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 static void cleanup_async_copy(struct nfsd4_copy *copy);
  61
  62 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
  63 static int nfsd4_ssc_umount_timeout = 900000;           /* default to 15 mins */
  64 module_param(nfsd4_ssc_umount_timeout, int, 0644);
  65 MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
  66                 "idle msecs before unmount export from source server");
  67 #endif
  68
  69 #define NFSDDBG_FACILITY                NFSDDBG_PROC
  70
  71 static u32 nfsd_attrmask[] = {
  72         NFSD_WRITEABLE_ATTRS_WORD0,
  73         NFSD_WRITEABLE_ATTRS_WORD1,
  74         NFSD_WRITEABLE_ATTRS_WORD2
  75 };
  76
  77 static u32 nfsd41_ex_attrmask[] = {
  78         NFSD_SUPPATTR_EXCLCREAT_WORD0,
  79         NFSD_SUPPATTR_EXCLCREAT_WORD1,
  80         NFSD_SUPPATTR_EXCLCREAT_WORD2
  81 };
  82
  83 static __be32
  84 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
  85                    u32 *bmval, u32 *writable)
  86 {
  87         struct dentry *dentry = cstate->current_fh.fh_dentry;
  88         struct svc_export *exp = cstate->current_fh.fh_export;
  89
  90         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
  91                 return nfserr_attrnotsupp;
  92         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
  93                 return nfserr_attrnotsupp;
  94         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
  95                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
  96                 return nfserr_attrnotsupp;
  97         if (writable && !bmval_is_subset(bmval, writable))
  98                 return nfserr_inval;
  99         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
 100                         (bmval[1] & FATTR4_WORD1_MODE))
 101                 return nfserr_inval;
 102         return nfs_ok;
 103 }
 104
 105 static __be32
 106 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
 107         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 108 {
 109         __be32 status = nfs_ok;
 110
 111         if (open->op_create == NFS4_OPEN_CREATE) {
 112                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
 113                     || open->op_createmode == NFS4_CREATE_GUARDED)
 114                         status = check_attr_support(rqstp, cstate,
 115                                         open->op_bmval, nfsd_attrmask);
 116                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
 117                         status = check_attr_support(rqstp, cstate,
 118                                         open->op_bmval, nfsd41_ex_attrmask);
 119         }
 120
 121         return status;
 122 }
 123
 124 static int
 125 is_create_with_attrs(struct nfsd4_open *open)
 126 {
 127         return open->op_create == NFS4_OPEN_CREATE
 128                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
 129                     || open->op_createmode == NFS4_CREATE_GUARDED
 130                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
 131 }
 132
 133 static inline void
 134 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
 135 {
 136         fh_put(dst);
 137         dget(src->fh_dentry);
 138         if (src->fh_export)
 139                 exp_get(src->fh_export);
 140         *dst = *src;
 141 }
 142
 143 static __be32
 144 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
 145 {
 146
 147         if (open->op_truncate &&
 148                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
 149                 return nfserr_inval;
 150
 151         accmode |= NFSD_MAY_READ_IF_EXEC;
 152
 153         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
 154                 accmode |= NFSD_MAY_READ;
 155         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
 156                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
 157         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
 158                 accmode |= NFSD_MAY_WRITE;
 159
 160         return fh_verify(rqstp, current_fh, S_IFREG, accmode);
 161 }
 162
 163 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh, u32 minor_version)
 164 {
 165         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
 166
 167         if (S_ISREG(mode))
 168                 return nfs_ok;
 169         if (S_ISDIR(mode))
 170                 return nfserr_isdir;
 171         if (S_ISLNK(mode))
 172                 return nfserr_symlink;
 173
 174         /* RFC 7530 - 16.16.6 */
 175         if (minor_version == 0)
 176                 return nfserr_symlink;
 177         else
 178                 return nfserr_wrong_type;
 179
 180 }
 181
 182 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
 183 {
 184         if (nfsd4_has_session(cstate))
 185                 return;
 186         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
 187                         &resfh->fh_handle);
 188 }
 189
 190 static inline bool nfsd4_create_is_exclusive(int createmode)
 191 {
 192         return createmode == NFS4_CREATE_EXCLUSIVE ||
 193                 createmode == NFS4_CREATE_EXCLUSIVE4_1;
 194 }
 195
 196 static __be32
 197 nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
 198                  struct nfsd4_open *open)
 199 {
 200         struct file *filp;
 201         struct path path;
 202         int oflags;
 203
 204         oflags = O_CREAT | O_LARGEFILE;
 205         switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
 206         case NFS4_SHARE_ACCESS_WRITE:
 207                 oflags |= O_WRONLY;
 208                 break;
 209         case NFS4_SHARE_ACCESS_BOTH:
 210                 oflags |= O_RDWR;
 211                 break;
 212         default:
 213                 oflags |= O_RDONLY;
 214         }
 215
 216         path.mnt = fhp->fh_export->ex_path.mnt;
 217         path.dentry = child;
 218         filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
 219                              current_cred());
 220         if (IS_ERR(filp))
 221                 return nfserrno(PTR_ERR(filp));
 222
 223         open->op_filp = filp;
 224         return nfs_ok;
 225 }
 226
 227 /*
 228  * Implement NFSv4's unchecked, guarded, and exclusive create
 229  * semantics for regular files. Open state for this new file is
 230  * subsequently fabricated in nfsd4_process_open2().
 231  *
 232  * Upon return, caller must release @fhp and @resfhp.
 233  */
 234 static __be32
 235 nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
 236                   struct svc_fh *resfhp, struct nfsd4_open *open)
 237 {
 238         struct iattr *iap = &open->op_iattr;
 239         struct nfsd_attrs attrs = {
 240                 .na_iattr       = iap,
 241                 .na_seclabel    = &open->op_label,
 242         };
 243         struct dentry *parent, *child;
 244         __u32 v_mtime, v_atime;
 245         struct inode *inode;
 246         __be32 status;
 247         int host_err;
 248
 249         if (isdotent(open->op_fname, open->op_fnamelen))
 250                 return nfserr_exist;
 251         if (!(iap->ia_valid & ATTR_MODE))
 252                 iap->ia_mode = 0;
 253
 254         status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
 255         if (status != nfs_ok)
 256                 return status;
 257         parent = fhp->fh_dentry;
 258         inode = d_inode(parent);
 259
 260         host_err = fh_want_write(fhp);
 261         if (host_err)
 262                 return nfserrno(host_err);
 263
 264         if (is_create_with_attrs(open))
 265                 nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
 266
 267         inode_lock_nested(inode, I_MUTEX_PARENT);
 268
 269         child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
 270         if (IS_ERR(child)) {
 271                 status = nfserrno(PTR_ERR(child));
 272                 goto out;
 273         }
 274
 275         if (d_really_is_negative(child)) {
 276                 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
 277                 if (status != nfs_ok)
 278                         goto out;
 279         }
 280
 281         status = fh_compose(resfhp, fhp->fh_export, child, fhp);
 282         if (status != nfs_ok)
 283                 goto out;
 284
 285         v_mtime = 0;
 286         v_atime = 0;
 287         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 288                 u32 *verifier = (u32 *)open->op_verf.data;
 289
 290                 /*
 291                  * Solaris 7 gets confused (bugid 4218508) if these have
 292                  * the high bit set, as do xfs filesystems without the
 293                  * "bigtime" feature. So just clear the high bits. If this
 294                  * is ever changed to use different attrs for storing the
 295                  * verifier, then do_open_lookup() will also need to be
 296                  * fixed accordingly.
