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 <[email protected]>
   8  *  Andy Adamson   <[email protected]>
   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         struct nfsd_file *nf;
 735         __be32 status;
 736
 737         status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
 738                                    NFSD_MAY_NOT_BREAK_LEASE, &nf);
 739         if (status != nfs_ok)
 740                 return status;
 741
 742         status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
 743                              commit->co_count,
 744                              (__be32 *)commit->co_verf.data);
 745         nfsd_file_put(nf);
 746         return status;
 747 }
 748
 749 static __be32
 750 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 751              union nfsd4_op_u *u)
 752 {
 753         struct nfsd4_create *create = &u->create;
 754         struct nfsd_attrs attrs = {
 755                 .na_iattr       = &create->cr_iattr,
 756                 .na_seclabel    = &create->cr_label,
 757         };
 758         struct svc_fh resfh;
 759         __be32 status;
 760         dev_t rdev;
 761
 762         fh_init(&resfh, NFS4_FHSIZE);
 763
 764         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 765         if (status)
 766                 return status;
 767
 768         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 769                                     nfsd_attrmask);
 770         if (status)
 771                 return status;
 772
 773         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 774         current->fs->umask = create->cr_umask;
 775         switch (create->cr_type) {
 776         case NF4LNK:
 777                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 778                                       create->cr_name, create->cr_namelen,
 779                                       create->cr_data, &attrs, &resfh);
 780                 break;
 781
 782         case NF4BLK:
 783                 status = nfserr_inval;
 784                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 785                 if (MAJOR(rdev) != create->cr_specdata1 ||
 786                     MINOR(rdev) != create->cr_specdata2)
 787                         goto out_umask;
 788                 status = nfsd_create(rqstp, &cstate->current_fh,
 789                                      create->cr_name, create->cr_namelen,
 790                                      &attrs, S_IFBLK, rdev, &resfh);
 791                 break;
 792
 793         case NF4CHR:
 794                 status = nfserr_inval;
 795                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 796                 if (MAJOR(rdev) != create->cr_specdata1 ||
 797                     MINOR(rdev) != create->cr_specdata2)
 798                         goto out_umask;
 799                 status = nfsd_create(rqstp, &cstate->current_fh,
 800                                      create->cr_name, create->cr_namelen,
 801                                      &attrs, S_IFCHR, rdev, &resfh);
 802                 break;
 803
 804         case NF4SOCK:
 805                 status = nfsd_create(rqstp, &cstate->current_fh,
 806                                      create->cr_name, create->cr_namelen,
 807                                      &attrs, S_IFSOCK, 0, &resfh);
 808                 break;
 809
 810         case NF4FIFO:
 811                 status = nfsd_create(rqstp, &cstate->current_fh,
 812                                      create->cr_name, create->cr_namelen,
 813                                      &attrs, S_IFIFO, 0, &resfh);
 814                 break;
 815
 816         case NF4DIR:
 817                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 818                 status = nfsd_create(rqstp, &cstate->current_fh,
 819                                      create->cr_name, create->cr_namelen,
 820                                      &attrs, S_IFDIR, 0, &resfh);
 821                 break;
 822
 823         default:
 824                 status = nfserr_badtype;
 825         }
 826
 827         if (status)
 828                 goto out;
 829
 830         if (attrs.na_labelerr)
 831                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 832         if (attrs.na_aclerr)
 833                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 834         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 835         fh_dup2(&cstate->current_fh, &resfh);
 836 out:
 837         fh_put(&resfh);
 838 out_umask:
 839         current->fs->umask = 0;
 840         nfsd_attrs_free(&attrs);
 841         return status;
 842 }
 843
 844 static __be32
 845 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 846               union nfsd4_op_u *u)
 847 {
 848         struct nfsd4_getattr *getattr = &u->getattr;
 849         __be32 status;
 850
 851         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 852         if (status)
 853                 return status;
 854
 855         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 856                 return nfserr_inval;
 857
 858         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 859         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 860         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 861
 862         getattr->ga_fhp = &cstate->current_fh;
 863         return nfs_ok;
 864 }
 865
 866 static __be32
 867 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 868            union nfsd4_op_u *u)
 869 {
 870         struct nfsd4_link *link = &u->link;
 871         __be32 status;
 872
 873         status = nfsd_link(rqstp, &cstate->current_fh,
 874                            link->li_name, link->li_namelen, &cstate->save_fh);
 875         if (!status)
 876                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 877         return status;
 878 }
 879
 880 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 881 {
 882         struct svc_fh tmp_fh;
 883         __be32 ret;
 884
 885         fh_init(&tmp_fh, NFS4_FHSIZE);
 886         ret = exp_pseudoroot(rqstp, &tmp_fh);
 887         if (ret)
 888                 return ret;
 889         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 890                 fh_put(&tmp_fh);
 891                 return nfserr_noent;
 892         }
 893         fh_put(&tmp_fh);
 894         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 895 }
 896
 897 static __be32
 898 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 899               union nfsd4_op_u *u)
 900 {
 901         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 902 }
 903
 904 static __be32
 905 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 906              union nfsd4_op_u *u)
 907 {
 908         return nfsd_lookup(rqstp, &cstate->current_fh,
 909                            u->lookup.lo_name, u->lookup.lo_len,
 910                            &cstate->current_fh);
 911 }
 912
 913 static __be32
 914 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 915            union nfsd4_op_u *u)
 916 {
 917         struct nfsd4_read *read = &u->read;
 918         __be32 status;
 919
 920         read->rd_nf = NULL;
 921
 922         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 923                               read->rd_offset, read->rd_length);
 924
 925         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 926         if (read->rd_offset > (u64)OFFSET_MAX)
 927                 read->rd_offset = (u64)OFFSET_MAX;
 928         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 929                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 930
 931         /*
 932          * If we do a zero copy read, then a client will see read data
 933          * that reflects the state of the file *after* performing the
 934          * following compound.
 935          *
 936          * To ensure proper ordering, we therefore turn off zero copy if
 937          * the client wants us to do more in this compound:
 938          */
 939         if (!nfsd4_last_compound_op(rqstp))
 940                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 941
 942         /* check stateid */
 943         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 944                                         &read->rd_stateid, RD_STATE,
 945                                         &read->rd_nf, NULL);
 946
 947         read->rd_rqstp = rqstp;
 948         read->rd_fhp = &cstate->current_fh;
 949         return status;
 950 }
 951
 952
 953 static void
 954 nfsd4_read_release(union nfsd4_op_u *u)
 955 {
 956         if (u->read.rd_nf)
 957                 nfsd_file_put(u->read.rd_nf);
 958         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 959                              u->read.rd_offset, u->read.rd_length);
 960 }
 961
 962 static __be32
 963 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 964               union nfsd4_op_u *u)
 965 {
 966         struct nfsd4_readdir *readdir = &u->readdir;
 967         u64 cookie = readdir->rd_cookie;
 968         static const nfs4_verifier zeroverf;
 969
 970         /* no need to check permission - this will be done in nfsd_readdir() */
 971
 972         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 973                 return nfserr_inval;
 974
 975         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 976         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 977         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 978
 979         if ((cookie == 1) || (cookie == 2) ||
 980             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
 981                 return nfserr_bad_cookie;
 982
 983         readdir->rd_rqstp = rqstp;
 984         readdir->rd_fhp = &cstate->current_fh;
 985         return nfs_ok;
 986 }
 987
 988 static __be32
 989 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 990                union nfsd4_op_u *u)
 991 {
 992         u->readlink.rl_rqstp = rqstp;
 993         u->readlink.rl_fhp = &cstate->current_fh;
 994         return nfs_ok;
 995 }
 996
 997 static __be32
 998 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 999              union nfsd4_op_u *u)
1000 {
1001         struct nfsd4_remove *remove = &u->remove;
1002         __be32 status;
1003
1004         if (opens_in_grace(SVC_NET(rqstp)))
1005                 return nfserr_grace;
1006         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1007                              remove->rm_name, remove->rm_namelen);
1008         if (!status)
1009                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1010         return status;
1011 }
1012
1013 static __be32
1014 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1015              union nfsd4_op_u *u)
1016 {
1017         struct nfsd4_rename *rename = &u->rename;
1018         __be32 status;
1019
1020         if (opens_in_grace(SVC_NET(rqstp)))
1021                 return nfserr_grace;
1022         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1023                              rename->rn_snamelen, &cstate->current_fh,
1024                              rename->rn_tname, rename->rn_tnamelen);
1025         if (status)
1026                 return status;
1027         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
1028         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
1029         return nfs_ok;
1030 }
1031
1032 static __be32
1033 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1034               union nfsd4_op_u *u)
1035 {
1036         struct nfsd4_secinfo *secinfo = &u->secinfo;
1037         struct svc_export *exp;
1038         struct dentry *dentry;
1039         __be32 err;
1040
1041         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1042         if (err)
1043                 return err;
1044         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1045                                     secinfo->si_name, secinfo->si_namelen,
1046                                     &exp, &dentry);
1047         if (err)
1048                 return err;
1049         if (d_really_is_negative(dentry)) {
1050                 exp_put(exp);
1051                 err = nfserr_noent;
1052         } else
1053                 secinfo->si_exp = exp;
1054         dput(dentry);
1055         if (cstate->minorversion)
1056                 /* See rfc 5661 section 2.6.3.1.1.8 */
1057                 fh_put(&cstate->current_fh);
1058         return err;
1059 }
1060
1061 static __be32
1062 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1063                 union nfsd4_op_u *u)
1064 {
1065         __be32 err;
1066
1067         switch (u->secinfo_no_name.sin_style) {
1068         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1069                 break;
1070         case NFS4_SECINFO_STYLE4_PARENT:
1071                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1072                 if (err)
1073                         return err;
1074                 break;
