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
 735         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
 736                              commit->co_count,
 737                              (__be32 *)commit->co_verf.data);
 738 }
 739
 740 static __be32
 741 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 742              union nfsd4_op_u *u)
 743 {
 744         struct nfsd4_create *create = &u->create;
 745         struct nfsd_attrs attrs = {
 746                 .na_iattr       = &create->cr_iattr,
 747                 .na_seclabel    = &create->cr_label,
 748         };
 749         struct svc_fh resfh;
 750         __be32 status;
 751         dev_t rdev;
 752
 753         fh_init(&resfh, NFS4_FHSIZE);
 754
 755         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 756         if (status)
 757                 return status;
 758
 759         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 760                                     nfsd_attrmask);
 761         if (status)
 762                 return status;
 763
 764         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 765         current->fs->umask = create->cr_umask;
 766         switch (create->cr_type) {
 767         case NF4LNK:
 768                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 769                                       create->cr_name, create->cr_namelen,
 770                                       create->cr_data, &attrs, &resfh);
 771                 break;
 772
 773         case NF4BLK:
 774                 status = nfserr_inval;
 775                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 776                 if (MAJOR(rdev) != create->cr_specdata1 ||
 777                     MINOR(rdev) != create->cr_specdata2)
 778                         goto out_umask;
 779                 status = nfsd_create(rqstp, &cstate->current_fh,
 780                                      create->cr_name, create->cr_namelen,
 781                                      &attrs, S_IFBLK, rdev, &resfh);
 782                 break;
 783
 784         case NF4CHR:
 785                 status = nfserr_inval;
 786                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 787                 if (MAJOR(rdev) != create->cr_specdata1 ||
 788                     MINOR(rdev) != create->cr_specdata2)
 789                         goto out_umask;
 790                 status = nfsd_create(rqstp, &cstate->current_fh,
 791                                      create->cr_name, create->cr_namelen,
 792                                      &attrs, S_IFCHR, rdev, &resfh);
 793                 break;
 794
 795         case NF4SOCK:
 796                 status = nfsd_create(rqstp, &cstate->current_fh,
 797                                      create->cr_name, create->cr_namelen,
 798                                      &attrs, S_IFSOCK, 0, &resfh);
 799                 break;
 800
 801         case NF4FIFO:
 802                 status = nfsd_create(rqstp, &cstate->current_fh,
 803                                      create->cr_name, create->cr_namelen,
 804                                      &attrs, S_IFIFO, 0, &resfh);
 805                 break;
 806
 807         case NF4DIR:
 808                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 809                 status = nfsd_create(rqstp, &cstate->current_fh,
 810                                      create->cr_name, create->cr_namelen,
 811                                      &attrs, S_IFDIR, 0, &resfh);
 812                 break;
 813
 814         default:
 815                 status = nfserr_badtype;
 816         }
 817
 818         if (status)
 819                 goto out;
 820
 821         if (attrs.na_labelerr)
 822                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 823         if (attrs.na_aclerr)
 824                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 825         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 826         fh_dup2(&cstate->current_fh, &resfh);
 827 out:
 828         fh_put(&resfh);
 829 out_umask:
 830         current->fs->umask = 0;
 831         nfsd_attrs_free(&attrs);
 832         return status;
 833 }
 834
 835 static __be32
 836 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 837               union nfsd4_op_u *u)
 838 {
 839         struct nfsd4_getattr *getattr = &u->getattr;
 840         __be32 status;
 841
 842         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 843         if (status)
 844                 return status;
 845
 846         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 847                 return nfserr_inval;
 848
 849         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 850         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 851         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 852
 853         getattr->ga_fhp = &cstate->current_fh;
 854         return nfs_ok;
 855 }
 856
 857 static __be32
 858 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 859            union nfsd4_op_u *u)
 860 {
 861         struct nfsd4_link *link = &u->link;
 862         __be32 status;
 863
 864         status = nfsd_link(rqstp, &cstate->current_fh,
 865                            link->li_name, link->li_namelen, &cstate->save_fh);
 866         if (!status)
 867                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 868         return status;
 869 }
 870
 871 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 872 {
 873         struct svc_fh tmp_fh;
 874         __be32 ret;
 875
 876         fh_init(&tmp_fh, NFS4_FHSIZE);
 877         ret = exp_pseudoroot(rqstp, &tmp_fh);
 878         if (ret)
 879                 return ret;
 880         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 881                 fh_put(&tmp_fh);
 882                 return nfserr_noent;
 883         }
 884         fh_put(&tmp_fh);
 885         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 886 }
 887
 888 static __be32
 889 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 890               union nfsd4_op_u *u)
 891 {
 892         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 893 }
 894
 895 static __be32
 896 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 897              union nfsd4_op_u *u)
 898 {
 899         return nfsd_lookup(rqstp, &cstate->current_fh,
 900                            u->lookup.lo_name, u->lookup.lo_len,
 901                            &cstate->current_fh);
 902 }
 903
 904 static __be32
 905 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 906            union nfsd4_op_u *u)
 907 {
 908         struct nfsd4_read *read = &u->read;
 909         __be32 status;
 910
 911         read->rd_nf = NULL;
 912
 913         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 914                               read->rd_offset, read->rd_length);
 915
 916         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 917         if (read->rd_offset > (u64)OFFSET_MAX)
 918                 read->rd_offset = (u64)OFFSET_MAX;
 919         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 920                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 921
 922         /*
 923          * If we do a zero copy read, then a client will see read data
 924          * that reflects the state of the file *after* performing the
 925          * following compound.
 926          *
 927          * To ensure proper ordering, we therefore turn off zero copy if
 928          * the client wants us to do more in this compound:
 929          */
 930         if (!nfsd4_last_compound_op(rqstp))
 931                 __clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 932
 933         /* check stateid */
 934         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 935                                         &read->rd_stateid, RD_STATE,
 936                                         &read->rd_nf, NULL);
 937         if (status) {
 938                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
 939                 goto out;
 940         }
 941         status = nfs_ok;
 942 out:
 943         read->rd_rqstp = rqstp;
 944         read->rd_fhp = &cstate->current_fh;
 945         return status;
 946 }
 947
 948
 949 static void
 950 nfsd4_read_release(union nfsd4_op_u *u)
 951 {
 952         if (u->read.rd_nf)
 953                 nfsd_file_put(u->read.rd_nf);
 954         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 955                              u->read.rd_offset, u->read.rd_length);
 956 }
 957
 958 static __be32
 959 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 960               union nfsd4_op_u *u)
 961 {
 962         struct nfsd4_readdir *readdir = &u->readdir;
 963         u64 cookie = readdir->rd_cookie;
 964         static const nfs4_verifier zeroverf;
 965
 966         /* no need to check permission - this will be done in nfsd_readdir() */
 967
 968         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 969                 return nfserr_inval;
 970
 971         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 972         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 973         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 974
 975         if ((cookie == 1) || (cookie == 2) ||
 976             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
 977                 return nfserr_bad_cookie;
 978
 979         readdir->rd_rqstp = rqstp;
 980         readdir->rd_fhp = &cstate->current_fh;
 981         return nfs_ok;
 982 }
 983
 984 static __be32
 985 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 986                union nfsd4_op_u *u)
 987 {
 988         u->readlink.rl_rqstp = rqstp;
 989         u->readlink.rl_fhp = &cstate->current_fh;
 990         return nfs_ok;
 991 }
 992
 993 static __be32
 994 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 995              union nfsd4_op_u *u)
 996 {
 997         struct nfsd4_remove *remove = &u->remove;
 998         __be32 status;
 999
1000         if (opens_in_grace(SVC_NET(rqstp)))
1001                 return nfserr_grace;
1002         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1003                              remove->rm_name, remove->rm_namelen);
1004         if (!status)
1005                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1006         return status;
1007 }
1008
1009 static __be32
1010 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1011              union nfsd4_op_u *u)
1012 {
1013         struct nfsd4_rename *rename = &u->rename;
1014         __be32 status;
1015
1016         if (opens_in_grace(SVC_NET(rqstp)))
1017                 return nfserr_grace;
1018         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1019                              rename->rn_snamelen, &cstate->current_fh,
1020                              rename->rn_tname, rename->rn_tnamelen);
1021         if (status)
1022                 return status;
1023         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
1024         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
1025         return nfs_ok;
1026 }
1027
1028 static __be32
1029 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1030               union nfsd4_op_u *u)
1031 {
1032         struct nfsd4_secinfo *secinfo = &u->secinfo;
1033         struct svc_export *exp;
1034         struct dentry *dentry;
1035         __be32 err;
1036
1037         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1038         if (err)
1039                 return err;
1040         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1041                                     secinfo->si_name, secinfo->si_namelen,
1042                                     &exp, &dentry);
1043         if (err)
1044                 return err;
1045         if (d_really_is_negative(dentry)) {
1046                 exp_put(exp);
1047                 err = nfserr_noent;
1048         } else
1049                 secinfo->si_exp = exp;
1050         dput(dentry);
1051         if (cstate->minorversion)
1052                 /* See rfc 5661 section 2.6.3.1.1.8 */
1053                 fh_put(&cstate->current_fh);
1054         return err;
1055 }
1056
1057 static __be32
1058 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1059                 union nfsd4_op_u *u)
1060 {
1061         __be32 err;
1062
1063         switch (u->secinfo_no_name.sin_style) {
1064         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1065                 break;
1066         case NFS4_SECINFO_STYLE4_PARENT:
1067                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1068                 if (err)
1069                         return err;
1070                 break;
1071         default:
1072                 return nfserr_inval;
1073         }
1074
1075         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1076         fh_put(&cstate->current_fh);
1077         return nfs_ok;
1078 }
1079
1080 static void
1081 nfsd4_secinfo_release(union nfsd4_op_u *u)
1082 {
1083         if (u->secinfo.si_exp)
1084                 exp_put(u->secinfo.si_exp);
1085 }
1086
1087 static void
1088 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1089 {
1090         if (u->secinfo_no_name.sin_exp)
1091                 exp_put(u->secinfo_no_name.sin_exp);
1092 }
1093
1094 static __be32
