_client *clp = server->nfs_client;
 685
 686         if (!nfs4_has_session(clp))
 687                 do_renew_lease(clp, timestamp);
 688 }
 689
 690 struct nfs4_call_sync_data {
 691         const struct nfs_server *seq_server;
 692         struct nfs4_sequence_args *seq_args;
 693         struct nfs4_sequence_res *seq_res;
 694 };
 695
 696 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 697                         struct nfs4_sequence_res *res, int cache_reply,
 698                         int privileged)
 699 {
 700         args->sa_slot = NULL;
 701         args->sa_cache_this = cache_reply;
 702         args->sa_privileged = privileged;
 703
 704         res->sr_slot = NULL;
 705 }
 706
 707 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 708 {
 709         struct nfs4_slot *slot = res->sr_slot;
 710         struct nfs4_slot_table *tbl;
 711
 712         tbl = slot->table;
 713         spin_lock(&tbl->slot_tbl_lock);
 714         if (!nfs41_wake_and_assign_slot(tbl, slot))
 715                 nfs4_free_slot(tbl, slot);
 716         spin_unlock(&tbl->slot_tbl_lock);
 717
 718         res->sr_slot = NULL;
 719 }
 720
 721 static int nfs40_sequence_done(struct rpc_task *task,
 722                                struct nfs4_sequence_res *res)
 723 {
 724         if (res->sr_slot != NULL)
 725                 nfs40_sequence_free_slot(res);
 726         return 1;
 727 }
 728
 729 #if defined(CONFIG_NFS_V4_1)
 730
 731 static void nfs41_release_slot(struct nfs4_slot *slot)
 732 {
 733         struct nfs4_session *session;
 734         struct nfs4_slot_table *tbl;
 735         bool send_new_highest_used_slotid = false;
 736
 737         if (!slot)
 738                 return;
 739         tbl = slot->table;
 740         session = tbl->session;
 741
 742         /* Bump the slot sequence number */
 743         if (slot->seq_done)
 744                 slot->seq_nr++;
 745         slot->seq_done = 0;
 746
 747         spin_lock(&tbl->slot_tbl_lock);
 748         /* Be nice to the server: try to ensure that the last transmitted
 749          * value for highest_user_slotid <= target_highest_slotid
 750          */
 751         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 752                 send_new_highest_used_slotid = true;
 753
 754         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 755                 send_new_highest_used_slotid = false;
 756                 goto out_unlock;
 757         }
 758         nfs4_free_slot(tbl, slot);
 759
 760         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 761                 send_new_highest_used_slotid = false;
 762 out_unlock:
 763         spin_unlock(&tbl->slot_tbl_lock);
 764         if (send_new_highest_used_slotid)
 765                 nfs41_notify_server(session->clp);
 766         if (waitqueue_active(&tbl->slot_waitq))
 767                 wake_up_all(&tbl->slot_waitq);
 768 }
 769
 770 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 771 {
 772         nfs41_release_slot(res->sr_slot);
 773         res->sr_slot = NULL;
 774 }
 775
 776 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 777                 u32 seqnr)
 778 {
 779         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 780                 slot->seq_nr_highest_sent = seqnr;
 781 }
 782 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 783                 u32 seqnr)
 784 {
 785         slot->seq_nr_highest_sent = seqnr;
 786         slot->seq_nr_last_acked = seqnr;
 787 }
 788
 789 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 790                                 struct nfs4_slot *slot)
 791 {
 792         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 793         if (!IS_ERR(task))
 794                 rpc_put_task_async(task);
 795 }
 796
 797 static int nfs41_sequence_process(struct rpc_task *task,
 798                 struct nfs4_sequence_res *res)
 799 {
 800         struct nfs4_session *session;
 801         struct nfs4_slot *slot = res->sr_slot;
 802         struct nfs_client *clp;
 803         int status;
 804         int ret = 1;
 805
 806         if (slot == NULL)
 807                 goto out_noaction;
 808         /* don't increment the sequence number if the task wasn't sent */
 809         if (!RPC_WAS_SENT(task) || slot->seq_done)
 810                 goto out;
 811
 812         session = slot->table->session;
 813         clp = session->clp;
 814
 815         trace_nfs4_sequence_done(session, res);
 816
 817         status = res->sr_status;
 818         if (task->tk_status == -NFS4ERR_DEADSESSION)
 819                 status = -NFS4ERR_DEADSESSION;
 820
 821         /* Check the SEQUENCE operation status */
 822         switch (status) {
 823         case 0:
 824                 /* Mark this sequence number as having been acked */
 825                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 826                 /* Update the slot's sequence and clientid lease timer */
 827                 slot->seq_done = 1;
 828                 do_renew_lease(clp, res->sr_timestamp);
 829                 /* Check sequence flags */
 830                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 831                                 !!slot->privileged);
 832                 nfs41_update_target_slotid(slot->table, slot, res);
 833                 break;
 834         case 1:
 835                 /*
 836                  * sr_status remains 1 if an RPC level error occurred.
 837                  * The server may or may not have processed the sequence
 838                  * operation..
 839                  */
 840                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 841                 slot->seq_done = 1;
 842                 goto out;
 843         case -NFS4ERR_DELAY:
 844                 /* The server detected a resend of the RPC call and
 845                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 846                  * of RFC5661.
 847                  */
 848                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 849                         __func__,
 850                         slot->slot_nr,
 851                         slot->seq_nr);
 852                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 853                 goto out_retry;
 854         case -NFS4ERR_RETRY_UNCACHED_REP:
 855         case -NFS4ERR_SEQ_FALSE_RETRY:
 856                 /*
 857                  * The server thinks we tried to replay a request.
 858                  * Retry the call after bumping the sequence ID.
 859                  */
 860                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 861                 goto retry_new_seq;
 862         case -NFS4ERR_BADSLOT:
 863                 /*
 864                  * The slot id we used was probably retired. Try again
 865                  * using a different slot id.
 866                  */
 867                 if (slot->slot_nr < slot->table->target_highest_slotid)
 868                         goto session_recover;
 869                 goto retry_nowait;
 870         case -NFS4ERR_SEQ_MISORDERED:
 871                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 872                 /*
 873                  * Were one or more calls using this slot interrupted?
 874                  * If the server never received the request, then our
 875                  * transmitted slot sequence number may be too high. However,
 876                  * if the server did receive the request then it might
 877                  * accidentally give us a reply with a mismatched operation.
 878                  * We can sort this out by sending a lone sequence operation
 879                  * to the server on the same slot.
 880                  */
 881                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 882                         slot->seq_nr--;
 883                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 884                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 885                                 res->sr_slot = NULL;
 886                         }
 887                         goto retry_nowait;
 888                 }
 889                 /*
 890                  * RFC5661:
 891                  * A retry might be sent while the original request is
 892                  * still in progress on the replier. The replier SHOULD
 893                  * deal with the issue by returning NFS4ERR_DELAY as the
 894                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 895                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 896                  *
 897                  * Restart the search after a delay.
 898                  */
 899                 slot->seq_nr = slot->seq_nr_highest_sent;
 900                 goto out_retry;
 901         case -NFS4ERR_BADSESSION:
 902         case -NFS4ERR_DEADSESSION:
 903         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 904                 goto session_recover;
 905         default:
 906                 /* Just update the slot sequence no. */
 907                 slot->seq_done = 1;
 908         }
 909 out:
 910         /* The session may be reset by one of the error handlers. */
 911         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 912 out_noaction:
 913         return ret;
 914 session_recover:
 915         nfs4_schedule_session_recovery(session, status);
 916         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 917         nfs41_sequence_free_slot(res);
 918         goto out;
 919 retry_new_seq:
 920         ++slot->seq_nr;
 921 retry_nowait:
 922         if (rpc_restart_call_prepare(task)) {
 923                 nfs41_sequence_free_slot(res);
 924                 task->tk_status = 0;
 925                 ret = 0;
 926         }
 927         goto out;
 928 out_retry:
 929         if (!rpc_restart_call(task))
 930                 goto out;
 931         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 932         return 0;
 933 }
 934
 935 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 936 {
 937         if (!nfs41_sequence_process(task, res))
 938                 return 0;
 939         if (res->sr_slot != NULL)
 940                 nfs41_sequence_free_slot(res);
 941         return 1;
 942
 943 }
 944 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 945
 946 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 947 {
 948         if (res->sr_slot == NULL)
 949                 return 1;
 950         if (res->sr_slot->table->session != NULL)
 951                 return nfs41_sequence_process(task, res);
 952         return nfs40_sequence_done(task, res);
 953 }
 954
 955 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 956 {
 957         if (res->sr_slot != NULL) {
 958                 if (res->sr_slot->table->session != NULL)
 959                         nfs41_sequence_free_slot(res);
 960                 else
 961                         nfs40_sequence_free_slot(res);
 962         }
 963 }
 964
 965 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 966 {
 967         if (res->sr_slot == NULL)
 968                 return 1;
 969         if (!res->sr_slot->table->session)
 970                 return nfs40_sequence_done(task, res);
 971         return nfs41_sequence_done(task, res);
 972 }
 973 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 974
 975 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 976 {
 977         struct nfs4_call_sync_data *data = calldata;
 978
 979         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 980
 981         nfs4_setup_sequence(data->seq_server->nfs_client,
 982                             data->seq_args, data->seq_res, task);
 983 }
 984
 985 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 986 {
 987         struct nfs4_call_sync_data *data = calldata;
 988
 989         nfs41_sequence_done(task, data->seq_res);
 990 }
 991
 992 static const struct rpc_call_ops nfs41_call_sync_ops = {
 993         .rpc_call_prepare = nfs41_call_sync_prepare,
 994         .rpc_call_done = nfs41_call_sync_done,
 995 };
 996
 997 #else   /* !CONFIG_NFS_V4_1 */
 998
 999 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1000 {
1001         return nfs40_sequence_done(task, res);
1002 }
1003
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006         if (res->sr_slot != NULL)
1007                 nfs40_sequence_free_slot(res);
1008 }
1009
1010 int nfs4_sequence_done(struct rpc_task *task,
1011                        struct nfs4_sequence_res *res)
1012 {
1013         return nfs40_sequence_done(task, res);
1014 }
1015 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1016
1017 #endif  /* !CONFIG_NFS_V4_1 */
1018
1019 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1020 {
1021         res->sr_timestamp = jiffies;
1022         res->sr_status_flags = 0;
1023         res->sr_status = 1;
1024 }
1025
1026 static
1027 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1028                 struct nfs4_sequence_res *res,
1029                 struct nfs4_slot *slot)
1030 {
1031         if (!slot)
1032                 return;
1033         slot->privileged = args->sa_privileged ? 1 : 0;
1034         args->sa_slot = slot;
1035
1036         res->sr_slot = slot;
1037 }
1038
1039 int nfs4_setup_sequence(struct nfs_client *client,
1040                         struct nfs4_sequence_args *args,
1041                         struct nfs4_sequence_res *res,
1042                         struct rpc_task *task)
1043 {
1044         struct nfs4_session *session = nfs4_get_session(client);
1045         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1046         struct nfs4_slot *slot;
1047
1048         /* slot already allocated? */
1049         if (res->sr_slot != NULL)
1050                 goto out_start;
1051
1052         if (session)
1053                 tbl = &session->fc_slot_table;
1054
1055         spin_lock(&tbl->slot_tbl_lock);
1056         /* The state manager will wait until the slot table is empty */
1057         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1058                 goto out_sleep;
1059
1060         slot = nfs4_alloc_slot(tbl);
1061         if (IS_ERR(slot)) {
1062                 if (slot == ERR_PTR(-ENOMEM))
1063                         goto out_sleep_timeout;
1064                 goto out_sleep;
1065         }
1066         spin_unlock(&tbl->slot_tbl_lock);
1067
1068         nfs4_sequence_attach_slot(args, res, slot);
1069
1070         trace_nfs4_setup_sequence(session, args);
1071 out_start:
1072         nfs41_sequence_res_init(res);
1073         rpc_call_start(task);
1074         return 0;
1075 out_sleep_timeout:
1076         /* Try again in 1/4 second */
1077         if (args->sa_privileged)
1078                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1079                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1080         else
1081                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1082                                 NULL, jiffies + (HZ >> 2));
1083         spin_unlock(&tbl->slot_tbl_lock);
1084         return -EAGAIN;
1085 out_sleep:
1086         if (args->sa_privileged)
1087                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1088                                 RPC_PRIORITY_PRIVILEGED);
1089         else
1090                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1091         spin_unlock(&tbl->slot_tbl_lock);
1092         return -EAGAIN;
1093 }
1094 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1095
1096 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1097 {
1098         struct nfs4_call_sync_data *data = calldata;
1099         nfs4_setup_sequence(data->seq_server->nfs_client,
1100                                 data->seq_args, data->seq_res, task);
1101 }
1102
1103 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1104 {
1105         struct nfs4_call_sync_data *data = calldata;
1106         nfs4_sequence_done(task, data->seq_res);
1107 }
1108
1109 static const struct rpc_call_ops nfs40_call_sync_ops = {
1110         .rpc_call_prepare = nfs40_call_sync_prepare,
1111         .rpc_call_done = nfs40_call_sync_done,
1112 };
1113
1114 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1115 {
1116         int ret;
1117         struct rpc_task *task;
1118
1119         task = rpc_run_task(task_setup);
1120         if (IS_ERR(task))
1121                 return PTR_ERR(task);
1122
1123         ret = task->tk_status;
1124         rpc_put_task(task);
1125         return ret;
1126 }
1127
1128 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1129                              struct nfs_server *server,
1130                              struct rpc_message *msg,
1131                              struct nfs4_sequence_args *args,
1132                              struct nfs4_sequence_res *res,
1133                              unsigned short task_flags)
1134 {
1135         struct nfs_client *clp = server->nfs_client;
1136         struct nfs4_call_sync_data data = {
1137                 .seq_server = server,
1138                 .seq_args = args,
1139                 .seq_res = res,
1140         };
1141         struct rpc_task_setup task_setup = {
1142                 .rpc_client = clnt,
1143                 .rpc_message = msg,
1144                 .callback_ops = clp->cl_mvops->call_sync_ops,
1145                 .callback_data = &data,
1146                 .flags = task_flags,
1147         };
1148
1149         return nfs4_call_sync_custom(&task_setup);
1150 }
1151
1152 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1153                                    struct nfs_server *server,
1154                                    struct rpc_message *msg,
1155                                    struct nfs4_sequence_args *args,
1156                                    struct nfs4_sequence_res *res)
1157 {
1158         unsigned short task_flags = 0;
1159
1160         if (server->nfs_client->cl_minorversion)
1161                 task_flags = RPC_TASK_MOVEABLE;
1162         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1163 }
1164
1165
1166 int nfs4_call_sync(struct rpc_clnt *clnt,
1167                    struct nfs_server *server,
1168                    struct rpc_message *msg,
1169                    struct nfs4_sequence_args *args,
1170                    struct nfs4_sequence_res *res,
1171                    int cache_reply)
1172 {
1173         nfs4_init_sequence(args, res, cache_reply, 0);
1174         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1175 }
1176
1177 static void
1178 nfs4_inc_nlink_locked(struct inode *inode)
1179 {
1180         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1181                                              NFS_INO_INVALID_CTIME |
1182                                              NFS_INO_INVALID_NLINK);
1183         inc_nlink(inode);
1184 }
1185
1186 static void
1187 nfs4_inc_nlink(struct inode *inode)
1188 {
1189         spin_lock(&inode->i_lock);
1190         nfs4_inc_nlink_locked(inode);
1191         spin_unlock(&inode->i_lock);
1192 }
1193
1194 static void
1195 nfs4_dec_nlink_locked(struct inode *inode)
1196 {
1197         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1198                                              NFS_INO_INVALID_CTIME |
1199                                              NFS_INO_INVALID_NLINK);
1200         drop_nlink(inode);
1201 }
1202
1203 static void
1204 nfs4_update_changeattr_locked(struct inode *inode,
1205                 struct nfs4_change_info *cinfo,
1206                 unsigned long timestamp, unsigned long cache_validity)
1207 {
1208         struct nfs_inode *nfsi = NFS_I(inode);
1209         u64 change_attr = inode_peek_iversion_raw(inode);
1210
1211         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1212         if (S_ISDIR(inode->i_mode))
1213                 cache_validity |= NFS_INO_INVALID_DATA;
1214
1215         switch (NFS_SERVER(inode)->change_attr_type) {
1216         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1217                 if (cinfo->after == change_attr)
1218                         goto out;
1219                 break;
1220         default:
1221                 if ((s64)(change_attr - cinfo->after) >= 0)
1222                         goto out;
1223         }
1224
1225         inode_set_iversion_raw(inode, cinfo->after);
1226         if (!cinfo->atomic || cinfo->before != change_attr) {
1227                 if (S_ISDIR(inode->i_mode))
1228                         nfs_force_lookup_revalidate(inode);
1229
1230                 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1231                         cache_validity |=
1232                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1233                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1234                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1235                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR |
1236                                 NFS_INO_REVAL_PAGECACHE;
1237                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1238         }
1239         nfsi->attrtimeo_timestamp = jiffies;
1240         nfsi->read_cache_jiffies = timestamp;
1241         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1242         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1243 out:
1244         nfs_set_cache_invalid(inode, cache_validity);
1245 }
1246
1247 void
1248 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1249                 unsigned long timestamp, unsigned long cache_validity)
1250 {
1251         spin_lock(&dir->i_lock);
1252         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1253         spin_unlock(&dir->i_lock);
1254 }
1255
1256 struct nfs4_open_createattrs {
1257         struct nfs4_label *label;
1258         struct iattr *sattr;
1259         const __u32 verf[2];
1260 };
1261
1262 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1263                 int err, struct nfs4_exception *exception)
1264 {
1265         if (err != -EINVAL)
1266                 return false;
1267         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1268                 return false;
1269         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1270         exception->retry = 1;
1271         return true;
1272 }
1273
1274 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1275 {
1276          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1277 }
1278
1279 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1280 {
1281         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1282
1283         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1284 }
1285
1286 static u32
1287 nfs4_map_atomic_open_share(struct nfs_server *server,
1288                 fmode_t fmode, int openflags)
1289 {
1290         u32 res = 0;
1291
1292         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1293         case FMODE_READ:
1294                 res = NFS4_SHARE_ACCESS_READ;
1295                 break;
1296         case FMODE_WRITE:
1297                 res = NFS4_SHARE_ACCESS_WRITE;
1298                 break;
1299         case FMODE_READ|FMODE_WRITE:
1300                 res = NFS4_SHARE_ACCESS_BOTH;
1301         }
1302         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1303                 goto out;
1304         /* Want no delegation if we're using O_DIRECT */
1305         if (openflags & O_DIRECT)
1306                 res |= NFS4_SHARE_WANT_NO_DELEG;
1307 out:
1308         return res;
1309 }
1310
1311 static enum open_claim_type4
1312 nfs4_map_atomic_open_claim(struct nfs_server *server,
1313                 enum open_claim_type4 claim)
1314 {
1315         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1316                 return claim;
1317         switch (claim) {
1318         default:
1319                 return claim;
1320         case NFS4_OPEN_CLAIM_FH:
1321                 return NFS4_OPEN_CLAIM_NULL;
1322         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1323                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1324         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1325                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1326         }
1327 }
1328
1329 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1330 {
1331         p->o_res.f_attr = &p->f_attr;
1332         p->o_res.f_label = p->f_label;
1333         p->o_res.seqid = p->o_arg.seqid;
1334         p->c_res.seqid = p->c_arg.seqid;
1335         p->o_res.server = p->o_arg.server;
1336         p->o_res.access_request = p->o_arg.access;
1337         nfs_fattr_init(&p->f_attr);
1338         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1339 }
1340
1341 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1342                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1343                 const struct nfs4_open_createattrs *c,
1344                 enum open_claim_type4 claim,
1345                 gfp_t gfp_mask)
1346 {
1347         struct dentry *parent = dget_parent(dentry);
1348         struct inode *dir = d_inode(parent);
1349         struct nfs_server *server = NFS_SERVER(dir);
1350         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1351         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1352         struct nfs4_opendata *p;
1353
1354         p = kzalloc(sizeof(*p), gfp_mask);
1355         if (p == NULL)
1356                 goto err;
1357
1358         p->f_label = nfs4_label_alloc(server, gfp_mask);
1359         if (IS_ERR(p->f_label))
1360                 goto err_free_p;
1361
1362         p->a_label = nfs4_label_alloc(server, gfp_mask);
1363         if (IS_ERR(p->a_label))
1364                 goto err_free_f;
1365
1366         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1367         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1368         if (IS_ERR(p->o_arg.seqid))
1369                 goto err_free_label;
1370         nfs_sb_active(dentry->d_sb);
1371         p->dentry = dget(dentry);
1372         p->dir = parent;
1373         p->owner = sp;
1374         atomic_inc(&sp->so_count);
1375         p->o_arg.open_flags = flags;
1376         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1377         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1378         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1379                         fmode, flags);
1380         if (flags & O_CREAT) {
1381                 p->o_arg.umask = current_umask();
1382                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1383                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1384                         p->o_arg.u.attrs = &p->attrs;
1385                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1386
1387                         memcpy(p->o_arg.u.verifier.data, c->verf,
1388                                         sizeof(p->o_arg.u.verifier.data));
1389                 }
1390         }
1391         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1392          * will return permission denied for all bits until close */
1393         if (!(flags & O_EXCL)) {
1394                 /* ask server to check for all possible rights as results
1395                  * are cached */
1396                 switch (p->o_arg.claim) {
1397                 default:
1398                         break;
1399                 case NFS4_OPEN_CLAIM_NULL:
1400                 case NFS4_OPEN_CLAIM_FH:
1401                         p->o_arg.access = NFS4_ACCESS_READ |
1402                                 NFS4_ACCESS_MODIFY |
1403                                 NFS4_ACCESS_EXTEND |
1404                                 NFS4_ACCESS_EXECUTE;
1405 #ifdef CONFIG_NFS_V4_2
1406                         if (server->caps & NFS_CAP_XATTR)
1407                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1408                                     NFS4_ACCESS_XAWRITE |
1409                                     NFS4_ACCESS_XALIST;
1410 #endif
1411                 }
1412         }
1413         p->o_arg.clientid = server->nfs_client->cl_clientid;
1414         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1415         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1416         p->o_arg.name = &dentry->d_name;
1417         p->o_arg.server = server;
1418         p->o_arg.bitmask = nfs4_bitmask(server, label);
1419         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1420         switch (p->o_arg.claim) {
1421         case NFS4_OPEN_CLAIM_NULL:
1422         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1423         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1424                 p->o_arg.fh = NFS_FH(dir);
1425                 break;
1426         case NFS4_OPEN_CLAIM_PREVIOUS:
1427         case NFS4_OPEN_CLAIM_FH:
1428         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1429         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1430                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1431         }
1432         p->c_arg.fh = &p->o_res.fh;
1433         p->c_arg.stateid = &p->o_res.stateid;
1434         p->c_arg.seqid = p->o_arg.seqid;
1435         nfs4_init_opendata_res(p);
1436         kref_init(&p->kref);
1437         return p;
1438
1439 err_free_label:
1440         nfs4_label_free(p->a_label);
1441 err_free_f:
1442         nfs4_label_free(p->f_label);
1443 err_free_p:
1444         kfree(p);
1445 err:
1446         dput(parent);
1447         return NULL;
1448 }
1449
1450 static void nfs4_opendata_free(struct kref *kref)
1451 {
1452         struct nfs4_opendata *p = container_of(kref,
1453                         struct nfs4_opendata, kref);
1454         struct super_block *sb = p->dentry->d_sb;
1455
1456         nfs4_lgopen_release(p->lgp);
1457         nfs_free_seqid(p->o_arg.seqid);
1458         nfs4_sequence_free_slot(&p->o_res.seq_res);
1459         if (p->state != NULL)
1460                 nfs4_put_open_state(p->state);
1461         nfs4_put_state_owner(p->owner);
1462
1463         nfs4_label_free(p->a_label);
1464         nfs4_label_free(p->f_label);
1465
1466         dput(p->dir);
1467         dput(p->dentry);
1468         nfs_sb_deactive(sb);
1469         nfs_fattr_free_names(&p->f_attr);
1470         kfree(p->f_attr.mdsthreshold);
1471         kfree(p);
1472 }
1473
1474 static void nfs4_opendata_put(struct nfs4_opendata *p)
1475 {
1476         if (p != NULL)
1477                 kref_put(&p->kref, nfs4_opendata_free);
1478 }
1479
1480 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1481                 fmode_t fmode)
1482 {
1483         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1484         case FMODE_READ|FMODE_WRITE:
1485                 return state->n_rdwr != 0;
1486         case FMODE_WRITE:
1487                 return state->n_wronly != 0;
1488         case FMODE_READ:
1489                 return state->n_rdonly != 0;
1490         }
1491         WARN_ON_ONCE(1);
1492         return false;
1493 }
1494
1495 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1496                 int open_mode, enum open_claim_type4 claim)
1497 {
1498         int ret = 0;
1499
1500         if (open_mode & (O_EXCL|O_TRUNC))
1501                 goto out;
1502         switch (claim) {
1503         case NFS4_OPEN_CLAIM_NULL:
1504         case NFS4_OPEN_CLAIM_FH:
1505                 goto out;
1506         default:
1507                 break;
1508         }
1509         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1510                 case FMODE_READ:
1511                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1512                                 && state->n_rdonly != 0;
1513                         break;
1514                 case FMODE_WRITE:
1515                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1516                                 && state->n_wronly != 0;
1517                         break;
1518                 case FMODE_READ|FMODE_WRITE:
1519                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1520                                 && state->n_rdwr != 0;
1521         }
1522 out:
1523         return ret;
1524 }
1525
1526 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1527                 enum open_claim_type4 claim)
1528 {
1529         if (delegation == NULL)
1530                 return 0;
1531         if ((delegation->type & fmode) != fmode)
1532                 return 0;
1533         switch (claim) {
1534         case NFS4_OPEN_CLAIM_NULL:
1535         case NFS4_OPEN_CLAIM_FH:
1536                 break;
1537         case NFS4_OPEN_CLAIM_PREVIOUS:
1538                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1539                         break;
1540                 fallthrough;
1541         default:
1542                 return 0;
1543         }
1544         nfs_mark_delegation_referenced(delegation);
1545         return 1;
1546 }
1547
1548 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1549 {
1550         switch (fmode) {
1551                 case FMODE_WRITE:
1552                         state->n_wronly++;
1553                         break;
1554                 case FMODE_READ:
1555                         state->n_rdonly++;
1556                         break;
1557                 case FMODE_READ|FMODE_WRITE:
1558                         state->n_rdwr++;
1559         }
1560         nfs4_state_set_mode_locked(state, state->state | fmode);
1561 }
1562
1563 #ifdef CONFIG_NFS_V4_1
1564 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1565 {
1566         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1567                 return true;
1568         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1569                 return true;
1570         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1571                 return true;
1572         return false;
1573 }
1574 #endif /* CONFIG_NFS_V4_1 */
1575
1576 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1577 {
1578         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1579                 wake_up_all(&state->waitq);
1580 }
1581
1582 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1583 {
1584         struct nfs_client *clp = state->owner->so_server->nfs_client;
1585         bool need_recover = false;
1586
1587         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1588                 need_recover = true;
1589         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1590                 need_recover = true;
1591         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1592                 need_recover = true;
1593         if (need_recover)
1594                 nfs4_state_mark_reclaim_nograce(clp, state);
1595 }
1596
1597 /*
1598  * Check for whether or not the caller may update the open stateid
1599  * to the value passed in by stateid.
1600  *
1601  * Note: This function relies heavily on the server implementing
1602  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1603  * correctly.
1604  * i.e. The stateid seqids have to be initialised to 1, and
1605  * are then incremented on every state transition.
1606  */
1607 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1608                 const nfs4_stateid *stateid)
1609 {
1610         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1611                 /* The common case - we're updating to a new sequence number */
1612                 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1613                         nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1614                         return true;
1615                 }
1616         } else {
1617                 /* This is the first OPEN in this generation */
1618                 if (stateid->seqid == cpu_to_be32(1))
1619                         return true;
1620         }
1621         return false;
1622 }
1623
1624 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1625 {
1626         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1627                 return;
1628         if (state->n_wronly)
1629                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1630         if (state->n_rdonly)
1631                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1632         if (state->n_rdwr)
1633                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1634         set_bit(NFS_OPEN_STATE, &state->flags);
1635 }
1636
1637 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1638                 nfs4_stateid *stateid, fmode_t fmode)
1639 {
1640         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1641         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1642         case FMODE_WRITE:
1643                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1644                 break;
1645         case FMODE_READ:
1646                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1647                 break;
1648         case 0:
1649                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1650                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1651                 clear_bit(NFS_OPEN_STATE, &state->flags);
1652         }
1653         if (stateid == NULL)
1654                 return;
1655         /* Handle OPEN+OPEN_DOWNGRADE races */
1656         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1657             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1658                 nfs_resync_open_stateid_locked(state);
1659                 goto out;
1660         }
1661         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1662                 nfs4_stateid_copy(&state->stateid, stateid);
1663         nfs4_stateid_copy(&state->open_stateid, stateid);
1664         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1665 out:
1666         nfs_state_log_update_open_stateid(state);
1667 }
1668
1669 static void nfs_clear_open_stateid(struct nfs4_state *state,
1670         nfs4_stateid *arg_stateid,
1671         nfs4_stateid *stateid, fmode_t fmode)
1672 {
1673         write_seqlock(&state->seqlock);
1674         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1675         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1676                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1677         write_sequnlock(&state->seqlock);
1678         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1679                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1680 }
1681
1682 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1683                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1684         __must_hold(&state->owner->so_lock)
1685         __must_hold(&state->seqlock)
1686         __must_hold(RCU)
1687
1688 {
1689         DEFINE_WAIT(wait);
1690         int status = 0;
1691         for (;;) {
1692
1693                 if (nfs_stateid_is_sequential(state, stateid))
1694                         break;
1695
1696                 if (status)
1697                         break;
1698                 /* Rely on seqids for serialisation with NFSv4.0 */
1699                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1700                         break;
1701
1702                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1703                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1704                 /*
1705                  * Ensure we process the state changes in the same order
1706                  * in which the server processed them by delaying the
1707                  * update of the stateid until we are in sequence.
1708                  */
1709                 write_sequnlock(&state->seqlock);
1710                 spin_unlock(&state->owner->so_lock);
1711                 rcu_read_unlock();
1712                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1713
1714                 if (!fatal_signal_pending(current)) {
1715                         if (schedule_timeout(5*HZ) == 0)
1716                                 status = -EAGAIN;
1717                         else
1718                                 status = 0;
1719                 } else
1720                         status = -EINTR;
1721                 finish_wait(&state->waitq, &wait);
1722                 rcu_read_lock();
1723                 spin_lock(&state->owner->so_lock);
1724                 write_seqlock(&state->seqlock);
1725         }
1726
1727         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1728             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1729                 nfs4_stateid_copy(freeme, &state->open_stateid);
1730                 nfs_test_and_clear_all_open_stateid(state);
1731         }
1732
1733         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1734                 nfs4_stateid_copy(&state->stateid, stateid);
1735         nfs4_stateid_copy(&state->open_stateid, stateid);
1736         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1737         nfs_state_log_update_open_stateid(state);
1738 }
1739
1740 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1741                 const nfs4_stateid *open_stateid,
1742                 fmode_t fmode,
1743                 nfs4_stateid *freeme)
1744 {
1745         /*
1746          * Protect the call to nfs4_state_set_mode_locked and
1747          * serialise the stateid update
1748          */
1749         write_seqlock(&state->seqlock);
1750         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1751         switch (fmode) {
1752         case FMODE_READ:
1753                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1754                 break;
1755         case FMODE_WRITE:
1756                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1757                 break;
1758         case FMODE_READ|FMODE_WRITE:
1759                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1760         }
1761         set_bit(NFS_OPEN_STATE, &state->flags);
1762         write_sequnlock(&state->seqlock);
1763 }
1764
1765 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1766 {
1767         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1768         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1769         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1770         clear_bit(NFS_OPEN_STATE, &state->flags);
1771 }
1772
1773 static void nfs_state_set_delegation(struct nfs4_state *state,
1774                 const nfs4_stateid *deleg_stateid,
1775                 fmode_t fmode)
1776 {
1777         /*
1778          * Protect the call to nfs4_state_set_mode_locked and
1779          * serialise the stateid update
1780          */
1781         write_seqlock(&state->seqlock);
1782         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1783         set_bit(NFS_DELEGATED_STATE, &state->flags);
1784         write_sequnlock(&state->seqlock);
1785 }
1786
1787 static void nfs_state_clear_delegation(struct nfs4_state *state)
1788 {
1789         write_seqlock(&state->seqlock);
1790         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1791         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1792         write_sequnlock(&state->seqlock);
1793 }
1794
1795 int update_open_stateid(struct nfs4_state *state,
1796                 const nfs4_stateid *open_stateid,
1797                 const nfs4_stateid *delegation,
1798                 fmode_t fmode)
1799 {
1800         struct nfs_server *server = NFS_SERVER(state->inode);
1801         struct nfs_client *clp = server->nfs_client;
1802         struct nfs_inode *nfsi = NFS_I(state->inode);
1803         struct nfs_delegation *deleg_cur;
1804         nfs4_stateid freeme = { };
1805         int ret = 0;
1806
1807         fmode &= (FMODE_READ|FMODE_WRITE);
1808
1809         rcu_read_lock();
1810         spin_lock(&state->owner->so_lock);
1811         if (open_stateid != NULL) {
1812                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1813                 ret = 1;
1814         }
1815
1816         deleg_cur = nfs4_get_valid_delegation(state->inode);
1817         if (deleg_cur == NULL)
1818                 goto no_delegation;
1819
1820         spin_lock(&deleg_cur->lock);
1821         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1822            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1823             (deleg_cur->type & fmode) != fmode)
1824                 goto no_delegation_unlock;
1825
1826         if (delegation == NULL)
1827                 delegation = &deleg_cur->stateid;
1828         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1829                 goto no_delegation_unlock;
1830
1831         nfs_mark_delegation_referenced(deleg_cur);
1832         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1833         ret = 1;
1834 no_delegation_unlock:
1835         spin_unlock(&deleg_cur->lock);
1836 no_delegation:
1837         if (ret)
1838                 update_open_stateflags(state, fmode);
1839         spin_unlock(&state->owner->so_lock);
1840         rcu_read_unlock();
1841
1842         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1843                 nfs4_schedule_state_manager(clp);
1844         if (freeme.type != 0)
1845                 nfs4_test_and_free_stateid(server, &freeme,
1846                                 state->owner->so_cred);
1847
1848         return ret;
1849 }
1850
1851 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1852                 const nfs4_stateid *stateid)
1853 {
1854         struct nfs4_state *state = lsp->ls_state;
1855         bool ret = false;
1856
1857         spin_lock(&state->state_lock);
1858         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1859                 goto out_noupdate;
1860         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1861                 goto out_noupdate;
1862         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1863         ret = true;
1864 out_noupdate:
1865         spin_unlock(&state->state_lock);
1866         return ret;
1867 }
1868
1869 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1870 {
1871         struct nfs_delegation *delegation;
1872
1873         fmode &= FMODE_READ|FMODE_WRITE;
1874         rcu_read_lock();
1875         delegation = nfs4_get_valid_delegation(inode);
1876         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1877                 rcu_read_unlock();
1878                 return;
1879         }
1880         rcu_read_unlock();
1881         nfs4_inode_return_delegation(inode);
1882 }
1883
1884 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1885 {
1886         struct nfs4_state *state = opendata->state;
1887         struct nfs_delegation *delegation;
1888         int open_mode = opendata->o_arg.open_flags;
1889         fmode_t fmode = opendata->o_arg.fmode;
1890         enum open_claim_type4 claim = opendata->o_arg.claim;
1891         nfs4_stateid stateid;
1892         int ret = -EAGAIN;
1893
1894         for (;;) {
1895                 spin_lock(&state->owner->so_lock);
1896                 if (can_open_cached(state, fmode, open_mode, claim)) {
1897                         update_open_stateflags(state, fmode);
1898                         spin_unlock(&state->owner->so_lock);
1899                         goto out_return_state;
1900                 }
1901                 spin_unlock(&state->owner->so_lock);
1902                 rcu_read_lock();
1903                 delegation = nfs4_get_valid_delegation(state->inode);
1904                 if (!can_open_delegated(delegation, fmode, claim)) {
1905                         rcu_read_unlock();
1906                         break;
1907                 }
1908                 /* Save the delegation */
1909                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1910                 rcu_read_unlock();
1911                 nfs_release_seqid(opendata->o_arg.seqid);
1912                 if (!opendata->is_recover) {
1913                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1914                         if (ret != 0)
1915                                 goto out;
1916                 }
1917                 ret = -EAGAIN;
1918
1919                 /* Try to update the stateid using the delegation */
1920                 if (update_open_stateid(state, NULL, &stateid, fmode))
1921                         goto out_return_state;
1922         }
1923 out:
1924         return ERR_PTR(ret);
1925 out_return_state:
1926         refcount_inc(&state->count);
1927         return state;
1928 }
1929
1930 static void
1931 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1932 {
1933         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1934         struct nfs_delegation *delegation;
1935         int delegation_flags = 0;
1936
1937         rcu_read_lock();
1938         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1939         if (delegation)
1940                 delegation_flags = delegation->flags;
1941         rcu_read_unlock();
1942         switch (data->o_arg.claim) {
1943         default:
1944                 break;
1945         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1946         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1947                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1948                                    "returning a delegation for "
1949                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1950                                    clp->cl_hostname);
1951                 return;
1952         }
1953         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1954                 nfs_inode_set_delegation(state->inode,
1955                                 data->owner->so_cred,
1956                                 data->o_res.delegation_type,
1957                                 &data->o_res.delegation,
1958                                 data->o_res.pagemod_limit);
1959         else
1960                 nfs_inode_reclaim_delegation(state->inode,
1961                                 data->owner->so_cred,
1962                                 data->o_res.delegation_type,
1963                                 &data->o_res.delegation,
1964                                 data->o_res.pagemod_limit);
1965
1966         if (data->o_res.do_recall)
1967                 nfs_async_inode_return_delegation(state->inode,
1968                                                   &data->o_res.delegation);
1969 }
1970
1971 /*
1972  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1973  * and update the nfs4_state.
