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                 /* NFSv4 protocol requires change attributes even though
 301                  * no change happened.
 302                  */
 303                 status = fh_fill_both_attrs(fhp);
 304                 if (status != nfs_ok)
 305                         goto out;
 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 (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
 326                             inode_get_atime_sec(d_inode(child)) == 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 (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
 335                             inode_get_atime_sec(d_inode(child)) == 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         status = fh_fill_pre_attrs(fhp);
 349         if (status != nfs_ok)
 350                 goto out;
 351         status = nfsd4_vfs_create(fhp, child, open);
 352         if (status != nfs_ok)
 353                 goto out;
 354         open->op_created = true;
 355         fh_fill_post_attrs(fhp);
 356
 357         /* A newly created file already has a file size of zero. */
 358         if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
 359                 iap->ia_valid &= ~ATTR_SIZE;
 360         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 361                 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
 362                                 ATTR_MTIME_SET|ATTR_ATIME_SET;
 363                 iap->ia_mtime.tv_sec = v_mtime;
 364                 iap->ia_atime.tv_sec = v_atime;
 365                 iap->ia_mtime.tv_nsec = 0;
 366                 iap->ia_atime.tv_nsec = 0;
 367         }
 368
 369 set_attr:
 370         status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
 371
 372         if (attrs.na_labelerr)
 373                 open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 374         if (attrs.na_aclerr)
 375                 open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
 376 out:
 377         inode_unlock(inode);
 378         nfsd_attrs_free(&attrs);
 379         if (child && !IS_ERR(child))
 380                 dput(child);
 381         fh_drop_write(fhp);
 382         return status;
 383 }
 384
 385 /**
 386  * set_change_info - set up the change_info4 for a reply
 387  * @cinfo: pointer to nfsd4_change_info to be populated
 388  * @fhp: pointer to svc_fh to use as source
 389  *
 390  * Many operations in NFSv4 require change_info4 in the reply. This function
 391  * populates that from the info that we (should!) have already collected. In
 392  * the event that we didn't get any pre-attrs, just zero out both.
 393  */
 394 static void
 395 set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp)
 396 {
 397         cinfo->atomic = (u32)(fhp->fh_pre_saved && fhp->fh_post_saved && !fhp->fh_no_atomic_attr);
 398         cinfo->before_change = fhp->fh_pre_change;
 399         cinfo->after_change = fhp->fh_post_change;
 400
 401         /*
 402          * If fetching the pre-change attributes failed, then we should
 403          * have already failed the whole operation. We could have still
 404          * failed to fetch post-change attributes however.
 405          *
 406          * If we didn't get post-op attrs, just zero-out the after
 407          * field since we don't know what it should be. If the pre_saved
 408          * field isn't set for some reason, throw warning and just copy
 409          * whatever is in the after field.
 410          */
 411         if (WARN_ON_ONCE(!fhp->fh_pre_saved))
 412                 cinfo->before_change = 0;
 413         if (!fhp->fh_post_saved)
 414                 cinfo->after_change = cinfo->before_change + 1;
 415 }
 416
 417 static __be32
 418 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
 419 {
 420         struct svc_fh *current_fh = &cstate->current_fh;
 421         int accmode;
 422         __be32 status;
 423
 424         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
 425         if (!*resfh)
 426                 return nfserr_jukebox;
 427         fh_init(*resfh, NFS4_FHSIZE);
 428         open->op_truncate = false;
 429
 430         if (open->op_create) {
 431                 /* FIXME: check session persistence and pnfs flags.
 432                  * The nfsv4.1 spec requires the following semantics:
 433                  *
 434                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
 435                  * Reply Cache  | server |                 |
 436                  * -------------+--------+-----------------+--------------------
 437                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
 438                  *              |        |                 | (SHOULD)
 439                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
 440                  *              |        |                 | (SHOULD NOT)
 441                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
 442                  * yes          | no     | GUARDED4        | GUARDED4
 443                  * yes          | yes    | GUARDED4        | GUARDED4
 444                  */
 445
 446                 current->fs->umask = open->op_umask;
 447                 status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
 448                 current->fs->umask = 0;
 449
 450                 /*
 451                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
 452                  * use the returned bitmask to indicate which attributes
 453                  * we used to store the verifier:
 454                  */
 455                 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
 456                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
 457                                                 FATTR4_WORD1_TIME_MODIFY);
 458         } else {
 459                 status = nfsd_lookup(rqstp, current_fh,
 460                                      open->op_fname, open->op_fnamelen, *resfh);
 461                 if (status == nfs_ok)
 462                         /* NFSv4 protocol requires change attributes even though
 463                          * no change happened.
 464                          */
 465                         status = fh_fill_both_attrs(current_fh);
 466         }
 467         if (status)
 468                 goto out;
 469         status = nfsd_check_obj_isreg(*resfh);
 470         if (status)
 471                 goto out;
 472
 473         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
 474         accmode = NFSD_MAY_NOP;
 475         if (open->op_created ||
 476                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
 477                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
 478         status = do_open_permission(rqstp, *resfh, open, accmode);
 479         set_change_info(&open->op_cinfo, current_fh);
 480 out:
 481         return status;
 482 }
 483
 484 static __be32
 485 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 486 {
 487         struct svc_fh *current_fh = &cstate->current_fh;
 488         int accmode = 0;
 489
 490         /* We don't know the target directory, and therefore can not
 491         * set the change info
 492         */
 493
 494         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
 495
 496         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
 497
 498         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
 499                 (open->op_iattr.ia_size == 0);
 500         /*
 501          * In the delegation case, the client is telling us about an
 502          * open that it *already* performed locally, some time ago.  We
 503          * should let it succeed now if possible.
 504          *
 505          * In the case of a CLAIM_FH open, on the other hand, the client
 506          * may be counting on us to enforce permissions (the Linux 4.1
 507          * client uses this for normal opens, for example).
 508          */
 509         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
 510                 accmode = NFSD_MAY_OWNER_OVERRIDE;
 511
 512         return do_open_permission(rqstp, current_fh, open, accmode);
 513 }
 514
 515 static void
 516 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
 517 {
 518         struct nfsd4_sessionid *sid =
 519                         (struct nfsd4_sessionid *)session->se_sessionid.data;
 520
 521         clid->cl_boot = sid->clientid.cl_boot;
 522         clid->cl_id = sid->clientid.cl_id;
 523 }
 524
 525 static __be32
 526 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 527            union nfsd4_op_u *u)
 528 {
 529         struct nfsd4_open *open = &u->open;
 530         __be32 status;
 531         struct svc_fh *resfh = NULL;
 532         struct net *net = SVC_NET(rqstp);
 533         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 534         bool reclaim = false;
 535
 536         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
 537                 (int)open->op_fnamelen, open->op_fname,
 538                 open->op_openowner);
 539
 540         open->op_filp = NULL;
 541         open->op_rqstp = rqstp;
 542
 543         /* This check required by spec. */
 544         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
 545                 return nfserr_inval;
 546
 547         open->op_created = false;
 548         /*
 549          * RFC5661 18.51.3
 550          * Before RECLAIM_COMPLETE done, server should deny new lock
 551          */
 552         if (nfsd4_has_session(cstate) &&
 553             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
 554             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 555                 return nfserr_grace;
 556
 557         if (nfsd4_has_session(cstate))
 558                 copy_clientid(&open->op_clientid, cstate->session);
 559
 560         /* check seqid for replay. set nfs4_owner */
 561         status = nfsd4_process_open1(cstate, open, nn);
 562         if (status == nfserr_replay_me) {
 563                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
 564                 fh_put(&cstate->current_fh);
 565                 fh_copy_shallow(&cstate->current_fh.fh_handle,
 566                                 &rp->rp_openfh);
 567                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 568                 if (status)
 569                         dprintk("nfsd4_open: replay failed"
 570                                 " restoring previous filehandle\n");
 571                 else
 572                         status = nfserr_replay_me;
 573         }
 574         if (status)
 575                 goto out;
 576         if (open->op_xdr_error) {
 577                 status = open->op_xdr_error;
 578                 goto out;
 579         }
 580
 581         status = nfsd4_check_open_attributes(rqstp, cstate, open);
 582         if (status)
 583                 goto out;
 584
 585         /* Openowner is now set, so sequence id will get bumped.  Now we need
 586          * these checks before we do any creates: */
 587         status = nfserr_grace;
 588         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 589                 goto out;
 590         status = nfserr_no_grace;
 591         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
 592                 goto out;
 593
 594         switch (open->op_claim_type) {
 595         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 596         case NFS4_OPEN_CLAIM_NULL:
 597                 status = do_open_lookup(rqstp, cstate, open, &resfh);
 598                 if (status)
 599                         goto out;
 600                 break;
 601         case NFS4_OPEN_CLAIM_PREVIOUS:
 602                 status = nfs4_check_open_reclaim(cstate->clp);
 603                 if (status)
 604                         goto out;
 605                 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
 606                 reclaim = true;
 607                 fallthrough;
 608         case NFS4_OPEN_CLAIM_FH:
 609         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 610                 status = do_open_fhandle(rqstp, cstate, open);
 611                 if (status)
 612                         goto out;
 613                 resfh = &cstate->current_fh;
 614                 break;
 615         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 616         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 617                 status = nfserr_notsupp;
 618                 goto out;
 619         default:
 620                 status = nfserr_inval;
 621                 goto out;
 622         }
 623
 624         status = nfsd4_process_open2(rqstp, resfh, open);
 625         if (status && open->op_created)
 626                 pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
 627                         be32_to_cpu(status));
 628         if (reclaim && !status)
 629                 nn->somebody_reclaimed = true;
 630 out:
 631         if (open->op_filp) {
 632                 fput(open->op_filp);
 633                 open->op_filp = NULL;
 634         }
 635         if (resfh && resfh != &cstate->current_fh) {
 636                 fh_dup2(&cstate->current_fh, resfh);
 637                 fh_put(resfh);
 638                 kfree(resfh);
 639         }
 640         nfsd4_cleanup_open_state(cstate, open);
 641         nfsd4_bump_seqid(cstate, status);
 642         return status;
 643 }
 644
 645 /*
 646  * OPEN is the only seqid-mutating operation whose decoding can fail
 647  * with a seqid-mutating error (specifically, decoding of user names in
 648  * the attributes).  Therefore we have to do some processing to look up
 649  * the stateowner so that we can bump the seqid.
 650  */
 651 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
 652 {
 653         struct nfsd4_open *open = &op->u.open;
 654
 655         if (!seqid_mutating_err(ntohl(op->status)))
 656                 return op->status;
 657         if (nfsd4_has_session(cstate))
 658                 return op->status;
 659         open->op_xdr_error = op->status;
 660         return nfsd4_open(rqstp, cstate, &op->u);
 661 }
 662
 663 /*
 664  * filehandle-manipulating ops.