1095 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1096               union nfsd4_op_u *u)
1097 {
1098         struct nfsd4_setattr *setattr = &u->setattr;
1099         struct nfsd_attrs attrs = {
1100                 .na_iattr       = &setattr->sa_iattr,
1101                 .na_seclabel    = &setattr->sa_label,
1102         };
1103         struct inode *inode;
1104         __be32 status = nfs_ok;
1105         int err;
1106
1107         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1108                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1109                                 &cstate->current_fh, &setattr->sa_stateid,
1110                                 WR_STATE, NULL, NULL);
1111                 if (status) {
1112                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
1113                         return status;
1114                 }
1115         }
1116         err = fh_want_write(&cstate->current_fh);
1117         if (err)
1118                 return nfserrno(err);
1119         status = nfs_ok;
1120
1121         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1122                                     nfsd_attrmask);
1123         if (status)
1124                 goto out;
1125
1126         inode = cstate->current_fh.fh_dentry->d_inode;
1127         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1128                                    setattr->sa_acl, &attrs);
1129
1130         if (status)
1131                 goto out;
1132         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1133                                 0, (time64_t)0);
1134         if (!status)
1135                 status = nfserrno(attrs.na_labelerr);
1136 out:
1137         nfsd_attrs_free(&attrs);
1138         fh_drop_write(&cstate->current_fh);
1139         return status;
1140 }
1141
1142 static __be32
1143 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1144             union nfsd4_op_u *u)
1145 {
1146         struct nfsd4_write *write = &u->write;
1147         stateid_t *stateid = &write->wr_stateid;
1148         struct nfsd_file *nf = NULL;
1149         __be32 status = nfs_ok;
1150         unsigned long cnt;
1151         int nvecs;
1152
1153         if (write->wr_offset > (u64)OFFSET_MAX ||
1154             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1155                 return nfserr_fbig;
1156
1157         cnt = write->wr_buflen;
1158         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1159                                write->wr_offset, cnt);
1160         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1161                                                 stateid, WR_STATE, &nf, NULL);
1162         if (status) {
1163                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1164                 return status;
1165         }
1166
1167         write->wr_how_written = write->wr_stable_how;
1168
1169         nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1170         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1171
1172         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1173                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1174                                 write->wr_how_written,
1175                                 (__be32 *)write->wr_verifier.data);
1176         nfsd_file_put(nf);
1177
1178         write->wr_bytes_written = cnt;
1179         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1180                               write->wr_offset, cnt);
1181         return status;
1182 }
1183
1184 static __be32
1185 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1186                   stateid_t *src_stateid, struct nfsd_file **src,
1187                   stateid_t *dst_stateid, struct nfsd_file **dst)
1188 {
1189         __be32 status;
1190
1191         if (!cstate->save_fh.fh_dentry)
1192                 return nfserr_nofilehandle;
1193
1194         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1195                                             src_stateid, RD_STATE, src, NULL);
1196         if (status) {
1197                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1198                 goto out;
1199         }
1200
1201         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1202                                             dst_stateid, WR_STATE, dst, NULL);
1203         if (status) {
1204                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1205                 goto out_put_src;
1206         }
1207
1208         /* fix up for NFS-specific error code */
1209         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1210             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1211                 status = nfserr_wrong_type;
1212                 goto out_put_dst;
1213         }
1214
1215 out:
1216         return status;
1217 out_put_dst:
1218         nfsd_file_put(*dst);
1219 out_put_src:
1220         nfsd_file_put(*src);
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 **retwork, struct vfsmount **ss_mnt)
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
1305         *ss_mnt = NULL;
1306         *retwork = 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                                 kfree(work);
1324                                 return nfserr_eagain;
1325                         }
1326                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1327                         goto try_again;
1328                 }
1329                 *ss_mnt = ni->nsui_vfsmount;
1330                 refcount_inc(&ni->nsui_refcnt);
1331                 spin_unlock(&nn->nfsd_ssc_lock);
1332                 kfree(work);
1333
1334                 /* return vfsmount in ss_mnt */
1335                 return 0;
1336         }
1337         if (work) {
1338                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1339                 refcount_set(&work->nsui_refcnt, 2);
1340                 work->nsui_busy = true;
1341                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1342                 *retwork = work;
1343         }
1344         spin_unlock(&nn->nfsd_ssc_lock);
1345         return 0;
1346 }
1347
1348 static void nfsd4_ssc_update_dul_work(struct nfsd_net *nn,
1349                 struct nfsd4_ssc_umount_item *work, struct vfsmount *ss_mnt)
1350 {
1351         /* set nsui_vfsmount, clear busy flag and wakeup waiters */
1352         spin_lock(&nn->nfsd_ssc_lock);
1353         work->nsui_vfsmount = ss_mnt;
1354         work->nsui_busy = false;
1355         wake_up_all(&nn->nfsd_ssc_waitq);
1356         spin_unlock(&nn->nfsd_ssc_lock);
1357 }
1358
1359 static void nfsd4_ssc_cancel_dul_work(struct nfsd_net *nn,
1360                 struct nfsd4_ssc_umount_item *work)
1361 {
1362         spin_lock(&nn->nfsd_ssc_lock);
1363         list_del(&work->nsui_list);
1364         wake_up_all(&nn->nfsd_ssc_waitq);
1365         spin_unlock(&nn->nfsd_ssc_lock);
1366         kfree(work);
1367 }
1368
1369 /*
1370  * Support one copy source server for now.
1371  */
1372 static __be32
1373 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1374                        struct vfsmount **mount)
1375 {
1376         struct file_system_type *type;
1377         struct vfsmount *ss_mnt;
1378         struct nfs42_netaddr *naddr;
1379         struct sockaddr_storage tmp_addr;
1380         size_t tmp_addrlen, match_netid_len = 3;
1381         char *startsep = "", *endsep = "", *match_netid = "tcp";
1382         char *ipaddr, *dev_name, *raw_data;
1383         int len, raw_len;
1384         __be32 status = nfserr_inval;
1385         struct nfsd4_ssc_umount_item *work = NULL;
1386         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1387
1388         naddr = &nss->u.nl4_addr;
1389         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1390                                          naddr->addr_len,
1391                                          (struct sockaddr *)&tmp_addr,
1392                                          sizeof(tmp_addr));
1393         if (tmp_addrlen == 0)
1394                 goto out_err;
1395
1396         if (tmp_addr.ss_family == AF_INET6) {
1397                 startsep = "[";
1398                 endsep = "]";
1399                 match_netid = "tcp6";
1400                 match_netid_len = 4;
1401         }
1402
1403         if (naddr->netid_len != match_netid_len ||
1404                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1405                 goto out_err;
1406
1407         /* Construct the raw data for the vfs_kern_mount call */
1408         len = RPC_MAX_ADDRBUFLEN + 1;
1409         ipaddr = kzalloc(len, GFP_KERNEL);
1410         if (!ipaddr)
1411                 goto out_err;
1412
1413         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1414
1415         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1416
1417         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1418         raw_data = kzalloc(raw_len, GFP_KERNEL);
1419         if (!raw_data)
1420                 goto out_free_ipaddr;
1421
1422         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1423
1424         status = nfserr_nodev;
1425         type = get_fs_type("nfs");
1426         if (!type)
1427                 goto out_free_rawdata;
1428
1429         /* Set the server:<export> for the vfs_kern_mount call */
1430         dev_name = kzalloc(len + 5, GFP_KERNEL);
1431         if (!dev_name)
1432                 goto out_free_rawdata;
1433         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1434
1435         status = nfsd4_ssc_setup_dul(nn, ipaddr, &work, &ss_mnt);
1436         if (status)
1437                 goto out_free_devname;
1438         if (ss_mnt)
1439                 goto out_done;
1440
1441         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1442         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1443         module_put(type->owner);
1444         if (IS_ERR(ss_mnt)) {
1445                 status = nfserr_nodev;
1446                 if (work)
1447                         nfsd4_ssc_cancel_dul_work(nn, work);
1448                 goto out_free_devname;
1449         }
1450         if (work)
1451                 nfsd4_ssc_update_dul_work(nn, work, ss_mnt);
1452 out_done:
1453         status = 0;
1454         *mount = ss_mnt;
1455
1456 out_free_devname:
1457         kfree(dev_name);
1458 out_free_rawdata:
1459         kfree(raw_data);
1460 out_free_ipaddr:
1461         kfree(ipaddr);
1462 out_err:
1463         return status;
1464 }
1465
1466 static void
1467 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1468 {
1469         nfs_do_sb_deactive(ss_mnt->mnt_sb);
1470         mntput(ss_mnt);
1471 }
1472
1473 /*
1474  * Verify COPY destination stateid.
1475  *
1476  * Connect to the source server with NFSv4.1.
1477  * Create the source struct file for nfsd_copy_range.
1478  * Called with COPY cstate:
1479  *    SAVED_FH: source filehandle
1480  *    CURRENT_FH: destination filehandle
1481  */
1482 static __be32
1483 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1484                       struct nfsd4_compound_state *cstate,
1485                       struct nfsd4_copy *copy, struct vfsmount **mount)
1486 {
1487         struct svc_fh *s_fh = NULL;
1488         stateid_t *s_stid = &copy->cp_src_stateid;
1489         __be32 status = nfserr_inval;
1490
1491         /* Verify the destination stateid and set dst struct file*/
1492         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1493                                             &copy->cp_dst_stateid,
1494                                             WR_STATE, &copy->nf_dst, NULL);
1495         if (status)
1496                 goto out;
1497
1498         status = nfsd4_interssc_connect(copy->cp_src, rqstp, mount);
1499         if (status)
1500                 goto out;
1501
1502         s_fh = &cstate->save_fh;
1503
1504         copy->c_fh.size = s_fh->fh_handle.fh_size;
1505         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1506         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1507         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1508                sizeof(stateid_opaque_t));
1509
1510         status = 0;
1511 out:
1512         return status;
1513 }
1514
1515 static void
1516 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct file *filp,
1517                         struct nfsd_file *dst)
1518 {
1519         bool found = false;
1520         long timeout;
1521         struct nfsd4_ssc_umount_item *tmp;
1522         struct nfsd4_ssc_umount_item *ni = NULL;
1523         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1524
1525         nfs42_ssc_close(filp);
1526         nfsd_file_put(dst);
1527         fput(filp);
1528
1529         if (!nn) {
1530                 mntput(ss_mnt);
1531                 return;
1532         }
1533         spin_lock(&nn->nfsd_ssc_lock);
1534         timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1535         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1536                 if (ni->nsui_vfsmount->mnt_sb == ss_mnt->mnt_sb) {
1537                         list_del(&ni->nsui_list);
1538                         /*
1539                          * vfsmount can be shared by multiple exports,
1540                          * decrement refcnt. If the count drops to 1 it
1541                          * will be unmounted when nsui_expire expires.
