fs/nfsd/nfs4proc.c

   1 /*
   2  *  Server-side procedures for NFSv4.
   3  *
   4  *  Copyright (c) 2002 The Regents of the University of Michigan.
   5  *  All rights reserved.
   6  *
   7  *  Kendrick Smith <kmsmith@umich.edu>
   8  *  Andy Adamson   <andros@umich.edu>
   9  *
  10  *  Redistribution and use in source and binary forms, with or without
  11  *  modification, are permitted provided that the following conditions
  12  *  are met:
  13  *
  14  *  1. Redistributions of source code must retain the above copyright
  15  *     notice, this list of conditions and the following disclaimer.
  16  *  2. Redistributions in binary form must reproduce the above copyright
  17  *     notice, this list of conditions and the following disclaimer in the
  18  *     documentation and/or other materials provided with the distribution.
  19  *  3. Neither the name of the University nor the names of its
  20  *     contributors may be used to endorse or promote products derived
  21  *     from this software without specific prior written permission.
  22  *
  23  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  24  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  25  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  26  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  27  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  30  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  31  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  32  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  33  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  34  */
  35 #include <linux/fs_struct.h>
  36 #include <linux/file.h>
  37 #include <linux/falloc.h>
  38 #include <linux/slab.h>
  39 #include <linux/kthread.h>
  40 #include <linux/namei.h>
  41
  42 #include <linux/sunrpc/addr.h>
  43 #include <linux/nfs_ssc.h>
  44
  45 #include "idmap.h"
  46 #include "cache.h"
  47 #include "xdr4.h"
  48 #include "vfs.h"
  49 #include "current_stateid.h"
  50 #include "netns.h"
  51 #include "acl.h"
  52 #include "pnfs.h"
  53 #include "trace.h"
  54
  55 static bool inter_copy_offload_enable;
  56 module_param(inter_copy_offload_enable, bool, 0644);
  57 MODULE_PARM_DESC(inter_copy_offload_enable,
  58                  "Enable inter server to server copy offload. Default: false");
  59
  60 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
  61 static int nfsd4_ssc_umount_timeout = 900000;           /* default to 15 mins */
  62 module_param(nfsd4_ssc_umount_timeout, int, 0644);
  63 MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
  64                 "idle msecs before unmount export from source server");
  65 #endif
  66
  67 #define NFSDDBG_FACILITY                NFSDDBG_PROC
  68
  69 static u32 nfsd_attrmask[] = {
  70         NFSD_WRITEABLE_ATTRS_WORD0,
  71         NFSD_WRITEABLE_ATTRS_WORD1,
  72         NFSD_WRITEABLE_ATTRS_WORD2
  73 };
  74
  75 static u32 nfsd41_ex_attrmask[] = {
  76         NFSD_SUPPATTR_EXCLCREAT_WORD0,
  77         NFSD_SUPPATTR_EXCLCREAT_WORD1,
  78         NFSD_SUPPATTR_EXCLCREAT_WORD2
  79 };
  80
  81 static __be32
  82 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
  83                    u32 *bmval, u32 *writable)
  84 {
  85         struct dentry *dentry = cstate->current_fh.fh_dentry;
  86         struct svc_export *exp = cstate->current_fh.fh_export;
  87
  88         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
  89                 return nfserr_attrnotsupp;
  90         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
  91                 return nfserr_attrnotsupp;
  92         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
  93                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
  94                 return nfserr_attrnotsupp;
  95         if (writable && !bmval_is_subset(bmval, writable))
  96                 return nfserr_inval;
  97         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
  98                         (bmval[1] & FATTR4_WORD1_MODE))
  99                 return nfserr_inval;
 100         return nfs_ok;
 101 }
 102
 103 static __be32
 104 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
 105         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 106 {
 107         __be32 status = nfs_ok;
 108
 109         if (open->op_create == NFS4_OPEN_CREATE) {
 110                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
 111                     || open->op_createmode == NFS4_CREATE_GUARDED)
 112                         status = check_attr_support(rqstp, cstate,
 113                                         open->op_bmval, nfsd_attrmask);
 114                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
 115                         status = check_attr_support(rqstp, cstate,
 116                                         open->op_bmval, nfsd41_ex_attrmask);
 117         }
 118
 119         return status;
 120 }
 121
 122 static int
 123 is_create_with_attrs(struct nfsd4_open *open)
 124 {
 125         return open->op_create == NFS4_OPEN_CREATE
 126                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
 127                     || open->op_createmode == NFS4_CREATE_GUARDED
 128                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
 129 }
 130
 131 static inline void
 132 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
 133 {
 134         fh_put(dst);
 135         dget(src->fh_dentry);
 136         if (src->fh_export)
 137                 exp_get(src->fh_export);
 138         *dst = *src;
 139 }
 140
 141 static __be32
 142 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
 143 {
 144
 145         if (open->op_truncate &&
 146                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
 147                 return nfserr_inval;
 148
 149         accmode |= NFSD_MAY_READ_IF_EXEC;
 150
 151         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
 152                 accmode |= NFSD_MAY_READ;
 153         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
 154                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
 155         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
 156                 accmode |= NFSD_MAY_WRITE;
 157
 158         return fh_verify(rqstp, current_fh, S_IFREG, accmode);
 159 }
 160
 161 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
 162 {
 163         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
 164
 165         if (S_ISREG(mode))
 166                 return nfs_ok;
 167         if (S_ISDIR(mode))
 168                 return nfserr_isdir;
 169         /*
 170          * Using err_symlink as our catch-all case may look odd; but
 171          * there's no other obvious error for this case in 4.0, and we
 172          * happen to know that it will cause the linux v4 client to do
 173          * the right thing on attempts to open something other than a
 174          * regular file.
 175          */
 176         return nfserr_symlink;
 177 }
 178
 179 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
 180 {
 181         if (nfsd4_has_session(cstate))
 182                 return;
 183         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
 184                         &resfh->fh_handle);
 185 }
 186
 187 static inline bool nfsd4_create_is_exclusive(int createmode)
 188 {
 189         return createmode == NFS4_CREATE_EXCLUSIVE ||
 190                 createmode == NFS4_CREATE_EXCLUSIVE4_1;
 191 }
 192
 193 static __be32
 194 nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
 195                  struct nfsd4_open *open)
 196 {
 197         struct file *filp;
 198         struct path path;
 199         int oflags;
 200
 201         oflags = O_CREAT | O_LARGEFILE;
 202         switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
 203         case NFS4_SHARE_ACCESS_WRITE:
 204                 oflags |= O_WRONLY;
 205                 break;
 206         case NFS4_SHARE_ACCESS_BOTH:
 207                 oflags |= O_RDWR;
 208                 break;
 209         default:
 210                 oflags |= O_RDONLY;
 211         }
 212
 213         path.mnt = fhp->fh_export->ex_path.mnt;
 214         path.dentry = child;
 215         filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
 216                              current_cred());
 217         if (IS_ERR(filp))
 218                 return nfserrno(PTR_ERR(filp));
 219
 220         open->op_filp = filp;
 221         return nfs_ok;
 222 }
 223
 224 /*
 225  * Implement NFSv4's unchecked, guarded, and exclusive create
 226  * semantics for regular files. Open state for this new file is
 227  * subsequently fabricated in nfsd4_process_open2().
 228  *
 229  * Upon return, caller must release @fhp and @resfhp.
 230  */
 231 static __be32
 232 nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
 233                   struct svc_fh *resfhp, struct nfsd4_open *open)
 234 {
 235         struct iattr *iap = &open->op_iattr;
 236         struct nfsd_attrs attrs = {
 237                 .na_iattr       = iap,
 238                 .na_seclabel    = &open->op_label,
 239         };
 240         struct dentry *parent, *child;
 241         __u32 v_mtime, v_atime;
 242         struct inode *inode;
 243         __be32 status;
 244         int host_err;
 245
 246         if (isdotent(open->op_fname, open->op_fnamelen))
 247                 return nfserr_exist;
 248         if (!(iap->ia_valid & ATTR_MODE))
 249                 iap->ia_mode = 0;
 250
 251         status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
 252         if (status != nfs_ok)
 253                 return status;
 254         parent = fhp->fh_dentry;
 255         inode = d_inode(parent);
 256
 257         host_err = fh_want_write(fhp);
 258         if (host_err)
 259                 return nfserrno(host_err);
 260
 261         if (is_create_with_attrs(open))
 262                 nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
 263
 264         inode_lock_nested(inode, I_MUTEX_PARENT);
 265
 266         child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
 267         if (IS_ERR(child)) {
 268                 status = nfserrno(PTR_ERR(child));
 269                 goto out;
 270         }
 271
 272         if (d_really_is_negative(child)) {
 273                 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
 274                 if (status != nfs_ok)
 275                         goto out;
 276         }
 277
 278         status = fh_compose(resfhp, fhp->fh_export, child, fhp);
 279         if (status != nfs_ok)
 280                 goto out;
 281
 282         v_mtime = 0;
 283         v_atime = 0;
 284         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 285                 u32 *verifier = (u32 *)open->op_verf.data;
 286
 287                 /*
 288                  * Solaris 7 gets confused (bugid 4218508) if these have
 289                  * the high bit set, as do xfs filesystems without the
 290                  * "bigtime" feature. So just clear the high bits. If this
 291                  * is ever changed to use different attrs for storing the
 292                  * verifier, then do_open_lookup() will also need to be
 293                  * fixed accordingly.
 294                  */
 295                 v_mtime = verifier[0] & 0x7fffffff;
 296                 v_atime = verifier[1] & 0x7fffffff;
 297         }
 298
 299         if (d_really_is_positive(child)) {
 300                 status = nfs_ok;
 301
 302                 /* NFSv4 protocol requires change attributes even though
 303                  * no change happened.
 304                  */
 305                 fh_fill_both_attrs(fhp);
 306
 307                 switch (open->op_createmode) {
 308                 case NFS4_CREATE_UNCHECKED:
 309                         if (!d_is_reg(child))
 310                                 break;
 311
 312                         /*
 313                          * In NFSv4, we don't want to truncate the file
 314                          * now. This would be wrong if the OPEN fails for
 315                          * some other reason. Furthermore, if the size is
 316                          * nonzero, we should ignore it according to spec!
 317                          */
 318                         open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
 319                                                 !iap->ia_size;
 320                         break;
 321                 case NFS4_CREATE_GUARDED:
 322                         status = nfserr_exist;
 323                         break;
 324                 case NFS4_CREATE_EXCLUSIVE:
 325                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
 326                             d_inode(child)->i_atime.tv_sec == v_atime &&
 327                             d_inode(child)->i_size == 0) {
 328                                 open->op_created = true;
 329                                 break;          /* subtle */
 330                         }
 331                         status = nfserr_exist;
 332                         break;
 333                 case NFS4_CREATE_EXCLUSIVE4_1:
 334                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
 335                             d_inode(child)->i_atime.tv_sec == v_atime &&
 336                             d_inode(child)->i_size == 0) {
 337                                 open->op_created = true;
 338                                 goto set_attr;  /* subtle */
 339                         }
 340                         status = nfserr_exist;
 341                 }
 342                 goto out;
 343         }
 344
 345         if (!IS_POSIXACL(inode))
 346                 iap->ia_mode &= ~current_umask();
 347
 348         fh_fill_pre_attrs(fhp);
 349         status = nfsd4_vfs_create(fhp, child, open);
 350         if (status != nfs_ok)
 351                 goto out;
 352         open->op_created = true;
 353         fh_fill_post_attrs(fhp);
 354
 355         /* A newly created file already has a file size of zero. */
 356         if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
 357                 iap->ia_valid &= ~ATTR_SIZE;
 358         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 359                 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
 360                                 ATTR_MTIME_SET|ATTR_ATIME_SET;
 361                 iap->ia_mtime.tv_sec = v_mtime;
 362                 iap->ia_atime.tv_sec = v_atime;
 363                 iap->ia_mtime.tv_nsec = 0;
 364                 iap->ia_atime.tv_nsec = 0;
 365         }
 366
 367 set_attr:
 368         status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
 369
 370         if (attrs.na_labelerr)
 371                 open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 372         if (attrs.na_aclerr)
 373                 open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
 374 out:
 375         inode_unlock(inode);
 376         nfsd_attrs_free(&attrs);
 377         if (child && !IS_ERR(child))
 378                 dput(child);
 379         fh_drop_write(fhp);
 380         return status;
 381 }
 382
 383 static __be32
 384 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
 385 {
 386         struct svc_fh *current_fh = &cstate->current_fh;
 387         int accmode;
 388         __be32 status;
 389
 390         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
 391         if (!*resfh)
 392                 return nfserr_jukebox;
 393         fh_init(*resfh, NFS4_FHSIZE);
 394         open->op_truncate = false;
 395
 396         if (open->op_create) {
 397                 /* FIXME: check session persistence and pnfs flags.
 398                  * The nfsv4.1 spec requires the following semantics:
 399                  *
 400                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
 401                  * Reply Cache  | server |                 |
 402                  * -------------+--------+-----------------+--------------------
 403                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
 404                  *              |        |                 | (SHOULD)
 405                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
 406                  *              |        |                 | (SHOULD NOT)
 407                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
 408                  * yes          | no     | GUARDED4        | GUARDED4
 409                  * yes          | yes    | GUARDED4        | GUARDED4
 410                  */
 411
 412                 current->fs->umask = open->op_umask;
 413                 status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
 414                 current->fs->umask = 0;
 415
 416                 /*
 417                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
 418                  * use the returned bitmask to indicate which attributes
 419                  * we used to store the verifier:
 420                  */
 421                 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
 422                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
 423                                                 FATTR4_WORD1_TIME_MODIFY);
 424         } else {
 425                 status = nfsd_lookup(rqstp, current_fh,
 426                                      open->op_fname, open->op_fnamelen, *resfh);
 427                 if (!status)
 428                         /* NFSv4 protocol requires change attributes even though
 429                          * no change happened.