1974  */
1975 static struct nfs4_state *
1976 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1977 {
1978         struct inode *inode = data->state->inode;
1979         struct nfs4_state *state = data->state;
1980         int ret;
1981
1982         if (!data->rpc_done) {
1983                 if (data->rpc_status)
1984                         return ERR_PTR(data->rpc_status);
1985                 /* cached opens have already been processed */
1986                 goto update;
1987         }
1988
1989         ret = nfs_refresh_inode(inode, &data->f_attr);
1990         if (ret)
1991                 return ERR_PTR(ret);
1992
1993         if (data->o_res.delegation_type != 0)
1994                 nfs4_opendata_check_deleg(data, state);
1995 update:
1996         if (!update_open_stateid(state, &data->o_res.stateid,
1997                                 NULL, data->o_arg.fmode))
1998                 return ERR_PTR(-EAGAIN);
1999         refcount_inc(&state->count);
2000
2001         return state;
2002 }
2003
2004 static struct inode *
2005 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2006 {
2007         struct inode *inode;
2008
2009         switch (data->o_arg.claim) {
2010         case NFS4_OPEN_CLAIM_NULL:
2011         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2012         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2013                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2014                         return ERR_PTR(-EAGAIN);
2015                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2016                                 &data->f_attr, data->f_label);
2017                 break;
2018         default:
2019                 inode = d_inode(data->dentry);
2020                 ihold(inode);
2021                 nfs_refresh_inode(inode, &data->f_attr);
2022         }
2023         return inode;
2024 }
2025
2026 static struct nfs4_state *
2027 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2028 {
2029         struct nfs4_state *state;
2030         struct inode *inode;
2031
2032         inode = nfs4_opendata_get_inode(data);
2033         if (IS_ERR(inode))
2034                 return ERR_CAST(inode);
2035         if (data->state != NULL && data->state->inode == inode) {
2036                 state = data->state;
2037                 refcount_inc(&state->count);
2038         } else
2039                 state = nfs4_get_open_state(inode, data->owner);
2040         iput(inode);
2041         if (state == NULL)
2042                 state = ERR_PTR(-ENOMEM);
2043         return state;
2044 }
2045
2046 static struct nfs4_state *
2047 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2048 {
2049         struct nfs4_state *state;
2050
2051         if (!data->rpc_done) {
2052                 state = nfs4_try_open_cached(data);
2053                 trace_nfs4_cached_open(data->state);
2054                 goto out;
2055         }
2056
2057         state = nfs4_opendata_find_nfs4_state(data);
2058         if (IS_ERR(state))
2059                 goto out;
2060
2061         if (data->o_res.delegation_type != 0)
2062                 nfs4_opendata_check_deleg(data, state);
2063         if (!update_open_stateid(state, &data->o_res.stateid,
2064                                 NULL, data->o_arg.fmode)) {
2065                 nfs4_put_open_state(state);
2066                 state = ERR_PTR(-EAGAIN);
2067         }
2068 out:
2069         nfs_release_seqid(data->o_arg.seqid);
2070         return state;
2071 }
2072
2073 static struct nfs4_state *
2074 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2075 {
2076         struct nfs4_state *ret;
2077
2078         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2079                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2080         else
2081                 ret = _nfs4_opendata_to_nfs4_state(data);
2082         nfs4_sequence_free_slot(&data->o_res.seq_res);
2083         return ret;
2084 }
2085
2086 static struct nfs_open_context *
2087 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2088 {
2089         struct nfs_inode *nfsi = NFS_I(state->inode);
2090         struct nfs_open_context *ctx;
2091
2092         rcu_read_lock();
2093         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2094                 if (ctx->state != state)
2095                         continue;
2096                 if ((ctx->mode & mode) != mode)
2097                         continue;
2098                 if (!get_nfs_open_context(ctx))
2099                         continue;
2100                 rcu_read_unlock();
2101                 return ctx;
2102         }
2103         rcu_read_unlock();
2104         return ERR_PTR(-ENOENT);
2105 }
2106
2107 static struct nfs_open_context *
2108 nfs4_state_find_open_context(struct nfs4_state *state)
2109 {
2110         struct nfs_open_context *ctx;
2111
2112         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2113         if (!IS_ERR(ctx))
2114                 return ctx;
2115         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2116         if (!IS_ERR(ctx))
2117                 return ctx;
2118         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2119 }
2120
2121 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2122                 struct nfs4_state *state, enum open_claim_type4 claim)
2123 {
2124         struct nfs4_opendata *opendata;
2125
2126         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2127                         NULL, claim, GFP_NOFS);
2128         if (opendata == NULL)
2129                 return ERR_PTR(-ENOMEM);
2130         opendata->state = state;
2131         refcount_inc(&state->count);
2132         return opendata;
2133 }
2134
2135 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2136                 fmode_t fmode)
2137 {
2138         struct nfs4_state *newstate;
2139         int ret;
2140
2141         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2142                 return 0;
2143         opendata->o_arg.open_flags = 0;
2144         opendata->o_arg.fmode = fmode;
2145         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2146                         NFS_SB(opendata->dentry->d_sb),
2147                         fmode, 0);
2148         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2149         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2150         nfs4_init_opendata_res(opendata);
2151         ret = _nfs4_recover_proc_open(opendata);
2152         if (ret != 0)
2153                 return ret;
2154         newstate = nfs4_opendata_to_nfs4_state(opendata);
2155         if (IS_ERR(newstate))
2156                 return PTR_ERR(newstate);
2157         if (newstate != opendata->state)
2158                 ret = -ESTALE;
2159         nfs4_close_state(newstate, fmode);
2160         return ret;
2161 }
2162
2163 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2164 {
2165         int ret;
2166
2167         /* memory barrier prior to reading state->n_* */
2168         smp_rmb();
2169         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2170         if (ret != 0)
2171                 return ret;
2172         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2173         if (ret != 0)
2174                 return ret;
2175         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2176         if (ret != 0)
2177                 return ret;
2178         /*
2179          * We may have performed cached opens for all three recoveries.
2180          * Check if we need to update the current stateid.
2181          */
2182         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2183             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2184                 write_seqlock(&state->seqlock);
2185                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2186                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2187                 write_sequnlock(&state->seqlock);
2188         }
2189         return 0;
2190 }
2191
2192 /*
2193  * OPEN_RECLAIM:
2194  *      reclaim state on the server after a reboot.
2195  */
2196 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2197 {
2198         struct nfs_delegation *delegation;
2199         struct nfs4_opendata *opendata;
2200         fmode_t delegation_type = 0;
2201         int status;
2202
2203         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2204                         NFS4_OPEN_CLAIM_PREVIOUS);
2205         if (IS_ERR(opendata))
2206                 return PTR_ERR(opendata);
2207         rcu_read_lock();
2208         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2209         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2210                 delegation_type = delegation->type;
2211         rcu_read_unlock();
2212         opendata->o_arg.u.delegation_type = delegation_type;
2213         status = nfs4_open_recover(opendata, state);
2214         nfs4_opendata_put(opendata);
2215         return status;
2216 }
2217
2218 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2219 {
2220         struct nfs_server *server = NFS_SERVER(state->inode);
2221         struct nfs4_exception exception = { };
2222         int err;
2223         do {
2224                 err = _nfs4_do_open_reclaim(ctx, state);
2225                 trace_nfs4_open_reclaim(ctx, 0, err);
2226                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2227                         continue;
2228                 if (err != -NFS4ERR_DELAY)
2229                         break;
2230                 nfs4_handle_exception(server, err, &exception);
2231         } while (exception.retry);
2232         return err;
2233 }
2234
2235 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2236 {
2237         struct nfs_open_context *ctx;
2238         int ret;
2239
2240         ctx = nfs4_state_find_open_context(state);
2241         if (IS_ERR(ctx))
2242                 return -EAGAIN;
2243         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2244         nfs_state_clear_open_state_flags(state);
2245         ret = nfs4_do_open_reclaim(ctx, state);
2246         put_nfs_open_context(ctx);
2247         return ret;
2248 }
2249
2250 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)
2251 {
2252         switch (err) {
2253                 default:
2254                         printk(KERN_ERR "NFS: %s: unhandled error "
2255                                         "%d.\n", __func__, err);
2256                         fallthrough;
2257                 case 0:
2258                 case -ENOENT:
2259                 case -EAGAIN:
2260                 case -ESTALE:
2261                 case -ETIMEDOUT:
2262                         break;
2263                 case -NFS4ERR_BADSESSION:
2264                 case -NFS4ERR_BADSLOT:
2265                 case -NFS4ERR_BAD_HIGH_SLOT:
2266                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2267                 case -NFS4ERR_DEADSESSION:
2268                         return -EAGAIN;
2269                 case -NFS4ERR_STALE_CLIENTID:
2270                 case -NFS4ERR_STALE_STATEID:
2271                         /* Don't recall a delegation if it was lost */
2272                         nfs4_schedule_lease_recovery(server->nfs_client);
2273                         return -EAGAIN;
2274                 case -NFS4ERR_MOVED:
2275                         nfs4_schedule_migration_recovery(server);
2276                         return -EAGAIN;
2277                 case -NFS4ERR_LEASE_MOVED:
2278                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2279                         return -EAGAIN;
2280                 case -NFS4ERR_DELEG_REVOKED:
2281                 case -NFS4ERR_ADMIN_REVOKED:
2282                 case -NFS4ERR_EXPIRED:
2283                 case -NFS4ERR_BAD_STATEID:
2284                 case -NFS4ERR_OPENMODE:
2285                         nfs_inode_find_state_and_recover(state->inode,
2286                                         stateid);
2287                         nfs4_schedule_stateid_recovery(server, state);
2288                         return -EAGAIN;
2289                 case -NFS4ERR_DELAY:
2290                 case -NFS4ERR_GRACE:
2291                         ssleep(1);
2292                         return -EAGAIN;
2293                 case -ENOMEM:
2294                 case -NFS4ERR_DENIED:
2295                         if (fl) {
2296                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2297                                 if (lsp)
2298                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2299                         }
2300                         return 0;
2301         }
2302         return err;
2303 }
2304
2305 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2306                 struct nfs4_state *state, const nfs4_stateid *stateid)
2307 {
2308         struct nfs_server *server = NFS_SERVER(state->inode);
2309         struct nfs4_opendata *opendata;
2310         int err = 0;
2311
2312         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2313                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2314         if (IS_ERR(opendata))
2315                 return PTR_ERR(opendata);
2316         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2317         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2318                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2319                 if (err)
2320                         goto out;
2321         }
2322         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2323                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2324                 if (err)
2325                         goto out;
2326         }
2327         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2328                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2329                 if (err)
2330                         goto out;
2331         }
2332         nfs_state_clear_delegation(state);
2333 out:
2334         nfs4_opendata_put(opendata);
2335         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2336 }
2337
2338 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2339 {
2340         struct nfs4_opendata *data = calldata;
2341
2342         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2343                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2344 }
2345
2346 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2347 {
2348         struct nfs4_opendata *data = calldata;
2349
2350         nfs40_sequence_done(task, &data->c_res.seq_res);
2351
2352         data->rpc_status = task->tk_status;
2353         if (data->rpc_status == 0) {
2354                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2355                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2356                 renew_lease(data->o_res.server, data->timestamp);
2357                 data->rpc_done = true;
2358         }
2359 }
2360
2361 static void nfs4_open_confirm_release(void *calldata)
2362 {
2363         struct nfs4_opendata *data = calldata;
2364         struct nfs4_state *state = NULL;
2365
2366         /* If this request hasn't been cancelled, do nothing */
2367         if (!data->cancelled)
2368                 goto out_free;
2369         /* In case of error, no cleanup! */
2370         if (!data->rpc_done)
2371                 goto out_free;
2372         state = nfs4_opendata_to_nfs4_state(data);
2373         if (!IS_ERR(state))
2374                 nfs4_close_state(state, data->o_arg.fmode);
2375 out_free:
2376         nfs4_opendata_put(data);
2377 }
2378
2379 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2380         .rpc_call_prepare = nfs4_open_confirm_prepare,
2381         .rpc_call_done = nfs4_open_confirm_done,
2382         .rpc_release = nfs4_open_confirm_release,
2383 };
2384
2385 /*
2386  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2387  */
2388 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2389 {
2390         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2391         struct rpc_task *task;
2392         struct  rpc_message msg = {
2393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2394                 .rpc_argp = &data->c_arg,
2395                 .rpc_resp = &data->c_res,
2396                 .rpc_cred = data->owner->so_cred,
2397         };
2398         struct rpc_task_setup task_setup_data = {
2399                 .rpc_client = server->client,
2400                 .rpc_message = &msg,
2401                 .callback_ops = &nfs4_open_confirm_ops,
2402                 .callback_data = data,
2403                 .workqueue = nfsiod_workqueue,
2404                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2405         };
2406         int status;
2407
2408         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2409                                 data->is_recover);
2410         kref_get(&data->kref);
2411         data->rpc_done = false;
2412         data->rpc_status = 0;
2413         data->timestamp = jiffies;
2414         task = rpc_run_task(&task_setup_data);
2415         if (IS_ERR(task))
2416                 return PTR_ERR(task);
2417         status = rpc_wait_for_completion_task(task);
2418         if (status != 0) {
2419                 data->cancelled = true;
2420                 smp_wmb();
2421         } else
2422                 status = data->rpc_status;
2423         rpc_put_task(task);
2424         return status;
2425 }
2426
2427 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2428 {
2429         struct nfs4_opendata *data = calldata;
2430         struct nfs4_state_owner *sp = data->owner;
2431         struct nfs_client *clp = sp->so_server->nfs_client;
2432         enum open_claim_type4 claim = data->o_arg.claim;
2433
2434         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2435                 goto out_wait;
2436         /*
2437          * Check if we still need to send an OPEN call, or if we can use
2438          * a delegation instead.
2439          */
2440         if (data->state != NULL) {
2441                 struct nfs_delegation *delegation;
2442
2443                 if (can_open_cached(data->state, data->o_arg.fmode,
2444                                         data->o_arg.open_flags, claim))
2445                         goto out_no_action;
2446                 rcu_read_lock();
2447                 delegation = nfs4_get_valid_delegation(data->state->inode);
2448                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2449                         goto unlock_no_action;
2450                 rcu_read_unlock();
2451         }
2452         /* Update client id. */
2453         data->o_arg.clientid = clp->cl_clientid;
2454         switch (claim) {
2455         default:
2456                 break;
2457         case NFS4_OPEN_CLAIM_PREVIOUS:
2458         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2459         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2460                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2461                 fallthrough;
2462         case NFS4_OPEN_CLAIM_FH:
2463                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2464         }
2465         data->timestamp = jiffies;
2466         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2467                                 &data->o_arg.seq_args,
2468                                 &data->o_res.seq_res,
2469                                 task) != 0)
2470                 nfs_release_seqid(data->o_arg.seqid);
2471
2472         /* Set the create mode (note dependency on the session type) */
2473         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2474         if (data->o_arg.open_flags & O_EXCL) {
2475                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2476                 if (nfs4_has_persistent_session(clp))
2477                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2478                 else if (clp->cl_mvops->minor_version > 0)
2479                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2480         }
2481         return;
2482 unlock_no_action:
2483         trace_nfs4_cached_open(data->state);
2484         rcu_read_unlock();
2485 out_no_action:
2486         task->tk_action = NULL;
2487 out_wait:
2488         nfs4_sequence_done(task, &data->o_res.seq_res);
2489 }
2490
2491 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2492 {
2493         struct nfs4_opendata *data = calldata;
2494
2495         data->rpc_status = task->tk_status;
2496
2497         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2498                 return;
2499
2500         if (task->tk_status == 0) {
2501                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2502                         switch (data->o_res.f_attr->mode & S_IFMT) {
2503                         case S_IFREG:
2504                                 break;
2505                         case S_IFLNK:
2506                                 data->rpc_status = -ELOOP;
2507                                 break;
2508                         case S_IFDIR:
2509                                 data->rpc_status = -EISDIR;
2510                                 break;
2511                         default:
2512                                 data->rpc_status = -ENOTDIR;
2513                         }
2514                 }
2515                 renew_lease(data->o_res.server, data->timestamp);
2516                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2517                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2518         }
2519         data->rpc_done = true;
2520 }
2521
2522 static void nfs4_open_release(void *calldata)
2523 {
2524         struct nfs4_opendata *data = calldata;
2525         struct nfs4_state *state = NULL;
2526
2527         /* If this request hasn't been cancelled, do nothing */
2528         if (!data->cancelled)
2529                 goto out_free;
2530         /* In case of error, no cleanup! */
2531         if (data->rpc_status != 0 || !data->rpc_done)
2532                 goto out_free;
2533         /* In case we need an open_confirm, no cleanup! */
2534         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2535                 goto out_free;
2536         state = nfs4_opendata_to_nfs4_state(data);
2537         if (!IS_ERR(state))
2538                 nfs4_close_state(state, data->o_arg.fmode);
2539 out_free:
2540         nfs4_opendata_put(data);
2541 }
2542
2543 static const struct rpc_call_ops nfs4_open_ops = {
2544         .rpc_call_prepare = nfs4_open_prepare,
2545         .rpc_call_done = nfs4_open_done,
2546         .rpc_release = nfs4_open_release,
2547 };
2548
2549 static int nfs4_run_open_task(struct nfs4_opendata *data,
2550                               struct nfs_open_context *ctx)
2551 {
2552         struct inode *dir = d_inode(data->dir);
2553         struct nfs_server *server = NFS_SERVER(dir);
2554         struct nfs_openargs *o_arg = &data->o_arg;
2555         struct nfs_openres *o_res = &data->o_res;
2556         struct rpc_task *task;
2557         struct rpc_message msg = {
2558                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2559                 .rpc_argp = o_arg,
2560                 .rpc_resp = o_res,
2561                 .rpc_cred = data->owner->so_cred,
2562         };
2563         struct rpc_task_setup task_setup_data = {
2564                 .rpc_client = server->client,
2565                 .rpc_message = &msg,
2566                 .callback_ops = &nfs4_open_ops,
2567                 .callback_data = data,
2568                 .workqueue = nfsiod_workqueue,
2569                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2570         };
2571         int status;
2572
2573         if (server->nfs_client->cl_minorversion)
2574                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2575
2576         kref_get(&data->kref);
2577         data->rpc_done = false;
2578         data->rpc_status = 0;
2579         data->cancelled = false;
2580         data->is_recover = false;
2581         if (!ctx) {
2582                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2583                 data->is_recover = true;
2584                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2585         } else {
2586                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2587                 pnfs_lgopen_prepare(data, ctx);
2588         }
2589         task = rpc_run_task(&task_setup_data);
2590         if (IS_ERR(task))
2591                 return PTR_ERR(task);
2592         status = rpc_wait_for_completion_task(task);
2593         if (status != 0) {
2594                 data->cancelled = true;
2595                 smp_wmb();
2596         } else
2597                 status = data->rpc_status;
2598         rpc_put_task(task);
2599
2600         return status;
2601 }
2602
2603 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2604 {
2605         struct inode *dir = d_inode(data->dir);
2606         struct nfs_openres *o_res = &data->o_res;
2607         int status;
2608
2609         status = nfs4_run_open_task(data, NULL);
2610         if (status != 0 || !data->rpc_done)
2611                 return status;
2612
2613         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2614
2615         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2616                 status = _nfs4_proc_open_confirm(data);
2617
2618         return status;
2619 }
2620
2621 /*
2622  * Additional permission checks in order to distinguish between an
2623  * open for read, and an open for execute. This works around the
2624  * fact that NFSv4 OPEN treats read and execute permissions as being
2625  * the same.
2626  * Note that in the non-execute case, we want to turn off permission
2627  * checking if we just created a new file (POSIX open() semantics).
2628  */
2629 static int nfs4_opendata_access(const struct cred *cred,
2630                                 struct nfs4_opendata *opendata,
2631                                 struct nfs4_state *state, fmode_t fmode,
2632                                 int openflags)
2633 {
2634         struct nfs_access_entry cache;
2635         u32 mask, flags;
2636
2637         /* access call failed or for some reason the server doesn't
2638          * support any access modes -- defer access call until later */
2639         if (opendata->o_res.access_supported == 0)
2640                 return 0;
2641
2642         mask = 0;
2643         /*
2644          * Use openflags to check for exec, because fmode won't
2645          * always have FMODE_EXEC set when file open for exec.
2646          */
2647         if (openflags & __FMODE_EXEC) {
2648                 /* ONLY check for exec rights */
2649                 if (S_ISDIR(state->inode->i_mode))
2650                         mask = NFS4_ACCESS_LOOKUP;
2651                 else
2652                         mask = NFS4_ACCESS_EXECUTE;
2653         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2654                 mask = NFS4_ACCESS_READ;
2655
2656         cache.cred = cred;
2657         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2658         nfs_access_add_cache(state->inode, &cache);
2659
2660         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2661         if ((mask & ~cache.mask & flags) == 0)
2662                 return 0;
2663
2664         return -EACCES;
2665 }
2666
2667 /*
2668  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2669  */
2670 static int _nfs4_proc_open(struct nfs4_opendata *data,
2671                            struct nfs_open_context *ctx)
2672 {
2673         struct inode *dir = d_inode(data->dir);
2674         struct nfs_server *server = NFS_SERVER(dir);
2675         struct nfs_openargs *o_arg = &data->o_arg;
2676         struct nfs_openres *o_res = &data->o_res;
2677         int status;
2678
2679         status = nfs4_run_open_task(data, ctx);
2680         if (!data->rpc_done)
2681                 return status;
2682         if (status != 0) {
2683                 if (status == -NFS4ERR_BADNAME &&
2684                                 !(o_arg->open_flags & O_CREAT))
2685                         return -ENOENT;
2686                 return status;
2687         }
2688
2689         nfs_fattr_map_and_free_names(server, &data->f_attr);
2690
2691         if (o_arg->open_flags & O_CREAT) {
2692                 if (o_arg->open_flags & O_EXCL)
2693                         data->file_created = true;
2694                 else if (o_res->cinfo.before != o_res->cinfo.after)
2695                         data->file_created = true;
2696                 if (data->file_created ||
2697                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2698                         nfs4_update_changeattr(dir, &o_res->cinfo,
2699                                         o_res->f_attr->time_start,
2700                                         NFS_INO_INVALID_DATA);
2701         }
2702         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2703                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2704         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2705                 status = _nfs4_proc_open_confirm(data);
2706                 if (status != 0)
2707                         return status;
2708         }
2709         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2710                 nfs4_sequence_free_slot(&o_res->seq_res);
2711                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2712                                 o_res->f_label, NULL);
2713         }
2714         return 0;
2715 }
2716
2717 /*
2718  * OPEN_EXPIRED:
2719  *      reclaim state on the server after a network partition.
2720  *      Assumes caller holds the appropriate lock
2721  */
2722 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2723 {
2724         struct nfs4_opendata *opendata;
2725         int ret;
2726
2727         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2728                         NFS4_OPEN_CLAIM_FH);
2729         if (IS_ERR(opendata))
2730                 return PTR_ERR(opendata);
2731         ret = nfs4_open_recover(opendata, state);
2732         if (ret == -ESTALE)
2733                 d_drop(ctx->dentry);
2734         nfs4_opendata_put(opendata);
2735         return ret;
2736 }
2737
2738 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2739 {
2740         struct nfs_server *server = NFS_SERVER(state->inode);
2741         struct nfs4_exception exception = { };
2742         int err;
2743
2744         do {
2745                 err = _nfs4_open_expired(ctx, state);
2746                 trace_nfs4_open_expired(ctx, 0, err);
2747                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2748                         continue;
2749                 switch (err) {
2750                 default:
2751                         goto out;
2752                 case -NFS4ERR_GRACE:
2753                 case -NFS4ERR_DELAY:
2754                         nfs4_handle_exception(server, err, &exception);
2755                         err = 0;
2756                 }
2757         } while (exception.retry);
2758 out:
2759         return err;
2760 }
2761
2762 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2763 {
2764         struct nfs_open_context *ctx;
2765         int ret;
2766
2767         ctx = nfs4_state_find_open_context(state);
2768         if (IS_ERR(ctx))
2769                 return -EAGAIN;
2770         ret = nfs4_do_open_expired(ctx, state);
2771         put_nfs_open_context(ctx);
2772         return ret;
2773 }
2774
2775 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2776                 const nfs4_stateid *stateid)
2777 {
2778         nfs_remove_bad_delegation(state->inode, stateid);
2779         nfs_state_clear_delegation(state);
2780 }
2781
2782 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2783 {
2784         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2785                 nfs_finish_clear_delegation_stateid(state, NULL);
2786 }
2787
2788 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2789 {
2790         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2791         nfs40_clear_delegation_stateid(state);
2792         nfs_state_clear_open_state_flags(state);
2793         return nfs4_open_expired(sp, state);
2794 }
2795
2796 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2797                 nfs4_stateid *stateid,
2798                 const struct cred *cred)
2799 {
2800         return -NFS4ERR_BAD_STATEID;
2801 }
2802
2803 #if defined(CONFIG_NFS_V4_1)
2804 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2805                 nfs4_stateid *stateid,
2806                 const struct cred *cred)
2807 {
2808         int status;
2809
2810         switch (stateid->type) {
2811         default:
2812                 break;
2813         case NFS4_INVALID_STATEID_TYPE:
2814         case NFS4_SPECIAL_STATEID_TYPE:
2815                 return -NFS4ERR_BAD_STATEID;
2816         case NFS4_REVOKED_STATEID_TYPE:
2817                 goto out_free;
2818         }
2819
2820         status = nfs41_test_stateid(server, stateid, cred);
2821         switch (status) {
2822         case -NFS4ERR_EXPIRED:
2823         case -NFS4ERR_ADMIN_REVOKED:
2824         case -NFS4ERR_DELEG_REVOKED:
2825                 break;
2826         default:
2827                 return status;
2828         }
2829 out_free:
2830         /* Ack the revoked state to the server */
2831         nfs41_free_stateid(server, stateid, cred, true);
2832         return -NFS4ERR_EXPIRED;
2833 }
2834
2835 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2836 {
2837         struct nfs_server *server = NFS_SERVER(state->inode);
2838         nfs4_stateid stateid;
2839         struct nfs_delegation *delegation;
2840         const struct cred *cred = NULL;
2841         int status, ret = NFS_OK;
2842
2843         /* Get the delegation credential for use by test/free_stateid */
2844         rcu_read_lock();
2845         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2846         if (delegation == NULL) {
2847                 rcu_read_unlock();
2848                 nfs_state_clear_delegation(state);
2849                 return NFS_OK;
2850         }
2851
2852         spin_lock(&delegation->lock);
2853         nfs4_stateid_copy(&stateid, &delegation->stateid);
2854
2855         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2856                                 &delegation->flags)) {
2857                 spin_unlock(&delegation->lock);
2858                 rcu_read_unlock();
2859                 return NFS_OK;
2860         }
2861
2862         if (delegation->cred)
2863                 cred = get_cred(delegation->cred);
2864         spin_unlock(&delegation->lock);
2865         rcu_read_unlock();
2866         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2867         trace_nfs4_test_delegation_stateid(state, NULL, status);
2868         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2869                 nfs_finish_clear_delegation_stateid(state, &stateid);
2870         else
2871                 ret = status;
2872
2873         put_cred(cred);
2874         return ret;
2875 }
2876
2877 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2878 {
2879         nfs4_stateid tmp;
2880
2881         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2882             nfs4_copy_delegation_stateid(state->inode, state->state,
2883                                 &tmp, NULL) &&
2884             nfs4_stateid_match_other(&state->stateid, &tmp))
2885                 nfs_state_set_delegation(state, &tmp, state->state);
2886         else
2887                 nfs_state_clear_delegation(state);
2888 }
2889
2890 /**
2891  * nfs41_check_expired_locks - possibly free a lock stateid
2892  *
2893  * @state: NFSv4 state for an inode
2894  *
2895  * Returns NFS_OK if recovery for this stateid is now finished.
2896  * Otherwise a negative NFS4ERR value is returned.
2897  */
2898 static int nfs41_check_expired_locks(struct nfs4_state *state)
2899 {
2900         int status, ret = NFS_OK;
2901         struct nfs4_lock_state *lsp, *prev = NULL;
2902         struct nfs_server *server = NFS_SERVER(state->inode);
2903
2904         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2905                 goto out;
2906
2907         spin_lock(&state->state_lock);
2908         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2909                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2910                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2911
2912                         refcount_inc(&lsp->ls_count);
2913                         spin_unlock(&state->state_lock);
2914
2915                         nfs4_put_lock_state(prev);
2916                         prev = lsp;
2917
2918                         status = nfs41_test_and_free_expired_stateid(server,
2919                                         &lsp->ls_stateid,
2920                                         cred);
2921                         trace_nfs4_test_lock_stateid(state, lsp, status);
2922                         if (status == -NFS4ERR_EXPIRED ||
2923                             status == -NFS4ERR_BAD_STATEID) {
2924                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2925                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2926                                 if (!recover_lost_locks)
2927                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2928                         } else if (status != NFS_OK) {
2929                                 ret = status;
2930                                 nfs4_put_lock_state(prev);
2931                                 goto out;
2932                         }
2933                         spin_lock(&state->state_lock);
2934                 }
2935         }
2936         spin_unlock(&state->state_lock);
2937         nfs4_put_lock_state(prev);
2938 out:
2939         return ret;
2940 }
2941
2942 /**
2943  * nfs41_check_open_stateid - possibly free an open stateid
2944  *
2945  * @state: NFSv4 state for an inode
2946  *
2947  * Returns NFS_OK if recovery for this stateid is now finished.
2948  * Otherwise a negative NFS4ERR value is returned.
2949  */
2950 static int nfs41_check_open_stateid(struct nfs4_state *state)
2951 {
2952         struct nfs_server *server = NFS_SERVER(state->inode);
2953         nfs4_stateid *stateid = &state->open_stateid;
2954         const struct cred *cred = state->owner->so_cred;
2955         int status;
2956
2957         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2958                 return -NFS4ERR_BAD_STATEID;
2959         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2960         trace_nfs4_test_open_stateid(state, NULL, status);
2961         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2962                 nfs_state_clear_open_state_flags(state);
2963                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2964                 return status;
2965         }
2966         if (nfs_open_stateid_recover_openmode(state))
2967                 return -NFS4ERR_OPENMODE;
2968         return NFS_OK;
2969 }
2970
2971 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2972 {
2973         int status;
2974
2975         status = nfs41_check_delegation_stateid(state);
2976         if (status != NFS_OK)
2977                 return status;
2978         nfs41_delegation_recover_stateid(state);
2979
2980         status = nfs41_check_expired_locks(state);
2981         if (status != NFS_OK)
2982                 return status;
2983         status = nfs41_check_open_stateid(state);
2984         if (status != NFS_OK)
2985                 status = nfs4_open_expired(sp, state);
2986         return status;
2987 }
2988 #endif
2989
2990 /*
2991  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2992  * fields corresponding to attributes that were used to store the verifier.
2993  * Make sure we clobber those fields in the later setattr call
2994  */
2995 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2996                                 struct iattr *sattr, struct nfs4_label **label)
2997 {
2998         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2999         __u32 attrset[3];
3000         unsigned ret;
3001         unsigned i;
3002
3003         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3004                 attrset[i] = opendata->o_res.attrset[i];
3005                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3006                         attrset[i] &= ~bitmask[i];
3007         }
3008
3009         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3010                 sattr->ia_valid : 0;
3011
3012         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3013                 if (sattr->ia_valid & ATTR_ATIME_SET)
3014                         ret |= ATTR_ATIME_SET;
3015                 else
3016                         ret |= ATTR_ATIME;
3017         }
3018
3019         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3020                 if (sattr->ia_valid & ATTR_MTIME_SET)
3021                         ret |= ATTR_MTIME_SET;
3022                 else
3023                         ret |= ATTR_MTIME;
3024         }
3025
3026         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3027                 *label = NULL;
3028         return ret;
3029 }
3030
3031 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3032                 int flags, struct nfs_open_context *ctx)
3033 {
3034         struct nfs4_state_owner *sp = opendata->owner;
3035         struct nfs_server *server = sp->so_server;
3036         struct dentry *dentry;
3037         struct nfs4_state *state;
3038         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3039         struct inode *dir = d_inode(opendata->dir);
3040         unsigned long dir_verifier;
3041         unsigned int seq;
3042         int ret;
3043
3044         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3045         dir_verifier = nfs_save_change_attribute(dir);
3046
3047         ret = _nfs4_proc_open(opendata, ctx);
3048         if (ret != 0)
3049                 goto out;
3050
3051         state = _nfs4_opendata_to_nfs4_state(opendata);
3052         ret = PTR_ERR(state);
3053         if (IS_ERR(state))
3054                 goto out;
3055         ctx->state = state;
3056         if (server->caps & NFS_CAP_POSIX_LOCK)
3057                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3058         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3059                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3060
3061         dentry = opendata->dentry;
3062         if (d_really_is_negative(dentry)) {
3063                 struct dentry *alias;
3064                 d_drop(dentry);
3065                 alias = d_exact_alias(dentry, state->inode);
3066                 if (!alias)
3067                         alias = d_splice_alias(igrab(state->inode), dentry);
3068                 /* d_splice_alias() can't fail here - it's a non-directory */
3069                 if (alias) {
3070                         dput(ctx->dentry);
3071                         ctx->dentry = dentry = alias;
3072                 }
3073         }
3074
3075         switch(opendata->o_arg.claim) {
3076         default:
3077                 break;
3078         case NFS4_OPEN_CLAIM_NULL:
3079         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3080         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3081                 if (!opendata->rpc_done)
3082                         break;
3083                 if (opendata->o_res.delegation_type != 0)
3084                         dir_verifier = nfs_save_change_attribute(dir);
3085                 nfs_set_verifier(dentry, dir_verifier);
3086         }
3087
3088         /* Parse layoutget results before we check for access */
3089         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3090
3091         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3092                         acc_mode, flags);
3093         if (ret != 0)
3094                 goto out;
3095
3096         if (d_inode(dentry) == state->inode) {
3097                 nfs_inode_attach_open_context(ctx);
3098                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3099                         nfs4_schedule_stateid_recovery(server, state);
3100         }
3101
3102 out:
3103         if (!opendata->cancelled)
3104                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3105         return ret;
3106 }
3107
3108 /*
3109  * Returns a referenced nfs4_state
3110  */
3111 static int _nfs4_do_open(struct inode *dir,
3112                         struct nfs_open_context *ctx,
3113                         int flags,
3114                         const struct nfs4_open_createattrs *c,
3115                         int *opened)
3116 {
3117         struct nfs4_state_owner  *sp;
3118         struct nfs4_state     *state = NULL;
3119         struct nfs_server       *server = NFS_SERVER(dir);
3120         struct nfs4_opendata *opendata;
3121         struct dentry *dentry = ctx->dentry;
3122         const struct cred *cred = ctx->cred;
3123         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3124         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3125         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3126         struct iattr *sattr = c->sattr;
3127         struct nfs4_label *label = c->label;
3128         struct nfs4_label *olabel = NULL;
3129         int status;
3130
3131         /* Protect against reboot recovery conflicts */
3132         status = -ENOMEM;
3133         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3134         if (sp == NULL) {
3135                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3136                 goto out_err;
3137         }
3138         status = nfs4_client_recover_expired_lease(server->nfs_client);
3139         if (status != 0)
3140                 goto err_put_state_owner;
3141         if (d_really_is_positive(dentry))
3142                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3143         status = -ENOMEM;
3144         if (d_really_is_positive(dentry))
3145                 claim = NFS4_OPEN_CLAIM_FH;
3146         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3147                         c, claim, GFP_KERNEL);
3148         if (opendata == NULL)
3149                 goto err_put_state_owner;
3150
3151         if (label) {
3152                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3153                 if (IS_ERR(olabel)) {
3154                         status = PTR_ERR(olabel);
3155                         goto err_opendata_put;
3156                 }
3157         }
3158
3159         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3160                 if (!opendata->f_attr.mdsthreshold) {
3161                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3162                         if (!opendata->f_attr.mdsthreshold)
3163                                 goto err_free_label;
3164                 }
3165                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3166         }
3167         if (d_really_is_positive(dentry))
3168                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3169
3170         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3171         if (status != 0)
3172                 goto err_free_label;
3173         state = ctx->state;
3174
3175         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3176             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3177                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3178                 /*
3179                  * send create attributes which was not set by open
3180                  * with an extra setattr.