1542                          */
1543                         refcount_dec(&ni->nsui_refcnt);
1544                         ni->nsui_expire = jiffies + timeout;
1545                         list_add_tail(&ni->nsui_list, &nn->nfsd_ssc_mount_list);
1546                         found = true;
1547                         break;
1548                 }
1549         }
1550         spin_unlock(&nn->nfsd_ssc_lock);
1551         if (!found) {
1552                 mntput(ss_mnt);
1553                 return;
1554         }
1555 }
1556
1557 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1558
1559 static __be32
1560 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1561                       struct nfsd4_compound_state *cstate,
1562                       struct nfsd4_copy *copy,
1563                       struct vfsmount **mount)
1564 {
1565         *mount = NULL;
1566         return nfserr_inval;
1567 }
1568
1569 static void
1570 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct file *filp,
1571                         struct nfsd_file *dst)
1572 {
1573 }
1574
1575 static void
1576 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1577 {
1578 }
1579
1580 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1581                                    struct nfs_fh *src_fh,
1582                                    nfs4_stateid *stateid)
1583 {
1584         return NULL;
1585 }
1586 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1587
1588 static __be32
1589 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1590                       struct nfsd4_compound_state *cstate,
1591                       struct nfsd4_copy *copy)
1592 {
1593         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1594                                  &copy->nf_src, &copy->cp_dst_stateid,
1595                                  &copy->nf_dst);
1596 }
1597
1598 static void
1599 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1600 {
1601         nfsd_file_put(src);
1602         nfsd_file_put(dst);
1603 }
1604
1605 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1606 {
1607         struct nfsd4_cb_offload *cbo =
1608                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1609
1610         kfree(cbo);
1611 }
1612
1613 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1614                                  struct rpc_task *task)
1615 {
1616         struct nfsd4_cb_offload *cbo =
1617                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1618
1619         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1620         return 1;
1621 }
1622
1623 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1624         .release = nfsd4_cb_offload_release,
1625         .done = nfsd4_cb_offload_done
1626 };
1627
1628 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1629 {
1630         copy->cp_res.wr_stable_how =
1631                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1632                         NFS_FILE_SYNC : NFS_UNSTABLE;
1633         nfsd4_copy_set_sync(copy, sync);
1634         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1635 }
1636
1637 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1638                                      struct file *dst,
1639                                      struct file *src)
1640 {
1641         errseq_t since;
1642         ssize_t bytes_copied = 0;
1643         u64 bytes_total = copy->cp_count;
1644         u64 src_pos = copy->cp_src_pos;
1645         u64 dst_pos = copy->cp_dst_pos;
1646         int status;
1647
1648         /* See RFC 7862 p.67: */
1649         if (bytes_total == 0)
1650                 bytes_total = ULLONG_MAX;
1651         do {
1652                 if (kthread_should_stop())
1653                         break;
1654                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1655                                                     bytes_total);
1656                 if (bytes_copied <= 0)
1657                         break;
1658                 bytes_total -= bytes_copied;
1659                 copy->cp_res.wr_bytes_written += bytes_copied;
1660                 src_pos += bytes_copied;
1661                 dst_pos += bytes_copied;
1662         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1663         /* for a non-zero asynchronous copy do a commit of data */
1664         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1665                 since = READ_ONCE(dst->f_wb_err);
1666                 status = vfs_fsync_range(dst, copy->cp_dst_pos,
1667                                          copy->cp_res.wr_bytes_written, 0);
1668                 if (!status)
1669                         status = filemap_check_wb_err(dst->f_mapping, since);
1670                 if (!status)
1671                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1672         }
1673         return bytes_copied;
1674 }
1675
1676 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1677                             struct file *src, struct file *dst,
1678                             bool sync)
1679 {
1680         __be32 status;
1681         ssize_t bytes;
1682
1683         bytes = _nfsd_copy_file_range(copy, dst, src);
1684
1685         /* for async copy, we ignore the error, client can always retry
1686          * to get the error
1687          */
1688         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1689                 status = nfserrno(bytes);
1690         else {
1691                 nfsd4_init_copy_res(copy, sync);
1692                 status = nfs_ok;
1693         }
1694         return status;
1695 }
1696
1697 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1698 {
1699         dst->cp_src_pos = src->cp_src_pos;
1700         dst->cp_dst_pos = src->cp_dst_pos;
1701         dst->cp_count = src->cp_count;
1702         dst->cp_flags = src->cp_flags;
1703         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1704         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1705         dst->cp_clp = src->cp_clp;
1706         dst->nf_dst = nfsd_file_get(src->nf_dst);
1707         /* for inter, nf_src doesn't exist yet */
1708         if (!nfsd4_ssc_is_inter(src))
1709                 dst->nf_src = nfsd_file_get(src->nf_src);
1710
1711         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1712         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1713         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1714         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1715         dst->ss_mnt = src->ss_mnt;
1716 }
1717
1718 static void cleanup_async_copy(struct nfsd4_copy *copy)
1719 {
1720         nfs4_free_copy_state(copy);
1721         nfsd_file_put(copy->nf_dst);
1722         if (!nfsd4_ssc_is_inter(copy))
1723                 nfsd_file_put(copy->nf_src);
1724         spin_lock(&copy->cp_clp->async_lock);
1725         list_del(&copy->copies);
1726         spin_unlock(&copy->cp_clp->async_lock);
1727         nfs4_put_copy(copy);
1728 }
1729
1730 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1731 {
1732         struct nfsd4_cb_offload *cbo;
1733
1734         cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1735         if (!cbo)
1736                 return;
1737
1738         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1739         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1740         cbo->co_nfserr = nfserr;
1741
1742         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1743                       NFSPROC4_CLNT_CB_OFFLOAD);
1744         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1745                               &cbo->co_fh, copy->cp_count, nfserr);
1746         nfsd4_run_cb(&cbo->co_cb);
1747 }
1748
1749 /**
1750  * nfsd4_do_async_copy - kthread function for background server-side COPY
1751  * @data: arguments for COPY operation
1752  *
1753  * Return values:
1754  *   %0: Copy operation is done.
1755  */
1756 static int nfsd4_do_async_copy(void *data)
1757 {
1758         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1759         __be32 nfserr;
1760
1761         if (nfsd4_ssc_is_inter(copy)) {
1762                 struct file *filp;
1763
1764                 filp = nfs42_ssc_open(copy->ss_mnt, &copy->c_fh,
1765                                       &copy->stateid);
1766                 if (IS_ERR(filp)) {
1767                         switch (PTR_ERR(filp)) {
1768                         case -EBADF:
1769                                 nfserr = nfserr_wrong_type;
1770                                 break;
1771                         default:
1772                                 nfserr = nfserr_offload_denied;
1773                         }
1774                         nfsd4_interssc_disconnect(copy->ss_mnt);
1775                         goto do_callback;
1776                 }
1777                 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1778                                        false);
1779                 nfsd4_cleanup_inter_ssc(copy->ss_mnt, filp, copy->nf_dst);
1780         } else {
1781                 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1782                                        copy->nf_dst->nf_file, false);
1783                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1784         }
1785
1786 do_callback:
1787         nfsd4_send_cb_offload(copy, nfserr);
1788         cleanup_async_copy(copy);
1789         return 0;
1790 }
1791
1792 static __be32
1793 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1794                 union nfsd4_op_u *u)
1795 {
1796         struct nfsd4_copy *copy = &u->copy;
1797         __be32 status;
1798         struct nfsd4_copy *async_copy = NULL;
1799
1800         if (nfsd4_ssc_is_inter(copy)) {
1801                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1802                         status = nfserr_notsupp;
1803                         goto out;
1804                 }
1805                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1806                                 &copy->ss_mnt);
1807                 if (status)
1808                         return nfserr_offload_denied;
1809         } else {
1810                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1811                 if (status)
1812                         return status;
1813         }
1814
1815         copy->cp_clp = cstate->clp;
1816         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1817                 sizeof(struct knfsd_fh));
1818         if (nfsd4_copy_is_async(copy)) {
1819                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1820
1821                 status = nfserrno(-ENOMEM);
1822                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1823                 if (!async_copy)
1824                         goto out_err;
1825                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1826                 if (!async_copy->cp_src)
1827                         goto out_err;
1828                 if (!nfs4_init_copy_state(nn, copy))
1829                         goto out_err;
1830                 refcount_set(&async_copy->refcount, 1);
1831                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1832                         sizeof(copy->cp_res.cb_stateid));
1833                 dup_copy_fields(copy, async_copy);
1834                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1835                                 async_copy, "%s", "copy thread");
1836                 if (IS_ERR(async_copy->copy_task))
1837                         goto out_err;
1838                 spin_lock(&async_copy->cp_clp->async_lock);
1839                 list_add(&async_copy->copies,
1840                                 &async_copy->cp_clp->async_copies);
1841                 spin_unlock(&async_copy->cp_clp->async_lock);
1842                 wake_up_process(async_copy->copy_task);
1843                 status = nfs_ok;
1844         } else {
1845                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1846                                        copy->nf_dst->nf_file, true);
1847                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1848         }
1849 out:
1850         return status;
1851 out_err:
1852         if (async_copy)
1853                 cleanup_async_copy(async_copy);
1854         status = nfserrno(-ENOMEM);
1855         if (nfsd4_ssc_is_inter(copy))
1856                 nfsd4_interssc_disconnect(copy->ss_mnt);
1857         goto out;
1858 }
1859
1860 struct nfsd4_copy *
1861 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1862 {
1863         struct nfsd4_copy *copy;
1864
1865         spin_lock(&clp->async_lock);
1866         list_for_each_entry(copy, &clp->async_copies, copies) {
1867                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1868                         continue;
1869                 refcount_inc(&copy->refcount);
1870                 spin_unlock(&clp->async_lock);
1871                 return copy;
1872         }
1873         spin_unlock(&clp->async_lock);
1874         return NULL;
1875 }
1876
1877 static __be32
1878 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1879                      struct nfsd4_compound_state *cstate,
1880                      union nfsd4_op_u *u)
1881 {
1882         struct nfsd4_offload_status *os = &u->offload_status;
1883         struct nfsd4_copy *copy;
1884         struct nfs4_client *clp = cstate->clp;
1885
1886         copy = find_async_copy(clp, &os->stateid);
1887         if (!copy) {
1888                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1889
1890                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1891         } else
1892                 nfsd4_stop_copy(copy);
1893
1894         return nfs_ok;
1895 }
1896
1897 static __be32
1898 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1899                   union nfsd4_op_u *u)
1900 {
1901         struct nfsd4_copy_notify *cn = &u->copy_notify;
1902         __be32 status;
1903         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1904         struct nfs4_stid *stid;
1905         struct nfs4_cpntf_state *cps;
1906         struct nfs4_client *clp = cstate->clp;
1907
1908         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1909                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1910                                         &stid);
1911         if (status)
1912                 return status;
1913
1914         cn->cpn_sec = nn->nfsd4_lease;
1915         cn->cpn_nsec = 0;
1916
1917         status = nfserrno(-ENOMEM);
1918         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1919         if (!cps)
1920                 goto out;
1921         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1922         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1923         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1924
1925         /* For now, only return one server address in cpn_src, the
1926          * address used by the client to connect to this server.