 430                          */
 431                         fh_fill_both_attrs(current_fh);
 432         }
 433         if (status)
 434                 goto out;
 435         status = nfsd_check_obj_isreg(*resfh);
 436         if (status)
 437                 goto out;
 438
 439         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
 440         accmode = NFSD_MAY_NOP;
 441         if (open->op_created ||
 442                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
 443                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
 444         status = do_open_permission(rqstp, *resfh, open, accmode);
 445         set_change_info(&open->op_cinfo, current_fh);
 446 out:
 447         return status;
 448 }
 449
 450 static __be32
 451 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 452 {
 453         struct svc_fh *current_fh = &cstate->current_fh;
 454         int accmode = 0;
 455
 456         /* We don't know the target directory, and therefore can not
 457         * set the change info
 458         */
 459
 460         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
 461
 462         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
 463
 464         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
 465                 (open->op_iattr.ia_size == 0);
 466         /*
 467          * In the delegation case, the client is telling us about an
 468          * open that it *already* performed locally, some time ago.  We
 469          * should let it succeed now if possible.
 470          *
 471          * In the case of a CLAIM_FH open, on the other hand, the client
 472          * may be counting on us to enforce permissions (the Linux 4.1
 473          * client uses this for normal opens, for example).
 474          */
 475         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
 476                 accmode = NFSD_MAY_OWNER_OVERRIDE;
 477
 478         return do_open_permission(rqstp, current_fh, open, accmode);
 479 }
 480
 481 static void
 482 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
 483 {
 484         struct nfsd4_sessionid *sid =
 485                         (struct nfsd4_sessionid *)session->se_sessionid.data;
 486
 487         clid->cl_boot = sid->clientid.cl_boot;
 488         clid->cl_id = sid->clientid.cl_id;
 489 }
 490
 491 static __be32
 492 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 493            union nfsd4_op_u *u)
 494 {
 495         struct nfsd4_open *open = &u->open;
 496         __be32 status;
 497         struct svc_fh *resfh = NULL;
 498         struct net *net = SVC_NET(rqstp);
 499         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 500         bool reclaim = false;
 501
 502         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
 503                 (int)open->op_fnamelen, open->op_fname,
 504                 open->op_openowner);
 505
 506         open->op_filp = NULL;
 507         open->op_rqstp = rqstp;
 508
 509         /* This check required by spec. */
 510         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
 511                 return nfserr_inval;
 512
 513         open->op_created = false;
 514         /*
 515          * RFC5661 18.51.3
 516          * Before RECLAIM_COMPLETE done, server should deny new lock
 517          */
 518         if (nfsd4_has_session(cstate) &&
 519             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
 520             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 521                 return nfserr_grace;
 522
 523         if (nfsd4_has_session(cstate))
 524                 copy_clientid(&open->op_clientid, cstate->session);
 525
 526         /* check seqid for replay. set nfs4_owner */
 527         status = nfsd4_process_open1(cstate, open, nn);
 528         if (status == nfserr_replay_me) {
 529                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
 530                 fh_put(&cstate->current_fh);
 531                 fh_copy_shallow(&cstate->current_fh.fh_handle,
 532                                 &rp->rp_openfh);
 533                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 534                 if (status)
 535                         dprintk("nfsd4_open: replay failed"
 536                                 " restoring previous filehandle\n");
 537                 else
 538                         status = nfserr_replay_me;
 539         }
 540         if (status)
 541                 goto out;
 542         if (open->op_xdr_error) {
 543                 status = open->op_xdr_error;
 544                 goto out;
 545         }
 546
 547         status = nfsd4_check_open_attributes(rqstp, cstate, open);
 548         if (status)
 549                 goto out;
 550
 551         /* Openowner is now set, so sequence id will get bumped.  Now we need
 552          * these checks before we do any creates: */
 553         status = nfserr_grace;
 554         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 555                 goto out;
 556         status = nfserr_no_grace;
 557         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
 558                 goto out;
 559
 560         switch (open->op_claim_type) {
 561         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 562         case NFS4_OPEN_CLAIM_NULL:
 563                 status = do_open_lookup(rqstp, cstate, open, &resfh);
 564                 if (status)
 565                         goto out;
 566                 break;
 567         case NFS4_OPEN_CLAIM_PREVIOUS:
 568                 status = nfs4_check_open_reclaim(cstate->clp);
 569                 if (status)
 570                         goto out;
 571                 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
 572                 reclaim = true;
 573                 fallthrough;
 574         case NFS4_OPEN_CLAIM_FH:
 575         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 576                 status = do_open_fhandle(rqstp, cstate, open);
 577                 if (status)
 578                         goto out;
 579                 resfh = &cstate->current_fh;
 580                 break;
 581         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 582         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 583                 status = nfserr_notsupp;
 584                 goto out;
 585         default:
 586                 status = nfserr_inval;
 587                 goto out;
 588         }
 589
 590         status = nfsd4_process_open2(rqstp, resfh, open);
 591         if (status && open->op_created)
 592                 pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
 593                         be32_to_cpu(status));
 594         if (reclaim && !status)
 595                 nn->somebody_reclaimed = true;
 596 out:
 597         if (open->op_filp) {
 598                 fput(open->op_filp);
 599                 open->op_filp = NULL;
 600         }
 601         if (resfh && resfh != &cstate->current_fh) {
 602                 fh_dup2(&cstate->current_fh, resfh);
 603                 fh_put(resfh);
 604                 kfree(resfh);
 605         }
 606         nfsd4_cleanup_open_state(cstate, open);
 607         nfsd4_bump_seqid(cstate, status);
 608         return status;
 609 }
 610
 611 /*
 612  * OPEN is the only seqid-mutating operation whose decoding can fail
 613  * with a seqid-mutating error (specifically, decoding of user names in
 614  * the attributes).  Therefore we have to do some processing to look up
 615  * the stateowner so that we can bump the seqid.
 616  */
 617 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
 618 {
 619         struct nfsd4_open *open = &op->u.open;
 620
 621         if (!seqid_mutating_err(ntohl(op->status)))
 622                 return op->status;
 623         if (nfsd4_has_session(cstate))
 624                 return op->status;
 625         open->op_xdr_error = op->status;
 626         return nfsd4_open(rqstp, cstate, &op->u);
 627 }
 628
 629 /*
 630  * filehandle-manipulating ops.
 631  */
 632 static __be32
 633 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 634             union nfsd4_op_u *u)
 635 {
 636         u->getfh = &cstate->current_fh;
 637         return nfs_ok;
 638 }
 639
 640 static __be32
 641 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 642             union nfsd4_op_u *u)
 643 {
 644         struct nfsd4_putfh *putfh = &u->putfh;
 645         __be32 ret;
 646
 647         fh_put(&cstate->current_fh);
 648         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
 649         memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
 650                putfh->pf_fhlen);
 651         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
 652 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
 653         if (ret == nfserr_stale && putfh->no_verify) {
 654                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
 655                 ret = 0;
 656         }
 657 #endif
 658         return ret;
 659 }
 660
 661 static __be32
 662 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 663                 union nfsd4_op_u *u)
 664 {
 665         fh_put(&cstate->current_fh);
 666
 667         return exp_pseudoroot(rqstp, &cstate->current_fh);
 668 }
 669
 670 static __be32
 671 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 672                 union nfsd4_op_u *u)
 673 {
 674         if (!cstate->save_fh.fh_dentry)
 675                 return nfserr_restorefh;
 676
 677         fh_dup2(&cstate->current_fh, &cstate->save_fh);
 678         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
 679                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
 680                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
 681         }
 682         return nfs_ok;
 683 }
 684
 685 static __be32
 686 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 687              union nfsd4_op_u *u)
 688 {
 689         fh_dup2(&cstate->save_fh, &cstate->current_fh);
 690         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
 691                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
 692                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
 693         }
 694         return nfs_ok;
 695 }
 696
 697 /*
 698  * misc nfsv4 ops
 699  */
 700 static __be32
 701 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 702              union nfsd4_op_u *u)
 703 {
 704         struct nfsd4_access *access = &u->access;
 705         u32 access_full;
 706
 707         access_full = NFS3_ACCESS_FULL;
 708         if (cstate->minorversion >= 2)
 709                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
 710                                NFS4_ACCESS_XAWRITE;
 711
 712         if (access->ac_req_access & ~access_full)
 713                 return nfserr_inval;
 714
 715         access->ac_resp_access = access->ac_req_access;
 716         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
 717                            &access->ac_supported);
 718 }
 719
 720 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
 721 {
 722         __be32 *verf = (__be32 *)verifier->data;
 723
 724         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
 725
 726         nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
 727 }
 728
 729 static __be32
 730 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 731              union nfsd4_op_u *u)
 732 {
 733         struct nfsd4_commit *commit = &u->commit;
 734
 735         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
 736                              commit->co_count,
 737                              (__be32 *)commit->co_verf.data);
 738 }
 739
 740 static __be32
 741 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 742              union nfsd4_op_u *u)
 743 {
 744         struct nfsd4_create *create = &u->create;
 745         struct nfsd_attrs attrs = {
 746                 .na_iattr       = &create->cr_iattr,
 747                 .na_seclabel    = &create->cr_label,
 748         };
 749         struct svc_fh resfh;
 750         __be32 status;
 751         dev_t rdev;
 752
 753         fh_init(&resfh, NFS4_FHSIZE);
 754
 755         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 756         if (status)
 757                 return status;
 758
 759         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 760                                     nfsd_attrmask);
 761         if (status)
 762                 return status;
 763
 764         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 765         current->fs->umask = create->cr_umask;
 766         switch (create->cr_type) {
 767         case NF4LNK:
 768                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 769                                       create->cr_name, create->cr_namelen,
 770                                       create->cr_data, &attrs, &resfh);
 771                 break;
 772
 773         case NF4BLK:
 774                 status = nfserr_inval;
 775                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 776                 if (MAJOR(rdev) != create->cr_specdata1 ||
 777                     MINOR(rdev) != create->cr_specdata2)
 778                         goto out_umask;
 779                 status = nfsd_create(rqstp, &cstate->current_fh,
 780                                      create->cr_name, create->cr_namelen,
 781                                      &attrs, S_IFBLK, rdev, &resfh);
 782                 break;
 783
 784         case NF4CHR:
 785                 status = nfserr_inval;
 786                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 787                 if (MAJOR(rdev) != create->cr_specdata1 ||
 788                     MINOR(rdev) != create->cr_specdata2)
 789                         goto out_umask;
 790                 status = nfsd_create(rqstp, &cstate->current_fh,
 791                                      create->cr_name, create->cr_namelen,
 792                                      &attrs, S_IFCHR, rdev, &resfh);
 793                 break;
 794
 795         case NF4SOCK:
 796                 status = nfsd_create(rqstp, &cstate->current_fh,
 797                                      create->cr_name, create->cr_namelen,
 798                                      &attrs, S_IFSOCK, 0, &resfh);
 799                 break;
 800
 801         case NF4FIFO:
 802                 status = nfsd_create(rqstp, &cstate->current_fh,
 803                                      create->cr_name, create->cr_namelen,
 804                                      &attrs, S_IFIFO, 0, &resfh);
 805                 break;
 806
 807         case NF4DIR:
 808                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 809                 status = nfsd_create(rqstp, &cstate->current_fh,
 810                                      create->cr_name, create->cr_namelen,
 811                                      &attrs, S_IFDIR, 0, &resfh);
 812                 break;
 813
 814         default:
 815                 status = nfserr_badtype;
 816         }
 817
 818         if (status)
 819                 goto out;
 820
 821         if (attrs.na_labelerr)
 822                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 823         if (attrs.na_aclerr)
 824                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 825         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 826         fh_dup2(&cstate->current_fh, &resfh);
 827 out:
 828         fh_put(&resfh);
 829 out_umask:
 830         current->fs->umask = 0;
 831         nfsd_attrs_free(&attrs);
 832         return status;
 833 }
 834
 835 static __be32
 836 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 837               union nfsd4_op_u *u)
 838 {
 839         struct nfsd4_getattr *getattr = &u->getattr;
 840         __be32 status;
 841
 842         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 843         if (status)
 844                 return status;
 845
 846         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 847                 return nfserr_inval;
 848
 849         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 850         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 851         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 852
 853         getattr->ga_fhp = &cstate->current_fh;
 854         return nfs_ok;
 855 }
 856
 857 static __be32
 858 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 859            union nfsd4_op_u *u)
 860 {
 861         struct nfsd4_link *link = &u->link;
 862         __be32 status;
 863
 864         status = nfsd_link(rqstp, &cstate->current_fh,
 865                            link->li_name, link->li_namelen, &cstate->save_fh);
 866         if (!status)
 867                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 868         return status;
 869 }
 870
 871 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 872 {
 873         struct svc_fh tmp_fh;
 874         __be32 ret;
 875
 876         fh_init(&tmp_fh, NFS4_FHSIZE);
 877         ret = exp_pseudoroot(rqstp, &tmp_fh);
 878         if (ret)
 879                 return ret;
 880         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 881                 fh_put(&tmp_fh);
 882                 return nfserr_noent;
 883         }
 884         fh_put(&tmp_fh);
 885         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 886 }
 887
 888 static __be32
 889 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 890               union nfsd4_op_u *u)
 891 {
 892         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 893 }
 894
 895 static __be32
 896 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 897              union nfsd4_op_u *u)
 898 {
 899         return nfsd_lookup(rqstp, &cstate->current_fh,
 900                            u->lookup.lo_name, u->lookup.lo_len,
 901                            &cstate->current_fh);
 902 }
 903
 904 static __be32
 905 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 906            union nfsd4_op_u *u)
 907 {
 908         struct nfsd4_read *read = &u->read;
 909         __be32 status;
 910
 911         read->rd_nf = NULL;
 912
 913         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 914                               read->rd_offset, read->rd_length);
 915
 916         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 917         if (read->rd_offset > (u64)OFFSET_MAX)
 918                 read->rd_offset = (u64)OFFSET_MAX;
 919         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 920                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 921
 922         /*
 923          * If we do a zero copy read, then a client will see read data
 924          * that reflects the state of the file *after* performing the
 925          * following compound.