3181                  */
3182                 if (attrs || label) {
3183                         unsigned ia_old = sattr->ia_valid;
3184
3185                         sattr->ia_valid = attrs;
3186                         nfs_fattr_init(opendata->o_res.f_attr);
3187                         status = nfs4_do_setattr(state->inode, cred,
3188                                         opendata->o_res.f_attr, sattr,
3189                                         ctx, label, olabel);
3190                         if (status == 0) {
3191                                 nfs_setattr_update_inode(state->inode, sattr,
3192                                                 opendata->o_res.f_attr);
3193                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3194                         }
3195                         sattr->ia_valid = ia_old;
3196                 }
3197         }
3198         if (opened && opendata->file_created)
3199                 *opened = 1;
3200
3201         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3202                 *ctx_th = opendata->f_attr.mdsthreshold;
3203                 opendata->f_attr.mdsthreshold = NULL;
3204         }
3205
3206         nfs4_label_free(olabel);
3207
3208         nfs4_opendata_put(opendata);
3209         nfs4_put_state_owner(sp);
3210         return 0;
3211 err_free_label:
3212         nfs4_label_free(olabel);
3213 err_opendata_put:
3214         nfs4_opendata_put(opendata);
3215 err_put_state_owner:
3216         nfs4_put_state_owner(sp);
3217 out_err:
3218         return status;
3219 }
3220
3221
3222 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3223                                         struct nfs_open_context *ctx,
3224                                         int flags,
3225                                         struct iattr *sattr,
3226                                         struct nfs4_label *label,
3227                                         int *opened)
3228 {
3229         struct nfs_server *server = NFS_SERVER(dir);
3230         struct nfs4_exception exception = {
3231                 .interruptible = true,
3232         };
3233         struct nfs4_state *res;
3234         struct nfs4_open_createattrs c = {
3235                 .label = label,
3236                 .sattr = sattr,
3237                 .verf = {
3238                         [0] = (__u32)jiffies,
3239                         [1] = (__u32)current->pid,
3240                 },
3241         };
3242         int status;
3243
3244         do {
3245                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3246                 res = ctx->state;
3247                 trace_nfs4_open_file(ctx, flags, status);
3248                 if (status == 0)
3249                         break;
3250                 /* NOTE: BAD_SEQID means the server and client disagree about the
3251                  * book-keeping w.r.t. state-changing operations
3252                  * (OPEN/CLOSE/LOCK/LOCKU...)
3253                  * It is actually a sign of a bug on the client or on the server.
3254                  *
3255                  * If we receive a BAD_SEQID error in the particular case of
3256                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3257                  * have unhashed the old state_owner for us, and that we can
3258                  * therefore safely retry using a new one. We should still warn
3259                  * the user though...
3260                  */
3261                 if (status == -NFS4ERR_BAD_SEQID) {
3262                         pr_warn_ratelimited("NFS: v4 server %s "
3263                                         " returned a bad sequence-id error!\n",
3264                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3265                         exception.retry = 1;
3266                         continue;
3267                 }
3268                 /*
3269                  * BAD_STATEID on OPEN means that the server cancelled our
3270                  * state before it received the OPEN_CONFIRM.
3271                  * Recover by retrying the request as per the discussion
3272                  * on Page 181 of RFC3530.
3273                  */
3274                 if (status == -NFS4ERR_BAD_STATEID) {
3275                         exception.retry = 1;
3276                         continue;
3277                 }
3278                 if (status == -NFS4ERR_EXPIRED) {
3279                         nfs4_schedule_lease_recovery(server->nfs_client);
3280                         exception.retry = 1;
3281                         continue;
3282                 }
3283                 if (status == -EAGAIN) {
3284                         /* We must have found a delegation */
3285                         exception.retry = 1;
3286                         continue;
3287                 }
3288                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3289                         continue;
3290                 res = ERR_PTR(nfs4_handle_exception(server,
3291                                         status, &exception));
3292         } while (exception.retry);
3293         return res;
3294 }
3295
3296 static int _nfs4_do_setattr(struct inode *inode,
3297                             struct nfs_setattrargs *arg,
3298                             struct nfs_setattrres *res,
3299                             const struct cred *cred,
3300                             struct nfs_open_context *ctx)
3301 {
3302         struct nfs_server *server = NFS_SERVER(inode);
3303         struct rpc_message msg = {
3304                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3305                 .rpc_argp       = arg,
3306                 .rpc_resp       = res,
3307                 .rpc_cred       = cred,
3308         };
3309         const struct cred *delegation_cred = NULL;
3310         unsigned long timestamp = jiffies;
3311         bool truncate;
3312         int status;
3313
3314         nfs_fattr_init(res->fattr);
3315
3316         /* Servers should only apply open mode checks for file size changes */
3317         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3318         if (!truncate) {
3319                 nfs4_inode_make_writeable(inode);
3320                 goto zero_stateid;
3321         }
3322
3323         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3324                 /* Use that stateid */
3325         } else if (ctx != NULL && ctx->state) {
3326                 struct nfs_lock_context *l_ctx;
3327                 if (!nfs4_valid_open_stateid(ctx->state))
3328                         return -EBADF;
3329                 l_ctx = nfs_get_lock_context(ctx);
3330                 if (IS_ERR(l_ctx))
3331                         return PTR_ERR(l_ctx);
3332                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3333                                                 &arg->stateid, &delegation_cred);
3334                 nfs_put_lock_context(l_ctx);
3335                 if (status == -EIO)
3336                         return -EBADF;
3337                 else if (status == -EAGAIN)
3338                         goto zero_stateid;
3339         } else {
3340 zero_stateid:
3341                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3342         }
3343         if (delegation_cred)
3344                 msg.rpc_cred = delegation_cred;
3345
3346         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3347
3348         put_cred(delegation_cred);
3349         if (status == 0 && ctx != NULL)
3350                 renew_lease(server, timestamp);
3351         trace_nfs4_setattr(inode, &arg->stateid, status);
3352         return status;
3353 }
3354
3355 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3356                            struct nfs_fattr *fattr, struct iattr *sattr,
3357                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3358                            struct nfs4_label *olabel)
3359 {
3360         struct nfs_server *server = NFS_SERVER(inode);
3361         __u32 bitmask[NFS4_BITMASK_SZ];
3362         struct nfs4_state *state = ctx ? ctx->state : NULL;
3363         struct nfs_setattrargs  arg = {
3364                 .fh             = NFS_FH(inode),
3365                 .iap            = sattr,
3366                 .server         = server,
3367                 .bitmask = bitmask,
3368                 .label          = ilabel,
3369         };
3370         struct nfs_setattrres  res = {
3371                 .fattr          = fattr,
3372                 .label          = olabel,
3373                 .server         = server,
3374         };
3375         struct nfs4_exception exception = {
3376                 .state = state,
3377                 .inode = inode,
3378                 .stateid = &arg.stateid,
3379         };
3380         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3381         int err;
3382
3383         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3384                 adjust_flags |= NFS_INO_INVALID_MODE;
3385         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3386                 adjust_flags |= NFS_INO_INVALID_OTHER;
3387
3388         do {
3389                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, olabel),
3390                                         inode, adjust_flags);
3391
3392                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3393                 switch (err) {
3394                 case -NFS4ERR_OPENMODE:
3395                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3396                                 pr_warn_once("NFSv4: server %s is incorrectly "
3397                                                 "applying open mode checks to "
3398                                                 "a SETATTR that is not "
3399                                                 "changing file size.\n",
3400                                                 server->nfs_client->cl_hostname);
3401                         }
3402                         if (state && !(state->state & FMODE_WRITE)) {
3403                                 err = -EBADF;
3404                                 if (sattr->ia_valid & ATTR_OPEN)
3405                                         err = -EACCES;
3406                                 goto out;
3407                         }
3408                 }
3409                 err = nfs4_handle_exception(server, err, &exception);
3410         } while (exception.retry);
3411 out:
3412         return err;
3413 }
3414
3415 static bool
3416 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3417 {
3418         if (inode == NULL || !nfs_have_layout(inode))
3419                 return false;
3420
3421         return pnfs_wait_on_layoutreturn(inode, task);
3422 }
3423
3424 /*
3425  * Update the seqid of an open stateid
3426  */
3427 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3428                 struct nfs4_state *state)
3429 {
3430         __be32 seqid_open;
3431         u32 dst_seqid;
3432         int seq;
3433
3434         for (;;) {
3435                 if (!nfs4_valid_open_stateid(state))
3436                         break;
3437                 seq = read_seqbegin(&state->seqlock);
3438                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3439                         nfs4_stateid_copy(dst, &state->open_stateid);
3440                         if (read_seqretry(&state->seqlock, seq))
3441                                 continue;
3442                         break;
3443                 }
3444                 seqid_open = state->open_stateid.seqid;
3445                 if (read_seqretry(&state->seqlock, seq))
3446                         continue;
3447
3448                 dst_seqid = be32_to_cpu(dst->seqid);
3449                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3450                         dst->seqid = seqid_open;
3451                 break;
3452         }
3453 }
3454
3455 /*
3456  * Update the seqid of an open stateid after receiving
3457  * NFS4ERR_OLD_STATEID
3458  */
3459 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3460                 struct nfs4_state *state)
3461 {
3462         __be32 seqid_open;
3463         u32 dst_seqid;
3464         bool ret;
3465         int seq, status = -EAGAIN;
3466         DEFINE_WAIT(wait);
3467
3468         for (;;) {
3469                 ret = false;
3470                 if (!nfs4_valid_open_stateid(state))
3471                         break;
3472                 seq = read_seqbegin(&state->seqlock);
3473                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3474                         if (read_seqretry(&state->seqlock, seq))
3475                                 continue;
3476                         break;
3477                 }
3478
3479                 write_seqlock(&state->seqlock);
3480                 seqid_open = state->open_stateid.seqid;
3481
3482                 dst_seqid = be32_to_cpu(dst->seqid);
3483
3484                 /* Did another OPEN bump the state's seqid?  try again: */
3485                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3486                         dst->seqid = seqid_open;
3487                         write_sequnlock(&state->seqlock);
3488                         ret = true;
3489                         break;
3490                 }
3491
3492                 /* server says we're behind but we haven't seen the update yet */
3493                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3494                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3495                 write_sequnlock(&state->seqlock);
3496                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3497
3498                 if (fatal_signal_pending(current))
3499                         status = -EINTR;
3500                 else
3501                         if (schedule_timeout(5*HZ) != 0)
3502                                 status = 0;
3503
3504                 finish_wait(&state->waitq, &wait);
3505
3506                 if (!status)
3507                         continue;
3508                 if (status == -EINTR)
3509                         break;
3510
3511                 /* we slept the whole 5 seconds, we must have lost a seqid */
3512                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3513                 ret = true;
3514                 break;
3515         }
3516
3517         return ret;
3518 }
3519
3520 struct nfs4_closedata {
3521         struct inode *inode;
3522         struct nfs4_state *state;
3523         struct nfs_closeargs arg;
3524         struct nfs_closeres res;
3525         struct {
3526                 struct nfs4_layoutreturn_args arg;
3527                 struct nfs4_layoutreturn_res res;
3528                 struct nfs4_xdr_opaque_data ld_private;
3529                 u32 roc_barrier;
3530                 bool roc;
3531         } lr;
3532         struct nfs_fattr fattr;
3533         unsigned long timestamp;
3534 };
3535
3536 static void nfs4_free_closedata(void *data)
3537 {
3538         struct nfs4_closedata *calldata = data;
3539         struct nfs4_state_owner *sp = calldata->state->owner;
3540         struct super_block *sb = calldata->state->inode->i_sb;
3541
3542         if (calldata->lr.roc)
3543                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3544                                 calldata->res.lr_ret);
3545         nfs4_put_open_state(calldata->state);
3546         nfs_free_seqid(calldata->arg.seqid);
3547         nfs4_put_state_owner(sp);
3548         nfs_sb_deactive(sb);
3549         kfree(calldata);
3550 }
3551
3552 static void nfs4_close_done(struct rpc_task *task, void *data)
3553 {
3554         struct nfs4_closedata *calldata = data;
3555         struct nfs4_state *state = calldata->state;
3556         struct nfs_server *server = NFS_SERVER(calldata->inode);
3557         nfs4_stateid *res_stateid = NULL;
3558         struct nfs4_exception exception = {
3559                 .state = state,
3560                 .inode = calldata->inode,
3561                 .stateid = &calldata->arg.stateid,
3562         };
3563
3564         dprintk("%s: begin!\n", __func__);
3565         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3566                 return;
3567         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3568
3569         /* Handle Layoutreturn errors */
3570         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3571                           &calldata->res.lr_ret) == -EAGAIN)
3572                 goto out_restart;
3573
3574         /* hmm. we are done with the inode, and in the process of freeing
3575          * the state_owner. we keep this around to process errors
3576          */
3577         switch (task->tk_status) {
3578                 case 0:
3579                         res_stateid = &calldata->res.stateid;
3580                         renew_lease(server, calldata->timestamp);
3581                         break;
3582                 case -NFS4ERR_ACCESS:
3583                         if (calldata->arg.bitmask != NULL) {
3584                                 calldata->arg.bitmask = NULL;
3585                                 calldata->res.fattr = NULL;
3586                                 goto out_restart;
3587
3588                         }
3589                         break;
3590                 case -NFS4ERR_OLD_STATEID:
3591                         /* Did we race with OPEN? */
3592                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3593                                                 state))
3594                                 goto out_restart;
3595                         goto out_release;
3596                 case -NFS4ERR_ADMIN_REVOKED:
3597                 case -NFS4ERR_STALE_STATEID:
3598                 case -NFS4ERR_EXPIRED:
3599                         nfs4_free_revoked_stateid(server,
3600                                         &calldata->arg.stateid,
3601                                         task->tk_msg.rpc_cred);
3602                         fallthrough;
3603                 case -NFS4ERR_BAD_STATEID:
3604                         if (calldata->arg.fmode == 0)
3605                                 break;
3606                         fallthrough;
3607                 default:
3608                         task->tk_status = nfs4_async_handle_exception(task,
3609                                         server, task->tk_status, &exception);
3610                         if (exception.retry)
3611                                 goto out_restart;
3612         }
3613         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3614                         res_stateid, calldata->arg.fmode);
3615 out_release:
3616         task->tk_status = 0;
3617         nfs_release_seqid(calldata->arg.seqid);
3618         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3619         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3620         return;
3621 out_restart:
3622         task->tk_status = 0;
3623         rpc_restart_call_prepare(task);
3624         goto out_release;
3625 }
3626
3627 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3628 {
3629         struct nfs4_closedata *calldata = data;
3630         struct nfs4_state *state = calldata->state;
3631         struct inode *inode = calldata->inode;
3632         struct nfs_server *server = NFS_SERVER(inode);
3633         struct pnfs_layout_hdr *lo;
3634         bool is_rdonly, is_wronly, is_rdwr;
3635         int call_close = 0;
3636
3637         dprintk("%s: begin!\n", __func__);
3638         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3639                 goto out_wait;
3640
3641         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3642         spin_lock(&state->owner->so_lock);
3643         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3644         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3645         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3646         /* Calculate the change in open mode */
3647         calldata->arg.fmode = 0;
3648         if (state->n_rdwr == 0) {
3649                 if (state->n_rdonly == 0)
3650                         call_close |= is_rdonly;
3651                 else if (is_rdonly)
3652                         calldata->arg.fmode |= FMODE_READ;
3653                 if (state->n_wronly == 0)
3654                         call_close |= is_wronly;
3655                 else if (is_wronly)
3656                         calldata->arg.fmode |= FMODE_WRITE;
3657                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3658                         call_close |= is_rdwr;
3659         } else if (is_rdwr)
3660                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3661
3662         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3663         if (!nfs4_valid_open_stateid(state))
3664                 call_close = 0;
3665         spin_unlock(&state->owner->so_lock);
3666
3667         if (!call_close) {
3668                 /* Note: exit _without_ calling nfs4_close_done */
3669                 goto out_no_action;
3670         }
3671
3672         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3673                 nfs_release_seqid(calldata->arg.seqid);
3674                 goto out_wait;
3675         }
3676
3677         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3678         if (lo && !pnfs_layout_is_valid(lo)) {
3679                 calldata->arg.lr_args = NULL;
3680                 calldata->res.lr_res = NULL;
3681         }
3682
3683         if (calldata->arg.fmode == 0)
3684                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3685
3686         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3687                 /* Close-to-open cache consistency revalidation */
3688                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3689                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3690                                          server->cache_consistency_bitmask,
3691                                          inode, server, NULL);
3692                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3693                 } else
3694                         calldata->arg.bitmask = NULL;
3695         }
3696
3697         calldata->arg.share_access =
3698                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3699                                 calldata->arg.fmode, 0);
3700
3701         if (calldata->res.fattr == NULL)
3702                 calldata->arg.bitmask = NULL;
3703         else if (calldata->arg.bitmask == NULL)
3704                 calldata->res.fattr = NULL;
3705         calldata->timestamp = jiffies;
3706         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3707                                 &calldata->arg.seq_args,
3708                                 &calldata->res.seq_res,
3709                                 task) != 0)
3710                 nfs_release_seqid(calldata->arg.seqid);
3711         dprintk("%s: done!\n", __func__);
3712         return;
3713 out_no_action:
3714         task->tk_action = NULL;
3715 out_wait:
3716         nfs4_sequence_done(task, &calldata->res.seq_res);
3717 }
3718
3719 static const struct rpc_call_ops nfs4_close_ops = {
3720         .rpc_call_prepare = nfs4_close_prepare,
3721         .rpc_call_done = nfs4_close_done,
3722         .rpc_release = nfs4_free_closedata,
3723 };
3724
3725 /*
3726  * It is possible for data to be read/written from a mem-mapped file
3727  * after the sys_close call (which hits the vfs layer as a flush).
3728  * This means that we can't safely call nfsv4 close on a file until
3729  * the inode is cleared. This in turn means that we are not good
3730  * NFSv4 citizens - we do not indicate to the server to update the file's
3731  * share state even when we are done with one of the three share
3732  * stateid's in the inode.
3733  *
3734  * NOTE: Caller must be holding the sp->so_owner semaphore!
3735  */
3736 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3737 {
3738         struct nfs_server *server = NFS_SERVER(state->inode);
3739         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3740         struct nfs4_closedata *calldata;
3741         struct nfs4_state_owner *sp = state->owner;
3742         struct rpc_task *task;
3743         struct rpc_message msg = {
3744                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3745                 .rpc_cred = state->owner->so_cred,
3746         };
3747         struct rpc_task_setup task_setup_data = {
3748                 .rpc_client = server->client,
3749                 .rpc_message = &msg,
3750                 .callback_ops = &nfs4_close_ops,
3751                 .workqueue = nfsiod_workqueue,
3752                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3753         };
3754         int status = -ENOMEM;
3755
3756         if (server->nfs_client->cl_minorversion)
3757                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3758
3759         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3760                 &task_setup_data.rpc_client, &msg);
3761
3762         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3763         if (calldata == NULL)
3764                 goto out;
3765         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3766         calldata->inode = state->inode;
3767         calldata->state = state;
3768         calldata->arg.fh = NFS_FH(state->inode);
3769         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3770                 goto out_free_calldata;
3771         /* Serialization for the sequence id */
3772         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3773         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3774         if (IS_ERR(calldata->arg.seqid))
3775                 goto out_free_calldata;
3776         nfs_fattr_init(&calldata->fattr);
3777         calldata->arg.fmode = 0;
3778         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3779         calldata->res.fattr = &calldata->fattr;
3780         calldata->res.seqid = calldata->arg.seqid;
3781         calldata->res.server = server;
3782         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3783         calldata->lr.roc = pnfs_roc(state->inode,
3784                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3785         if (calldata->lr.roc) {
3786                 calldata->arg.lr_args = &calldata->lr.arg;
3787                 calldata->res.lr_res = &calldata->lr.res;
3788         }
3789         nfs_sb_active(calldata->inode->i_sb);
3790
3791         msg.rpc_argp = &calldata->arg;
3792         msg.rpc_resp = &calldata->res;
3793         task_setup_data.callback_data = calldata;
3794         task = rpc_run_task(&task_setup_data);
3795         if (IS_ERR(task))
3796                 return PTR_ERR(task);
3797         status = 0;
3798         if (wait)
3799                 status = rpc_wait_for_completion_task(task);
3800         rpc_put_task(task);
3801         return status;
3802 out_free_calldata:
3803         kfree(calldata);
3804 out:
3805         nfs4_put_open_state(state);
3806         nfs4_put_state_owner(sp);
3807         return status;
3808 }
3809
3810 static struct inode *
3811 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3812                 int open_flags, struct iattr *attr, int *opened)
3813 {
3814         struct nfs4_state *state;
3815         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3816
3817         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3818
3819         /* Protect against concurrent sillydeletes */
3820         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3821
3822         nfs4_label_release_security(label);
3823
3824         if (IS_ERR(state))
3825                 return ERR_CAST(state);
3826         return state->inode;
3827 }
3828
3829 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3830 {
3831         if (ctx->state == NULL)
3832                 return;
3833         if (is_sync)
3834                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3835         else
3836                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3837 }
3838
3839 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3840 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3841 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3842
3843 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3844 {
3845         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3846         struct nfs4_server_caps_arg args = {
3847                 .fhandle = fhandle,
3848                 .bitmask = bitmask,
3849         };
3850         struct nfs4_server_caps_res res = {};
3851         struct rpc_message msg = {
3852                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3853                 .rpc_argp = &args,
3854                 .rpc_resp = &res,
3855         };
3856         int status;
3857         int i;
3858
3859         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3860                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3861                      FATTR4_WORD0_LINK_SUPPORT |
3862                      FATTR4_WORD0_SYMLINK_SUPPORT |
3863                      FATTR4_WORD0_ACLSUPPORT;
3864         if (minorversion)
3865                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3866
3867         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3868         if (status == 0) {
3869                 /* Sanity check the server answers */
3870                 switch (minorversion) {
3871                 case 0:
3872                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3873                         res.attr_bitmask[2] = 0;
3874                         break;
3875                 case 1:
3876                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3877                         break;
3878                 case 2:
3879                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3880                 }
3881                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3882                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3883                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3884                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3885                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3886                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3887                         server->caps |= NFS_CAP_ACLS;
3888                 if (res.has_links != 0)
3889                         server->caps |= NFS_CAP_HARDLINKS;
3890                 if (res.has_symlinks != 0)
3891                         server->caps |= NFS_CAP_SYMLINKS;
3892 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3893                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3894                         server->caps |= NFS_CAP_SECURITY_LABEL;
3895 #endif
3896                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3897                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3898                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3899                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3900                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3901                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3902                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3903                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3904                                 NFS_ATTR_FATTR_OWNER_NAME);
3905                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3906                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3907                                 NFS_ATTR_FATTR_GROUP_NAME);
3908                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3909                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3910                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3911                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3912                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3913                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3914                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3915                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3916                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3917                                 sizeof(server->attr_bitmask));
3918                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3919
3920                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3921                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3922                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3923                 server->cache_consistency_bitmask[2] = 0;
3924
3925                 /* Avoid a regression due to buggy server */
3926                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3927                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3928                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3929                         sizeof(server->exclcreat_bitmask));
3930
3931                 server->acl_bitmask = res.acl_bitmask;
3932                 server->fh_expire_type = res.fh_expire_type;
3933         }
3934
3935         return status;
3936 }
3937
3938 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3939 {
3940         struct nfs4_exception exception = {
3941                 .interruptible = true,
3942         };
3943         int err;
3944         do {
3945                 err = nfs4_handle_exception(server,
3946                                 _nfs4_server_capabilities(server, fhandle),
3947                                 &exception);
3948         } while (exception.retry);
3949         return err;
3950 }
3951
3952 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3953                 struct nfs_fsinfo *info)
3954 {
3955         u32 bitmask[3];
3956         struct nfs4_lookup_root_arg args = {
3957                 .bitmask = bitmask,
3958         };
3959         struct nfs4_lookup_res res = {
3960                 .server = server,
3961                 .fattr = info->fattr,
3962                 .fh = fhandle,
3963         };
3964         struct rpc_message msg = {
3965                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3966                 .rpc_argp = &args,
3967                 .rpc_resp = &res,
3968         };
3969
3970         bitmask[0] = nfs4_fattr_bitmap[0];
3971         bitmask[1] = nfs4_fattr_bitmap[1];
3972         /*
3973          * Process the label in the upcoming getfattr
3974          */
3975         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3976
3977         nfs_fattr_init(info->fattr);
3978         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3979 }
3980
3981 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3982                 struct nfs_fsinfo *info)
3983 {
3984         struct nfs4_exception exception = {
3985                 .interruptible = true,
3986         };
3987         int err;
3988         do {
3989                 err = _nfs4_lookup_root(server, fhandle, info);
3990                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3991                 switch (err) {
3992                 case 0:
3993                 case -NFS4ERR_WRONGSEC:
3994                         goto out;
3995                 default:
3996                         err = nfs4_handle_exception(server, err, &exception);
3997                 }
3998         } while (exception.retry);
3999 out:
4000         return err;
4001 }
4002
4003 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4004                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4005 {
4006         struct rpc_auth_create_args auth_args = {
4007                 .pseudoflavor = flavor,
4008         };
4009         struct rpc_auth *auth;
4010
4011         auth = rpcauth_create(&auth_args, server->client);
4012         if (IS_ERR(auth))
4013                 return -EACCES;
4014         return nfs4_lookup_root(server, fhandle, info);
4015 }
4016
4017 /*
4018  * Retry pseudoroot lookup with various security flavors.  We do this when:
4019  *
4020  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4021  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4022  *
4023  * Returns zero on success, or a negative NFS4ERR value, or a
4024  * negative errno value.
4025  */
4026 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4027                               struct nfs_fsinfo *info)
4028 {
4029         /* Per 3530bis 15.33.5 */
4030         static const rpc_authflavor_t flav_array[] = {
4031                 RPC_AUTH_GSS_KRB5P,
4032                 RPC_AUTH_GSS_KRB5I,
4033                 RPC_AUTH_GSS_KRB5,
4034                 RPC_AUTH_UNIX,                  /* courtesy */
4035                 RPC_AUTH_NULL,
4036         };
4037         int status = -EPERM;
4038         size_t i;
4039
4040         if (server->auth_info.flavor_len > 0) {
4041                 /* try each flavor specified by user */
4042                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4043                         status = nfs4_lookup_root_sec(server, fhandle, info,
4044                                                 server->auth_info.flavors[i]);
4045                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4046                                 continue;
4047                         break;
4048                 }
4049         } else {
4050                 /* no flavors specified by user, try default list */
4051                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4052                         status = nfs4_lookup_root_sec(server, fhandle, info,
4053                                                       flav_array[i]);
4054                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4055                                 continue;
4056                         break;
4057                 }
4058         }
4059
4060         /*
4061          * -EACCES could mean that the user doesn't have correct permissions
4062          * to access the mount.  It could also mean that we tried to mount
4063          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4064          * existing mount programs don't handle -EACCES very well so it should
4065          * be mapped to -EPERM instead.
4066          */
4067         if (status == -EACCES)
4068                 status = -EPERM;
4069         return status;
4070 }
4071
4072 /**
4073  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4074  * @server: initialized nfs_server handle
4075  * @fhandle: we fill in the pseudo-fs root file handle
4076  * @info: we fill in an FSINFO struct
4077  * @auth_probe: probe the auth flavours
4078  *
4079  * Returns zero on success, or a negative errno.
4080  */
4081 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4082                          struct nfs_fsinfo *info,
4083                          bool auth_probe)
4084 {
4085         int status = 0;
4086
4087         if (!auth_probe)
4088                 status = nfs4_lookup_root(server, fhandle, info);
4089
4090         if (auth_probe || status == NFS4ERR_WRONGSEC)
4091                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4092                                 fhandle, info);
4093
4094         if (status == 0)
4095                 status = nfs4_server_capabilities(server, fhandle);
4096         if (status == 0)
4097                 status = nfs4_do_fsinfo(server, fhandle, info);
4098
4099         return nfs4_map_errors(status);
4100 }
4101
4102 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4103                               struct nfs_fsinfo *info)
4104 {
4105         int error;
4106         struct nfs_fattr *fattr = info->fattr;
4107         struct nfs4_label *label = fattr->label;
4108
4109         error = nfs4_server_capabilities(server, mntfh);
4110         if (error < 0) {
4111                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4112                 return error;
4113         }
4114
4115         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4116         if (error < 0) {
4117                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4118                 goto out;
4119         }
4120
4121         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4122             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4123                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4124
4125 out:
4126         return error;
4127 }
4128
4129 /*
4130  * Get locations and (maybe) other attributes of a referral.
4131  * Note that we'll actually follow the referral later when
4132  * we detect fsid mismatch in inode revalidation
4133  */
4134 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4135                              const struct qstr *name, struct nfs_fattr *fattr,
4136                              struct nfs_fh *fhandle)
4137 {
4138         int status = -ENOMEM;
4139         struct page *page = NULL;
4140         struct nfs4_fs_locations *locations = NULL;
4141
4142         page = alloc_page(GFP_KERNEL);
4143         if (page == NULL)
4144                 goto out;
4145         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4146         if (locations == NULL)
4147                 goto out;
4148
4149         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4150         if (status != 0)
4151                 goto out;
4152
4153         /*
4154          * If the fsid didn't change, this is a migration event, not a
4155          * referral.  Cause us to drop into the exception handler, which
4156          * will kick off migration recovery.
4157          */
4158         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4159                 dprintk("%s: server did not return a different fsid for"
4160                         " a referral at %s\n", __func__, name->name);
4161                 status = -NFS4ERR_MOVED;
4162                 goto out;
4163         }
4164         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4165         nfs_fixup_referral_attributes(&locations->fattr);
4166
4167         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4168         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4169         memset(fhandle, 0, sizeof(struct nfs_fh));
4170 out:
4171         if (page)
4172                 __free_page(page);
4173         kfree(locations);
4174         return status;
4175 }
4176
4177 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4178                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4179                                 struct inode *inode)
4180 {
4181         __u32 bitmask[NFS4_BITMASK_SZ];
4182         struct nfs4_getattr_arg args = {
4183                 .fh = fhandle,
4184                 .bitmask = bitmask,
4185         };
4186         struct nfs4_getattr_res res = {
4187                 .fattr = fattr,
4188                 .label = label,
4189                 .server = server,
4190         };
4191         struct rpc_message msg = {
4192                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4193                 .rpc_argp = &args,
4194                 .rpc_resp = &res,
4195         };
4196         unsigned short task_flags = 0;
4197
4198         if (nfs4_has_session(server->nfs_client))
4199                 task_flags = RPC_TASK_MOVEABLE;
4200
4201         /* Is this is an attribute revalidation, subject to softreval? */
4202         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4203                 task_flags |= RPC_TASK_TIMEOUT;
4204
4205         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode, 0);
4206         nfs_fattr_init(fattr);
4207         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4208         return nfs4_do_call_sync(server->client, server, &msg,
4209                         &args.seq_args, &res.seq_res, task_flags);
4210 }
4211
4212 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4213                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4214                                 struct inode *inode)
4215 {
4216         struct nfs4_exception exception = {
4217                 .interruptible = true,
4218         };
4219         int err;
4220         do {
4221                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4222                 trace_nfs4_getattr(server, fhandle, fattr, err);
4223                 err = nfs4_handle_exception(server, err,
4224                                 &exception);
4225         } while (exception.retry);
4226         return err;
4227 }
4228
4229 /*
4230  * The file is not closed if it is opened due to the a request to change
4231  * the size of the file. The open call will not be needed once the
4232  * VFS layer lookup-intents are implemented.
4233  *
4234  * Close is called when the inode is destroyed.
4235  * If we haven't opened the file for O_WRONLY, we
4236  * need to in the size_change case to obtain a stateid.
4237  *
4238  * Got race?
4239  * Because OPEN is always done by name in nfsv4, it is
4240  * possible that we opened a different file by the same
4241  * name.  We can recognize this race condition, but we
4242  * can't do anything about it besides returning an error.
4243  *
4244  * This will be fixed with VFS changes (lookup-intent).