 297                  */
 298                 v_mtime = verifier[0] & 0x7fffffff;
 299                 v_atime = verifier[1] & 0x7fffffff;
 300         }
 301
 302         if (d_really_is_positive(child)) {
 303                 /* NFSv4 protocol requires change attributes even though
 304                  * no change happened.
 305                  */
 306                 status = fh_fill_both_attrs(fhp);
 307                 if (status != nfs_ok)
 308                         goto out;
 309
 310                 switch (open->op_createmode) {
 311                 case NFS4_CREATE_UNCHECKED:
 312                         if (!d_is_reg(child))
 313                                 break;
 314
 315                         /*
 316                          * In NFSv4, we don't want to truncate the file
 317                          * now. This would be wrong if the OPEN fails for
 318                          * some other reason. Furthermore, if the size is
 319                          * nonzero, we should ignore it according to spec!
 320                          */
 321                         open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
 322                                                 !iap->ia_size;
 323                         break;
 324                 case NFS4_CREATE_GUARDED:
 325                         status = nfserr_exist;
 326                         break;
 327                 case NFS4_CREATE_EXCLUSIVE:
 328                         if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
 329                             inode_get_atime_sec(d_inode(child)) == v_atime &&
 330                             d_inode(child)->i_size == 0) {
 331                                 open->op_created = true;
 332                                 break;          /* subtle */
 333                         }
 334                         status = nfserr_exist;
 335                         break;
 336                 case NFS4_CREATE_EXCLUSIVE4_1:
 337                         if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
 338                             inode_get_atime_sec(d_inode(child)) == v_atime &&
 339                             d_inode(child)->i_size == 0) {
 340                                 open->op_created = true;
 341                                 goto set_attr;  /* subtle */
 342                         }
 343                         status = nfserr_exist;
 344                 }
 345                 goto out;
 346         }
 347
 348         if (!IS_POSIXACL(inode))
 349                 iap->ia_mode &= ~current_umask();
 350
 351         status = fh_fill_pre_attrs(fhp);
 352         if (status != nfs_ok)
 353                 goto out;
 354         status = nfsd4_vfs_create(fhp, child, open);
 355         if (status != nfs_ok)
 356                 goto out;
 357         open->op_created = true;
 358         fh_fill_post_attrs(fhp);
 359
 360         /* A newly created file already has a file size of zero. */
 361         if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
 362                 iap->ia_valid &= ~ATTR_SIZE;
 363         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 364                 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
 365                                 ATTR_MTIME_SET|ATTR_ATIME_SET;
 366                 iap->ia_mtime.tv_sec = v_mtime;
 367                 iap->ia_atime.tv_sec = v_atime;
 368                 iap->ia_mtime.tv_nsec = 0;
 369                 iap->ia_atime.tv_nsec = 0;
 370         }
 371
 372 set_attr:
 373         status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
 374
 375         if (attrs.na_labelerr)
 376                 open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 377         if (attrs.na_aclerr)
 378                 open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
 379 out:
 380         inode_unlock(inode);
 381         nfsd_attrs_free(&attrs);
 382         if (child && !IS_ERR(child))
 383                 dput(child);
 384         fh_drop_write(fhp);
 385         return status;
 386 }
 387
 388 /**
 389  * set_change_info - set up the change_info4 for a reply
 390  * @cinfo: pointer to nfsd4_change_info to be populated
 391  * @fhp: pointer to svc_fh to use as source
 392  *
 393  * Many operations in NFSv4 require change_info4 in the reply. This function
 394  * populates that from the info that we (should!) have already collected. In
 395  * the event that we didn't get any pre-attrs, just zero out both.
 396  */
 397 static void
 398 set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp)
 399 {
 400         cinfo->atomic = (u32)(fhp->fh_pre_saved && fhp->fh_post_saved && !fhp->fh_no_atomic_attr);
 401         cinfo->before_change = fhp->fh_pre_change;
 402         cinfo->after_change = fhp->fh_post_change;
 403
 404         /*
 405          * If fetching the pre-change attributes failed, then we should
 406          * have already failed the whole operation. We could have still
 407          * failed to fetch post-change attributes however.
 408          *
 409          * If we didn't get post-op attrs, just zero-out the after
 410          * field since we don't know what it should be. If the pre_saved
 411          * field isn't set for some reason, throw warning and just copy
 412          * whatever is in the after field.
 413          */
 414         if (WARN_ON_ONCE(!fhp->fh_pre_saved))
 415                 cinfo->before_change = 0;
 416         if (!fhp->fh_post_saved)
 417                 cinfo->after_change = cinfo->before_change + 1;
 418 }
 419
 420 static __be32
 421 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
 422 {
 423         struct svc_fh *current_fh = &cstate->current_fh;
 424         int accmode;
 425         __be32 status;
 426
 427         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
 428         if (!*resfh)
 429                 return nfserr_jukebox;
 430         fh_init(*resfh, NFS4_FHSIZE);
 431         open->op_truncate = false;
 432
 433         if (open->op_create) {
 434                 /* FIXME: check session persistence and pnfs flags.
 435                  * The nfsv4.1 spec requires the following semantics:
 436                  *
 437                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
 438                  * Reply Cache  | server |                 |
 439                  * -------------+--------+-----------------+--------------------
 440                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
 441                  *              |        |                 | (SHOULD)
 442                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
 443                  *              |        |                 | (SHOULD NOT)
 444                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
 445                  * yes          | no     | GUARDED4        | GUARDED4
 446                  * yes          | yes    | GUARDED4        | GUARDED4
 447                  */
 448
 449                 current->fs->umask = open->op_umask;
 450                 status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
 451                 current->fs->umask = 0;
 452
 453                 /*
 454                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
 455                  * use the returned bitmask to indicate which attributes
 456                  * we used to store the verifier:
 457                  */
 458                 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
 459                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
 460                                                 FATTR4_WORD1_TIME_MODIFY);
 461         } else {
 462                 status = nfsd_lookup(rqstp, current_fh,
 463                                      open->op_fname, open->op_fnamelen, *resfh);
 464                 if (status == nfs_ok)
 465                         /* NFSv4 protocol requires change attributes even though
 466                          * no change happened.