 926          *
 927          * To ensure proper ordering, we therefore turn off zero copy if
 928          * the client wants us to do more in this compound:
 929          */
 930         if (!nfsd4_last_compound_op(rqstp))
 931                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 932
 933         /* check stateid */
 934         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 935                                         &read->rd_stateid, RD_STATE,
 936                                         &read->rd_nf, NULL);
 937         if (status) {
 938                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
 939                 goto out;
 940         }
 941         status = nfs_ok;
 942 out:
 943         read->rd_rqstp = rqstp;
 944         read->rd_fhp = &cstate->current_fh;
 945         return status;
 946 }
 947
 948
 949 static void
 950 nfsd4_read_release(union nfsd4_op_u *u)
 951 {
 952         if (u->read.rd_nf)
 953                 nfsd_file_put(u->read.rd_nf);
 954         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 955                              u->read.rd_offset, u->read.rd_length);
 956 }
 957
 958 static __be32
 959 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 960               union nfsd4_op_u *u)
 961 {
 962         struct nfsd4_readdir *readdir = &u->readdir;
 963         u64 cookie = readdir->rd_cookie;
 964         static const nfs4_verifier zeroverf;
 965
 966         /* no need to check permission - this will be done in nfsd_readdir() */
 967
 968         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 969                 return nfserr_inval;
 970
 971         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 972         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 973         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 974
 975         if ((cookie == 1) || (cookie == 2) ||
 976             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
 977                 return nfserr_bad_cookie;
 978
 979         readdir->rd_rqstp = rqstp;
 980         readdir->rd_fhp = &cstate->current_fh;
 981         return nfs_ok;
 982 }
 983
 984 static __be32
 985 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 986                union nfsd4_op_u *u)
 987 {
 988         u->readlink.rl_rqstp = rqstp;
 989         u->readlink.rl_fhp = &cstate->current_fh;
 990         return nfs_ok;
 991 }
 992
 993 static __be32
 994 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 995              union nfsd4_op_u *u)
 996 {
 997         struct nfsd4_remove *remove = &u->remove;
 998         __be32 status;
 999
1000         if (opens_in_grace(SVC_NET(rqstp)))
1001                 return nfserr_grace;
1002         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1003                              remove->rm_name, remove->rm_namelen);
1004         if (!status)
1005                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1006         return status;
1007 }
1008
1009 static __be32
1010 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1011              union nfsd4_op_u *u)
1012 {
1013         struct nfsd4_rename *rename = &u->rename;
1014         __be32 status;
1015
1016         if (opens_in_grace(SVC_NET(rqstp)))
1017                 return nfserr_grace;
1018         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1019                              rename->rn_snamelen, &cstate->current_fh,
1020                              rename->rn_tname, rename->rn_tnamelen);
1021         if (status)
1022                 return status;
1023         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
1024         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
1025         return nfs_ok;
1026 }
1027
1028 static __be32
1029 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1030               union nfsd4_op_u *u)
1031 {
1032         struct nfsd4_secinfo *secinfo = &u->secinfo;
1033         struct svc_export *exp;
1034         struct dentry *dentry;
1035         __be32 err;
1036
1037         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1038         if (err)
1039                 return err;
1040         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1041                                     secinfo->si_name, secinfo->si_namelen,
1042                                     &exp, &dentry);
1043         if (err)
1044                 return err;
1045         if (d_really_is_negative(dentry)) {
1046                 exp_put(exp);
1047                 err = nfserr_noent;
1048         } else
1049                 secinfo->si_exp = exp;
1050         dput(dentry);
1051         if (cstate->minorversion)
1052                 /* See rfc 5661 section 2.6.3.1.1.8 */
1053                 fh_put(&cstate->current_fh);
1054         return err;
1055 }
1056
1057 static __be32
1058 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1059                 union nfsd4_op_u *u)
1060 {
1061         __be32 err;
1062
1063         switch (u->secinfo_no_name.sin_style) {
1064         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1065                 break;
1066         case NFS4_SECINFO_STYLE4_PARENT:
1067                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1068                 if (err)
1069                         return err;
1070                 break;
1071         default:
1072                 return nfserr_inval;
1073         }
1074
1075         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1076         fh_put(&cstate->current_fh);
1077         return nfs_ok;
1078 }
1079
1080 static void
1081 nfsd4_secinfo_release(union nfsd4_op_u *u)
1082 {
1083         if (u->secinfo.si_exp)
1084                 exp_put(u->secinfo.si_exp);
1085 }
1086
1087 static void
1088 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1089 {
1090         if (u->secinfo_no_name.sin_exp)
1091                 exp_put(u->secinfo_no_name.sin_exp);
1092 }
1093
1094 static __be32
1095 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1096               union nfsd4_op_u *u)
1097 {
1098         struct nfsd4_setattr *setattr = &u->setattr;
1099         struct nfsd_attrs attrs = {
1100                 .na_iattr       = &setattr->sa_iattr,
1101                 .na_seclabel    = &setattr->sa_label,
1102         };
1103         struct inode *inode;
1104         __be32 status = nfs_ok;
1105         int err;
1106
1107         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1108                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1109                                 &cstate->current_fh, &setattr->sa_stateid,
1110                                 WR_STATE, NULL, NULL);
1111                 if (status) {
1112                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
1113                         return status;
1114                 }
1115         }
1116         err = fh_want_write(&cstate->current_fh);
1117         if (err)
1118                 return nfserrno(err);
1119         status = nfs_ok;
1120
1121         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1122                                     nfsd_attrmask);
1123         if (status)
1124                 goto out;
1125
1126         inode = cstate->current_fh.fh_dentry->d_inode;
1127         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1128                                    setattr->sa_acl, &attrs);
1129
1130         if (status)
1131                 goto out;
1132         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1133                                 0, (time64_t)0);
1134         if (!status)
1135                 status = nfserrno(attrs.na_labelerr);
1136         if (!status)
1137                 status = nfserrno(attrs.na_aclerr);
1138 out:
1139         nfsd_attrs_free(&attrs);
1140         fh_drop_write(&cstate->current_fh);
1141         return status;
1142 }
1143
1144 static __be32
1145 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1146             union nfsd4_op_u *u)
1147 {
1148         struct nfsd4_write *write = &u->write;
1149         stateid_t *stateid = &write->wr_stateid;
1150         struct nfsd_file *nf = NULL;
1151         __be32 status = nfs_ok;
1152         unsigned long cnt;
1153         int nvecs;
1154
1155         if (write->wr_offset > (u64)OFFSET_MAX ||
1156             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1157                 return nfserr_fbig;
1158
1159         cnt = write->wr_buflen;
1160         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1161                                write->wr_offset, cnt);
1162         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1163                                                 stateid, WR_STATE, &nf, NULL);
1164         if (status) {
1165                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1166                 return status;
1167         }
1168
1169         write->wr_how_written = write->wr_stable_how;
1170
1171         nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1172         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1173
1174         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1175                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1176                                 write->wr_how_written,
1177                                 (__be32 *)write->wr_verifier.data);
1178         nfsd_file_put(nf);
1179
1180         write->wr_bytes_written = cnt;
1181         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1182                               write->wr_offset, cnt);
1183         return status;
1184 }
1185
1186 static __be32
1187 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1188                   stateid_t *src_stateid, struct nfsd_file **src,
1189                   stateid_t *dst_stateid, struct nfsd_file **dst)
1190 {
1191         __be32 status;
1192
1193         if (!cstate->save_fh.fh_dentry)
1194                 return nfserr_nofilehandle;
1195
1196         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1197                                             src_stateid, RD_STATE, src, NULL);
1198         if (status) {
1199                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1200                 goto out;
1201         }
1202
1203         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1204                                             dst_stateid, WR_STATE, dst, NULL);
1205         if (status) {
1206                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1207                 goto out_put_src;
1208         }
1209
1210         /* fix up for NFS-specific error code */
1211         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1212             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1213                 status = nfserr_wrong_type;
1214                 goto out_put_dst;
1215         }
1216
1217 out:
1218         return status;
1219 out_put_dst:
1220         nfsd_file_put(*dst);
1221 out_put_src:
1222         nfsd_file_put(*src);
1223         goto out;
1224 }
1225
1226 static __be32
1227 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1228                 union nfsd4_op_u *u)
1229 {
1230         struct nfsd4_clone *clone = &u->clone;
1231         struct nfsd_file *src, *dst;
1232         __be32 status;
1233
1234         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1235                                    &clone->cl_dst_stateid, &dst);
1236         if (status)
1237                 goto out;
1238
1239         status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1240                         dst, clone->cl_dst_pos, clone->cl_count,
1241                         EX_ISSYNC(cstate->current_fh.fh_export));
1242
1243         nfsd_file_put(dst);
1244         nfsd_file_put(src);
1245 out:
1246         return status;
1247 }
1248
1249 static void nfs4_put_copy(struct nfsd4_copy *copy)
1250 {
1251         if (!refcount_dec_and_test(&copy->refcount))
1252                 return;
1253         kfree(copy->cp_src);
1254         kfree(copy);
1255 }
1256
1257 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1258 {
1259         if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1260                 kthread_stop(copy->copy_task);
1261         nfs4_put_copy(copy);
1262 }
1263
1264 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1265 {
1266         struct nfsd4_copy *copy = NULL;
1267
1268         spin_lock(&clp->async_lock);
1269         if (!list_empty(&clp->async_copies)) {
1270                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1271                                         copies);
1272                 refcount_inc(&copy->refcount);
1273         }
1274         spin_unlock(&clp->async_lock);
1275         return copy;
1276 }
1277
1278 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1279 {
1280         struct nfsd4_copy *copy;
1281
1282         while ((copy = nfsd4_get_copy(clp)) != NULL)
1283                 nfsd4_stop_copy(copy);
1284 }
1285 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1286
1287 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1288                                    struct nfs_fh *src_fh,
1289                                    nfs4_stateid *stateid);
1290 extern void nfs42_ssc_close(struct file *filep);
1291
1292 extern void nfs_sb_deactive(struct super_block *sb);
1293
1294 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1295
1296 /*
1297  * setup a work entry in the ssc delayed unmount list.
1298  */
1299 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1300                 struct nfsd4_ssc_umount_item **retwork, struct vfsmount **ss_mnt)
1301 {
1302         struct nfsd4_ssc_umount_item *ni = NULL;
1303         struct nfsd4_ssc_umount_item *work = NULL;
1304         struct nfsd4_ssc_umount_item *tmp;
1305         DEFINE_WAIT(wait);
1306
1307         *ss_mnt = NULL;
1308         *retwork = NULL;
1309         work = kzalloc(sizeof(*work), GFP_KERNEL);
1310 try_again:
1311         spin_lock(&nn->nfsd_ssc_lock);
1312         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1313                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1314                         continue;
1315                 /* found a match */
1316                 if (ni->nsui_busy) {
1317                         /*  wait - and try again */
1318                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait,
1319                                 TASK_INTERRUPTIBLE);
1320                         spin_unlock(&nn->nfsd_ssc_lock);
1321
1322                         /* allow 20secs for mount/unmount for now - revisit */
1323                         if (signal_pending(current) ||
1324                                         (schedule_timeout(20*HZ) == 0)) {
1325                                 kfree(work);
1326                                 return nfserr_eagain;
1327                         }
1328                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1329                         goto try_again;
1330                 }
1331                 *ss_mnt = ni->nsui_vfsmount;
1332                 refcount_inc(&ni->nsui_refcnt);
1333                 spin_unlock(&nn->nfsd_ssc_lock);
1334                 kfree(work);
1335
1336                 /* return vfsmount in ss_mnt */
1337                 return 0;
1338         }
1339         if (work) {
1340                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1341                 refcount_set(&work->nsui_refcnt, 2);
1342                 work->nsui_busy = true;
1343                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1344                 *retwork = work;
1345         }
1346         spin_unlock(&nn->nfsd_ssc_lock);
1347         return 0;
1348 }
1349
1350 static void nfsd4_ssc_update_dul_work(struct nfsd_net *nn,
1351                 struct nfsd4_ssc_umount_item *work, struct vfsmount *ss_mnt)
1352 {
1353         /* set nsui_vfsmount, clear busy flag and wakeup waiters */
1354         spin_lock(&nn->nfsd_ssc_lock);
1355         work->nsui_vfsmount = ss_mnt;
1356         work->nsui_busy = false;
1357         wake_up_all(&nn->nfsd_ssc_waitq);
1358         spin_unlock(&nn->nfsd_ssc_lock);
1359 }
1360
1361 static void nfsd4_ssc_cancel_dul_work(struct nfsd_net *nn,
1362                 struct nfsd4_ssc_umount_item *work)
1363 {
1364         spin_lock(&nn->nfsd_ssc_lock);
1365         list_del(&work->nsui_list);
1366         wake_up_all(&nn->nfsd_ssc_waitq);
1367         spin_unlock(&nn->nfsd_ssc_lock);
1368         kfree(work);
1369 }
1370
1371 /*
1372  * Support one copy source server for now.