4245  */
4246 static int
4247 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4248                   struct iattr *sattr)
4249 {
4250         struct inode *inode = d_inode(dentry);
4251         const struct cred *cred = NULL;
4252         struct nfs_open_context *ctx = NULL;
4253         struct nfs4_label *label = NULL;
4254         int status;
4255
4256         if (pnfs_ld_layoutret_on_setattr(inode) &&
4257             sattr->ia_valid & ATTR_SIZE &&
4258             sattr->ia_size < i_size_read(inode))
4259                 pnfs_commit_and_return_layout(inode);
4260
4261         nfs_fattr_init(fattr);
4262
4263         /* Deal with open(O_TRUNC) */
4264         if (sattr->ia_valid & ATTR_OPEN)
4265                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4266
4267         /* Optimization: if the end result is no change, don't RPC */
4268         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4269                 return 0;
4270
4271         /* Search for an existing open(O_WRITE) file */
4272         if (sattr->ia_valid & ATTR_FILE) {
4273
4274                 ctx = nfs_file_open_context(sattr->ia_file);
4275                 if (ctx)
4276                         cred = ctx->cred;
4277         }
4278
4279         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4280         if (IS_ERR(label))
4281                 return PTR_ERR(label);
4282
4283         /* Return any delegations if we're going to change ACLs */
4284         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4285                 nfs4_inode_make_writeable(inode);
4286
4287         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4288         if (status == 0) {
4289                 nfs_setattr_update_inode(inode, sattr, fattr);
4290                 nfs_setsecurity(inode, fattr, label);
4291         }
4292         nfs4_label_free(label);
4293         return status;
4294 }
4295
4296 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4297                 struct dentry *dentry, struct nfs_fh *fhandle,
4298                 struct nfs_fattr *fattr, struct nfs4_label *label)
4299 {
4300         struct nfs_server *server = NFS_SERVER(dir);
4301         int                    status;
4302         struct nfs4_lookup_arg args = {
4303                 .bitmask = server->attr_bitmask,
4304                 .dir_fh = NFS_FH(dir),
4305                 .name = &dentry->d_name,
4306         };
4307         struct nfs4_lookup_res res = {
4308                 .server = server,
4309                 .fattr = fattr,
4310                 .label = label,
4311                 .fh = fhandle,
4312         };
4313         struct rpc_message msg = {
4314                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4315                 .rpc_argp = &args,
4316                 .rpc_resp = &res,
4317         };
4318         unsigned short task_flags = 0;
4319
4320         if (server->nfs_client->cl_minorversion)
4321                 task_flags = RPC_TASK_MOVEABLE;
4322
4323         /* Is this is an attribute revalidation, subject to softreval? */
4324         if (nfs_lookup_is_soft_revalidate(dentry))
4325                 task_flags |= RPC_TASK_TIMEOUT;
4326
4327         args.bitmask = nfs4_bitmask(server, label);
4328
4329         nfs_fattr_init(fattr);
4330
4331         dprintk("NFS call  lookup %pd2\n", dentry);
4332         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4333         status = nfs4_do_call_sync(clnt, server, &msg,
4334                         &args.seq_args, &res.seq_res, task_flags);
4335         dprintk("NFS reply lookup: %d\n", status);
4336         return status;
4337 }
4338
4339 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4340 {
4341         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4342                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4343         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4344         fattr->nlink = 2;
4345 }
4346
4347 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4348                                    struct dentry *dentry, struct nfs_fh *fhandle,
4349                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4350 {
4351         struct nfs4_exception exception = {
4352                 .interruptible = true,
4353         };
4354         struct rpc_clnt *client = *clnt;
4355         const struct qstr *name = &dentry->d_name;
4356         int err;
4357         do {
4358                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4359                 trace_nfs4_lookup(dir, name, err);
4360                 switch (err) {
4361                 case -NFS4ERR_BADNAME:
4362                         err = -ENOENT;
4363                         goto out;
4364                 case -NFS4ERR_MOVED:
4365                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4366                         if (err == -NFS4ERR_MOVED)
4367                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4368                         goto out;
4369                 case -NFS4ERR_WRONGSEC:
4370                         err = -EPERM;
4371                         if (client != *clnt)
4372                                 goto out;
4373                         client = nfs4_negotiate_security(client, dir, name);
4374                         if (IS_ERR(client))
4375                                 return PTR_ERR(client);
4376
4377                         exception.retry = 1;
4378                         break;
4379                 default:
4380                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4381                 }
4382         } while (exception.retry);
4383
4384 out:
4385         if (err == 0)
4386                 *clnt = client;
4387         else if (client != *clnt)
4388                 rpc_shutdown_client(client);
4389
4390         return err;
4391 }
4392
4393 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4394                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4395                             struct nfs4_label *label)
4396 {
4397         int status;
4398         struct rpc_clnt *client = NFS_CLIENT(dir);
4399
4400         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4401         if (client != NFS_CLIENT(dir)) {
4402                 rpc_shutdown_client(client);
4403                 nfs_fixup_secinfo_attributes(fattr);
4404         }
4405         return status;
4406 }
4407
4408 struct rpc_clnt *
4409 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4410                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4411 {
4412         struct rpc_clnt *client = NFS_CLIENT(dir);
4413         int status;
4414
4415         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4416         if (status < 0)
4417                 return ERR_PTR(status);
4418         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4419 }
4420
4421 static int _nfs4_proc_lookupp(struct inode *inode,
4422                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4423                 struct nfs4_label *label)
4424 {
4425         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4426         struct nfs_server *server = NFS_SERVER(inode);
4427         int                    status;
4428         struct nfs4_lookupp_arg args = {
4429                 .bitmask = server->attr_bitmask,
4430                 .fh = NFS_FH(inode),
4431         };
4432         struct nfs4_lookupp_res res = {
4433                 .server = server,
4434                 .fattr = fattr,
4435                 .label = label,
4436                 .fh = fhandle,
4437         };
4438         struct rpc_message msg = {
4439                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4440                 .rpc_argp = &args,
4441                 .rpc_resp = &res,
4442         };
4443         unsigned short task_flags = 0;
4444
4445         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4446                 task_flags |= RPC_TASK_TIMEOUT;
4447
4448         args.bitmask = nfs4_bitmask(server, label);
4449
4450         nfs_fattr_init(fattr);
4451
4452         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4453         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4454                                 &res.seq_res, task_flags);
4455         dprintk("NFS reply lookupp: %d\n", status);
4456         return status;
4457 }
4458
4459 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4460                              struct nfs_fattr *fattr, struct nfs4_label *label)
4461 {
4462         struct nfs4_exception exception = {
4463                 .interruptible = true,
4464         };
4465         int err;
4466         do {
4467                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4468                 trace_nfs4_lookupp(inode, err);
4469                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4470                                 &exception);
4471         } while (exception.retry);
4472         return err;
4473 }
4474
4475 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4476 {
4477         struct nfs_server *server = NFS_SERVER(inode);
4478         struct nfs4_accessargs args = {
4479                 .fh = NFS_FH(inode),
4480                 .access = entry->mask,
4481         };
4482         struct nfs4_accessres res = {
4483                 .server = server,
4484         };
4485         struct rpc_message msg = {
4486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4487                 .rpc_argp = &args,
4488                 .rpc_resp = &res,
4489                 .rpc_cred = entry->cred,
4490         };
4491         int status = 0;
4492
4493         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4494                 res.fattr = nfs_alloc_fattr();
4495                 if (res.fattr == NULL)
4496                         return -ENOMEM;
4497                 args.bitmask = server->cache_consistency_bitmask;
4498         }
4499         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4500         if (!status) {
4501                 nfs_access_set_mask(entry, res.access);
4502                 if (res.fattr)
4503                         nfs_refresh_inode(inode, res.fattr);
4504         }
4505         nfs_free_fattr(res.fattr);
4506         return status;
4507 }
4508
4509 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4510 {
4511         struct nfs4_exception exception = {
4512                 .interruptible = true,
4513         };
4514         int err;
4515         do {
4516                 err = _nfs4_proc_access(inode, entry);
4517                 trace_nfs4_access(inode, err);
4518                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4519                                 &exception);
4520         } while (exception.retry);
4521         return err;
4522 }
4523
4524 /*
4525  * TODO: For the time being, we don't try to get any attributes
4526  * along with any of the zero-copy operations READ, READDIR,
4527  * READLINK, WRITE.
4528  *
4529  * In the case of the first three, we want to put the GETATTR
4530  * after the read-type operation -- this is because it is hard
4531  * to predict the length of a GETATTR response in v4, and thus
4532  * align the READ data correctly.  This means that the GETATTR
4533  * may end up partially falling into the page cache, and we should
4534  * shift it into the 'tail' of the xdr_buf before processing.
4535  * To do this efficiently, we need to know the total length
4536  * of data received, which doesn't seem to be available outside
4537  * of the RPC layer.
4538  *
4539  * In the case of WRITE, we also want to put the GETATTR after
4540  * the operation -- in this case because we want to make sure
4541  * we get the post-operation mtime and size.
4542  *
4543  * Both of these changes to the XDR layer would in fact be quite
4544  * minor, but I decided to leave them for a subsequent patch.
4545  */
4546 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4547                 unsigned int pgbase, unsigned int pglen)
4548 {
4549         struct nfs4_readlink args = {
4550                 .fh       = NFS_FH(inode),
4551                 .pgbase   = pgbase,
4552                 .pglen    = pglen,
4553                 .pages    = &page,
4554         };
4555         struct nfs4_readlink_res res;
4556         struct rpc_message msg = {
4557                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4558                 .rpc_argp = &args,
4559                 .rpc_resp = &res,
4560         };
4561
4562         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4563 }
4564
4565 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4566                 unsigned int pgbase, unsigned int pglen)
4567 {
4568         struct nfs4_exception exception = {
4569                 .interruptible = true,
4570         };
4571         int err;
4572         do {
4573                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4574                 trace_nfs4_readlink(inode, err);
4575                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4576                                 &exception);
4577         } while (exception.retry);
4578         return err;
4579 }
4580
4581 /*
4582  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4583  */
4584 static int
4585 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4586                  int flags)
4587 {
4588         struct nfs_server *server = NFS_SERVER(dir);
4589         struct nfs4_label l, *ilabel = NULL;
4590         struct nfs_open_context *ctx;
4591         struct nfs4_state *state;
4592         int status = 0;
4593
4594         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4595         if (IS_ERR(ctx))
4596                 return PTR_ERR(ctx);
4597
4598         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4599
4600         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4601                 sattr->ia_mode &= ~current_umask();
4602         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4603         if (IS_ERR(state)) {
4604                 status = PTR_ERR(state);
4605                 goto out;
4606         }
4607 out:
4608         nfs4_label_release_security(ilabel);
4609         put_nfs_open_context(ctx);
4610         return status;
4611 }
4612
4613 static int
4614 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4615 {
4616         struct nfs_server *server = NFS_SERVER(dir);
4617         struct nfs_removeargs args = {
4618                 .fh = NFS_FH(dir),
4619                 .name = *name,
4620         };
4621         struct nfs_removeres res = {
4622                 .server = server,
4623         };
4624         struct rpc_message msg = {
4625                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4626                 .rpc_argp = &args,
4627                 .rpc_resp = &res,
4628         };
4629         unsigned long timestamp = jiffies;
4630         int status;
4631
4632         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4633         if (status == 0) {
4634                 spin_lock(&dir->i_lock);
4635                 /* Removing a directory decrements nlink in the parent */
4636                 if (ftype == NF4DIR && dir->i_nlink > 2)
4637                         nfs4_dec_nlink_locked(dir);
4638                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4639                                               NFS_INO_INVALID_DATA);
4640                 spin_unlock(&dir->i_lock);
4641         }
4642         return status;
4643 }
4644
4645 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4646 {
4647         struct nfs4_exception exception = {
4648                 .interruptible = true,
4649         };
4650         struct inode *inode = d_inode(dentry);
4651         int err;
4652
4653         if (inode) {
4654                 if (inode->i_nlink == 1)
4655                         nfs4_inode_return_delegation(inode);
4656                 else
4657                         nfs4_inode_make_writeable(inode);
4658         }
4659         do {
4660                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4661                 trace_nfs4_remove(dir, &dentry->d_name, err);
4662                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4663                                 &exception);
4664         } while (exception.retry);
4665         return err;
4666 }
4667
4668 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4669 {
4670         struct nfs4_exception exception = {
4671                 .interruptible = true,
4672         };
4673         int err;
4674
4675         do {
4676                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4677                 trace_nfs4_remove(dir, name, err);
4678                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4679                                 &exception);
4680         } while (exception.retry);
4681         return err;
4682 }
4683
4684 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4685                 struct dentry *dentry,
4686                 struct inode *inode)
4687 {
4688         struct nfs_removeargs *args = msg->rpc_argp;
4689         struct nfs_removeres *res = msg->rpc_resp;
4690
4691         res->server = NFS_SB(dentry->d_sb);
4692         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4693         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4694
4695         nfs_fattr_init(res->dir_attr);
4696
4697         if (inode)
4698                 nfs4_inode_return_delegation(inode);
4699 }
4700
4701 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4702 {
4703         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4704                         &data->args.seq_args,
4705                         &data->res.seq_res,
4706                         task);
4707 }
4708
4709 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4710 {
4711         struct nfs_unlinkdata *data = task->tk_calldata;
4712         struct nfs_removeres *res = &data->res;
4713
4714         if (!nfs4_sequence_done(task, &res->seq_res))
4715                 return 0;
4716         if (nfs4_async_handle_error(task, res->server, NULL,
4717                                     &data->timeout) == -EAGAIN)
4718                 return 0;
4719         if (task->tk_status == 0)
4720                 nfs4_update_changeattr(dir, &res->cinfo,
4721                                 res->dir_attr->time_start,
4722                                 NFS_INO_INVALID_DATA);
4723         return 1;
4724 }
4725
4726 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4727                 struct dentry *old_dentry,
4728                 struct dentry *new_dentry)
4729 {
4730         struct nfs_renameargs *arg = msg->rpc_argp;
4731         struct nfs_renameres *res = msg->rpc_resp;
4732         struct inode *old_inode = d_inode(old_dentry);
4733         struct inode *new_inode = d_inode(new_dentry);
4734
4735         if (old_inode)
4736                 nfs4_inode_make_writeable(old_inode);
4737         if (new_inode)
4738                 nfs4_inode_return_delegation(new_inode);
4739         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4740         res->server = NFS_SB(old_dentry->d_sb);
4741         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4742 }
4743
4744 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4745 {
4746         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4747                         &data->args.seq_args,
4748                         &data->res.seq_res,
4749                         task);
4750 }
4751
4752 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4753                                  struct inode *new_dir)
4754 {
4755         struct nfs_renamedata *data = task->tk_calldata;
4756         struct nfs_renameres *res = &data->res;
4757
4758         if (!nfs4_sequence_done(task, &res->seq_res))
4759                 return 0;
4760         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4761                 return 0;
4762
4763         if (task->tk_status == 0) {
4764                 if (new_dir != old_dir) {
4765                         /* Note: If we moved a directory, nlink will change */
4766                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4767                                         res->old_fattr->time_start,
4768                                         NFS_INO_INVALID_NLINK |
4769                                             NFS_INO_INVALID_DATA);
4770                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4771                                         res->new_fattr->time_start,
4772                                         NFS_INO_INVALID_NLINK |
4773                                             NFS_INO_INVALID_DATA);
4774                 } else
4775                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4776                                         res->old_fattr->time_start,
4777                                         NFS_INO_INVALID_DATA);
4778         }
4779         return 1;
4780 }
4781
4782 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4783 {
4784         struct nfs_server *server = NFS_SERVER(inode);
4785         __u32 bitmask[NFS4_BITMASK_SZ];
4786         struct nfs4_link_arg arg = {
4787                 .fh     = NFS_FH(inode),
4788                 .dir_fh = NFS_FH(dir),
4789                 .name   = name,
4790                 .bitmask = bitmask,
4791         };
4792         struct nfs4_link_res res = {
4793                 .server = server,
4794                 .label = NULL,
4795         };
4796         struct rpc_message msg = {
4797                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4798                 .rpc_argp = &arg,
4799                 .rpc_resp = &res,
4800         };
4801         int status = -ENOMEM;
4802
4803         res.fattr = nfs_alloc_fattr();
4804         if (res.fattr == NULL)
4805                 goto out;
4806
4807         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4808         if (IS_ERR(res.label)) {
4809                 status = PTR_ERR(res.label);
4810                 goto out;
4811         }
4812
4813         nfs4_inode_make_writeable(inode);
4814         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.label), inode,
4815                                 NFS_INO_INVALID_CHANGE);
4816         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4817         if (!status) {
4818                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4819                                        NFS_INO_INVALID_DATA);
4820                 nfs4_inc_nlink(inode);
4821                 status = nfs_post_op_update_inode(inode, res.fattr);
4822                 if (!status)
4823                         nfs_setsecurity(inode, res.fattr, res.label);
4824         }
4825
4826
4827         nfs4_label_free(res.label);
4828
4829 out:
4830         nfs_free_fattr(res.fattr);
4831         return status;
4832 }
4833
4834 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4835 {
4836         struct nfs4_exception exception = {
4837                 .interruptible = true,
4838         };
4839         int err;
4840         do {
4841                 err = nfs4_handle_exception(NFS_SERVER(inode),
4842                                 _nfs4_proc_link(inode, dir, name),
4843                                 &exception);
4844         } while (exception.retry);
4845         return err;
4846 }
4847
4848 struct nfs4_createdata {
4849         struct rpc_message msg;
4850         struct nfs4_create_arg arg;
4851         struct nfs4_create_res res;
4852         struct nfs_fh fh;
4853         struct nfs_fattr fattr;
4854         struct nfs4_label *label;
4855 };
4856
4857 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4858                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4859 {
4860         struct nfs4_createdata *data;
4861
4862         data = kzalloc(sizeof(*data), GFP_KERNEL);
4863         if (data != NULL) {
4864                 struct nfs_server *server = NFS_SERVER(dir);
4865
4866                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4867                 if (IS_ERR(data->label))
4868                         goto out_free;
4869
4870                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4871                 data->msg.rpc_argp = &data->arg;
4872                 data->msg.rpc_resp = &data->res;
4873                 data->arg.dir_fh = NFS_FH(dir);
4874                 data->arg.server = server;
4875                 data->arg.name = name;
4876                 data->arg.attrs = sattr;
4877                 data->arg.ftype = ftype;
4878                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4879                 data->arg.umask = current_umask();
4880                 data->res.server = server;
4881                 data->res.fh = &data->fh;
4882                 data->res.fattr = &data->fattr;
4883                 data->res.label = data->label;
4884                 nfs_fattr_init(data->res.fattr);
4885         }
4886         return data;
4887 out_free:
4888         kfree(data);
4889         return NULL;
4890 }
4891
4892 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4893 {
4894         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4895                                     &data->arg.seq_args, &data->res.seq_res, 1);
4896         if (status == 0) {
4897                 spin_lock(&dir->i_lock);
4898                 /* Creating a directory bumps nlink in the parent */
4899                 if (data->arg.ftype == NF4DIR)
4900                         nfs4_inc_nlink_locked(dir);
4901                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4902                                               data->res.fattr->time_start,
4903                                               NFS_INO_INVALID_DATA);
4904                 spin_unlock(&dir->i_lock);
4905                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4906         }
4907         return status;
4908 }
4909
4910 static void nfs4_free_createdata(struct nfs4_createdata *data)
4911 {
4912         nfs4_label_free(data->label);
4913         kfree(data);
4914 }
4915
4916 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4917                 struct page *page, unsigned int len, struct iattr *sattr,
4918                 struct nfs4_label *label)
4919 {
4920         struct nfs4_createdata *data;
4921         int status = -ENAMETOOLONG;
4922
4923         if (len > NFS4_MAXPATHLEN)
4924                 goto out;
4925
4926         status = -ENOMEM;
4927         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4928         if (data == NULL)
4929                 goto out;
4930
4931         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4932         data->arg.u.symlink.pages = &page;
4933         data->arg.u.symlink.len = len;
4934         data->arg.label = label;
4935
4936         status = nfs4_do_create(dir, dentry, data);
4937
4938         nfs4_free_createdata(data);
4939 out:
4940         return status;
4941 }
4942
4943 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4944                 struct page *page, unsigned int len, struct iattr *sattr)
4945 {
4946         struct nfs4_exception exception = {
4947                 .interruptible = true,
4948         };
4949         struct nfs4_label l, *label = NULL;
4950         int err;
4951
4952         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4953
4954         do {
4955                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4956                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4957                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4958                                 &exception);
4959         } while (exception.retry);
4960
4961         nfs4_label_release_security(label);
4962         return err;
4963 }
4964
4965 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4966                 struct iattr *sattr, struct nfs4_label *label)
4967 {
4968         struct nfs4_createdata *data;
4969         int status = -ENOMEM;
4970
4971         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4972         if (data == NULL)
4973                 goto out;
4974
4975         data->arg.label = label;
4976         status = nfs4_do_create(dir, dentry, data);
4977
4978         nfs4_free_createdata(data);
4979 out:
4980         return status;
4981 }
4982
4983 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4984                 struct iattr *sattr)
4985 {
4986         struct nfs_server *server = NFS_SERVER(dir);
4987         struct nfs4_exception exception = {
4988                 .interruptible = true,
4989         };
4990         struct nfs4_label l, *label = NULL;
4991         int err;
4992
4993         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4994
4995         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4996                 sattr->ia_mode &= ~current_umask();
4997         do {
4998                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4999                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5000                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5001                                 &exception);
5002         } while (exception.retry);
5003         nfs4_label_release_security(label);
5004
5005         return err;
5006 }
5007
5008 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5009                               struct nfs_readdir_res *nr_res)
5010 {
5011         struct inode            *dir = d_inode(nr_arg->dentry);
5012         struct nfs_server       *server = NFS_SERVER(dir);
5013         struct nfs4_readdir_arg args = {
5014                 .fh = NFS_FH(dir),
5015                 .pages = nr_arg->pages,
5016                 .pgbase = 0,
5017                 .count = nr_arg->page_len,
5018                 .plus = nr_arg->plus,
5019         };
5020         struct nfs4_readdir_res res;
5021         struct rpc_message msg = {
5022                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5023                 .rpc_argp = &args,
5024                 .rpc_resp = &res,
5025                 .rpc_cred = nr_arg->cred,
5026         };
5027         int                     status;
5028
5029         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5030                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5031         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5032                 args.bitmask = server->attr_bitmask_nl;
5033         else
5034                 args.bitmask = server->attr_bitmask;
5035
5036         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5037         res.pgbase = args.pgbase;
5038         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5039                         &res.seq_res, 0);
5040         if (status >= 0) {
5041                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5042                 status += args.pgbase;
5043         }
5044
5045         nfs_invalidate_atime(dir);
5046
5047         dprintk("%s: returns %d\n", __func__, status);
5048         return status;
5049 }
5050
5051 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5052                              struct nfs_readdir_res *res)
5053 {
5054         struct nfs4_exception exception = {
5055                 .interruptible = true,
5056         };
5057         int err;
5058         do {
5059                 err = _nfs4_proc_readdir(arg, res);
5060                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5061                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5062                                             err, &exception);
5063         } while (exception.retry);
5064         return err;
5065 }
5066
5067 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5068                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5069 {
5070         struct nfs4_createdata *data;
5071         int mode = sattr->ia_mode;
5072         int status = -ENOMEM;
5073
5074         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5075         if (data == NULL)
5076                 goto out;
5077
5078         if (S_ISFIFO(mode))
5079                 data->arg.ftype = NF4FIFO;
5080         else if (S_ISBLK(mode)) {
5081                 data->arg.ftype = NF4BLK;
5082                 data->arg.u.device.specdata1 = MAJOR(rdev);
5083                 data->arg.u.device.specdata2 = MINOR(rdev);
5084         }
5085         else if (S_ISCHR(mode)) {
5086                 data->arg.ftype = NF4CHR;
5087                 data->arg.u.device.specdata1 = MAJOR(rdev);
5088                 data->arg.u.device.specdata2 = MINOR(rdev);
5089         } else if (!S_ISSOCK(mode)) {
5090                 status = -EINVAL;
5091                 goto out_free;
5092         }
5093
5094         data->arg.label = label;
5095         status = nfs4_do_create(dir, dentry, data);
5096 out_free:
5097         nfs4_free_createdata(data);
5098 out:
5099         return status;
5100 }
5101
5102 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5103                 struct iattr *sattr, dev_t rdev)
5104 {
5105         struct nfs_server *server = NFS_SERVER(dir);
5106         struct nfs4_exception exception = {
5107                 .interruptible = true,
5108         };
5109         struct nfs4_label l, *label = NULL;
5110         int err;
5111
5112         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5113
5114         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5115                 sattr->ia_mode &= ~current_umask();
5116         do {
5117                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5118                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5119                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5120                                 &exception);
5121         } while (exception.retry);
5122
5123         nfs4_label_release_security(label);
5124
5125         return err;
5126 }
5127
5128 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5129                  struct nfs_fsstat *fsstat)
5130 {
5131         struct nfs4_statfs_arg args = {
5132                 .fh = fhandle,
5133                 .bitmask = server->attr_bitmask,
5134         };
5135         struct nfs4_statfs_res res = {
5136                 .fsstat = fsstat,
5137         };
5138         struct rpc_message msg = {
5139                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5140                 .rpc_argp = &args,
5141                 .rpc_resp = &res,
5142         };
5143
5144         nfs_fattr_init(fsstat->fattr);
5145         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5146 }
5147
5148 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5149 {
5150         struct nfs4_exception exception = {
5151                 .interruptible = true,
5152         };
5153         int err;
5154         do {
5155                 err = nfs4_handle_exception(server,
5156                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5157                                 &exception);
5158         } while (exception.retry);
5159         return err;
5160 }
5161
5162 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5163                 struct nfs_fsinfo *fsinfo)
5164 {
5165         struct nfs4_fsinfo_arg args = {
5166                 .fh = fhandle,
5167                 .bitmask = server->attr_bitmask,
5168         };
5169         struct nfs4_fsinfo_res res = {
5170                 .fsinfo = fsinfo,
5171         };
5172         struct rpc_message msg = {
5173                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5174                 .rpc_argp = &args,
5175                 .rpc_resp = &res,
5176         };
5177
5178         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5179 }
5180
5181 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5182 {
5183         struct nfs4_exception exception = {
5184                 .interruptible = true,
5185         };
5186         int err;
5187
5188         do {
5189                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5190                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5191                 if (err == 0) {
5192                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5193                         break;
5194                 }
5195                 err = nfs4_handle_exception(server, err, &exception);
5196         } while (exception.retry);
5197         return err;
5198 }
5199
5200 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5201 {
5202         int error;
5203
5204         nfs_fattr_init(fsinfo->fattr);
5205         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5206         if (error == 0) {
5207                 /* block layout checks this! */
5208                 server->pnfs_blksize = fsinfo->blksize;
5209                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5210         }
5211
5212         return error;
5213 }
5214
5215 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5216                 struct nfs_pathconf *pathconf)
5217 {
5218         struct nfs4_pathconf_arg args = {
5219                 .fh = fhandle,
5220                 .bitmask = server->attr_bitmask,
5221         };
5222         struct nfs4_pathconf_res res = {
5223                 .pathconf = pathconf,
5224         };
5225         struct rpc_message msg = {
5226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5227                 .rpc_argp = &args,
5228                 .rpc_resp = &res,
5229         };
5230
5231         /* None of the pathconf attributes are mandatory to implement */
5232         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5233                 memset(pathconf, 0, sizeof(*pathconf));
5234                 return 0;
5235         }
5236
5237         nfs_fattr_init(pathconf->fattr);
5238         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5239 }
5240
5241 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5242                 struct nfs_pathconf *pathconf)
5243 {
5244         struct nfs4_exception exception = {
5245                 .interruptible = true,
5246         };
5247         int err;
5248
5249         do {
5250                 err = nfs4_handle_exception(server,
5251                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5252                                 &exception);
5253         } while (exception.retry);
5254         return err;
5255 }
5256
5257 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5258                 const struct nfs_open_context *ctx,
5259                 const struct nfs_lock_context *l_ctx,
5260                 fmode_t fmode)
5261 {
5262         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5263 }
5264 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5265
5266 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5267                 const struct nfs_open_context *ctx,
5268                 const struct nfs_lock_context *l_ctx,
5269                 fmode_t fmode)
5270 {
5271         nfs4_stateid _current_stateid;
5272
5273         /* If the current stateid represents a lost lock, then exit */
5274         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5275                 return true;
5276         return nfs4_stateid_match(stateid, &_current_stateid);
5277 }
5278
5279 static bool nfs4_error_stateid_expired(int err)
5280 {
5281         switch (err) {
5282         case -NFS4ERR_DELEG_REVOKED:
5283         case -NFS4ERR_ADMIN_REVOKED:
5284         case -NFS4ERR_BAD_STATEID:
5285         case -NFS4ERR_STALE_STATEID:
5286         case -NFS4ERR_OLD_STATEID:
5287         case -NFS4ERR_OPENMODE:
5288         case -NFS4ERR_EXPIRED:
5289                 return true;
5290         }
5291         return false;
5292 }
5293
5294 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5295 {
5296         struct nfs_server *server = NFS_SERVER(hdr->inode);
5297
5298         trace_nfs4_read(hdr, task->tk_status);
5299         if (task->tk_status < 0) {
5300                 struct nfs4_exception exception = {
5301                         .inode = hdr->inode,
5302                         .state = hdr->args.context->state,
5303                         .stateid = &hdr->args.stateid,
5304                 };
5305                 task->tk_status = nfs4_async_handle_exception(task,
5306                                 server, task->tk_status, &exception);
5307                 if (exception.retry) {
5308                         rpc_restart_call_prepare(task);
5309                         return -EAGAIN;
5310                 }
5311         }
5312
5313         if (task->tk_status > 0)
5314                 renew_lease(server, hdr->timestamp);
5315         return 0;
5316 }
5317
5318 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5319                 struct nfs_pgio_args *args)
5320 {
5321
5322         if (!nfs4_error_stateid_expired(task->tk_status) ||
5323                 nfs4_stateid_is_current(&args->stateid,
5324                                 args->context,
5325                                 args->lock_context,
5326                                 FMODE_READ))
5327                 return false;
5328         rpc_restart_call_prepare(task);
5329         return true;
5330 }
5331
5332 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5333                                          struct nfs_pgio_header *hdr)
5334 {
5335         struct nfs_server *server = NFS_SERVER(hdr->inode);
5336         struct rpc_message *msg = &task->tk_msg;
5337
5338         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5339             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5340                 server->caps &= ~NFS_CAP_READ_PLUS;
5341                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5342                 rpc_restart_call_prepare(task);
5343                 return true;
5344         }
5345         return false;
5346 }
5347
5348 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5349 {
5350         dprintk("--> %s\n", __func__);
5351
5352         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5353                 return -EAGAIN;
5354         if (nfs4_read_stateid_changed(task, &hdr->args))
5355                 return -EAGAIN;
5356         if (nfs4_read_plus_not_supported(task, hdr))
5357                 return -EAGAIN;
5358         if (task->tk_status > 0)
5359                 nfs_invalidate_atime(hdr->inode);
5360         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5361                                     nfs4_read_done_cb(task, hdr);
5362 }
5363
5364 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5365 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5366                                     struct rpc_message *msg)
5367 {
5368         /* Note: We don't use READ_PLUS with pNFS yet */
5369         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5370                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5371 }
5372 #else
5373 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5374                                     struct rpc_message *msg)
5375 {
5376 }
5377 #endif /* CONFIG_NFS_V4_2 */
5378
5379 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5380                                  struct rpc_message *msg)
5381 {
5382         hdr->timestamp   = jiffies;
5383         if (!hdr->pgio_done_cb)
5384                 hdr->pgio_done_cb = nfs4_read_done_cb;
5385         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5386         nfs42_read_plus_support(hdr, msg);
5387         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5388 }
5389
5390 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5391                                       struct nfs_pgio_header *hdr)
5392 {
5393         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5394                         &hdr->args.seq_args,
5395                         &hdr->res.seq_res,
5396                         task))
5397                 return 0;
5398         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5399                                 hdr->args.lock_context,
5400                                 hdr->rw_mode) == -EIO)
5401                 return -EIO;
5402         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5403                 return -EIO;
5404         return 0;
5405 }
5406
5407 static int nfs4_write_done_cb(struct rpc_task *task,
5408                               struct nfs_pgio_header *hdr)
5409 {
5410         struct inode *inode = hdr->inode;
5411
5412         trace_nfs4_write(hdr, task->tk_status);
5413         if (task->tk_status < 0) {
5414                 struct nfs4_exception exception = {
5415                         .inode = hdr->inode,
5416                         .state = hdr->args.context->state,
5417                         .stateid = &hdr->args.stateid,
5418                 };
5419                 task->tk_status = nfs4_async_handle_exception(task,
5420                                 NFS_SERVER(inode), task->tk_status,
5421                                 &exception);
5422                 if (exception.retry) {
5423                         rpc_restart_call_prepare(task);
5424                         return -EAGAIN;
5425                 }
5426         }
5427         if (task->tk_status >= 0) {
5428                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5429                 nfs_writeback_update_inode(hdr);
5430         }
5431         return 0;
5432 }
5433
5434 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5435                 struct nfs_pgio_args *args)
5436 {
5437
5438         if (!nfs4_error_stateid_expired(task->tk_status) ||
5439                 nfs4_stateid_is_current(&args->stateid,
5440                                 args->context,
5441                                 args->lock_context,
5442                                 FMODE_WRITE))
5443                 return false;
5444         rpc_restart_call_prepare(task);
5445         return true;
5446 }
5447
5448 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5449 {
5450         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5451                 return -EAGAIN;
5452         if (nfs4_write_stateid_changed(task, &hdr->args))
5453                 return -EAGAIN;
5454         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5455                 nfs4_write_done_cb(task, hdr);
5456 }
5457
5458 static
5459 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5460 {
5461         /* Don't request attributes for pNFS or O_DIRECT writes */
5462         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5463                 return false;
5464         /* Otherwise, request attributes if and only if we don't hold
5465          * a delegation
5466          */
5467         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5468 }
5469
5470 static void nfs4_bitmask_set(__u32 bitmask[NFS4_BITMASK_SZ], const __u32 *src,
5471                              struct inode *inode, struct nfs_server *server,
5472                              struct nfs4_label *label)
5473 {
5474         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5475         unsigned int i;
5476
5477         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5478
5479         if (cache_validity & NFS_INO_INVALID_CHANGE)
5480                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5481         if (cache_validity & NFS_INO_INVALID_ATIME)
5482                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5483         if (cache_validity & NFS_INO_INVALID_MODE)
5484                 bitmask[1] |= FATTR4_WORD1_MODE;
5485         if (cache_validity & NFS_INO_INVALID_OTHER)
5486                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5487         if (cache_validity & NFS_INO_INVALID_NLINK)
5488                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5489         if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL)
5490                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5491         if (cache_validity & NFS_INO_INVALID_CTIME)
5492                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5493         if (cache_validity & NFS_INO_INVALID_MTIME)
5494                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5495         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5496                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5497
5498         if (cache_validity & NFS_INO_INVALID_SIZE)
5499                 bitmask[0] |= FATTR4_WORD0_SIZE;
5500
5501         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5502                 bitmask[i] &= server->attr_bitmask[i];
5503 }
5504
5505 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5506                                   struct rpc_message *msg,
5507                                   struct rpc_clnt **clnt)
5508 {
5509         struct nfs_server *server = NFS_SERVER(hdr->inode);
5510
5511         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5512                 hdr->args.bitmask = NULL;
5513                 hdr->res.fattr = NULL;
5514         } else {
5515                 nfs4_bitmask_set(hdr->args.bitmask_store,
5516                                  server->cache_consistency_bitmask,
5517                                  hdr->inode, server, NULL);
5518                 hdr->args.bitmask = hdr->args.bitmask_store;
5519         }
5520
5521         if (!hdr->pgio_done_cb)
5522                 hdr->pgio_done_cb = nfs4_write_done_cb;
5523         hdr->res.server = server;
5524         hdr->timestamp   = jiffies;
5525
5526         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5527         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5528         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5529 }
5530
5531 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5532 {
5533         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5534                         &data->args.seq_args,
5535                         &data->res.seq_res,
5536                         task);
5537 }
5538
5539 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5540 {
5541         struct inode *inode = data->inode;
5542
5543         trace_nfs4_commit(data, task->tk_status);
5544         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5545                                     NULL, NULL) == -EAGAIN) {
5546                 rpc_restart_call_prepare(task);
5547                 return -EAGAIN;
5548         }
5549         return 0;
5550 }
5551
5552 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5553 {
5554         if (!nfs4_sequence_done(task, &data->res.seq_res))
5555                 return -EAGAIN;
5556         return data->commit_done_cb(task, data);
5557 }
5558
5559 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5560                                    struct rpc_clnt **clnt)
5561 {
5562         struct nfs_server *server = NFS_SERVER(data->inode);
5563
5564         if (data->commit_done_cb == NULL)
5565                 data->commit_done_cb = nfs4_commit_done_cb;
5566         data->res.server = server;
5567         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5568         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5569         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5570 }
5571
5572 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5573                                 struct nfs_commitres *res)
5574 {
5575         struct inode *dst_inode = file_inode(dst);
5576         struct nfs_server *server = NFS_SERVER(dst_inode);
5577         struct rpc_message msg = {
5578                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5579                 .rpc_argp = args,
5580                 .rpc_resp = res,
5581         };
5582
5583         args->fh = NFS_FH(dst_inode);
5584         return nfs4_call_sync(server->client, server, &msg,
5585                         &args->seq_args, &res->seq_res, 1);
5586 }
5587
5588 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5589 {
5590         struct nfs_commitargs args = {
5591                 .offset = offset,
5592                 .count = count,
5593         };
5594         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5595         struct nfs4_exception exception = { };
5596         int status;
5597
5598         do {
5599                 status = _nfs4_proc_commit(dst, &args, res);
5600                 status = nfs4_handle_exception(dst_server, status, &exception);
5601         } while (exception.retry);
5602
5603         return status;
5604 }
5605
5606 struct nfs4_renewdata {
5607         struct nfs_client       *client;
5608         unsigned long           timestamp;
5609 };
5610
5611 /*
5612  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5613  * standalone procedure for queueing an asynchronous RENEW.