 467                          */
 468                         status = fh_fill_both_attrs(current_fh);
 469         }
 470         if (status)
 471                 goto out;
 472         status = nfsd_check_obj_isreg(*resfh, cstate->minorversion);
 473         if (status)
 474                 goto out;
 475
 476         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
 477         accmode = NFSD_MAY_NOP;
 478         if (open->op_created ||
 479                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
 480                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
 481         status = do_open_permission(rqstp, *resfh, open, accmode);
 482         set_change_info(&open->op_cinfo, current_fh);
 483 out:
 484         return status;
 485 }
 486
 487 static __be32
 488 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 489 {
 490         struct svc_fh *current_fh = &cstate->current_fh;
 491         int accmode = 0;
 492
 493         /* We don't know the target directory, and therefore can not
 494         * set the change info
 495         */
 496
 497         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
 498
 499         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
 500
 501         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
 502                 (open->op_iattr.ia_size == 0);
 503         /*
 504          * In the delegation case, the client is telling us about an
 505          * open that it *already* performed locally, some time ago.  We
 506          * should let it succeed now if possible.
 507          *
 508          * In the case of a CLAIM_FH open, on the other hand, the client
 509          * may be counting on us to enforce permissions (the Linux 4.1
 510          * client uses this for normal opens, for example).
 511          */
 512         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
 513                 accmode = NFSD_MAY_OWNER_OVERRIDE;
 514
 515         return do_open_permission(rqstp, current_fh, open, accmode);
 516 }
 517
 518 static void
 519 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
 520 {
 521         struct nfsd4_sessionid *sid =
 522                         (struct nfsd4_sessionid *)session->se_sessionid.data;
 523
 524         clid->cl_boot = sid->clientid.cl_boot;
 525         clid->cl_id = sid->clientid.cl_id;
 526 }
 527
 528 static __be32
 529 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 530            union nfsd4_op_u *u)
 531 {
 532         struct nfsd4_open *open = &u->open;
 533         __be32 status;
 534         struct svc_fh *resfh = NULL;
 535         struct net *net = SVC_NET(rqstp);
 536         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 537         bool reclaim = false;
 538
 539         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
 540                 (int)open->op_fnamelen, open->op_fname,
 541                 open->op_openowner);
 542
 543         open->op_filp = NULL;
 544         open->op_rqstp = rqstp;
 545
 546         /* This check required by spec. */
 547         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
 548                 return nfserr_inval;
 549
 550         open->op_created = false;
 551         /*
 552          * RFC5661 18.51.3
 553          * Before RECLAIM_COMPLETE done, server should deny new lock
 554          */
 555         if (nfsd4_has_session(cstate) &&
 556             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
 557             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 558                 return nfserr_grace;
 559
 560         if (nfsd4_has_session(cstate))
 561                 copy_clientid(&open->op_clientid, cstate->session);
 562
 563         /* check seqid for replay. set nfs4_owner */
 564         status = nfsd4_process_open1(cstate, open, nn);
 565         if (status == nfserr_replay_me) {
 566                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
 567                 fh_put(&cstate->current_fh);
 568                 fh_copy_shallow(&cstate->current_fh.fh_handle,
 569                                 &rp->rp_openfh);
 570                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 571                 if (status)
 572                         dprintk("nfsd4_open: replay failed"
 573                                 " restoring previous filehandle\n");
 574                 else
 575                         status = nfserr_replay_me;
 576         }
 577         if (status)
 578                 goto out;
 579         if (open->op_xdr_error) {
 580                 status = open->op_xdr_error;
 581                 goto out;
 582         }
 583
 584         status = nfsd4_check_open_attributes(rqstp, cstate, open);
 585         if (status)
 586                 goto out;
 587
 588         /* Openowner is now set, so sequence id will get bumped.  Now we need
 589          * these checks before we do any creates: */
 590         status = nfserr_grace;
 591         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 592                 goto out;
 593         status = nfserr_no_grace;
 594         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
 595                 goto out;
 596
 597         switch (open->op_claim_type) {
 598         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 599         case NFS4_OPEN_CLAIM_NULL:
 600                 status = do_open_lookup(rqstp, cstate, open, &resfh);
 601                 if (status)
 602                         goto out;
 603                 break;
 604         case NFS4_OPEN_CLAIM_PREVIOUS:
 605                 status = nfs4_check_open_reclaim(cstate->clp);
 606                 if (status)
 607                         goto out;
 608                 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
 609                 reclaim = true;
 610                 fallthrough;
 611         case NFS4_OPEN_CLAIM_FH:
 612         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 613                 status = do_open_fhandle(rqstp, cstate, open);
 614                 if (status)
 615                         goto out;
 616                 resfh = &cstate->current_fh;
 617                 break;
 618         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 619         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 620                 status = nfserr_notsupp;
 621                 goto out;
 622         default:
 623                 status = nfserr_inval;
 624                 goto out;
 625         }
 626
 627         status = nfsd4_process_open2(rqstp, resfh, open);
 628         if (status && open->op_created)
 629                 pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
 630                         be32_to_cpu(status));
 631         if (reclaim && !status)
 632                 nn->somebody_reclaimed = true;
 633 out:
 634         if (open->op_filp) {
 635                 fput(open->op_filp);
 636                 open->op_filp = NULL;
 637         }
 638         if (resfh && resfh != &cstate->current_fh) {
 639                 fh_dup2(&cstate->current_fh, resfh);
 640                 fh_put(resfh);
 641                 kfree(resfh);
 642         }
 643         nfsd4_cleanup_open_state(cstate, open);
 644         nfsd4_bump_seqid(cstate, status);
 645         return status;
 646 }
 647
 648 /*
 649  * OPEN is the only seqid-mutating operation whose decoding can fail
 650  * with a seqid-mutating error (specifically, decoding of user names in
 651  * the attributes).  Therefore we have to do some processing to look up
 652  * the stateowner so that we can bump the seqid.
 653  */
 654 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
 655 {
 656         struct nfsd4_open *open = &op->u.open;
 657
 658         if (!seqid_mutating_err(ntohl(op->status)))
 659                 return op->status;
 660         if (nfsd4_has_session(cstate))
 661                 return op->status;
 662         open->op_xdr_error = op->status;
 663         return nfsd4_open(rqstp, cstate, &op->u);
 664 }
 665
 666 /*
 667  * filehandle-manipulating ops.