 665  */
 666 static __be32
 667 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 668             union nfsd4_op_u *u)
 669 {
 670         u->getfh = &cstate->current_fh;
 671         return nfs_ok;
 672 }
 673
 674 static __be32
 675 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 676             union nfsd4_op_u *u)
 677 {
 678         struct nfsd4_putfh *putfh = &u->putfh;
 679         __be32 ret;
 680
 681         fh_put(&cstate->current_fh);
 682         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
 683         memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
 684                putfh->pf_fhlen);
 685         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
 686 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
 687         if (ret == nfserr_stale && putfh->no_verify) {
 688                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
 689                 ret = 0;
 690         }
 691 #endif
 692         return ret;
 693 }
 694
 695 static __be32
 696 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 697                 union nfsd4_op_u *u)
 698 {
 699         fh_put(&cstate->current_fh);
 700
 701         return exp_pseudoroot(rqstp, &cstate->current_fh);
 702 }
 703
 704 static __be32
 705 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 706                 union nfsd4_op_u *u)
 707 {
 708         if (!cstate->save_fh.fh_dentry)
 709                 return nfserr_restorefh;
 710
 711         fh_dup2(&cstate->current_fh, &cstate->save_fh);
 712         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
 713                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
 714                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
 715         }
 716         return nfs_ok;
 717 }
 718
 719 static __be32
 720 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 721              union nfsd4_op_u *u)
 722 {
 723         fh_dup2(&cstate->save_fh, &cstate->current_fh);
 724         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
 725                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
 726                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
 727         }
 728         return nfs_ok;
 729 }
 730
 731 /*
 732  * misc nfsv4 ops
 733  */
 734 static __be32
 735 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 736              union nfsd4_op_u *u)
 737 {
 738         struct nfsd4_access *access = &u->access;
 739         u32 access_full;
 740
 741         access_full = NFS3_ACCESS_FULL;
 742         if (cstate->minorversion >= 2)
 743                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
 744                                NFS4_ACCESS_XAWRITE;
 745
 746         if (access->ac_req_access & ~access_full)
 747                 return nfserr_inval;
 748
 749         access->ac_resp_access = access->ac_req_access;
 750         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
 751                            &access->ac_supported);
 752 }
 753
 754 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
 755 {
 756         __be32 *verf = (__be32 *)verifier->data;
 757
 758         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
 759
 760         nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
 761 }
 762
 763 static __be32
 764 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 765              union nfsd4_op_u *u)
 766 {
 767         struct nfsd4_commit *commit = &u->commit;
 768         struct nfsd_file *nf;
 769         __be32 status;
 770
 771         status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
 772                                    NFSD_MAY_NOT_BREAK_LEASE, &nf);
 773         if (status != nfs_ok)
 774                 return status;
 775
 776         status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
 777                              commit->co_count,
 778                              (__be32 *)commit->co_verf.data);
 779         nfsd_file_put(nf);
 780         return status;
 781 }
 782
 783 static __be32
 784 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 785              union nfsd4_op_u *u)
 786 {
 787         struct nfsd4_create *create = &u->create;
 788         struct nfsd_attrs attrs = {
 789                 .na_iattr       = &create->cr_iattr,
 790                 .na_seclabel    = &create->cr_label,
 791         };
 792         struct svc_fh resfh;
 793         __be32 status;
 794         dev_t rdev;
 795
 796         fh_init(&resfh, NFS4_FHSIZE);
 797
 798         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 799         if (status)
 800                 return status;
 801
 802         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 803                                     nfsd_attrmask);
 804         if (status)
 805                 return status;
 806
 807         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 808         current->fs->umask = create->cr_umask;
 809         switch (create->cr_type) {
 810         case NF4LNK:
 811                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 812                                       create->cr_name, create->cr_namelen,
 813                                       create->cr_data, &attrs, &resfh);
 814                 break;
 815
 816         case NF4BLK:
 817                 status = nfserr_inval;
 818                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 819                 if (MAJOR(rdev) != create->cr_specdata1 ||
 820                     MINOR(rdev) != create->cr_specdata2)
 821                         goto out_umask;
 822                 status = nfsd_create(rqstp, &cstate->current_fh,
 823                                      create->cr_name, create->cr_namelen,
 824                                      &attrs, S_IFBLK, rdev, &resfh);
 825                 break;
 826
 827         case NF4CHR:
 828                 status = nfserr_inval;
 829                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 830                 if (MAJOR(rdev) != create->cr_specdata1 ||
 831                     MINOR(rdev) != create->cr_specdata2)
 832                         goto out_umask;
 833                 status = nfsd_create(rqstp, &cstate->current_fh,
 834                                      create->cr_name, create->cr_namelen,
 835                                      &attrs, S_IFCHR, rdev, &resfh);
 836                 break;
 837
 838         case NF4SOCK:
 839                 status = nfsd_create(rqstp, &cstate->current_fh,
 840                                      create->cr_name, create->cr_namelen,
 841                                      &attrs, S_IFSOCK, 0, &resfh);
 842                 break;
 843
 844         case NF4FIFO:
 845                 status = nfsd_create(rqstp, &cstate->current_fh,
 846                                      create->cr_name, create->cr_namelen,
 847                                      &attrs, S_IFIFO, 0, &resfh);
 848                 break;
 849
 850         case NF4DIR:
 851                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 852                 status = nfsd_create(rqstp, &cstate->current_fh,
 853                                      create->cr_name, create->cr_namelen,
 854                                      &attrs, S_IFDIR, 0, &resfh);
 855                 break;
 856
 857         default:
 858                 status = nfserr_badtype;
 859         }
 860
 861         if (status)
 862                 goto out;
 863
 864         if (attrs.na_labelerr)
 865                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 866         if (attrs.na_aclerr)
 867                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 868         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 869         fh_dup2(&cstate->current_fh, &resfh);
 870 out:
 871         fh_put(&resfh);
 872 out_umask:
 873         current->fs->umask = 0;
 874         nfsd_attrs_free(&attrs);
 875         return status;
 876 }
 877
 878 static __be32
 879 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 880               union nfsd4_op_u *u)
 881 {
 882         struct nfsd4_getattr *getattr = &u->getattr;
 883         __be32 status;
 884
 885         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 886         if (status)
 887                 return status;
 888
 889         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 890                 return nfserr_inval;
 891
 892         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 893         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 894         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 895
 896         getattr->ga_fhp = &cstate->current_fh;
 897         return nfs_ok;
 898 }
 899
 900 static __be32
 901 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 902            union nfsd4_op_u *u)
 903 {
 904         struct nfsd4_link *link = &u->link;
 905         __be32 status;
 906
 907         status = nfsd_link(rqstp, &cstate->current_fh,
 908                            link->li_name, link->li_namelen, &cstate->save_fh);
 909         if (!status)
 910                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 911         return status;
 912 }
 913
 914 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 915 {
 916         struct svc_fh tmp_fh;
 917         __be32 ret;
 918
 919         fh_init(&tmp_fh, NFS4_FHSIZE);
 920         ret = exp_pseudoroot(rqstp, &tmp_fh);
 921         if (ret)
 922                 return ret;
 923         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 924                 fh_put(&tmp_fh);
 925                 return nfserr_noent;
 926         }
 927         fh_put(&tmp_fh);
 928         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 929 }
 930
 931 static __be32
 932 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 933               union nfsd4_op_u *u)
 934 {
 935         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 936 }
 937
 938 static __be32
 939 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 940              union nfsd4_op_u *u)
 941 {
 942         return nfsd_lookup(rqstp, &cstate->current_fh,
 943                            u->lookup.lo_name, u->lookup.lo_len,
 944                            &cstate->current_fh);
 945 }
 946
 947 static __be32
 948 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 949            union nfsd4_op_u *u)
 950 {
 951         struct nfsd4_read *read = &u->read;
 952         __be32 status;
 953
 954         read->rd_nf = NULL;
 955
 956         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 957                               read->rd_offset, read->rd_length);
 958
 959         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 960         if (read->rd_offset > (u64)OFFSET_MAX)
 961                 read->rd_offset = (u64)OFFSET_MAX;
 962         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 963                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 964
 965         /*
 966          * If we do a zero copy read, then a client will see read data
 967          * that reflects the state of the file *after* performing the
 968          * following compound.
 969          *
 970          * To ensure proper ordering, we therefore turn off zero copy if
 971          * the client wants us to do more in this compound:
 972          */
 973         if (!nfsd4_last_compound_op(rqstp)) {
 974                 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
 975
 976                 argp->splice_ok = false;
 977         }
 978
 979         /* check stateid */
 980         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 981                                         &read->rd_stateid, RD_STATE,
 982                                         &read->rd_nf, NULL);
 983
 984         read->rd_rqstp = rqstp;
 985         read->rd_fhp = &cstate->current_fh;
 986         return status;
 987 }
 988
 989
 990 static void
 991 nfsd4_read_release(union nfsd4_op_u *u)
 992 {
 993         if (u->read.rd_nf)
 994                 nfsd_file_put(u->read.rd_nf);
 995         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 996                              u->read.rd_offset, u->read.rd_length);
 997 }
 998
 999 static __be32
1000 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1001               union nfsd4_op_u *u)
1002 {
1003         struct nfsd4_readdir *readdir = &u->readdir;
1004         u64 cookie = readdir->rd_cookie;
1005         static const nfs4_verifier zeroverf;
1006
1007         /* no need to check permission - this will be done in nfsd_readdir() */
1008
1009         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1010                 return nfserr_inval;
1011
1012         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1013         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1014         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1015
1016         if ((cookie == 1) || (cookie == 2) ||
1017             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1018                 return nfserr_bad_cookie;
1019
1020         readdir->rd_rqstp = rqstp;
1021         readdir->rd_fhp = &cstate->current_fh;
1022         return nfs_ok;
1023 }
1024
1025 static __be32
1026 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1027                union nfsd4_op_u *u)
1028 {
1029         u->readlink.rl_rqstp = rqstp;
1030         u->readlink.rl_fhp = &cstate->current_fh;
1031         return nfs_ok;
1032 }
1033
1034 static __be32
1035 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1036              union nfsd4_op_u *u)
1037 {
1038         struct nfsd4_remove *remove = &u->remove;
1039         __be32 status;
1040
1041         if (opens_in_grace(SVC_NET(rqstp)))
1042                 return nfserr_grace;
1043         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1044                              remove->rm_name, remove->rm_namelen);
1045         if (!status)
1046                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1047         return status;
1048 }
1049
1050 static __be32
1051 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1052              union nfsd4_op_u *u)
1053 {
1054         struct nfsd4_rename *rename = &u->rename;
1055         __be32 status;
1056
1057         if (opens_in_grace(SVC_NET(rqstp)))
1058                 return nfserr_grace;
1059         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1060                              rename->rn_snamelen, &cstate->current_fh,
1061                              rename->rn_tname, rename->rn_tnamelen);
1062         if (status)
1063                 return status;
1064         set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1065         set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1066         return nfs_ok;
1067 }
1068
1069 static __be32
1070 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1071               union nfsd4_op_u *u)
1072 {
1073         struct nfsd4_secinfo *secinfo = &u->secinfo;
1074         struct svc_export *exp;
1075         struct dentry *dentry;
1076         __be32 err;
1077
1078         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1079         if (err)
1080                 return err;
1081         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1082                                     secinfo->si_name, secinfo->si_namelen,
1083                                     &exp, &dentry);
1084         if (err)
1085                 return err;
1086         if (d_really_is_negative(dentry)) {
1087                 exp_put(exp);
1088                 err = nfserr_noent;
1089         } else
1090                 secinfo->si_exp = exp;
1091         dput(dentry);
1092         if (cstate->minorversion)
1093                 /* See rfc 5661 section 2.6.3.1.1.8 */
1094                 fh_put(&cstate->current_fh);
1095         return err;
1096 }
1097
1098 static __be32
1099 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1100                 union nfsd4_op_u *u)
1101 {
1102         __be32 err;
1103
1104         switch (u->secinfo_no_name.sin_style) {
1105         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1106                 break;
1107         case NFS4_SECINFO_STYLE4_PARENT:
1108                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1109                 if (err)
1110                         return err;
1111                 break;
1112         default:
1113                 return nfserr_inval;
1114         }
1115
1116         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1117         fh_put(&cstate->current_fh);
1118         return nfs_ok;
1119 }
1120
1121 static void
1122 nfsd4_secinfo_release(union nfsd4_op_u *u)
1123 {
1124         if (u->secinfo.si_exp)
1125                 exp_put(u->secinfo.si_exp);
1126 }
1127
1128 static void
1129 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1130 {
1131         if (u->secinfo_no_name.sin_exp)
1132                 exp_put(u->secinfo_no_name.sin_exp);
1133 }
1134
1135 static __be32
1136 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1137               union nfsd4_op_u *u)
1138 {
1139         struct nfsd4_setattr *setattr = &u->setattr;
1140         struct nfsd_attrs attrs = {
1141                 .na_iattr       = &setattr->sa_iattr,
1142                 .na_seclabel    = &setattr->sa_label,
1143         };
1144         struct inode *inode;
1145         __be32 status = nfs_ok;
1146         int err;
1147
1148         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1149                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1150                                 &cstate->current_fh, &setattr->sa_stateid,
1151                                 WR_STATE, NULL, NULL);
1152                 if (status)
1153                         return status;
1154         }
1155         err = fh_want_write(&cstate->current_fh);
1156         if (err)
1157                 return nfserrno(err);
1158         status = nfs_ok;
1159
1160         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1161                                     nfsd_attrmask);
1162         if (status)
1163                 goto out;
1164
1165         inode = cstate->current_fh.fh_dentry->d_inode;
1166         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1167                                    setattr->sa_acl, &attrs);
1168
1169         if (status)
1170                 goto out;
1171         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1172                                 0, (time64_t)0);
1173         if (!status)
1174                 status = nfserrno(attrs.na_labelerr);
1175         if (!status)
1176                 status = nfserrno(attrs.na_aclerr);
1177 out:
1178         nfsd_attrs_free(&attrs);
1179         fh_drop_write(&cstate->current_fh);
1180         return status;
1181 }
1182
1183 static __be32
1184 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1185             union nfsd4_op_u *u)
1186 {
1187         struct nfsd4_write *write = &u->write;
1188         stateid_t *stateid = &write->wr_stateid;
1189         struct nfsd_file *nf = NULL;
1190         __be32 status = nfs_ok;
1191         unsigned long cnt;
1192         int nvecs;
1193
1194         if (write->wr_offset > (u64)OFFSET_MAX ||
1195             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1196                 return nfserr_fbig;
1197
1198         cnt = write->wr_buflen;
1199         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1200                                write->wr_offset, cnt);
1201         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1202                                                 stateid, WR_STATE, &nf, NULL);
1203         if (status)
1204                 return status;
1205
1206         write->wr_how_written = write->wr_stable_how;
1207
1208         nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1209         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1210
1211         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1212                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1213                                 write->wr_how_written,
1214                                 (__be32 *)write->wr_verifier.data);
1215         nfsd_file_put(nf);
1216
1217         write->wr_bytes_written = cnt;
1218         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1219                               write->wr_offset, cnt);
1220         return status;
1221 }
1222
1223 static __be32
1224 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1225                   stateid_t *src_stateid, struct nfsd_file **src,
1226                   stateid_t *dst_stateid, struct nfsd_file **dst)
1227 {
1228         __be32 status;
1229
1230         if (!cstate->save_fh.fh_dentry)
1231                 return nfserr_nofilehandle;
1232
1233         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1234                                             src_stateid, RD_STATE, src, NULL);
1235         if (status)
1236                 goto out;
1237
1238         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1239                                             dst_stateid, WR_STATE, dst, NULL);
1240         if (status)
1241                 goto out_put_src;
1242
1243         /* fix up for NFS-specific error code */
1244         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1245             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1246                 status = nfserr_wrong_type;
1247                 goto out_put_dst;
1248         }
1249
1250 out:
1251         return status;
1252 out_put_dst:
1253         nfsd_file_put(*dst);
1254         *dst = NULL;
1255 out_put_src:
1256         nfsd_file_put(*src);
1257         *src = NULL;
1258         goto out;
1259 }
1260
1261 static __be32
1262 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1263                 union nfsd4_op_u *u)
1264 {
1265         struct nfsd4_clone *clone = &u->clone;
1266         struct nfsd_file *src, *dst;
1267         __be32 status;
1268
1269         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1270                                    &clone->cl_dst_stateid, &dst);
1271         if (status)
1272                 goto out;
1273
1274         status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1275                         dst, clone->cl_dst_pos, clone->cl_count,
1276                         EX_ISSYNC(cstate->current_fh.fh_export));
1277
1278         nfsd_file_put(dst);
1279         nfsd_file_put(src);
1280 out:
1281         return status;
1282 }
1283
1284 static void nfs4_put_copy(struct nfsd4_copy *copy)
1285 {
1286         if (!refcount_dec_and_test(&copy->refcount))
1287                 return;
1288         kfree(copy->cp_src);
1289         kfree(copy);
1290 }
1291
1292 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1293 {
1294         if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1295                 kthread_stop(copy->copy_task);
1296         nfs4_put_copy(copy);
1297 }
1298
1299 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1300 {
1301         struct nfsd4_copy *copy = NULL;
1302
1303         spin_lock(&clp->async_lock);
1304         if (!list_empty(&clp->async_copies)) {
1305                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1306                                         copies);
1307                 refcount_inc(&copy->refcount);
1308         }
1309         spin_unlock(&clp->async_lock);
1310         return copy;
1311 }
1312
1313 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1314 {
1315         struct nfsd4_copy *copy;
1316
1317         while ((copy = nfsd4_get_copy(clp)) != NULL)
1318                 nfsd4_stop_copy(copy);
1319 }
1320 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1321
1322 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1323                                    struct nfs_fh *src_fh,
1324                                    nfs4_stateid *stateid);
1325 extern void nfs42_ssc_close(struct file *filep);
1326
1327 extern void nfs_sb_deactive(struct super_block *sb);
1328
1329 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1330
1331 /*
1332  * setup a work entry in the ssc delayed unmount list.