5614  */
5615 static void nfs4_renew_release(void *calldata)
5616 {
5617         struct nfs4_renewdata *data = calldata;
5618         struct nfs_client *clp = data->client;
5619
5620         if (refcount_read(&clp->cl_count) > 1)
5621                 nfs4_schedule_state_renewal(clp);
5622         nfs_put_client(clp);
5623         kfree(data);
5624 }
5625
5626 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5627 {
5628         struct nfs4_renewdata *data = calldata;
5629         struct nfs_client *clp = data->client;
5630         unsigned long timestamp = data->timestamp;
5631
5632         trace_nfs4_renew_async(clp, task->tk_status);
5633         switch (task->tk_status) {
5634         case 0:
5635                 break;
5636         case -NFS4ERR_LEASE_MOVED:
5637                 nfs4_schedule_lease_moved_recovery(clp);
5638                 break;
5639         default:
5640                 /* Unless we're shutting down, schedule state recovery! */
5641                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5642                         return;
5643                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5644                         nfs4_schedule_lease_recovery(clp);
5645                         return;
5646                 }
5647                 nfs4_schedule_path_down_recovery(clp);
5648         }
5649         do_renew_lease(clp, timestamp);
5650 }
5651
5652 static const struct rpc_call_ops nfs4_renew_ops = {
5653         .rpc_call_done = nfs4_renew_done,
5654         .rpc_release = nfs4_renew_release,
5655 };
5656
5657 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5658 {
5659         struct rpc_message msg = {
5660                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5661                 .rpc_argp       = clp,
5662                 .rpc_cred       = cred,
5663         };
5664         struct nfs4_renewdata *data;
5665
5666         if (renew_flags == 0)
5667                 return 0;
5668         if (!refcount_inc_not_zero(&clp->cl_count))
5669                 return -EIO;
5670         data = kmalloc(sizeof(*data), GFP_NOFS);
5671         if (data == NULL) {
5672                 nfs_put_client(clp);
5673                 return -ENOMEM;
5674         }
5675         data->client = clp;
5676         data->timestamp = jiffies;
5677         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5678                         &nfs4_renew_ops, data);
5679 }
5680
5681 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5682 {
5683         struct rpc_message msg = {
5684                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5685                 .rpc_argp       = clp,
5686                 .rpc_cred       = cred,
5687         };
5688         unsigned long now = jiffies;
5689         int status;
5690
5691         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5692         if (status < 0)
5693                 return status;
5694         do_renew_lease(clp, now);
5695         return 0;
5696 }
5697
5698 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5699 {
5700         return server->caps & NFS_CAP_ACLS;
5701 }
5702
5703 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5704  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5705  * the stack.
5706  */
5707 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5708
5709 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5710                 struct page **pages)
5711 {
5712         struct page *newpage, **spages;
5713         int rc = 0;
5714         size_t len;
5715         spages = pages;
5716
5717         do {
5718                 len = min_t(size_t, PAGE_SIZE, buflen);
5719                 newpage = alloc_page(GFP_KERNEL);
5720
5721                 if (newpage == NULL)
5722                         goto unwind;
5723                 memcpy(page_address(newpage), buf, len);
5724                 buf += len;
5725                 buflen -= len;
5726                 *pages++ = newpage;
5727                 rc++;
5728         } while (buflen != 0);
5729
5730         return rc;
5731
5732 unwind:
5733         for(; rc > 0; rc--)
5734                 __free_page(spages[rc-1]);
5735         return -ENOMEM;
5736 }
5737
5738 struct nfs4_cached_acl {
5739         int cached;
5740         size_t len;
5741         char data[];
5742 };
5743
5744 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5745 {
5746         struct nfs_inode *nfsi = NFS_I(inode);
5747
5748         spin_lock(&inode->i_lock);
5749         kfree(nfsi->nfs4_acl);
5750         nfsi->nfs4_acl = acl;
5751         spin_unlock(&inode->i_lock);
5752 }
5753
5754 static void nfs4_zap_acl_attr(struct inode *inode)
5755 {
5756         nfs4_set_cached_acl(inode, NULL);
5757 }
5758
5759 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5760 {
5761         struct nfs_inode *nfsi = NFS_I(inode);
5762         struct nfs4_cached_acl *acl;
5763         int ret = -ENOENT;
5764
5765         spin_lock(&inode->i_lock);
5766         acl = nfsi->nfs4_acl;
5767         if (acl == NULL)
5768                 goto out;
5769         if (buf == NULL) /* user is just asking for length */
5770                 goto out_len;
5771         if (acl->cached == 0)
5772                 goto out;
5773         ret = -ERANGE; /* see getxattr(2) man page */
5774         if (acl->len > buflen)
5775                 goto out;
5776         memcpy(buf, acl->data, acl->len);
5777 out_len:
5778         ret = acl->len;
5779 out:
5780         spin_unlock(&inode->i_lock);
5781         return ret;
5782 }
5783
5784 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5785 {
5786         struct nfs4_cached_acl *acl;
5787         size_t buflen = sizeof(*acl) + acl_len;
5788
5789         if (buflen <= PAGE_SIZE) {
5790                 acl = kmalloc(buflen, GFP_KERNEL);
5791                 if (acl == NULL)
5792                         goto out;
5793                 acl->cached = 1;
5794                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5795         } else {
5796                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5797                 if (acl == NULL)
5798                         goto out;
5799                 acl->cached = 0;
5800         }
5801         acl->len = acl_len;
5802 out:
5803         nfs4_set_cached_acl(inode, acl);
5804 }
5805
5806 /*
5807  * The getxattr API returns the required buffer length when called with a
5808  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5809  * the required buf.  On a NULL buf, we send a page of data to the server
5810  * guessing that the ACL request can be serviced by a page. If so, we cache
5811  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5812  * the cache. If not so, we throw away the page, and cache the required
5813  * length. The next getxattr call will then produce another round trip to
5814  * the server, this time with the input buf of the required size.
5815  */
5816 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5817 {
5818         struct page **pages;
5819         struct nfs_getaclargs args = {
5820                 .fh = NFS_FH(inode),
5821                 .acl_len = buflen,
5822         };
5823         struct nfs_getaclres res = {
5824                 .acl_len = buflen,
5825         };
5826         struct rpc_message msg = {
5827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5828                 .rpc_argp = &args,
5829                 .rpc_resp = &res,
5830         };
5831         unsigned int npages;
5832         int ret = -ENOMEM, i;
5833         struct nfs_server *server = NFS_SERVER(inode);
5834
5835         if (buflen == 0)
5836                 buflen = server->rsize;
5837
5838         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5839         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5840         if (!pages)
5841                 return -ENOMEM;
5842
5843         args.acl_pages = pages;
5844
5845         for (i = 0; i < npages; i++) {
5846                 pages[i] = alloc_page(GFP_KERNEL);
5847                 if (!pages[i])
5848                         goto out_free;
5849         }
5850
5851         /* for decoding across pages */
5852         res.acl_scratch = alloc_page(GFP_KERNEL);
5853         if (!res.acl_scratch)
5854                 goto out_free;
5855
5856         args.acl_len = npages * PAGE_SIZE;
5857
5858         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5859                 __func__, buf, buflen, npages, args.acl_len);
5860         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5861                              &msg, &args.seq_args, &res.seq_res, 0);
5862         if (ret)
5863                 goto out_free;
5864
5865         /* Handle the case where the passed-in buffer is too short */
5866         if (res.acl_flags & NFS4_ACL_TRUNC) {
5867                 /* Did the user only issue a request for the acl length? */
5868                 if (buf == NULL)
5869                         goto out_ok;
5870                 ret = -ERANGE;
5871                 goto out_free;
5872         }
5873         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5874         if (buf) {
5875                 if (res.acl_len > buflen) {
5876                         ret = -ERANGE;
5877                         goto out_free;
5878                 }
5879                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5880         }
5881 out_ok:
5882         ret = res.acl_len;
5883 out_free:
5884         for (i = 0; i < npages; i++)
5885                 if (pages[i])
5886                         __free_page(pages[i]);
5887         if (res.acl_scratch)
5888                 __free_page(res.acl_scratch);
5889         kfree(pages);
5890         return ret;
5891 }
5892
5893 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5894 {
5895         struct nfs4_exception exception = {
5896                 .interruptible = true,
5897         };
5898         ssize_t ret;
5899         do {
5900                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5901                 trace_nfs4_get_acl(inode, ret);
5902                 if (ret >= 0)
5903                         break;
5904                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5905         } while (exception.retry);
5906         return ret;
5907 }
5908
5909 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5910 {
5911         struct nfs_server *server = NFS_SERVER(inode);
5912         int ret;
5913
5914         if (!nfs4_server_supports_acls(server))
5915                 return -EOPNOTSUPP;
5916         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
5917         if (ret < 0)
5918                 return ret;
5919         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5920                 nfs_zap_acl_cache(inode);
5921         ret = nfs4_read_cached_acl(inode, buf, buflen);
5922         if (ret != -ENOENT)
5923                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5924                  * but no cached acl data, just the acl length */
5925                 return ret;
5926         return nfs4_get_acl_uncached(inode, buf, buflen);
5927 }
5928
5929 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5930 {
5931         struct nfs_server *server = NFS_SERVER(inode);
5932         struct page *pages[NFS4ACL_MAXPAGES];
5933         struct nfs_setaclargs arg = {
5934                 .fh             = NFS_FH(inode),
5935                 .acl_pages      = pages,
5936                 .acl_len        = buflen,
5937         };
5938         struct nfs_setaclres res;
5939         struct rpc_message msg = {
5940                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5941                 .rpc_argp       = &arg,
5942                 .rpc_resp       = &res,
5943         };
5944         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5945         int ret, i;
5946
5947         /* You can't remove system.nfs4_acl: */
5948         if (buflen == 0)
5949                 return -EINVAL;
5950         if (!nfs4_server_supports_acls(server))
5951                 return -EOPNOTSUPP;
5952         if (npages > ARRAY_SIZE(pages))
5953                 return -ERANGE;
5954         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5955         if (i < 0)
5956                 return i;
5957         nfs4_inode_make_writeable(inode);
5958         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5959
5960         /*
5961          * Free each page after tx, so the only ref left is
5962          * held by the network stack
5963          */
5964         for (; i > 0; i--)
5965                 put_page(pages[i-1]);
5966
5967         /*
5968          * Acl update can result in inode attribute update.
5969          * so mark the attribute cache invalid.
5970          */
5971         spin_lock(&inode->i_lock);
5972         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
5973                                              NFS_INO_INVALID_CTIME |
5974                                              NFS_INO_REVAL_FORCED);
5975         spin_unlock(&inode->i_lock);
5976         nfs_access_zap_cache(inode);
5977         nfs_zap_acl_cache(inode);
5978         return ret;
5979 }
5980
5981 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5982 {
5983         struct nfs4_exception exception = { };
5984         int err;
5985         do {
5986                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5987                 trace_nfs4_set_acl(inode, err);
5988                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
5989                         /*
5990                          * no need to retry since the kernel
5991                          * isn't involved in encoding the ACEs.
5992                          */
5993                         err = -EINVAL;
5994                         break;
5995                 }
5996                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5997                                 &exception);
5998         } while (exception.retry);
5999         return err;
6000 }
6001
6002 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6003 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6004                                         size_t buflen)
6005 {
6006         struct nfs_server *server = NFS_SERVER(inode);
6007         struct nfs_fattr fattr;
6008         struct nfs4_label label = {0, 0, buflen, buf};
6009
6010         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6011         struct nfs4_getattr_arg arg = {
6012                 .fh             = NFS_FH(inode),
6013                 .bitmask        = bitmask,
6014         };
6015         struct nfs4_getattr_res res = {
6016                 .fattr          = &fattr,
6017                 .label          = &label,
6018                 .server         = server,
6019         };
6020         struct rpc_message msg = {
6021                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6022                 .rpc_argp       = &arg,
6023                 .rpc_resp       = &res,
6024         };
6025         int ret;
6026
6027         nfs_fattr_init(&fattr);
6028
6029         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6030         if (ret)
6031                 return ret;
6032         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6033                 return -ENOENT;
6034         return label.len;
6035 }
6036
6037 static int nfs4_get_security_label(struct inode *inode, void *buf,
6038                                         size_t buflen)
6039 {
6040         struct nfs4_exception exception = {
6041                 .interruptible = true,
6042         };
6043         int err;
6044
6045         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6046                 return -EOPNOTSUPP;
6047
6048         do {
6049                 err = _nfs4_get_security_label(inode, buf, buflen);
6050                 trace_nfs4_get_security_label(inode, err);
6051                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6052                                 &exception);
6053         } while (exception.retry);
6054         return err;
6055 }
6056
6057 static int _nfs4_do_set_security_label(struct inode *inode,
6058                 struct nfs4_label *ilabel,
6059                 struct nfs_fattr *fattr,
6060                 struct nfs4_label *olabel)
6061 {
6062
6063         struct iattr sattr = {0};
6064         struct nfs_server *server = NFS_SERVER(inode);
6065         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6066         struct nfs_setattrargs arg = {
6067                 .fh             = NFS_FH(inode),
6068                 .iap            = &sattr,
6069                 .server         = server,
6070                 .bitmask        = bitmask,
6071                 .label          = ilabel,
6072         };
6073         struct nfs_setattrres res = {
6074                 .fattr          = fattr,
6075                 .label          = olabel,
6076                 .server         = server,
6077         };
6078         struct rpc_message msg = {
6079                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6080                 .rpc_argp       = &arg,
6081                 .rpc_resp       = &res,
6082         };
6083         int status;
6084
6085         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6086
6087         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6088         if (status)
6089                 dprintk("%s failed: %d\n", __func__, status);
6090
6091         return status;
6092 }
6093
6094 static int nfs4_do_set_security_label(struct inode *inode,
6095                 struct nfs4_label *ilabel,
6096                 struct nfs_fattr *fattr,
6097                 struct nfs4_label *olabel)
6098 {
6099         struct nfs4_exception exception = { };
6100         int err;
6101
6102         do {
6103                 err = _nfs4_do_set_security_label(inode, ilabel,
6104                                 fattr, olabel);
6105                 trace_nfs4_set_security_label(inode, err);
6106                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6107                                 &exception);
6108         } while (exception.retry);
6109         return err;
6110 }
6111
6112 static int
6113 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6114 {
6115         struct nfs4_label ilabel, *olabel = NULL;
6116         struct nfs_fattr fattr;
6117         int status;
6118
6119         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6120                 return -EOPNOTSUPP;
6121
6122         nfs_fattr_init(&fattr);
6123
6124         ilabel.pi = 0;
6125         ilabel.lfs = 0;
6126         ilabel.label = (char *)buf;
6127         ilabel.len = buflen;
6128
6129         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6130         if (IS_ERR(olabel)) {
6131                 status = -PTR_ERR(olabel);
6132                 goto out;
6133         }
6134
6135         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6136         if (status == 0)
6137                 nfs_setsecurity(inode, &fattr, olabel);
6138
6139         nfs4_label_free(olabel);
6140 out:
6141         return status;
6142 }
6143 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6144
6145
6146 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6147                                     nfs4_verifier *bootverf)
6148 {
6149         __be32 verf[2];
6150
6151         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6152                 /* An impossible timestamp guarantees this value
6153                  * will never match a generated boot time. */
6154                 verf[0] = cpu_to_be32(U32_MAX);
6155                 verf[1] = cpu_to_be32(U32_MAX);
6156         } else {
6157                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6158                 u64 ns = ktime_to_ns(nn->boot_time);
6159
6160                 verf[0] = cpu_to_be32(ns >> 32);
6161                 verf[1] = cpu_to_be32(ns);
6162         }
6163         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6164 }
6165
6166 static size_t
6167 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6168 {
6169         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6170         struct nfs_netns_client *nn_clp = nn->nfs_client;
6171         const char *id;
6172
6173         buf[0] = '\0';
6174
6175         if (nn_clp) {
6176                 rcu_read_lock();
6177                 id = rcu_dereference(nn_clp->identifier);
6178                 if (id)
6179                         strscpy(buf, id, buflen);
6180                 rcu_read_unlock();
6181         }
6182
6183         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6184                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6185
6186         return strlen(buf);
6187 }
6188
6189 static int
6190 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6191 {
6192         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6193         size_t buflen;
6194         size_t len;
6195         char *str;
6196
6197         if (clp->cl_owner_id != NULL)
6198                 return 0;
6199
6200         rcu_read_lock();
6201         len = 14 +
6202                 strlen(clp->cl_rpcclient->cl_nodename) +
6203                 1 +
6204                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6205                 1;
6206         rcu_read_unlock();
6207
6208         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6209         if (buflen)
6210                 len += buflen + 1;
6211
6212         if (len > NFS4_OPAQUE_LIMIT + 1)
6213                 return -EINVAL;
6214
6215         /*
6216          * Since this string is allocated at mount time, and held until the
6217          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6218          * about a memory-reclaim deadlock.
6219          */
6220         str = kmalloc(len, GFP_KERNEL);
6221         if (!str)
6222                 return -ENOMEM;
6223
6224         rcu_read_lock();
6225         if (buflen)
6226                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6227                           clp->cl_rpcclient->cl_nodename, buf,
6228                           rpc_peeraddr2str(clp->cl_rpcclient,
6229                                            RPC_DISPLAY_ADDR));
6230         else
6231                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6232                           clp->cl_rpcclient->cl_nodename,
6233                           rpc_peeraddr2str(clp->cl_rpcclient,
6234                                            RPC_DISPLAY_ADDR));
6235         rcu_read_unlock();
6236
6237         clp->cl_owner_id = str;
6238         return 0;
6239 }
6240
6241 static int
6242 nfs4_init_uniform_client_string(struct nfs_client *clp)
6243 {
6244         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6245         size_t buflen;
6246         size_t len;
6247         char *str;
6248
6249         if (clp->cl_owner_id != NULL)
6250                 return 0;
6251
6252         len = 10 + 10 + 1 + 10 + 1 +
6253                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6254
6255         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6256         if (buflen)
6257                 len += buflen + 1;
6258
6259         if (len > NFS4_OPAQUE_LIMIT + 1)
6260                 return -EINVAL;
6261
6262         /*
6263          * Since this string is allocated at mount time, and held until the
6264          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6265          * about a memory-reclaim deadlock.
6266          */
6267         str = kmalloc(len, GFP_KERNEL);
6268         if (!str)
6269                 return -ENOMEM;
6270
6271         if (buflen)
6272                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6273                           clp->rpc_ops->version, clp->cl_minorversion,
6274                           buf, clp->cl_rpcclient->cl_nodename);
6275         else
6276                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6277                           clp->rpc_ops->version, clp->cl_minorversion,
6278                           clp->cl_rpcclient->cl_nodename);
6279         clp->cl_owner_id = str;
6280         return 0;
6281 }
6282
6283 /*
6284  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6285  * services.  Advertise one based on the address family of the
6286  * clientaddr.
6287  */
6288 static unsigned int
6289 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6290 {
6291         if (strchr(clp->cl_ipaddr, ':') != NULL)
6292                 return scnprintf(buf, len, "tcp6");
6293         else
6294                 return scnprintf(buf, len, "tcp");
6295 }
6296
6297 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6298 {
6299         struct nfs4_setclientid *sc = calldata;
6300
6301         if (task->tk_status == 0)
6302                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6303 }
6304
6305 static const struct rpc_call_ops nfs4_setclientid_ops = {
6306         .rpc_call_done = nfs4_setclientid_done,
6307 };
6308
6309 /**
6310  * nfs4_proc_setclientid - Negotiate client ID
6311  * @clp: state data structure
6312  * @program: RPC program for NFSv4 callback service
6313  * @port: IP port number for NFS4 callback service
6314  * @cred: credential to use for this call
6315  * @res: where to place the result
6316  *
6317  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6318  */
6319 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6320                 unsigned short port, const struct cred *cred,
6321                 struct nfs4_setclientid_res *res)
6322 {
6323         nfs4_verifier sc_verifier;
6324         struct nfs4_setclientid setclientid = {
6325                 .sc_verifier = &sc_verifier,
6326                 .sc_prog = program,
6327                 .sc_clnt = clp,
6328         };
6329         struct rpc_message msg = {
6330                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6331                 .rpc_argp = &setclientid,
6332                 .rpc_resp = res,
6333                 .rpc_cred = cred,
6334         };
6335         struct rpc_task_setup task_setup_data = {
6336                 .rpc_client = clp->cl_rpcclient,
6337                 .rpc_message = &msg,
6338                 .callback_ops = &nfs4_setclientid_ops,
6339                 .callback_data = &setclientid,
6340                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6341         };
6342         unsigned long now = jiffies;
6343         int status;
6344
6345         /* nfs_client_id4 */
6346         nfs4_init_boot_verifier(clp, &sc_verifier);
6347
6348         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6349                 status = nfs4_init_uniform_client_string(clp);
6350         else
6351                 status = nfs4_init_nonuniform_client_string(clp);
6352
6353         if (status)
6354                 goto out;
6355
6356         /* cb_client4 */
6357         setclientid.sc_netid_len =
6358                                 nfs4_init_callback_netid(clp,
6359                                                 setclientid.sc_netid,
6360                                                 sizeof(setclientid.sc_netid));
6361         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6362                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6363                                 clp->cl_ipaddr, port >> 8, port & 255);
6364
6365         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6366                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6367                 clp->cl_owner_id);
6368
6369         status = nfs4_call_sync_custom(&task_setup_data);
6370         if (setclientid.sc_cred) {
6371                 kfree(clp->cl_acceptor);
6372                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6373                 put_rpccred(setclientid.sc_cred);
6374         }
6375
6376         if (status == 0)
6377                 do_renew_lease(clp, now);
6378 out:
6379         trace_nfs4_setclientid(clp, status);
6380         dprintk("NFS reply setclientid: %d\n", status);
6381         return status;
6382 }
6383
6384 /**
6385  * nfs4_proc_setclientid_confirm - Confirm client ID
6386  * @clp: state data structure
6387  * @arg: result of a previous SETCLIENTID
6388  * @cred: credential to use for this call
6389  *
6390  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6391  */
6392 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6393                 struct nfs4_setclientid_res *arg,
6394                 const struct cred *cred)
6395 {
6396         struct rpc_message msg = {
6397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6398                 .rpc_argp = arg,
6399                 .rpc_cred = cred,
6400         };
6401         int status;
6402
6403         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6404                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6405                 clp->cl_clientid);
6406         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6407                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6408         trace_nfs4_setclientid_confirm(clp, status);
6409         dprintk("NFS reply setclientid_confirm: %d\n", status);
6410         return status;
6411 }
6412
6413 struct nfs4_delegreturndata {
6414         struct nfs4_delegreturnargs args;
6415         struct nfs4_delegreturnres res;
6416         struct nfs_fh fh;
6417         nfs4_stateid stateid;
6418         unsigned long timestamp;
6419         struct {
6420                 struct nfs4_layoutreturn_args arg;
6421                 struct nfs4_layoutreturn_res res;
6422                 struct nfs4_xdr_opaque_data ld_private;
6423                 u32 roc_barrier;
6424                 bool roc;
6425         } lr;
6426         struct nfs_fattr fattr;
6427         int rpc_status;
6428         struct inode *inode;
6429 };
6430
6431 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6432 {
6433         struct nfs4_delegreturndata *data = calldata;
6434         struct nfs4_exception exception = {
6435                 .inode = data->inode,
6436                 .stateid = &data->stateid,
6437                 .task_is_privileged = data->args.seq_args.sa_privileged,
6438         };
6439
6440         if (!nfs4_sequence_done(task, &data->res.seq_res))
6441                 return;
6442
6443         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6444
6445         /* Handle Layoutreturn errors */
6446         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6447                           &data->res.lr_ret) == -EAGAIN)
6448                 goto out_restart;
6449
6450         switch (task->tk_status) {
6451         case 0:
6452                 renew_lease(data->res.server, data->timestamp);
6453                 break;
6454         case -NFS4ERR_ADMIN_REVOKED:
6455         case -NFS4ERR_DELEG_REVOKED:
6456         case -NFS4ERR_EXPIRED:
6457                 nfs4_free_revoked_stateid(data->res.server,
6458                                 data->args.stateid,
6459                                 task->tk_msg.rpc_cred);
6460                 fallthrough;
6461         case -NFS4ERR_BAD_STATEID:
6462         case -NFS4ERR_STALE_STATEID:
6463         case -ETIMEDOUT:
6464                 task->tk_status = 0;
6465                 break;
6466         case -NFS4ERR_OLD_STATEID:
6467                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6468                         nfs4_stateid_seqid_inc(&data->stateid);
6469                 if (data->args.bitmask) {
6470                         data->args.bitmask = NULL;
6471                         data->res.fattr = NULL;
6472                 }
6473                 goto out_restart;
6474         case -NFS4ERR_ACCESS:
6475                 if (data->args.bitmask) {
6476                         data->args.bitmask = NULL;
6477                         data->res.fattr = NULL;
6478                         goto out_restart;
6479                 }
6480                 fallthrough;
6481         default:
6482                 task->tk_status = nfs4_async_handle_exception(task,
6483                                 data->res.server, task->tk_status,
6484                                 &exception);
6485                 if (exception.retry)
6486                         goto out_restart;
6487         }
6488         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6489         data->rpc_status = task->tk_status;
6490         return;
6491 out_restart:
6492         task->tk_status = 0;
6493         rpc_restart_call_prepare(task);
6494 }
6495
6496 static void nfs4_delegreturn_release(void *calldata)
6497 {
6498         struct nfs4_delegreturndata *data = calldata;
6499         struct inode *inode = data->inode;
6500
6501         if (data->lr.roc)
6502                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6503                                  data->res.lr_ret);
6504         if (inode) {
6505                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6506                 nfs_iput_and_deactive(inode);
6507         }
6508         kfree(calldata);
6509 }
6510
6511 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6512 {
6513         struct nfs4_delegreturndata *d_data;
6514         struct pnfs_layout_hdr *lo;
6515
6516         d_data = (struct nfs4_delegreturndata *)data;
6517
6518         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6519                 nfs4_sequence_done(task, &d_data->res.seq_res);
6520                 return;
6521         }
6522
6523         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6524         if (lo && !pnfs_layout_is_valid(lo)) {
6525                 d_data->args.lr_args = NULL;
6526                 d_data->res.lr_res = NULL;
6527         }
6528
6529         nfs4_setup_sequence(d_data->res.server->nfs_client,
6530                         &d_data->args.seq_args,
6531                         &d_data->res.seq_res,
6532                         task);
6533 }
6534
6535 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6536         .rpc_call_prepare = nfs4_delegreturn_prepare,
6537         .rpc_call_done = nfs4_delegreturn_done,
6538         .rpc_release = nfs4_delegreturn_release,
6539 };
6540
6541 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6542 {
6543         struct nfs4_delegreturndata *data;
6544         struct nfs_server *server = NFS_SERVER(inode);
6545         struct rpc_task *task;
6546         struct rpc_message msg = {
6547                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6548                 .rpc_cred = cred,
6549         };
6550         struct rpc_task_setup task_setup_data = {
6551                 .rpc_client = server->client,
6552                 .rpc_message = &msg,
6553                 .callback_ops = &nfs4_delegreturn_ops,
6554                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
6555         };
6556         int status = 0;
6557
6558         data = kzalloc(sizeof(*data), GFP_NOFS);
6559         if (data == NULL)
6560                 return -ENOMEM;
6561
6562         nfs4_state_protect(server->nfs_client,
6563                         NFS_SP4_MACH_CRED_CLEANUP,
6564                         &task_setup_data.rpc_client, &msg);
6565
6566         data->args.fhandle = &data->fh;
6567         data->args.stateid = &data->stateid;
6568         nfs4_bitmask_set(data->args.bitmask_store,
6569                          server->cache_consistency_bitmask, inode, server,
6570                          NULL);
6571         data->args.bitmask = data->args.bitmask_store;
6572         nfs_copy_fh(&data->fh, NFS_FH(inode));
6573         nfs4_stateid_copy(&data->stateid, stateid);
6574         data->res.fattr = &data->fattr;
6575         data->res.server = server;
6576         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6577         data->lr.arg.ld_private = &data->lr.ld_private;
6578         nfs_fattr_init(data->res.fattr);
6579         data->timestamp = jiffies;
6580         data->rpc_status = 0;
6581         data->inode = nfs_igrab_and_active(inode);
6582         if (data->inode || issync) {
6583                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6584                                         cred);
6585                 if (data->lr.roc) {
6586                         data->args.lr_args = &data->lr.arg;
6587                         data->res.lr_res = &data->lr.res;
6588                 }
6589         }
6590
6591         if (!data->inode)
6592                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6593                                    1);
6594         else
6595                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6596                                    0);
6597         task_setup_data.callback_data = data;
6598         msg.rpc_argp = &data->args;
6599         msg.rpc_resp = &data->res;
6600         task = rpc_run_task(&task_setup_data);
6601         if (IS_ERR(task))
6602                 return PTR_ERR(task);
6603         if (!issync)
6604                 goto out;
6605         status = rpc_wait_for_completion_task(task);
6606         if (status != 0)
6607                 goto out;
6608         status = data->rpc_status;
6609 out:
6610         rpc_put_task(task);
6611         return status;
6612 }
6613
6614 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6615 {
6616         struct nfs_server *server = NFS_SERVER(inode);
6617         struct nfs4_exception exception = { };
6618         int err;
6619         do {
6620                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6621                 trace_nfs4_delegreturn(inode, stateid, err);
6622                 switch (err) {
6623                         case -NFS4ERR_STALE_STATEID:
6624                         case -NFS4ERR_EXPIRED:
6625                         case 0:
6626                                 return 0;
6627                 }
6628                 err = nfs4_handle_exception(server, err, &exception);
6629         } while (exception.retry);
6630         return err;
6631 }
6632
6633 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6634 {
6635         struct inode *inode = state->inode;
6636         struct nfs_server *server = NFS_SERVER(inode);
6637         struct nfs_client *clp = server->nfs_client;
6638         struct nfs_lockt_args arg = {
6639                 .fh = NFS_FH(inode),
6640                 .fl = request,
6641         };
6642         struct nfs_lockt_res res = {
6643                 .denied = request,
6644         };
6645         struct rpc_message msg = {
6646                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6647                 .rpc_argp       = &arg,
6648                 .rpc_resp       = &res,
6649                 .rpc_cred       = state->owner->so_cred,
6650         };
6651         struct nfs4_lock_state *lsp;
6652         int status;
6653
6654         arg.lock_owner.clientid = clp->cl_clientid;
6655         status = nfs4_set_lock_state(state, request);
6656         if (status != 0)
6657                 goto out;
6658         lsp = request->fl_u.nfs4_fl.owner;
6659         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6660         arg.lock_owner.s_dev = server->s_dev;
6661         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6662         switch (status) {
6663                 case 0:
6664                         request->fl_type = F_UNLCK;
6665                         break;
6666                 case -NFS4ERR_DENIED:
6667                         status = 0;
6668         }
6669         request->fl_ops->fl_release_private(request);
6670         request->fl_ops = NULL;
6671 out:
6672         return status;
6673 }
6674
6675 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6676 {
6677         struct nfs4_exception exception = {
6678                 .interruptible = true,
6679         };
6680         int err;
6681
6682         do {
6683                 err = _nfs4_proc_getlk(state, cmd, request);
6684                 trace_nfs4_get_lock(request, state, cmd, err);
6685                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6686                                 &exception);
6687         } while (exception.retry);
6688         return err;
6689 }
6690
6691 /*
6692  * Update the seqid of a lock stateid after receiving
6693  * NFS4ERR_OLD_STATEID
6694  */
6695 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6696                 struct nfs4_lock_state *lsp)
6697 {
6698         struct nfs4_state *state = lsp->ls_state;
6699         bool ret = false;
6700
6701         spin_lock(&state->state_lock);
6702         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6703                 goto out;
6704         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6705                 nfs4_stateid_seqid_inc(dst);
6706         else
6707                 dst->seqid = lsp->ls_stateid.seqid;
6708         ret = true;
6709 out:
6710         spin_unlock(&state->state_lock);
6711         return ret;
6712 }
6713
6714 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6715                 struct nfs4_lock_state *lsp)
6716 {
6717         struct nfs4_state *state = lsp->ls_state;
6718         bool ret;
6719
6720         spin_lock(&state->state_lock);
6721         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6722         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6723         spin_unlock(&state->state_lock);
6724         return ret;
6725 }
6726
6727 struct nfs4_unlockdata {
6728         struct nfs_locku_args arg;
6729         struct nfs_locku_res res;
6730         struct nfs4_lock_state *lsp;
6731         struct nfs_open_context *ctx;
6732         struct nfs_lock_context *l_ctx;
6733         struct file_lock fl;
6734         struct nfs_server *server;
6735         unsigned long timestamp;
6736 };
6737
6738 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6739                 struct nfs_open_context *ctx,
6740                 struct nfs4_lock_state *lsp,
6741                 struct nfs_seqid *seqid)
6742 {
6743         struct nfs4_unlockdata *p;
6744         struct nfs4_state *state = lsp->ls_state;
6745         struct inode *inode = state->inode;
6746
6747         p = kzalloc(sizeof(*p), GFP_NOFS);
6748         if (p == NULL)
6749                 return NULL;
6750         p->arg.fh = NFS_FH(inode);
6751         p->arg.fl = &p->fl;
6752         p->arg.seqid = seqid;
6753         p->res.seqid = seqid;
6754         p->lsp = lsp;
6755         /* Ensure we don't close file until we're done freeing locks! */
6756         p->ctx = get_nfs_open_context(ctx);
6757         p->l_ctx = nfs_get_lock_context(ctx);
6758         locks_init_lock(&p->fl);
6759         locks_copy_lock(&p->fl, fl);
6760         p->server = NFS_SERVER(inode);
6761         spin_lock(&state->state_lock);
6762         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6763         spin_unlock(&state->state_lock);
6764         return p;
6765 }
6766
6767 static void nfs4_locku_release_calldata(void *data)
6768 {
6769         struct nfs4_unlockdata *calldata = data;
6770         nfs_free_seqid(calldata->arg.seqid);
6771         nfs4_put_lock_state(calldata->lsp);
6772         nfs_put_lock_context(calldata->l_ctx);
6773         put_nfs_open_context(calldata->ctx);
6774         kfree(calldata);
6775 }
6776
6777 static void nfs4_locku_done(struct rpc_task *task, void *data)
6778 {
6779         struct nfs4_unlockdata *calldata = data;
6780         struct nfs4_exception exception = {
6781                 .inode = calldata->lsp->ls_state->inode,
6782                 .stateid = &calldata->arg.stateid,
6783         };
6784
6785         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6786                 return;
6787         switch (task->tk_status) {
6788                 case 0:
6789                         renew_lease(calldata->server, calldata->timestamp);
6790                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6791                         if (nfs4_update_lock_stateid(calldata->lsp,
6792                                         &calldata->res.stateid))
6793                                 break;
6794                         fallthrough;
6795                 case -NFS4ERR_ADMIN_REVOKED:
6796                 case -NFS4ERR_EXPIRED:
6797                         nfs4_free_revoked_stateid(calldata->server,
6798                                         &calldata->arg.stateid,
6799                                         task->tk_msg.rpc_cred);
6800                         fallthrough;
6801                 case -NFS4ERR_BAD_STATEID:
6802                 case -NFS4ERR_STALE_STATEID:
6803                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6804                                                 calldata->lsp))
6805                                 rpc_restart_call_prepare(task);
6806                         break;
6807                 case -NFS4ERR_OLD_STATEID:
6808                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6809                                                 calldata->lsp))
6810                                 rpc_restart_call_prepare(task);
6811                         break;
6812                 default:
6813                         task->tk_status = nfs4_async_handle_exception(task,
6814                                         calldata->server, task->tk_status,
6815                                         &exception);
6816                         if (exception.retry)
6817                                 rpc_restart_call_prepare(task);
6818         }
6819         nfs_release_seqid(calldata->arg.seqid);
6820 }
6821
6822 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6823 {
6824         struct nfs4_unlockdata *calldata = data;
6825
6826         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6827                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6828                 return;
6829
6830         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6831                 goto out_wait;
6832         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6833                 /* Note: exit _without_ running nfs4_locku_done */
6834                 goto out_no_action;
6835         }
6836         calldata->timestamp = jiffies;
6837         if (nfs4_setup_sequence(calldata->server->nfs_client,
6838                                 &calldata->arg.seq_args,
6839                                 &calldata->res.seq_res,
6840                                 task) != 0)
6841                 nfs_release_seqid(calldata->arg.seqid);
6842         return;
6843 out_no_action:
6844         task->tk_action = NULL;
6845 out_wait:
6846         nfs4_sequence_done(task, &calldata->res.seq_res);
6847 }
6848
6849 static const struct rpc_call_ops nfs4_locku_ops = {
6850         .rpc_call_prepare = nfs4_locku_prepare,
6851         .rpc_call_done = nfs4_locku_done,
6852         .rpc_release = nfs4_locku_release_calldata,
6853 };
6854
6855 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6856                 struct nfs_open_context *ctx,
6857                 struct nfs4_lock_state *lsp,
6858                 struct nfs_seqid *seqid)
6859 {
6860         struct nfs4_unlockdata *data;
6861         struct rpc_message msg = {
6862                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6863                 .rpc_cred = ctx->cred,
6864         };
6865         struct rpc_task_setup task_setup_data = {
6866                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6867                 .rpc_message = &msg,
6868                 .callback_ops = &nfs4_locku_ops,
6869                 .workqueue = nfsiod_workqueue,
6870                 .flags = RPC_TASK_ASYNC,
6871         };
6872         struct nfs_client *client =
6873                 NFS_SERVER(lsp->ls_state->inode)->nfs_client;
6874
6875         if (client->cl_minorversion)
6876                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6877
6878         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6879                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6880
6881         /* Ensure this is an unlock - when canceling a lock, the
6882          * canceled lock is passed in, and it won't be an unlock.