 668  */
 669 static __be32
 670 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 671             union nfsd4_op_u *u)
 672 {
 673         u->getfh = &cstate->current_fh;
 674         return nfs_ok;
 675 }
 676
 677 static __be32
 678 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 679             union nfsd4_op_u *u)
 680 {
 681         struct nfsd4_putfh *putfh = &u->putfh;
 682         __be32 ret;
 683
 684         fh_put(&cstate->current_fh);
 685         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
 686         memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
 687                putfh->pf_fhlen);
 688         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
 689 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
 690         if (ret == nfserr_stale && putfh->no_verify) {
 691                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
 692                 ret = 0;
 693         }
 694 #endif
 695         return ret;
 696 }
 697
 698 static __be32
 699 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 700                 union nfsd4_op_u *u)
 701 {
 702         fh_put(&cstate->current_fh);
 703
 704         return exp_pseudoroot(rqstp, &cstate->current_fh);
 705 }
 706
 707 static __be32
 708 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 709                 union nfsd4_op_u *u)
 710 {
 711         if (!cstate->save_fh.fh_dentry)
 712                 return nfserr_restorefh;
 713
 714         fh_dup2(&cstate->current_fh, &cstate->save_fh);
 715         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
 716                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
 717                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
 718         }
 719         return nfs_ok;
 720 }
 721
 722 static __be32
 723 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 724              union nfsd4_op_u *u)
 725 {
 726         fh_dup2(&cstate->save_fh, &cstate->current_fh);
 727         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
 728                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
 729                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
 730         }
 731         return nfs_ok;
 732 }
 733
 734 /*
 735  * misc nfsv4 ops
 736  */
 737 static __be32
 738 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 739              union nfsd4_op_u *u)
 740 {
 741         struct nfsd4_access *access = &u->access;
 742         u32 access_full;
 743
 744         access_full = NFS3_ACCESS_FULL;
 745         if (cstate->minorversion >= 2)
 746                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
 747                                NFS4_ACCESS_XAWRITE;
 748
 749         if (access->ac_req_access & ~access_full)
 750                 return nfserr_inval;
 751
 752         access->ac_resp_access = access->ac_req_access;
 753         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
 754                            &access->ac_supported);
 755 }
 756
 757 static __be32
 758 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 759              union nfsd4_op_u *u)
 760 {
 761         struct nfsd4_commit *commit = &u->commit;
 762         struct nfsd_file *nf;
 763         __be32 status;
 764
 765         status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
 766                                    NFSD_MAY_NOT_BREAK_LEASE, &nf);
 767         if (status != nfs_ok)
 768                 return status;
 769
 770         status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
 771                              commit->co_count,
 772                              (__be32 *)commit->co_verf.data);
 773         nfsd_file_put(nf);
 774         return status;
 775 }
 776
 777 static __be32
 778 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 779              union nfsd4_op_u *u)
 780 {
 781         struct nfsd4_create *create = &u->create;
 782         struct nfsd_attrs attrs = {
 783                 .na_iattr       = &create->cr_iattr,
 784                 .na_seclabel    = &create->cr_label,
 785         };
 786         struct svc_fh resfh;
 787         __be32 status;
 788         dev_t rdev;
 789
 790         fh_init(&resfh, NFS4_FHSIZE);
 791
 792         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 793         if (status)
 794                 return status;
 795
 796         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 797                                     nfsd_attrmask);
 798         if (status)
 799                 return status;
 800
 801         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 802         current->fs->umask = create->cr_umask;
 803         switch (create->cr_type) {
 804         case NF4LNK:
 805                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 806                                       create->cr_name, create->cr_namelen,
 807                                       create->cr_data, &attrs, &resfh);
 808                 break;
 809
 810         case NF4BLK:
 811                 status = nfserr_inval;
 812                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 813                 if (MAJOR(rdev) != create->cr_specdata1 ||
 814                     MINOR(rdev) != create->cr_specdata2)
 815                         goto out_umask;
 816                 status = nfsd_create(rqstp, &cstate->current_fh,
 817                                      create->cr_name, create->cr_namelen,
 818                                      &attrs, S_IFBLK, rdev, &resfh);
 819                 break;
 820
 821         case NF4CHR:
 822                 status = nfserr_inval;
 823                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 824                 if (MAJOR(rdev) != create->cr_specdata1 ||
 825                     MINOR(rdev) != create->cr_specdata2)
 826                         goto out_umask;
 827                 status = nfsd_create(rqstp, &cstate->current_fh,
 828                                      create->cr_name, create->cr_namelen,
 829                                      &attrs, S_IFCHR, rdev, &resfh);
 830                 break;
 831
 832         case NF4SOCK:
 833                 status = nfsd_create(rqstp, &cstate->current_fh,
 834                                      create->cr_name, create->cr_namelen,
 835                                      &attrs, S_IFSOCK, 0, &resfh);
 836                 break;
 837
 838         case NF4FIFO:
 839                 status = nfsd_create(rqstp, &cstate->current_fh,
 840                                      create->cr_name, create->cr_namelen,
 841                                      &attrs, S_IFIFO, 0, &resfh);
 842                 break;
 843
 844         case NF4DIR:
 845                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 846                 status = nfsd_create(rqstp, &cstate->current_fh,
 847                                      create->cr_name, create->cr_namelen,
 848                                      &attrs, S_IFDIR, 0, &resfh);
 849                 break;
 850
 851         default:
 852                 status = nfserr_badtype;
 853         }
 854
 855         if (status)
 856                 goto out;
 857
 858         if (attrs.na_labelerr)
 859                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 860         if (attrs.na_aclerr)
 861                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 862         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 863         fh_dup2(&cstate->current_fh, &resfh);
 864 out:
 865         fh_put(&resfh);
 866 out_umask:
 867         current->fs->umask = 0;
 868         nfsd_attrs_free(&attrs);
 869         return status;
 870 }
 871
 872 static __be32
 873 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 874               union nfsd4_op_u *u)
 875 {
 876         struct nfsd4_getattr *getattr = &u->getattr;
 877         __be32 status;
 878
 879         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 880         if (status)
 881                 return status;
 882
 883         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 884                 return nfserr_inval;
 885
 886         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 887         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 888         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 889
 890         getattr->ga_fhp = &cstate->current_fh;
 891         return nfs_ok;
 892 }
 893
 894 static __be32
 895 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 896            union nfsd4_op_u *u)
 897 {
 898         struct nfsd4_link *link = &u->link;
 899         __be32 status;
 900
 901         status = nfsd_link(rqstp, &cstate->current_fh,
 902                            link->li_name, link->li_namelen, &cstate->save_fh);
 903         if (!status)
 904                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 905         return status;
 906 }
 907
 908 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 909 {
 910         struct svc_fh tmp_fh;
 911         __be32 ret;
 912
 913         fh_init(&tmp_fh, NFS4_FHSIZE);
 914         ret = exp_pseudoroot(rqstp, &tmp_fh);
 915         if (ret)
 916                 return ret;
 917         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 918                 fh_put(&tmp_fh);
 919                 return nfserr_noent;
 920         }
 921         fh_put(&tmp_fh);
 922         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 923 }
 924
 925 static __be32
 926 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 927               union nfsd4_op_u *u)
 928 {
 929         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 930 }
 931
 932 static __be32
 933 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 934              union nfsd4_op_u *u)
 935 {
 936         return nfsd_lookup(rqstp, &cstate->current_fh,
 937                            u->lookup.lo_name, u->lookup.lo_len,
 938                            &cstate->current_fh);
 939 }
 940
 941 static __be32
 942 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 943            union nfsd4_op_u *u)
 944 {
 945         struct nfsd4_read *read = &u->read;
 946         __be32 status;
 947
 948         read->rd_nf = NULL;
 949
 950         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 951                               read->rd_offset, read->rd_length);
 952
 953         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 954         if (read->rd_offset > (u64)OFFSET_MAX)
 955                 read->rd_offset = (u64)OFFSET_MAX;
 956         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 957                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 958
 959         /*
 960          * If we do a zero copy read, then a client will see read data
 961          * that reflects the state of the file *after* performing the
 962          * following compound.