1333  */
1334 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1335                                   struct nfsd4_ssc_umount_item **nsui,
1336                                   struct svc_rqst *rqstp)
1337 {
1338         struct nfsd4_ssc_umount_item *ni = NULL;
1339         struct nfsd4_ssc_umount_item *work = NULL;
1340         struct nfsd4_ssc_umount_item *tmp;
1341         DEFINE_WAIT(wait);
1342         __be32 status = 0;
1343
1344         *nsui = NULL;
1345         work = kzalloc(sizeof(*work), GFP_KERNEL);
1346 try_again:
1347         spin_lock(&nn->nfsd_ssc_lock);
1348         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1349                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1350                         continue;
1351                 /* found a match */
1352                 if (ni->nsui_busy) {
1353                         /*  wait - and try again */
1354                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1355                         spin_unlock(&nn->nfsd_ssc_lock);
1356
1357                         /* allow 20secs for mount/unmount for now - revisit */
1358                         if (svc_thread_should_stop(rqstp) ||
1359                                         (schedule_timeout(20*HZ) == 0)) {
1360                                 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1361                                 kfree(work);
1362                                 return nfserr_eagain;
1363                         }
1364                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1365                         goto try_again;
1366                 }
1367                 *nsui = ni;
1368                 refcount_inc(&ni->nsui_refcnt);
1369                 spin_unlock(&nn->nfsd_ssc_lock);
1370                 kfree(work);
1371
1372                 /* return vfsmount in (*nsui)->nsui_vfsmount */
1373                 return 0;
1374         }
1375         if (work) {
1376                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1377                 refcount_set(&work->nsui_refcnt, 2);
1378                 work->nsui_busy = true;
1379                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1380                 *nsui = work;
1381         } else
1382                 status = nfserr_resource;
1383         spin_unlock(&nn->nfsd_ssc_lock);
1384         return status;
1385 }
1386
1387 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1388                                  struct nfsd4_ssc_umount_item *nsui,
1389                                  struct vfsmount *ss_mnt)
1390 {
1391         spin_lock(&nn->nfsd_ssc_lock);
1392         nsui->nsui_vfsmount = ss_mnt;
1393         nsui->nsui_busy = false;
1394         wake_up_all(&nn->nfsd_ssc_waitq);
1395         spin_unlock(&nn->nfsd_ssc_lock);
1396 }
1397
1398 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1399                                  struct nfsd4_ssc_umount_item *nsui)
1400 {
1401         spin_lock(&nn->nfsd_ssc_lock);
1402         list_del(&nsui->nsui_list);
1403         wake_up_all(&nn->nfsd_ssc_waitq);
1404         spin_unlock(&nn->nfsd_ssc_lock);
1405         kfree(nsui);
1406 }
1407
1408 /*
1409  * Support one copy source server for now.
1410  */
1411 static __be32
1412 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1413                        struct nfsd4_ssc_umount_item **nsui)
1414 {
1415         struct file_system_type *type;
1416         struct vfsmount *ss_mnt;
1417         struct nfs42_netaddr *naddr;
1418         struct sockaddr_storage tmp_addr;
1419         size_t tmp_addrlen, match_netid_len = 3;
1420         char *startsep = "", *endsep = "", *match_netid = "tcp";
1421         char *ipaddr, *dev_name, *raw_data;
1422         int len, raw_len;
1423         __be32 status = nfserr_inval;
1424         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1425
1426         naddr = &nss->u.nl4_addr;
1427         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1428                                          naddr->addr_len,
1429                                          (struct sockaddr *)&tmp_addr,
1430                                          sizeof(tmp_addr));
1431         *nsui = NULL;
1432         if (tmp_addrlen == 0)
1433                 goto out_err;
1434
1435         if (tmp_addr.ss_family == AF_INET6) {
1436                 startsep = "[";
1437                 endsep = "]";
1438                 match_netid = "tcp6";
1439                 match_netid_len = 4;
1440         }
1441
1442         if (naddr->netid_len != match_netid_len ||
1443                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1444                 goto out_err;
1445
1446         /* Construct the raw data for the vfs_kern_mount call */
1447         len = RPC_MAX_ADDRBUFLEN + 1;
1448         ipaddr = kzalloc(len, GFP_KERNEL);
1449         if (!ipaddr)
1450                 goto out_err;
1451
1452         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1453
1454         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1455
1456         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1457         raw_data = kzalloc(raw_len, GFP_KERNEL);
1458         if (!raw_data)
1459                 goto out_free_ipaddr;
1460
1461         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1462
1463         status = nfserr_nodev;
1464         type = get_fs_type("nfs");
1465         if (!type)
1466                 goto out_free_rawdata;
1467
1468         /* Set the server:<export> for the vfs_kern_mount call */
1469         dev_name = kzalloc(len + 5, GFP_KERNEL);
1470         if (!dev_name)
1471                 goto out_free_rawdata;
1472         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1473
1474         status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui, rqstp);
1475         if (status)
1476                 goto out_free_devname;
1477         if ((*nsui)->nsui_vfsmount)
1478                 goto out_done;
1479
1480         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1481         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1482         module_put(type->owner);
1483         if (IS_ERR(ss_mnt)) {
1484                 status = nfserr_nodev;
1485                 nfsd4_ssc_cancel_dul(nn, *nsui);
1486                 goto out_free_devname;
1487         }
1488         nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1489 out_done:
1490         status = 0;
1491
1492 out_free_devname:
1493         kfree(dev_name);
1494 out_free_rawdata:
1495         kfree(raw_data);
1496 out_free_ipaddr:
1497         kfree(ipaddr);
1498 out_err:
1499         return status;
1500 }
1501
1502 /*
1503  * Verify COPY destination stateid.
1504  *
1505  * Connect to the source server with NFSv4.1.
1506  * Create the source struct file for nfsd_copy_range.
1507  * Called with COPY cstate:
1508  *    SAVED_FH: source filehandle
1509  *    CURRENT_FH: destination filehandle
1510  */
1511 static __be32
1512 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1513                       struct nfsd4_compound_state *cstate,
1514                       struct nfsd4_copy *copy)
1515 {
1516         struct svc_fh *s_fh = NULL;
1517         stateid_t *s_stid = &copy->cp_src_stateid;
1518         __be32 status = nfserr_inval;
1519
1520         /* Verify the destination stateid and set dst struct file*/
1521         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1522                                             &copy->cp_dst_stateid,
1523                                             WR_STATE, &copy->nf_dst, NULL);
1524         if (status)
1525                 goto out;
1526
1527         status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1528         if (status)
1529                 goto out;
1530
1531         s_fh = &cstate->save_fh;
1532
1533         copy->c_fh.size = s_fh->fh_handle.fh_size;
1534         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1535         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1536         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1537                sizeof(stateid_opaque_t));
1538
1539         status = 0;
1540 out:
1541         return status;
1542 }
1543
1544 static void
1545 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1546                         struct nfsd_file *dst)
1547 {
1548         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1549         long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1550
1551         nfs42_ssc_close(filp);
1552         fput(filp);
1553
1554         spin_lock(&nn->nfsd_ssc_lock);
1555         list_del(&nsui->nsui_list);
1556         /*
1557          * vfsmount can be shared by multiple exports,
1558          * decrement refcnt. If the count drops to 1 it
1559          * will be unmounted when nsui_expire expires.