6883          */
6884         fl->fl_type = F_UNLCK;
6885         if (fl->fl_flags & FL_CLOSE)
6886                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6887
6888         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6889         if (data == NULL) {
6890                 nfs_free_seqid(seqid);
6891                 return ERR_PTR(-ENOMEM);
6892         }
6893
6894         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6895         msg.rpc_argp = &data->arg;
6896         msg.rpc_resp = &data->res;
6897         task_setup_data.callback_data = data;
6898         return rpc_run_task(&task_setup_data);
6899 }
6900
6901 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6902 {
6903         struct inode *inode = state->inode;
6904         struct nfs4_state_owner *sp = state->owner;
6905         struct nfs_inode *nfsi = NFS_I(inode);
6906         struct nfs_seqid *seqid;
6907         struct nfs4_lock_state *lsp;
6908         struct rpc_task *task;
6909         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6910         int status = 0;
6911         unsigned char fl_flags = request->fl_flags;
6912
6913         status = nfs4_set_lock_state(state, request);
6914         /* Unlock _before_ we do the RPC call */
6915         request->fl_flags |= FL_EXISTS;
6916         /* Exclude nfs_delegation_claim_locks() */
6917         mutex_lock(&sp->so_delegreturn_mutex);
6918         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6919         down_read(&nfsi->rwsem);
6920         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6921                 up_read(&nfsi->rwsem);
6922                 mutex_unlock(&sp->so_delegreturn_mutex);
6923                 goto out;
6924         }
6925         up_read(&nfsi->rwsem);
6926         mutex_unlock(&sp->so_delegreturn_mutex);
6927         if (status != 0)
6928                 goto out;
6929         /* Is this a delegated lock? */
6930         lsp = request->fl_u.nfs4_fl.owner;
6931         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6932                 goto out;
6933         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6934         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6935         status = -ENOMEM;
6936         if (IS_ERR(seqid))
6937                 goto out;
6938         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6939         status = PTR_ERR(task);
6940         if (IS_ERR(task))
6941                 goto out;
6942         status = rpc_wait_for_completion_task(task);
6943         rpc_put_task(task);
6944 out:
6945         request->fl_flags = fl_flags;
6946         trace_nfs4_unlock(request, state, F_SETLK, status);
6947         return status;
6948 }
6949
6950 struct nfs4_lockdata {
6951         struct nfs_lock_args arg;
6952         struct nfs_lock_res res;
6953         struct nfs4_lock_state *lsp;
6954         struct nfs_open_context *ctx;
6955         struct file_lock fl;
6956         unsigned long timestamp;
6957         int rpc_status;
6958         int cancelled;
6959         struct nfs_server *server;
6960 };
6961
6962 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6963                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6964                 gfp_t gfp_mask)
6965 {
6966         struct nfs4_lockdata *p;
6967         struct inode *inode = lsp->ls_state->inode;
6968         struct nfs_server *server = NFS_SERVER(inode);
6969         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6970
6971         p = kzalloc(sizeof(*p), gfp_mask);
6972         if (p == NULL)
6973                 return NULL;
6974
6975         p->arg.fh = NFS_FH(inode);
6976         p->arg.fl = &p->fl;
6977         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6978         if (IS_ERR(p->arg.open_seqid))
6979                 goto out_free;
6980         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6981         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6982         if (IS_ERR(p->arg.lock_seqid))
6983                 goto out_free_seqid;
6984         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6985         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6986         p->arg.lock_owner.s_dev = server->s_dev;
6987         p->res.lock_seqid = p->arg.lock_seqid;
6988         p->lsp = lsp;
6989         p->server = server;
6990         p->ctx = get_nfs_open_context(ctx);
6991         locks_init_lock(&p->fl);
6992         locks_copy_lock(&p->fl, fl);
6993         return p;
6994 out_free_seqid:
6995         nfs_free_seqid(p->arg.open_seqid);
6996 out_free:
6997         kfree(p);
6998         return NULL;
6999 }
7000
7001 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7002 {
7003         struct nfs4_lockdata *data = calldata;
7004         struct nfs4_state *state = data->lsp->ls_state;
7005
7006         dprintk("%s: begin!\n", __func__);
7007         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7008                 goto out_wait;
7009         /* Do we need to do an open_to_lock_owner? */
7010         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7011                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7012                         goto out_release_lock_seqid;
7013                 }
7014                 nfs4_stateid_copy(&data->arg.open_stateid,
7015                                 &state->open_stateid);
7016                 data->arg.new_lock_owner = 1;
7017                 data->res.open_seqid = data->arg.open_seqid;
7018         } else {
7019                 data->arg.new_lock_owner = 0;
7020                 nfs4_stateid_copy(&data->arg.lock_stateid,
7021                                 &data->lsp->ls_stateid);
7022         }
7023         if (!nfs4_valid_open_stateid(state)) {
7024                 data->rpc_status = -EBADF;
7025                 task->tk_action = NULL;
7026                 goto out_release_open_seqid;
7027         }
7028         data->timestamp = jiffies;
7029         if (nfs4_setup_sequence(data->server->nfs_client,
7030                                 &data->arg.seq_args,
7031                                 &data->res.seq_res,
7032                                 task) == 0)
7033                 return;
7034 out_release_open_seqid:
7035         nfs_release_seqid(data->arg.open_seqid);
7036 out_release_lock_seqid:
7037         nfs_release_seqid(data->arg.lock_seqid);
7038 out_wait:
7039         nfs4_sequence_done(task, &data->res.seq_res);
7040         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
7041 }
7042
7043 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7044 {
7045         struct nfs4_lockdata *data = calldata;
7046         struct nfs4_lock_state *lsp = data->lsp;
7047
7048         dprintk("%s: begin!\n", __func__);
7049
7050         if (!nfs4_sequence_done(task, &data->res.seq_res))
7051                 return;
7052
7053         data->rpc_status = task->tk_status;
7054         switch (task->tk_status) {
7055         case 0:
7056                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7057                                 data->timestamp);
7058                 if (data->arg.new_lock && !data->cancelled) {
7059                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7060                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7061                                 goto out_restart;
7062                 }
7063                 if (data->arg.new_lock_owner != 0) {
7064                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7065                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7066                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7067                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7068                         goto out_restart;
7069                 break;
7070         case -NFS4ERR_BAD_STATEID:
7071         case -NFS4ERR_OLD_STATEID:
7072         case -NFS4ERR_STALE_STATEID:
7073         case -NFS4ERR_EXPIRED:
7074                 if (data->arg.new_lock_owner != 0) {
7075                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7076                                                 &lsp->ls_state->open_stateid))
7077                                 goto out_restart;
7078                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7079                                                 &lsp->ls_stateid))
7080                                 goto out_restart;
7081         }
7082 out_done:
7083         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7084         return;
7085 out_restart:
7086         if (!data->cancelled)
7087                 rpc_restart_call_prepare(task);
7088         goto out_done;
7089 }
7090
7091 static void nfs4_lock_release(void *calldata)
7092 {
7093         struct nfs4_lockdata *data = calldata;
7094
7095         dprintk("%s: begin!\n", __func__);
7096         nfs_free_seqid(data->arg.open_seqid);
7097         if (data->cancelled && data->rpc_status == 0) {
7098                 struct rpc_task *task;
7099                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7100                                 data->arg.lock_seqid);
7101                 if (!IS_ERR(task))
7102                         rpc_put_task_async(task);
7103                 dprintk("%s: cancelling lock!\n", __func__);
7104         } else
7105                 nfs_free_seqid(data->arg.lock_seqid);
7106         nfs4_put_lock_state(data->lsp);
7107         put_nfs_open_context(data->ctx);
7108         kfree(data);
7109         dprintk("%s: done!\n", __func__);
7110 }
7111
7112 static const struct rpc_call_ops nfs4_lock_ops = {
7113         .rpc_call_prepare = nfs4_lock_prepare,
7114         .rpc_call_done = nfs4_lock_done,
7115         .rpc_release = nfs4_lock_release,
7116 };
7117
7118 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7119 {
7120         switch (error) {
7121         case -NFS4ERR_ADMIN_REVOKED:
7122         case -NFS4ERR_EXPIRED:
7123         case -NFS4ERR_BAD_STATEID:
7124                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7125                 if (new_lock_owner != 0 ||
7126                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7127                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7128                 break;
7129         case -NFS4ERR_STALE_STATEID:
7130                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7131                 nfs4_schedule_lease_recovery(server->nfs_client);
7132         }
7133 }
7134
7135 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7136 {
7137         struct nfs4_lockdata *data;
7138         struct rpc_task *task;
7139         struct rpc_message msg = {
7140                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7141                 .rpc_cred = state->owner->so_cred,
7142         };
7143         struct rpc_task_setup task_setup_data = {
7144                 .rpc_client = NFS_CLIENT(state->inode),
7145                 .rpc_message = &msg,
7146                 .callback_ops = &nfs4_lock_ops,
7147                 .workqueue = nfsiod_workqueue,
7148                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7149         };
7150         int ret;
7151         struct nfs_client *client = NFS_SERVER(state->inode)->nfs_client;
7152
7153         if (client->cl_minorversion)
7154                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7155
7156         dprintk("%s: begin!\n", __func__);
7157         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7158                         fl->fl_u.nfs4_fl.owner,
7159                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7160         if (data == NULL)
7161                 return -ENOMEM;
7162         if (IS_SETLKW(cmd))
7163                 data->arg.block = 1;
7164         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7165                                 recovery_type > NFS_LOCK_NEW);
7166         msg.rpc_argp = &data->arg;
7167         msg.rpc_resp = &data->res;
7168         task_setup_data.callback_data = data;
7169         if (recovery_type > NFS_LOCK_NEW) {
7170                 if (recovery_type == NFS_LOCK_RECLAIM)
7171                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7172         } else
7173                 data->arg.new_lock = 1;
7174         task = rpc_run_task(&task_setup_data);
7175         if (IS_ERR(task))
7176                 return PTR_ERR(task);
7177         ret = rpc_wait_for_completion_task(task);
7178         if (ret == 0) {
7179                 ret = data->rpc_status;
7180                 if (ret)
7181                         nfs4_handle_setlk_error(data->server, data->lsp,
7182                                         data->arg.new_lock_owner, ret);
7183         } else
7184                 data->cancelled = true;
7185         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7186         rpc_put_task(task);
7187         dprintk("%s: done, ret = %d!\n", __func__, ret);
7188         return ret;
7189 }
7190
7191 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7192 {
7193         struct nfs_server *server = NFS_SERVER(state->inode);
7194         struct nfs4_exception exception = {
7195                 .inode = state->inode,
7196         };
7197         int err;
7198
7199         do {
7200                 /* Cache the lock if possible... */
7201                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7202                         return 0;
7203                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7204                 if (err != -NFS4ERR_DELAY)
7205                         break;
7206                 nfs4_handle_exception(server, err, &exception);
7207         } while (exception.retry);
7208         return err;
7209 }
7210
7211 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7212 {
7213         struct nfs_server *server = NFS_SERVER(state->inode);
7214         struct nfs4_exception exception = {
7215                 .inode = state->inode,
7216         };
7217         int err;
7218
7219         err = nfs4_set_lock_state(state, request);
7220         if (err != 0)
7221                 return err;
7222         if (!recover_lost_locks) {
7223                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7224                 return 0;
7225         }
7226         do {
7227                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7228                         return 0;
7229                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7230                 switch (err) {
7231                 default:
7232                         goto out;
7233                 case -NFS4ERR_GRACE:
7234                 case -NFS4ERR_DELAY:
7235                         nfs4_handle_exception(server, err, &exception);
7236                         err = 0;
7237                 }
7238         } while (exception.retry);
7239 out:
7240         return err;
7241 }
7242
7243 #if defined(CONFIG_NFS_V4_1)
7244 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7245 {
7246         struct nfs4_lock_state *lsp;
7247         int status;
7248
7249         status = nfs4_set_lock_state(state, request);
7250         if (status != 0)
7251                 return status;
7252         lsp = request->fl_u.nfs4_fl.owner;
7253         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7254             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7255                 return 0;
7256         return nfs4_lock_expired(state, request);
7257 }
7258 #endif
7259
7260 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7261 {
7262         struct nfs_inode *nfsi = NFS_I(state->inode);
7263         struct nfs4_state_owner *sp = state->owner;
7264         unsigned char fl_flags = request->fl_flags;
7265         int status;
7266
7267         request->fl_flags |= FL_ACCESS;
7268         status = locks_lock_inode_wait(state->inode, request);
7269         if (status < 0)
7270                 goto out;
7271         mutex_lock(&sp->so_delegreturn_mutex);
7272         down_read(&nfsi->rwsem);
7273         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7274                 /* Yes: cache locks! */
7275                 /* ...but avoid races with delegation recall... */
7276                 request->fl_flags = fl_flags & ~FL_SLEEP;
7277                 status = locks_lock_inode_wait(state->inode, request);
7278                 up_read(&nfsi->rwsem);
7279                 mutex_unlock(&sp->so_delegreturn_mutex);
7280                 goto out;
7281         }
7282         up_read(&nfsi->rwsem);
7283         mutex_unlock(&sp->so_delegreturn_mutex);
7284         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7285 out:
7286         request->fl_flags = fl_flags;
7287         return status;
7288 }
7289
7290 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7291 {
7292         struct nfs4_exception exception = {
7293                 .state = state,
7294                 .inode = state->inode,
7295                 .interruptible = true,
7296         };
7297         int err;
7298
7299         do {
7300                 err = _nfs4_proc_setlk(state, cmd, request);
7301                 if (err == -NFS4ERR_DENIED)
7302                         err = -EAGAIN;
7303                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7304                                 err, &exception);
7305         } while (exception.retry);
7306         return err;
7307 }
7308
7309 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7310 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7311
7312 static int
7313 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7314                         struct file_lock *request)
7315 {
7316         int             status = -ERESTARTSYS;
7317         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7318
7319         while(!signalled()) {
7320                 status = nfs4_proc_setlk(state, cmd, request);
7321                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7322                         break;
7323                 freezable_schedule_timeout_interruptible(timeout);
7324                 timeout *= 2;
7325                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7326                 status = -ERESTARTSYS;
7327         }
7328         return status;
7329 }
7330
7331 #ifdef CONFIG_NFS_V4_1
7332 struct nfs4_lock_waiter {
7333         struct inode            *inode;
7334         struct nfs_lowner       owner;
7335         wait_queue_entry_t      wait;
7336 };
7337
7338 static int
7339 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7340 {
7341         struct nfs4_lock_waiter *waiter =
7342                 container_of(wait, struct nfs4_lock_waiter, wait);
7343
7344         /* NULL key means to wake up everyone */
7345         if (key) {
7346                 struct cb_notify_lock_args      *cbnl = key;
7347                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7348                                                 *wowner = &waiter->owner;
7349
7350                 /* Only wake if the callback was for the same owner. */
7351                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7352                         return 0;
7353
7354                 /* Make sure it's for the right inode */
7355                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7356                         return 0;
7357         }
7358
7359         return woken_wake_function(wait, mode, flags, key);
7360 }
7361
7362 static int
7363 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7364 {
7365         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7366         struct nfs_server *server = NFS_SERVER(state->inode);
7367         struct nfs_client *clp = server->nfs_client;
7368         wait_queue_head_t *q = &clp->cl_lock_waitq;
7369         struct nfs4_lock_waiter waiter = {
7370                 .inode = state->inode,
7371                 .owner = { .clientid = clp->cl_clientid,
7372                            .id = lsp->ls_seqid.owner_id,
7373                            .s_dev = server->s_dev },
7374         };
7375         int status;
7376
7377         /* Don't bother with waitqueue if we don't expect a callback */
7378         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7379                 return nfs4_retry_setlk_simple(state, cmd, request);
7380
7381         init_wait(&waiter.wait);
7382         waiter.wait.func = nfs4_wake_lock_waiter;
7383         add_wait_queue(q, &waiter.wait);
7384
7385         do {
7386                 status = nfs4_proc_setlk(state, cmd, request);
7387                 if (status != -EAGAIN || IS_SETLK(cmd))
7388                         break;
7389
7390                 status = -ERESTARTSYS;
7391                 freezer_do_not_count();
7392                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7393                            NFS4_LOCK_MAXTIMEOUT);
7394                 freezer_count();
7395         } while (!signalled());
7396
7397         remove_wait_queue(q, &waiter.wait);
7398
7399         return status;
7400 }
7401 #else /* !CONFIG_NFS_V4_1 */
7402 static inline int
7403 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7404 {
7405         return nfs4_retry_setlk_simple(state, cmd, request);
7406 }
7407 #endif
7408
7409 static int
7410 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7411 {
7412         struct nfs_open_context *ctx;
7413         struct nfs4_state *state;
7414         int status;
7415
7416         /* verify open state */
7417         ctx = nfs_file_open_context(filp);
7418         state = ctx->state;
7419
7420         if (IS_GETLK(cmd)) {
7421                 if (state != NULL)
7422                         return nfs4_proc_getlk(state, F_GETLK, request);
7423                 return 0;
7424         }
7425
7426         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7427                 return -EINVAL;
7428
7429         if (request->fl_type == F_UNLCK) {
7430                 if (state != NULL)
7431                         return nfs4_proc_unlck(state, cmd, request);
7432                 return 0;
7433         }
7434
7435         if (state == NULL)
7436                 return -ENOLCK;
7437
7438         if ((request->fl_flags & FL_POSIX) &&
7439             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7440                 return -ENOLCK;
7441
7442         /*
7443          * Don't rely on the VFS having checked the file open mode,
7444          * since it won't do this for flock() locks.
7445          */
7446         switch (request->fl_type) {
7447         case F_RDLCK:
7448                 if (!(filp->f_mode & FMODE_READ))
7449                         return -EBADF;
7450                 break;
7451         case F_WRLCK:
7452                 if (!(filp->f_mode & FMODE_WRITE))
7453                         return -EBADF;
7454         }
7455
7456         status = nfs4_set_lock_state(state, request);
7457         if (status != 0)
7458                 return status;
7459
7460         return nfs4_retry_setlk(state, cmd, request);
7461 }
7462
7463 static int nfs4_delete_lease(struct file *file, void **priv)
7464 {
7465         return generic_setlease(file, F_UNLCK, NULL, priv);
7466 }
7467
7468 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7469                           void **priv)
7470 {
7471         struct inode *inode = file_inode(file);
7472         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7473         int ret;
7474
7475         /* No delegation, no lease */
7476         if (!nfs4_have_delegation(inode, type))
7477                 return -EAGAIN;
7478         ret = generic_setlease(file, arg, lease, priv);
7479         if (ret || nfs4_have_delegation(inode, type))
7480                 return ret;
7481         /* We raced with a delegation return */
7482         nfs4_delete_lease(file, priv);
7483         return -EAGAIN;
7484 }
7485
7486 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7487                        void **priv)
7488 {
7489         switch (arg) {
7490         case F_RDLCK:
7491         case F_WRLCK:
7492                 return nfs4_add_lease(file, arg, lease, priv);
7493         case F_UNLCK:
7494                 return nfs4_delete_lease(file, priv);
7495         default:
7496                 return -EINVAL;
7497         }
7498 }
7499
7500 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7501 {
7502         struct nfs_server *server = NFS_SERVER(state->inode);
7503         int err;
7504
7505         err = nfs4_set_lock_state(state, fl);
7506         if (err != 0)
7507                 return err;
7508         do {
7509                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7510                 if (err != -NFS4ERR_DELAY)
7511                         break;
7512                 ssleep(1);
7513         } while (err == -NFS4ERR_DELAY);
7514         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7515 }
7516
7517 struct nfs_release_lockowner_data {
7518         struct nfs4_lock_state *lsp;
7519         struct nfs_server *server;
7520         struct nfs_release_lockowner_args args;
7521         struct nfs_release_lockowner_res res;
7522         unsigned long timestamp;
7523 };
7524
7525 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7526 {
7527         struct nfs_release_lockowner_data *data = calldata;
7528         struct nfs_server *server = data->server;
7529         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7530                            &data->res.seq_res, task);
7531         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7532         data->timestamp = jiffies;
7533 }
7534
7535 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7536 {
7537         struct nfs_release_lockowner_data *data = calldata;
7538         struct nfs_server *server = data->server;
7539
7540         nfs40_sequence_done(task, &data->res.seq_res);
7541
7542         switch (task->tk_status) {
7543         case 0:
7544                 renew_lease(server, data->timestamp);
7545                 break;
7546         case -NFS4ERR_STALE_CLIENTID:
7547         case -NFS4ERR_EXPIRED:
7548                 nfs4_schedule_lease_recovery(server->nfs_client);
7549                 break;
7550         case -NFS4ERR_LEASE_MOVED:
7551         case -NFS4ERR_DELAY:
7552                 if (nfs4_async_handle_error(task, server,
7553                                             NULL, NULL) == -EAGAIN)
7554                         rpc_restart_call_prepare(task);
7555         }
7556 }
7557
7558 static void nfs4_release_lockowner_release(void *calldata)
7559 {
7560         struct nfs_release_lockowner_data *data = calldata;
7561         nfs4_free_lock_state(data->server, data->lsp);
7562         kfree(calldata);
7563 }
7564
7565 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7566         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7567         .rpc_call_done = nfs4_release_lockowner_done,
7568         .rpc_release = nfs4_release_lockowner_release,
7569 };
7570
7571 static void
7572 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7573 {
7574         struct nfs_release_lockowner_data *data;
7575         struct rpc_message msg = {
7576                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7577         };
7578
7579         if (server->nfs_client->cl_mvops->minor_version != 0)
7580                 return;
7581
7582         data = kmalloc(sizeof(*data), GFP_NOFS);
7583         if (!data)
7584                 return;
7585         data->lsp = lsp;
7586         data->server = server;
7587         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7588         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7589         data->args.lock_owner.s_dev = server->s_dev;
7590
7591         msg.rpc_argp = &data->args;
7592         msg.rpc_resp = &data->res;
7593         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7594         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7595 }
7596
7597 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7598
7599 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7600                                    struct user_namespace *mnt_userns,
7601                                    struct dentry *unused, struct inode *inode,
7602                                    const char *key, const void *buf,
7603                                    size_t buflen, int flags)
7604 {
7605         return nfs4_proc_set_acl(inode, buf, buflen);
7606 }
7607
7608 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7609                                    struct dentry *unused, struct inode *inode,
7610                                    const char *key, void *buf, size_t buflen)
7611 {
7612         return nfs4_proc_get_acl(inode, buf, buflen);
7613 }
7614
7615 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7616 {
7617         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7618 }
7619
7620 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7621
7622 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7623                                      struct user_namespace *mnt_userns,
7624                                      struct dentry *unused, struct inode *inode,
7625                                      const char *key, const void *buf,
7626                                      size_t buflen, int flags)
7627 {
7628         if (security_ismaclabel(key))
7629                 return nfs4_set_security_label(inode, buf, buflen);
7630
7631         return -EOPNOTSUPP;
7632 }
7633
7634 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7635                                      struct dentry *unused, struct inode *inode,
7636                                      const char *key, void *buf, size_t buflen)
7637 {
7638         if (security_ismaclabel(key))
7639                 return nfs4_get_security_label(inode, buf, buflen);
7640         return -EOPNOTSUPP;
7641 }
7642
7643 static ssize_t
7644 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7645 {
7646         int len = 0;
7647
7648         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7649                 len = security_inode_listsecurity(inode, list, list_len);
7650                 if (len >= 0 && list_len && len > list_len)
7651                         return -ERANGE;
7652         }
7653         return len;
7654 }
7655
7656 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7657         .prefix = XATTR_SECURITY_PREFIX,
7658         .get    = nfs4_xattr_get_nfs4_label,
7659         .set    = nfs4_xattr_set_nfs4_label,
7660 };
7661
7662 #else
7663
7664 static ssize_t
7665 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7666 {
7667         return 0;
7668 }
7669
7670 #endif
7671
7672 #ifdef CONFIG_NFS_V4_2
7673 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7674                                     struct user_namespace *mnt_userns,
7675                                     struct dentry *unused, struct inode *inode,
7676                                     const char *key, const void *buf,
7677                                     size_t buflen, int flags)
7678 {
7679         struct nfs_access_entry cache;
7680         int ret;
7681
7682         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7683                 return -EOPNOTSUPP;
7684
7685         /*
7686          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7687          * flags right now. Handling of xattr operations use the normal
7688          * file read/write permissions.
7689          *
7690          * Just in case the server has other ideas (which RFC 8276 allows),
7691          * do a cached access check for the XA* flags to possibly avoid
7692          * doing an RPC and getting EACCES back.
7693          */
7694         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7695                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7696                         return -EACCES;
7697         }
7698
7699         if (buf == NULL) {
7700                 ret = nfs42_proc_removexattr(inode, key);
7701                 if (!ret)
7702                         nfs4_xattr_cache_remove(inode, key);
7703         } else {
7704                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7705                 if (!ret)
7706                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7707         }
7708
7709         return ret;
7710 }
7711
7712 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7713                                     struct dentry *unused, struct inode *inode,
7714                                     const char *key, void *buf, size_t buflen)
7715 {
7716         struct nfs_access_entry cache;
7717         ssize_t ret;
7718
7719         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7720                 return -EOPNOTSUPP;
7721
7722         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7723                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7724                         return -EACCES;
7725         }
7726
7727         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7728         if (ret)
7729                 return ret;
7730
7731         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7732         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7733                 return ret;
7734
7735         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7736
7737         return ret;
7738 }
7739
7740 static ssize_t
7741 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7742 {
7743         u64 cookie;
7744         bool eof;
7745         ssize_t ret, size;
7746         char *buf;
7747         size_t buflen;
7748         struct nfs_access_entry cache;
7749
7750         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7751                 return 0;
7752
7753         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7754                 if (!(cache.mask & NFS_ACCESS_XALIST))
7755                         return 0;
7756         }
7757
7758         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7759         if (ret)
7760                 return ret;
7761
7762         ret = nfs4_xattr_cache_list(inode, list, list_len);
7763         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7764                 return ret;
7765
7766         cookie = 0;
7767         eof = false;
7768         buflen = list_len ? list_len : XATTR_LIST_MAX;
7769         buf = list_len ? list : NULL;
7770         size = 0;
7771
7772         while (!eof) {
7773                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7774                     &cookie, &eof);
7775                 if (ret < 0)
7776                         return ret;
7777
7778                 if (list_len) {
7779                         buf += ret;
7780                         buflen -= ret;
7781                 }
7782                 size += ret;
7783         }
7784
7785         if (list_len)
7786                 nfs4_xattr_cache_set_list(inode, list, size);
7787
7788         return size;
7789 }
7790
7791 #else
7792
7793 static ssize_t
7794 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7795 {
7796         return 0;
7797 }
7798 #endif /* CONFIG_NFS_V4_2 */
7799
7800 /*
7801  * nfs_fhget will use either the mounted_on_fileid or the fileid
7802  */
7803 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7804 {
7805         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7806                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7807               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7808               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7809                 return;
7810
7811         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7812                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7813         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7814         fattr->nlink = 2;
7815 }
7816
7817 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7818                                    const struct qstr *name,
7819                                    struct nfs4_fs_locations *fs_locations,
7820                                    struct page *page)
7821 {
7822         struct nfs_server *server = NFS_SERVER(dir);
7823         u32 bitmask[3];
7824         struct nfs4_fs_locations_arg args = {
7825                 .dir_fh = NFS_FH(dir),
7826                 .name = name,
7827                 .page = page,
7828                 .bitmask = bitmask,
7829         };
7830         struct nfs4_fs_locations_res res = {
7831                 .fs_locations = fs_locations,
7832         };
7833         struct rpc_message msg = {
7834                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7835                 .rpc_argp = &args,
7836                 .rpc_resp = &res,
7837         };
7838         int status;
7839
7840         dprintk("%s: start\n", __func__);
7841
7842         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7843         bitmask[1] = nfs4_fattr_bitmap[1];
7844
7845         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7846          * is not supported */
7847         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7848                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7849         else
7850                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7851
7852         nfs_fattr_init(&fs_locations->fattr);
7853         fs_locations->server = server;
7854         fs_locations->nlocations = 0;
7855         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7856         dprintk("%s: returned status = %d\n", __func__, status);
7857         return status;
7858 }
7859
7860 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7861                            const struct qstr *name,
7862                            struct nfs4_fs_locations *fs_locations,
7863                            struct page *page)
7864 {
7865         struct nfs4_exception exception = {
7866                 .interruptible = true,
7867         };
7868         int err;
7869         do {
7870                 err = _nfs4_proc_fs_locations(client, dir, name,
7871                                 fs_locations, page);
7872                 trace_nfs4_get_fs_locations(dir, name, err);
7873                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7874                                 &exception);
7875         } while (exception.retry);
7876         return err;
7877 }
7878
7879 /*
7880  * This operation also signals the server that this client is
7881  * performing migration recovery.  The server can stop returning
7882  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7883  * appended to this compound to identify the client ID which is
7884  * performing recovery.
7885  */
7886 static int _nfs40_proc_get_locations(struct inode *inode,
7887                                      struct nfs4_fs_locations *locations,
7888                                      struct page *page, const struct cred *cred)
7889 {
7890         struct nfs_server *server = NFS_SERVER(inode);
7891         struct rpc_clnt *clnt = server->client;
7892         u32 bitmask[2] = {
7893                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7894         };
7895         struct nfs4_fs_locations_arg args = {
7896                 .clientid       = server->nfs_client->cl_clientid,
7897                 .fh             = NFS_FH(inode),
7898                 .page           = page,
7899                 .bitmask        = bitmask,
7900                 .migration      = 1,            /* skip LOOKUP */
7901                 .renew          = 1,            /* append RENEW */
7902         };
7903         struct nfs4_fs_locations_res res = {
7904                 .fs_locations   = locations,
7905                 .migration      = 1,
7906                 .renew          = 1,
7907         };
7908         struct rpc_message msg = {
7909                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7910                 .rpc_argp       = &args,
7911                 .rpc_resp       = &res,
7912                 .rpc_cred       = cred,
7913         };
7914         unsigned long now = jiffies;
7915         int status;
7916
7917         nfs_fattr_init(&locations->fattr);
7918         locations->server = server;
7919         locations->nlocations = 0;
7920
7921         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7922         status = nfs4_call_sync_sequence(clnt, server, &msg,
7923                                         &args.seq_args, &res.seq_res);
7924         if (status)
7925                 return status;
7926
7927         renew_lease(server, now);
7928         return 0;
7929 }
7930
7931 #ifdef CONFIG_NFS_V4_1
7932
7933 /*
7934  * This operation also signals the server that this client is
7935  * performing migration recovery.  The server can stop asserting
7936  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7937  * performing this operation is identified in the SEQUENCE
7938  * operation in this compound.
7939  *
7940  * When the client supports GETATTR(fs_locations_info), it can
7941  * be plumbed in here.
7942  */
7943 static int _nfs41_proc_get_locations(struct inode *inode,
7944                                      struct nfs4_fs_locations *locations,
7945                                      struct page *page, const struct cred *cred)
7946 {
7947         struct nfs_server *server = NFS_SERVER(inode);
7948         struct rpc_clnt *clnt = server->client;
7949         u32 bitmask[2] = {
7950                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7951         };
7952         struct nfs4_fs_locations_arg args = {
7953                 .fh             = NFS_FH(inode),
7954                 .page           = page,
7955                 .bitmask        = bitmask,
7956                 .migration      = 1,            /* skip LOOKUP */
7957         };
7958         struct nfs4_fs_locations_res res = {
7959                 .fs_locations   = locations,
7960                 .migration      = 1,
7961         };
7962         struct rpc_message msg = {
7963                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7964                 .rpc_argp       = &args,
7965                 .rpc_resp       = &res,
7966                 .rpc_cred       = cred,
7967         };
7968         int status;
7969
7970         nfs_fattr_init(&locations->fattr);
7971         locations->server = server;
7972         locations->nlocations = 0;
7973
7974         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7975         status = nfs4_call_sync_sequence(clnt, server, &msg,
7976                                         &args.seq_args, &res.seq_res);
7977         if (status == NFS4_OK &&
7978             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7979                 status = -NFS4ERR_LEASE_MOVED;
7980         return status;
7981 }
7982
7983 #endif  /* CONFIG_NFS_V4_1 */
7984
7985 /**
7986  * nfs4_proc_get_locations - discover locations for a migrated FSID
7987  * @inode: inode on FSID that is migrating
7988  * @locations: result of query
7989  * @page: buffer
7990  * @cred: credential to use for this operation
7991  *
7992  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7993  * operation failed, or a negative errno if a local error occurred.
7994  *
7995  * On success, "locations" is filled in, but if the server has
7996  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7997  * asserted.
7998  *
7999  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8000  * from this client that require migration recovery.
8001  */
8002 int nfs4_proc_get_locations(struct inode *inode,
8003                             struct nfs4_fs_locations *locations,
8004                             struct page *page, const struct cred *cred)
8005 {
8006         struct nfs_server *server = NFS_SERVER(inode);
8007         struct nfs_client *clp = server->nfs_client;
8008         const struct nfs4_mig_recovery_ops *ops =
8009                                         clp->cl_mvops->mig_recovery_ops;
8010         struct nfs4_exception exception = {
8011                 .interruptible = true,
8012         };
8013         int status;
8014
8015         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8016                 (unsigned long long)server->fsid.major,
8017                 (unsigned long long)server->fsid.minor,
8018                 clp->cl_hostname);
8019         nfs_display_fhandle(NFS_FH(inode), __func__);
8020
8021         do {
8022                 status = ops->get_locations(inode, locations, page, cred);
8023                 if (status != -NFS4ERR_DELAY)
8024                         break;
8025                 nfs4_handle_exception(server, status, &exception);
8026         } while (exception.retry);
8027         return status;
8028 }
8029
8030 /*
8031  * This operation also signals the server that this client is
8032  * performing "lease moved" recovery.  The server can stop
8033  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8034  * is appended to this compound to identify the client ID which is
8035  * performing recovery.
8036  */
8037 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8038 {
8039         struct nfs_server *server = NFS_SERVER(inode);
8040         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8041         struct rpc_clnt *clnt = server->client;
8042         struct nfs4_fsid_present_arg args = {
8043                 .fh             = NFS_FH(inode),
8044                 .clientid       = clp->cl_clientid,
8045                 .renew          = 1,            /* append RENEW */
8046         };
8047         struct nfs4_fsid_present_res res = {
8048                 .renew          = 1,
8049         };
8050         struct rpc_message msg = {
8051                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8052                 .rpc_argp       = &args,
8053                 .rpc_resp       = &res,
8054                 .rpc_cred       = cred,
8055         };
8056         unsigned long now = jiffies;
8057         int status;
8058
8059         res.fh = nfs_alloc_fhandle();
8060         if (res.fh == NULL)
8061                 return -ENOMEM;
8062
8063         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8064         status = nfs4_call_sync_sequence(clnt, server, &msg,
8065                                                 &args.seq_args, &res.seq_res);
8066         nfs_free_fhandle(res.fh);
8067         if (status)
8068                 return status;
8069
8070         do_renew_lease(clp, now);
8071         return 0;
8072 }
8073
8074 #ifdef CONFIG_NFS_V4_1
8075
8076 /*
8077  * This operation also signals the server that this client is
8078  * performing "lease moved" recovery.  The server can stop asserting
8079  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8080  * this operation is identified in the SEQUENCE operation in this
8081  * compound.
8082  */
8083 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8084 {
8085         struct nfs_server *server = NFS_SERVER(inode);
8086         struct rpc_clnt *clnt = server->client;
8087         struct nfs4_fsid_present_arg args = {
8088                 .fh             = NFS_FH(inode),
8089         };
8090         struct nfs4_fsid_present_res res = {
8091         };
8092         struct rpc_message msg = {
8093                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8094                 .rpc_argp       = &args,
8095                 .rpc_resp       = &res,
8096                 .rpc_cred       = cred,
8097         };
8098         int status;
8099
8100         res.fh = nfs_alloc_fhandle();
8101         if (res.fh == NULL)
8102                 return -ENOMEM;
8103
8104         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8105         status = nfs4_call_sync_sequence(clnt, server, &msg,
8106                                                 &args.seq_args, &res.seq_res);
8107         nfs_free_fhandle(res.fh);
8108         if (status == NFS4_OK &&
8109             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8110                 status = -NFS4ERR_LEASE_MOVED;
8111         return status;
8112 }
8113
8114 #endif  /* CONFIG_NFS_V4_1 */
8115
8116 /**
8117  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8118  * @inode: inode on FSID to check
8119  * @cred: credential to use for this operation
8120  *
8121  * Server indicates whether the FSID is present, moved, or not
8122  * recognized.  This operation is necessary to clear a LEASE_MOVED
8123  * condition for this client ID.