1373  */
1374 static __be32
1375 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1376                        struct vfsmount **mount)
1377 {
1378         struct file_system_type *type;
1379         struct vfsmount *ss_mnt;
1380         struct nfs42_netaddr *naddr;
1381         struct sockaddr_storage tmp_addr;
1382         size_t tmp_addrlen, match_netid_len = 3;
1383         char *startsep = "", *endsep = "", *match_netid = "tcp";
1384         char *ipaddr, *dev_name, *raw_data;
1385         int len, raw_len;
1386         __be32 status = nfserr_inval;
1387         struct nfsd4_ssc_umount_item *work = NULL;
1388         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1389
1390         naddr = &nss->u.nl4_addr;
1391         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1392                                          naddr->addr_len,
1393                                          (struct sockaddr *)&tmp_addr,
1394                                          sizeof(tmp_addr));
1395         if (tmp_addrlen == 0)
1396                 goto out_err;
1397
1398         if (tmp_addr.ss_family == AF_INET6) {
1399                 startsep = "[";
1400                 endsep = "]";
1401                 match_netid = "tcp6";
1402                 match_netid_len = 4;
1403         }
1404
1405         if (naddr->netid_len != match_netid_len ||
1406                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1407                 goto out_err;
1408
1409         /* Construct the raw data for the vfs_kern_mount call */
1410         len = RPC_MAX_ADDRBUFLEN + 1;
1411         ipaddr = kzalloc(len, GFP_KERNEL);
1412         if (!ipaddr)
1413                 goto out_err;
1414
1415         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1416
1417         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1418
1419         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1420         raw_data = kzalloc(raw_len, GFP_KERNEL);
1421         if (!raw_data)
1422                 goto out_free_ipaddr;
1423
1424         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1425
1426         status = nfserr_nodev;
1427         type = get_fs_type("nfs");
1428         if (!type)
1429                 goto out_free_rawdata;
1430
1431         /* Set the server:<export> for the vfs_kern_mount call */
1432         dev_name = kzalloc(len + 5, GFP_KERNEL);
1433         if (!dev_name)
1434                 goto out_free_rawdata;
1435         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1436
1437         status = nfsd4_ssc_setup_dul(nn, ipaddr, &work, &ss_mnt);
1438         if (status)
1439                 goto out_free_devname;
1440         if (ss_mnt)
1441                 goto out_done;
1442
1443         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1444         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1445         module_put(type->owner);
1446         if (IS_ERR(ss_mnt)) {
1447                 status = nfserr_nodev;
1448                 if (work)
1449                         nfsd4_ssc_cancel_dul_work(nn, work);
1450                 goto out_free_devname;
1451         }
1452         if (work)
1453                 nfsd4_ssc_update_dul_work(nn, work, ss_mnt);
1454 out_done:
1455         status = 0;
1456         *mount = ss_mnt;
1457
1458 out_free_devname:
1459         kfree(dev_name);
1460 out_free_rawdata:
1461         kfree(raw_data);
1462 out_free_ipaddr:
1463         kfree(ipaddr);
1464 out_err:
1465         return status;
1466 }
1467
1468 /*
1469  * Verify COPY destination stateid.
1470  *
1471  * Connect to the source server with NFSv4.1.
1472  * Create the source struct file for nfsd_copy_range.
1473  * Called with COPY cstate:
1474  *    SAVED_FH: source filehandle
1475  *    CURRENT_FH: destination filehandle
1476  */
1477 static __be32
1478 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1479                       struct nfsd4_compound_state *cstate,
1480                       struct nfsd4_copy *copy, struct vfsmount **mount)
1481 {
1482         struct svc_fh *s_fh = NULL;
1483         stateid_t *s_stid = &copy->cp_src_stateid;
1484         __be32 status = nfserr_inval;
1485
1486         /* Verify the destination stateid and set dst struct file*/
1487         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1488                                             &copy->cp_dst_stateid,
1489                                             WR_STATE, &copy->nf_dst, NULL);
1490         if (status)
1491                 goto out;
1492
1493         status = nfsd4_interssc_connect(copy->cp_src, rqstp, mount);
1494         if (status)
1495                 goto out;
1496
1497         s_fh = &cstate->save_fh;
1498
1499         copy->c_fh.size = s_fh->fh_handle.fh_size;
1500         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1501         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1502         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1503                sizeof(stateid_opaque_t));
1504
1505         status = 0;
1506 out:
1507         return status;
1508 }
1509
1510 static void
1511 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct file *filp,
1512                         struct nfsd_file *dst)
1513 {
1514         bool found = false;
1515         long timeout;
1516         struct nfsd4_ssc_umount_item *tmp;
1517         struct nfsd4_ssc_umount_item *ni = NULL;
1518         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1519
1520         nfs42_ssc_close(filp);
1521         nfsd_file_put(dst);
1522         fput(filp);
1523
1524         if (!nn) {
1525                 mntput(ss_mnt);
1526                 return;
1527         }
1528         spin_lock(&nn->nfsd_ssc_lock);
1529         timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1530         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1531                 if (ni->nsui_vfsmount->mnt_sb == ss_mnt->mnt_sb) {
1532                         list_del(&ni->nsui_list);
1533                         /*
1534                          * vfsmount can be shared by multiple exports,
1535                          * decrement refcnt. If the count drops to 1 it
1536                          * will be unmounted when nsui_expire expires.
1537                          */
1538                         refcount_dec(&ni->nsui_refcnt);
1539                         ni->nsui_expire = jiffies + timeout;
1540                         list_add_tail(&ni->nsui_list, &nn->nfsd_ssc_mount_list);
1541                         found = true;
1542                         break;
1543                 }
1544         }
1545         spin_unlock(&nn->nfsd_ssc_lock);
1546         if (!found) {
1547                 mntput(ss_mnt);
1548                 return;
1549         }
1550 }
1551
1552 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1553
1554 static __be32
1555 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1556                       struct nfsd4_compound_state *cstate,
1557                       struct nfsd4_copy *copy,
1558                       struct vfsmount **mount)
1559 {
1560         *mount = NULL;
1561         return nfserr_inval;
1562 }
1563
1564 static void
1565 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct file *filp,
1566                         struct nfsd_file *dst)
1567 {
1568 }
1569
1570 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1571                                    struct nfs_fh *src_fh,
1572                                    nfs4_stateid *stateid)
1573 {
1574         return NULL;
1575 }
1576 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1577
1578 static __be32
1579 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1580                       struct nfsd4_compound_state *cstate,
1581                       struct nfsd4_copy *copy)
1582 {
1583         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1584                                  &copy->nf_src, &copy->cp_dst_stateid,
1585                                  &copy->nf_dst);
1586 }
1587
1588 static void
1589 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1590 {
1591         nfsd_file_put(src);
1592         nfsd_file_put(dst);
1593 }
1594
1595 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1596 {
1597         struct nfsd4_cb_offload *cbo =
1598                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1599
1600         kfree(cbo);
1601 }
1602
1603 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1604                                  struct rpc_task *task)
1605 {
1606         struct nfsd4_cb_offload *cbo =
1607                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1608
1609         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1610         return 1;
1611 }
1612
1613 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1614         .release = nfsd4_cb_offload_release,
1615         .done = nfsd4_cb_offload_done
1616 };
1617
1618 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1619 {
1620         copy->cp_res.wr_stable_how =
1621                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1622                         NFS_FILE_SYNC : NFS_UNSTABLE;
1623         nfsd4_copy_set_sync(copy, sync);
1624         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1625 }
1626
1627 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1628                                      struct file *dst,
1629                                      struct file *src)
1630 {
1631         errseq_t since;
1632         ssize_t bytes_copied = 0;
1633         u64 bytes_total = copy->cp_count;
1634         u64 src_pos = copy->cp_src_pos;
1635         u64 dst_pos = copy->cp_dst_pos;
1636         int status;
1637         loff_t end;
1638
1639         /* See RFC 7862 p.67: */
1640         if (bytes_total == 0)
1641                 bytes_total = ULLONG_MAX;
1642         do {
1643                 if (kthread_should_stop())
1644                         break;
1645                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1646                                                     bytes_total);
1647                 if (bytes_copied <= 0)
1648                         break;
1649                 bytes_total -= bytes_copied;
1650                 copy->cp_res.wr_bytes_written += bytes_copied;
1651                 src_pos += bytes_copied;
1652                 dst_pos += bytes_copied;
1653         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1654         /* for a non-zero asynchronous copy do a commit of data */
1655         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1656                 since = READ_ONCE(dst->f_wb_err);
1657                 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1658                 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1659                 if (!status)
1660                         status = filemap_check_wb_err(dst->f_mapping, since);
1661                 if (!status)
1662                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1663         }
1664         return bytes_copied;
1665 }
1666
1667 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1668                             struct file *src, struct file *dst,
1669                             bool sync)
1670 {
1671         __be32 status;
1672         ssize_t bytes;
1673
1674         bytes = _nfsd_copy_file_range(copy, dst, src);
1675
1676         /* for async copy, we ignore the error, client can always retry
1677          * to get the error
1678          */
1679         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1680                 status = nfserrno(bytes);
1681         else {
1682                 nfsd4_init_copy_res(copy, sync);
1683                 status = nfs_ok;
1684         }
1685         return status;
1686 }
1687
1688 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1689 {
1690         dst->cp_src_pos = src->cp_src_pos;
1691         dst->cp_dst_pos = src->cp_dst_pos;
1692         dst->cp_count = src->cp_count;
1693         dst->cp_flags = src->cp_flags;
1694         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1695         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1696         dst->cp_clp = src->cp_clp;
1697         dst->nf_dst = nfsd_file_get(src->nf_dst);
1698         /* for inter, nf_src doesn't exist yet */
1699         if (!nfsd4_ssc_is_inter(src))
1700                 dst->nf_src = nfsd_file_get(src->nf_src);
1701
1702         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1703         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1704         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1705         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1706         dst->ss_mnt = src->ss_mnt;
1707 }
1708
1709 static void cleanup_async_copy(struct nfsd4_copy *copy)
1710 {
1711         nfs4_free_copy_state(copy);
1712         nfsd_file_put(copy->nf_dst);
1713         if (!nfsd4_ssc_is_inter(copy))
1714                 nfsd_file_put(copy->nf_src);
1715         spin_lock(&copy->cp_clp->async_lock);
1716         list_del(&copy->copies);
1717         spin_unlock(&copy->cp_clp->async_lock);
1718         nfs4_put_copy(copy);
1719 }
1720
1721 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1722 {
1723         struct nfsd4_cb_offload *cbo;
1724
1725         cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1726         if (!cbo)
1727                 return;
1728
1729         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1730         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1731         cbo->co_nfserr = nfserr;
1732
1733         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1734                       NFSPROC4_CLNT_CB_OFFLOAD);
1735         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1736                               &cbo->co_fh, copy->cp_count, nfserr);
1737         nfsd4_run_cb(&cbo->co_cb);
1738 }
1739
1740 /**
1741  * nfsd4_do_async_copy - kthread function for background server-side COPY
1742  * @data: arguments for COPY operation
1743  *
1744  * Return values:
1745  *   %0: Copy operation is done.
1746  */
1747 static int nfsd4_do_async_copy(void *data)
1748 {
1749         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1750         __be32 nfserr;
1751
1752         if (nfsd4_ssc_is_inter(copy)) {
1753                 struct file *filp;
1754
1755                 filp = nfs42_ssc_open(copy->ss_mnt, &copy->c_fh,
1756                                       &copy->stateid);
1757                 if (IS_ERR(filp)) {
1758                         switch (PTR_ERR(filp)) {
1759                         case -EBADF:
1760                                 nfserr = nfserr_wrong_type;
1761                                 break;
1762                         default:
1763                                 nfserr = nfserr_offload_denied;
1764                         }
1765                         /* ss_mnt will be unmounted by the laundromat */
1766                         goto do_callback;
1767                 }
1768                 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1769                                        false);
1770                 nfsd4_cleanup_inter_ssc(copy->ss_mnt, filp, copy->nf_dst);
1771         } else {
1772                 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1773                                        copy->nf_dst->nf_file, false);
1774                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1775         }
1776
1777 do_callback:
1778         nfsd4_send_cb_offload(copy, nfserr);
1779         cleanup_async_copy(copy);
1780         return 0;
1781 }
1782
1783 static __be32
1784 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1785                 union nfsd4_op_u *u)
1786 {
1787         struct nfsd4_copy *copy = &u->copy;
1788         __be32 status;
1789         struct nfsd4_copy *async_copy = NULL;
1790
1791         if (nfsd4_ssc_is_inter(copy)) {
1792                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1793                         status = nfserr_notsupp;
1794                         goto out;
1795                 }
1796                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1797                                 &copy->ss_mnt);
1798                 if (status)
1799                         return nfserr_offload_denied;
1800         } else {
1801                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1802                 if (status)
1803                         return status;
1804         }
1805
1806         copy->cp_clp = cstate->clp;
1807         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1808                 sizeof(struct knfsd_fh));
1809         if (nfsd4_copy_is_async(copy)) {
1810                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1811
1812                 status = nfserrno(-ENOMEM);
1813                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1814                 if (!async_copy)
1815                         goto out_err;
1816                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1817                 if (!async_copy->cp_src)
1818                         goto out_err;
1819                 if (!nfs4_init_copy_state(nn, copy))
1820                         goto out_err;
1821                 refcount_set(&async_copy->refcount, 1);
1822                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1823                         sizeof(copy->cp_res.cb_stateid));
1824                 dup_copy_fields(copy, async_copy);
1825                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1826                                 async_copy, "%s", "copy thread");
1827                 if (IS_ERR(async_copy->copy_task))
1828                         goto out_err;
1829                 spin_lock(&async_copy->cp_clp->async_lock);
1830                 list_add(&async_copy->copies,
1831                                 &async_copy->cp_clp->async_copies);
1832                 spin_unlock(&async_copy->cp_clp->async_lock);
1833                 wake_up_process(async_copy->copy_task);
1834                 status = nfs_ok;
1835         } else {
1836                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1837                                        copy->nf_dst->nf_file, true);
1838                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1839         }
1840 out:
1841         return status;
1842 out_err:
1843         if (async_copy)
1844                 cleanup_async_copy(async_copy);
1845         status = nfserrno(-ENOMEM);
1846         /*
1847          * source's vfsmount of inter-copy will be unmounted
1848          * by the laundromat
1849          */
1850         goto out;
1851 }
1852
1853 struct nfsd4_copy *
1854 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1855 {
1856         struct nfsd4_copy *copy;
1857
1858         spin_lock(&clp->async_lock);
1859         list_for_each_entry(copy, &clp->async_copies, copies) {
1860                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1861                         continue;
1862                 refcount_inc(&copy->refcount);
1863                 spin_unlock(&clp->async_lock);
1864                 return copy;
1865         }
1866         spin_unlock(&clp->async_lock);
1867         return NULL;
1868 }
1869
1870 static __be32
1871 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1872                      struct nfsd4_compound_state *cstate,
1873                      union nfsd4_op_u *u)
1874 {
1875         struct nfsd4_offload_status *os = &u->offload_status;
1876         struct nfsd4_copy *copy;
1877         struct nfs4_client *clp = cstate->clp;
1878
1879         copy = find_async_copy(clp, &os->stateid);
1880         if (!copy) {
1881                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1882
1883                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1884         } else
1885                 nfsd4_stop_copy(copy);
1886
1887         return nfs_ok;
1888 }
1889
1890 static __be32
1891 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1892                   union nfsd4_op_u *u)
1893 {
1894         struct nfsd4_copy_notify *cn = &u->copy_notify;
1895         __be32 status;
1896         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1897         struct nfs4_stid *stid;
1898         struct nfs4_cpntf_state *cps;
1899         struct nfs4_client *clp = cstate->clp;
1900
1901         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1902                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1903                                         &stid);
1904         if (status)
1905                 return status;
1906
1907         cn->cpn_sec = nn->nfsd4_lease;
1908         cn->cpn_nsec = 0;
1909
1910         status = nfserrno(-ENOMEM);
1911         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1912         if (!cps)
1913                 goto out;
1914         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1915         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1916         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1917
1918         /* For now, only return one server address in cpn_src, the
1919          * address used by the client to connect to this server.