1560          */
1561         refcount_dec(&nsui->nsui_refcnt);
1562         nsui->nsui_expire = jiffies + timeout;
1563         list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1564         spin_unlock(&nn->nfsd_ssc_lock);
1565 }
1566
1567 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1568
1569 static __be32
1570 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1571                       struct nfsd4_compound_state *cstate,
1572                       struct nfsd4_copy *copy)
1573 {
1574         return nfserr_inval;
1575 }
1576
1577 static void
1578 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1579                         struct nfsd_file *dst)
1580 {
1581 }
1582
1583 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1584                                    struct nfs_fh *src_fh,
1585                                    nfs4_stateid *stateid)
1586 {
1587         return NULL;
1588 }
1589 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1590
1591 static __be32
1592 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1593                       struct nfsd4_compound_state *cstate,
1594                       struct nfsd4_copy *copy)
1595 {
1596         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1597                                  &copy->nf_src, &copy->cp_dst_stateid,
1598                                  &copy->nf_dst);
1599 }
1600
1601 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1602 {
1603         struct nfsd4_cb_offload *cbo =
1604                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1605
1606         kfree(cbo);
1607 }
1608
1609 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1610                                  struct rpc_task *task)
1611 {
1612         struct nfsd4_cb_offload *cbo =
1613                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1614
1615         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1616         return 1;
1617 }
1618
1619 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1620         .release = nfsd4_cb_offload_release,
1621         .done = nfsd4_cb_offload_done
1622 };
1623
1624 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1625 {
1626         copy->cp_res.wr_stable_how =
1627                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1628                         NFS_FILE_SYNC : NFS_UNSTABLE;
1629         nfsd4_copy_set_sync(copy, sync);
1630         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1631 }
1632
1633 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1634                                      struct file *dst,
1635                                      struct file *src)
1636 {
1637         errseq_t since;
1638         ssize_t bytes_copied = 0;
1639         u64 bytes_total = copy->cp_count;
1640         u64 src_pos = copy->cp_src_pos;
1641         u64 dst_pos = copy->cp_dst_pos;
1642         int status;
1643         loff_t end;
1644
1645         /* See RFC 7862 p.67: */
1646         if (bytes_total == 0)
1647                 bytes_total = ULLONG_MAX;
1648         do {
1649                 /* Only async copies can be stopped here */
1650                 if (kthread_should_stop())
1651                         break;
1652                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1653                                                     bytes_total);
1654                 if (bytes_copied <= 0)
1655                         break;
1656                 bytes_total -= bytes_copied;
1657                 copy->cp_res.wr_bytes_written += bytes_copied;
1658                 src_pos += bytes_copied;
1659                 dst_pos += bytes_copied;
1660         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1661         /* for a non-zero asynchronous copy do a commit of data */
1662         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1663                 since = READ_ONCE(dst->f_wb_err);
1664                 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1665                 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1666                 if (!status)
1667                         status = filemap_check_wb_err(dst->f_mapping, since);
1668                 if (!status)
1669                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1670         }
1671         return bytes_copied;
1672 }
1673
1674 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1675                             struct file *src, struct file *dst,
1676                             bool sync)
1677 {
1678         __be32 status;
1679         ssize_t bytes;
1680
1681         bytes = _nfsd_copy_file_range(copy, dst, src);
1682
1683         /* for async copy, we ignore the error, client can always retry
1684          * to get the error
1685          */
1686         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1687                 status = nfserrno(bytes);
1688         else {
1689                 nfsd4_init_copy_res(copy, sync);
1690                 status = nfs_ok;
1691         }
1692         return status;
1693 }
1694
1695 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1696 {
1697         dst->cp_src_pos = src->cp_src_pos;
1698         dst->cp_dst_pos = src->cp_dst_pos;
1699         dst->cp_count = src->cp_count;
1700         dst->cp_flags = src->cp_flags;
1701         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1702         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1703         dst->cp_clp = src->cp_clp;
1704         dst->nf_dst = nfsd_file_get(src->nf_dst);
1705         /* for inter, nf_src doesn't exist yet */
1706         if (!nfsd4_ssc_is_inter(src))
1707                 dst->nf_src = nfsd_file_get(src->nf_src);
1708
1709         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1710         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1711         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1712         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1713         dst->ss_nsui = src->ss_nsui;
1714 }
1715
1716 static void release_copy_files(struct nfsd4_copy *copy)
1717 {
1718         if (copy->nf_src)
1719                 nfsd_file_put(copy->nf_src);
1720         if (copy->nf_dst)
1721                 nfsd_file_put(copy->nf_dst);
1722 }
1723
1724 static void cleanup_async_copy(struct nfsd4_copy *copy)
1725 {
1726         nfs4_free_copy_state(copy);
1727         release_copy_files(copy);
1728         if (copy->cp_clp) {
1729                 spin_lock(&copy->cp_clp->async_lock);
1730                 if (!list_empty(&copy->copies))
1731                         list_del_init(&copy->copies);
1732                 spin_unlock(&copy->cp_clp->async_lock);
1733         }
1734         nfs4_put_copy(copy);
1735 }
1736
1737 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1738 {
1739         struct nfsd4_cb_offload *cbo;
1740
1741         cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1742         if (!cbo)
1743                 return;
1744
1745         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1746         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1747         cbo->co_nfserr = nfserr;
1748
1749         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1750                       NFSPROC4_CLNT_CB_OFFLOAD);
1751         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1752                               &cbo->co_fh, copy->cp_count, nfserr);
1753         nfsd4_run_cb(&cbo->co_cb);
1754 }
1755
1756 /**
1757  * nfsd4_do_async_copy - kthread function for background server-side COPY
1758  * @data: arguments for COPY operation
1759  *
1760  * Return values:
1761  *   %0: Copy operation is done.
1762  */
1763 static int nfsd4_do_async_copy(void *data)
1764 {
1765         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1766         __be32 nfserr;
1767
1768         trace_nfsd_copy_do_async(copy);
1769         if (nfsd4_ssc_is_inter(copy)) {
1770                 struct file *filp;
1771
1772                 filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1773                                       &copy->c_fh, &copy->stateid);
1774                 if (IS_ERR(filp)) {
1775                         switch (PTR_ERR(filp)) {
1776                         case -EBADF:
1777                                 nfserr = nfserr_wrong_type;
1778                                 break;
1779                         default:
1780                                 nfserr = nfserr_offload_denied;
1781                         }
1782                         /* ss_mnt will be unmounted by the laundromat */
1783                         goto do_callback;
1784                 }
1785                 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1786                                        false);
1787                 nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1788         } else {
1789                 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1790                                        copy->nf_dst->nf_file, false);
1791         }
1792
1793 do_callback:
1794         nfsd4_send_cb_offload(copy, nfserr);
1795         cleanup_async_copy(copy);
1796         return 0;
1797 }
1798
1799 static __be32
1800 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1801                 union nfsd4_op_u *u)
1802 {
1803         struct nfsd4_copy *copy = &u->copy;
1804         __be32 status;
1805         struct nfsd4_copy *async_copy = NULL;
1806
1807         copy->cp_clp = cstate->clp;
1808         if (nfsd4_ssc_is_inter(copy)) {
1809                 trace_nfsd_copy_inter(copy);
1810                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1811                         status = nfserr_notsupp;
1812                         goto out;
1813                 }
1814                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1815                 if (status) {
1816                         trace_nfsd_copy_done(copy, status);
1817                         return nfserr_offload_denied;
1818                 }
1819         } else {
1820                 trace_nfsd_copy_intra(copy);
1821                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1822                 if (status) {
1823                         trace_nfsd_copy_done(copy, status);
1824                         return status;
1825                 }
1826         }
1827
1828         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1829                 sizeof(struct knfsd_fh));
1830         if (nfsd4_copy_is_async(copy)) {
1831                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1832
1833                 status = nfserrno(-ENOMEM);
1834                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1835                 if (!async_copy)
1836                         goto out_err;
1837                 INIT_LIST_HEAD(&async_copy->copies);
1838                 refcount_set(&async_copy->refcount, 1);
1839                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1840                 if (!async_copy->cp_src)
1841                         goto out_err;
1842                 if (!nfs4_init_copy_state(nn, copy))
1843                         goto out_err;
1844                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1845                         sizeof(copy->cp_res.cb_stateid));
1846                 dup_copy_fields(copy, async_copy);
1847                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1848                                 async_copy, "%s", "copy thread");
1849                 if (IS_ERR(async_copy->copy_task))
1850                         goto out_err;
1851                 spin_lock(&async_copy->cp_clp->async_lock);
1852                 list_add(&async_copy->copies,
1853                                 &async_copy->cp_clp->async_copies);
1854                 spin_unlock(&async_copy->cp_clp->async_lock);
1855                 wake_up_process(async_copy->copy_task);
1856                 status = nfs_ok;
1857         } else {
1858                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1859                                        copy->nf_dst->nf_file, true);
1860         }
1861 out:
1862         trace_nfsd_copy_done(copy, status);
1863         release_copy_files(copy);
1864         return status;
1865 out_err:
1866         if (nfsd4_ssc_is_inter(copy)) {
1867                 /*
1868                  * Source's vfsmount of inter-copy will be unmounted
1869                  * by the laundromat. Use copy instead of async_copy
1870                  * since async_copy->ss_nsui might not be set yet.
1871                  */
1872                 refcount_dec(&copy->ss_nsui->nsui_refcnt);
1873         }
1874         if (async_copy)
1875                 cleanup_async_copy(async_copy);
1876         status = nfserrno(-ENOMEM);
1877         goto out;
1878 }
1879
1880 static struct nfsd4_copy *
1881 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1882 {
1883         struct nfsd4_copy *copy;
1884
1885         lockdep_assert_held(&clp->async_lock);
1886
1887         list_for_each_entry(copy, &clp->async_copies, copies) {
1888                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1889                         continue;
1890                 return copy;
1891         }
1892         return NULL;
1893 }
1894
1895 static struct nfsd4_copy *
1896 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1897 {
1898         struct nfsd4_copy *copy;
1899
1900         spin_lock(&clp->async_lock);
1901         copy = find_async_copy_locked(clp, stateid);
1902         if (copy)
1903                 refcount_inc(&copy->refcount);
1904         spin_unlock(&clp->async_lock);
1905         return copy;
1906 }
1907
1908 static __be32
1909 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1910                      struct nfsd4_compound_state *cstate,
1911                      union nfsd4_op_u *u)
1912 {
1913         struct nfsd4_offload_status *os = &u->offload_status;
1914         struct nfsd4_copy *copy;
1915         struct nfs4_client *clp = cstate->clp;
1916
1917         copy = find_async_copy(clp, &os->stateid);
1918         if (!copy) {
1919                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1920
1921                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1922         } else
1923                 nfsd4_stop_copy(copy);
1924
1925         return nfs_ok;
1926 }
1927
1928 static __be32
1929 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1930                   union nfsd4_op_u *u)
1931 {
1932         struct nfsd4_copy_notify *cn = &u->copy_notify;
1933         __be32 status;
1934         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1935         struct nfs4_stid *stid;
1936         struct nfs4_cpntf_state *cps;
1937         struct nfs4_client *clp = cstate->clp;
1938
1939         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1940                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1941                                         &stid);
1942         if (status)
1943                 return status;
1944
1945         cn->cpn_lease_time.tv_sec = nn->nfsd4_lease;
1946         cn->cpn_lease_time.tv_nsec = 0;
1947
1948         status = nfserrno(-ENOMEM);
1949         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1950         if (!cps)
1951                 goto out;
1952         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1953         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1954         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1955
1956         /* For now, only return one server address in cpn_src, the
1957          * address used by the client to connect to this server.
1958          */
1959         cn->cpn_src->nl4_type = NL4_NETADDR;
1960         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1961                                  &cn->cpn_src->u.nl4_addr);
1962         WARN_ON_ONCE(status);
1963         if (status) {
1964                 nfs4_put_cpntf_state(nn, cps);
1965                 goto out;
1966         }
1967 out:
1968         nfs4_put_stid(stid);
1969         return status;
1970 }
1971
1972 static __be32
1973 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1974                 struct nfsd4_fallocate *fallocate, int flags)
1975 {
1976         __be32 status;
1977         struct nfsd_file *nf;
1978
1979         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1980                                             &fallocate->falloc_stateid,
1981                                             WR_STATE, &nf, NULL);
1982         if (status != nfs_ok)
1983                 return status;
1984
1985         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1986                                      fallocate->falloc_offset,
1987                                      fallocate->falloc_length,
1988                                      flags);
1989         nfsd_file_put(nf);
1990         return status;
1991 }
1992
1993 static __be32
1994 nfsd4_offload_status(struct svc_rqst *rqstp,
1995                      struct nfsd4_compound_state *cstate,
1996                      union nfsd4_op_u *u)
1997 {
1998         struct nfsd4_offload_status *os = &u->offload_status;
1999         __be32 status = nfs_ok;
2000         struct nfsd4_copy *copy;
2001         struct nfs4_client *clp = cstate->clp;
2002
2003         spin_lock(&clp->async_lock);
2004         copy = find_async_copy_locked(clp, &os->stateid);
2005         if (copy)
2006                 os->count = copy->cp_res.wr_bytes_written;
2007         else
2008                 status = nfserr_bad_stateid;
2009         spin_unlock(&clp->async_lock);
2010
2011         return status;
2012 }
2013
2014 static __be32
2015 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2016                union nfsd4_op_u *u)
2017 {
2018         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2019 }
2020
2021 static __be32
2022 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2023                  union nfsd4_op_u *u)
2024 {
2025         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2026                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2027 }
2028
2029 static __be32
2030 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2031            union nfsd4_op_u *u)
2032 {
2033         struct nfsd4_seek *seek = &u->seek;
2034         int whence;
2035         __be32 status;
2036         struct nfsd_file *nf;
2037
2038         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2039                                             &seek->seek_stateid,
2040                                             RD_STATE, &nf, NULL);
2041         if (status)
2042                 return status;
2043
2044         switch (seek->seek_whence) {
2045         case NFS4_CONTENT_DATA:
2046                 whence = SEEK_DATA;
2047                 break;
2048         case NFS4_CONTENT_HOLE:
2049                 whence = SEEK_HOLE;
2050                 break;
2051         default:
2052                 status = nfserr_union_notsupp;
2053                 goto out;
2054         }
2055
2056         /*
2057          * Note:  This call does change file->f_pos, but nothing in NFSD
2058          *        should ever file->f_pos.