1075         default:
1076                 return nfserr_inval;
1077         }
1078
1079         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1080         fh_put(&cstate->current_fh);
1081         return nfs_ok;
1082 }
1083
1084 static void
1085 nfsd4_secinfo_release(union nfsd4_op_u *u)
1086 {
1087         if (u->secinfo.si_exp)
1088                 exp_put(u->secinfo.si_exp);
1089 }
1090
1091 static void
1092 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1093 {
1094         if (u->secinfo_no_name.sin_exp)
1095                 exp_put(u->secinfo_no_name.sin_exp);
1096 }
1097
1098 static __be32
1099 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1100               union nfsd4_op_u *u)
1101 {
1102         struct nfsd4_setattr *setattr = &u->setattr;
1103         struct nfsd_attrs attrs = {
1104                 .na_iattr       = &setattr->sa_iattr,
1105                 .na_seclabel    = &setattr->sa_label,
1106         };
1107         struct inode *inode;
1108         __be32 status = nfs_ok;
1109         int err;
1110
1111         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1112                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1113                                 &cstate->current_fh, &setattr->sa_stateid,
1114                                 WR_STATE, NULL, NULL);
1115                 if (status)
1116                         return status;
1117         }
1118         err = fh_want_write(&cstate->current_fh);
1119         if (err)
1120                 return nfserrno(err);
1121         status = nfs_ok;
1122
1123         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1124                                     nfsd_attrmask);
1125         if (status)
1126                 goto out;
1127
1128         inode = cstate->current_fh.fh_dentry->d_inode;
1129         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1130                                    setattr->sa_acl, &attrs);
1131
1132         if (status)
1133                 goto out;
1134         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1135                                 0, (time64_t)0);
1136         if (!status)
1137                 status = nfserrno(attrs.na_labelerr);
1138         if (!status)
1139                 status = nfserrno(attrs.na_aclerr);
1140 out:
1141         nfsd_attrs_free(&attrs);
1142         fh_drop_write(&cstate->current_fh);
1143         return status;
1144 }
1145
1146 static __be32
1147 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1148             union nfsd4_op_u *u)
1149 {
1150         struct nfsd4_write *write = &u->write;
1151         stateid_t *stateid = &write->wr_stateid;
1152         struct nfsd_file *nf = NULL;
1153         __be32 status = nfs_ok;
1154         unsigned long cnt;
1155         int nvecs;
1156
1157         if (write->wr_offset > (u64)OFFSET_MAX ||
1158             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1159                 return nfserr_fbig;
1160
1161         cnt = write->wr_buflen;
1162         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1163                                write->wr_offset, cnt);
1164         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1165                                                 stateid, WR_STATE, &nf, NULL);
1166         if (status)
1167                 return status;
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                 goto out;
1200
1201         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1202                                             dst_stateid, WR_STATE, dst, NULL);
1203         if (status)
1204                 goto out_put_src;
1205
1206         /* fix up for NFS-specific error code */
1207         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1208             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1209                 status = nfserr_wrong_type;
1210                 goto out_put_dst;
1211         }
1212
1213 out:
1214         return status;
1215 out_put_dst:
1216         nfsd_file_put(*dst);
1217         *dst = NULL;
1218 out_put_src:
1219         nfsd_file_put(*src);
1220         *src = NULL;
1221         goto out;
1222 }
1223
1224 static __be32
1225 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1226                 union nfsd4_op_u *u)
1227 {
1228         struct nfsd4_clone *clone = &u->clone;
1229         struct nfsd_file *src, *dst;
1230         __be32 status;
1231
1232         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1233                                    &clone->cl_dst_stateid, &dst);
1234         if (status)
1235                 goto out;
1236
1237         status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1238                         dst, clone->cl_dst_pos, clone->cl_count,
1239                         EX_ISSYNC(cstate->current_fh.fh_export));
1240
1241         nfsd_file_put(dst);
1242         nfsd_file_put(src);
1243 out:
1244         return status;
1245 }
1246
1247 static void nfs4_put_copy(struct nfsd4_copy *copy)
1248 {
1249         if (!refcount_dec_and_test(&copy->refcount))
1250                 return;
1251         kfree(copy->cp_src);
1252         kfree(copy);
1253 }
1254
1255 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1256 {
1257         if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1258                 kthread_stop(copy->copy_task);
1259         nfs4_put_copy(copy);
1260 }
1261
1262 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1263 {
1264         struct nfsd4_copy *copy = NULL;
1265
1266         spin_lock(&clp->async_lock);
1267         if (!list_empty(&clp->async_copies)) {
1268                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1269                                         copies);
1270                 refcount_inc(&copy->refcount);
1271         }
1272         spin_unlock(&clp->async_lock);
1273         return copy;
1274 }
1275
1276 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1277 {
1278         struct nfsd4_copy *copy;
1279
1280         while ((copy = nfsd4_get_copy(clp)) != NULL)
1281                 nfsd4_stop_copy(copy);
1282 }
1283 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1284
1285 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1286                                    struct nfs_fh *src_fh,
1287                                    nfs4_stateid *stateid);
1288 extern void nfs42_ssc_close(struct file *filep);
1289
1290 extern void nfs_sb_deactive(struct super_block *sb);
1291
1292 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1293
1294 /*
1295  * setup a work entry in the ssc delayed unmount list.
1296  */
1297 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1298                                   struct nfsd4_ssc_umount_item **nsui)
1299 {
1300         struct nfsd4_ssc_umount_item *ni = NULL;
1301         struct nfsd4_ssc_umount_item *work = NULL;
1302         struct nfsd4_ssc_umount_item *tmp;
1303         DEFINE_WAIT(wait);
1304         __be32 status = 0;
1305
1306         *nsui = NULL;
1307         work = kzalloc(sizeof(*work), GFP_KERNEL);
1308 try_again:
1309         spin_lock(&nn->nfsd_ssc_lock);
1310         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1311                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1312                         continue;
1313                 /* found a match */
1314                 if (ni->nsui_busy) {
1315                         /*  wait - and try again */
1316                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait,
1317                                 TASK_INTERRUPTIBLE);
1318                         spin_unlock(&nn->nfsd_ssc_lock);
1319
1320                         /* allow 20secs for mount/unmount for now - revisit */
1321                         if (signal_pending(current) ||
1322                                         (schedule_timeout(20*HZ) == 0)) {
1323                                 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1324                                 kfree(work);
1325                                 return nfserr_eagain;
1326                         }
1327                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1328                         goto try_again;
1329                 }
1330                 *nsui = ni;
1331                 refcount_inc(&ni->nsui_refcnt);
1332                 spin_unlock(&nn->nfsd_ssc_lock);
1333                 kfree(work);
1334
1335                 /* return vfsmount in (*nsui)->nsui_vfsmount */
1336                 return 0;
1337         }
1338         if (work) {
1339                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1340                 refcount_set(&work->nsui_refcnt, 2);
1341                 work->nsui_busy = true;
1342                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1343                 *nsui = work;
1344         } else
1345                 status = nfserr_resource;
1346         spin_unlock(&nn->nfsd_ssc_lock);
1347         return status;
1348 }
1349
1350 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1351                                  struct nfsd4_ssc_umount_item *nsui,
1352                                  struct vfsmount *ss_mnt)
1353 {
1354         spin_lock(&nn->nfsd_ssc_lock);
1355         nsui->nsui_vfsmount = ss_mnt;
1356         nsui->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(struct nfsd_net *nn,
1362                                  struct nfsd4_ssc_umount_item *nsui)
1363 {
1364         spin_lock(&nn->nfsd_ssc_lock);
1365         list_del(&nsui->nsui_list);
1366         wake_up_all(&nn->nfsd_ssc_waitq);
1367         spin_unlock(&nn->nfsd_ssc_lock);
1368         kfree(nsui);
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 nfsd4_ssc_umount_item **nsui)
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 nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1388
1389         naddr = &nss->u.nl4_addr;
1390         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1391                                          naddr->addr_len,
1392                                          (struct sockaddr *)&tmp_addr,
1393                                          sizeof(tmp_addr));
1394         *nsui = NULL;
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, nsui);
1438         if (status)
1439                 goto out_free_devname;
1440         if ((*nsui)->nsui_vfsmount)
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                 nfsd4_ssc_cancel_dul(nn, *nsui);
1449                 goto out_free_devname;
1450         }
1451         nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1452 out_done:
1453         status = 0;
1454
1455 out_free_devname:
1456         kfree(dev_name);
1457 out_free_rawdata:
1458         kfree(raw_data);
1459 out_free_ipaddr:
1460         kfree(ipaddr);
1461 out_err:
1462         return status;
1463 }
1464
1465 /*
1466  * Verify COPY destination stateid.
1467  *
1468  * Connect to the source server with NFSv4.1.
1469  * Create the source struct file for nfsd_copy_range.
1470  * Called with COPY cstate:
1471  *    SAVED_FH: source filehandle
1472  *    CURRENT_FH: destination filehandle
1473  */
1474 static __be32
1475 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1476                       struct nfsd4_compound_state *cstate,
1477                       struct nfsd4_copy *copy)
1478 {
1479         struct svc_fh *s_fh = NULL;
1480         stateid_t *s_stid = &copy->cp_src_stateid;
1481         __be32 status = nfserr_inval;
1482
1483         /* Verify the destination stateid and set dst struct file*/
1484         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1485                                             &copy->cp_dst_stateid,
1486                                             WR_STATE, &copy->nf_dst, NULL);
1487         if (status)
1488                 goto out;
1489
1490         status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1491         if (status)
1492                 goto out;
1493
1494         s_fh = &cstate->save_fh;
1495
1496         copy->c_fh.size = s_fh->fh_handle.fh_size;
1497         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1498         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1499         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1500                sizeof(stateid_opaque_t));
1501
1502         status = 0;
1503 out:
1504         return status;
1505 }
1506
1507 static void
1508 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1509                         struct nfsd_file *dst)
1510 {
1511         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1512         long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1513
1514         nfs42_ssc_close(filp);
1515         fput(filp);
1516
1517         spin_lock(&nn->nfsd_ssc_lock);
1518         list_del(&nsui->nsui_list);
1519         /*
1520          * vfsmount can be shared by multiple exports,
1521          * decrement refcnt. If the count drops to 1 it
1522          * will be unmounted when nsui_expire expires.