 963          *
 964          * To ensure proper ordering, we therefore turn off zero copy if
 965          * the client wants us to do more in this compound:
 966          */
 967         if (!nfsd4_last_compound_op(rqstp)) {
 968                 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
 969
 970                 argp->splice_ok = false;
 971         }
 972
 973         /* check stateid */
 974         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 975                                         &read->rd_stateid, RD_STATE,
 976                                         &read->rd_nf, NULL);
 977
 978         read->rd_rqstp = rqstp;
 979         read->rd_fhp = &cstate->current_fh;
 980         return status;
 981 }
 982
 983
 984 static void
 985 nfsd4_read_release(union nfsd4_op_u *u)
 986 {
 987         if (u->read.rd_nf)
 988                 nfsd_file_put(u->read.rd_nf);
 989         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 990                              u->read.rd_offset, u->read.rd_length);
 991 }
 992
 993 static __be32
 994 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 995               union nfsd4_op_u *u)
 996 {
 997         struct nfsd4_readdir *readdir = &u->readdir;
 998         u64 cookie = readdir->rd_cookie;
 999         static const nfs4_verifier zeroverf;
1000
1001         /* no need to check permission - this will be done in nfsd_readdir() */
1002
1003         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1004                 return nfserr_inval;
1005
1006         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1007         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1008         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1009
1010         if ((cookie == 1) || (cookie == 2) ||
1011             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1012                 return nfserr_bad_cookie;
1013
1014         readdir->rd_rqstp = rqstp;
1015         readdir->rd_fhp = &cstate->current_fh;
1016         return nfs_ok;
1017 }
1018
1019 static __be32
1020 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1021                union nfsd4_op_u *u)
1022 {
1023         u->readlink.rl_rqstp = rqstp;
1024         u->readlink.rl_fhp = &cstate->current_fh;
1025         return nfs_ok;
1026 }
1027
1028 static __be32
1029 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1030              union nfsd4_op_u *u)
1031 {
1032         struct nfsd4_remove *remove = &u->remove;
1033         __be32 status;
1034
1035         if (opens_in_grace(SVC_NET(rqstp)))
1036                 return nfserr_grace;
1037         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1038                              remove->rm_name, remove->rm_namelen);
1039         if (!status)
1040                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1041         return status;
1042 }
1043
1044 static __be32
1045 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1046              union nfsd4_op_u *u)
1047 {
1048         struct nfsd4_rename *rename = &u->rename;
1049         __be32 status;
1050
1051         if (opens_in_grace(SVC_NET(rqstp)))
1052                 return nfserr_grace;
1053         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1054                              rename->rn_snamelen, &cstate->current_fh,
1055                              rename->rn_tname, rename->rn_tnamelen);
1056         if (status)
1057                 return status;
1058         set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1059         set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1060         return nfs_ok;
1061 }
1062
1063 static __be32
1064 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1065               union nfsd4_op_u *u)
1066 {
1067         struct nfsd4_secinfo *secinfo = &u->secinfo;
1068         struct svc_export *exp;
1069         struct dentry *dentry;
1070         __be32 err;
1071
1072         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1073         if (err)
1074                 return err;
1075         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1076                                     secinfo->si_name, secinfo->si_namelen,
1077                                     &exp, &dentry);
1078         if (err)
1079                 return err;
1080         if (d_really_is_negative(dentry)) {
1081                 exp_put(exp);
1082                 err = nfserr_noent;
1083         } else
1084                 secinfo->si_exp = exp;
1085         dput(dentry);
1086         if (cstate->minorversion)
1087                 /* See rfc 5661 section 2.6.3.1.1.8 */
1088                 fh_put(&cstate->current_fh);
1089         return err;
1090 }
1091
1092 static __be32
1093 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1094                 union nfsd4_op_u *u)
1095 {
1096         __be32 err;
1097
1098         switch (u->secinfo_no_name.sin_style) {
1099         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1100                 break;
1101         case NFS4_SECINFO_STYLE4_PARENT:
1102                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1103                 if (err)
1104                         return err;
1105                 break;
1106         default:
1107                 return nfserr_inval;
1108         }
1109
1110         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1111         fh_put(&cstate->current_fh);
1112         return nfs_ok;
1113 }
1114
1115 static void
1116 nfsd4_secinfo_release(union nfsd4_op_u *u)
1117 {
1118         if (u->secinfo.si_exp)
1119                 exp_put(u->secinfo.si_exp);
1120 }
1121
1122 static void
1123 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1124 {
1125         if (u->secinfo_no_name.sin_exp)
1126                 exp_put(u->secinfo_no_name.sin_exp);
1127 }
1128
1129 static __be32
1130 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1131               union nfsd4_op_u *u)
1132 {
1133         struct nfsd4_setattr *setattr = &u->setattr;
1134         struct nfsd_attrs attrs = {
1135                 .na_iattr       = &setattr->sa_iattr,
1136                 .na_seclabel    = &setattr->sa_label,
1137         };
1138         struct inode *inode;
1139         __be32 status = nfs_ok;
1140         bool save_no_wcc;
1141         int err;
1142
1143         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1144                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1145                                 &cstate->current_fh, &setattr->sa_stateid,
1146                                 WR_STATE, NULL, NULL);
1147                 if (status)
1148                         return status;
1149         }
1150         err = fh_want_write(&cstate->current_fh);
1151         if (err)
1152                 return nfserrno(err);
1153         status = nfs_ok;
1154
1155         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1156                                     nfsd_attrmask);
1157         if (status)
1158                 goto out;
1159
1160         inode = cstate->current_fh.fh_dentry->d_inode;
1161         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1162                                    setattr->sa_acl, &attrs);
1163
1164         if (status)
1165                 goto out;
1166         save_no_wcc = cstate->current_fh.fh_no_wcc;
1167         cstate->current_fh.fh_no_wcc = true;
1168         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs, NULL);
1169         cstate->current_fh.fh_no_wcc = save_no_wcc;
1170         if (!status)
1171                 status = nfserrno(attrs.na_labelerr);
1172         if (!status)
1173                 status = nfserrno(attrs.na_aclerr);
1174 out:
1175         nfsd_attrs_free(&attrs);
1176         fh_drop_write(&cstate->current_fh);
1177         return status;
1178 }
1179
1180 static __be32
1181 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1182             union nfsd4_op_u *u)
1183 {
1184         struct nfsd4_write *write = &u->write;
1185         stateid_t *stateid = &write->wr_stateid;