8124  *
8125  * Returns NFS4_OK if the FSID is present on this server,
8126  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8127  *  NFS4ERR code if some error occurred on the server, or a
8128  *  negative errno if a local failure occurred.
8129  */
8130 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8131 {
8132         struct nfs_server *server = NFS_SERVER(inode);
8133         struct nfs_client *clp = server->nfs_client;
8134         const struct nfs4_mig_recovery_ops *ops =
8135                                         clp->cl_mvops->mig_recovery_ops;
8136         struct nfs4_exception exception = {
8137                 .interruptible = true,
8138         };
8139         int status;
8140
8141         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8142                 (unsigned long long)server->fsid.major,
8143                 (unsigned long long)server->fsid.minor,
8144                 clp->cl_hostname);
8145         nfs_display_fhandle(NFS_FH(inode), __func__);
8146
8147         do {
8148                 status = ops->fsid_present(inode, cred);
8149                 if (status != -NFS4ERR_DELAY)
8150                         break;
8151                 nfs4_handle_exception(server, status, &exception);
8152         } while (exception.retry);
8153         return status;
8154 }
8155
8156 /*
8157  * If 'use_integrity' is true and the state managment nfs_client
8158  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8159  * and the machine credential as per RFC3530bis and RFC5661 Security
8160  * Considerations sections. Otherwise, just use the user cred with the
8161  * filesystem's rpc_client.
8162  */
8163 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8164 {
8165         int status;
8166         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8167         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8168         struct nfs4_secinfo_arg args = {
8169                 .dir_fh = NFS_FH(dir),
8170                 .name   = name,
8171         };
8172         struct nfs4_secinfo_res res = {
8173                 .flavors     = flavors,
8174         };
8175         struct rpc_message msg = {
8176                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8177                 .rpc_argp = &args,
8178                 .rpc_resp = &res,
8179         };
8180         struct nfs4_call_sync_data data = {
8181                 .seq_server = NFS_SERVER(dir),
8182                 .seq_args = &args.seq_args,
8183                 .seq_res = &res.seq_res,
8184         };
8185         struct rpc_task_setup task_setup = {
8186                 .rpc_client = clnt,
8187                 .rpc_message = &msg,
8188                 .callback_ops = clp->cl_mvops->call_sync_ops,
8189                 .callback_data = &data,
8190                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8191         };
8192         const struct cred *cred = NULL;
8193
8194         if (use_integrity) {
8195                 clnt = clp->cl_rpcclient;
8196                 task_setup.rpc_client = clnt;
8197
8198                 cred = nfs4_get_clid_cred(clp);
8199                 msg.rpc_cred = cred;
8200         }
8201
8202         dprintk("NFS call  secinfo %s\n", name->name);
8203
8204         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8205         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8206         status = nfs4_call_sync_custom(&task_setup);
8207
8208         dprintk("NFS reply  secinfo: %d\n", status);
8209
8210         put_cred(cred);
8211         return status;
8212 }
8213
8214 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8215                       struct nfs4_secinfo_flavors *flavors)
8216 {
8217         struct nfs4_exception exception = {
8218                 .interruptible = true,
8219         };
8220         int err;
8221         do {
8222                 err = -NFS4ERR_WRONGSEC;
8223
8224                 /* try to use integrity protection with machine cred */
8225                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8226                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8227
8228                 /*
8229                  * if unable to use integrity protection, or SECINFO with
8230                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8231                  * disallowed by spec, but exists in deployed servers) use
8232                  * the current filesystem's rpc_client and the user cred.
8233                  */
8234                 if (err == -NFS4ERR_WRONGSEC)
8235                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8236
8237                 trace_nfs4_secinfo(dir, name, err);
8238                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8239                                 &exception);
8240         } while (exception.retry);
8241         return err;
8242 }
8243
8244 #ifdef CONFIG_NFS_V4_1
8245 /*
8246  * Check the exchange flags returned by the server for invalid flags, having
8247  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8248  * DS flags set.
8249  */
8250 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8251 {
8252         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8253                 goto out_inval;
8254         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8255                 goto out_inval;
8256         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8257             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8258                 goto out_inval;
8259         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8260                 goto out_inval;
8261         return NFS_OK;
8262 out_inval:
8263         return -NFS4ERR_INVAL;
8264 }
8265
8266 static bool
8267 nfs41_same_server_scope(struct nfs41_server_scope *a,
8268                         struct nfs41_server_scope *b)
8269 {
8270         if (a->server_scope_sz != b->server_scope_sz)
8271                 return false;
8272         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8273 }
8274
8275 static void
8276 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8277 {
8278         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8279         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8280         struct nfs_client *clp = args->client;
8281
8282         switch (task->tk_status) {
8283         case -NFS4ERR_BADSESSION:
8284         case -NFS4ERR_DEADSESSION:
8285                 nfs4_schedule_session_recovery(clp->cl_session,
8286                                 task->tk_status);
8287         }
8288         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8289                         res->dir != NFS4_CDFS4_BOTH) {
8290                 rpc_task_close_connection(task);
8291                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8292                         rpc_restart_call(task);
8293         }
8294 }
8295
8296 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8297         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8298 };
8299
8300 /*
8301  * nfs4_proc_bind_one_conn_to_session()
8302  *
8303  * The 4.1 client currently uses the same TCP connection for the
8304  * fore and backchannel.
8305  */
8306 static
8307 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8308                 struct rpc_xprt *xprt,
8309                 struct nfs_client *clp,
8310                 const struct cred *cred)
8311 {
8312         int status;
8313         struct nfs41_bind_conn_to_session_args args = {
8314                 .client = clp,
8315                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8316                 .retries = 0,
8317         };
8318         struct nfs41_bind_conn_to_session_res res;
8319         struct rpc_message msg = {
8320                 .rpc_proc =
8321                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8322                 .rpc_argp = &args,
8323                 .rpc_resp = &res,
8324                 .rpc_cred = cred,
8325         };
8326         struct rpc_task_setup task_setup_data = {
8327                 .rpc_client = clnt,
8328                 .rpc_xprt = xprt,
8329                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8330                 .rpc_message = &msg,
8331                 .flags = RPC_TASK_TIMEOUT,
8332         };
8333         struct rpc_task *task;
8334
8335         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8336         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8337                 args.dir = NFS4_CDFC4_FORE;
8338
8339         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8340         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8341                 args.dir = NFS4_CDFC4_FORE;
8342
8343         task = rpc_run_task(&task_setup_data);
8344         if (!IS_ERR(task)) {
8345                 status = task->tk_status;
8346                 rpc_put_task(task);
8347         } else
8348                 status = PTR_ERR(task);
8349         trace_nfs4_bind_conn_to_session(clp, status);
8350         if (status == 0) {
8351                 if (memcmp(res.sessionid.data,
8352                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8353                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8354                         return -EIO;
8355                 }
8356                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8357                         dprintk("NFS: %s: Unexpected direction from server\n",
8358                                 __func__);
8359                         return -EIO;
8360                 }
8361                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8362                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8363                                 __func__);
8364                         return -EIO;
8365                 }
8366         }
8367
8368         return status;
8369 }
8370
8371 struct rpc_bind_conn_calldata {
8372         struct nfs_client *clp;
8373         const struct cred *cred;
8374 };
8375
8376 static int
8377 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8378                 struct rpc_xprt *xprt,
8379                 void *calldata)
8380 {
8381         struct rpc_bind_conn_calldata *p = calldata;
8382
8383         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8384 }
8385
8386 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8387 {
8388         struct rpc_bind_conn_calldata data = {
8389                 .clp = clp,
8390                 .cred = cred,
8391         };
8392         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8393                         nfs4_proc_bind_conn_to_session_callback, &data);
8394 }
8395
8396 /*
8397  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8398  * and operations we'd like to see to enable certain features in the allow map
8399  */
8400 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8401         .how = SP4_MACH_CRED,
8402         .enforce.u.words = {
8403                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8404                       1 << (OP_EXCHANGE_ID - 32) |
8405                       1 << (OP_CREATE_SESSION - 32) |
8406                       1 << (OP_DESTROY_SESSION - 32) |
8407                       1 << (OP_DESTROY_CLIENTID - 32)
8408         },
8409         .allow.u.words = {
8410                 [0] = 1 << (OP_CLOSE) |
8411                       1 << (OP_OPEN_DOWNGRADE) |
8412                       1 << (OP_LOCKU) |
8413                       1 << (OP_DELEGRETURN) |
8414                       1 << (OP_COMMIT),
8415                 [1] = 1 << (OP_SECINFO - 32) |
8416                       1 << (OP_SECINFO_NO_NAME - 32) |
8417                       1 << (OP_LAYOUTRETURN - 32) |
8418                       1 << (OP_TEST_STATEID - 32) |
8419                       1 << (OP_FREE_STATEID - 32) |
8420                       1 << (OP_WRITE - 32)
8421         }
8422 };
8423
8424 /*
8425  * Select the state protection mode for client `clp' given the server results
8426  * from exchange_id in `sp'.
8427  *
8428  * Returns 0 on success, negative errno otherwise.
8429  */
8430 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8431                                  struct nfs41_state_protection *sp)
8432 {
8433         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8434                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8435                       1 << (OP_EXCHANGE_ID - 32) |
8436                       1 << (OP_CREATE_SESSION - 32) |
8437                       1 << (OP_DESTROY_SESSION - 32) |
8438                       1 << (OP_DESTROY_CLIENTID - 32)
8439         };
8440         unsigned long flags = 0;
8441         unsigned int i;
8442         int ret = 0;
8443
8444         if (sp->how == SP4_MACH_CRED) {
8445                 /* Print state protect result */
8446                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8447                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8448                         if (test_bit(i, sp->enforce.u.longs))
8449                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8450                         if (test_bit(i, sp->allow.u.longs))
8451                                 dfprintk(MOUNT, "  allow op %d\n", i);
8452                 }
8453
8454                 /* make sure nothing is on enforce list that isn't supported */
8455                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8456                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8457                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8458                                 ret = -EINVAL;
8459                                 goto out;
8460                         }
8461                 }
8462
8463                 /*
8464                  * Minimal mode - state operations are allowed to use machine
8465                  * credential.  Note this already happens by default, so the
8466                  * client doesn't have to do anything more than the negotiation.
8467                  *
8468                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8469                  *       we're already using the machine cred for exchange_id
8470                  *       and will never use a different cred.
8471                  */
8472                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8473                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8474                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8475                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8476                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8477                         dfprintk(MOUNT, "  minimal mode enabled\n");
8478                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8479                 } else {
8480                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8481                         ret = -EINVAL;
8482                         goto out;
8483                 }
8484
8485                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8486                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8487                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8488                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8489                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8490                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8491                 }
8492
8493                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8494                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8495                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8496                 }
8497
8498                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8499                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8500                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8501                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8502                 }
8503
8504                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8505                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8506                         dfprintk(MOUNT, "  stateid mode enabled\n");
8507                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8508                 }
8509
8510                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8511                         dfprintk(MOUNT, "  write mode enabled\n");
8512                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8513                 }
8514
8515                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8516                         dfprintk(MOUNT, "  commit mode enabled\n");
8517                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8518                 }
8519         }
8520 out:
8521         clp->cl_sp4_flags = flags;
8522         return ret;
8523 }
8524
8525 struct nfs41_exchange_id_data {
8526         struct nfs41_exchange_id_res res;
8527         struct nfs41_exchange_id_args args;
8528 };
8529
8530 static void nfs4_exchange_id_release(void *data)
8531 {
8532         struct nfs41_exchange_id_data *cdata =
8533                                         (struct nfs41_exchange_id_data *)data;
8534
8535         nfs_put_client(cdata->args.client);
8536         kfree(cdata->res.impl_id);
8537         kfree(cdata->res.server_scope);
8538         kfree(cdata->res.server_owner);
8539         kfree(cdata);
8540 }
8541
8542 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8543         .rpc_release = nfs4_exchange_id_release,
8544 };
8545
8546 /*
8547  * _nfs4_proc_exchange_id()
8548  *
8549  * Wrapper for EXCHANGE_ID operation.
8550  */
8551 static struct rpc_task *
8552 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8553                         u32 sp4_how, struct rpc_xprt *xprt)
8554 {
8555         struct rpc_message msg = {
8556                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8557                 .rpc_cred = cred,
8558         };
8559         struct rpc_task_setup task_setup_data = {
8560                 .rpc_client = clp->cl_rpcclient,
8561                 .callback_ops = &nfs4_exchange_id_call_ops,
8562                 .rpc_message = &msg,
8563                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8564         };
8565         struct nfs41_exchange_id_data *calldata;
8566         int status;
8567
8568         if (!refcount_inc_not_zero(&clp->cl_count))
8569                 return ERR_PTR(-EIO);
8570
8571         status = -ENOMEM;
8572         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8573         if (!calldata)
8574                 goto out;
8575
8576         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8577
8578         status = nfs4_init_uniform_client_string(clp);
8579         if (status)
8580                 goto out_calldata;
8581
8582         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8583                                                 GFP_NOFS);
8584         status = -ENOMEM;
8585         if (unlikely(calldata->res.server_owner == NULL))
8586                 goto out_calldata;
8587
8588         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8589                                         GFP_NOFS);
8590         if (unlikely(calldata->res.server_scope == NULL))
8591                 goto out_server_owner;
8592
8593         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8594         if (unlikely(calldata->res.impl_id == NULL))
8595                 goto out_server_scope;
8596
8597         switch (sp4_how) {
8598         case SP4_NONE:
8599                 calldata->args.state_protect.how = SP4_NONE;
8600                 break;
8601
8602         case SP4_MACH_CRED:
8603                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8604                 break;
8605
8606         default:
8607                 /* unsupported! */
8608                 WARN_ON_ONCE(1);
8609                 status = -EINVAL;
8610                 goto out_impl_id;
8611         }
8612         if (xprt) {
8613                 task_setup_data.rpc_xprt = xprt;
8614                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8615                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8616                                 sizeof(calldata->args.verifier.data));
8617         }
8618         calldata->args.client = clp;
8619         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8620         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8621 #ifdef CONFIG_NFS_V4_1_MIGRATION
8622         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8623 #endif
8624         msg.rpc_argp = &calldata->args;
8625         msg.rpc_resp = &calldata->res;
8626         task_setup_data.callback_data = calldata;
8627
8628         return rpc_run_task(&task_setup_data);
8629
8630 out_impl_id:
8631         kfree(calldata->res.impl_id);
8632 out_server_scope:
8633         kfree(calldata->res.server_scope);
8634 out_server_owner:
8635         kfree(calldata->res.server_owner);
8636 out_calldata:
8637         kfree(calldata);
8638 out:
8639         nfs_put_client(clp);
8640         return ERR_PTR(status);
8641 }
8642
8643 /*
8644  * _nfs4_proc_exchange_id()
8645  *
8646  * Wrapper for EXCHANGE_ID operation.
8647  */
8648 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8649                         u32 sp4_how)
8650 {
8651         struct rpc_task *task;
8652         struct nfs41_exchange_id_args *argp;
8653         struct nfs41_exchange_id_res *resp;
8654         unsigned long now = jiffies;
8655         int status;
8656
8657         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8658         if (IS_ERR(task))
8659                 return PTR_ERR(task);
8660
8661         argp = task->tk_msg.rpc_argp;
8662         resp = task->tk_msg.rpc_resp;
8663         status = task->tk_status;
8664         if (status  != 0)
8665                 goto out;
8666
8667         status = nfs4_check_cl_exchange_flags(resp->flags,
8668                         clp->cl_mvops->minor_version);
8669         if (status  != 0)
8670                 goto out;
8671
8672         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8673         if (status != 0)
8674                 goto out;
8675
8676         do_renew_lease(clp, now);
8677
8678         clp->cl_clientid = resp->clientid;
8679         clp->cl_exchange_flags = resp->flags;
8680         clp->cl_seqid = resp->seqid;
8681         /* Client ID is not confirmed */
8682         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8683                 clear_bit(NFS4_SESSION_ESTABLISHED,
8684                           &clp->cl_session->session_state);
8685
8686         if (clp->cl_serverscope != NULL &&
8687             !nfs41_same_server_scope(clp->cl_serverscope,
8688                                 resp->server_scope)) {
8689                 dprintk("%s: server_scope mismatch detected\n",
8690                         __func__);
8691                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8692         }
8693
8694         swap(clp->cl_serverowner, resp->server_owner);
8695         swap(clp->cl_serverscope, resp->server_scope);
8696         swap(clp->cl_implid, resp->impl_id);
8697
8698         /* Save the EXCHANGE_ID verifier session trunk tests */
8699         memcpy(clp->cl_confirm.data, argp->verifier.data,
8700                sizeof(clp->cl_confirm.data));
8701 out:
8702         trace_nfs4_exchange_id(clp, status);
8703         rpc_put_task(task);
8704         return status;
8705 }
8706
8707 /*
8708  * nfs4_proc_exchange_id()
8709  *
8710  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8711  *
8712  * Since the clientid has expired, all compounds using sessions
8713  * associated with the stale clientid will be returning
8714  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8715  * be in some phase of session reset.
8716  *
8717  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8718  */
8719 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8720 {
8721         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8722         int status;
8723
8724         /* try SP4_MACH_CRED if krb5i/p */
8725         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8726             authflavor == RPC_AUTH_GSS_KRB5P) {
8727                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8728                 if (!status)
8729                         return 0;
8730         }
8731
8732         /* try SP4_NONE */
8733         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8734 }
8735
8736 /**
8737  * nfs4_test_session_trunk
8738  *
8739  * This is an add_xprt_test() test function called from
8740  * rpc_clnt_setup_test_and_add_xprt.
8741  *
8742  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8743  * and is dereferrenced in nfs4_exchange_id_release
8744  *
8745  * Upon success, add the new transport to the rpc_clnt
8746  *
8747  * @clnt: struct rpc_clnt to get new transport
8748  * @xprt: the rpc_xprt to test
8749  * @data: call data for _nfs4_proc_exchange_id.
8750  */
8751 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8752                             void *data)
8753 {
8754         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8755         struct rpc_task *task;
8756         int status;
8757
8758         u32 sp4_how;
8759
8760         dprintk("--> %s try %s\n", __func__,
8761                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8762
8763         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8764
8765         /* Test connection for session trunking. Async exchange_id call */
8766         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8767         if (IS_ERR(task))
8768                 return;
8769
8770         status = task->tk_status;
8771         if (status == 0)
8772                 status = nfs4_detect_session_trunking(adata->clp,
8773                                 task->tk_msg.rpc_resp, xprt);
8774
8775         if (status == 0)
8776                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8777
8778         rpc_put_task(task);
8779 }
8780 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8781
8782 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8783                 const struct cred *cred)
8784 {
8785         struct rpc_message msg = {
8786                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8787                 .rpc_argp = clp,
8788                 .rpc_cred = cred,
8789         };
8790         int status;
8791
8792         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8793                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8794         trace_nfs4_destroy_clientid(clp, status);
8795         if (status)
8796                 dprintk("NFS: Got error %d from the server %s on "
8797                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8798         return status;
8799 }
8800
8801 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8802                 const struct cred *cred)
8803 {
8804         unsigned int loop;
8805         int ret;
8806
8807         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8808                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8809                 switch (ret) {
8810                 case -NFS4ERR_DELAY:
8811                 case -NFS4ERR_CLIENTID_BUSY:
8812                         ssleep(1);
8813                         break;
8814                 default:
8815                         return ret;
8816                 }
8817         }
8818         return 0;
8819 }
8820
8821 int nfs4_destroy_clientid(struct nfs_client *clp)
8822 {
8823         const struct cred *cred;
8824         int ret = 0;
8825
8826         if (clp->cl_mvops->minor_version < 1)
8827                 goto out;
8828         if (clp->cl_exchange_flags == 0)
8829                 goto out;
8830         if (clp->cl_preserve_clid)
8831                 goto out;
8832         cred = nfs4_get_clid_cred(clp);
8833         ret = nfs4_proc_destroy_clientid(clp, cred);
8834         put_cred(cred);
8835         switch (ret) {
8836         case 0:
8837         case -NFS4ERR_STALE_CLIENTID:
8838                 clp->cl_exchange_flags = 0;
8839         }
8840 out:
8841         return ret;
8842 }
8843
8844 #endif /* CONFIG_NFS_V4_1 */
8845
8846 struct nfs4_get_lease_time_data {
8847         struct nfs4_get_lease_time_args *args;
8848         struct nfs4_get_lease_time_res *res;
8849         struct nfs_client *clp;
8850 };
8851
8852 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8853                                         void *calldata)
8854 {
8855         struct nfs4_get_lease_time_data *data =
8856                         (struct nfs4_get_lease_time_data *)calldata;
8857
8858         dprintk("--> %s\n", __func__);
8859         /* just setup sequence, do not trigger session recovery
8860            since we're invoked within one */
8861         nfs4_setup_sequence(data->clp,
8862                         &data->args->la_seq_args,
8863                         &data->res->lr_seq_res,
8864                         task);
8865         dprintk("<-- %s\n", __func__);
8866 }
8867
8868 /*
8869  * Called from nfs4_state_manager thread for session setup, so don't recover
8870  * from sequence operation or clientid errors.
8871  */
8872 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8873 {
8874         struct nfs4_get_lease_time_data *data =
8875                         (struct nfs4_get_lease_time_data *)calldata;
8876
8877         dprintk("--> %s\n", __func__);
8878         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8879                 return;
8880         switch (task->tk_status) {
8881         case -NFS4ERR_DELAY:
8882         case -NFS4ERR_GRACE:
8883                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8884                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8885                 task->tk_status = 0;
8886                 fallthrough;
8887         case -NFS4ERR_RETRY_UNCACHED_REP:
8888                 rpc_restart_call_prepare(task);
8889                 return;
8890         }
8891         dprintk("<-- %s\n", __func__);
8892 }
8893
8894 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8895         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8896         .rpc_call_done = nfs4_get_lease_time_done,
8897 };
8898
8899 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8900 {
8901         struct nfs4_get_lease_time_args args;
8902         struct nfs4_get_lease_time_res res = {
8903                 .lr_fsinfo = fsinfo,
8904         };
8905         struct nfs4_get_lease_time_data data = {
8906                 .args = &args,
8907                 .res = &res,
8908                 .clp = clp,
8909         };
8910         struct rpc_message msg = {
8911                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8912                 .rpc_argp = &args,
8913                 .rpc_resp = &res,
8914         };
8915         struct rpc_task_setup task_setup = {
8916                 .rpc_client = clp->cl_rpcclient,
8917                 .rpc_message = &msg,
8918                 .callback_ops = &nfs4_get_lease_time_ops,
8919                 .callback_data = &data,
8920                 .flags = RPC_TASK_TIMEOUT,
8921         };
8922
8923         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8924         return nfs4_call_sync_custom(&task_setup);
8925 }
8926
8927 #ifdef CONFIG_NFS_V4_1
8928
8929 /*
8930  * Initialize the values to be used by the client in CREATE_SESSION
8931  * If nfs4_init_session set the fore channel request and response sizes,
8932  * use them.
8933  *
8934  * Set the back channel max_resp_sz_cached to zero to force the client to
8935  * always set csa_cachethis to FALSE because the current implementation
8936  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8937  */
8938 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8939                                     struct rpc_clnt *clnt)
8940 {
8941         unsigned int max_rqst_sz, max_resp_sz;
8942         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8943         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8944
8945         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8946         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8947
8948         /* Fore channel attributes */
8949         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8950         args->fc_attrs.max_resp_sz = max_resp_sz;
8951         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8952         args->fc_attrs.max_reqs = max_session_slots;
8953
8954         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8955                 "max_ops=%u max_reqs=%u\n",
8956                 __func__,
8957                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8958                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8959
8960         /* Back channel attributes */
8961         args->bc_attrs.max_rqst_sz = max_bc_payload;
8962         args->bc_attrs.max_resp_sz = max_bc_payload;
8963         args->bc_attrs.max_resp_sz_cached = 0;
8964         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8965         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8966         if (args->bc_attrs.max_reqs > max_bc_slots)
8967                 args->bc_attrs.max_reqs = max_bc_slots;
8968
8969         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8970                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8971                 __func__,
8972                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8973                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8974                 args->bc_attrs.max_reqs);
8975 }
8976
8977 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8978                 struct nfs41_create_session_res *res)
8979 {
8980         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8981         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8982
8983         if (rcvd->max_resp_sz > sent->max_resp_sz)
8984                 return -EINVAL;
8985         /*
8986          * Our requested max_ops is the minimum we need; we're not
8987          * prepared to break up compounds into smaller pieces than that.
8988          * So, no point even trying to continue if the server won't
8989          * cooperate:
8990          */
8991         if (rcvd->max_ops < sent->max_ops)
8992                 return -EINVAL;
8993         if (rcvd->max_reqs == 0)
8994                 return -EINVAL;
8995         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8996                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8997         return 0;
8998 }
8999
9000 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9001                 struct nfs41_create_session_res *res)
9002 {
9003         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9004         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9005
9006         if (!(res->flags & SESSION4_BACK_CHAN))
9007                 goto out;
9008         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9009                 return -EINVAL;
9010         if (rcvd->max_resp_sz < sent->max_resp_sz)
9011                 return -EINVAL;
9012         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9013                 return -EINVAL;
9014         if (rcvd->max_ops > sent->max_ops)
9015                 return -EINVAL;
9016         if (rcvd->max_reqs > sent->max_reqs)
9017                 return -EINVAL;
9018 out:
9019         return 0;
9020 }
9021
9022 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9023                                      struct nfs41_create_session_res *res)
9024 {
9025         int ret;
9026
9027         ret = nfs4_verify_fore_channel_attrs(args, res);
9028         if (ret)
9029                 return ret;
9030         return nfs4_verify_back_channel_attrs(args, res);
9031 }
9032
9033 static void nfs4_update_session(struct nfs4_session *session,
9034                 struct nfs41_create_session_res *res)
9035 {
9036         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9037         /* Mark client id and session as being confirmed */
9038         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9039         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9040         session->flags = res->flags;
9041         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9042         if (res->flags & SESSION4_BACK_CHAN)
9043                 memcpy(&session->bc_attrs, &res->bc_attrs,
9044                                 sizeof(session->bc_attrs));
9045 }
9046
9047 static int _nfs4_proc_create_session(struct nfs_client *clp,
9048                 const struct cred *cred)
9049 {
9050         struct nfs4_session *session = clp->cl_session;
9051         struct nfs41_create_session_args args = {
9052                 .client = clp,
9053                 .clientid = clp->cl_clientid,
9054                 .seqid = clp->cl_seqid,
9055                 .cb_program = NFS4_CALLBACK,
9056         };
9057         struct nfs41_create_session_res res;
9058
9059         struct rpc_message msg = {
9060                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9061                 .rpc_argp = &args,
9062                 .rpc_resp = &res,
9063                 .rpc_cred = cred,
9064         };
9065         int status;
9066
9067         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9068         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9069
9070         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9071                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9072         trace_nfs4_create_session(clp, status);
9073
9074         switch (status) {
9075         case -NFS4ERR_STALE_CLIENTID:
9076         case -NFS4ERR_DELAY:
9077         case -ETIMEDOUT:
9078         case -EACCES:
9079         case -EAGAIN:
9080                 goto out;
9081         }
9082
9083         clp->cl_seqid++;
9084         if (!status) {
9085                 /* Verify the session's negotiated channel_attrs values */
9086                 status = nfs4_verify_channel_attrs(&args, &res);
9087                 /* Increment the clientid slot sequence id */
9088                 if (status)
9089                         goto out;
9090                 nfs4_update_session(session, &res);
9091         }
9092 out:
9093         return status;
9094 }
9095
9096 /*
9097  * Issues a CREATE_SESSION operation to the server.
9098  * It is the responsibility of the caller to verify the session is
9099  * expired before calling this routine.
9100  */
9101 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9102 {
9103         int status;
9104         unsigned *ptr;
9105         struct nfs4_session *session = clp->cl_session;
9106
9107         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9108
9109         status = _nfs4_proc_create_session(clp, cred);
9110         if (status)
9111                 goto out;
9112
9113         /* Init or reset the session slot tables */
9114         status = nfs4_setup_session_slot_tables(session);
9115         dprintk("slot table setup returned %d\n", status);
9116         if (status)
9117                 goto out;
9118
9119         ptr = (unsigned *)&session->sess_id.data[0];
9120         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9121                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9122 out:
9123         dprintk("<-- %s\n", __func__);
9124         return status;
9125 }
9126
9127 /*
9128  * Issue the over-the-wire RPC DESTROY_SESSION.
9129  * The caller must serialize access to this routine.
9130  */
9131 int nfs4_proc_destroy_session(struct nfs4_session *session,
9132                 const struct cred *cred)
9133 {
9134         struct rpc_message msg = {
9135                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9136                 .rpc_argp = session,
9137                 .rpc_cred = cred,
9138         };
9139         int status = 0;
9140
9141         dprintk("--> nfs4_proc_destroy_session\n");
9142
9143         /* session is still being setup */
9144         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9145                 return 0;
9146
9147         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9148                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9149         trace_nfs4_destroy_session(session->clp, status);
9150
9151         if (status)
9152                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9153                         "Session has been destroyed regardless...\n", status);
9154
9155         dprintk("<-- nfs4_proc_destroy_session\n");
9156         return status;
9157 }
9158
9159 /*
9160  * Renew the cl_session lease.
9161  */
9162 struct nfs4_sequence_data {
9163         struct nfs_client *clp;
9164         struct nfs4_sequence_args args;
9165         struct nfs4_sequence_res res;
9166 };
9167
9168 static void nfs41_sequence_release(void *data)
9169 {
9170         struct nfs4_sequence_data *calldata = data;
9171         struct nfs_client *clp = calldata->clp;
9172
9173         if (refcount_read(&clp->cl_count) > 1)
9174                 nfs4_schedule_state_renewal(clp);
9175         nfs_put_client(clp);
9176         kfree(calldata);
9177 }
9178
9179 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9180 {
9181         switch(task->tk_status) {
9182         case -NFS4ERR_DELAY:
9183                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9184                 return -EAGAIN;
9185         default:
9186                 nfs4_schedule_lease_recovery(clp);
9187         }
9188         return 0;
9189 }
9190
9191 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9192 {
9193         struct nfs4_sequence_data *calldata = data;
9194         struct nfs_client *clp = calldata->clp;
9195
9196         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9197                 return;
9198
9199         trace_nfs4_sequence(clp, task->tk_status);
9200         if (task->tk_status < 0) {
9201                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9202                 if (refcount_read(&clp->cl_count) == 1)
9203                         goto out;
9204
9205                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9206                         rpc_restart_call_prepare(task);
9207                         return;
9208                 }
9209         }
9210         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9211 out:
9212         dprintk("<-- %s\n", __func__);
9213 }
9214
9215 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9216 {
9217         struct nfs4_sequence_data *calldata = data;
9218         struct nfs_client *clp = calldata->clp;
9219         struct nfs4_sequence_args *args;
9220         struct nfs4_sequence_res *res;
9221
9222         args = task->tk_msg.rpc_argp;
9223         res = task->tk_msg.rpc_resp;
9224
9225         nfs4_setup_sequence(clp, args, res, task);
9226 }
9227
9228 static const struct rpc_call_ops nfs41_sequence_ops = {
9229         .rpc_call_done = nfs41_sequence_call_done,
9230         .rpc_call_prepare = nfs41_sequence_prepare,
9231         .rpc_release = nfs41_sequence_release,
9232 };
9233
9234 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9235                 const struct cred *cred,
9236                 struct nfs4_slot *slot,
9237                 bool is_privileged)
9238 {
9239         struct nfs4_sequence_data *calldata;
9240         struct rpc_message msg = {
9241                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9242                 .rpc_cred = cred,
9243         };
9244         struct rpc_task_setup task_setup_data = {
9245                 .rpc_client = clp->cl_rpcclient,
9246                 .rpc_message = &msg,
9247                 .callback_ops = &nfs41_sequence_ops,
9248                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9249         };
9250         struct rpc_task *ret;
9251
9252         ret = ERR_PTR(-EIO);
9253         if (!refcount_inc_not_zero(&clp->cl_count))
9254                 goto out_err;
9255
9256         ret = ERR_PTR(-ENOMEM);
9257         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9258         if (calldata == NULL)
9259                 goto out_put_clp;
9260         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9261         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9262         msg.rpc_argp = &calldata->args;
9263         msg.rpc_resp = &calldata->res;
9264         calldata->clp = clp;
9265         task_setup_data.callback_data = calldata;
9266
9267         ret = rpc_run_task(&task_setup_data);
9268         if (IS_ERR(ret))
9269                 goto out_err;
9270         return ret;
9271 out_put_clp:
9272         nfs_put_client(clp);
9273 out_err:
9274         nfs41_release_slot(slot);
9275         return ret;
9276 }
9277
9278 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9279 {
9280         struct rpc_task *task;
9281         int ret = 0;
9282
9283         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9284                 return -EAGAIN;
9285         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9286         if (IS_ERR(task))
9287                 ret = PTR_ERR(task);
9288         else
9289                 rpc_put_task_async(task);
9290         dprintk("<-- %s status=%d\n", __func__, ret);
9291         return ret;
9292 }
9293
9294 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9295 {
9296         struct rpc_task *task;
9297         int ret;
9298
9299         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9300         if (IS_ERR(task)) {
9301                 ret = PTR_ERR(task);
9302                 goto out;
9303         }
9304         ret = rpc_wait_for_completion_task(task);
9305         if (!ret)
9306                 ret = task->tk_status;
9307         rpc_put_task(task);
9308 out:
9309         dprintk("<-- %s status=%d\n", __func__, ret);
9310         return ret;
9311 }
9312
9313 struct nfs4_reclaim_complete_data {
9314         struct nfs_client *clp;
9315         struct nfs41_reclaim_complete_args arg;
9316         struct nfs41_reclaim_complete_res res;
9317 };
9318
9319 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9320 {
9321         struct nfs4_reclaim_complete_data *calldata = data;
9322
9323         nfs4_setup_sequence(calldata->clp,
9324                         &calldata->arg.seq_args,
9325                         &calldata->res.seq_res,
9326                         task);
9327 }
9328
9329 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9330 {
9331         switch(task->tk_status) {
9332         case 0:
9333                 wake_up_all(&clp->cl_lock_waitq);
9334                 fallthrough;
9335         case -NFS4ERR_COMPLETE_ALREADY:
9336         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9337                 break;
9338         case -NFS4ERR_DELAY:
9339                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9340                 fallthrough;
9341         case -NFS4ERR_RETRY_UNCACHED_REP:
9342                 return -EAGAIN;
9343         case -NFS4ERR_BADSESSION:
9344         case -NFS4ERR_DEADSESSION:
9345         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9346                 break;
9347         default:
9348                 nfs4_schedule_lease_recovery(clp);
9349         }
9350         return 0;
9351 }
9352
9353 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9354 {
9355         struct nfs4_reclaim_complete_data *calldata = data;
9356         struct nfs_client *clp = calldata->clp;
9357         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9358
9359         dprintk("--> %s\n", __func__);
9360         if (!nfs41_sequence_done(task, res))
9361                 return;
9362
9363         trace_nfs4_reclaim_complete(clp, task->tk_status);
9364         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9365                 rpc_restart_call_prepare(task);
9366                 return;
9367         }
9368         dprintk("<-- %s\n", __func__);
9369 }
9370
9371 static void nfs4_free_reclaim_complete_data(void *data)
9372 {
9373         struct nfs4_reclaim_complete_data *calldata = data;
9374
9375         kfree(calldata);
9376 }
9377
9378 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9379         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9380         .rpc_call_done = nfs4_reclaim_complete_done,
9381         .rpc_release = nfs4_free_reclaim_complete_data,
9382 };
9383
9384 /*
9385  * Issue a global reclaim complete.