1523          */
1524         refcount_dec(&nsui->nsui_refcnt);
1525         nsui->nsui_expire = jiffies + timeout;
1526         list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1527         spin_unlock(&nn->nfsd_ssc_lock);
1528 }
1529
1530 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1531
1532 static __be32
1533 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1534                       struct nfsd4_compound_state *cstate,
1535                       struct nfsd4_copy *copy)
1536 {
1537         return nfserr_inval;
1538 }
1539
1540 static void
1541 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1542                         struct nfsd_file *dst)
1543 {
1544 }
1545
1546 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1547                                    struct nfs_fh *src_fh,
1548                                    nfs4_stateid *stateid)
1549 {
1550         return NULL;
1551 }
1552 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1553
1554 static __be32
1555 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1556                       struct nfsd4_compound_state *cstate,
1557                       struct nfsd4_copy *copy)
1558 {
1559         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1560                                  &copy->nf_src, &copy->cp_dst_stateid,
1561                                  &copy->nf_dst);
1562 }
1563
1564 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1565 {
1566         struct nfsd4_cb_offload *cbo =
1567                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1568
1569         kfree(cbo);
1570 }
1571
1572 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1573                                  struct rpc_task *task)
1574 {
1575         struct nfsd4_cb_offload *cbo =
1576                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1577
1578         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1579         return 1;
1580 }
1581
1582 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1583         .release = nfsd4_cb_offload_release,
1584         .done = nfsd4_cb_offload_done
1585 };
1586
1587 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1588 {
1589         copy->cp_res.wr_stable_how =
1590                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1591                         NFS_FILE_SYNC : NFS_UNSTABLE;
1592         nfsd4_copy_set_sync(copy, sync);
1593         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1594 }
1595
1596 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1597                                      struct file *dst,
1598                                      struct file *src)
1599 {
1600         errseq_t since;
1601         ssize_t bytes_copied = 0;
1602         u64 bytes_total = copy->cp_count;
1603         u64 src_pos = copy->cp_src_pos;
1604         u64 dst_pos = copy->cp_dst_pos;
1605         int status;
1606         loff_t end;
1607
1608         /* See RFC 7862 p.67: */
1609         if (bytes_total == 0)
1610                 bytes_total = ULLONG_MAX;
1611         do {
1612                 if (kthread_should_stop())
1613                         break;
1614                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1615                                                     bytes_total);
1616                 if (bytes_copied <= 0)
1617                         break;
1618                 bytes_total -= bytes_copied;
1619                 copy->cp_res.wr_bytes_written += bytes_copied;
1620                 src_pos += bytes_copied;
1621                 dst_pos += bytes_copied;
1622         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1623         /* for a non-zero asynchronous copy do a commit of data */
1624         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1625                 since = READ_ONCE(dst->f_wb_err);
1626                 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1627                 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1628                 if (!status)
1629                         status = filemap_check_wb_err(dst->f_mapping, since);
1630                 if (!status)
1631                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1632         }
1633         return bytes_copied;
1634 }
1635
1636 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1637                             struct file *src, struct file *dst,
1638                             bool sync)
1639 {
1640         __be32 status;
1641         ssize_t bytes;
1642
1643         bytes = _nfsd_copy_file_range(copy, dst, src);
1644
1645         /* for async copy, we ignore the error, client can always retry
1646          * to get the error
1647          */
1648         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1649                 status = nfserrno(bytes);
1650         else {
1651                 nfsd4_init_copy_res(copy, sync);
1652                 status = nfs_ok;
1653         }
1654         return status;
1655 }
1656
1657 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1658 {
1659         dst->cp_src_pos = src->cp_src_pos;
1660         dst->cp_dst_pos = src->cp_dst_pos;
1661         dst->cp_count = src->cp_count;
1662         dst->cp_flags = src->cp_flags;
1663         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1664         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1665         dst->cp_clp = src->cp_clp;
1666         dst->nf_dst = nfsd_file_get(src->nf_dst);
1667         /* for inter, nf_src doesn't exist yet */
1668         if (!nfsd4_ssc_is_inter(src))
1669                 dst->nf_src = nfsd_file_get(src->nf_src);
1670
1671         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1672         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1673         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1674         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1675         dst->ss_nsui = src->ss_nsui;
1676 }
1677
1678 static void release_copy_files(struct nfsd4_copy *copy)
1679 {
1680         if (copy->nf_src)
1681                 nfsd_file_put(copy->nf_src);
1682         if (copy->nf_dst)
1683                 nfsd_file_put(copy->nf_dst);
1684 }
1685
1686 static void cleanup_async_copy(struct nfsd4_copy *copy)
1687 {
1688         nfs4_free_copy_state(copy);
1689         release_copy_files(copy);
1690         if (copy->cp_clp) {
1691                 spin_lock(&copy->cp_clp->async_lock);
1692                 if (!list_empty(&copy->copies))
1693                         list_del_init(&copy->copies);
1694                 spin_unlock(&copy->cp_clp->async_lock);
1695         }
1696         nfs4_put_copy(copy);
1697 }
1698
1699 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1700 {
1701         struct nfsd4_cb_offload *cbo;
1702
1703         cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1704         if (!cbo)
1705                 return;
1706
1707         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1708         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1709         cbo->co_nfserr = nfserr;
1710
1711         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1712                       NFSPROC4_CLNT_CB_OFFLOAD);
1713         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1714                               &cbo->co_fh, copy->cp_count, nfserr);
1715         nfsd4_run_cb(&cbo->co_cb);
1716 }
1717
1718 /**
1719  * nfsd4_do_async_copy - kthread function for background server-side COPY
1720  * @data: arguments for COPY operation
1721  *
1722  * Return values:
1723  *   %0: Copy operation is done.
1724  */
1725 static int nfsd4_do_async_copy(void *data)
1726 {
1727         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1728         __be32 nfserr;
1729
1730         if (nfsd4_ssc_is_inter(copy)) {
1731                 struct file *filp;
1732
1733                 filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1734                                       &copy->c_fh, &copy->stateid);
1735                 if (IS_ERR(filp)) {
1736                         switch (PTR_ERR(filp)) {
1737                         case -EBADF:
1738                                 nfserr = nfserr_wrong_type;
1739                                 break;
1740                         default:
1741                                 nfserr = nfserr_offload_denied;
1742                         }
1743                         /* ss_mnt will be unmounted by the laundromat */
1744                         goto do_callback;
1745                 }
1746                 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1747                                        false);
1748                 nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1749         } else {
1750                 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1751                                        copy->nf_dst->nf_file, false);
1752         }
1753
1754 do_callback:
1755         nfsd4_send_cb_offload(copy, nfserr);
1756         cleanup_async_copy(copy);
1757         return 0;
1758 }
1759
1760 static __be32
1761 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1762                 union nfsd4_op_u *u)
1763 {
1764         struct nfsd4_copy *copy = &u->copy;
1765         __be32 status;
1766         struct nfsd4_copy *async_copy = NULL;
1767
1768         if (nfsd4_ssc_is_inter(copy)) {
1769                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1770                         status = nfserr_notsupp;
1771                         goto out;
1772                 }
1773                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1774                 if (status)
1775                         return nfserr_offload_denied;
1776         } else {
1777                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1778                 if (status)
1779                         return status;
1780         }
1781
1782         copy->cp_clp = cstate->clp;
1783         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1784                 sizeof(struct knfsd_fh));
1785         if (nfsd4_copy_is_async(copy)) {
1786                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1787
1788                 status = nfserrno(-ENOMEM);
1789                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1790                 if (!async_copy)
1791                         goto out_err;
1792                 INIT_LIST_HEAD(&async_copy->copies);
1793                 refcount_set(&async_copy->refcount, 1);
1794                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1795                 if (!async_copy->cp_src)
1796                         goto out_err;
1797                 if (!nfs4_init_copy_state(nn, copy))
1798                         goto out_err;
1799                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1800                         sizeof(copy->cp_res.cb_stateid));
1801                 dup_copy_fields(copy, async_copy);
1802                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1803                                 async_copy, "%s", "copy thread");
1804                 if (IS_ERR(async_copy->copy_task))
1805                         goto out_err;
1806                 spin_lock(&async_copy->cp_clp->async_lock);
1807                 list_add(&async_copy->copies,
1808                                 &async_copy->cp_clp->async_copies);
1809                 spin_unlock(&async_copy->cp_clp->async_lock);
1810                 wake_up_process(async_copy->copy_task);
1811                 status = nfs_ok;
1812         } else {
1813                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1814                                        copy->nf_dst->nf_file, true);
1815         }
1816 out:
1817         release_copy_files(copy);
1818         return status;
1819 out_err:
1820         if (nfsd4_ssc_is_inter(copy)) {
1821                 /*
1822                  * Source's vfsmount of inter-copy will be unmounted
1823                  * by the laundromat. Use copy instead of async_copy
1824                  * since async_copy->ss_nsui might not be set yet.
1825                  */
1826                 refcount_dec(&copy->ss_nsui->nsui_refcnt);
1827         }
1828         if (async_copy)
1829                 cleanup_async_copy(async_copy);
1830         status = nfserrno(-ENOMEM);
1831         goto out;
1832 }
1833
1834 static struct nfsd4_copy *
1835 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1836 {
1837         struct nfsd4_copy *copy;
1838
1839         lockdep_assert_held(&clp->async_lock);
1840
1841         list_for_each_entry(copy, &clp->async_copies, copies) {
1842                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1843                         continue;
1844                 return copy;
1845         }
1846         return NULL;
1847 }
1848
1849 static struct nfsd4_copy *
1850 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1851 {
1852         struct nfsd4_copy *copy;
1853
1854         spin_lock(&clp->async_lock);
1855         copy = find_async_copy_locked(clp, stateid);
1856         if (copy)
1857                 refcount_inc(&copy->refcount);
1858         spin_unlock(&clp->async_lock);
1859         return copy;
1860 }
1861
1862 static __be32
1863 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1864                      struct nfsd4_compound_state *cstate,
1865                      union nfsd4_op_u *u)
1866 {
1867         struct nfsd4_offload_status *os = &u->offload_status;
1868         struct nfsd4_copy *copy;
1869         struct nfs4_client *clp = cstate->clp;
1870
1871         copy = find_async_copy(clp, &os->stateid);
1872         if (!copy) {
1873                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1874
1875                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1876         } else
1877                 nfsd4_stop_copy(copy);
1878
1879         return nfs_ok;
1880 }
1881
1882 static __be32
1883 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1884                   union nfsd4_op_u *u)
1885 {
1886         struct nfsd4_copy_notify *cn = &u->copy_notify;
1887         __be32 status;
1888         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1889         struct nfs4_stid *stid;
1890         struct nfs4_cpntf_state *cps;
1891         struct nfs4_client *clp = cstate->clp;
1892
1893         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1894                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1895                                         &stid);
1896         if (status)
1897                 return status;
1898
1899         cn->cpn_sec = nn->nfsd4_lease;
1900         cn->cpn_nsec = 0;
1901
1902         status = nfserrno(-ENOMEM);
1903         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1904         if (!cps)
1905                 goto out;
1906         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1907         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1908         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1909
1910         /* For now, only return one server address in cpn_src, the
1911          * address used by the client to connect to this server.