1927          */
1928         cn->cpn_src->nl4_type = NL4_NETADDR;
1929         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1930                                  &cn->cpn_src->u.nl4_addr);
1931         WARN_ON_ONCE(status);
1932         if (status) {
1933                 nfs4_put_cpntf_state(nn, cps);
1934                 goto out;
1935         }
1936 out:
1937         nfs4_put_stid(stid);
1938         return status;
1939 }
1940
1941 static __be32
1942 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1943                 struct nfsd4_fallocate *fallocate, int flags)
1944 {
1945         __be32 status;
1946         struct nfsd_file *nf;
1947
1948         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1949                                             &fallocate->falloc_stateid,
1950                                             WR_STATE, &nf, NULL);
1951         if (status != nfs_ok) {
1952                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1953                 return status;
1954         }
1955
1956         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1957                                      fallocate->falloc_offset,
1958                                      fallocate->falloc_length,
1959                                      flags);
1960         nfsd_file_put(nf);
1961         return status;
1962 }
1963 static __be32
1964 nfsd4_offload_status(struct svc_rqst *rqstp,
1965                      struct nfsd4_compound_state *cstate,
1966                      union nfsd4_op_u *u)
1967 {
1968         struct nfsd4_offload_status *os = &u->offload_status;
1969         __be32 status = 0;
1970         struct nfsd4_copy *copy;
1971         struct nfs4_client *clp = cstate->clp;
1972
1973         copy = find_async_copy(clp, &os->stateid);
1974         if (copy) {
1975                 os->count = copy->cp_res.wr_bytes_written;
1976                 nfs4_put_copy(copy);
1977         } else
1978                 status = nfserr_bad_stateid;
1979
1980         return status;
1981 }
1982
1983 static __be32
1984 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1985                union nfsd4_op_u *u)
1986 {
1987         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1988 }
1989
1990 static __be32
1991 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1992                  union nfsd4_op_u *u)
1993 {
1994         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1995                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1996 }
1997
1998 static __be32
1999 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2000            union nfsd4_op_u *u)
2001 {
2002         struct nfsd4_seek *seek = &u->seek;
2003         int whence;
2004         __be32 status;
2005         struct nfsd_file *nf;
2006
2007         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2008                                             &seek->seek_stateid,
2009                                             RD_STATE, &nf, NULL);
2010         if (status) {
2011                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
2012                 return status;
2013         }
2014
2015         switch (seek->seek_whence) {
2016         case NFS4_CONTENT_DATA:
2017                 whence = SEEK_DATA;
2018                 break;
2019         case NFS4_CONTENT_HOLE:
2020                 whence = SEEK_HOLE;
2021                 break;
2022         default:
2023                 status = nfserr_union_notsupp;
2024                 goto out;
2025         }
2026
2027         /*
2028          * Note:  This call does change file->f_pos, but nothing in NFSD
2029          *        should ever file->f_pos.
2030          */
2031         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2032         if (seek->seek_pos < 0)
2033                 status = nfserrno(seek->seek_pos);
2034         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2035                 seek->seek_eof = true;
2036
2037 out:
2038         nfsd_file_put(nf);
2039         return status;
2040 }
2041
2042 /* This routine never returns NFS_OK!  If there are no other errors, it
2043  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2044  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2045  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2046  */
2047 static __be32
2048 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2049              struct nfsd4_verify *verify)
2050 {
2051         __be32 *buf, *p;
2052         int count;
2053         __be32 status;
2054
2055         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2056         if (status)
2057                 return status;
2058
2059         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2060         if (status)
2061                 return status;
2062
2063         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2064             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2065                 return nfserr_inval;
2066         if (verify->ve_attrlen & 3)
2067                 return nfserr_inval;
2068
2069         /* count in words:
2070          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2071          */
2072         count = 4 + (verify->ve_attrlen >> 2);
2073         buf = kmalloc(count << 2, GFP_KERNEL);
2074         if (!buf)
2075                 return nfserr_jukebox;
2076
2077         p = buf;
2078         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2079                                     cstate->current_fh.fh_export,
2080                                     cstate->current_fh.fh_dentry,
2081                                     verify->ve_bmval,
2082                                     rqstp, 0);
2083         /*
2084          * If nfsd4_encode_fattr() ran out of space, assume that's because
2085          * the attributes are longer (hence different) than those given:
2086          */
2087         if (status == nfserr_resource)
2088                 status = nfserr_not_same;
2089         if (status)
2090                 goto out_kfree;
2091
2092         /* skip bitmap */
2093         p = buf + 1 + ntohl(buf[0]);
2094         status = nfserr_not_same;
2095         if (ntohl(*p++) != verify->ve_attrlen)
2096                 goto out_kfree;
2097         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2098                 status = nfserr_same;
2099
2100 out_kfree:
2101         kfree(buf);
2102         return status;
2103 }
2104
2105 static __be32
2106 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2107               union nfsd4_op_u *u)
2108 {
2109         __be32 status;
2110
2111         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2112         return status == nfserr_not_same ? nfs_ok : status;
2113 }
2114
2115 static __be32
2116 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2117              union nfsd4_op_u *u)
2118 {
2119         __be32 status;
2120
2121         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2122         return status == nfserr_same ? nfs_ok : status;
2123 }
2124
2125 #ifdef CONFIG_NFSD_PNFS
2126 static const struct nfsd4_layout_ops *
2127 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2128 {
2129         if (!exp->ex_layout_types) {
2130                 dprintk("%s: export does not support pNFS\n", __func__);
2131                 return NULL;
2132         }
2133
2134         if (layout_type >= LAYOUT_TYPE_MAX ||
2135             !(exp->ex_layout_types & (1 << layout_type))) {
2136                 dprintk("%s: layout type %d not supported\n",
2137                         __func__, layout_type);
2138                 return NULL;
2139         }
2140
2141         return nfsd4_layout_ops[layout_type];
2142 }
2143
2144 static __be32
2145 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2146                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2147 {
2148         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2149         const struct nfsd4_layout_ops *ops;
2150         struct nfsd4_deviceid_map *map;
2151         struct svc_export *exp;
2152         __be32 nfserr;
2153
2154         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2155                __func__,
2156                gdp->gd_layout_type,
2157                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2158                gdp->gd_maxcount);
2159
2160         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2161         if (!map) {
2162                 dprintk("%s: couldn't find device ID to export mapping!\n",
2163                         __func__);
2164                 return nfserr_noent;
2165         }
2166
2167         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2168         if (IS_ERR(exp)) {
2169                 dprintk("%s: could not find device id\n", __func__);
2170                 return nfserr_noent;
2171         }
2172
2173         nfserr = nfserr_layoutunavailable;
2174         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2175         if (!ops)
2176                 goto out;
2177
2178         nfserr = nfs_ok;
2179         if (gdp->gd_maxcount != 0) {
2180                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2181                                 rqstp, cstate->clp, gdp);
2182         }
2183
2184         gdp->gd_notify_types &= ops->notify_types;
2185 out:
2186         exp_put(exp);
2187         return nfserr;
2188 }
2189
2190 static void
2191 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2192 {
2193         kfree(u->getdeviceinfo.gd_device);
2194 }
2195
2196 static __be32
2197 nfsd4_layoutget(struct svc_rqst *rqstp,
2198                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2199 {
2200         struct nfsd4_layoutget *lgp = &u->layoutget;
2201         struct svc_fh *current_fh = &cstate->current_fh;
2202         const struct nfsd4_layout_ops *ops;
2203         struct nfs4_layout_stateid *ls;
2204         __be32 nfserr;
2205         int accmode = NFSD_MAY_READ_IF_EXEC;
2206
2207         switch (lgp->lg_seg.iomode) {
2208         case IOMODE_READ:
2209                 accmode |= NFSD_MAY_READ;
2210                 break;
2211         case IOMODE_RW:
2212                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2213                 break;
2214         default:
2215                 dprintk("%s: invalid iomode %d\n",
2216                         __func__, lgp->lg_seg.iomode);
2217                 nfserr = nfserr_badiomode;
2218                 goto out;
2219         }
2220
2221         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2222         if (nfserr)
2223                 goto out;
2224
2225         nfserr = nfserr_layoutunavailable;
2226         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2227         if (!ops)
2228                 goto out;
2229
2230         /*
2231          * Verify minlength and range as per RFC5661:
2232          *  o  If loga_length is less than loga_minlength,
2233          *     the metadata server MUST return NFS4ERR_INVAL.
2234          *  o  If the sum of loga_offset and loga_minlength exceeds
2235          *     NFS4_UINT64_MAX, and loga_minlength is not
2236          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2237          *  o  If the sum of loga_offset and loga_length exceeds
2238          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2239          *     the error NFS4ERR_INVAL MUST result.