1186         struct nfsd_file *nf = NULL;
1187         __be32 status = nfs_ok;
1188         unsigned long cnt;
1189         int nvecs;
1190
1191         if (write->wr_offset > (u64)OFFSET_MAX ||
1192             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1193                 return nfserr_fbig;
1194
1195         cnt = write->wr_buflen;
1196         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1197                                write->wr_offset, cnt);
1198         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1199                                                 stateid, WR_STATE, &nf, NULL);
1200         if (status)
1201                 return status;
1202
1203         write->wr_how_written = write->wr_stable_how;
1204
1205         nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1206         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1207
1208         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1209                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1210                                 write->wr_how_written,
1211                                 (__be32 *)write->wr_verifier.data);
1212         nfsd_file_put(nf);
1213
1214         write->wr_bytes_written = cnt;
1215         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1216                               write->wr_offset, cnt);
1217         return status;
1218 }
1219
1220 static __be32
1221 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1222                   stateid_t *src_stateid, struct nfsd_file **src,
1223                   stateid_t *dst_stateid, struct nfsd_file **dst)
1224 {
1225         __be32 status;
1226
1227         if (!cstate->save_fh.fh_dentry)
1228                 return nfserr_nofilehandle;
1229
1230         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1231                                             src_stateid, RD_STATE, src, NULL);
1232         if (status)
1233                 goto out;
1234
1235         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1236                                             dst_stateid, WR_STATE, dst, NULL);
1237         if (status)
1238                 goto out_put_src;
1239
1240         /* fix up for NFS-specific error code */
1241         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1242             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1243                 status = nfserr_wrong_type;
1244                 goto out_put_dst;
1245         }
1246
1247 out:
1248         return status;
1249 out_put_dst:
1250         nfsd_file_put(*dst);
1251         *dst = NULL;
1252 out_put_src:
1253         nfsd_file_put(*src);
1254         *src = NULL;
1255         goto out;
1256 }
1257
1258 static __be32
1259 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1260                 union nfsd4_op_u *u)
1261 {
1262         struct nfsd4_clone *clone = &u->clone;
1263         struct nfsd_file *src, *dst;
1264         __be32 status;
1265
1266         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1267                                    &clone->cl_dst_stateid, &dst);
1268         if (status)
1269                 goto out;
1270
1271         status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1272                         dst, clone->cl_dst_pos, clone->cl_count,
1273                         EX_ISSYNC(cstate->current_fh.fh_export));
1274
1275         nfsd_file_put(dst);
1276         nfsd_file_put(src);
1277 out:
1278         return status;
1279 }
1280
1281 /**
1282  * nfsd4_has_active_async_copies - Check for ongoing copy operations
1283  * @clp: Client to be checked
1284  *
1285  * NFSD maintains state for async COPY operations after they complete,
1286  * and this state remains in the nfs4_client's async_copies list.
1287  * Ongoing copies should block the destruction of the nfs4_client, but
1288  * completed copies should not.
1289  *
1290  * Return values:
1291  *   %true: At least one active async COPY is ongoing
1292  *   %false: No active async COPY operations were found
1293  */
1294 bool nfsd4_has_active_async_copies(struct nfs4_client *clp)
1295 {
1296         struct nfsd4_copy *copy;
1297         bool result = false;
1298
1299         spin_lock(&clp->async_lock);
1300         list_for_each_entry(copy, &clp->async_copies, copies) {
1301                 if (!test_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags) &&
1302                     !test_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags)) {
1303                         result = true;
1304                         break;
1305                 }
1306         }
1307         spin_unlock(&clp->async_lock);
1308         return result;
1309 }
1310
1311 /**
1312  * nfsd4_async_copy_reaper - Purge completed copies
1313  * @nn: Network namespace with possible active copy information
1314  */
1315 void nfsd4_async_copy_reaper(struct nfsd_net *nn)
1316 {
1317         struct nfs4_client *clp;
1318         struct nfsd4_copy *copy;
1319         LIST_HEAD(reaplist);
1320
1321         spin_lock(&nn->client_lock);
1322         list_for_each_entry(clp, &nn->client_lru, cl_lru) {
1323                 struct list_head *pos, *next;
1324
1325                 spin_lock(&clp->async_lock);
1326                 list_for_each_safe(pos, next, &clp->async_copies) {
1327                         copy = list_entry(pos, struct nfsd4_copy, copies);
1328                         if (test_bit(NFSD4_COPY_F_OFFLOAD_DONE, &copy->cp_flags)) {
1329                                 if (--copy->cp_ttl) {
1330                                         list_del_init(&copy->copies);
1331                                         list_add(&copy->copies, &reaplist);
1332                                 }
1333                         }
1334                 }
1335                 spin_unlock(&clp->async_lock);
1336         }
1337         spin_unlock(&nn->client_lock);
1338
1339         while (!list_empty(&reaplist)) {
1340                 copy = list_first_entry(&reaplist, struct nfsd4_copy, copies);
1341                 list_del_init(&copy->copies);
1342                 cleanup_async_copy(copy);
1343         }
1344 }
1345
1346 static void nfs4_put_copy(struct nfsd4_copy *copy)
1347 {
1348         if (!refcount_dec_and_test(&copy->refcount))
1349                 return;
1350         kfree(copy->cp_src);
1351         kfree(copy);
1352 }
1353
1354 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1355 {
1356         trace_nfsd_copy_async_cancel(copy);
1357         if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1358                 kthread_stop(copy->copy_task);
1359         nfs4_put_copy(copy);
1360 }
1361
1362 static struct nfsd4_copy *nfsd4_unhash_copy(struct nfs4_client *clp)
1363 {
1364         struct nfsd4_copy *copy = NULL;
1365
1366         spin_lock(&clp->async_lock);
1367         if (!list_empty(&clp->async_copies)) {
1368                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1369                                         copies);
1370                 refcount_inc(&copy->refcount);
1371                 copy->cp_clp = NULL;
1372                 if (!list_empty(&copy->copies))
1373                         list_del_init(&copy->copies);
1374         }
1375         spin_unlock(&clp->async_lock);
1376         return copy;
1377 }
1378
1379 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1380 {
1381         struct nfsd4_copy *copy;
1382
1383         while ((copy = nfsd4_unhash_copy(clp)) != NULL)
1384                 nfsd4_stop_copy(copy);
1385 }
1386 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1387
1388 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1389                                    struct nfs_fh *src_fh,
1390                                    nfs4_stateid *stateid);
1391 extern void nfs42_ssc_close(struct file *filep);
1392
1393 extern void nfs_sb_deactive(struct super_block *sb);
1394
1395 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1396
1397 /*
1398  * setup a work entry in the ssc delayed unmount list.