1912          */
1913         cn->cpn_src->nl4_type = NL4_NETADDR;
1914         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1915                                  &cn->cpn_src->u.nl4_addr);
1916         WARN_ON_ONCE(status);
1917         if (status) {
1918                 nfs4_put_cpntf_state(nn, cps);
1919                 goto out;
1920         }
1921 out:
1922         nfs4_put_stid(stid);
1923         return status;
1924 }
1925
1926 static __be32
1927 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1928                 struct nfsd4_fallocate *fallocate, int flags)
1929 {
1930         __be32 status;
1931         struct nfsd_file *nf;
1932
1933         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1934                                             &fallocate->falloc_stateid,
1935                                             WR_STATE, &nf, NULL);
1936         if (status != nfs_ok)
1937                 return status;
1938
1939         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1940                                      fallocate->falloc_offset,
1941                                      fallocate->falloc_length,
1942                                      flags);
1943         nfsd_file_put(nf);
1944         return status;
1945 }
1946
1947 static __be32
1948 nfsd4_offload_status(struct svc_rqst *rqstp,
1949                      struct nfsd4_compound_state *cstate,
1950                      union nfsd4_op_u *u)
1951 {
1952         struct nfsd4_offload_status *os = &u->offload_status;
1953         __be32 status = nfs_ok;
1954         struct nfsd4_copy *copy;
1955         struct nfs4_client *clp = cstate->clp;
1956
1957         spin_lock(&clp->async_lock);
1958         copy = find_async_copy_locked(clp, &os->stateid);
1959         if (copy)
1960                 os->count = copy->cp_res.wr_bytes_written;
1961         else
1962                 status = nfserr_bad_stateid;
1963         spin_unlock(&clp->async_lock);
1964
1965         return status;
1966 }
1967
1968 static __be32
1969 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1970                union nfsd4_op_u *u)
1971 {
1972         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1973 }
1974
1975 static __be32
1976 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1977                  union nfsd4_op_u *u)
1978 {
1979         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1980                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1981 }
1982
1983 static __be32
1984 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1985            union nfsd4_op_u *u)
1986 {
1987         struct nfsd4_seek *seek = &u->seek;
1988         int whence;
1989         __be32 status;
1990         struct nfsd_file *nf;
1991
1992         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1993                                             &seek->seek_stateid,
1994                                             RD_STATE, &nf, NULL);
1995         if (status)
1996                 return status;
1997
1998         switch (seek->seek_whence) {
1999         case NFS4_CONTENT_DATA:
2000                 whence = SEEK_DATA;
2001                 break;
2002         case NFS4_CONTENT_HOLE:
2003                 whence = SEEK_HOLE;
2004                 break;
2005         default:
2006                 status = nfserr_union_notsupp;
2007                 goto out;
2008         }
2009
2010         /*
2011          * Note:  This call does change file->f_pos, but nothing in NFSD
2012          *        should ever file->f_pos.
2013          */
2014         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2015         if (seek->seek_pos < 0)
2016                 status = nfserrno(seek->seek_pos);
2017         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2018                 seek->seek_eof = true;
2019
2020 out:
2021         nfsd_file_put(nf);
2022         return status;
2023 }
2024
2025 /* This routine never returns NFS_OK!  If there are no other errors, it
2026  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2027  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2028  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2029  */
2030 static __be32
2031 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2032              struct nfsd4_verify *verify)
2033 {
2034         __be32 *buf, *p;
2035         int count;
2036         __be32 status;
2037
2038         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2039         if (status)
2040                 return status;
2041
2042         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2043         if (status)
2044                 return status;
2045
2046         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2047             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2048                 return nfserr_inval;
2049         if (verify->ve_attrlen & 3)
2050                 return nfserr_inval;
2051
2052         /* count in words:
2053          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2054          */
2055         count = 4 + (verify->ve_attrlen >> 2);
2056         buf = kmalloc(count << 2, GFP_KERNEL);
2057         if (!buf)
2058                 return nfserr_jukebox;
2059
2060         p = buf;
2061         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2062                                     cstate->current_fh.fh_export,
2063                                     cstate->current_fh.fh_dentry,
2064                                     verify->ve_bmval,
2065                                     rqstp, 0);
2066         /*
2067          * If nfsd4_encode_fattr() ran out of space, assume that's because
2068          * the attributes are longer (hence different) than those given:
2069          */
2070         if (status == nfserr_resource)
2071                 status = nfserr_not_same;
2072         if (status)
2073                 goto out_kfree;
2074
2075         /* skip bitmap */
2076         p = buf + 1 + ntohl(buf[0]);
2077         status = nfserr_not_same;
2078         if (ntohl(*p++) != verify->ve_attrlen)
2079                 goto out_kfree;
2080         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2081                 status = nfserr_same;
2082
2083 out_kfree:
2084         kfree(buf);
2085         return status;
2086 }
2087
2088 static __be32
2089 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2090               union nfsd4_op_u *u)
2091 {
2092         __be32 status;
2093
2094         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2095         return status == nfserr_not_same ? nfs_ok : status;
2096 }
2097
2098 static __be32
2099 nfsd4_verify(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->nverify);
2105         return status == nfserr_same ? nfs_ok : status;
2106 }
2107
2108 #ifdef CONFIG_NFSD_PNFS
2109 static const struct nfsd4_layout_ops *
2110 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2111 {
2112         if (!exp->ex_layout_types) {
2113                 dprintk("%s: export does not support pNFS\n", __func__);
2114                 return NULL;
2115         }
2116
2117         if (layout_type >= LAYOUT_TYPE_MAX ||
2118             !(exp->ex_layout_types & (1 << layout_type))) {
2119                 dprintk("%s: layout type %d not supported\n",
2120                         __func__, layout_type);
2121                 return NULL;
2122         }
2123
2124         return nfsd4_layout_ops[layout_type];
2125 }
2126
2127 static __be32
2128 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2129                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2130 {
2131         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2132         const struct nfsd4_layout_ops *ops;
2133         struct nfsd4_deviceid_map *map;
2134         struct svc_export *exp;
2135         __be32 nfserr;
2136
2137         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2138                __func__,
2139                gdp->gd_layout_type,
2140                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2141                gdp->gd_maxcount);
2142
2143         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2144         if (!map) {
2145                 dprintk("%s: couldn't find device ID to export mapping!\n",
2146                         __func__);
2147                 return nfserr_noent;
2148         }
2149
2150         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2151         if (IS_ERR(exp)) {
2152                 dprintk("%s: could not find device id\n", __func__);
2153                 return nfserr_noent;
2154         }
2155
2156         nfserr = nfserr_layoutunavailable;
2157         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2158         if (!ops)
2159                 goto out;
2160
2161         nfserr = nfs_ok;
2162         if (gdp->gd_maxcount != 0) {
2163                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2164                                 rqstp, cstate->clp, gdp);
2165         }
2166
2167         gdp->gd_notify_types &= ops->notify_types;
2168 out:
2169         exp_put(exp);
2170         return nfserr;
2171 }
2172
2173 static void
2174 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2175 {
2176         kfree(u->getdeviceinfo.gd_device);
2177 }
2178
2179 static __be32
2180 nfsd4_layoutget(struct svc_rqst *rqstp,
2181                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2182 {
2183         struct nfsd4_layoutget *lgp = &u->layoutget;
2184         struct svc_fh *current_fh = &cstate->current_fh;
2185         const struct nfsd4_layout_ops *ops;
2186         struct nfs4_layout_stateid *ls;
2187         __be32 nfserr;
2188         int accmode = NFSD_MAY_READ_IF_EXEC;
2189
2190         switch (lgp->lg_seg.iomode) {
2191         case IOMODE_READ:
2192                 accmode |= NFSD_MAY_READ;
2193                 break;
2194         case IOMODE_RW:
2195                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2196                 break;
2197         default:
2198                 dprintk("%s: invalid iomode %d\n",
2199                         __func__, lgp->lg_seg.iomode);
2200                 nfserr = nfserr_badiomode;
2201                 goto out;
2202         }
2203
2204         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2205         if (nfserr)
2206                 goto out;
2207
2208         nfserr = nfserr_layoutunavailable;
2209         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2210         if (!ops)
2211                 goto out;
2212
2213         /*
2214          * Verify minlength and range as per RFC5661:
2215          *  o  If loga_length is less than loga_minlength,
2216          *     the metadata server MUST return NFS4ERR_INVAL.
2217          *  o  If the sum of loga_offset and loga_minlength exceeds
2218          *     NFS4_UINT64_MAX, and loga_minlength is not
2219          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2220          *  o  If the sum of loga_offset and loga_length exceeds
2221          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2222          *     the error NFS4ERR_INVAL MUST result.
2223          */
2224         nfserr = nfserr_inval;
2225         if (lgp->lg_seg.length < lgp->lg_minlength ||
2226             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2227              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2228             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2229              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2230                 goto out;
2231         if (lgp->lg_seg.length == 0)
2232                 goto out;
2233
2234         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2235                                                 true, lgp->lg_layout_type, &ls);
2236         if (nfserr) {
2237                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2238                 goto out;
2239         }
2240
2241         nfserr = nfserr_recallconflict;
2242         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2243                 goto out_put_stid;
2244
2245         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2246                                      current_fh, lgp);
2247         if (nfserr)
2248                 goto out_put_stid;
2249
2250         nfserr = nfsd4_insert_layout(lgp, ls);
2251
2252 out_put_stid:
2253         mutex_unlock(&ls->ls_mutex);
2254         nfs4_put_stid(&ls->ls_stid);
2255 out:
2256         return nfserr;
2257 }
2258
2259 static void
2260 nfsd4_layoutget_release(union nfsd4_op_u *u)
2261 {
2262         kfree(u->layoutget.lg_content);
2263 }
2264
2265 static __be32
2266 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2267                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2268 {
2269         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2270         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2271         struct svc_fh *current_fh = &cstate->current_fh;
2272         const struct nfsd4_layout_ops *ops;
2273         loff_t new_size = lcp->lc_last_wr + 1;
2274         struct inode *inode;
2275         struct nfs4_layout_stateid *ls;
2276         __be32 nfserr;
2277
2278         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2279         if (nfserr)
2280                 goto out;
2281
2282         nfserr = nfserr_layoutunavailable;
2283         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2284         if (!ops)
2285                 goto out;
2286         inode = d_inode(current_fh->fh_dentry);
2287
2288         nfserr = nfserr_inval;
2289         if (new_size <= seg->offset) {
2290                 dprintk("pnfsd: last write before layout segment\n");
2291                 goto out;
2292         }
2293         if (new_size > seg->offset + seg->length) {
2294                 dprintk("pnfsd: last write beyond layout segment\n");
2295                 goto out;
2296         }
2297         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2298                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2299                 goto out;
2300         }
2301
2302         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2303                                                 false, lcp->lc_layout_type,
2304                                                 &ls);
2305         if (nfserr) {
2306                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2307                 /* fixup error code as per RFC5661 */
2308                 if (nfserr == nfserr_bad_stateid)
2309                         nfserr = nfserr_badlayout;
2310                 goto out;
2311         }
2312
2313         /* LAYOUTCOMMIT does not require any serialization */
2314         mutex_unlock(&ls->ls_mutex);
2315
2316         if (new_size > i_size_read(inode)) {
2317                 lcp->lc_size_chg = 1;
2318                 lcp->lc_newsize = new_size;
2319         } else {
2320                 lcp->lc_size_chg = 0;
2321         }
2322
2323         nfserr = ops->proc_layoutcommit(inode, lcp);
2324         nfs4_put_stid(&ls->ls_stid);
2325 out:
2326         return nfserr;
2327 }
2328
2329 static __be32
2330 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2331                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2332 {
2333         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2334         struct svc_fh *current_fh = &cstate->current_fh;
2335         __be32 nfserr;
2336
2337         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2338         if (nfserr)
2339                 goto out;
2340
2341         nfserr = nfserr_layoutunavailable;
2342         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2343                 goto out;
2344
2345         switch (lrp->lr_seg.iomode) {
2346         case IOMODE_READ:
2347         case IOMODE_RW:
2348         case IOMODE_ANY:
2349                 break;
2350         default:
2351                 dprintk("%s: invalid iomode %d\n", __func__,
2352                         lrp->lr_seg.iomode);
2353                 nfserr = nfserr_inval;
2354                 goto out;
2355         }
2356
2357         switch (lrp->lr_return_type) {
2358         case RETURN_FILE:
2359                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2360                 break;
2361         case RETURN_FSID:
2362         case RETURN_ALL:
2363                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2364                 break;
2365         default:
2366                 dprintk("%s: invalid return_type %d\n", __func__,
2367                         lrp->lr_return_type);
2368                 nfserr = nfserr_inval;
2369                 break;
2370         }
2371 out:
2372         return nfserr;
2373 }
2374 #endif /* CONFIG_NFSD_PNFS */
2375
2376 static __be32
2377 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2378                union nfsd4_op_u *u)
2379 {
2380         struct nfsd4_getxattr *getxattr = &u->getxattr;
2381
2382         return nfsd_getxattr(rqstp, &cstate->current_fh,
2383                              getxattr->getxa_name, &getxattr->getxa_buf,
2384                              &getxattr->getxa_len);
2385 }
2386
2387 static __be32
2388 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2389            union nfsd4_op_u *u)
2390 {
2391         struct nfsd4_setxattr *setxattr = &u->setxattr;
2392         __be32 ret;
2393
2394         if (opens_in_grace(SVC_NET(rqstp)))
2395                 return nfserr_grace;
2396
2397         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2398                             setxattr->setxa_buf, setxattr->setxa_len,
2399                             setxattr->setxa_flags);
2400
2401         if (!ret)
2402                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2403
2404         return ret;
2405 }
2406
2407 static __be32
2408 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2409            union nfsd4_op_u *u)
2410 {
2411         /*
2412          * Get the entire list, then copy out only the user attributes
2413          * in the encode function.