2240          */
2241         nfserr = nfserr_inval;
2242         if (lgp->lg_seg.length < lgp->lg_minlength ||
2243             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2244              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2245             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2246              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2247                 goto out;
2248         if (lgp->lg_seg.length == 0)
2249                 goto out;
2250
2251         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2252                                                 true, lgp->lg_layout_type, &ls);
2253         if (nfserr) {
2254                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2255                 goto out;
2256         }
2257
2258         nfserr = nfserr_recallconflict;
2259         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2260                 goto out_put_stid;
2261
2262         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2263                                      current_fh, lgp);
2264         if (nfserr)
2265                 goto out_put_stid;
2266
2267         nfserr = nfsd4_insert_layout(lgp, ls);
2268
2269 out_put_stid:
2270         mutex_unlock(&ls->ls_mutex);
2271         nfs4_put_stid(&ls->ls_stid);
2272 out:
2273         return nfserr;
2274 }
2275
2276 static void
2277 nfsd4_layoutget_release(union nfsd4_op_u *u)
2278 {
2279         kfree(u->layoutget.lg_content);
2280 }
2281
2282 static __be32
2283 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2284                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2285 {
2286         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2287         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2288         struct svc_fh *current_fh = &cstate->current_fh;
2289         const struct nfsd4_layout_ops *ops;
2290         loff_t new_size = lcp->lc_last_wr + 1;
2291         struct inode *inode;
2292         struct nfs4_layout_stateid *ls;
2293         __be32 nfserr;
2294
2295         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2296         if (nfserr)
2297                 goto out;
2298
2299         nfserr = nfserr_layoutunavailable;
2300         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2301         if (!ops)
2302                 goto out;
2303         inode = d_inode(current_fh->fh_dentry);
2304
2305         nfserr = nfserr_inval;
2306         if (new_size <= seg->offset) {
2307                 dprintk("pnfsd: last write before layout segment\n");
2308                 goto out;
2309         }
2310         if (new_size > seg->offset + seg->length) {
2311                 dprintk("pnfsd: last write beyond layout segment\n");
2312                 goto out;
2313         }
2314         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2315                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2316                 goto out;
2317         }
2318
2319         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2320                                                 false, lcp->lc_layout_type,
2321                                                 &ls);
2322         if (nfserr) {
2323                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2324                 /* fixup error code as per RFC5661 */
2325                 if (nfserr == nfserr_bad_stateid)
2326                         nfserr = nfserr_badlayout;
2327                 goto out;
2328         }
2329
2330         /* LAYOUTCOMMIT does not require any serialization */
2331         mutex_unlock(&ls->ls_mutex);
2332
2333         if (new_size > i_size_read(inode)) {
2334                 lcp->lc_size_chg = 1;
2335                 lcp->lc_newsize = new_size;
2336         } else {
2337                 lcp->lc_size_chg = 0;
2338         }
2339
2340         nfserr = ops->proc_layoutcommit(inode, lcp);
2341         nfs4_put_stid(&ls->ls_stid);
2342 out:
2343         return nfserr;
2344 }
2345
2346 static __be32
2347 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2348                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2349 {
2350         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2351         struct svc_fh *current_fh = &cstate->current_fh;
2352         __be32 nfserr;
2353
2354         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2355         if (nfserr)
2356                 goto out;
2357
2358         nfserr = nfserr_layoutunavailable;
2359         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2360                 goto out;
2361
2362         switch (lrp->lr_seg.iomode) {
2363         case IOMODE_READ:
2364         case IOMODE_RW:
2365         case IOMODE_ANY:
2366                 break;
2367         default:
2368                 dprintk("%s: invalid iomode %d\n", __func__,
2369                         lrp->lr_seg.iomode);
2370                 nfserr = nfserr_inval;
2371                 goto out;
2372         }
2373
2374         switch (lrp->lr_return_type) {
2375         case RETURN_FILE:
2376                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2377                 break;
2378         case RETURN_FSID:
2379         case RETURN_ALL:
2380                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2381                 break;
2382         default:
2383                 dprintk("%s: invalid return_type %d\n", __func__,
2384                         lrp->lr_return_type);
2385                 nfserr = nfserr_inval;
2386                 break;
2387         }
2388 out:
2389         return nfserr;
2390 }
2391 #endif /* CONFIG_NFSD_PNFS */
2392
2393 static __be32
2394 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2395                union nfsd4_op_u *u)
2396 {
2397         struct nfsd4_getxattr *getxattr = &u->getxattr;
2398
2399         return nfsd_getxattr(rqstp, &cstate->current_fh,
2400                              getxattr->getxa_name, &getxattr->getxa_buf,
2401                              &getxattr->getxa_len);
2402 }
2403
2404 static __be32
2405 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2406            union nfsd4_op_u *u)
2407 {
2408         struct nfsd4_setxattr *setxattr = &u->setxattr;
2409         __be32 ret;
2410
2411         if (opens_in_grace(SVC_NET(rqstp)))
2412                 return nfserr_grace;
2413
2414         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2415                             setxattr->setxa_buf, setxattr->setxa_len,
2416                             setxattr->setxa_flags);
2417
2418         if (!ret)
2419                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2420
2421         return ret;
2422 }
2423
2424 static __be32
2425 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2426            union nfsd4_op_u *u)
2427 {
2428         /*
2429          * Get the entire list, then copy out only the user attributes
2430          * in the encode function.
2431          */
2432         return nfsd_listxattr(rqstp, &cstate->current_fh,
2433                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2434 }
2435
2436 static __be32
2437 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2438            union nfsd4_op_u *u)
2439 {
2440         struct nfsd4_removexattr *removexattr = &u->removexattr;
2441         __be32 ret;
2442
2443         if (opens_in_grace(SVC_NET(rqstp)))
2444                 return nfserr_grace;
2445
2446         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2447             removexattr->rmxa_name);
2448
2449         if (!ret)
2450                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2451
2452         return ret;
2453 }
2454
2455 /*
2456  * NULL call.
2457  */
2458 static __be32
2459 nfsd4_proc_null(struct svc_rqst *rqstp)
2460 {
2461         return rpc_success;
2462 }
2463
2464 static inline void nfsd4_increment_op_stats(u32 opnum)
2465 {
2466         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2467                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2468 }
2469
2470 static const struct nfsd4_operation nfsd4_ops[];
2471
2472 static const char *nfsd4_op_name(unsigned opnum);
2473
2474 /*
2475  * Enforce NFSv4.1 COMPOUND ordering rules:
2476  *
2477  * Also note, enforced elsewhere:
2478  *      - SEQUENCE other than as first op results in
2479  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2480  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2481  *        (Enforced in nfsd4_bind_conn_to_session().)
2482  *      - DESTROY_SESSION must be the final operation in a compound, if
2483  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2484  *        (Enforced in nfsd4_destroy_session().)
2485  */
2486 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2487 {
2488         struct nfsd4_op *first_op = &args->ops[0];
2489
2490         /* These ordering requirements don't apply to NFSv4.0: */
2491         if (args->minorversion == 0)
2492                 return nfs_ok;
2493         /* This is weird, but OK, not our problem: */
2494         if (args->opcnt == 0)
2495                 return nfs_ok;
2496         if (first_op->status == nfserr_op_illegal)
2497                 return nfs_ok;
2498         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2499                 return nfserr_op_not_in_session;
2500         if (first_op->opnum == OP_SEQUENCE)
2501                 return nfs_ok;
2502         /*
2503          * So first_op is something allowed outside a session, like
2504          * EXCHANGE_ID; but then it has to be the only op in the
2505          * compound:
2506          */
2507         if (args->opcnt != 1)
2508                 return nfserr_not_only_op;
2509         return nfs_ok;
2510 }
2511
2512 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2513 {
2514         return &nfsd4_ops[op->opnum];
2515 }
2516
2517 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2518 {
2519         if (op->opnum == OP_ILLEGAL)
2520                 return false;
2521         return OPDESC(op)->op_flags & OP_CACHEME;
2522 }
2523
2524 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2525 {
2526         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2527         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2528         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2529         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2530         const struct nfsd4_operation *thisd = OPDESC(this);
2531         const struct nfsd4_operation *nextd;
2532
2533         /*
2534          * Most ops check wronsec on our own; only the putfh-like ops
2535          * have special rules.
2536          */
2537         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2538                 return false;
2539         /*
2540          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2541          * put-filehandle operation if we're not going to use the
2542          * result:
2543          */
2544         if (argp->opcnt == resp->opcnt)
2545                 return false;
2546         if (next->opnum == OP_ILLEGAL)
2547                 return false;
2548         nextd = OPDESC(next);
2549         /*
2550          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2551          * errors themselves as necessary; others should check for them
2552          * now:
2553          */
2554         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2555 }
2556
2557 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2558 static void
2559 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2560 {
2561         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2562         struct nfsd4_copy *copy;
2563         struct nfsd4_putfh *putfh;
2564         int i;
2565
2566         /* traverse all operation and if it's a COPY compound, mark the
2567          * source filehandle to skip verification
2568          */
2569         for (i = 0; i < args->opcnt; i++) {
2570                 op = &args->ops[i];
2571                 if (op->opnum == OP_PUTFH)
2572                         current_op = op;
2573                 else if (op->opnum == OP_SAVEFH)
2574                         saved_op = current_op;
2575                 else if (op->opnum == OP_RESTOREFH)
2576                         current_op = saved_op;
2577                 else if (op->opnum == OP_COPY) {
2578                         copy = (struct nfsd4_copy *)&op->u;
2579                         if (!saved_op) {
2580                                 op->status = nfserr_nofilehandle;
2581                                 return;
2582                         }
2583                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2584                         if (nfsd4_ssc_is_inter(copy))
2585                                 putfh->no_verify = true;
2586                 }
2587         }
2588 }
2589 #else
2590 static void
2591 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2592 {
2593 }
2594 #endif
2595
2596 /*
2597  * COMPOUND call.
2598  */
2599 static __be32
2600 nfsd4_proc_compound(struct svc_rqst *rqstp)
2601 {
2602         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2603         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2604         struct nfsd4_op *op;
2605         struct nfsd4_compound_state *cstate = &resp->cstate;
2606         struct svc_fh *current_fh = &cstate->current_fh;
2607         struct svc_fh *save_fh = &cstate->save_fh;
2608         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2609         __be32          status;
2610
2611         resp->xdr = &rqstp->rq_res_stream;
2612         resp->statusp = resp->xdr->p;
2613
2614         /* reserve space for: NFS status code */
2615         xdr_reserve_space(resp->xdr, XDR_UNIT);
2616
2617         /* reserve space for: taglen, tag, and opcnt */
2618         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2619         resp->taglen = args->taglen;
2620         resp->tag = args->tag;
2621         resp->rqstp = rqstp;
2622         cstate->minorversion = args->minorversion;
2623         fh_init(current_fh, NFS4_FHSIZE);
2624         fh_init(save_fh, NFS4_FHSIZE);
2625
2626         /*
2627          * Don't use the deferral mechanism for NFSv4; compounds make it
2628          * too hard to avoid non-idempotency problems.
2629          */
2630         __clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2631
2632         /*
2633          * According to RFC3010, this takes precedence over all other errors.
2634          */
2635         status = nfserr_minor_vers_mismatch;
2636         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2637                 goto out;
2638
2639         status = nfs41_check_op_ordering(args);
2640         if (status) {
2641                 op = &args->ops[0];
2642                 op->status = status;
2643                 resp->opcnt = 1;
2644                 goto encode_op;
2645         }
2646         check_if_stalefh_allowed(args);
2647
2648         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2649
2650         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2651         while (!status && resp->opcnt < args->opcnt) {
2652                 op = &args->ops[resp->opcnt++];
2653
2654                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2655                         /* If there are still more operations to process,
2656                          * stop here and report NFS4ERR_RESOURCE. */
2657                         if (cstate->minorversion == 0 &&
2658                             args->client_opcnt > resp->opcnt) {
2659                                 op->status = nfserr_resource;
2660                                 goto encode_op;
2661                         }
2662                 }
2663
2664                 /*
2665                  * The XDR decode routines may have pre-set op->status;
2666                  * for example, if there is a miscellaneous XDR error
2667                  * it will be set to nfserr_bad_xdr.