1399  */
1400 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1401                                   struct nfsd4_ssc_umount_item **nsui,
1402                                   struct svc_rqst *rqstp)
1403 {
1404         struct nfsd4_ssc_umount_item *ni = NULL;
1405         struct nfsd4_ssc_umount_item *work = NULL;
1406         struct nfsd4_ssc_umount_item *tmp;
1407         DEFINE_WAIT(wait);
1408         __be32 status = 0;
1409
1410         *nsui = NULL;
1411         work = kzalloc(sizeof(*work), GFP_KERNEL);
1412 try_again:
1413         spin_lock(&nn->nfsd_ssc_lock);
1414         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1415                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1416                         continue;
1417                 /* found a match */
1418                 if (ni->nsui_busy) {
1419                         /*  wait - and try again */
1420                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1421                         spin_unlock(&nn->nfsd_ssc_lock);
1422
1423                         /* allow 20secs for mount/unmount for now - revisit */
1424                         if (svc_thread_should_stop(rqstp) ||
1425                                         (schedule_timeout(20*HZ) == 0)) {
1426                                 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1427                                 kfree(work);
1428                                 return nfserr_eagain;
1429                         }
1430                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1431                         goto try_again;
1432                 }
1433                 *nsui = ni;
1434                 refcount_inc(&ni->nsui_refcnt);
1435                 spin_unlock(&nn->nfsd_ssc_lock);
1436                 kfree(work);
1437
1438                 /* return vfsmount in (*nsui)->nsui_vfsmount */
1439                 return 0;
1440         }
1441         if (work) {
1442                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1443                 refcount_set(&work->nsui_refcnt, 2);
1444                 work->nsui_busy = true;
1445                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1446                 *nsui = work;
1447         } else
1448                 status = nfserr_resource;
1449         spin_unlock(&nn->nfsd_ssc_lock);
1450         return status;
1451 }
1452
1453 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1454                                  struct nfsd4_ssc_umount_item *nsui,
1455                                  struct vfsmount *ss_mnt)
1456 {
1457         spin_lock(&nn->nfsd_ssc_lock);
1458         nsui->nsui_vfsmount = ss_mnt;
1459         nsui->nsui_busy = false;
1460         wake_up_all(&nn->nfsd_ssc_waitq);
1461         spin_unlock(&nn->nfsd_ssc_lock);
1462 }
1463
1464 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1465                                  struct nfsd4_ssc_umount_item *nsui)
1466 {
1467         spin_lock(&nn->nfsd_ssc_lock);
1468         list_del(&nsui->nsui_list);
1469         wake_up_all(&nn->nfsd_ssc_waitq);
1470         spin_unlock(&nn->nfsd_ssc_lock);
1471         kfree(nsui);
1472 }
1473
1474 /*
1475  * Support one copy source server for now.
1476  */
1477 static __be32
1478 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1479                        struct nfsd4_ssc_umount_item **nsui)
1480 {
1481         struct file_system_type *type;
1482         struct vfsmount *ss_mnt;
1483         struct nfs42_netaddr *naddr;
1484         struct sockaddr_storage tmp_addr;
1485         size_t tmp_addrlen, match_netid_len = 3;
1486         char *startsep = "", *endsep = "", *match_netid = "tcp";
1487         char *ipaddr, *dev_name, *raw_data;
1488         int len, raw_len;
1489         __be32 status = nfserr_inval;
1490         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1491
1492         naddr = &nss->u.nl4_addr;
1493         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1494                                          naddr->addr_len,
1495                                          (struct sockaddr *)&tmp_addr,
1496                                          sizeof(tmp_addr));
1497         *nsui = NULL;
1498         if (tmp_addrlen == 0)
1499                 goto out_err;
1500
1501         if (tmp_addr.ss_family == AF_INET6) {
1502                 startsep = "[";
1503                 endsep = "]";
1504                 match_netid = "tcp6";
1505                 match_netid_len = 4;
1506         }
1507
1508         if (naddr->netid_len != match_netid_len ||
1509                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1510                 goto out_err;
1511
1512         /* Construct the raw data for the vfs_kern_mount call */
1513         len = RPC_MAX_ADDRBUFLEN + 1;
1514         ipaddr = kzalloc(len, GFP_KERNEL);
1515         if (!ipaddr)
1516                 goto out_err;
1517
1518         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1519
1520         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1521
1522         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1523         raw_data = kzalloc(raw_len, GFP_KERNEL);
1524         if (!raw_data)
1525                 goto out_free_ipaddr;
1526
1527         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1528
1529         status = nfserr_nodev;
1530         type = get_fs_type("nfs");
1531         if (!type)
1532                 goto out_free_rawdata;
1533
1534         /* Set the server:<export> for the vfs_kern_mount call */
1535         dev_name = kzalloc(len + 5, GFP_KERNEL);
1536         if (!dev_name)
1537                 goto out_free_rawdata;
1538         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1539
1540         status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui, rqstp);
1541         if (status)
1542                 goto out_free_devname;
1543         if ((*nsui)->nsui_vfsmount)
1544                 goto out_done;
1545
1546         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1547         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1548         module_put(type->owner);
1549         if (IS_ERR(ss_mnt)) {
1550                 status = nfserr_nodev;
1551                 nfsd4_ssc_cancel_dul(nn, *nsui);
1552                 goto out_free_devname;
1553         }
1554         nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1555 out_done:
1556         status = 0;
1557
1558 out_free_devname:
1559         kfree(dev_name);
1560 out_free_rawdata:
1561         kfree(raw_data);
1562 out_free_ipaddr:
1563         kfree(ipaddr);
1564 out_err:
1565         return status;
1566 }
1567
1568 /*
1569  * Verify COPY destination stateid.
1570  *
1571  * Connect to the source server with NFSv4.1.
1572  * Create the source struct file for nfsd_copy_range.