2414          */
2415         return nfsd_listxattr(rqstp, &cstate->current_fh,
2416                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2417 }
2418
2419 static __be32
2420 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2421            union nfsd4_op_u *u)
2422 {
2423         struct nfsd4_removexattr *removexattr = &u->removexattr;
2424         __be32 ret;
2425
2426         if (opens_in_grace(SVC_NET(rqstp)))
2427                 return nfserr_grace;
2428
2429         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2430             removexattr->rmxa_name);
2431
2432         if (!ret)
2433                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2434
2435         return ret;
2436 }
2437
2438 /*
2439  * NULL call.
2440  */
2441 static __be32
2442 nfsd4_proc_null(struct svc_rqst *rqstp)
2443 {
2444         return rpc_success;
2445 }
2446
2447 static inline void nfsd4_increment_op_stats(u32 opnum)
2448 {
2449         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2450                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2451 }
2452
2453 static const struct nfsd4_operation nfsd4_ops[];
2454
2455 static const char *nfsd4_op_name(unsigned opnum);
2456
2457 /*
2458  * Enforce NFSv4.1 COMPOUND ordering rules:
2459  *
2460  * Also note, enforced elsewhere:
2461  *      - SEQUENCE other than as first op results in
2462  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2463  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2464  *        (Enforced in nfsd4_bind_conn_to_session().)
2465  *      - DESTROY_SESSION must be the final operation in a compound, if
2466  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2467  *        (Enforced in nfsd4_destroy_session().)
2468  */
2469 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2470 {
2471         struct nfsd4_op *first_op = &args->ops[0];
2472
2473         /* These ordering requirements don't apply to NFSv4.0: */
2474         if (args->minorversion == 0)
2475                 return nfs_ok;
2476         /* This is weird, but OK, not our problem: */
2477         if (args->opcnt == 0)
2478                 return nfs_ok;
2479         if (first_op->status == nfserr_op_illegal)
2480                 return nfs_ok;
2481         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2482                 return nfserr_op_not_in_session;
2483         if (first_op->opnum == OP_SEQUENCE)
2484                 return nfs_ok;
2485         /*
2486          * So first_op is something allowed outside a session, like
2487          * EXCHANGE_ID; but then it has to be the only op in the
2488          * compound:
2489          */
2490         if (args->opcnt != 1)
2491                 return nfserr_not_only_op;
2492         return nfs_ok;
2493 }
2494
2495 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2496 {
2497         return &nfsd4_ops[op->opnum];
2498 }
2499
2500 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2501 {
2502         if (op->opnum == OP_ILLEGAL)
2503                 return false;
2504         return OPDESC(op)->op_flags & OP_CACHEME;
2505 }
2506
2507 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2508 {
2509         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2510         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2511         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2512         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2513         const struct nfsd4_operation *thisd = OPDESC(this);
2514         const struct nfsd4_operation *nextd;
2515
2516         /*
2517          * Most ops check wronsec on our own; only the putfh-like ops
2518          * have special rules.
2519          */
2520         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2521                 return false;
2522         /*
2523          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2524          * put-filehandle operation if we're not going to use the
2525          * result:
2526          */
2527         if (argp->opcnt == resp->opcnt)
2528                 return false;
2529         if (next->opnum == OP_ILLEGAL)
2530                 return false;
2531         nextd = OPDESC(next);
2532         /*
2533          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2534          * errors themselves as necessary; others should check for them
2535          * now:
2536          */
2537         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2538 }
2539
2540 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2541 static void
2542 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2543 {
2544         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2545         struct nfsd4_copy *copy;
2546         struct nfsd4_putfh *putfh;
2547         int i;
2548
2549         /* traverse all operation and if it's a COPY compound, mark the
2550          * source filehandle to skip verification
2551          */
2552         for (i = 0; i < args->opcnt; i++) {
2553                 op = &args->ops[i];
2554                 if (op->opnum == OP_PUTFH)
2555                         current_op = op;
2556                 else if (op->opnum == OP_SAVEFH)
2557                         saved_op = current_op;
2558                 else if (op->opnum == OP_RESTOREFH)
2559                         current_op = saved_op;
2560                 else if (op->opnum == OP_COPY) {
2561                         copy = (struct nfsd4_copy *)&op->u;
2562                         if (!saved_op) {
2563                                 op->status = nfserr_nofilehandle;
2564                                 return;
2565                         }
2566                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2567                         if (nfsd4_ssc_is_inter(copy))
2568                                 putfh->no_verify = true;
2569                 }
2570         }
2571 }
2572 #else
2573 static void
2574 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2575 {
2576 }
2577 #endif
2578
2579 /*
2580  * COMPOUND call.
2581  */
2582 static __be32
2583 nfsd4_proc_compound(struct svc_rqst *rqstp)
2584 {
2585         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2586         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2587         struct nfsd4_op *op;
2588         struct nfsd4_compound_state *cstate = &resp->cstate;
2589         struct svc_fh *current_fh = &cstate->current_fh;
2590         struct svc_fh *save_fh = &cstate->save_fh;
2591         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2592         __be32          status;
2593
2594         resp->xdr = &rqstp->rq_res_stream;
2595         resp->statusp = resp->xdr->p;
2596
2597         /* reserve space for: NFS status code */
2598         xdr_reserve_space(resp->xdr, XDR_UNIT);
2599
2600         /* reserve space for: taglen, tag, and opcnt */
2601         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2602         resp->taglen = args->taglen;
2603         resp->tag = args->tag;
2604         resp->rqstp = rqstp;
2605         cstate->minorversion = args->minorversion;
2606         fh_init(current_fh, NFS4_FHSIZE);
2607         fh_init(save_fh, NFS4_FHSIZE);
2608         /*
2609          * Don't use the deferral mechanism for NFSv4; compounds make it
2610          * too hard to avoid non-idempotency problems.
2611          */
2612         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2613
2614         /*
2615          * According to RFC3010, this takes precedence over all other errors.
2616          */
2617         status = nfserr_minor_vers_mismatch;
2618         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2619                 goto out;
2620
2621         status = nfs41_check_op_ordering(args);
2622         if (status) {
2623                 op = &args->ops[0];
2624                 op->status = status;
2625                 resp->opcnt = 1;
2626                 goto encode_op;
2627         }
2628         check_if_stalefh_allowed(args);
2629
2630         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2631
2632         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2633         while (!status && resp->opcnt < args->opcnt) {
2634                 op = &args->ops[resp->opcnt++];
2635
2636                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2637                         /* If there are still more operations to process,
2638                          * stop here and report NFS4ERR_RESOURCE. */
2639                         if (cstate->minorversion == 0 &&
2640                             args->client_opcnt > resp->opcnt) {
2641                                 op->status = nfserr_resource;
2642                                 goto encode_op;
2643                         }
2644                 }
2645
2646                 /*
2647                  * The XDR decode routines may have pre-set op->status;
2648                  * for example, if there is a miscellaneous XDR error
2649                  * it will be set to nfserr_bad_xdr.