1920          */
1921         cn->cpn_src->nl4_type = NL4_NETADDR;
1922         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1923                                  &cn->cpn_src->u.nl4_addr);
1924         WARN_ON_ONCE(status);
1925         if (status) {
1926                 nfs4_put_cpntf_state(nn, cps);
1927                 goto out;
1928         }
1929 out:
1930         nfs4_put_stid(stid);
1931         return status;
1932 }
1933
1934 static __be32
1935 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1936                 struct nfsd4_fallocate *fallocate, int flags)
1937 {
1938         __be32 status;
1939         struct nfsd_file *nf;
1940
1941         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1942                                             &fallocate->falloc_stateid,
1943                                             WR_STATE, &nf, NULL);
1944         if (status != nfs_ok) {
1945                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1946                 return status;
1947         }
1948
1949         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1950                                      fallocate->falloc_offset,
1951                                      fallocate->falloc_length,
1952                                      flags);
1953         nfsd_file_put(nf);
1954         return status;
1955 }
1956 static __be32
1957 nfsd4_offload_status(struct svc_rqst *rqstp,
1958                      struct nfsd4_compound_state *cstate,
1959                      union nfsd4_op_u *u)
1960 {
1961         struct nfsd4_offload_status *os = &u->offload_status;
1962         __be32 status = 0;
1963         struct nfsd4_copy *copy;
1964         struct nfs4_client *clp = cstate->clp;
1965
1966         copy = find_async_copy(clp, &os->stateid);
1967         if (copy) {
1968                 os->count = copy->cp_res.wr_bytes_written;
1969                 nfs4_put_copy(copy);
1970         } else
1971                 status = nfserr_bad_stateid;
1972
1973         return status;
1974 }
1975
1976 static __be32
1977 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1978                union nfsd4_op_u *u)
1979 {
1980         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1981 }
1982
1983 static __be32
1984 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1985                  union nfsd4_op_u *u)
1986 {
1987         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1988                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1989 }
1990
1991 static __be32
1992 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1993            union nfsd4_op_u *u)
1994 {
1995         struct nfsd4_seek *seek = &u->seek;
1996         int whence;
1997         __be32 status;
1998         struct nfsd_file *nf;
1999
2000         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2001                                             &seek->seek_stateid,
2002                                             RD_STATE, &nf, NULL);
2003         if (status) {
2004                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
2005                 return status;
2006         }
2007
2008         switch (seek->seek_whence) {
2009         case NFS4_CONTENT_DATA:
2010                 whence = SEEK_DATA;
2011                 break;
2012         case NFS4_CONTENT_HOLE:
2013                 whence = SEEK_HOLE;
2014                 break;
2015         default:
2016                 status = nfserr_union_notsupp;
2017                 goto out;
2018         }
2019
2020         /*
2021          * Note:  This call does change file->f_pos, but nothing in NFSD
2022          *        should ever file->f_pos.
2023          */
2024         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2025         if (seek->seek_pos < 0)
2026                 status = nfserrno(seek->seek_pos);
2027         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2028                 seek->seek_eof = true;
2029
2030 out:
2031         nfsd_file_put(nf);
2032         return status;
2033 }
2034
2035 /* This routine never returns NFS_OK!  If there are no other errors, it
2036  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2037  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2038  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2039  */
2040 static __be32
2041 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2042              struct nfsd4_verify *verify)
2043 {
2044         __be32 *buf, *p;
2045         int count;
2046         __be32 status;
2047
2048         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2049         if (status)
2050                 return status;
2051
2052         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2053         if (status)
2054                 return status;
2055
2056         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2057             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2058                 return nfserr_inval;
2059         if (verify->ve_attrlen & 3)
2060                 return nfserr_inval;
2061
2062         /* count in words:
2063          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2064          */
2065         count = 4 + (verify->ve_attrlen >> 2);
2066         buf = kmalloc(count << 2, GFP_KERNEL);
2067         if (!buf)
2068                 return nfserr_jukebox;
2069
2070         p = buf;
2071         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2072                                     cstate->current_fh.fh_export,
2073                                     cstate->current_fh.fh_dentry,
2074                                     verify->ve_bmval,
2075                                     rqstp, 0);
2076         /*
2077          * If nfsd4_encode_fattr() ran out of space, assume that's because
2078          * the attributes are longer (hence different) than those given:
2079          */
2080         if (status == nfserr_resource)
2081                 status = nfserr_not_same;
2082         if (status)
2083                 goto out_kfree;
2084
2085         /* skip bitmap */
2086         p = buf + 1 + ntohl(buf[0]);
2087         status = nfserr_not_same;
2088         if (ntohl(*p++) != verify->ve_attrlen)
2089                 goto out_kfree;
2090         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2091                 status = nfserr_same;
2092
2093 out_kfree:
2094         kfree(buf);
2095         return status;
2096 }
2097
2098 static __be32
2099 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2100               union nfsd4_op_u *u)
2101 {
2102         __be32 status;
2103
2104         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2105         return status == nfserr_not_same ? nfs_ok : status;
2106 }
2107
2108 static __be32
2109 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2110              union nfsd4_op_u *u)
2111 {
2112         __be32 status;
2113
2114         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2115         return status == nfserr_same ? nfs_ok : status;
2116 }
2117
2118 #ifdef CONFIG_NFSD_PNFS
2119 static const struct nfsd4_layout_ops *
2120 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2121 {
2122         if (!exp->ex_layout_types) {
2123                 dprintk("%s: export does not support pNFS\n", __func__);
2124                 return NULL;
2125         }
2126
2127         if (layout_type >= LAYOUT_TYPE_MAX ||
2128             !(exp->ex_layout_types & (1 << layout_type))) {
2129                 dprintk("%s: layout type %d not supported\n",
2130                         __func__, layout_type);
2131                 return NULL;
2132         }
2133
2134         return nfsd4_layout_ops[layout_type];
2135 }
2136
2137 static __be32
2138 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2139                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2140 {
2141         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2142         const struct nfsd4_layout_ops *ops;
2143         struct nfsd4_deviceid_map *map;
2144         struct svc_export *exp;
2145         __be32 nfserr;
2146
2147         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2148                __func__,
2149                gdp->gd_layout_type,
2150                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2151                gdp->gd_maxcount);
2152
2153         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2154         if (!map) {
2155                 dprintk("%s: couldn't find device ID to export mapping!\n",
2156                         __func__);
2157                 return nfserr_noent;
2158         }
2159
2160         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2161         if (IS_ERR(exp)) {
2162                 dprintk("%s: could not find device id\n", __func__);
2163                 return nfserr_noent;
2164         }
2165
2166         nfserr = nfserr_layoutunavailable;
2167         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2168         if (!ops)
2169                 goto out;
2170
2171         nfserr = nfs_ok;
2172         if (gdp->gd_maxcount != 0) {
2173                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2174                                 rqstp, cstate->clp, gdp);
2175         }
2176
2177         gdp->gd_notify_types &= ops->notify_types;
2178 out:
2179         exp_put(exp);
2180         return nfserr;
2181 }
2182
2183 static void
2184 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2185 {
2186         kfree(u->getdeviceinfo.gd_device);
2187 }
2188
2189 static __be32
2190 nfsd4_layoutget(struct svc_rqst *rqstp,
2191                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2192 {
2193         struct nfsd4_layoutget *lgp = &u->layoutget;
2194         struct svc_fh *current_fh = &cstate->current_fh;
2195         const struct nfsd4_layout_ops *ops;
2196         struct nfs4_layout_stateid *ls;
2197         __be32 nfserr;
2198         int accmode = NFSD_MAY_READ_IF_EXEC;
2199
2200         switch (lgp->lg_seg.iomode) {
2201         case IOMODE_READ:
2202                 accmode |= NFSD_MAY_READ;
2203                 break;
2204         case IOMODE_RW:
2205                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2206                 break;
2207         default:
2208                 dprintk("%s: invalid iomode %d\n",
2209                         __func__, lgp->lg_seg.iomode);
2210                 nfserr = nfserr_badiomode;
2211                 goto out;
2212         }
2213
2214         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2215         if (nfserr)
2216                 goto out;
2217
2218         nfserr = nfserr_layoutunavailable;
2219         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2220         if (!ops)
2221                 goto out;
2222
2223         /*
2224          * Verify minlength and range as per RFC5661:
2225          *  o  If loga_length is less than loga_minlength,
2226          *     the metadata server MUST return NFS4ERR_INVAL.
2227          *  o  If the sum of loga_offset and loga_minlength exceeds
2228          *     NFS4_UINT64_MAX, and loga_minlength is not
2229          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2230          *  o  If the sum of loga_offset and loga_length exceeds
2231          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2232          *     the error NFS4ERR_INVAL MUST result.
2233          */
2234         nfserr = nfserr_inval;
2235         if (lgp->lg_seg.length < lgp->lg_minlength ||
2236             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2237              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2238             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2239              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2240                 goto out;
2241         if (lgp->lg_seg.length == 0)
2242                 goto out;
2243
2244         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2245                                                 true, lgp->lg_layout_type, &ls);
2246         if (nfserr) {
2247                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2248                 goto out;
2249         }
2250
2251         nfserr = nfserr_recallconflict;
2252         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2253                 goto out_put_stid;
2254
2255         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2256                                      current_fh, lgp);
2257         if (nfserr)
2258                 goto out_put_stid;
2259
2260         nfserr = nfsd4_insert_layout(lgp, ls);
2261
2262 out_put_stid:
2263         mutex_unlock(&ls->ls_mutex);
2264         nfs4_put_stid(&ls->ls_stid);
2265 out:
2266         return nfserr;
2267 }
2268
2269 static void
2270 nfsd4_layoutget_release(union nfsd4_op_u *u)
2271 {
2272         kfree(u->layoutget.lg_content);
2273 }
2274
2275 static __be32
2276 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2277                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2278 {
2279         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2280         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2281         struct svc_fh *current_fh = &cstate->current_fh;
2282         const struct nfsd4_layout_ops *ops;
2283         loff_t new_size = lcp->lc_last_wr + 1;
2284         struct inode *inode;
2285         struct nfs4_layout_stateid *ls;
2286         __be32 nfserr;
2287
2288         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2289         if (nfserr)
2290                 goto out;
2291
2292         nfserr = nfserr_layoutunavailable;
2293         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2294         if (!ops)
2295                 goto out;
2296         inode = d_inode(current_fh->fh_dentry);
2297
2298         nfserr = nfserr_inval;
2299         if (new_size <= seg->offset) {
2300                 dprintk("pnfsd: last write before layout segment\n");
2301                 goto out;
2302         }
2303         if (new_size > seg->offset + seg->length) {
2304                 dprintk("pnfsd: last write beyond layout segment\n");
2305                 goto out;
2306         }
2307         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2308                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2309                 goto out;
2310         }
2311
2312         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2313                                                 false, lcp->lc_layout_type,
2314                                                 &ls);
2315         if (nfserr) {
2316                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2317                 /* fixup error code as per RFC5661 */
2318                 if (nfserr == nfserr_bad_stateid)
2319                         nfserr = nfserr_badlayout;
2320                 goto out;
2321         }
2322
2323         /* LAYOUTCOMMIT does not require any serialization */
2324         mutex_unlock(&ls->ls_mutex);
2325
2326         if (new_size > i_size_read(inode)) {
2327                 lcp->lc_size_chg = 1;
2328                 lcp->lc_newsize = new_size;
2329         } else {
2330                 lcp->lc_size_chg = 0;
2331         }
2332
2333         nfserr = ops->proc_layoutcommit(inode, lcp);
2334         nfs4_put_stid(&ls->ls_stid);
2335 out:
2336         return nfserr;
2337 }
2338
2339 static __be32
2340 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2341                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2342 {
2343         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2344         struct svc_fh *current_fh = &cstate->current_fh;
2345         __be32 nfserr;
2346
2347         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2348         if (nfserr)
2349                 goto out;
2350
2351         nfserr = nfserr_layoutunavailable;
2352         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2353                 goto out;
2354
2355         switch (lrp->lr_seg.iomode) {
2356         case IOMODE_READ:
2357         case IOMODE_RW:
2358         case IOMODE_ANY:
2359                 break;
2360         default:
2361                 dprintk("%s: invalid iomode %d\n", __func__,
2362                         lrp->lr_seg.iomode);
2363                 nfserr = nfserr_inval;
2364                 goto out;
2365         }
2366
2367         switch (lrp->lr_return_type) {
2368         case RETURN_FILE:
2369                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2370                 break;
2371         case RETURN_FSID:
2372         case RETURN_ALL:
2373                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2374                 break;
2375         default:
2376                 dprintk("%s: invalid return_type %d\n", __func__,
2377                         lrp->lr_return_type);
2378                 nfserr = nfserr_inval;
2379                 break;
2380         }
2381 out:
2382         return nfserr;
2383 }
2384 #endif /* CONFIG_NFSD_PNFS */
2385
2386 static __be32
2387 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2388                union nfsd4_op_u *u)
2389 {
2390         struct nfsd4_getxattr *getxattr = &u->getxattr;
2391
2392         return nfsd_getxattr(rqstp, &cstate->current_fh,
2393                              getxattr->getxa_name, &getxattr->getxa_buf,
2394                              &getxattr->getxa_len);
2395 }
2396
2397 static __be32
2398 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2399            union nfsd4_op_u *u)
2400 {
2401         struct nfsd4_setxattr *setxattr = &u->setxattr;
2402         __be32 ret;
2403
2404         if (opens_in_grace(SVC_NET(rqstp)))
2405                 return nfserr_grace;
2406
2407         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2408                             setxattr->setxa_buf, setxattr->setxa_len,
2409                             setxattr->setxa_flags);
2410
2411         if (!ret)
2412                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2413
2414         return ret;
2415 }
2416
2417 static __be32
2418 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2419            union nfsd4_op_u *u)
2420 {
2421         /*
2422          * Get the entire list, then copy out only the user attributes
2423          * in the encode function.