2059          */
2060         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2061         if (seek->seek_pos < 0)
2062                 status = nfserrno(seek->seek_pos);
2063         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2064                 seek->seek_eof = true;
2065
2066 out:
2067         nfsd_file_put(nf);
2068         return status;
2069 }
2070
2071 /* This routine never returns NFS_OK!  If there are no other errors, it
2072  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2073  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2074  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2075  */
2076 static __be32
2077 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2078              struct nfsd4_verify *verify)
2079 {
2080         __be32 *buf, *p;
2081         int count;
2082         __be32 status;
2083
2084         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2085         if (status)
2086                 return status;
2087
2088         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2089         if (status)
2090                 return status;
2091
2092         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2093             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2094                 return nfserr_inval;
2095         if (verify->ve_attrlen & 3)
2096                 return nfserr_inval;
2097
2098         /* count in words:
2099          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2100          */
2101         count = 4 + (verify->ve_attrlen >> 2);
2102         buf = kmalloc(count << 2, GFP_KERNEL);
2103         if (!buf)
2104                 return nfserr_jukebox;
2105
2106         p = buf;
2107         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2108                                     cstate->current_fh.fh_export,
2109                                     cstate->current_fh.fh_dentry,
2110                                     verify->ve_bmval,
2111                                     rqstp, 0);
2112         /*
2113          * If nfsd4_encode_fattr() ran out of space, assume that's because
2114          * the attributes are longer (hence different) than those given:
2115          */
2116         if (status == nfserr_resource)
2117                 status = nfserr_not_same;
2118         if (status)
2119                 goto out_kfree;
2120
2121         /* skip bitmap */
2122         p = buf + 1 + ntohl(buf[0]);
2123         status = nfserr_not_same;
2124         if (ntohl(*p++) != verify->ve_attrlen)
2125                 goto out_kfree;
2126         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2127                 status = nfserr_same;
2128
2129 out_kfree:
2130         kfree(buf);
2131         return status;
2132 }
2133
2134 static __be32
2135 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2136               union nfsd4_op_u *u)
2137 {
2138         __be32 status;
2139
2140         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2141         return status == nfserr_not_same ? nfs_ok : status;
2142 }
2143
2144 static __be32
2145 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2146              union nfsd4_op_u *u)
2147 {
2148         __be32 status;
2149
2150         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2151         return status == nfserr_same ? nfs_ok : status;
2152 }
2153
2154 #ifdef CONFIG_NFSD_PNFS
2155 static const struct nfsd4_layout_ops *
2156 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2157 {
2158         if (!exp->ex_layout_types) {
2159                 dprintk("%s: export does not support pNFS\n", __func__);
2160                 return NULL;
2161         }
2162
2163         if (layout_type >= LAYOUT_TYPE_MAX ||
2164             !(exp->ex_layout_types & (1 << layout_type))) {
2165                 dprintk("%s: layout type %d not supported\n",
2166                         __func__, layout_type);
2167                 return NULL;
2168         }
2169
2170         return nfsd4_layout_ops[layout_type];
2171 }
2172
2173 static __be32
2174 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2175                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2176 {
2177         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2178         const struct nfsd4_layout_ops *ops;
2179         struct nfsd4_deviceid_map *map;
2180         struct svc_export *exp;
2181         __be32 nfserr;
2182
2183         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2184                __func__,
2185                gdp->gd_layout_type,
2186                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2187                gdp->gd_maxcount);
2188
2189         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2190         if (!map) {
2191                 dprintk("%s: couldn't find device ID to export mapping!\n",
2192                         __func__);
2193                 return nfserr_noent;
2194         }
2195
2196         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2197         if (IS_ERR(exp)) {
2198                 dprintk("%s: could not find device id\n", __func__);
2199                 return nfserr_noent;
2200         }
2201
2202         nfserr = nfserr_layoutunavailable;
2203         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2204         if (!ops)
2205                 goto out;
2206
2207         nfserr = nfs_ok;
2208         if (gdp->gd_maxcount != 0) {
2209                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2210                                 rqstp, cstate->clp, gdp);
2211         }
2212
2213         gdp->gd_notify_types &= ops->notify_types;
2214 out:
2215         exp_put(exp);
2216         return nfserr;
2217 }
2218
2219 static void
2220 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2221 {
2222         kfree(u->getdeviceinfo.gd_device);
2223 }
2224
2225 static __be32
2226 nfsd4_layoutget(struct svc_rqst *rqstp,
2227                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2228 {
2229         struct nfsd4_layoutget *lgp = &u->layoutget;
2230         struct svc_fh *current_fh = &cstate->current_fh;
2231         const struct nfsd4_layout_ops *ops;
2232         struct nfs4_layout_stateid *ls;
2233         __be32 nfserr;
2234         int accmode = NFSD_MAY_READ_IF_EXEC;
2235
2236         switch (lgp->lg_seg.iomode) {
2237         case IOMODE_READ:
2238                 accmode |= NFSD_MAY_READ;
2239                 break;
2240         case IOMODE_RW:
2241                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2242                 break;
2243         default:
2244                 dprintk("%s: invalid iomode %d\n",
2245                         __func__, lgp->lg_seg.iomode);
2246                 nfserr = nfserr_badiomode;
2247                 goto out;
2248         }
2249
2250         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2251         if (nfserr)
2252                 goto out;
2253
2254         nfserr = nfserr_layoutunavailable;
2255         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2256         if (!ops)
2257                 goto out;
2258
2259         /*
2260          * Verify minlength and range as per RFC5661:
2261          *  o  If loga_length is less than loga_minlength,
2262          *     the metadata server MUST return NFS4ERR_INVAL.
2263          *  o  If the sum of loga_offset and loga_minlength exceeds
2264          *     NFS4_UINT64_MAX, and loga_minlength is not
2265          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2266          *  o  If the sum of loga_offset and loga_length exceeds
2267          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2268          *     the error NFS4ERR_INVAL MUST result.
2269          */
2270         nfserr = nfserr_inval;
2271         if (lgp->lg_seg.length < lgp->lg_minlength ||
2272             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2273              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2274             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2275              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2276                 goto out;
2277         if (lgp->lg_seg.length == 0)
2278                 goto out;
2279
2280         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2281                                                 true, lgp->lg_layout_type, &ls);
2282         if (nfserr) {
2283                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2284                 goto out;
2285         }
2286
2287         nfserr = nfserr_recallconflict;
2288         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2289                 goto out_put_stid;
2290
2291         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2292                                      current_fh, lgp);
2293         if (nfserr)
2294                 goto out_put_stid;
2295
2296         nfserr = nfsd4_insert_layout(lgp, ls);
2297
2298 out_put_stid:
2299         mutex_unlock(&ls->ls_mutex);
2300         nfs4_put_stid(&ls->ls_stid);
2301 out:
2302         return nfserr;
2303 }
2304
2305 static void
2306 nfsd4_layoutget_release(union nfsd4_op_u *u)
2307 {
2308         kfree(u->layoutget.lg_content);
2309 }
2310
2311 static __be32
2312 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2313                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2314 {
2315         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2316         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2317         struct svc_fh *current_fh = &cstate->current_fh;
2318         const struct nfsd4_layout_ops *ops;
2319         loff_t new_size = lcp->lc_last_wr + 1;
2320         struct inode *inode;
2321         struct nfs4_layout_stateid *ls;
2322         __be32 nfserr;
2323
2324         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2325         if (nfserr)
2326                 goto out;
2327
2328         nfserr = nfserr_layoutunavailable;
2329         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2330         if (!ops)
2331                 goto out;
2332         inode = d_inode(current_fh->fh_dentry);
2333
2334         nfserr = nfserr_inval;
2335         if (new_size <= seg->offset) {
2336                 dprintk("pnfsd: last write before layout segment\n");
2337                 goto out;
2338         }
2339         if (new_size > seg->offset + seg->length) {
2340                 dprintk("pnfsd: last write beyond layout segment\n");
2341                 goto out;
2342         }
2343         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2344                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2345                 goto out;
2346         }
2347
2348         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2349                                                 false, lcp->lc_layout_type,
2350                                                 &ls);
2351         if (nfserr) {
2352                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2353                 /* fixup error code as per RFC5661 */
2354                 if (nfserr == nfserr_bad_stateid)
2355                         nfserr = nfserr_badlayout;
2356                 goto out;
2357         }
2358
2359         /* LAYOUTCOMMIT does not require any serialization */
2360         mutex_unlock(&ls->ls_mutex);
2361
2362         if (new_size > i_size_read(inode)) {
2363                 lcp->lc_size_chg = true;
2364                 lcp->lc_newsize = new_size;
2365         } else {
2366                 lcp->lc_size_chg = false;
2367         }
2368
2369         nfserr = ops->proc_layoutcommit(inode, lcp);
2370         nfs4_put_stid(&ls->ls_stid);
2371 out:
2372         return nfserr;
2373 }
2374
2375 static __be32
2376 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2377                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2378 {
2379         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2380         struct svc_fh *current_fh = &cstate->current_fh;
2381         __be32 nfserr;
2382
2383         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2384         if (nfserr)
2385                 goto out;
2386
2387         nfserr = nfserr_layoutunavailable;
2388         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2389                 goto out;
2390
2391         switch (lrp->lr_seg.iomode) {
2392         case IOMODE_READ:
2393         case IOMODE_RW:
2394         case IOMODE_ANY:
2395                 break;
2396         default:
2397                 dprintk("%s: invalid iomode %d\n", __func__,
2398                         lrp->lr_seg.iomode);
2399                 nfserr = nfserr_inval;
2400                 goto out;
2401         }
2402
2403         switch (lrp->lr_return_type) {
2404         case RETURN_FILE:
2405                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2406                 break;
2407         case RETURN_FSID:
2408         case RETURN_ALL:
2409                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2410                 break;
2411         default:
2412                 dprintk("%s: invalid return_type %d\n", __func__,
2413                         lrp->lr_return_type);
2414                 nfserr = nfserr_inval;
2415                 break;
2416         }
2417 out:
2418         return nfserr;
2419 }
2420 #endif /* CONFIG_NFSD_PNFS */
2421
2422 static __be32
2423 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2424                union nfsd4_op_u *u)
2425 {
2426         struct nfsd4_getxattr *getxattr = &u->getxattr;
2427
2428         return nfsd_getxattr(rqstp, &cstate->current_fh,
2429                              getxattr->getxa_name, &getxattr->getxa_buf,
2430                              &getxattr->getxa_len);
2431 }
2432
2433 static __be32
2434 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2435            union nfsd4_op_u *u)
2436 {
2437         struct nfsd4_setxattr *setxattr = &u->setxattr;
2438         __be32 ret;
2439
2440         if (opens_in_grace(SVC_NET(rqstp)))
2441                 return nfserr_grace;
2442
2443         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2444                             setxattr->setxa_buf, setxattr->setxa_len,
2445                             setxattr->setxa_flags);
2446
2447         if (!ret)
2448                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2449
2450         return ret;
2451 }
2452
2453 static __be32
2454 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2455            union nfsd4_op_u *u)
2456 {
2457         /*
2458          * Get the entire list, then copy out only the user attributes
2459          * in the encode function.
2460          */
2461         return nfsd_listxattr(rqstp, &cstate->current_fh,
2462                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2463 }
2464
2465 static __be32
2466 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2467            union nfsd4_op_u *u)
2468 {
2469         struct nfsd4_removexattr *removexattr = &u->removexattr;
2470         __be32 ret;
2471
2472         if (opens_in_grace(SVC_NET(rqstp)))
2473                 return nfserr_grace;
2474
2475         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2476             removexattr->rmxa_name);
2477
2478         if (!ret)
2479                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2480
2481         return ret;
2482 }
2483
2484 /*
2485  * NULL call.
2486  */
2487 static __be32
2488 nfsd4_proc_null(struct svc_rqst *rqstp)
2489 {
2490         return rpc_success;
2491 }
2492
2493 static inline void nfsd4_increment_op_stats(u32 opnum)
2494 {
2495         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2496                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2497 }
2498
2499 static const struct nfsd4_operation nfsd4_ops[];
2500
2501 static const char *nfsd4_op_name(unsigned opnum);
2502
2503 /*
2504  * Enforce NFSv4.1 COMPOUND ordering rules:
2505  *
2506  * Also note, enforced elsewhere:
2507  *      - SEQUENCE other than as first op results in
2508  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2509  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2510  *        (Enforced in nfsd4_bind_conn_to_session().)
2511  *      - DESTROY_SESSION must be the final operation in a compound, if
2512  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2513  *        (Enforced in nfsd4_destroy_session().)