2668                  */
2669                 if (op->status) {
2670                         if (op->opnum == OP_OPEN)
2671                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2672                         goto encode_op;
2673                 }
2674                 if (!current_fh->fh_dentry &&
2675                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2676                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2677                                 op->status = nfserr_nofilehandle;
2678                                 goto encode_op;
2679                         }
2680                 } else if (current_fh->fh_export &&
2681                            current_fh->fh_export->ex_fslocs.migrated &&
2682                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2683                         op->status = nfserr_moved;
2684                         goto encode_op;
2685                 }
2686
2687                 fh_clear_pre_post_attrs(current_fh);
2688
2689                 /* If op is non-idempotent */
2690                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2691                         /*
2692                          * Don't execute this op if we couldn't encode a
2693                          * successful reply:
2694                          */
2695                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2696                         /*
2697                          * Plus if there's another operation, make sure
2698                          * we'll have space to at least encode an error:
2699                          */
2700                         if (resp->opcnt < args->opcnt)
2701                                 plen += COMPOUND_ERR_SLACK_SPACE;
2702                         op->status = nfsd4_check_resp_size(resp, plen);
2703                 }
2704
2705                 if (op->status)
2706                         goto encode_op;
2707
2708                 if (op->opdesc->op_get_currentstateid)
2709                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2710                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2711
2712                 /* Only from SEQUENCE */
2713                 if (cstate->status == nfserr_replay_cache) {
2714                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2715                         status = op->status;
2716                         goto out;
2717                 }
2718                 if (!op->status) {
2719                         if (op->opdesc->op_set_currentstateid)
2720                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2721
2722                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2723                                 clear_current_stateid(cstate);
2724
2725                         if (current_fh->fh_export &&
2726                                         need_wrongsec_check(rqstp))
2727                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2728                 }
2729 encode_op:
2730                 if (op->status == nfserr_replay_me) {
2731                         op->replay = &cstate->replay_owner->so_replay;
2732                         nfsd4_encode_replay(resp->xdr, op);
2733                         status = op->status = op->replay->rp_status;
2734                 } else {
2735                         nfsd4_encode_operation(resp, op);
2736                         status = op->status;
2737                 }
2738
2739                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2740                                            status, nfsd4_op_name(op->opnum));
2741
2742                 nfsd4_cstate_clear_replay(cstate);
2743                 nfsd4_increment_op_stats(op->opnum);
2744         }
2745
2746         fh_put(current_fh);
2747         fh_put(save_fh);
2748         BUG_ON(cstate->replay_owner);
2749 out:
2750         cstate->status = status;
2751         /* Reset deferral mechanism for RPC deferrals */
2752         __set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2753         return rpc_success;
2754 }
2755
2756 #define op_encode_hdr_size              (2)
2757 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2758 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2759 #define op_encode_change_info_maxsz     (5)
2760 #define nfs4_fattr_bitmap_maxsz         (4)
2761
2762 /* We'll fall back on returning no lockowner if run out of space: */
2763 #define op_encode_lockowner_maxsz       (0)
2764 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2765
2766 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2767
2768 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2769 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2770                                          op_encode_ace_maxsz)
2771
2772 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2773
2774 /*
2775  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2776  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2777  * the maximum payload size here, based on transport limits and the size
2778  * of the remaining space in the rq_pages array.
2779  */
2780 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2781 {
2782         u32 buflen;
2783
2784         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2785         buflen -= rqstp->rq_auth_slack;
2786         buflen -= rqstp->rq_res.head[0].iov_len;
2787         return min_t(u32, buflen, svc_max_payload(rqstp));
2788 }
2789
2790 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2791                                    const struct nfsd4_op *op)
2792 {
2793         return (op_encode_hdr_size) * sizeof(__be32);
2794 }
2795
2796 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2797                                       const struct nfsd4_op *op)
2798 {
2799         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2800 }
2801
2802 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2803                               const struct nfsd4_op *op)
2804 {
2805         /* ac_supported, ac_resp_access */
2806         return (op_encode_hdr_size + 2)* sizeof(__be32);
2807 }
2808
2809 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2810                               const struct nfsd4_op *op)
2811 {
2812         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2813 }
2814
2815 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2816                               const struct nfsd4_op *op)
2817 {
2818         return (op_encode_hdr_size + op_encode_change_info_maxsz
2819                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2820 }
2821
2822 /*
2823  * Note since this is an idempotent operation we won't insist on failing
2824  * the op prematurely if the estimate is too large.  We may turn off splice
2825  * reads unnecessarily.
2826  */
2827 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2828                                const struct nfsd4_op *op)
2829 {
2830         const u32 *bmap = op->u.getattr.ga_bmval;
2831         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2832         u32 ret = 0;
2833
2834         if (bmap0 & FATTR4_WORD0_ACL)
2835                 return nfsd4_max_payload(rqstp);
2836         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2837                 return nfsd4_max_payload(rqstp);
2838
2839         if (bmap1 & FATTR4_WORD1_OWNER) {
2840                 ret += IDMAP_NAMESZ + 4;
2841                 bmap1 &= ~FATTR4_WORD1_OWNER;
2842         }
2843         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2844                 ret += IDMAP_NAMESZ + 4;
2845                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2846         }
2847         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2848                 ret += NFS4_FHSIZE + 4;
2849                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2850         }
2851         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2852                 ret += NFS4_MAXLABELLEN + 12;
2853                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2854         }
2855         /*
2856          * Largest of remaining attributes are 16 bytes (e.g.,
2857          * supported_attributes)
2858          */
2859         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2860         /* bitmask, length */
2861         ret += 20;
2862         return ret;
2863 }
2864
2865 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2866                              const struct nfsd4_op *op)
2867 {
2868         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2869 }
2870
2871 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2872                             const struct nfsd4_op *op)
2873 {
2874         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2875                 * sizeof(__be32);
2876 }
2877
2878 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2879                             const struct nfsd4_op *op)
2880 {
2881         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2882                 * sizeof(__be32);
2883 }
2884
2885 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2886                             const struct nfsd4_op *op)
2887 {
2888         return (op_encode_hdr_size + op_encode_stateid_maxsz
2889                 + op_encode_change_info_maxsz + 1
2890                 + nfs4_fattr_bitmap_maxsz
2891                 + op_encode_delegation_maxsz) * sizeof(__be32);
2892 }
2893
2894 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2895                             const struct nfsd4_op *op)
2896 {
2897         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2898
2899         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2900 }
2901
2902 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2903                                  const struct nfsd4_op *op)
2904 {
2905         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2906         /*
2907          * If we detect that the file changed during hole encoding, then we
2908          * recover by encoding the remaining reply as data. This means we need
2909          * to set aside enough room to encode two data segments.
2910          */
2911         u32 seg_len = 2 * (1 + 2 + 1);
2912
2913         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2914 }
2915
2916 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2917                                const struct nfsd4_op *op)
2918 {
2919         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2920
2921         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2922                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2923 }
2924
2925 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2926                                 const struct nfsd4_op *op)
2927 {
2928         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2929 }
2930
2931 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2932                               const struct nfsd4_op *op)
2933 {
2934         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2935                 * sizeof(__be32);
2936 }
2937
2938 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2939                               const struct nfsd4_op *op)
2940 {
2941         return (op_encode_hdr_size + op_encode_change_info_maxsz
2942                 + op_encode_change_info_maxsz) * sizeof(__be32);
2943 }
2944
2945 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2946                                 const struct nfsd4_op *op)
2947 {
2948         return (op_encode_hdr_size
2949                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2950 }
2951
2952 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2953                                     const struct nfsd4_op *op)
2954 {
2955         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2956                 * sizeof(__be32);
2957 }
2958
2959 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2960                                const struct nfsd4_op *op)
2961 {
2962         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2963 }
2964
2965 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2966                                const struct nfsd4_op *op)
2967 {
2968         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2969                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2970 }
2971
2972 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2973                                    const struct nfsd4_op *op)
2974 {
2975         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2976                                                                 sizeof(__be32);
2977 }
2978
2979 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
2980                              const struct nfsd4_op *op)
2981 {
2982         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2983 }
2984
2985 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
2986                                    const struct nfsd4_op *op)
2987 {
2988         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2989                 1 + 1 + /* eir_flags, spr_how */\
2990                 4 + /* spo_must_enforce & _allow with bitmap */\
2991                 2 + /*eir_server_owner.so_minor_id */\
2992                 /* eir_server_owner.so_major_id<> */\
2993                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2994                 /* eir_server_scope<> */\
2995                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2996                 1 + /* eir_server_impl_id array length */\
2997                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2998 }
2999
3000 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3001                                             const struct nfsd4_op *op)
3002 {
3003         return (op_encode_hdr_size + \
3004                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3005                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3006 }
3007
3008 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3009                                       const struct nfsd4_op *op)
3010 {
3011         return (op_encode_hdr_size + \
3012                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3013                 2 + /* csr_sequence, csr_flags */\
3014                 op_encode_channel_attrs_maxsz + \
3015                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3016 }
3017
3018 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3019                             const struct nfsd4_op *op)
3020 {
3021         return (op_encode_hdr_size +
3022                 1 /* wr_callback */ +
3023                 op_encode_stateid_maxsz /* wr_callback */ +
3024                 2 /* wr_count */ +
3025                 1 /* wr_committed */ +
3026                 op_encode_verifier_maxsz +
3027                 1 /* cr_consecutive */ +
3028                 1 /* cr_synchronous */) * sizeof(__be32);
3029 }
3030
3031 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3032                                       const struct nfsd4_op *op)
3033 {
3034         return (op_encode_hdr_size +
3035                 2 /* osr_count */ +
3036                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3037 }
3038
3039 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3040                                    const struct nfsd4_op *op)
3041 {
3042         return (op_encode_hdr_size +
3043                 3 /* cnr_lease_time */ +
3044                 1 /* We support one cnr_source_server */ +
3045                 1 /* cnr_stateid seq */ +
3046                 op_encode_stateid_maxsz /* cnr_stateid */ +
3047                 1 /* num cnr_source_server*/ +
3048                 1 /* nl4_type */ +
3049                 1 /* nl4 size */ +
3050                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3051                 * sizeof(__be32);
3052 }
3053
3054 #ifdef CONFIG_NFSD_PNFS
3055 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3056                                      const struct nfsd4_op *op)
3057 {
3058         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3059
3060         return (op_encode_hdr_size +
3061                 1 /* gd_layout_type*/ +
3062                 XDR_QUADLEN(rlen) +
3063                 2 /* gd_notify_types */) * sizeof(__be32);
3064 }
3065
3066 /*
3067  * At this stage we don't really know what layout driver will handle the request,
3068  * so we need to define an arbitrary upper bound here.