2424          */
2425         return nfsd_listxattr(rqstp, &cstate->current_fh,
2426                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2427 }
2428
2429 static __be32
2430 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2431            union nfsd4_op_u *u)
2432 {
2433         struct nfsd4_removexattr *removexattr = &u->removexattr;
2434         __be32 ret;
2435
2436         if (opens_in_grace(SVC_NET(rqstp)))
2437                 return nfserr_grace;
2438
2439         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2440             removexattr->rmxa_name);
2441
2442         if (!ret)
2443                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2444
2445         return ret;
2446 }
2447
2448 /*
2449  * NULL call.
2450  */
2451 static __be32
2452 nfsd4_proc_null(struct svc_rqst *rqstp)
2453 {
2454         return rpc_success;
2455 }
2456
2457 static inline void nfsd4_increment_op_stats(u32 opnum)
2458 {
2459         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2460                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2461 }
2462
2463 static const struct nfsd4_operation nfsd4_ops[];
2464
2465 static const char *nfsd4_op_name(unsigned opnum);
2466
2467 /*
2468  * Enforce NFSv4.1 COMPOUND ordering rules:
2469  *
2470  * Also note, enforced elsewhere:
2471  *      - SEQUENCE other than as first op results in
2472  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2473  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2474  *        (Enforced in nfsd4_bind_conn_to_session().)
2475  *      - DESTROY_SESSION must be the final operation in a compound, if
2476  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2477  *        (Enforced in nfsd4_destroy_session().)
2478  */
2479 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2480 {
2481         struct nfsd4_op *first_op = &args->ops[0];
2482
2483         /* These ordering requirements don't apply to NFSv4.0: */
2484         if (args->minorversion == 0)
2485                 return nfs_ok;
2486         /* This is weird, but OK, not our problem: */
2487         if (args->opcnt == 0)
2488                 return nfs_ok;
2489         if (first_op->status == nfserr_op_illegal)
2490                 return nfs_ok;
2491         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2492                 return nfserr_op_not_in_session;
2493         if (first_op->opnum == OP_SEQUENCE)
2494                 return nfs_ok;
2495         /*
2496          * So first_op is something allowed outside a session, like
2497          * EXCHANGE_ID; but then it has to be the only op in the
2498          * compound:
2499          */
2500         if (args->opcnt != 1)
2501                 return nfserr_not_only_op;
2502         return nfs_ok;
2503 }
2504
2505 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2506 {
2507         return &nfsd4_ops[op->opnum];
2508 }
2509
2510 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2511 {
2512         if (op->opnum == OP_ILLEGAL)
2513                 return false;
2514         return OPDESC(op)->op_flags & OP_CACHEME;
2515 }
2516
2517 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2518 {
2519         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2520         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2521         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2522         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2523         const struct nfsd4_operation *thisd = OPDESC(this);
2524         const struct nfsd4_operation *nextd;
2525
2526         /*
2527          * Most ops check wronsec on our own; only the putfh-like ops
2528          * have special rules.
2529          */
2530         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2531                 return false;
2532         /*
2533          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2534          * put-filehandle operation if we're not going to use the
2535          * result:
2536          */
2537         if (argp->opcnt == resp->opcnt)
2538                 return false;
2539         if (next->opnum == OP_ILLEGAL)
2540                 return false;
2541         nextd = OPDESC(next);
2542         /*
2543          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2544          * errors themselves as necessary; others should check for them
2545          * now:
2546          */
2547         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2548 }
2549
2550 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2551 static void
2552 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2553 {
2554         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2555         struct nfsd4_copy *copy;
2556         struct nfsd4_putfh *putfh;
2557         int i;
2558
2559         /* traverse all operation and if it's a COPY compound, mark the
2560          * source filehandle to skip verification
2561          */
2562         for (i = 0; i < args->opcnt; i++) {
2563                 op = &args->ops[i];
2564                 if (op->opnum == OP_PUTFH)
2565                         current_op = op;
2566                 else if (op->opnum == OP_SAVEFH)
2567                         saved_op = current_op;
2568                 else if (op->opnum == OP_RESTOREFH)
2569                         current_op = saved_op;
2570                 else if (op->opnum == OP_COPY) {
2571                         copy = (struct nfsd4_copy *)&op->u;
2572                         if (!saved_op) {
2573                                 op->status = nfserr_nofilehandle;
2574                                 return;
2575                         }
2576                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2577                         if (nfsd4_ssc_is_inter(copy))
2578                                 putfh->no_verify = true;
2579                 }
2580         }
2581 }
2582 #else
2583 static void
2584 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2585 {
2586 }
2587 #endif
2588
2589 /*
2590  * COMPOUND call.
2591  */
2592 static __be32
2593 nfsd4_proc_compound(struct svc_rqst *rqstp)
2594 {
2595         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2596         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2597         struct nfsd4_op *op;
2598         struct nfsd4_compound_state *cstate = &resp->cstate;
2599         struct svc_fh *current_fh = &cstate->current_fh;
2600         struct svc_fh *save_fh = &cstate->save_fh;
2601         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2602         __be32          status;
2603
2604         resp->xdr = &rqstp->rq_res_stream;
2605         resp->statusp = resp->xdr->p;
2606
2607         /* reserve space for: NFS status code */
2608         xdr_reserve_space(resp->xdr, XDR_UNIT);
2609
2610         /* reserve space for: taglen, tag, and opcnt */
2611         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2612         resp->taglen = args->taglen;
2613         resp->tag = args->tag;
2614         resp->rqstp = rqstp;
2615         cstate->minorversion = args->minorversion;
2616         fh_init(current_fh, NFS4_FHSIZE);
2617         fh_init(save_fh, NFS4_FHSIZE);
2618         /*
2619          * Don't use the deferral mechanism for NFSv4; compounds make it
2620          * too hard to avoid non-idempotency problems.
2621          */
2622         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2623
2624         /*
2625          * According to RFC3010, this takes precedence over all other errors.
2626          */
2627         status = nfserr_minor_vers_mismatch;
2628         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2629                 goto out;
2630
2631         status = nfs41_check_op_ordering(args);
2632         if (status) {
2633                 op = &args->ops[0];
2634                 op->status = status;
2635                 resp->opcnt = 1;
2636                 goto encode_op;
2637         }
2638         check_if_stalefh_allowed(args);
2639
2640         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2641
2642         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2643         while (!status && resp->opcnt < args->opcnt) {
2644                 op = &args->ops[resp->opcnt++];
2645
2646                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2647                         /* If there are still more operations to process,
2648                          * stop here and report NFS4ERR_RESOURCE. */
2649                         if (cstate->minorversion == 0 &&
2650                             args->client_opcnt > resp->opcnt) {
2651                                 op->status = nfserr_resource;
2652                                 goto encode_op;
2653                         }
2654                 }
2655
2656                 /*
2657                  * The XDR decode routines may have pre-set op->status;
2658                  * for example, if there is a miscellaneous XDR error
2659                  * it will be set to nfserr_bad_xdr.
2660                  */
2661                 if (op->status) {
2662                         if (op->opnum == OP_OPEN)
2663                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2664                         goto encode_op;
2665                 }
2666                 if (!current_fh->fh_dentry &&
2667                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2668                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2669                                 op->status = nfserr_nofilehandle;
2670                                 goto encode_op;
2671                         }
2672                 } else if (current_fh->fh_export &&
2673                            current_fh->fh_export->ex_fslocs.migrated &&
2674                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2675                         op->status = nfserr_moved;
2676                         goto encode_op;
2677                 }
2678
2679                 fh_clear_pre_post_attrs(current_fh);
2680
2681                 /* If op is non-idempotent */
2682                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2683                         /*
2684                          * Don't execute this op if we couldn't encode a
2685                          * successful reply:
2686                          */
2687                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2688                         /*
2689                          * Plus if there's another operation, make sure
2690                          * we'll have space to at least encode an error:
2691                          */
2692                         if (resp->opcnt < args->opcnt)
2693                                 plen += COMPOUND_ERR_SLACK_SPACE;
2694                         op->status = nfsd4_check_resp_size(resp, plen);
2695                 }
2696
2697                 if (op->status)
2698                         goto encode_op;
2699
2700                 if (op->opdesc->op_get_currentstateid)
2701                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2702                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2703
2704                 /* Only from SEQUENCE */
2705                 if (cstate->status == nfserr_replay_cache) {
2706                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2707                         status = op->status;
2708                         goto out;
2709                 }
2710                 if (!op->status) {
2711                         if (op->opdesc->op_set_currentstateid)
2712                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2713
2714                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2715                                 clear_current_stateid(cstate);
2716
2717                         if (current_fh->fh_export &&
2718                                         need_wrongsec_check(rqstp))
2719                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2720                 }
2721 encode_op:
2722                 if (op->status == nfserr_replay_me) {
2723                         op->replay = &cstate->replay_owner->so_replay;
2724                         nfsd4_encode_replay(resp->xdr, op);
2725                         status = op->status = op->replay->rp_status;
2726                 } else {
2727                         nfsd4_encode_operation(resp, op);
2728                         status = op->status;
2729                 }
2730
2731                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2732                                            status, nfsd4_op_name(op->opnum));
2733
2734                 nfsd4_cstate_clear_replay(cstate);
2735                 nfsd4_increment_op_stats(op->opnum);
2736         }
2737
2738         fh_put(current_fh);
2739         fh_put(save_fh);
2740         BUG_ON(cstate->replay_owner);
2741 out:
2742         cstate->status = status;
2743         /* Reset deferral mechanism for RPC deferrals */
2744         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2745         return rpc_success;
2746 }
2747
2748 #define op_encode_hdr_size              (2)
2749 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2750 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2751 #define op_encode_change_info_maxsz     (5)
2752 #define nfs4_fattr_bitmap_maxsz         (4)
2753
2754 /* We'll fall back on returning no lockowner if run out of space: */
2755 #define op_encode_lockowner_maxsz       (0)
2756 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2757
2758 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2759
2760 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2761 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2762                                          op_encode_ace_maxsz)
2763
2764 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2765
2766 /*
2767  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2768  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2769  * the maximum payload size here, based on transport limits and the size
2770  * of the remaining space in the rq_pages array.
2771  */
2772 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2773 {
2774         u32 buflen;
2775
2776         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2777         buflen -= rqstp->rq_auth_slack;
2778         buflen -= rqstp->rq_res.head[0].iov_len;
2779         return min_t(u32, buflen, svc_max_payload(rqstp));
2780 }
2781
2782 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2783                                    const struct nfsd4_op *op)
2784 {
2785         return (op_encode_hdr_size) * sizeof(__be32);
2786 }
2787
2788 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2789                                       const struct nfsd4_op *op)
2790 {
2791         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2792 }
2793
2794 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2795                               const struct nfsd4_op *op)
2796 {
2797         /* ac_supported, ac_resp_access */
2798         return (op_encode_hdr_size + 2)* sizeof(__be32);
2799 }
2800
2801 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2802                               const struct nfsd4_op *op)
2803 {
2804         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2805 }
2806
2807 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2808                               const struct nfsd4_op *op)
2809 {
2810         return (op_encode_hdr_size + op_encode_change_info_maxsz
2811                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2812 }
2813
2814 /*
2815  * Note since this is an idempotent operation we won't insist on failing
2816  * the op prematurely if the estimate is too large.  We may turn off splice
2817  * reads unnecessarily.