9386  */
9387 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9388                 const struct cred *cred)
9389 {
9390         struct nfs4_reclaim_complete_data *calldata;
9391         struct rpc_message msg = {
9392                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9393                 .rpc_cred = cred,
9394         };
9395         struct rpc_task_setup task_setup_data = {
9396                 .rpc_client = clp->cl_rpcclient,
9397                 .rpc_message = &msg,
9398                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9399                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9400         };
9401         int status = -ENOMEM;
9402
9403         dprintk("--> %s\n", __func__);
9404         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9405         if (calldata == NULL)
9406                 goto out;
9407         calldata->clp = clp;
9408         calldata->arg.one_fs = 0;
9409
9410         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9411         msg.rpc_argp = &calldata->arg;
9412         msg.rpc_resp = &calldata->res;
9413         task_setup_data.callback_data = calldata;
9414         status = nfs4_call_sync_custom(&task_setup_data);
9415 out:
9416         dprintk("<-- %s status=%d\n", __func__, status);
9417         return status;
9418 }
9419
9420 static void
9421 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9422 {
9423         struct nfs4_layoutget *lgp = calldata;
9424         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9425
9426         dprintk("--> %s\n", __func__);
9427         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9428                                 &lgp->res.seq_res, task);
9429         dprintk("<-- %s\n", __func__);
9430 }
9431
9432 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9433 {
9434         struct nfs4_layoutget *lgp = calldata;
9435
9436         dprintk("--> %s\n", __func__);
9437         nfs41_sequence_process(task, &lgp->res.seq_res);
9438         dprintk("<-- %s\n", __func__);
9439 }
9440
9441 static int
9442 nfs4_layoutget_handle_exception(struct rpc_task *task,
9443                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9444 {
9445         struct inode *inode = lgp->args.inode;
9446         struct nfs_server *server = NFS_SERVER(inode);
9447         struct pnfs_layout_hdr *lo = lgp->lo;
9448         int nfs4err = task->tk_status;
9449         int err, status = 0;
9450         LIST_HEAD(head);
9451
9452         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9453
9454         nfs4_sequence_free_slot(&lgp->res.seq_res);
9455
9456         switch (nfs4err) {
9457         case 0:
9458                 goto out;
9459
9460         /*
9461          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9462          * on the file. set tk_status to -ENODATA to tell upper layer to
9463          * retry go inband.
9464          */
9465         case -NFS4ERR_LAYOUTUNAVAILABLE:
9466                 status = -ENODATA;
9467                 goto out;
9468         /*
9469          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9470          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9471          */
9472         case -NFS4ERR_BADLAYOUT:
9473                 status = -EOVERFLOW;
9474                 goto out;
9475         /*
9476          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9477          * (or clients) writing to the same RAID stripe except when
9478          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9479          *
9480          * Treat it like we would RECALLCONFLICT -- we retry for a little
9481          * while, and then eventually give up.
9482          */
9483         case -NFS4ERR_LAYOUTTRYLATER:
9484                 if (lgp->args.minlength == 0) {
9485                         status = -EOVERFLOW;
9486                         goto out;
9487                 }
9488                 status = -EBUSY;
9489                 break;
9490         case -NFS4ERR_RECALLCONFLICT:
9491                 status = -ERECALLCONFLICT;
9492                 break;
9493         case -NFS4ERR_DELEG_REVOKED:
9494         case -NFS4ERR_ADMIN_REVOKED:
9495         case -NFS4ERR_EXPIRED:
9496         case -NFS4ERR_BAD_STATEID:
9497                 exception->timeout = 0;
9498                 spin_lock(&inode->i_lock);
9499                 /* If the open stateid was bad, then recover it. */
9500                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9501                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9502                         spin_unlock(&inode->i_lock);
9503                         exception->state = lgp->args.ctx->state;
9504                         exception->stateid = &lgp->args.stateid;
9505                         break;
9506                 }
9507
9508                 /*
9509                  * Mark the bad layout state as invalid, then retry
9510                  */
9511                 pnfs_mark_layout_stateid_invalid(lo, &head);
9512                 spin_unlock(&inode->i_lock);
9513                 nfs_commit_inode(inode, 0);
9514                 pnfs_free_lseg_list(&head);
9515                 status = -EAGAIN;
9516                 goto out;
9517         }
9518
9519         err = nfs4_handle_exception(server, nfs4err, exception);
9520         if (!status) {
9521                 if (exception->retry)
9522                         status = -EAGAIN;
9523                 else
9524                         status = err;
9525         }
9526 out:
9527         dprintk("<-- %s\n", __func__);
9528         return status;
9529 }
9530
9531 size_t max_response_pages(struct nfs_server *server)
9532 {
9533         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9534         return nfs_page_array_len(0, max_resp_sz);
9535 }
9536
9537 static void nfs4_layoutget_release(void *calldata)
9538 {
9539         struct nfs4_layoutget *lgp = calldata;
9540
9541         dprintk("--> %s\n", __func__);
9542         nfs4_sequence_free_slot(&lgp->res.seq_res);
9543         pnfs_layoutget_free(lgp);
9544         dprintk("<-- %s\n", __func__);
9545 }
9546
9547 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9548         .rpc_call_prepare = nfs4_layoutget_prepare,
9549         .rpc_call_done = nfs4_layoutget_done,
9550         .rpc_release = nfs4_layoutget_release,
9551 };
9552
9553 struct pnfs_layout_segment *
9554 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9555 {
9556         struct inode *inode = lgp->args.inode;
9557         struct nfs_server *server = NFS_SERVER(inode);
9558         struct rpc_task *task;
9559         struct rpc_message msg = {
9560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9561                 .rpc_argp = &lgp->args,
9562                 .rpc_resp = &lgp->res,
9563                 .rpc_cred = lgp->cred,
9564         };
9565         struct rpc_task_setup task_setup_data = {
9566                 .rpc_client = server->client,
9567                 .rpc_message = &msg,
9568                 .callback_ops = &nfs4_layoutget_call_ops,
9569                 .callback_data = lgp,
9570                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9571                          RPC_TASK_MOVEABLE,
9572         };
9573         struct pnfs_layout_segment *lseg = NULL;
9574         struct nfs4_exception exception = {
9575                 .inode = inode,
9576                 .timeout = *timeout,
9577         };
9578         int status = 0;
9579
9580         dprintk("--> %s\n", __func__);
9581
9582         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9583
9584         task = rpc_run_task(&task_setup_data);
9585
9586         status = rpc_wait_for_completion_task(task);
9587         if (status != 0)
9588                 goto out;
9589
9590         if (task->tk_status < 0) {
9591                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9592                 *timeout = exception.timeout;
9593         } else if (lgp->res.layoutp->len == 0) {
9594                 status = -EAGAIN;
9595                 *timeout = nfs4_update_delay(&exception.timeout);
9596         } else
9597                 lseg = pnfs_layout_process(lgp);
9598 out:
9599         trace_nfs4_layoutget(lgp->args.ctx,
9600                         &lgp->args.range,
9601                         &lgp->res.range,
9602                         &lgp->res.stateid,
9603                         status);
9604
9605         rpc_put_task(task);
9606         dprintk("<-- %s status=%d\n", __func__, status);
9607         if (status)
9608                 return ERR_PTR(status);
9609         return lseg;
9610 }
9611
9612 static void
9613 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9614 {
9615         struct nfs4_layoutreturn *lrp = calldata;
9616
9617         dprintk("--> %s\n", __func__);
9618         nfs4_setup_sequence(lrp->clp,
9619                         &lrp->args.seq_args,
9620                         &lrp->res.seq_res,
9621                         task);
9622         if (!pnfs_layout_is_valid(lrp->args.layout))
9623                 rpc_exit(task, 0);
9624 }
9625
9626 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9627 {
9628         struct nfs4_layoutreturn *lrp = calldata;
9629         struct nfs_server *server;
9630
9631         dprintk("--> %s\n", __func__);
9632
9633         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9634                 return;
9635
9636         /*
9637          * Was there an RPC level error? Assume the call succeeded,
9638          * and that we need to release the layout
9639          */
9640         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9641                 lrp->res.lrs_present = 0;
9642                 return;
9643         }
9644
9645         server = NFS_SERVER(lrp->args.inode);
9646         switch (task->tk_status) {
9647         case -NFS4ERR_OLD_STATEID:
9648                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9649                                         &lrp->args.range,
9650                                         lrp->args.inode))
9651                         goto out_restart;
9652                 fallthrough;
9653         default:
9654                 task->tk_status = 0;
9655                 fallthrough;
9656         case 0:
9657                 break;
9658         case -NFS4ERR_DELAY:
9659                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9660                         break;
9661                 goto out_restart;
9662         }
9663         dprintk("<-- %s\n", __func__);
9664         return;
9665 out_restart:
9666         task->tk_status = 0;
9667         nfs4_sequence_free_slot(&lrp->res.seq_res);
9668         rpc_restart_call_prepare(task);
9669 }
9670
9671 static void nfs4_layoutreturn_release(void *calldata)
9672 {
9673         struct nfs4_layoutreturn *lrp = calldata;
9674         struct pnfs_layout_hdr *lo = lrp->args.layout;
9675
9676         dprintk("--> %s\n", __func__);
9677         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9678                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9679         nfs4_sequence_free_slot(&lrp->res.seq_res);
9680         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9681                 lrp->ld_private.ops->free(&lrp->ld_private);
9682         pnfs_put_layout_hdr(lrp->args.layout);
9683         nfs_iput_and_deactive(lrp->inode);
9684         put_cred(lrp->cred);
9685         kfree(calldata);
9686         dprintk("<-- %s\n", __func__);
9687 }
9688
9689 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9690         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9691         .rpc_call_done = nfs4_layoutreturn_done,
9692         .rpc_release = nfs4_layoutreturn_release,
9693 };
9694
9695 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9696 {
9697         struct rpc_task *task;
9698         struct rpc_message msg = {
9699                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9700                 .rpc_argp = &lrp->args,
9701                 .rpc_resp = &lrp->res,
9702                 .rpc_cred = lrp->cred,
9703         };
9704         struct rpc_task_setup task_setup_data = {
9705                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9706                 .rpc_message = &msg,
9707                 .callback_ops = &nfs4_layoutreturn_call_ops,
9708                 .callback_data = lrp,
9709                 .flags = RPC_TASK_MOVEABLE,
9710         };
9711         int status = 0;
9712
9713         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9714                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9715                         &task_setup_data.rpc_client, &msg);
9716
9717         dprintk("--> %s\n", __func__);
9718         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9719         if (!sync) {
9720                 if (!lrp->inode) {
9721                         nfs4_layoutreturn_release(lrp);
9722                         return -EAGAIN;
9723                 }
9724                 task_setup_data.flags |= RPC_TASK_ASYNC;
9725         }
9726         if (!lrp->inode)
9727                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9728                                    1);
9729         else
9730                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9731                                    0);
9732         task = rpc_run_task(&task_setup_data);
9733         if (IS_ERR(task))
9734                 return PTR_ERR(task);
9735         if (sync)
9736                 status = task->tk_status;
9737         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9738         dprintk("<-- %s status=%d\n", __func__, status);
9739         rpc_put_task(task);
9740         return status;
9741 }
9742
9743 static int
9744 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9745                 struct pnfs_device *pdev,
9746                 const struct cred *cred)
9747 {
9748         struct nfs4_getdeviceinfo_args args = {
9749                 .pdev = pdev,
9750                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9751                         NOTIFY_DEVICEID4_DELETE,
9752         };
9753         struct nfs4_getdeviceinfo_res res = {
9754                 .pdev = pdev,
9755         };
9756         struct rpc_message msg = {
9757                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9758                 .rpc_argp = &args,
9759                 .rpc_resp = &res,
9760                 .rpc_cred = cred,
9761         };
9762         int status;
9763
9764         dprintk("--> %s\n", __func__);
9765         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9766         if (res.notification & ~args.notify_types)
9767                 dprintk("%s: unsupported notification\n", __func__);
9768         if (res.notification != args.notify_types)
9769                 pdev->nocache = 1;
9770
9771         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9772
9773         dprintk("<-- %s status=%d\n", __func__, status);
9774
9775         return status;
9776 }
9777
9778 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9779                 struct pnfs_device *pdev,
9780                 const struct cred *cred)
9781 {
9782         struct nfs4_exception exception = { };
9783         int err;
9784
9785         do {
9786                 err = nfs4_handle_exception(server,
9787                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9788                                         &exception);
9789         } while (exception.retry);
9790         return err;
9791 }
9792 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9793
9794 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9795 {
9796         struct nfs4_layoutcommit_data *data = calldata;
9797         struct nfs_server *server = NFS_SERVER(data->args.inode);
9798
9799         nfs4_setup_sequence(server->nfs_client,
9800                         &data->args.seq_args,
9801                         &data->res.seq_res,
9802                         task);
9803 }
9804
9805 static void
9806 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9807 {
9808         struct nfs4_layoutcommit_data *data = calldata;
9809         struct nfs_server *server = NFS_SERVER(data->args.inode);
9810
9811         if (!nfs41_sequence_done(task, &data->res.seq_res))
9812                 return;
9813
9814         switch (task->tk_status) { /* Just ignore these failures */
9815         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9816         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9817         case -NFS4ERR_BADLAYOUT:     /* no layout */
9818         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9819                 task->tk_status = 0;
9820                 break;
9821         case 0:
9822                 break;
9823         default:
9824                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9825                         rpc_restart_call_prepare(task);
9826                         return;
9827                 }
9828         }
9829 }
9830
9831 static void nfs4_layoutcommit_release(void *calldata)
9832 {
9833         struct nfs4_layoutcommit_data *data = calldata;
9834
9835         pnfs_cleanup_layoutcommit(data);
9836         nfs_post_op_update_inode_force_wcc(data->args.inode,
9837                                            data->res.fattr);
9838         put_cred(data->cred);
9839         nfs_iput_and_deactive(data->inode);
9840         kfree(data);
9841 }
9842
9843 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9844         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9845         .rpc_call_done = nfs4_layoutcommit_done,
9846         .rpc_release = nfs4_layoutcommit_release,
9847 };
9848
9849 int
9850 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9851 {
9852         struct rpc_message msg = {
9853                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9854                 .rpc_argp = &data->args,
9855                 .rpc_resp = &data->res,
9856                 .rpc_cred = data->cred,
9857         };
9858         struct rpc_task_setup task_setup_data = {
9859                 .task = &data->task,
9860                 .rpc_client = NFS_CLIENT(data->args.inode),
9861                 .rpc_message = &msg,
9862                 .callback_ops = &nfs4_layoutcommit_ops,
9863                 .callback_data = data,
9864                 .flags = RPC_TASK_MOVEABLE,
9865         };
9866         struct rpc_task *task;
9867         int status = 0;
9868
9869         dprintk("NFS: initiating layoutcommit call. sync %d "
9870                 "lbw: %llu inode %lu\n", sync,
9871                 data->args.lastbytewritten,
9872                 data->args.inode->i_ino);
9873
9874         if (!sync) {
9875                 data->inode = nfs_igrab_and_active(data->args.inode);
9876                 if (data->inode == NULL) {
9877                         nfs4_layoutcommit_release(data);
9878                         return -EAGAIN;
9879                 }
9880                 task_setup_data.flags = RPC_TASK_ASYNC;
9881         }
9882         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9883         task = rpc_run_task(&task_setup_data);
9884         if (IS_ERR(task))
9885                 return PTR_ERR(task);
9886         if (sync)
9887                 status = task->tk_status;
9888         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9889         dprintk("%s: status %d\n", __func__, status);
9890         rpc_put_task(task);
9891         return status;
9892 }
9893
9894 /*
9895  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9896  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9897  */
9898 static int
9899 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9900                     struct nfs_fsinfo *info,
9901                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9902 {
9903         struct nfs41_secinfo_no_name_args args = {
9904                 .style = SECINFO_STYLE_CURRENT_FH,
9905         };
9906         struct nfs4_secinfo_res res = {
9907                 .flavors = flavors,
9908         };
9909         struct rpc_message msg = {
9910                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9911                 .rpc_argp = &args,
9912                 .rpc_resp = &res,
9913         };
9914         struct nfs4_call_sync_data data = {
9915                 .seq_server = server,
9916                 .seq_args = &args.seq_args,
9917                 .seq_res = &res.seq_res,
9918         };
9919         struct rpc_task_setup task_setup = {
9920                 .rpc_client = server->client,
9921                 .rpc_message = &msg,
9922                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9923                 .callback_data = &data,
9924                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9925         };
9926         const struct cred *cred = NULL;
9927         int status;
9928
9929         if (use_integrity) {
9930                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9931
9932                 cred = nfs4_get_clid_cred(server->nfs_client);
9933                 msg.rpc_cred = cred;
9934         }
9935
9936         dprintk("--> %s\n", __func__);
9937         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9938         status = nfs4_call_sync_custom(&task_setup);
9939         dprintk("<-- %s status=%d\n", __func__, status);
9940
9941         put_cred(cred);
9942
9943         return status;
9944 }
9945
9946 static int
9947 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9948                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9949 {
9950         struct nfs4_exception exception = {
9951                 .interruptible = true,
9952         };
9953         int err;
9954         do {
9955                 /* first try using integrity protection */
9956                 err = -NFS4ERR_WRONGSEC;
9957
9958                 /* try to use integrity protection with machine cred */
9959                 if (_nfs4_is_integrity_protected(server->nfs_client))
9960                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9961                                                           flavors, true);
9962
9963                 /*
9964                  * if unable to use integrity protection, or SECINFO with
9965                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9966                  * disallowed by spec, but exists in deployed servers) use
9967                  * the current filesystem's rpc_client and the user cred.
9968                  */
9969                 if (err == -NFS4ERR_WRONGSEC)
9970                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9971                                                           flavors, false);
9972
9973                 switch (err) {
9974                 case 0:
9975                 case -NFS4ERR_WRONGSEC:
9976                 case -ENOTSUPP:
9977                         goto out;
9978                 default:
9979                         err = nfs4_handle_exception(server, err, &exception);
9980                 }
9981         } while (exception.retry);
9982 out:
9983         return err;
9984 }
9985
9986 static int
9987 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9988                     struct nfs_fsinfo *info)
9989 {
9990         int err;
9991         struct page *page;
9992         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9993         struct nfs4_secinfo_flavors *flavors;
9994         struct nfs4_secinfo4 *secinfo;
9995         int i;
9996
9997         page = alloc_page(GFP_KERNEL);
9998         if (!page) {
9999                 err = -ENOMEM;
10000                 goto out;
10001         }
10002
10003         flavors = page_address(page);
10004         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10005
10006         /*
10007          * Fall back on "guess and check" method if
10008          * the server doesn't support SECINFO_NO_NAME
10009          */
10010         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10011                 err = nfs4_find_root_sec(server, fhandle, info);
10012                 goto out_freepage;
10013         }
10014         if (err)
10015                 goto out_freepage;
10016
10017         for (i = 0; i < flavors->num_flavors; i++) {
10018                 secinfo = &flavors->flavors[i];
10019
10020                 switch (secinfo->flavor) {
10021                 case RPC_AUTH_NULL:
10022                 case RPC_AUTH_UNIX:
10023                 case RPC_AUTH_GSS:
10024                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10025                                         &secinfo->flavor_info);
10026                         break;
10027                 default:
10028                         flavor = RPC_AUTH_MAXFLAVOR;
10029                         break;
10030                 }
10031
10032                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10033                         flavor = RPC_AUTH_MAXFLAVOR;
10034
10035                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10036                         err = nfs4_lookup_root_sec(server, fhandle,
10037                                                    info, flavor);
10038                         if (!err)
10039                                 break;
10040                 }
10041         }
10042
10043         if (flavor == RPC_AUTH_MAXFLAVOR)
10044                 err = -EPERM;
10045
10046 out_freepage:
10047         put_page(page);
10048         if (err == -EACCES)
10049                 return -EPERM;
10050 out:
10051         return err;
10052 }
10053
10054 static int _nfs41_test_stateid(struct nfs_server *server,
10055                 nfs4_stateid *stateid,
10056                 const struct cred *cred)
10057 {
10058         int status;
10059         struct nfs41_test_stateid_args args = {
10060                 .stateid = stateid,
10061         };
10062         struct nfs41_test_stateid_res res;
10063         struct rpc_message msg = {
10064                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10065                 .rpc_argp = &args,
10066                 .rpc_resp = &res,
10067                 .rpc_cred = cred,
10068         };
10069         struct rpc_clnt *rpc_client = server->client;
10070
10071         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10072                 &rpc_client, &msg);
10073
10074         dprintk("NFS call  test_stateid %p\n", stateid);
10075         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10076         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10077                         &args.seq_args, &res.seq_res);
10078         if (status != NFS_OK) {
10079                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10080                 return status;
10081         }
10082         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10083         return -res.status;
10084 }
10085
10086 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10087                 int err, struct nfs4_exception *exception)
10088 {
10089         exception->retry = 0;
10090         switch(err) {
10091         case -NFS4ERR_DELAY:
10092         case -NFS4ERR_RETRY_UNCACHED_REP:
10093                 nfs4_handle_exception(server, err, exception);
10094                 break;
10095         case -NFS4ERR_BADSESSION:
10096         case -NFS4ERR_BADSLOT:
10097         case -NFS4ERR_BAD_HIGH_SLOT:
10098         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10099         case -NFS4ERR_DEADSESSION:
10100                 nfs4_do_handle_exception(server, err, exception);
10101         }
10102 }
10103
10104 /**
10105  * nfs41_test_stateid - perform a TEST_STATEID operation
10106  *
10107  * @server: server / transport on which to perform the operation
10108  * @stateid: state ID to test
10109  * @cred: credential
10110  *
10111  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10112  * Otherwise a negative NFS4ERR value is returned if the operation
10113  * failed or the state ID is not currently valid.
10114  */
10115 static int nfs41_test_stateid(struct nfs_server *server,
10116                 nfs4_stateid *stateid,
10117                 const struct cred *cred)
10118 {
10119         struct nfs4_exception exception = {
10120                 .interruptible = true,
10121         };
10122         int err;
10123         do {
10124                 err = _nfs41_test_stateid(server, stateid, cred);
10125                 nfs4_handle_delay_or_session_error(server, err, &exception);
10126         } while (exception.retry);
10127         return err;
10128 }
10129
10130 struct nfs_free_stateid_data {
10131         struct nfs_server *server;
10132         struct nfs41_free_stateid_args args;
10133         struct nfs41_free_stateid_res res;
10134 };
10135
10136 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10137 {
10138         struct nfs_free_stateid_data *data = calldata;
10139         nfs4_setup_sequence(data->server->nfs_client,
10140                         &data->args.seq_args,
10141                         &data->res.seq_res,
10142                         task);
10143 }
10144
10145 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10146 {
10147         struct nfs_free_stateid_data *data = calldata;
10148
10149         nfs41_sequence_done(task, &data->res.seq_res);
10150
10151         switch (task->tk_status) {
10152         case -NFS4ERR_DELAY:
10153                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10154                         rpc_restart_call_prepare(task);
10155         }
10156 }
10157
10158 static void nfs41_free_stateid_release(void *calldata)
10159 {
10160         kfree(calldata);
10161 }
10162
10163 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10164         .rpc_call_prepare = nfs41_free_stateid_prepare,
10165         .rpc_call_done = nfs41_free_stateid_done,
10166         .rpc_release = nfs41_free_stateid_release,
10167 };
10168
10169 /**
10170  * nfs41_free_stateid - perform a FREE_STATEID operation
10171  *
10172  * @server: server / transport on which to perform the operation
10173  * @stateid: state ID to release
10174  * @cred: credential
10175  * @privileged: set to true if this call needs to be privileged
10176  *
10177  * Note: this function is always asynchronous.
10178  */
10179 static int nfs41_free_stateid(struct nfs_server *server,
10180                 const nfs4_stateid *stateid,
10181                 const struct cred *cred,
10182                 bool privileged)
10183 {
10184         struct rpc_message msg = {
10185                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10186                 .rpc_cred = cred,
10187         };
10188         struct rpc_task_setup task_setup = {
10189                 .rpc_client = server->client,
10190                 .rpc_message = &msg,
10191                 .callback_ops = &nfs41_free_stateid_ops,
10192                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10193         };
10194         struct nfs_free_stateid_data *data;
10195         struct rpc_task *task;
10196
10197         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10198                 &task_setup.rpc_client, &msg);
10199
10200         dprintk("NFS call  free_stateid %p\n", stateid);
10201         data = kmalloc(sizeof(*data), GFP_NOFS);
10202         if (!data)
10203                 return -ENOMEM;
10204         data->server = server;
10205         nfs4_stateid_copy(&data->args.stateid, stateid);
10206
10207         task_setup.callback_data = data;
10208
10209         msg.rpc_argp = &data->args;
10210         msg.rpc_resp = &data->res;
10211         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10212         task = rpc_run_task(&task_setup);
10213         if (IS_ERR(task))
10214                 return PTR_ERR(task);
10215         rpc_put_task(task);
10216         return 0;
10217 }
10218
10219 static void
10220 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10221 {
10222         const struct cred *cred = lsp->ls_state->owner->so_cred;
10223
10224         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10225         nfs4_free_lock_state(server, lsp);
10226 }
10227
10228 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10229                 const nfs4_stateid *s2)
10230 {
10231         if (s1->type != s2->type)
10232                 return false;
10233
10234         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10235                 return false;
10236
10237         if (s1->seqid == s2->seqid)
10238                 return true;
10239
10240         return s1->seqid == 0 || s2->seqid == 0;
10241 }
10242
10243 #endif /* CONFIG_NFS_V4_1 */
10244
10245 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10246                 const nfs4_stateid *s2)
10247 {
10248         return nfs4_stateid_match(s1, s2);
10249 }
10250
10251
10252 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10253         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10254         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10255         .recover_open   = nfs4_open_reclaim,
10256         .recover_lock   = nfs4_lock_reclaim,
10257         .establish_clid = nfs4_init_clientid,
10258         .detect_trunking = nfs40_discover_server_trunking,
10259 };
10260
10261 #if defined(CONFIG_NFS_V4_1)
10262 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10263         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10264         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10265         .recover_open   = nfs4_open_reclaim,
10266         .recover_lock   = nfs4_lock_reclaim,
10267         .establish_clid = nfs41_init_clientid,
10268         .reclaim_complete = nfs41_proc_reclaim_complete,
10269         .detect_trunking = nfs41_discover_server_trunking,
10270 };
10271 #endif /* CONFIG_NFS_V4_1 */
10272
10273 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10274         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10275         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10276         .recover_open   = nfs40_open_expired,
10277         .recover_lock   = nfs4_lock_expired,
10278         .establish_clid = nfs4_init_clientid,
10279 };
10280
10281 #if defined(CONFIG_NFS_V4_1)
10282 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10283         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10284         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10285         .recover_open   = nfs41_open_expired,
10286         .recover_lock   = nfs41_lock_expired,
10287         .establish_clid = nfs41_init_clientid,
10288 };
10289 #endif /* CONFIG_NFS_V4_1 */
10290
10291 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10292         .sched_state_renewal = nfs4_proc_async_renew,
10293         .get_state_renewal_cred = nfs4_get_renew_cred,
10294         .renew_lease = nfs4_proc_renew,
10295 };
10296
10297 #if defined(CONFIG_NFS_V4_1)
10298 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10299         .sched_state_renewal = nfs41_proc_async_sequence,
10300         .get_state_renewal_cred = nfs4_get_machine_cred,
10301         .renew_lease = nfs4_proc_sequence,
10302 };
10303 #endif
10304
10305 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10306         .get_locations = _nfs40_proc_get_locations,
10307         .fsid_present = _nfs40_proc_fsid_present,
10308 };
10309
10310 #if defined(CONFIG_NFS_V4_1)
10311 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10312         .get_locations = _nfs41_proc_get_locations,
10313         .fsid_present = _nfs41_proc_fsid_present,
10314 };
10315 #endif  /* CONFIG_NFS_V4_1 */
10316
10317 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10318         .minor_version = 0,
10319         .init_caps = NFS_CAP_READDIRPLUS
10320                 | NFS_CAP_ATOMIC_OPEN
10321                 | NFS_CAP_POSIX_LOCK,
10322         .init_client = nfs40_init_client,
10323         .shutdown_client = nfs40_shutdown_client,
10324         .match_stateid = nfs4_match_stateid,
10325         .find_root_sec = nfs4_find_root_sec,
10326         .free_lock_state = nfs4_release_lockowner,
10327         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10328         .alloc_seqid = nfs_alloc_seqid,
10329         .call_sync_ops = &nfs40_call_sync_ops,
10330         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10331         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10332         .state_renewal_ops = &nfs40_state_renewal_ops,
10333         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10334 };
10335
10336 #if defined(CONFIG_NFS_V4_1)
10337 static struct nfs_seqid *
10338 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10339 {
10340         return NULL;
10341 }
10342
10343 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10344         .minor_version = 1,
10345         .init_caps = NFS_CAP_READDIRPLUS
10346                 | NFS_CAP_ATOMIC_OPEN
10347                 | NFS_CAP_POSIX_LOCK
10348                 | NFS_CAP_STATEID_NFSV41
10349                 | NFS_CAP_ATOMIC_OPEN_V1
10350                 | NFS_CAP_LGOPEN,
10351         .init_client = nfs41_init_client,
10352         .shutdown_client = nfs41_shutdown_client,
10353         .match_stateid = nfs41_match_stateid,
10354         .find_root_sec = nfs41_find_root_sec,
10355         .free_lock_state = nfs41_free_lock_state,
10356         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10357         .alloc_seqid = nfs_alloc_no_seqid,
10358         .session_trunk = nfs4_test_session_trunk,
10359         .call_sync_ops = &nfs41_call_sync_ops,
10360         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10361         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10362         .state_renewal_ops = &nfs41_state_renewal_ops,
10363         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10364 };
10365 #endif
10366
10367 #if defined(CONFIG_NFS_V4_2)
10368 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10369         .minor_version = 2,
10370         .init_caps = NFS_CAP_READDIRPLUS
10371                 | NFS_CAP_ATOMIC_OPEN
10372                 | NFS_CAP_POSIX_LOCK
10373                 | NFS_CAP_STATEID_NFSV41
10374                 | NFS_CAP_ATOMIC_OPEN_V1
10375                 | NFS_CAP_LGOPEN
10376                 | NFS_CAP_ALLOCATE
10377                 | NFS_CAP_COPY
10378                 | NFS_CAP_OFFLOAD_CANCEL
10379                 | NFS_CAP_COPY_NOTIFY
10380                 | NFS_CAP_DEALLOCATE
10381                 | NFS_CAP_SEEK
10382                 | NFS_CAP_LAYOUTSTATS
10383                 | NFS_CAP_CLONE
10384                 | NFS_CAP_LAYOUTERROR
10385                 | NFS_CAP_READ_PLUS,
10386         .init_client = nfs41_init_client,
10387         .shutdown_client = nfs41_shutdown_client,
10388         .match_stateid = nfs41_match_stateid,
10389         .find_root_sec = nfs41_find_root_sec,
10390         .free_lock_state = nfs41_free_lock_state,
10391         .call_sync_ops = &nfs41_call_sync_ops,
10392         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10393         .alloc_seqid = nfs_alloc_no_seqid,
10394         .session_trunk = nfs4_test_session_trunk,
10395         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10396         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10397         .state_renewal_ops = &nfs41_state_renewal_ops,
10398         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10399 };
10400 #endif
10401
10402 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10403         [0] = &nfs_v4_0_minor_ops,
10404 #if defined(CONFIG_NFS_V4_1)
10405         [1] = &nfs_v4_1_minor_ops,
10406 #endif
10407 #if defined(CONFIG_NFS_V4_2)
10408         [2] = &nfs_v4_2_minor_ops,
10409 #endif
10410 };
10411
10412 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10413 {
10414         ssize_t error, error2, error3;
10415
10416         error = generic_listxattr(dentry, list, size);
10417         if (error < 0)
10418                 return error;
10419         if (list) {
10420                 list += error;
10421                 size -= error;
10422         }
10423
10424         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10425         if (error2 < 0)
10426                 return error2;
10427
10428         if (list) {
10429                 list += error2;
10430                 size -= error2;
10431         }
10432
10433         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10434         if (error3 < 0)
10435                 return error3;
10436
10437         return error + error2 + error3;
10438 }
10439
10440 static const struct inode_operations nfs4_dir_inode_operations = {
10441         .create         = nfs_create,
10442         .lookup         = nfs_lookup,
10443         .atomic_open    = nfs_atomic_open,
10444         .link           = nfs_link,
10445         .unlink         = nfs_unlink,
10446         .symlink        = nfs_symlink,
10447         .mkdir          = nfs_mkdir,
10448         .rmdir          = nfs_rmdir,
10449         .mknod          = nfs_mknod,
10450         .rename         = nfs_rename,
10451         .permission     = nfs_permission,
10452         .getattr        = nfs_getattr,
10453         .setattr        = nfs_setattr,
10454         .listxattr      = nfs4_listxattr,
10455 };
10456
10457 static const struct inode_operations nfs4_file_inode_operations = {
10458         .permission     = nfs_permission,
10459         .getattr        = nfs_getattr,
10460         .setattr        = nfs_setattr,
10461         .listxattr      = nfs4_listxattr,
10462 };
10463
10464 const struct nfs_rpc_ops nfs_v4_clientops = {
10465         .version        = 4,                    /* protocol version */
10466         .dentry_ops     = &nfs4_dentry_operations,
10467         .dir_inode_ops  = &nfs4_dir_inode_operations,
10468         .file_inode_ops = &nfs4_file_inode_operations,
10469         .file_ops       = &nfs4_file_operations,
10470         .getroot        = nfs4_proc_get_root,
10471         .submount       = nfs4_submount,
10472         .try_get_tree   = nfs4_try_get_tree,
10473         .getattr        = nfs4_proc_getattr,
10474         .setattr        = nfs4_proc_setattr,
10475         .lookup         = nfs4_proc_lookup,
10476         .lookupp        = nfs4_proc_lookupp,
10477         .access         = nfs4_proc_access,
10478         .readlink       = nfs4_proc_readlink,
10479         .create         = nfs4_proc_create,
10480         .remove         = nfs4_proc_remove,
10481         .unlink_setup   = nfs4_proc_unlink_setup,
10482         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10483         .unlink_done    = nfs4_proc_unlink_done,
10484         .rename_setup   = nfs4_proc_rename_setup,
10485         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10486         .rename_done    = nfs4_proc_rename_done,
10487         .link           = nfs4_proc_link,
10488         .symlink        = nfs4_proc_symlink,
10489         .mkdir          = nfs4_proc_mkdir,
10490         .rmdir          = nfs4_proc_rmdir,
10491         .readdir        = nfs4_proc_readdir,
10492         .mknod          = nfs4_proc_mknod,
10493         .statfs         = nfs4_proc_statfs,
10494         .fsinfo         = nfs4_proc_fsinfo,
10495         .pathconf       = nfs4_proc_pathconf,
10496         .set_capabilities = nfs4_server_capabilities,
10497         .decode_dirent  = nfs4_decode_dirent,
10498         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10499         .read_setup     = nfs4_proc_read_setup,
10500         .read_done      = nfs4_read_done,
10501         .write_setup    = nfs4_proc_write_setup,
10502         .write_done     = nfs4_write_done,
10503         .commit_setup   = nfs4_proc_commit_setup,
10504         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10505         .commit_done    = nfs4_commit_done,
10506         .lock           = nfs4_proc_lock,
10507         .clear_acl_cache = nfs4_zap_acl_attr,
10508         .close_context  = nfs4_close_context,
10509         .open_context   = nfs4_atomic_open,
10510         .have_delegation = nfs4_have_delegation,
10511         .alloc_client   = nfs4_alloc_client,
10512         .init_client    = nfs4_init_client,
10513         .free_client    = nfs4_free_client,
10514         .create_server  = nfs4_create_server,
10515         .clone_server   = nfs_clone_server,
10516 };
10517
10518 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10519         .name   = XATTR_NAME_NFSV4_ACL,
10520         .list   = nfs4_xattr_list_nfs4_acl,
10521         .get    = nfs4_xattr_get_nfs4_acl,
10522         .set    = nfs4_xattr_set_nfs4_acl,
10523 };
10524
10525 #ifdef CONFIG_NFS_V4_2
10526 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10527         .prefix = XATTR_USER_PREFIX,
10528         .get    = nfs4_xattr_get_nfs4_user,
10529         .set    = nfs4_xattr_set_nfs4_user,
10530 };
10531 #endif
10532
10533 const struct xattr_handler *nfs4_xattr_handlers[] = {
10534         &nfs4_xattr_nfs4_acl_handler,
10535 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10536         &nfs4_xattr_nfs4_label_handler,
10537 #endif
10538 #ifdef CONFIG_NFS_V4_2
10539         &nfs4_xattr_nfs4_user_handler,
10540 #endif
10541         NULL
10542 };