3069  */
3070 #define MAX_LAYOUT_SIZE         128
3071 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3072                                  const struct nfsd4_op *op)
3073 {
3074         return (op_encode_hdr_size +
3075                 1 /* logr_return_on_close */ +
3076                 op_encode_stateid_maxsz +
3077                 1 /* nr of layouts */ +
3078                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3079 }
3080
3081 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3082                                     const struct nfsd4_op *op)
3083 {
3084         return (op_encode_hdr_size +
3085                 1 /* locr_newsize */ +
3086                 2 /* ns_size */) * sizeof(__be32);
3087 }
3088
3089 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3090                                     const struct nfsd4_op *op)
3091 {
3092         return (op_encode_hdr_size +
3093                 1 /* lrs_stateid */ +
3094                 op_encode_stateid_maxsz) * sizeof(__be32);
3095 }
3096 #endif /* CONFIG_NFSD_PNFS */
3097
3098
3099 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3100                             const struct nfsd4_op *op)
3101 {
3102         return (op_encode_hdr_size + 3) * sizeof(__be32);
3103 }
3104
3105 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3106                                 const struct nfsd4_op *op)
3107 {
3108         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3109
3110         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3111 }
3112
3113 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3114                                 const struct nfsd4_op *op)
3115 {
3116         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3117                 * sizeof(__be32);
3118 }
3119 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3120                                   const struct nfsd4_op *op)
3121 {
3122         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3123
3124         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3125 }
3126
3127 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3128                                    const struct nfsd4_op *op)
3129 {
3130         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3131                 * sizeof(__be32);
3132 }
3133
3134
3135 static const struct nfsd4_operation nfsd4_ops[] = {
3136         [OP_ACCESS] = {
3137                 .op_func = nfsd4_access,
3138                 .op_name = "OP_ACCESS",
3139                 .op_rsize_bop = nfsd4_access_rsize,
3140         },
3141         [OP_CLOSE] = {
3142                 .op_func = nfsd4_close,
3143                 .op_flags = OP_MODIFIES_SOMETHING,
3144                 .op_name = "OP_CLOSE",
3145                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3146                 .op_get_currentstateid = nfsd4_get_closestateid,
3147                 .op_set_currentstateid = nfsd4_set_closestateid,
3148         },
3149         [OP_COMMIT] = {
3150                 .op_func = nfsd4_commit,
3151                 .op_flags = OP_MODIFIES_SOMETHING,
3152                 .op_name = "OP_COMMIT",
3153                 .op_rsize_bop = nfsd4_commit_rsize,
3154         },
3155         [OP_CREATE] = {
3156                 .op_func = nfsd4_create,
3157                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3158                 .op_name = "OP_CREATE",
3159                 .op_rsize_bop = nfsd4_create_rsize,
3160         },
3161         [OP_DELEGRETURN] = {
3162                 .op_func = nfsd4_delegreturn,
3163                 .op_flags = OP_MODIFIES_SOMETHING,
3164                 .op_name = "OP_DELEGRETURN",
3165                 .op_rsize_bop = nfsd4_only_status_rsize,
3166                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3167         },
3168         [OP_GETATTR] = {
3169                 .op_func = nfsd4_getattr,
3170                 .op_flags = ALLOWED_ON_ABSENT_FS,
3171                 .op_rsize_bop = nfsd4_getattr_rsize,
3172                 .op_name = "OP_GETATTR",
3173         },
3174         [OP_GETFH] = {
3175                 .op_func = nfsd4_getfh,
3176                 .op_name = "OP_GETFH",
3177                 .op_rsize_bop = nfsd4_getfh_rsize,
3178         },
3179         [OP_LINK] = {
3180                 .op_func = nfsd4_link,
3181                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3182                                 | OP_CACHEME,
3183                 .op_name = "OP_LINK",
3184                 .op_rsize_bop = nfsd4_link_rsize,
3185         },
3186         [OP_LOCK] = {
3187                 .op_func = nfsd4_lock,
3188                 .op_flags = OP_MODIFIES_SOMETHING |
3189                                 OP_NONTRIVIAL_ERROR_ENCODE,
3190                 .op_name = "OP_LOCK",
3191                 .op_rsize_bop = nfsd4_lock_rsize,
3192                 .op_set_currentstateid = nfsd4_set_lockstateid,
3193         },
3194         [OP_LOCKT] = {
3195                 .op_func = nfsd4_lockt,
3196                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3197                 .op_name = "OP_LOCKT",
3198                 .op_rsize_bop = nfsd4_lock_rsize,
3199         },
3200         [OP_LOCKU] = {
3201                 .op_func = nfsd4_locku,
3202                 .op_flags = OP_MODIFIES_SOMETHING,
3203                 .op_name = "OP_LOCKU",
3204                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3205                 .op_get_currentstateid = nfsd4_get_lockustateid,
3206         },
3207         [OP_LOOKUP] = {
3208                 .op_func = nfsd4_lookup,
3209                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3210                 .op_name = "OP_LOOKUP",
3211                 .op_rsize_bop = nfsd4_only_status_rsize,
3212         },
3213         [OP_LOOKUPP] = {
3214                 .op_func = nfsd4_lookupp,
3215                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3216                 .op_name = "OP_LOOKUPP",
3217                 .op_rsize_bop = nfsd4_only_status_rsize,
3218         },
3219         [OP_NVERIFY] = {
3220                 .op_func = nfsd4_nverify,
3221                 .op_name = "OP_NVERIFY",
3222                 .op_rsize_bop = nfsd4_only_status_rsize,
3223         },
3224         [OP_OPEN] = {
3225                 .op_func = nfsd4_open,
3226                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3227                 .op_name = "OP_OPEN",
3228                 .op_rsize_bop = nfsd4_open_rsize,
3229                 .op_set_currentstateid = nfsd4_set_openstateid,
3230         },
3231         [OP_OPEN_CONFIRM] = {
3232                 .op_func = nfsd4_open_confirm,
3233                 .op_flags = OP_MODIFIES_SOMETHING,
3234                 .op_name = "OP_OPEN_CONFIRM",
3235                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3236         },
3237         [OP_OPEN_DOWNGRADE] = {
3238                 .op_func = nfsd4_open_downgrade,
3239                 .op_flags = OP_MODIFIES_SOMETHING,
3240                 .op_name = "OP_OPEN_DOWNGRADE",
3241                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3242                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3243                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3244         },
3245         [OP_PUTFH] = {
3246                 .op_func = nfsd4_putfh,
3247                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3248                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3249                 .op_name = "OP_PUTFH",
3250                 .op_rsize_bop = nfsd4_only_status_rsize,
3251         },
3252         [OP_PUTPUBFH] = {
3253                 .op_func = nfsd4_putrootfh,
3254                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3255                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3256                 .op_name = "OP_PUTPUBFH",
3257                 .op_rsize_bop = nfsd4_only_status_rsize,
3258         },
3259         [OP_PUTROOTFH] = {
3260                 .op_func = nfsd4_putrootfh,
3261                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3262                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3263                 .op_name = "OP_PUTROOTFH",
3264                 .op_rsize_bop = nfsd4_only_status_rsize,
3265         },
3266         [OP_READ] = {
3267                 .op_func = nfsd4_read,
3268                 .op_release = nfsd4_read_release,
3269                 .op_name = "OP_READ",
3270                 .op_rsize_bop = nfsd4_read_rsize,
3271                 .op_get_currentstateid = nfsd4_get_readstateid,
3272         },
3273         [OP_READDIR] = {
3274                 .op_func = nfsd4_readdir,
3275                 .op_name = "OP_READDIR",
3276                 .op_rsize_bop = nfsd4_readdir_rsize,
3277         },
3278         [OP_READLINK] = {
3279                 .op_func = nfsd4_readlink,
3280                 .op_name = "OP_READLINK",
3281                 .op_rsize_bop = nfsd4_readlink_rsize,
3282         },
3283         [OP_REMOVE] = {
3284                 .op_func = nfsd4_remove,
3285                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3286                 .op_name = "OP_REMOVE",
3287                 .op_rsize_bop = nfsd4_remove_rsize,
3288         },
3289         [OP_RENAME] = {
3290                 .op_func = nfsd4_rename,
3291                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3292                 .op_name = "OP_RENAME",
3293                 .op_rsize_bop = nfsd4_rename_rsize,
3294         },
3295         [OP_RENEW] = {
3296                 .op_func = nfsd4_renew,
3297                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3298                                 | OP_MODIFIES_SOMETHING,
3299                 .op_name = "OP_RENEW",
3300                 .op_rsize_bop = nfsd4_only_status_rsize,
3301
3302         },
3303         [OP_RESTOREFH] = {
3304                 .op_func = nfsd4_restorefh,
3305                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3306                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3307                 .op_name = "OP_RESTOREFH",
3308                 .op_rsize_bop = nfsd4_only_status_rsize,
3309         },
3310         [OP_SAVEFH] = {
3311                 .op_func = nfsd4_savefh,
3312                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3313                 .op_name = "OP_SAVEFH",
3314                 .op_rsize_bop = nfsd4_only_status_rsize,
3315         },
3316         [OP_SECINFO] = {
3317                 .op_func = nfsd4_secinfo,
3318                 .op_release = nfsd4_secinfo_release,
3319                 .op_flags = OP_HANDLES_WRONGSEC,
3320                 .op_name = "OP_SECINFO",
3321                 .op_rsize_bop = nfsd4_secinfo_rsize,
3322         },
3323         [OP_SETATTR] = {
3324                 .op_func = nfsd4_setattr,
3325                 .op_name = "OP_SETATTR",
3326                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3327                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3328                 .op_rsize_bop = nfsd4_setattr_rsize,
3329                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3330         },
3331         [OP_SETCLIENTID] = {
3332                 .op_func = nfsd4_setclientid,
3333                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3334                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3335                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3336                 .op_name = "OP_SETCLIENTID",
3337                 .op_rsize_bop = nfsd4_setclientid_rsize,
3338         },
3339         [OP_SETCLIENTID_CONFIRM] = {
3340                 .op_func = nfsd4_setclientid_confirm,
3341                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3342                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3343                 .op_name = "OP_SETCLIENTID_CONFIRM",
3344                 .op_rsize_bop = nfsd4_only_status_rsize,
3345         },
3346         [OP_VERIFY] = {
3347                 .op_func = nfsd4_verify,
3348                 .op_name = "OP_VERIFY",
3349                 .op_rsize_bop = nfsd4_only_status_rsize,
3350         },
3351         [OP_WRITE] = {
3352                 .op_func = nfsd4_write,
3353                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3354                 .op_name = "OP_WRITE",
3355                 .op_rsize_bop = nfsd4_write_rsize,
3356                 .op_get_currentstateid = nfsd4_get_writestateid,
3357         },
3358         [OP_RELEASE_LOCKOWNER] = {
3359                 .op_func = nfsd4_release_lockowner,