2818  */
2819 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2820                                const struct nfsd4_op *op)
2821 {
2822         const u32 *bmap = op->u.getattr.ga_bmval;
2823         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2824         u32 ret = 0;
2825
2826         if (bmap0 & FATTR4_WORD0_ACL)
2827                 return nfsd4_max_payload(rqstp);
2828         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2829                 return nfsd4_max_payload(rqstp);
2830
2831         if (bmap1 & FATTR4_WORD1_OWNER) {
2832                 ret += IDMAP_NAMESZ + 4;
2833                 bmap1 &= ~FATTR4_WORD1_OWNER;
2834         }
2835         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2836                 ret += IDMAP_NAMESZ + 4;
2837                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2838         }
2839         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2840                 ret += NFS4_FHSIZE + 4;
2841                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2842         }
2843         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2844                 ret += NFS4_MAXLABELLEN + 12;
2845                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2846         }
2847         /*
2848          * Largest of remaining attributes are 16 bytes (e.g.,
2849          * supported_attributes)
2850          */
2851         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2852         /* bitmask, length */
2853         ret += 20;
2854         return ret;
2855 }
2856
2857 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2858                              const struct nfsd4_op *op)
2859 {
2860         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2861 }
2862
2863 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2864                             const struct nfsd4_op *op)
2865 {
2866         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2867                 * sizeof(__be32);
2868 }
2869
2870 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2871                             const struct nfsd4_op *op)
2872 {
2873         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2874                 * sizeof(__be32);
2875 }
2876
2877 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2878                             const struct nfsd4_op *op)
2879 {
2880         return (op_encode_hdr_size + op_encode_stateid_maxsz
2881                 + op_encode_change_info_maxsz + 1
2882                 + nfs4_fattr_bitmap_maxsz
2883                 + op_encode_delegation_maxsz) * sizeof(__be32);
2884 }
2885
2886 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2887                             const struct nfsd4_op *op)
2888 {
2889         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2890
2891         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2892 }
2893
2894 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2895                                  const struct nfsd4_op *op)
2896 {
2897         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2898         /*
2899          * If we detect that the file changed during hole encoding, then we
2900          * recover by encoding the remaining reply as data. This means we need
2901          * to set aside enough room to encode two data segments.
2902          */
2903         u32 seg_len = 2 * (1 + 2 + 1);
2904
2905         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2906 }
2907
2908 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2909                                const struct nfsd4_op *op)
2910 {
2911         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2912
2913         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2914                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2915 }
2916
2917 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2918                                 const struct nfsd4_op *op)
2919 {
2920         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2921 }
2922
2923 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2924                               const struct nfsd4_op *op)
2925 {
2926         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2927                 * sizeof(__be32);
2928 }
2929
2930 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2931                               const struct nfsd4_op *op)
2932 {
2933         return (op_encode_hdr_size + op_encode_change_info_maxsz
2934                 + op_encode_change_info_maxsz) * sizeof(__be32);
2935 }
2936
2937 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2938                                 const struct nfsd4_op *op)
2939 {
2940         return (op_encode_hdr_size
2941                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2942 }
2943
2944 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2945                                     const struct nfsd4_op *op)
2946 {
2947         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2948                 * sizeof(__be32);
2949 }
2950
2951 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2952                                const struct nfsd4_op *op)
2953 {
2954         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2955 }
2956
2957 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2958                                const struct nfsd4_op *op)
2959 {
2960         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2961                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2962 }
2963
2964 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2965                                    const struct nfsd4_op *op)
2966 {
2967         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2968                                                                 sizeof(__be32);
2969 }
2970
2971 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
2972                              const struct nfsd4_op *op)
2973 {
2974         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2975 }
2976
2977 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
2978                                    const struct nfsd4_op *op)
2979 {
2980         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2981                 1 + 1 + /* eir_flags, spr_how */\
2982                 4 + /* spo_must_enforce & _allow with bitmap */\
2983                 2 + /*eir_server_owner.so_minor_id */\
2984                 /* eir_server_owner.so_major_id<> */\
2985                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2986                 /* eir_server_scope<> */\
2987                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2988                 1 + /* eir_server_impl_id array length */\
2989                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2990 }
2991
2992 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
2993                                             const struct nfsd4_op *op)
2994 {
2995         return (op_encode_hdr_size + \
2996                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2997                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2998 }
2999
3000 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3001                                       const struct nfsd4_op *op)
3002 {
3003         return (op_encode_hdr_size + \
3004                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3005                 2 + /* csr_sequence, csr_flags */\
3006                 op_encode_channel_attrs_maxsz + \
3007                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3008 }
3009
3010 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3011                             const struct nfsd4_op *op)
3012 {
3013         return (op_encode_hdr_size +
3014                 1 /* wr_callback */ +
3015                 op_encode_stateid_maxsz /* wr_callback */ +
3016                 2 /* wr_count */ +
3017                 1 /* wr_committed */ +
3018                 op_encode_verifier_maxsz +
3019                 1 /* cr_consecutive */ +
3020                 1 /* cr_synchronous */) * sizeof(__be32);
3021 }
3022
3023 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3024                                       const struct nfsd4_op *op)
3025 {
3026         return (op_encode_hdr_size +
3027                 2 /* osr_count */ +
3028                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3029 }
3030
3031 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3032                                    const struct nfsd4_op *op)
3033 {
3034         return (op_encode_hdr_size +
3035                 3 /* cnr_lease_time */ +
3036                 1 /* We support one cnr_source_server */ +
3037                 1 /* cnr_stateid seq */ +
3038                 op_encode_stateid_maxsz /* cnr_stateid */ +
3039                 1 /* num cnr_source_server*/ +
3040                 1 /* nl4_type */ +
3041                 1 /* nl4 size */ +
3042                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3043                 * sizeof(__be32);
3044 }
3045
3046 #ifdef CONFIG_NFSD_PNFS
3047 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3048                                      const struct nfsd4_op *op)
3049 {
3050         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3051
3052         return (op_encode_hdr_size +
3053                 1 /* gd_layout_type*/ +
3054                 XDR_QUADLEN(rlen) +
3055                 2 /* gd_notify_types */) * sizeof(__be32);
3056 }
3057
3058 /*
3059  * At this stage we don't really know what layout driver will handle the request,
3060  * so we need to define an arbitrary upper bound here.
3061  */
3062 #define MAX_LAYOUT_SIZE         128
3063 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3064                                  const struct nfsd4_op *op)
3065 {
3066         return (op_encode_hdr_size +
3067                 1 /* logr_return_on_close */ +
3068                 op_encode_stateid_maxsz +
3069                 1 /* nr of layouts */ +
3070                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3071 }
3072
3073 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3074                                     const struct nfsd4_op *op)
3075 {
3076         return (op_encode_hdr_size +
3077                 1 /* locr_newsize */ +
3078                 2 /* ns_size */) * sizeof(__be32);
3079 }
3080
3081 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3082                                     const struct nfsd4_op *op)
3083 {
3084         return (op_encode_hdr_size +
3085                 1 /* lrs_stateid */ +
3086                 op_encode_stateid_maxsz) * sizeof(__be32);
3087 }
3088 #endif /* CONFIG_NFSD_PNFS */
3089
3090
3091 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3092                             const struct nfsd4_op *op)
3093 {
3094         return (op_encode_hdr_size + 3) * sizeof(__be32);
3095 }
3096
3097 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3098                                 const struct nfsd4_op *op)
3099 {
3100         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3101
3102         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3103 }
3104
3105 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3106                                 const struct nfsd4_op *op)
3107 {
3108         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3109                 * sizeof(__be32);
3110 }
3111 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3112                                   const struct nfsd4_op *op)
3113 {
3114         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3115
3116         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3117 }
3118
3119 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3120                                    const struct nfsd4_op *op)
3121 {
3122         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3123                 * sizeof(__be32);
3124 }
3125
3126
3127 static const struct nfsd4_operation nfsd4_ops[] = {
3128         [OP_ACCESS] = {
3129                 .op_func = nfsd4_access,
3130                 .op_name = "OP_ACCESS",
3131                 .op_rsize_bop = nfsd4_access_rsize,
3132         },
3133         [OP_CLOSE] = {
3134                 .op_func = nfsd4_close,
3135                 .op_flags = OP_MODIFIES_SOMETHING,
3136                 .op_name = "OP_CLOSE",
3137                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3138                 .op_get_currentstateid = nfsd4_get_closestateid,
3139                 .op_set_currentstateid = nfsd4_set_closestateid,
3140         },
3141         [OP_COMMIT] = {
3142                 .op_func = nfsd4_commit,
3143                 .op_flags = OP_MODIFIES_SOMETHING,
3144                 .op_name = "OP_COMMIT",
3145                 .op_rsize_bop = nfsd4_commit_rsize,
3146         },
3147         [OP_CREATE] = {
3148                 .op_func = nfsd4_create,
3149                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3150                 .op_name = "OP_CREATE",
3151                 .op_rsize_bop = nfsd4_create_rsize,
3152         },
3153         [OP_DELEGRETURN] = {
3154                 .op_func = nfsd4_delegreturn,
3155                 .op_flags = OP_MODIFIES_SOMETHING,
3156                 .op_name = "OP_DELEGRETURN",
3157                 .op_rsize_bop = nfsd4_only_status_rsize,
3158                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3159         },
3160         [OP_GETATTR] = {
3161                 .op_func = nfsd4_getattr,
3162                 .op_flags = ALLOWED_ON_ABSENT_FS,
3163                 .op_rsize_bop = nfsd4_getattr_rsize,
3164                 .op_name = "OP_GETATTR",
3165         },
3166         [OP_GETFH] = {
3167                 .op_func = nfsd4_getfh,
3168                 .op_name = "OP_GETFH",
3169                 .op_rsize_bop = nfsd4_getfh_rsize,
3170         },
3171         [OP_LINK] = {
3172                 .op_func = nfsd4_link,
3173                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3174                                 | OP_CACHEME,
3175                 .op_name = "OP_LINK",
3176                 .op_rsize_bop = nfsd4_link_rsize,
3177         },
3178         [OP_LOCK] = {
3179                 .op_func = nfsd4_lock,
3180                 .op_flags = OP_MODIFIES_SOMETHING |
3181                                 OP_NONTRIVIAL_ERROR_ENCODE,
3182                 .op_name = "OP_LOCK",
3183                 .op_rsize_bop = nfsd4_lock_rsize,
3184                 .op_set_currentstateid = nfsd4_set_lockstateid,
3185         },
3186         [OP_LOCKT] = {
3187                 .op_func = nfsd4_lockt,
3188                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3189                 .op_name = "OP_LOCKT",
3190                 .op_rsize_bop = nfsd4_lock_rsize,
3191         },
3192         [OP_LOCKU] = {
3193                 .op_func = nfsd4_locku,
3194                 .op_flags = OP_MODIFIES_SOMETHING,
3195                 .op_name = "OP_LOCKU",
3196                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3197                 .op_get_currentstateid = nfsd4_get_lockustateid,
3198         },
3199         [OP_LOOKUP] = {
3200                 .op_func = nfsd4_lookup,
3201                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3202                 .op_name = "OP_LOOKUP",
3203                 .op_rsize_bop = nfsd4_only_status_rsize,
3204         },
3205         [OP_LOOKUPP] = {
3206                 .op_func = nfsd4_lookupp,
3207                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3208                 .op_name = "OP_LOOKUPP",
3209                 .op_rsize_bop = nfsd4_only_status_rsize,
3210         },
3211         [OP_NVERIFY] = {
3212                 .op_func = nfsd4_nverify,
3213                 .op_name = "OP_NVERIFY",
3214                 .op_rsize_bop = nfsd4_only_status_rsize,
3215         },
3216         [OP_OPEN] = {
3217                 .op_func = nfsd4_open,
3218                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3219                 .op_name = "OP_OPEN",
3220                 .op_rsize_bop = nfsd4_open_rsize,
3221                 .op_set_currentstateid = nfsd4_set_openstateid,
3222         },
3223         [OP_OPEN_CONFIRM] = {
3224                 .op_func = nfsd4_open_confirm,
3225                 .op_flags = OP_MODIFIES_SOMETHING,
3226                 .op_name = "OP_OPEN_CONFIRM",
3227                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3228         },
3229         [OP_OPEN_DOWNGRADE] = {
3230                 .op_func = nfsd4_open_downgrade,
3231                 .op_flags = OP_MODIFIES_SOMETHING,
3232                 .op_name = "OP_OPEN_DOWNGRADE",
3233                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3234                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3235                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3236         },
3237         [OP_PUTFH] = {
3238                 .op_func = nfsd4_putfh,
3239                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3240                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3241                 .op_name = "OP_PUTFH",
3242                 .op_rsize_bop = nfsd4_only_status_rsize,
3243         },
3244         [OP_PUTPUBFH] = {
3245                 .op_func = nfsd4_putrootfh,
3246                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3247                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3248                 .op_name = "OP_PUTPUBFH",