2514  */
2515 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2516 {
2517         struct nfsd4_op *first_op = &args->ops[0];
2518
2519         /* These ordering requirements don't apply to NFSv4.0: */
2520         if (args->minorversion == 0)
2521                 return nfs_ok;
2522         /* This is weird, but OK, not our problem: */
2523         if (args->opcnt == 0)
2524                 return nfs_ok;
2525         if (first_op->status == nfserr_op_illegal)
2526                 return nfs_ok;
2527         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2528                 return nfserr_op_not_in_session;
2529         if (first_op->opnum == OP_SEQUENCE)
2530                 return nfs_ok;
2531         /*
2532          * So first_op is something allowed outside a session, like
2533          * EXCHANGE_ID; but then it has to be the only op in the
2534          * compound:
2535          */
2536         if (args->opcnt != 1)
2537                 return nfserr_not_only_op;
2538         return nfs_ok;
2539 }
2540
2541 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2542 {
2543         return &nfsd4_ops[op->opnum];
2544 }
2545
2546 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2547 {
2548         if (op->opnum == OP_ILLEGAL)
2549                 return false;
2550         return OPDESC(op)->op_flags & OP_CACHEME;
2551 }
2552
2553 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2554 {
2555         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2556         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2557         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2558         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2559         const struct nfsd4_operation *thisd = OPDESC(this);
2560         const struct nfsd4_operation *nextd;
2561
2562         /*
2563          * Most ops check wronsec on our own; only the putfh-like ops
2564          * have special rules.
2565          */
2566         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2567                 return false;
2568         /*
2569          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2570          * put-filehandle operation if we're not going to use the
2571          * result:
2572          */
2573         if (argp->opcnt == resp->opcnt)
2574                 return false;
2575         if (next->opnum == OP_ILLEGAL)
2576                 return false;
2577         nextd = OPDESC(next);
2578         /*
2579          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2580          * errors themselves as necessary; others should check for them
2581          * now:
2582          */
2583         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2584 }
2585
2586 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2587 static void
2588 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2589 {
2590         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2591         struct nfsd4_copy *copy;
2592         struct nfsd4_putfh *putfh;
2593         int i;
2594
2595         /* traverse all operation and if it's a COPY compound, mark the
2596          * source filehandle to skip verification
2597          */
2598         for (i = 0; i < args->opcnt; i++) {
2599                 op = &args->ops[i];
2600                 if (op->opnum == OP_PUTFH)
2601                         current_op = op;
2602                 else if (op->opnum == OP_SAVEFH)
2603                         saved_op = current_op;
2604                 else if (op->opnum == OP_RESTOREFH)
2605                         current_op = saved_op;
2606                 else if (op->opnum == OP_COPY) {
2607                         copy = (struct nfsd4_copy *)&op->u;
2608                         if (!saved_op) {
2609                                 op->status = nfserr_nofilehandle;
2610                                 return;
2611                         }
2612                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2613                         if (nfsd4_ssc_is_inter(copy))
2614                                 putfh->no_verify = true;
2615                 }
2616         }
2617 }
2618 #else
2619 static void
2620 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2621 {
2622 }
2623 #endif
2624
2625 /*
2626  * COMPOUND call.
2627  */
2628 static __be32
2629 nfsd4_proc_compound(struct svc_rqst *rqstp)
2630 {
2631         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2632         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2633         struct nfsd4_op *op;
2634         struct nfsd4_compound_state *cstate = &resp->cstate;
2635         struct svc_fh *current_fh = &cstate->current_fh;
2636         struct svc_fh *save_fh = &cstate->save_fh;
2637         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2638         __be32          status;
2639
2640         resp->xdr = &rqstp->rq_res_stream;
2641         resp->statusp = resp->xdr->p;
2642
2643         /* reserve space for: NFS status code */
2644         xdr_reserve_space(resp->xdr, XDR_UNIT);
2645
2646         /* reserve space for: taglen, tag, and opcnt */
2647         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2648         resp->taglen = args->taglen;
2649         resp->tag = args->tag;
2650         resp->rqstp = rqstp;
2651         cstate->minorversion = args->minorversion;
2652         fh_init(current_fh, NFS4_FHSIZE);
2653         fh_init(save_fh, NFS4_FHSIZE);
2654         /*
2655          * Don't use the deferral mechanism for NFSv4; compounds make it
2656          * too hard to avoid non-idempotency problems.
2657          */
2658         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2659
2660         /*
2661          * According to RFC3010, this takes precedence over all other errors.
2662          */
2663         status = nfserr_minor_vers_mismatch;
2664         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2665                 goto out;
2666
2667         status = nfs41_check_op_ordering(args);
2668         if (status) {
2669                 op = &args->ops[0];
2670                 op->status = status;
2671                 resp->opcnt = 1;
2672                 goto encode_op;
2673         }
2674         check_if_stalefh_allowed(args);
2675
2676         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2677
2678         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2679         while (!status && resp->opcnt < args->opcnt) {
2680                 op = &args->ops[resp->opcnt++];
2681
2682                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2683                         /* If there are still more operations to process,
2684                          * stop here and report NFS4ERR_RESOURCE. */
2685                         if (cstate->minorversion == 0 &&
2686                             args->client_opcnt > resp->opcnt) {
2687                                 op->status = nfserr_resource;
2688                                 goto encode_op;
2689                         }
2690                 }
2691
2692                 /*
2693                  * The XDR decode routines may have pre-set op->status;
2694                  * for example, if there is a miscellaneous XDR error
2695                  * it will be set to nfserr_bad_xdr.
2696                  */
2697                 if (op->status) {
2698                         if (op->opnum == OP_OPEN)
2699                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2700                         goto encode_op;
2701                 }
2702                 if (!current_fh->fh_dentry &&
2703                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2704                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2705                                 op->status = nfserr_nofilehandle;
2706                                 goto encode_op;
2707                         }
2708                 } else if (current_fh->fh_export &&
2709                            current_fh->fh_export->ex_fslocs.migrated &&
2710                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2711                         op->status = nfserr_moved;
2712                         goto encode_op;
2713                 }
2714
2715                 fh_clear_pre_post_attrs(current_fh);
2716
2717                 /* If op is non-idempotent */
2718                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2719                         /*
2720                          * Don't execute this op if we couldn't encode a
2721                          * successful reply:
2722                          */
2723                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2724                         /*
2725                          * Plus if there's another operation, make sure
2726                          * we'll have space to at least encode an error:
2727                          */
2728                         if (resp->opcnt < args->opcnt)
2729                                 plen += COMPOUND_ERR_SLACK_SPACE;
2730                         op->status = nfsd4_check_resp_size(resp, plen);
2731                 }
2732
2733                 if (op->status)
2734                         goto encode_op;
2735
2736                 if (op->opdesc->op_get_currentstateid)
2737                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2738                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2739
2740                 /* Only from SEQUENCE */
2741                 if (cstate->status == nfserr_replay_cache) {
2742                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2743                         status = op->status;
2744                         goto out;
2745                 }
2746                 if (!op->status) {
2747                         if (op->opdesc->op_set_currentstateid)
2748                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2749
2750                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2751                                 clear_current_stateid(cstate);
2752
2753                         if (current_fh->fh_export &&
2754                                         need_wrongsec_check(rqstp))
2755                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2756                 }
2757 encode_op:
2758                 if (op->status == nfserr_replay_me) {
2759                         op->replay = &cstate->replay_owner->so_replay;
2760                         nfsd4_encode_replay(resp->xdr, op);
2761                         status = op->status = op->replay->rp_status;
2762                 } else {
2763                         nfsd4_encode_operation(resp, op);
2764                         status = op->status;
2765                 }
2766
2767                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2768                                            status, nfsd4_op_name(op->opnum));
2769
2770                 nfsd4_cstate_clear_replay(cstate);
2771                 nfsd4_increment_op_stats(op->opnum);
2772         }
2773
2774         fh_put(current_fh);
2775         fh_put(save_fh);
2776         BUG_ON(cstate->replay_owner);
2777 out:
2778         cstate->status = status;
2779         /* Reset deferral mechanism for RPC deferrals */
2780         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2781         return rpc_success;
2782 }
2783
2784 #define op_encode_hdr_size              (2)
2785 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2786 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2787 #define op_encode_change_info_maxsz     (5)
2788 #define nfs4_fattr_bitmap_maxsz         (4)
2789
2790 /* We'll fall back on returning no lockowner if run out of space: */
2791 #define op_encode_lockowner_maxsz       (0)
2792 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2793
2794 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2795
2796 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2797 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2798                                          op_encode_ace_maxsz)
2799
2800 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2801
2802 /*
2803  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2804  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2805  * the maximum payload size here, based on transport limits and the size
2806  * of the remaining space in the rq_pages array.
2807  */
2808 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2809 {
2810         u32 buflen;
2811
2812         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2813         buflen -= rqstp->rq_auth_slack;
2814         buflen -= rqstp->rq_res.head[0].iov_len;
2815         return min_t(u32, buflen, svc_max_payload(rqstp));
2816 }
2817
2818 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2819                                    const struct nfsd4_op *op)
2820 {
2821         return (op_encode_hdr_size) * sizeof(__be32);
2822 }
2823
2824 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2825                                       const struct nfsd4_op *op)
2826 {
2827         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2828 }
2829
2830 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2831                               const struct nfsd4_op *op)
2832 {
2833         /* ac_supported, ac_resp_access */
2834         return (op_encode_hdr_size + 2)* sizeof(__be32);
2835 }
2836
2837 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2838                               const struct nfsd4_op *op)
2839 {
2840         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2841 }
2842
2843 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2844                               const struct nfsd4_op *op)
2845 {
2846         return (op_encode_hdr_size + op_encode_change_info_maxsz
2847                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2848 }
2849
2850 /*
2851  * Note since this is an idempotent operation we won't insist on failing
2852  * the op prematurely if the estimate is too large.  We may turn off splice
2853  * reads unnecessarily.
2854  */
2855 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2856                                const struct nfsd4_op *op)
2857 {
2858         const u32 *bmap = op->u.getattr.ga_bmval;
2859         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2860         u32 ret = 0;
2861
2862         if (bmap0 & FATTR4_WORD0_ACL)
2863                 return nfsd4_max_payload(rqstp);
2864         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2865                 return nfsd4_max_payload(rqstp);
2866
2867         if (bmap1 & FATTR4_WORD1_OWNER) {
2868                 ret += IDMAP_NAMESZ + 4;
2869                 bmap1 &= ~FATTR4_WORD1_OWNER;
2870         }
2871         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2872                 ret += IDMAP_NAMESZ + 4;
2873                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2874         }
2875         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2876                 ret += NFS4_FHSIZE + 4;
2877                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2878         }
2879         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2880                 ret += NFS4_MAXLABELLEN + 12;
2881                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2882         }
2883         /*
2884          * Largest of remaining attributes are 16 bytes (e.g.,
2885          * supported_attributes)
2886          */
2887         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2888         /* bitmask, length */
2889         ret += 20;
2890         return ret;
2891 }
2892
2893 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2894                              const struct nfsd4_op *op)
2895 {
2896         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2897 }
2898
2899 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2900                             const struct nfsd4_op *op)
2901 {
2902         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2903                 * sizeof(__be32);
2904 }
2905
2906 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2907                             const struct nfsd4_op *op)
2908 {
2909         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2910                 * sizeof(__be32);
2911 }
2912
2913 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2914                             const struct nfsd4_op *op)
2915 {
2916         return (op_encode_hdr_size + op_encode_stateid_maxsz
2917                 + op_encode_change_info_maxsz + 1
2918                 + nfs4_fattr_bitmap_maxsz
2919                 + op_encode_delegation_maxsz) * sizeof(__be32);
2920 }
2921
2922 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2923                             const struct nfsd4_op *op)
2924 {
2925         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2926
2927         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2928 }
2929
2930 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2931                                  const struct nfsd4_op *op)
2932 {
2933         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2934         /*
2935          * If we detect that the file changed during hole encoding, then we
2936          * recover by encoding the remaining reply as data. This means we need
2937          * to set aside enough room to encode two data segments.