2650                  */
2651                 if (op->status) {
2652                         if (op->opnum == OP_OPEN)
2653                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2654                         goto encode_op;
2655                 }
2656                 if (!current_fh->fh_dentry &&
2657                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2658                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2659                                 op->status = nfserr_nofilehandle;
2660                                 goto encode_op;
2661                         }
2662                 } else if (current_fh->fh_export &&
2663                            current_fh->fh_export->ex_fslocs.migrated &&
2664                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2665                         op->status = nfserr_moved;
2666                         goto encode_op;
2667                 }
2668
2669                 fh_clear_pre_post_attrs(current_fh);
2670
2671                 /* If op is non-idempotent */
2672                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2673                         /*
2674                          * Don't execute this op if we couldn't encode a
2675                          * successful reply:
2676                          */
2677                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2678                         /*
2679                          * Plus if there's another operation, make sure
2680                          * we'll have space to at least encode an error:
2681                          */
2682                         if (resp->opcnt < args->opcnt)
2683                                 plen += COMPOUND_ERR_SLACK_SPACE;
2684                         op->status = nfsd4_check_resp_size(resp, plen);
2685                 }
2686
2687                 if (op->status)
2688                         goto encode_op;
2689
2690                 if (op->opdesc->op_get_currentstateid)
2691                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2692                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2693
2694                 /* Only from SEQUENCE */
2695                 if (cstate->status == nfserr_replay_cache) {
2696                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2697                         status = op->status;
2698                         goto out;
2699                 }
2700                 if (!op->status) {
2701                         if (op->opdesc->op_set_currentstateid)
2702                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2703
2704                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2705                                 clear_current_stateid(cstate);
2706
2707                         if (current_fh->fh_export &&
2708                                         need_wrongsec_check(rqstp))
2709                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2710                 }
2711 encode_op:
2712                 if (op->status == nfserr_replay_me) {
2713                         op->replay = &cstate->replay_owner->so_replay;
2714                         nfsd4_encode_replay(resp->xdr, op);
2715                         status = op->status = op->replay->rp_status;
2716                 } else {
2717                         nfsd4_encode_operation(resp, op);
2718                         status = op->status;
2719                 }
2720
2721                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2722                                            status, nfsd4_op_name(op->opnum));
2723
2724                 nfsd4_cstate_clear_replay(cstate);
2725                 nfsd4_increment_op_stats(op->opnum);
2726         }
2727
2728         fh_put(current_fh);
2729         fh_put(save_fh);
2730         BUG_ON(cstate->replay_owner);
2731 out:
2732         cstate->status = status;
2733         /* Reset deferral mechanism for RPC deferrals */
2734         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2735         return rpc_success;
2736 }
2737
2738 #define op_encode_hdr_size              (2)
2739 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2740 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2741 #define op_encode_change_info_maxsz     (5)
2742 #define nfs4_fattr_bitmap_maxsz         (4)
2743
2744 /* We'll fall back on returning no lockowner if run out of space: */
2745 #define op_encode_lockowner_maxsz       (0)
2746 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2747
2748 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2749
2750 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2751 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2752                                          op_encode_ace_maxsz)
2753
2754 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2755
2756 /*
2757  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2758  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2759  * the maximum payload size here, based on transport limits and the size
2760  * of the remaining space in the rq_pages array.
2761  */
2762 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2763 {
2764         u32 buflen;
2765
2766         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2767         buflen -= rqstp->rq_auth_slack;
2768         buflen -= rqstp->rq_res.head[0].iov_len;
2769         return min_t(u32, buflen, svc_max_payload(rqstp));
2770 }
2771
2772 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2773                                    const struct nfsd4_op *op)
2774 {
2775         return (op_encode_hdr_size) * sizeof(__be32);
2776 }
2777
2778 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2779                                       const struct nfsd4_op *op)
2780 {
2781         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2782 }
2783
2784 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2785                               const struct nfsd4_op *op)
2786 {
2787         /* ac_supported, ac_resp_access */
2788         return (op_encode_hdr_size + 2)* sizeof(__be32);
2789 }
2790
2791 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2792                               const struct nfsd4_op *op)
2793 {
2794         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2795 }
2796
2797 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2798                               const struct nfsd4_op *op)
2799 {
2800         return (op_encode_hdr_size + op_encode_change_info_maxsz
2801                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2802 }
2803
2804 /*
2805  * Note since this is an idempotent operation we won't insist on failing
2806  * the op prematurely if the estimate is too large.  We may turn off splice
2807  * reads unnecessarily.
2808  */
2809 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2810                                const struct nfsd4_op *op)
2811 {
2812         const u32 *bmap = op->u.getattr.ga_bmval;
2813         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2814         u32 ret = 0;
2815
2816         if (bmap0 & FATTR4_WORD0_ACL)
2817                 return nfsd4_max_payload(rqstp);
2818         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2819                 return nfsd4_max_payload(rqstp);
2820
2821         if (bmap1 & FATTR4_WORD1_OWNER) {
2822                 ret += IDMAP_NAMESZ + 4;
2823                 bmap1 &= ~FATTR4_WORD1_OWNER;
2824         }
2825         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2826                 ret += IDMAP_NAMESZ + 4;
2827                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2828         }
2829         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2830                 ret += NFS4_FHSIZE + 4;
2831                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2832         }
2833         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2834                 ret += NFS4_MAXLABELLEN + 12;
2835                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2836         }
2837         /*
2838          * Largest of remaining attributes are 16 bytes (e.g.,
2839          * supported_attributes)
2840          */
2841         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2842         /* bitmask, length */
2843         ret += 20;
2844         return ret;
2845 }
2846
2847 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2848                              const struct nfsd4_op *op)
2849 {
2850         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2851 }
2852
2853 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2854                             const struct nfsd4_op *op)
2855 {
2856         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2857                 * sizeof(__be32);
2858 }
2859
2860 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2861                             const struct nfsd4_op *op)
2862 {
2863         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2864                 * sizeof(__be32);
2865 }
2866
2867 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2868                             const struct nfsd4_op *op)
2869 {
2870         return (op_encode_hdr_size + op_encode_stateid_maxsz
2871                 + op_encode_change_info_maxsz + 1
2872                 + nfs4_fattr_bitmap_maxsz
2873                 + op_encode_delegation_maxsz) * sizeof(__be32);
2874 }
2875
2876 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2877                             const struct nfsd4_op *op)
2878 {
2879         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2880
2881         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2882 }
2883
2884 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2885                                  const struct nfsd4_op *op)
2886 {
2887         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2888         /*
2889          * If we detect that the file changed during hole encoding, then we
2890          * recover by encoding the remaining reply as data. This means we need
2891          * to set aside enough room to encode two data segments.
2892          */
2893         u32 seg_len = 2 * (1 + 2 + 1);
2894
2895         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2896 }
2897
2898 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2899                                const struct nfsd4_op *op)
2900 {
2901         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2902
2903         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2904                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2905 }
2906
2907 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2908                                 const struct nfsd4_op *op)
2909 {
2910         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2911 }
2912
2913 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2914                               const struct nfsd4_op *op)
2915 {
2916         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2917                 * sizeof(__be32);
2918 }
2919
2920 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2921                               const struct nfsd4_op *op)
2922 {
2923         return (op_encode_hdr_size + op_encode_change_info_maxsz
2924                 + op_encode_change_info_maxsz) * sizeof(__be32);
2925 }
2926
2927 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2928                                 const struct nfsd4_op *op)
2929 {
2930         return (op_encode_hdr_size
2931                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2932 }
2933
2934 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2935                                     const struct nfsd4_op *op)
2936 {
2937         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2938                 * sizeof(__be32);
2939 }
2940
2941 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2942                                const struct nfsd4_op *op)
2943 {
2944         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2945 }
2946
2947 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2948                                const struct nfsd4_op *op)
2949 {
2950         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2951                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2952 }
2953
2954 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2955                                    const struct nfsd4_op *op)
2956 {
2957         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2958                                                                 sizeof(__be32);
2959 }
2960
2961 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
2962                              const struct nfsd4_op *op)
2963 {
2964         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2965 }
2966
2967 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
2968                                    const struct nfsd4_op *op)
2969 {
2970         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2971                 1 + 1 + /* eir_flags, spr_how */\
2972                 4 + /* spo_must_enforce & _allow with bitmap */\
2973                 2 + /*eir_server_owner.so_minor_id */\
2974                 /* eir_server_owner.so_major_id<> */\
2975                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2976                 /* eir_server_scope<> */\
2977                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2978                 1 + /* eir_server_impl_id array length */\
2979                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2980 }
2981
2982 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
2983                                             const struct nfsd4_op *op)
2984 {
2985         return (op_encode_hdr_size + \
2986                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2987                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2988 }
2989
2990 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
2991                                       const struct nfsd4_op *op)
2992 {
2993         return (op_encode_hdr_size + \
2994                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2995                 2 + /* csr_sequence, csr_flags */\
2996                 op_encode_channel_attrs_maxsz + \
2997                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2998 }
2999
3000 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3001                             const struct nfsd4_op *op)
3002 {
3003         return (op_encode_hdr_size +
3004                 1 /* wr_callback */ +
3005                 op_encode_stateid_maxsz /* wr_callback */ +
3006                 2 /* wr_count */ +
3007                 1 /* wr_committed */ +
3008                 op_encode_verifier_maxsz +
3009                 1 /* cr_consecutive */ +
3010                 1 /* cr_synchronous */) * sizeof(__be32);
3011 }
3012
3013 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3014                                       const struct nfsd4_op *op)
3015 {
3016         return (op_encode_hdr_size +
3017                 2 /* osr_count */ +
3018                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3019 }
3020
3021 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3022                                    const struct nfsd4_op *op)
3023 {
3024         return (op_encode_hdr_size +
3025                 3 /* cnr_lease_time */ +
3026                 1 /* We support one cnr_source_server */ +
3027                 1 /* cnr_stateid seq */ +
3028                 op_encode_stateid_maxsz /* cnr_stateid */ +
3029                 1 /* num cnr_source_server*/ +
3030                 1 /* nl4_type */ +
3031                 1 /* nl4 size */ +
3032                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3033                 * sizeof(__be32);
3034 }
3035
3036 #ifdef CONFIG_NFSD_PNFS
3037 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3038                                      const struct nfsd4_op *op)
3039 {
3040         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3041
3042         return (op_encode_hdr_size +
3043                 1 /* gd_layout_type*/ +
3044                 XDR_QUADLEN(rlen) +
3045                 2 /* gd_notify_types */) * sizeof(__be32);
3046 }
3047
3048 /*
3049  * At this stage we don't really know what layout driver will handle the request,
3050  * so we need to define an arbitrary upper bound here.