3360                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3361                                 | OP_MODIFIES_SOMETHING,
3362                 .op_name = "OP_RELEASE_LOCKOWNER",
3363                 .op_rsize_bop = nfsd4_only_status_rsize,
3364         },
3365
3366         /* NFSv4.1 operations */
3367         [OP_EXCHANGE_ID] = {
3368                 .op_func = nfsd4_exchange_id,
3369                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3370                                 | OP_MODIFIES_SOMETHING,
3371                 .op_name = "OP_EXCHANGE_ID",
3372                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3373         },
3374         [OP_BACKCHANNEL_CTL] = {
3375                 .op_func = nfsd4_backchannel_ctl,
3376                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3377                 .op_name = "OP_BACKCHANNEL_CTL",
3378                 .op_rsize_bop = nfsd4_only_status_rsize,
3379         },
3380         [OP_BIND_CONN_TO_SESSION] = {
3381                 .op_func = nfsd4_bind_conn_to_session,
3382                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3383                                 | OP_MODIFIES_SOMETHING,
3384                 .op_name = "OP_BIND_CONN_TO_SESSION",
3385                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3386         },
3387         [OP_CREATE_SESSION] = {
3388                 .op_func = nfsd4_create_session,
3389                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3390                                 | OP_MODIFIES_SOMETHING,
3391                 .op_name = "OP_CREATE_SESSION",
3392                 .op_rsize_bop = nfsd4_create_session_rsize,
3393         },
3394         [OP_DESTROY_SESSION] = {
3395                 .op_func = nfsd4_destroy_session,
3396                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3397                                 | OP_MODIFIES_SOMETHING,
3398                 .op_name = "OP_DESTROY_SESSION",
3399                 .op_rsize_bop = nfsd4_only_status_rsize,
3400         },
3401         [OP_SEQUENCE] = {
3402                 .op_func = nfsd4_sequence,
3403                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3404                 .op_name = "OP_SEQUENCE",
3405                 .op_rsize_bop = nfsd4_sequence_rsize,
3406         },
3407         [OP_DESTROY_CLIENTID] = {
3408                 .op_func = nfsd4_destroy_clientid,
3409                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3410                                 | OP_MODIFIES_SOMETHING,
3411                 .op_name = "OP_DESTROY_CLIENTID",
3412                 .op_rsize_bop = nfsd4_only_status_rsize,
3413         },
3414         [OP_RECLAIM_COMPLETE] = {
3415                 .op_func = nfsd4_reclaim_complete,
3416                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3417                 .op_name = "OP_RECLAIM_COMPLETE",
3418                 .op_rsize_bop = nfsd4_only_status_rsize,
3419         },
3420         [OP_SECINFO_NO_NAME] = {
3421                 .op_func = nfsd4_secinfo_no_name,
3422                 .op_release = nfsd4_secinfo_no_name_release,
3423                 .op_flags = OP_HANDLES_WRONGSEC,
3424                 .op_name = "OP_SECINFO_NO_NAME",
3425                 .op_rsize_bop = nfsd4_secinfo_rsize,
3426         },
3427         [OP_TEST_STATEID] = {
3428                 .op_func = nfsd4_test_stateid,
3429                 .op_flags = ALLOWED_WITHOUT_FH,
3430                 .op_name = "OP_TEST_STATEID",
3431                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3432         },
3433         [OP_FREE_STATEID] = {
3434                 .op_func = nfsd4_free_stateid,
3435                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3436                 .op_name = "OP_FREE_STATEID",
3437                 .op_get_currentstateid = nfsd4_get_freestateid,
3438                 .op_rsize_bop = nfsd4_only_status_rsize,
3439         },
3440 #ifdef CONFIG_NFSD_PNFS
3441         [OP_GETDEVICEINFO] = {
3442                 .op_func = nfsd4_getdeviceinfo,
3443                 .op_release = nfsd4_getdeviceinfo_release,
3444                 .op_flags = ALLOWED_WITHOUT_FH,
3445                 .op_name = "OP_GETDEVICEINFO",
3446                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3447         },
3448         [OP_LAYOUTGET] = {
3449                 .op_func = nfsd4_layoutget,
3450                 .op_release = nfsd4_layoutget_release,
3451                 .op_flags = OP_MODIFIES_SOMETHING,
3452                 .op_name = "OP_LAYOUTGET",
3453                 .op_rsize_bop = nfsd4_layoutget_rsize,
3454         },
3455         [OP_LAYOUTCOMMIT] = {
3456                 .op_func = nfsd4_layoutcommit,
3457                 .op_flags = OP_MODIFIES_SOMETHING,
3458                 .op_name = "OP_LAYOUTCOMMIT",
3459                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3460         },
3461         [OP_LAYOUTRETURN] = {
3462                 .op_func = nfsd4_layoutreturn,
3463                 .op_flags = OP_MODIFIES_SOMETHING,
3464                 .op_name = "OP_LAYOUTRETURN",
3465                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3466         },
3467 #endif /* CONFIG_NFSD_PNFS */
3468
3469         /* NFSv4.2 operations */
3470         [OP_ALLOCATE] = {
3471                 .op_func = nfsd4_allocate,
3472                 .op_flags = OP_MODIFIES_SOMETHING,
3473                 .op_name = "OP_ALLOCATE",
3474                 .op_rsize_bop = nfsd4_only_status_rsize,
3475         },
3476         [OP_DEALLOCATE] = {
3477                 .op_func = nfsd4_deallocate,
3478                 .op_flags = OP_MODIFIES_SOMETHING,
3479                 .op_name = "OP_DEALLOCATE",
3480                 .op_rsize_bop = nfsd4_only_status_rsize,
3481         },
3482         [OP_CLONE] = {
3483                 .op_func = nfsd4_clone,
3484                 .op_flags = OP_MODIFIES_SOMETHING,
3485                 .op_name = "OP_CLONE",
3486                 .op_rsize_bop = nfsd4_only_status_rsize,
3487         },
3488         [OP_COPY] = {
3489                 .op_func = nfsd4_copy,
3490                 .op_flags = OP_MODIFIES_SOMETHING,
3491                 .op_name = "OP_COPY",
3492                 .op_rsize_bop = nfsd4_copy_rsize,
3493         },
3494         [OP_READ_PLUS] = {
3495                 .op_func = nfsd4_read,
3496                 .op_release = nfsd4_read_release,
3497                 .op_name = "OP_READ_PLUS",
3498                 .op_rsize_bop = nfsd4_read_plus_rsize,
3499                 .op_get_currentstateid = nfsd4_get_readstateid,
3500         },
3501         [OP_SEEK] = {
3502                 .op_func = nfsd4_seek,
3503                 .op_name = "OP_SEEK",
3504                 .op_rsize_bop = nfsd4_seek_rsize,
3505         },
3506         [OP_OFFLOAD_STATUS] = {
3507                 .op_func = nfsd4_offload_status,
3508                 .op_name = "OP_OFFLOAD_STATUS",
3509                 .op_rsize_bop = nfsd4_offload_status_rsize,
3510         },
3511         [OP_OFFLOAD_CANCEL] = {
3512                 .op_func = nfsd4_offload_cancel,
3513                 .op_flags = OP_MODIFIES_SOMETHING,
3514                 .op_name = "OP_OFFLOAD_CANCEL",
3515                 .op_rsize_bop = nfsd4_only_status_rsize,
3516         },
3517         [OP_COPY_NOTIFY] = {
3518                 .op_func = nfsd4_copy_notify,
3519                 .op_flags = OP_MODIFIES_SOMETHING,
3520                 .op_name = "OP_COPY_NOTIFY",
3521                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3522         },
3523         [OP_GETXATTR] = {
3524                 .op_func = nfsd4_getxattr,
3525                 .op_name = "OP_GETXATTR",
3526                 .op_rsize_bop = nfsd4_getxattr_rsize,
3527         },
3528         [OP_SETXATTR] = {
3529                 .op_func = nfsd4_setxattr,
3530                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3531                 .op_name = "OP_SETXATTR",
3532                 .op_rsize_bop = nfsd4_setxattr_rsize,
3533         },
3534         [OP_LISTXATTRS] = {
3535                 .op_func = nfsd4_listxattrs,
3536                 .op_name = "OP_LISTXATTRS",
3537                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3538         },
3539         [OP_REMOVEXATTR] = {
3540                 .op_func = nfsd4_removexattr,
3541                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3542                 .op_name = "OP_REMOVEXATTR",
3543                 .op_rsize_bop = nfsd4_removexattr_rsize,
3544         },
3545 };
3546
3547 /**
3548  * nfsd4_spo_must_allow - Determine if the compound op contains an
3549  * operation that is allowed to be sent with machine credentials
3550  *
3551  * @rqstp: a pointer to the struct svc_rqst
3552  *
3553  * Checks to see if the compound contains a spo_must_allow op
3554  * and confirms that it was sent with the proper machine creds.
3555  */
3556
3557 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3558 {
3559         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3560         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3561         struct nfsd4_op *this;
3562         struct nfsd4_compound_state *cstate = &resp->cstate;
3563         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3564         u32 opiter;
3565
3566         if (!cstate->minorversion)
3567                 return false;
3568
3569         if (cstate->spo_must_allowed)
3570                 return true;
3571
3572         opiter = resp->opcnt;
3573         while (opiter < argp->opcnt) {
3574                 this = &argp->ops[opiter++];
3575                 if (test_bit(this->opnum, allow->u.longs) &&
3576                         cstate->clp->cl_mach_cred &&
3577                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3578                         cstate->spo_must_allowed = true;
3579                         return true;
3580                 }
3581         }
3582         cstate->spo_must_allowed = false;
3583         return false;
3584 }
3585
3586 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3587 {
3588         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3589                 return op_encode_hdr_size * sizeof(__be32);
3590
3591         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3592         return OPDESC(op)->op_rsize_bop(rqstp, op);
3593 }
3594
3595 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3596 {
3597         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3598                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3599                         op->opnum, nfsd4_op_name(op->opnum));
3600                 WARN_ON_ONCE(1);
3601         }
3602 }
3603
3604 static const char *nfsd4_op_name(unsigned opnum)
3605 {
3606         if (opnum < ARRAY_SIZE(nfsd4_ops))
3607                 return nfsd4_ops[opnum].op_name;
3608         return "unknown_operation";
3609 }
3610
3611 static const struct svc_procedure nfsd_procedures4[2] = {
3612         [NFSPROC4_NULL] = {
3613                 .pc_func = nfsd4_proc_null,
3614                 .pc_decode = nfssvc_decode_voidarg,
3615                 .pc_encode = nfssvc_encode_voidres,
3616                 .pc_argsize = sizeof(struct nfsd_voidargs),
3617                 .pc_argzero = sizeof(struct nfsd_voidargs),
3618                 .pc_ressize = sizeof(struct nfsd_voidres),
3619                 .pc_cachetype = RC_NOCACHE,
3620                 .pc_xdrressize = 1,
3621                 .pc_name = "NULL",
3622         },
3623         [NFSPROC4_COMPOUND] = {
3624                 .pc_func = nfsd4_proc_compound,
3625                 .pc_decode = nfs4svc_decode_compoundargs,
3626                 .pc_encode = nfs4svc_encode_compoundres,
3627                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3628                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3629                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3630                 .pc_release = nfsd4_release_compoundargs,
3631                 .pc_cachetype = RC_NOCACHE,
3632                 .pc_xdrressize = NFSD_BUFSIZE/4,
3633                 .pc_name = "COMPOUND",
3634         },
3635 };
3636
3637 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3638 const struct svc_version nfsd_version4 = {
3639         .vs_vers                = 4,
3640         .vs_nproc               = 2,
3641         .vs_proc                = nfsd_procedures4,
3642         .vs_count               = nfsd_count3,
3643         .vs_dispatch            = nfsd_dispatch,
3644         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3645         .vs_rpcb_optnl          = true,
3646         .vs_need_cong_ctrl      = true,
3647 };