3249                 .op_rsize_bop = nfsd4_only_status_rsize,
3250         },
3251         [OP_PUTROOTFH] = {
3252                 .op_func = nfsd4_putrootfh,
3253                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3254                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3255                 .op_name = "OP_PUTROOTFH",
3256                 .op_rsize_bop = nfsd4_only_status_rsize,
3257         },
3258         [OP_READ] = {
3259                 .op_func = nfsd4_read,
3260                 .op_release = nfsd4_read_release,
3261                 .op_name = "OP_READ",
3262                 .op_rsize_bop = nfsd4_read_rsize,
3263                 .op_get_currentstateid = nfsd4_get_readstateid,
3264         },
3265         [OP_READDIR] = {
3266                 .op_func = nfsd4_readdir,
3267                 .op_name = "OP_READDIR",
3268                 .op_rsize_bop = nfsd4_readdir_rsize,
3269         },
3270         [OP_READLINK] = {
3271                 .op_func = nfsd4_readlink,
3272                 .op_name = "OP_READLINK",
3273                 .op_rsize_bop = nfsd4_readlink_rsize,
3274         },
3275         [OP_REMOVE] = {
3276                 .op_func = nfsd4_remove,
3277                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3278                 .op_name = "OP_REMOVE",
3279                 .op_rsize_bop = nfsd4_remove_rsize,
3280         },
3281         [OP_RENAME] = {
3282                 .op_func = nfsd4_rename,
3283                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3284                 .op_name = "OP_RENAME",
3285                 .op_rsize_bop = nfsd4_rename_rsize,
3286         },
3287         [OP_RENEW] = {
3288                 .op_func = nfsd4_renew,
3289                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3290                                 | OP_MODIFIES_SOMETHING,
3291                 .op_name = "OP_RENEW",
3292                 .op_rsize_bop = nfsd4_only_status_rsize,
3293
3294         },
3295         [OP_RESTOREFH] = {
3296                 .op_func = nfsd4_restorefh,
3297                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3298                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3299                 .op_name = "OP_RESTOREFH",
3300                 .op_rsize_bop = nfsd4_only_status_rsize,
3301         },
3302         [OP_SAVEFH] = {
3303                 .op_func = nfsd4_savefh,
3304                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3305                 .op_name = "OP_SAVEFH",
3306                 .op_rsize_bop = nfsd4_only_status_rsize,
3307         },
3308         [OP_SECINFO] = {
3309                 .op_func = nfsd4_secinfo,
3310                 .op_release = nfsd4_secinfo_release,
3311                 .op_flags = OP_HANDLES_WRONGSEC,
3312                 .op_name = "OP_SECINFO",
3313                 .op_rsize_bop = nfsd4_secinfo_rsize,
3314         },
3315         [OP_SETATTR] = {
3316                 .op_func = nfsd4_setattr,
3317                 .op_name = "OP_SETATTR",
3318                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3319                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3320                 .op_rsize_bop = nfsd4_setattr_rsize,
3321                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3322         },
3323         [OP_SETCLIENTID] = {
3324                 .op_func = nfsd4_setclientid,
3325                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3326                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3327                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3328                 .op_name = "OP_SETCLIENTID",
3329                 .op_rsize_bop = nfsd4_setclientid_rsize,
3330         },
3331         [OP_SETCLIENTID_CONFIRM] = {
3332                 .op_func = nfsd4_setclientid_confirm,
3333                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3334                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3335                 .op_name = "OP_SETCLIENTID_CONFIRM",
3336                 .op_rsize_bop = nfsd4_only_status_rsize,
3337         },
3338         [OP_VERIFY] = {
3339                 .op_func = nfsd4_verify,
3340                 .op_name = "OP_VERIFY",
3341                 .op_rsize_bop = nfsd4_only_status_rsize,
3342         },
3343         [OP_WRITE] = {
3344                 .op_func = nfsd4_write,
3345                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3346                 .op_name = "OP_WRITE",
3347                 .op_rsize_bop = nfsd4_write_rsize,
3348                 .op_get_currentstateid = nfsd4_get_writestateid,
3349         },
3350         [OP_RELEASE_LOCKOWNER] = {
3351                 .op_func = nfsd4_release_lockowner,
3352                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3353                                 | OP_MODIFIES_SOMETHING,
3354                 .op_name = "OP_RELEASE_LOCKOWNER",
3355                 .op_rsize_bop = nfsd4_only_status_rsize,
3356         },
3357
3358         /* NFSv4.1 operations */
3359         [OP_EXCHANGE_ID] = {
3360                 .op_func = nfsd4_exchange_id,
3361                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3362                                 | OP_MODIFIES_SOMETHING,
3363                 .op_name = "OP_EXCHANGE_ID",
3364                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3365         },
3366         [OP_BACKCHANNEL_CTL] = {
3367                 .op_func = nfsd4_backchannel_ctl,
3368                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3369                 .op_name = "OP_BACKCHANNEL_CTL",
3370                 .op_rsize_bop = nfsd4_only_status_rsize,
3371         },
3372         [OP_BIND_CONN_TO_SESSION] = {
3373                 .op_func = nfsd4_bind_conn_to_session,
3374                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3375                                 | OP_MODIFIES_SOMETHING,
3376                 .op_name = "OP_BIND_CONN_TO_SESSION",
3377                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3378         },
3379         [OP_CREATE_SESSION] = {
3380                 .op_func = nfsd4_create_session,
3381                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3382                                 | OP_MODIFIES_SOMETHING,
3383                 .op_name = "OP_CREATE_SESSION",
3384                 .op_rsize_bop = nfsd4_create_session_rsize,
3385         },
3386         [OP_DESTROY_SESSION] = {
3387                 .op_func = nfsd4_destroy_session,
3388                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3389                                 | OP_MODIFIES_SOMETHING,
3390                 .op_name = "OP_DESTROY_SESSION",
3391                 .op_rsize_bop = nfsd4_only_status_rsize,
3392         },
3393         [OP_SEQUENCE] = {
3394                 .op_func = nfsd4_sequence,
3395                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3396                 .op_name = "OP_SEQUENCE",
3397                 .op_rsize_bop = nfsd4_sequence_rsize,
3398         },
3399         [OP_DESTROY_CLIENTID] = {
3400                 .op_func = nfsd4_destroy_clientid,
3401                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3402                                 | OP_MODIFIES_SOMETHING,
3403                 .op_name = "OP_DESTROY_CLIENTID",
3404                 .op_rsize_bop = nfsd4_only_status_rsize,
3405         },
3406         [OP_RECLAIM_COMPLETE] = {
3407                 .op_func = nfsd4_reclaim_complete,
3408                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3409                 .op_name = "OP_RECLAIM_COMPLETE",
3410                 .op_rsize_bop = nfsd4_only_status_rsize,
3411         },
3412         [OP_SECINFO_NO_NAME] = {
3413                 .op_func = nfsd4_secinfo_no_name,
3414                 .op_release = nfsd4_secinfo_no_name_release,
3415                 .op_flags = OP_HANDLES_WRONGSEC,
3416                 .op_name = "OP_SECINFO_NO_NAME",
3417                 .op_rsize_bop = nfsd4_secinfo_rsize,
3418         },
3419         [OP_TEST_STATEID] = {
3420                 .op_func = nfsd4_test_stateid,
3421                 .op_flags = ALLOWED_WITHOUT_FH,
3422                 .op_name = "OP_TEST_STATEID",
3423                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3424         },
3425         [OP_FREE_STATEID] = {
3426                 .op_func = nfsd4_free_stateid,
3427                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3428                 .op_name = "OP_FREE_STATEID",
3429                 .op_get_currentstateid = nfsd4_get_freestateid,
3430                 .op_rsize_bop = nfsd4_only_status_rsize,
3431         },
3432 #ifdef CONFIG_NFSD_PNFS
3433         [OP_GETDEVICEINFO] = {
3434                 .op_func = nfsd4_getdeviceinfo,
3435                 .op_release = nfsd4_getdeviceinfo_release,
3436                 .op_flags = ALLOWED_WITHOUT_FH,
3437                 .op_name = "OP_GETDEVICEINFO",
3438                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3439         },
3440         [OP_LAYOUTGET] = {
3441                 .op_func = nfsd4_layoutget,
3442                 .op_release = nfsd4_layoutget_release,
3443                 .op_flags = OP_MODIFIES_SOMETHING,
3444                 .op_name = "OP_LAYOUTGET",
3445                 .op_rsize_bop = nfsd4_layoutget_rsize,
3446         },
3447         [OP_LAYOUTCOMMIT] = {
3448                 .op_func = nfsd4_layoutcommit,
3449                 .op_flags = OP_MODIFIES_SOMETHING,
3450                 .op_name = "OP_LAYOUTCOMMIT",
3451                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3452         },
3453         [OP_LAYOUTRETURN] = {
3454                 .op_func = nfsd4_layoutreturn,
3455                 .op_flags = OP_MODIFIES_SOMETHING,
3456                 .op_name = "OP_LAYOUTRETURN",
3457                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3458         },
3459 #endif /* CONFIG_NFSD_PNFS */
3460
3461         /* NFSv4.2 operations */
3462         [OP_ALLOCATE] = {
3463                 .op_func = nfsd4_allocate,
3464                 .op_flags = OP_MODIFIES_SOMETHING,
3465                 .op_name = "OP_ALLOCATE",
3466                 .op_rsize_bop = nfsd4_only_status_rsize,
3467         },
3468         [OP_DEALLOCATE] = {
3469                 .op_func = nfsd4_deallocate,
3470                 .op_flags = OP_MODIFIES_SOMETHING,
3471                 .op_name = "OP_DEALLOCATE",
3472                 .op_rsize_bop = nfsd4_only_status_rsize,
3473         },
3474         [OP_CLONE] = {
3475                 .op_func = nfsd4_clone,
3476                 .op_flags = OP_MODIFIES_SOMETHING,
3477                 .op_name = "OP_CLONE",
3478                 .op_rsize_bop = nfsd4_only_status_rsize,
3479         },
3480         [OP_COPY] = {
3481                 .op_func = nfsd4_copy,
3482                 .op_flags = OP_MODIFIES_SOMETHING,
3483                 .op_name = "OP_COPY",
3484                 .op_rsize_bop = nfsd4_copy_rsize,
3485         },
3486         [OP_READ_PLUS] = {
3487                 .op_func = nfsd4_read,
3488                 .op_release = nfsd4_read_release,
3489                 .op_name = "OP_READ_PLUS",
3490                 .op_rsize_bop = nfsd4_read_plus_rsize,
3491                 .op_get_currentstateid = nfsd4_get_readstateid,
3492         },
3493         [OP_SEEK] = {
3494                 .op_func = nfsd4_seek,
3495                 .op_name = "OP_SEEK",
3496                 .op_rsize_bop = nfsd4_seek_rsize,
3497         },
3498         [OP_OFFLOAD_STATUS] = {
3499                 .op_func = nfsd4_offload_status,
3500                 .op_name = "OP_OFFLOAD_STATUS",
3501                 .op_rsize_bop = nfsd4_offload_status_rsize,
3502         },
3503         [OP_OFFLOAD_CANCEL] = {
3504                 .op_func = nfsd4_offload_cancel,
3505                 .op_flags = OP_MODIFIES_SOMETHING,
3506                 .op_name = "OP_OFFLOAD_CANCEL",
3507                 .op_rsize_bop = nfsd4_only_status_rsize,
3508         },
3509         [OP_COPY_NOTIFY] = {
3510                 .op_func = nfsd4_copy_notify,
3511                 .op_flags = OP_MODIFIES_SOMETHING,
3512                 .op_name = "OP_COPY_NOTIFY",
3513                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3514         },
3515         [OP_GETXATTR] = {
3516                 .op_func = nfsd4_getxattr,
3517                 .op_name = "OP_GETXATTR",
3518                 .op_rsize_bop = nfsd4_getxattr_rsize,
3519         },
3520         [OP_SETXATTR] = {
3521                 .op_func = nfsd4_setxattr,
3522                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3523                 .op_name = "OP_SETXATTR",
3524                 .op_rsize_bop = nfsd4_setxattr_rsize,
3525         },
3526         [OP_LISTXATTRS] = {
3527                 .op_func = nfsd4_listxattrs,
3528                 .op_name = "OP_LISTXATTRS",
3529                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3530         },
3531         [OP_REMOVEXATTR] = {
3532                 .op_func = nfsd4_removexattr,
3533                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3534                 .op_name = "OP_REMOVEXATTR",
3535                 .op_rsize_bop = nfsd4_removexattr_rsize,
3536         },
3537 };
3538
3539 /**
3540  * nfsd4_spo_must_allow - Determine if the compound op contains an
3541  * operation that is allowed to be sent with machine credentials
3542  *
3543  * @rqstp: a pointer to the struct svc_rqst
3544  *
3545  * Checks to see if the compound contains a spo_must_allow op
3546  * and confirms that it was sent with the proper machine creds.
3547  */
3548
3549 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3550 {
3551         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3552         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3553         struct nfsd4_op *this;
3554         struct nfsd4_compound_state *cstate = &resp->cstate;
3555         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3556         u32 opiter;
3557
3558         if (!cstate->minorversion)
3559                 return false;
3560
3561         if (cstate->spo_must_allowed)
3562                 return true;
3563
3564         opiter = resp->opcnt;
3565         while (opiter < argp->opcnt) {
3566                 this = &argp->ops[opiter++];
3567                 if (test_bit(this->opnum, allow->u.longs) &&
3568                         cstate->clp->cl_mach_cred &&
3569                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3570                         cstate->spo_must_allowed = true;
3571                         return true;
3572                 }
3573         }
3574         cstate->spo_must_allowed = false;
3575         return false;
3576 }
3577
3578 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3579 {
3580         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3581                 return op_encode_hdr_size * sizeof(__be32);
3582
3583         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3584         return OPDESC(op)->op_rsize_bop(rqstp, op);
3585 }
3586
3587 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3588 {
3589         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3590                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3591                         op->opnum, nfsd4_op_name(op->opnum));
3592                 WARN_ON_ONCE(1);
3593         }
3594 }
3595
3596 static const char *nfsd4_op_name(unsigned opnum)
3597 {
3598         if (opnum < ARRAY_SIZE(nfsd4_ops))
3599                 return nfsd4_ops[opnum].op_name;
3600         return "unknown_operation";
3601 }
3602
3603 static const struct svc_procedure nfsd_procedures4[2] = {
3604         [NFSPROC4_NULL] = {
3605                 .pc_func = nfsd4_proc_null,
3606                 .pc_decode = nfssvc_decode_voidarg,
3607                 .pc_encode = nfssvc_encode_voidres,
3608                 .pc_argsize = sizeof(struct nfsd_voidargs),
3609                 .pc_argzero = sizeof(struct nfsd_voidargs),
3610                 .pc_ressize = sizeof(struct nfsd_voidres),
3611                 .pc_cachetype = RC_NOCACHE,
3612                 .pc_xdrressize = 1,
3613                 .pc_name = "NULL",
3614         },
3615         [NFSPROC4_COMPOUND] = {
3616                 .pc_func = nfsd4_proc_compound,
3617                 .pc_decode = nfs4svc_decode_compoundargs,
3618                 .pc_encode = nfs4svc_encode_compoundres,
3619                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3620                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3621                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3622                 .pc_release = nfsd4_release_compoundargs,
3623                 .pc_cachetype = RC_NOCACHE,
3624                 .pc_xdrressize = NFSD_BUFSIZE/4,
3625                 .pc_name = "COMPOUND",
3626         },
3627 };
3628
3629 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3630 const struct svc_version nfsd_version4 = {
3631         .vs_vers                = 4,
3632         .vs_nproc               = 2,
3633         .vs_proc                = nfsd_procedures4,
3634         .vs_count               = nfsd_count3,
3635         .vs_dispatch            = nfsd_dispatch,
3636         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3637         .vs_rpcb_optnl          = true,
3638         .vs_need_cong_ctrl      = true,
3639 };