3051  */
3052 #define MAX_LAYOUT_SIZE         128
3053 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3054                                  const struct nfsd4_op *op)
3055 {
3056         return (op_encode_hdr_size +
3057                 1 /* logr_return_on_close */ +
3058                 op_encode_stateid_maxsz +
3059                 1 /* nr of layouts */ +
3060                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3061 }
3062
3063 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3064                                     const struct nfsd4_op *op)
3065 {
3066         return (op_encode_hdr_size +
3067                 1 /* locr_newsize */ +
3068                 2 /* ns_size */) * sizeof(__be32);
3069 }
3070
3071 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3072                                     const struct nfsd4_op *op)
3073 {
3074         return (op_encode_hdr_size +
3075                 1 /* lrs_stateid */ +
3076                 op_encode_stateid_maxsz) * sizeof(__be32);
3077 }
3078 #endif /* CONFIG_NFSD_PNFS */
3079
3080
3081 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3082                             const struct nfsd4_op *op)
3083 {
3084         return (op_encode_hdr_size + 3) * sizeof(__be32);
3085 }
3086
3087 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3088                                 const struct nfsd4_op *op)
3089 {
3090         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3091
3092         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3093 }
3094
3095 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3096                                 const struct nfsd4_op *op)
3097 {
3098         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3099                 * sizeof(__be32);
3100 }
3101 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3102                                   const struct nfsd4_op *op)
3103 {
3104         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3105
3106         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3107 }
3108
3109 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3110                                    const struct nfsd4_op *op)
3111 {
3112         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3113                 * sizeof(__be32);
3114 }
3115
3116
3117 static const struct nfsd4_operation nfsd4_ops[] = {
3118         [OP_ACCESS] = {
3119                 .op_func = nfsd4_access,
3120                 .op_name = "OP_ACCESS",
3121                 .op_rsize_bop = nfsd4_access_rsize,
3122         },
3123         [OP_CLOSE] = {
3124                 .op_func = nfsd4_close,
3125                 .op_flags = OP_MODIFIES_SOMETHING,
3126                 .op_name = "OP_CLOSE",
3127                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3128                 .op_get_currentstateid = nfsd4_get_closestateid,
3129                 .op_set_currentstateid = nfsd4_set_closestateid,
3130         },
3131         [OP_COMMIT] = {
3132                 .op_func = nfsd4_commit,
3133                 .op_flags = OP_MODIFIES_SOMETHING,
3134                 .op_name = "OP_COMMIT",
3135                 .op_rsize_bop = nfsd4_commit_rsize,
3136         },
3137         [OP_CREATE] = {
3138                 .op_func = nfsd4_create,
3139                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3140                 .op_name = "OP_CREATE",
3141                 .op_rsize_bop = nfsd4_create_rsize,
3142         },
3143         [OP_DELEGRETURN] = {
3144                 .op_func = nfsd4_delegreturn,
3145                 .op_flags = OP_MODIFIES_SOMETHING,
3146                 .op_name = "OP_DELEGRETURN",
3147                 .op_rsize_bop = nfsd4_only_status_rsize,
3148                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3149         },
3150         [OP_GETATTR] = {
3151                 .op_func = nfsd4_getattr,
3152                 .op_flags = ALLOWED_ON_ABSENT_FS,
3153                 .op_rsize_bop = nfsd4_getattr_rsize,
3154                 .op_name = "OP_GETATTR",
3155         },
3156         [OP_GETFH] = {
3157                 .op_func = nfsd4_getfh,
3158                 .op_name = "OP_GETFH",
3159                 .op_rsize_bop = nfsd4_getfh_rsize,
3160         },
3161         [OP_LINK] = {
3162                 .op_func = nfsd4_link,
3163                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3164                                 | OP_CACHEME,
3165                 .op_name = "OP_LINK",
3166                 .op_rsize_bop = nfsd4_link_rsize,
3167         },
3168         [OP_LOCK] = {
3169                 .op_func = nfsd4_lock,
3170                 .op_flags = OP_MODIFIES_SOMETHING |
3171                                 OP_NONTRIVIAL_ERROR_ENCODE,
3172                 .op_name = "OP_LOCK",
3173                 .op_rsize_bop = nfsd4_lock_rsize,
3174                 .op_set_currentstateid = nfsd4_set_lockstateid,
3175         },
3176         [OP_LOCKT] = {
3177                 .op_func = nfsd4_lockt,
3178                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3179                 .op_name = "OP_LOCKT",
3180                 .op_rsize_bop = nfsd4_lock_rsize,
3181         },
3182         [OP_LOCKU] = {
3183                 .op_func = nfsd4_locku,
3184                 .op_flags = OP_MODIFIES_SOMETHING,
3185                 .op_name = "OP_LOCKU",
3186                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3187                 .op_get_currentstateid = nfsd4_get_lockustateid,
3188         },
3189         [OP_LOOKUP] = {
3190                 .op_func = nfsd4_lookup,
3191                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3192                 .op_name = "OP_LOOKUP",
3193                 .op_rsize_bop = nfsd4_only_status_rsize,
3194         },
3195         [OP_LOOKUPP] = {
3196                 .op_func = nfsd4_lookupp,
3197                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3198                 .op_name = "OP_LOOKUPP",
3199                 .op_rsize_bop = nfsd4_only_status_rsize,
3200         },
3201         [OP_NVERIFY] = {
3202                 .op_func = nfsd4_nverify,
3203                 .op_name = "OP_NVERIFY",
3204                 .op_rsize_bop = nfsd4_only_status_rsize,
3205         },
3206         [OP_OPEN] = {
3207                 .op_func = nfsd4_open,
3208                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3209                 .op_name = "OP_OPEN",
3210                 .op_rsize_bop = nfsd4_open_rsize,
3211                 .op_set_currentstateid = nfsd4_set_openstateid,
3212         },
3213         [OP_OPEN_CONFIRM] = {
3214                 .op_func = nfsd4_open_confirm,
3215                 .op_flags = OP_MODIFIES_SOMETHING,
3216                 .op_name = "OP_OPEN_CONFIRM",
3217                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3218         },
3219         [OP_OPEN_DOWNGRADE] = {
3220                 .op_func = nfsd4_open_downgrade,
3221                 .op_flags = OP_MODIFIES_SOMETHING,
3222                 .op_name = "OP_OPEN_DOWNGRADE",
3223                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3224                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3225                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3226         },
3227         [OP_PUTFH] = {
3228                 .op_func = nfsd4_putfh,
3229                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3230                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3231                 .op_name = "OP_PUTFH",
3232                 .op_rsize_bop = nfsd4_only_status_rsize,
3233         },
3234         [OP_PUTPUBFH] = {
3235                 .op_func = nfsd4_putrootfh,
3236                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3237                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3238                 .op_name = "OP_PUTPUBFH",
3239                 .op_rsize_bop = nfsd4_only_status_rsize,
3240         },
3241         [OP_PUTROOTFH] = {
3242                 .op_func = nfsd4_putrootfh,
3243                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3244                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3245                 .op_name = "OP_PUTROOTFH",
3246                 .op_rsize_bop = nfsd4_only_status_rsize,
3247         },
3248         [OP_READ] = {
3249                 .op_func = nfsd4_read,
3250                 .op_release = nfsd4_read_release,
3251                 .op_name = "OP_READ",
3252                 .op_rsize_bop = nfsd4_read_rsize,
3253                 .op_get_currentstateid = nfsd4_get_readstateid,
3254         },
3255         [OP_READDIR] = {
3256                 .op_func = nfsd4_readdir,
3257                 .op_name = "OP_READDIR",
3258                 .op_rsize_bop = nfsd4_readdir_rsize,
3259         },
3260         [OP_READLINK] = {
3261                 .op_func = nfsd4_readlink,
3262                 .op_name = "OP_READLINK",
3263                 .op_rsize_bop = nfsd4_readlink_rsize,
3264         },
3265         [OP_REMOVE] = {
3266                 .op_func = nfsd4_remove,
3267                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3268                 .op_name = "OP_REMOVE",
3269                 .op_rsize_bop = nfsd4_remove_rsize,
3270         },
3271         [OP_RENAME] = {
3272                 .op_func = nfsd4_rename,
3273                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3274                 .op_name = "OP_RENAME",
3275                 .op_rsize_bop = nfsd4_rename_rsize,
3276         },
3277         [OP_RENEW] = {
3278                 .op_func = nfsd4_renew,
3279                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3280                                 | OP_MODIFIES_SOMETHING,
3281                 .op_name = "OP_RENEW",
3282                 .op_rsize_bop = nfsd4_only_status_rsize,
3283
3284         },
3285         [OP_RESTOREFH] = {
3286                 .op_func = nfsd4_restorefh,
3287                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3288                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3289                 .op_name = "OP_RESTOREFH",
3290                 .op_rsize_bop = nfsd4_only_status_rsize,
3291         },
3292         [OP_SAVEFH] = {
3293                 .op_func = nfsd4_savefh,
3294                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3295                 .op_name = "OP_SAVEFH",
3296                 .op_rsize_bop = nfsd4_only_status_rsize,
3297         },
3298         [OP_SECINFO] = {
3299                 .op_func = nfsd4_secinfo,
3300                 .op_release = nfsd4_secinfo_release,
3301                 .op_flags = OP_HANDLES_WRONGSEC,
3302                 .op_name = "OP_SECINFO",
3303                 .op_rsize_bop = nfsd4_secinfo_rsize,
3304         },
3305         [OP_SETATTR] = {
3306                 .op_func = nfsd4_setattr,
3307                 .op_name = "OP_SETATTR",
3308                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3309                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3310                 .op_rsize_bop = nfsd4_setattr_rsize,
3311                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3312         },
3313         [OP_SETCLIENTID] = {
3314                 .op_func = nfsd4_setclientid,
3315                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3316                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3317                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3318                 .op_name = "OP_SETCLIENTID",
3319                 .op_rsize_bop = nfsd4_setclientid_rsize,
3320         },
3321         [OP_SETCLIENTID_CONFIRM] = {
3322                 .op_func = nfsd4_setclientid_confirm,
3323                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3324                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3325                 .op_name = "OP_SETCLIENTID_CONFIRM",
3326                 .op_rsize_bop = nfsd4_only_status_rsize,
3327         },
3328         [OP_VERIFY] = {
3329                 .op_func = nfsd4_verify,
3330                 .op_name = "OP_VERIFY",
3331                 .op_rsize_bop = nfsd4_only_status_rsize,
3332         },
3333         [OP_WRITE] = {
3334                 .op_func = nfsd4_write,
3335                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3336                 .op_name = "OP_WRITE",
3337                 .op_rsize_bop = nfsd4_write_rsize,
3338                 .op_get_currentstateid = nfsd4_get_writestateid,
3339         },
3340         [OP_RELEASE_LOCKOWNER] = {
3341                 .op_func = nfsd4_release_lockowner,
3342                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3343                                 | OP_MODIFIES_SOMETHING,