2938          */
2939         u32 seg_len = 2 * (1 + 2 + 1);
2940
2941         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2942 }
2943
2944 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2945                                const struct nfsd4_op *op)
2946 {
2947         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2948
2949         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2950                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2951 }
2952
2953 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2954                                 const struct nfsd4_op *op)
2955 {
2956         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2957 }
2958
2959 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2960                               const struct nfsd4_op *op)
2961 {
2962         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2963                 * sizeof(__be32);
2964 }
2965
2966 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2967                               const struct nfsd4_op *op)
2968 {
2969         return (op_encode_hdr_size + op_encode_change_info_maxsz
2970                 + op_encode_change_info_maxsz) * sizeof(__be32);
2971 }
2972
2973 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2974                                 const struct nfsd4_op *op)
2975 {
2976         return (op_encode_hdr_size
2977                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2978 }
2979
2980 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2981                                     const struct nfsd4_op *op)
2982 {
2983         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2984                 * sizeof(__be32);
2985 }
2986
2987 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2988                                const struct nfsd4_op *op)
2989 {
2990         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2991 }
2992
2993 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2994                                const struct nfsd4_op *op)
2995 {
2996         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2997                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2998 }
2999
3000 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
3001                                    const struct nfsd4_op *op)
3002 {
3003         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
3004                                                                 sizeof(__be32);
3005 }
3006
3007 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
3008                              const struct nfsd4_op *op)
3009 {
3010         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3011 }
3012
3013 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3014                                    const struct nfsd4_op *op)
3015 {
3016         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3017                 1 + 1 + /* eir_flags, spr_how */\
3018                 4 + /* spo_must_enforce & _allow with bitmap */\
3019                 2 + /*eir_server_owner.so_minor_id */\
3020                 /* eir_server_owner.so_major_id<> */\
3021                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3022                 /* eir_server_scope<> */\
3023                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3024                 1 + /* eir_server_impl_id array length */\
3025                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3026 }
3027
3028 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3029                                             const struct nfsd4_op *op)
3030 {
3031         return (op_encode_hdr_size + \
3032                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3033                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3034 }
3035
3036 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3037                                       const struct nfsd4_op *op)
3038 {
3039         return (op_encode_hdr_size + \
3040                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3041                 2 + /* csr_sequence, csr_flags */\
3042                 op_encode_channel_attrs_maxsz + \
3043                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3044 }
3045
3046 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3047                             const struct nfsd4_op *op)
3048 {
3049         return (op_encode_hdr_size +
3050                 1 /* wr_callback */ +
3051                 op_encode_stateid_maxsz /* wr_callback */ +
3052                 2 /* wr_count */ +
3053                 1 /* wr_committed */ +
3054                 op_encode_verifier_maxsz +
3055                 1 /* cr_consecutive */ +
3056                 1 /* cr_synchronous */) * sizeof(__be32);
3057 }
3058
3059 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3060                                       const struct nfsd4_op *op)
3061 {
3062         return (op_encode_hdr_size +
3063                 2 /* osr_count */ +
3064                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3065 }
3066
3067 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3068                                    const struct nfsd4_op *op)
3069 {
3070         return (op_encode_hdr_size +
3071                 3 /* cnr_lease_time */ +
3072                 1 /* We support one cnr_source_server */ +
3073                 1 /* cnr_stateid seq */ +
3074                 op_encode_stateid_maxsz /* cnr_stateid */ +
3075                 1 /* num cnr_source_server*/ +
3076                 1 /* nl4_type */ +
3077                 1 /* nl4 size */ +
3078                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3079                 * sizeof(__be32);
3080 }
3081
3082 #ifdef CONFIG_NFSD_PNFS
3083 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3084                                      const struct nfsd4_op *op)
3085 {
3086         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3087
3088         return (op_encode_hdr_size +
3089                 1 /* gd_layout_type*/ +
3090                 XDR_QUADLEN(rlen) +
3091                 2 /* gd_notify_types */) * sizeof(__be32);
3092 }
3093
3094 /*
3095  * At this stage we don't really know what layout driver will handle the request,
3096  * so we need to define an arbitrary upper bound here.
3097  */
3098 #define MAX_LAYOUT_SIZE         128
3099 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3100                                  const struct nfsd4_op *op)
3101 {
3102         return (op_encode_hdr_size +
3103                 1 /* logr_return_on_close */ +
3104                 op_encode_stateid_maxsz +
3105                 1 /* nr of layouts */ +
3106                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3107 }
3108
3109 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3110                                     const struct nfsd4_op *op)
3111 {
3112         return (op_encode_hdr_size +
3113                 1 /* locr_newsize */ +
3114                 2 /* ns_size */) * sizeof(__be32);
3115 }
3116
3117 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3118                                     const struct nfsd4_op *op)
3119 {
3120         return (op_encode_hdr_size +
3121                 1 /* lrs_stateid */ +
3122                 op_encode_stateid_maxsz) * sizeof(__be32);
3123 }
3124 #endif /* CONFIG_NFSD_PNFS */
3125
3126
3127 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3128                             const struct nfsd4_op *op)
3129 {
3130         return (op_encode_hdr_size + 3) * sizeof(__be32);
3131 }
3132
3133 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3134                                 const struct nfsd4_op *op)
3135 {
3136         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3137
3138         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3139 }
3140
3141 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3142                                 const struct nfsd4_op *op)
3143 {
3144         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3145                 * sizeof(__be32);
3146 }
3147 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3148                                   const struct nfsd4_op *op)
3149 {
3150         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3151
3152         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3153 }
3154
3155 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3156                                    const struct nfsd4_op *op)
3157 {
3158         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3159                 * sizeof(__be32);
3160 }
3161
3162
3163 static const struct nfsd4_operation nfsd4_ops[] = {
3164         [OP_ACCESS] = {
3165                 .op_func = nfsd4_access,
3166                 .op_name = "OP_ACCESS",
3167                 .op_rsize_bop = nfsd4_access_rsize,
3168         },
3169         [OP_CLOSE] = {
3170                 .op_func = nfsd4_close,
3171                 .op_flags = OP_MODIFIES_SOMETHING,
3172                 .op_name = "OP_CLOSE",
3173                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3174                 .op_get_currentstateid = nfsd4_get_closestateid,
3175                 .op_set_currentstateid = nfsd4_set_closestateid,
3176         },
3177         [OP_COMMIT] = {
3178                 .op_func = nfsd4_commit,
3179                 .op_flags = OP_MODIFIES_SOMETHING,
3180                 .op_name = "OP_COMMIT",
3181                 .op_rsize_bop = nfsd4_commit_rsize,
3182         },
3183         [OP_CREATE] = {
3184                 .op_func = nfsd4_create,
3185                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3186                 .op_name = "OP_CREATE",
3187                 .op_rsize_bop = nfsd4_create_rsize,
3188         },
3189         [OP_DELEGRETURN] = {
3190                 .op_func = nfsd4_delegreturn,
3191                 .op_flags = OP_MODIFIES_SOMETHING,
3192                 .op_name = "OP_DELEGRETURN",
3193                 .op_rsize_bop = nfsd4_only_status_rsize,
3194                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3195         },
3196         [OP_GETATTR] = {
3197                 .op_func = nfsd4_getattr,
3198                 .op_flags = ALLOWED_ON_ABSENT_FS,
3199                 .op_rsize_bop = nfsd4_getattr_rsize,
3200                 .op_name = "OP_GETATTR",
3201         },
3202         [OP_GETFH] = {
3203                 .op_func = nfsd4_getfh,
3204                 .op_name = "OP_GETFH",
3205                 .op_rsize_bop = nfsd4_getfh_rsize,
3206         },
3207         [OP_LINK] = {
3208                 .op_func = nfsd4_link,
3209                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3210                                 | OP_CACHEME,
3211                 .op_name = "OP_LINK",
3212                 .op_rsize_bop = nfsd4_link_rsize,
3213         },
3214         [OP_LOCK] = {
3215                 .op_func = nfsd4_lock,
3216                 .op_release = nfsd4_lock_release,
3217                 .op_flags = OP_MODIFIES_SOMETHING |
3218                                 OP_NONTRIVIAL_ERROR_ENCODE,
3219                 .op_name = "OP_LOCK",
3220                 .op_rsize_bop = nfsd4_lock_rsize,
3221                 .op_set_currentstateid = nfsd4_set_lockstateid,
3222         },
3223         [OP_LOCKT] = {
3224                 .op_func = nfsd4_lockt,
3225                 .op_release = nfsd4_lockt_release,
3226                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3227                 .op_name = "OP_LOCKT",
3228                 .op_rsize_bop = nfsd4_lock_rsize,
3229         },
3230         [OP_LOCKU] = {
3231                 .op_func = nfsd4_locku,
3232                 .op_flags = OP_MODIFIES_SOMETHING,
3233                 .op_name = "OP_LOCKU",
3234                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3235                 .op_get_currentstateid = nfsd4_get_lockustateid,
3236         },
3237         [OP_LOOKUP] = {
3238                 .op_func = nfsd4_lookup,
3239                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3240                 .op_name = "OP_LOOKUP",
3241                 .op_rsize_bop = nfsd4_only_status_rsize,
3242         },
3243         [OP_LOOKUPP] = {
3244                 .op_func = nfsd4_lookupp,
3245                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3246                 .op_name = "OP_LOOKUPP",
3247                 .op_rsize_bop = nfsd4_only_status_rsize,
3248         },
3249         [OP_NVERIFY] = {
3250                 .op_func = nfsd4_nverify,
3251                 .op_name = "OP_NVERIFY",
3252                 .op_rsize_bop = nfsd4_only_status_rsize,
3253         },
3254         [OP_OPEN] = {
3255                 .op_func = nfsd4_open,
3256                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3257                 .op_name = "OP_OPEN",
3258                 .op_rsize_bop = nfsd4_open_rsize,
3259                 .op_set_currentstateid = nfsd4_set_openstateid,
3260         },
3261         [OP_OPEN_CONFIRM] = {
3262                 .op_func = nfsd4_open_confirm,
3263                 .op_flags = OP_MODIFIES_SOMETHING,
3264                 .op_name = "OP_OPEN_CONFIRM",
3265                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3266         },
3267         [OP_OPEN_DOWNGRADE] = {
3268                 .op_func = nfsd4_open_downgrade,
3269                 .op_flags = OP_MODIFIES_SOMETHING,
3270                 .op_name = "OP_OPEN_DOWNGRADE",
3271                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3272                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3273                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3274         },
3275         [OP_PUTFH] = {
3276                 .op_func = nfsd4_putfh,
3277                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3278                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3279                 .op_name = "OP_PUTFH",
3280                 .op_rsize_bop = nfsd4_only_status_rsize,
3281         },
3282         [OP_PUTPUBFH] = {
3283                 .op_func = nfsd4_putrootfh,
3284                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3285                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3286                 .op_name = "OP_PUTPUBFH",
3287                 .op_rsize_bop = nfsd4_only_status_rsize,
3288         },
3289         [OP_PUTROOTFH] = {
3290                 .op_func = nfsd4_putrootfh,
3291                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3292                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3293                 .op_name = "OP_PUTROOTFH",
3294                 .op_rsize_bop = nfsd4_only_status_rsize,
3295         },
3296         [OP_READ] = {
3297                 .op_func = nfsd4_read,
3298                 .op_release = nfsd4_read_release,
3299                 .op_name = "OP_READ",
3300                 .op_rsize_bop = nfsd4_read_rsize,
3301                 .op_get_currentstateid = nfsd4_get_readstateid,
3302         },
3303         [OP_READDIR] = {
3304                 .op_func = nfsd4_readdir,
3305                 .op_name = "OP_READDIR",
3306                 .op_rsize_bop = nfsd4_readdir_rsize,
3307         },
3308         [OP_READLINK] = {
3309                 .op_func = nfsd4_readlink,
3310                 .op_name = "OP_READLINK",
3311                 .op_rsize_bop = nfsd4_readlink_rsize,
3312         },
3313         [OP_REMOVE] = {
3314                 .op_func = nfsd4_remove,
3315                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3316                 .op_name = "OP_REMOVE",
3317                 .op_rsize_bop = nfsd4_remove_rsize,
3318         },
3319         [OP_RENAME] = {
3320                 .op_func = nfsd4_rename,
3321                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3322                 .op_name = "OP_RENAME",
3323                 .op_rsize_bop = nfsd4_rename_rsize,