3344                 .op_name = "OP_RELEASE_LOCKOWNER",
3345                 .op_rsize_bop = nfsd4_only_status_rsize,
3346         },
3347
3348         /* NFSv4.1 operations */
3349         [OP_EXCHANGE_ID] = {
3350                 .op_func = nfsd4_exchange_id,
3351                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3352                                 | OP_MODIFIES_SOMETHING,
3353                 .op_name = "OP_EXCHANGE_ID",
3354                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3355         },
3356         [OP_BACKCHANNEL_CTL] = {
3357                 .op_func = nfsd4_backchannel_ctl,
3358                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3359                 .op_name = "OP_BACKCHANNEL_CTL",
3360                 .op_rsize_bop = nfsd4_only_status_rsize,
3361         },
3362         [OP_BIND_CONN_TO_SESSION] = {
3363                 .op_func = nfsd4_bind_conn_to_session,
3364                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3365                                 | OP_MODIFIES_SOMETHING,
3366                 .op_name = "OP_BIND_CONN_TO_SESSION",
3367                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3368         },
3369         [OP_CREATE_SESSION] = {
3370                 .op_func = nfsd4_create_session,
3371                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3372                                 | OP_MODIFIES_SOMETHING,
3373                 .op_name = "OP_CREATE_SESSION",
3374                 .op_rsize_bop = nfsd4_create_session_rsize,
3375         },
3376         [OP_DESTROY_SESSION] = {
3377                 .op_func = nfsd4_destroy_session,
3378                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3379                                 | OP_MODIFIES_SOMETHING,
3380                 .op_name = "OP_DESTROY_SESSION",
3381                 .op_rsize_bop = nfsd4_only_status_rsize,
3382         },
3383         [OP_SEQUENCE] = {
3384                 .op_func = nfsd4_sequence,
3385                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3386                 .op_name = "OP_SEQUENCE",
3387                 .op_rsize_bop = nfsd4_sequence_rsize,
3388         },
3389         [OP_DESTROY_CLIENTID] = {
3390                 .op_func = nfsd4_destroy_clientid,
3391                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3392                                 | OP_MODIFIES_SOMETHING,
3393                 .op_name = "OP_DESTROY_CLIENTID",
3394                 .op_rsize_bop = nfsd4_only_status_rsize,
3395         },
3396         [OP_RECLAIM_COMPLETE] = {
3397                 .op_func = nfsd4_reclaim_complete,
3398                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3399                 .op_name = "OP_RECLAIM_COMPLETE",
3400                 .op_rsize_bop = nfsd4_only_status_rsize,
3401         },
3402         [OP_SECINFO_NO_NAME] = {
3403                 .op_func = nfsd4_secinfo_no_name,
3404                 .op_release = nfsd4_secinfo_no_name_release,
3405                 .op_flags = OP_HANDLES_WRONGSEC,
3406                 .op_name = "OP_SECINFO_NO_NAME",
3407                 .op_rsize_bop = nfsd4_secinfo_rsize,
3408         },
3409         [OP_TEST_STATEID] = {
3410                 .op_func = nfsd4_test_stateid,
3411                 .op_flags = ALLOWED_WITHOUT_FH,
3412                 .op_name = "OP_TEST_STATEID",
3413                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3414         },
3415         [OP_FREE_STATEID] = {
3416                 .op_func = nfsd4_free_stateid,
3417                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3418                 .op_name = "OP_FREE_STATEID",
3419                 .op_get_currentstateid = nfsd4_get_freestateid,
3420                 .op_rsize_bop = nfsd4_only_status_rsize,
3421         },
3422 #ifdef CONFIG_NFSD_PNFS
3423         [OP_GETDEVICEINFO] = {
3424                 .op_func = nfsd4_getdeviceinfo,
3425                 .op_release = nfsd4_getdeviceinfo_release,
3426                 .op_flags = ALLOWED_WITHOUT_FH,
3427                 .op_name = "OP_GETDEVICEINFO",
3428                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3429         },
3430         [OP_LAYOUTGET] = {
3431                 .op_func = nfsd4_layoutget,
3432                 .op_release = nfsd4_layoutget_release,
3433                 .op_flags = OP_MODIFIES_SOMETHING,
3434                 .op_name = "OP_LAYOUTGET",
3435                 .op_rsize_bop = nfsd4_layoutget_rsize,
3436         },
3437         [OP_LAYOUTCOMMIT] = {
3438                 .op_func = nfsd4_layoutcommit,
3439                 .op_flags = OP_MODIFIES_SOMETHING,
3440                 .op_name = "OP_LAYOUTCOMMIT",
3441                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3442         },
3443         [OP_LAYOUTRETURN] = {
3444                 .op_func = nfsd4_layoutreturn,
3445                 .op_flags = OP_MODIFIES_SOMETHING,
3446                 .op_name = "OP_LAYOUTRETURN",
3447                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3448         },
3449 #endif /* CONFIG_NFSD_PNFS */
3450
3451         /* NFSv4.2 operations */
3452         [OP_ALLOCATE] = {
3453                 .op_func = nfsd4_allocate,
3454                 .op_flags = OP_MODIFIES_SOMETHING,
3455                 .op_name = "OP_ALLOCATE",
3456                 .op_rsize_bop = nfsd4_only_status_rsize,
3457         },
3458         [OP_DEALLOCATE] = {
3459                 .op_func = nfsd4_deallocate,
3460                 .op_flags = OP_MODIFIES_SOMETHING,
3461                 .op_name = "OP_DEALLOCATE",
3462                 .op_rsize_bop = nfsd4_only_status_rsize,
3463         },
3464         [OP_CLONE] = {
3465                 .op_func = nfsd4_clone,
3466                 .op_flags = OP_MODIFIES_SOMETHING,
3467                 .op_name = "OP_CLONE",
3468                 .op_rsize_bop = nfsd4_only_status_rsize,
3469         },
3470         [OP_COPY] = {
3471                 .op_func = nfsd4_copy,
3472                 .op_flags = OP_MODIFIES_SOMETHING,
3473                 .op_name = "OP_COPY",
3474                 .op_rsize_bop = nfsd4_copy_rsize,
3475         },
3476         [OP_READ_PLUS] = {
3477                 .op_func = nfsd4_read,
3478                 .op_release = nfsd4_read_release,
3479                 .op_name = "OP_READ_PLUS",
3480                 .op_rsize_bop = nfsd4_read_plus_rsize,
3481                 .op_get_currentstateid = nfsd4_get_readstateid,
3482         },
3483         [OP_SEEK] = {
3484                 .op_func = nfsd4_seek,
3485                 .op_name = "OP_SEEK",
3486                 .op_rsize_bop = nfsd4_seek_rsize,
3487         },
3488         [OP_OFFLOAD_STATUS] = {
3489                 .op_func = nfsd4_offload_status,
3490                 .op_name = "OP_OFFLOAD_STATUS",
3491                 .op_rsize_bop = nfsd4_offload_status_rsize,
3492         },
3493         [OP_OFFLOAD_CANCEL] = {
3494                 .op_func = nfsd4_offload_cancel,
3495                 .op_flags = OP_MODIFIES_SOMETHING,
3496                 .op_name = "OP_OFFLOAD_CANCEL",
3497                 .op_rsize_bop = nfsd4_only_status_rsize,
3498         },
3499         [OP_COPY_NOTIFY] = {
3500                 .op_func = nfsd4_copy_notify,
3501                 .op_flags = OP_MODIFIES_SOMETHING,
3502                 .op_name = "OP_COPY_NOTIFY",
3503                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3504         },
3505         [OP_GETXATTR] = {
3506                 .op_func = nfsd4_getxattr,
3507                 .op_name = "OP_GETXATTR",
3508                 .op_rsize_bop = nfsd4_getxattr_rsize,
3509         },
3510         [OP_SETXATTR] = {
3511                 .op_func = nfsd4_setxattr,
3512                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3513                 .op_name = "OP_SETXATTR",
3514                 .op_rsize_bop = nfsd4_setxattr_rsize,
3515         },
3516         [OP_LISTXATTRS] = {
3517                 .op_func = nfsd4_listxattrs,
3518                 .op_name = "OP_LISTXATTRS",
3519                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3520         },
3521         [OP_REMOVEXATTR] = {
3522                 .op_func = nfsd4_removexattr,
3523                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3524                 .op_name = "OP_REMOVEXATTR",
3525                 .op_rsize_bop = nfsd4_removexattr_rsize,
3526         },
3527 };
3528
3529 /**
3530  * nfsd4_spo_must_allow - Determine if the compound op contains an
3531  * operation that is allowed to be sent with machine credentials
3532  *
3533  * @rqstp: a pointer to the struct svc_rqst
3534  *
3535  * Checks to see if the compound contains a spo_must_allow op
3536  * and confirms that it was sent with the proper machine creds.
3537  */
3538
3539 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3540 {
3541         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3542         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3543         struct nfsd4_op *this;
3544         struct nfsd4_compound_state *cstate = &resp->cstate;
3545         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3546         u32 opiter;
3547
3548         if (!cstate->minorversion)
3549                 return false;
3550
3551         if (cstate->spo_must_allowed)
3552                 return true;
3553
3554         opiter = resp->opcnt;
3555         while (opiter < argp->opcnt) {
3556                 this = &argp->ops[opiter++];
3557                 if (test_bit(this->opnum, allow->u.longs) &&
3558                         cstate->clp->cl_mach_cred &&
3559                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3560                         cstate->spo_must_allowed = true;
3561                         return true;
3562                 }
3563         }
3564         cstate->spo_must_allowed = false;
3565         return false;
3566 }
3567
3568 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3569 {
3570         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3571                 return op_encode_hdr_size * sizeof(__be32);
3572
3573         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3574         return OPDESC(op)->op_rsize_bop(rqstp, op);
3575 }
3576
3577 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3578 {
3579         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3580                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3581                         op->opnum, nfsd4_op_name(op->opnum));
3582                 WARN_ON_ONCE(1);
3583         }
3584 }
3585
3586 static const char *nfsd4_op_name(unsigned opnum)
3587 {
3588         if (opnum < ARRAY_SIZE(nfsd4_ops))
3589                 return nfsd4_ops[opnum].op_name;
3590         return "unknown_operation";
3591 }
3592
3593 static const struct svc_procedure nfsd_procedures4[2] = {
3594         [NFSPROC4_NULL] = {
3595                 .pc_func = nfsd4_proc_null,
3596                 .pc_decode = nfssvc_decode_voidarg,
3597                 .pc_encode = nfssvc_encode_voidres,
3598                 .pc_argsize = sizeof(struct nfsd_voidargs),
3599                 .pc_argzero = sizeof(struct nfsd_voidargs),
3600                 .pc_ressize = sizeof(struct nfsd_voidres),
3601                 .pc_cachetype = RC_NOCACHE,
3602                 .pc_xdrressize = 1,
3603                 .pc_name = "NULL",
3604         },
3605         [NFSPROC4_COMPOUND] = {
3606                 .pc_func = nfsd4_proc_compound,
3607                 .pc_decode = nfs4svc_decode_compoundargs,
3608                 .pc_encode = nfs4svc_encode_compoundres,
3609                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3610                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3611                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3612                 .pc_release = nfsd4_release_compoundargs,
3613                 .pc_cachetype = RC_NOCACHE,
3614                 .pc_xdrressize = NFSD_BUFSIZE/4,
3615                 .pc_name = "COMPOUND",
3616         },
3617 };
3618
3619 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3620                               nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3621 const struct svc_version nfsd_version4 = {
3622         .vs_vers                = 4,
3623         .vs_nproc               = ARRAY_SIZE(nfsd_procedures4),
3624         .vs_proc                = nfsd_procedures4,
3625         .vs_count               = nfsd_count4,
3626         .vs_dispatch            = nfsd_dispatch,
3627         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3628         .vs_rpcb_optnl          = true,
3629         .vs_need_cong_ctrl      = true,
3630 };