3324         },
3325         [OP_RENEW] = {
3326                 .op_func = nfsd4_renew,
3327                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3328                                 | OP_MODIFIES_SOMETHING,
3329                 .op_name = "OP_RENEW",
3330                 .op_rsize_bop = nfsd4_only_status_rsize,
3331
3332         },
3333         [OP_RESTOREFH] = {
3334                 .op_func = nfsd4_restorefh,
3335                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3336                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3337                 .op_name = "OP_RESTOREFH",
3338                 .op_rsize_bop = nfsd4_only_status_rsize,
3339         },
3340         [OP_SAVEFH] = {
3341                 .op_func = nfsd4_savefh,
3342                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3343                 .op_name = "OP_SAVEFH",
3344                 .op_rsize_bop = nfsd4_only_status_rsize,
3345         },
3346         [OP_SECINFO] = {
3347                 .op_func = nfsd4_secinfo,
3348                 .op_release = nfsd4_secinfo_release,
3349                 .op_flags = OP_HANDLES_WRONGSEC,
3350                 .op_name = "OP_SECINFO",
3351                 .op_rsize_bop = nfsd4_secinfo_rsize,
3352         },
3353         [OP_SETATTR] = {
3354                 .op_func = nfsd4_setattr,
3355                 .op_name = "OP_SETATTR",
3356                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3357                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3358                 .op_rsize_bop = nfsd4_setattr_rsize,
3359                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3360         },
3361         [OP_SETCLIENTID] = {
3362                 .op_func = nfsd4_setclientid,
3363                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3364                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3365                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3366                 .op_name = "OP_SETCLIENTID",
3367                 .op_rsize_bop = nfsd4_setclientid_rsize,
3368         },
3369         [OP_SETCLIENTID_CONFIRM] = {
3370                 .op_func = nfsd4_setclientid_confirm,
3371                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3372                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3373                 .op_name = "OP_SETCLIENTID_CONFIRM",
3374                 .op_rsize_bop = nfsd4_only_status_rsize,
3375         },
3376         [OP_VERIFY] = {
3377                 .op_func = nfsd4_verify,
3378                 .op_name = "OP_VERIFY",
3379                 .op_rsize_bop = nfsd4_only_status_rsize,
3380         },
3381         [OP_WRITE] = {
3382                 .op_func = nfsd4_write,
3383                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3384                 .op_name = "OP_WRITE",
3385                 .op_rsize_bop = nfsd4_write_rsize,
3386                 .op_get_currentstateid = nfsd4_get_writestateid,
3387         },
3388         [OP_RELEASE_LOCKOWNER] = {
3389                 .op_func = nfsd4_release_lockowner,
3390                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3391                                 | OP_MODIFIES_SOMETHING,
3392                 .op_name = "OP_RELEASE_LOCKOWNER",
3393                 .op_rsize_bop = nfsd4_only_status_rsize,
3394         },
3395
3396         /* NFSv4.1 operations */
3397         [OP_EXCHANGE_ID] = {
3398                 .op_func = nfsd4_exchange_id,
3399                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3400                                 | OP_MODIFIES_SOMETHING,
3401                 .op_name = "OP_EXCHANGE_ID",
3402                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3403         },
3404         [OP_BACKCHANNEL_CTL] = {
3405                 .op_func = nfsd4_backchannel_ctl,
3406                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3407                 .op_name = "OP_BACKCHANNEL_CTL",
3408                 .op_rsize_bop = nfsd4_only_status_rsize,
3409         },
3410         [OP_BIND_CONN_TO_SESSION] = {
3411                 .op_func = nfsd4_bind_conn_to_session,
3412                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3413                                 | OP_MODIFIES_SOMETHING,
3414                 .op_name = "OP_BIND_CONN_TO_SESSION",
3415                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3416         },
3417         [OP_CREATE_SESSION] = {
3418                 .op_func = nfsd4_create_session,
3419                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3420                                 | OP_MODIFIES_SOMETHING,
3421                 .op_name = "OP_CREATE_SESSION",
3422                 .op_rsize_bop = nfsd4_create_session_rsize,
3423         },
3424         [OP_DESTROY_SESSION] = {
3425                 .op_func = nfsd4_destroy_session,
3426                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3427                                 | OP_MODIFIES_SOMETHING,
3428                 .op_name = "OP_DESTROY_SESSION",
3429                 .op_rsize_bop = nfsd4_only_status_rsize,
3430         },
3431         [OP_SEQUENCE] = {
3432                 .op_func = nfsd4_sequence,
3433                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3434                 .op_name = "OP_SEQUENCE",
3435                 .op_rsize_bop = nfsd4_sequence_rsize,
3436         },
3437         [OP_DESTROY_CLIENTID] = {
3438                 .op_func = nfsd4_destroy_clientid,
3439                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3440                                 | OP_MODIFIES_SOMETHING,
3441                 .op_name = "OP_DESTROY_CLIENTID",
3442                 .op_rsize_bop = nfsd4_only_status_rsize,
3443         },
3444         [OP_RECLAIM_COMPLETE] = {
3445                 .op_func = nfsd4_reclaim_complete,
3446                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3447                 .op_name = "OP_RECLAIM_COMPLETE",
3448                 .op_rsize_bop = nfsd4_only_status_rsize,
3449         },
3450         [OP_SECINFO_NO_NAME] = {
3451                 .op_func = nfsd4_secinfo_no_name,
3452                 .op_release = nfsd4_secinfo_no_name_release,
3453                 .op_flags = OP_HANDLES_WRONGSEC,
3454                 .op_name = "OP_SECINFO_NO_NAME",
3455                 .op_rsize_bop = nfsd4_secinfo_rsize,
3456         },
3457         [OP_TEST_STATEID] = {
3458                 .op_func = nfsd4_test_stateid,
3459                 .op_flags = ALLOWED_WITHOUT_FH,
3460                 .op_name = "OP_TEST_STATEID",
3461                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3462         },
3463         [OP_FREE_STATEID] = {
3464                 .op_func = nfsd4_free_stateid,
3465                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3466                 .op_name = "OP_FREE_STATEID",
3467                 .op_get_currentstateid = nfsd4_get_freestateid,
3468                 .op_rsize_bop = nfsd4_only_status_rsize,
3469         },
3470 #ifdef CONFIG_NFSD_PNFS
3471         [OP_GETDEVICEINFO] = {
3472                 .op_func = nfsd4_getdeviceinfo,
3473                 .op_release = nfsd4_getdeviceinfo_release,
3474                 .op_flags = ALLOWED_WITHOUT_FH,
3475                 .op_name = "OP_GETDEVICEINFO",
3476                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3477         },
3478         [OP_LAYOUTGET] = {
3479                 .op_func = nfsd4_layoutget,
3480                 .op_release = nfsd4_layoutget_release,
3481                 .op_flags = OP_MODIFIES_SOMETHING,
3482                 .op_name = "OP_LAYOUTGET",
3483                 .op_rsize_bop = nfsd4_layoutget_rsize,
3484         },
3485         [OP_LAYOUTCOMMIT] = {
3486                 .op_func = nfsd4_layoutcommit,
3487                 .op_flags = OP_MODIFIES_SOMETHING,
3488                 .op_name = "OP_LAYOUTCOMMIT",
3489                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3490         },
3491         [OP_LAYOUTRETURN] = {
3492                 .op_func = nfsd4_layoutreturn,
3493                 .op_flags = OP_MODIFIES_SOMETHING,
3494                 .op_name = "OP_LAYOUTRETURN",
3495                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3496         },
3497 #endif /* CONFIG_NFSD_PNFS */
3498
3499         /* NFSv4.2 operations */
3500         [OP_ALLOCATE] = {
3501                 .op_func = nfsd4_allocate,
3502                 .op_flags = OP_MODIFIES_SOMETHING,
3503                 .op_name = "OP_ALLOCATE",
3504                 .op_rsize_bop = nfsd4_only_status_rsize,
3505         },
3506         [OP_DEALLOCATE] = {
3507                 .op_func = nfsd4_deallocate,
3508                 .op_flags = OP_MODIFIES_SOMETHING,
3509                 .op_name = "OP_DEALLOCATE",
3510                 .op_rsize_bop = nfsd4_only_status_rsize,
3511         },
3512         [OP_CLONE] = {
3513                 .op_func = nfsd4_clone,
3514                 .op_flags = OP_MODIFIES_SOMETHING,
3515                 .op_name = "OP_CLONE",
3516                 .op_rsize_bop = nfsd4_only_status_rsize,
3517         },
3518         [OP_COPY] = {
3519                 .op_func = nfsd4_copy,
3520                 .op_flags = OP_MODIFIES_SOMETHING,
3521                 .op_name = "OP_COPY",
3522                 .op_rsize_bop = nfsd4_copy_rsize,
3523         },
3524         [OP_READ_PLUS] = {
3525                 .op_func = nfsd4_read,
3526                 .op_release = nfsd4_read_release,
3527                 .op_name = "OP_READ_PLUS",
3528                 .op_rsize_bop = nfsd4_read_plus_rsize,
3529                 .op_get_currentstateid = nfsd4_get_readstateid,
3530         },
3531         [OP_SEEK] = {
3532                 .op_func = nfsd4_seek,
3533                 .op_name = "OP_SEEK",
3534                 .op_rsize_bop = nfsd4_seek_rsize,
3535         },
3536         [OP_OFFLOAD_STATUS] = {
3537                 .op_func = nfsd4_offload_status,
3538                 .op_name = "OP_OFFLOAD_STATUS",
3539                 .op_rsize_bop = nfsd4_offload_status_rsize,
3540         },
3541         [OP_OFFLOAD_CANCEL] = {
3542                 .op_func = nfsd4_offload_cancel,
3543                 .op_flags = OP_MODIFIES_SOMETHING,
3544                 .op_name = "OP_OFFLOAD_CANCEL",
3545                 .op_rsize_bop = nfsd4_only_status_rsize,
3546         },
3547         [OP_COPY_NOTIFY] = {
3548                 .op_func = nfsd4_copy_notify,
3549                 .op_flags = OP_MODIFIES_SOMETHING,
3550                 .op_name = "OP_COPY_NOTIFY",
3551                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3552         },
3553         [OP_GETXATTR] = {
3554                 .op_func = nfsd4_getxattr,
3555                 .op_name = "OP_GETXATTR",
3556                 .op_rsize_bop = nfsd4_getxattr_rsize,
3557         },
3558         [OP_SETXATTR] = {
3559                 .op_func = nfsd4_setxattr,
3560                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3561                 .op_name = "OP_SETXATTR",
3562                 .op_rsize_bop = nfsd4_setxattr_rsize,
3563         },
3564         [OP_LISTXATTRS] = {
3565                 .op_func = nfsd4_listxattrs,
3566                 .op_name = "OP_LISTXATTRS",
3567                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3568         },
3569         [OP_REMOVEXATTR] = {
3570                 .op_func = nfsd4_removexattr,
3571                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3572                 .op_name = "OP_REMOVEXATTR",
3573                 .op_rsize_bop = nfsd4_removexattr_rsize,
3574         },
3575 };
3576
3577 /**
3578  * nfsd4_spo_must_allow - Determine if the compound op contains an
3579  * operation that is allowed to be sent with machine credentials
3580  *
3581  * @rqstp: a pointer to the struct svc_rqst
3582  *
3583  * Checks to see if the compound contains a spo_must_allow op
3584  * and confirms that it was sent with the proper machine creds.
3585  */
3586
3587 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3588 {
3589         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3590         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3591         struct nfsd4_op *this;
3592         struct nfsd4_compound_state *cstate = &resp->cstate;
3593         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3594         u32 opiter;
3595
3596         if (!cstate->minorversion)
3597                 return false;
3598
3599         if (cstate->spo_must_allowed)
3600                 return true;
3601
3602         opiter = resp->opcnt;
3603         while (opiter < argp->opcnt) {
3604                 this = &argp->ops[opiter++];
3605                 if (test_bit(this->opnum, allow->u.longs) &&
3606                         cstate->clp->cl_mach_cred &&
3607                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3608                         cstate->spo_must_allowed = true;
3609                         return true;
3610                 }
3611         }
3612         cstate->spo_must_allowed = false;
3613         return false;
3614 }
3615
3616 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3617 {
3618         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3619                 return op_encode_hdr_size * sizeof(__be32);
3620
3621         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3622         return OPDESC(op)->op_rsize_bop(rqstp, op);
3623 }
3624
3625 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3626 {
3627         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3628                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3629                         op->opnum, nfsd4_op_name(op->opnum));
3630                 WARN_ON_ONCE(1);
3631         }
3632 }
3633
3634 static const char *nfsd4_op_name(unsigned opnum)
3635 {
3636         if (opnum < ARRAY_SIZE(nfsd4_ops))
3637                 return nfsd4_ops[opnum].op_name;
3638         return "unknown_operation";
3639 }
3640
3641 static const struct svc_procedure nfsd_procedures4[2] = {
3642         [NFSPROC4_NULL] = {
3643                 .pc_func = nfsd4_proc_null,
3644                 .pc_decode = nfssvc_decode_voidarg,
3645                 .pc_encode = nfssvc_encode_voidres,
3646                 .pc_argsize = sizeof(struct nfsd_voidargs),
3647                 .pc_argzero = sizeof(struct nfsd_voidargs),
3648                 .pc_ressize = sizeof(struct nfsd_voidres),
3649                 .pc_cachetype = RC_NOCACHE,
3650                 .pc_xdrressize = 1,
3651                 .pc_name = "NULL",
3652         },
3653         [NFSPROC4_COMPOUND] = {
3654                 .pc_func = nfsd4_proc_compound,
3655                 .pc_decode = nfs4svc_decode_compoundargs,
3656                 .pc_encode = nfs4svc_encode_compoundres,
3657                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3658                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3659                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3660                 .pc_release = nfsd4_release_compoundargs,
3661                 .pc_cachetype = RC_NOCACHE,
3662                 .pc_xdrressize = NFSD_BUFSIZE/4,
3663                 .pc_name = "COMPOUND",
3664         },
3665 };
3666
3667 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3668                               nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3669 const struct svc_version nfsd_version4 = {
3670         .vs_vers                = 4,
3671         .vs_nproc               = ARRAY_SIZE(nfsd_procedures4),
3672         .vs_proc                = nfsd_procedures4,
3673         .vs_count               = nfsd_count4,
3674         .vs_dispatch            = nfsd_dispatch,
3675         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3676         .vs_rpcb_optnl          = true,
3677         .vs_need_cong_ctrl      = true,
3678 };