1573  * Called with COPY cstate:
1574  *    SAVED_FH: source filehandle
1575  *    CURRENT_FH: destination filehandle
1576  */
1577 static __be32
1578 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1579                       struct nfsd4_compound_state *cstate,
1580                       struct nfsd4_copy *copy)
1581 {
1582         struct svc_fh *s_fh = NULL;
1583         stateid_t *s_stid = &copy->cp_src_stateid;
1584         __be32 status = nfserr_inval;
1585
1586         /* Verify the destination stateid and set dst struct file*/
1587         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1588                                             &copy->cp_dst_stateid,
1589                                             WR_STATE, &copy->nf_dst, NULL);
1590         if (status)
1591                 goto out;
1592
1593         status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1594         if (status)
1595                 goto out;
1596
1597         s_fh = &cstate->save_fh;
1598
1599         copy->c_fh.size = s_fh->fh_handle.fh_size;
1600         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1601         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1602         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1603                sizeof(stateid_opaque_t));
1604
1605         status = 0;
1606 out:
1607         return status;
1608 }
1609
1610 static void
1611 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1612                         struct nfsd_file *dst)
1613 {
1614         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1615         long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1616
1617         nfs42_ssc_close(filp);
1618         fput(filp);
1619
1620         spin_lock(&nn->nfsd_ssc_lock);
1621         list_del(&nsui->nsui_list);
1622         /*
1623          * vfsmount can be shared by multiple exports,
1624          * decrement refcnt. If the count drops to 1 it
1625          * will be unmounted when nsui_expire expires.
1626          */
1627         refcount_dec(&nsui->nsui_refcnt);
1628         nsui->nsui_expire = jiffies + timeout;
1629         list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1630         spin_unlock(&nn->nfsd_ssc_lock);
1631 }
1632
1633 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1634
1635 static __be32
1636 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1637                       struct nfsd4_compound_state *cstate,
1638                       struct nfsd4_copy *copy)
1639 {
1640         return nfserr_inval;
1641 }
1642
1643 static void
1644 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1645                         struct nfsd_file *dst)
1646 {
1647 }
1648
1649 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1650                                    struct nfs_fh *src_fh,
1651                                    nfs4_stateid *stateid)
1652 {
1653         return NULL;
1654 }
1655 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1656
1657 static __be32
1658 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1659                       struct nfsd4_compound_state *cstate,
1660                       struct nfsd4_copy *copy)
1661 {
1662         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1663                                  &copy->nf_src, &copy->cp_dst_stateid,
1664                                  &copy->nf_dst);
1665 }
1666
1667 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1668 {
1669         struct nfsd4_cb_offload *cbo =
1670                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1671         struct nfsd4_copy *copy =
1672                 container_of(cbo, struct nfsd4_copy, cp_cb_offload);
1673
1674         set_bit(NFSD4_COPY_F_OFFLOAD_DONE, &copy->cp_flags);
1675 }
1676
1677 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1678                                  struct rpc_task *task)
1679 {
1680         struct nfsd4_cb_offload *cbo =
1681                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1682
1683         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1684         switch (task->tk_status) {
1685         case -NFS4ERR_DELAY:
1686                 if (cbo->co_retries--) {
1687                         rpc_delay(task, 1 * HZ);
1688                         return 0;
1689                 }
1690         }
1691         return 1;
1692 }
1693
1694 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1695         .release = nfsd4_cb_offload_release,
1696         .done = nfsd4_cb_offload_done,
1697         .opcode = OP_CB_OFFLOAD,
1698 };
1699
1700 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1701 {
1702         copy->cp_res.wr_stable_how =
1703                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1704                         NFS_FILE_SYNC : NFS_UNSTABLE;
1705         nfsd4_copy_set_sync(copy, sync);
1706 }
1707
1708 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1709                                      struct file *dst,
1710                                      struct file *src)
1711 {
1712         errseq_t since;
1713         ssize_t bytes_copied = 0;
1714         u64 bytes_total = copy->cp_count;
1715         u64 src_pos = copy->cp_src_pos;
1716         u64 dst_pos = copy->cp_dst_pos;
1717         int status;
1718         loff_t end;
1719
1720         /* See RFC 7862 p.67: */
1721         if (bytes_total == 0)
1722                 bytes_total = ULLONG_MAX;
1723         do {
1724                 /* Only async copies can be stopped here */
1725                 if (kthread_should_stop())
1726                         break;
1727                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1728                                                     bytes_total);
1729                 if (bytes_copied <= 0)
1730                         break;
1731                 bytes_total -= bytes_copied;
1732                 copy->cp_res.wr_bytes_written += bytes_copied;
1733                 src_pos += bytes_copied;
1734                 dst_pos += bytes_copied;
1735         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1736         /* for a non-zero asynchronous copy do a commit of data */
1737         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1738                 since = READ_ONCE(dst->f_wb_err);
1739                 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1740                 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1741                 if (!status)
1742                         status = filemap_check_wb_err(dst->f_mapping, since);
1743                 if (!status)
1744                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1745         }
1746         return bytes_copied;
1747 }
1748
1749 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1750                             struct file *src, struct file *dst,
1751                             bool sync)
1752 {
1753         __be32 status;
1754         ssize_t bytes;
1755
1756         bytes = _nfsd_copy_file_range(copy, dst, src);
1757
1758         /* for async copy, we ignore the error, client can always retry
1759          * to get the error
1760          */
1761         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1762                 status = nfserrno(bytes);
1763         else {
1764                 nfsd4_init_copy_res(copy, sync);
1765                 status = nfs_ok;
1766         }
1767         return status;
1768 }
1769
1770 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1771 {
1772         dst->cp_src_pos = src->cp_src_pos;
1773         dst->cp_dst_pos = src->cp_dst_pos;
1774         dst->cp_count = src->cp_count;
1775         dst->cp_flags = src->cp_flags;
1776         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1777         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1778         dst->cp_clp = src->cp_clp;
1779         dst->nf_dst = nfsd_file_get(src->nf_dst);
1780         /* for inter, nf_src doesn't exist yet */
1781         if (!nfsd4_ssc_is_inter(src))
1782                 dst->nf_src = nfsd_file_get(src->nf_src);
1783
1784         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1785         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1786         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1787         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1788         dst->ss_nsui = src->ss_nsui;
1789 }
1790
1791 static void release_copy_files(struct nfsd4_copy *copy)
1792 {
1793         if (copy->nf_src)
1794                 nfsd_file_put(copy->nf_src);
1795         if (copy->nf_dst)
1796                 nfsd_file_put(copy->nf_dst);
1797 }
1798
1799 static void cleanup_async_copy(struct nfsd4_copy *copy)
1800 {
1801         nfs4_free_copy_state(copy);
1802         release_copy_files(copy);
1803         if (copy->cp_clp) {
1804                 spin_lock(&copy->cp_clp->async_lock);
1805                 if (!list_empty(&copy->copies))
1806                         list_del_init(&copy->copies);
1807                 spin_unlock(&copy->cp_clp->async_lock);
1808         }
1809         nfs4_put_copy(copy);
1810 }
1811
1812 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy)
1813 {
1814         struct nfsd4_cb_offload *cbo = &copy->cp_cb_offload;
1815
1816         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1817         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1818         cbo->co_nfserr = copy->nfserr;
1819         cbo->co_retries = 5;
1820
1821         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1822                       NFSPROC4_CLNT_CB_OFFLOAD);
1823         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1824                               &cbo->co_fh, copy->cp_count, copy->nfserr);
1825         nfsd4_run_cb(&cbo->co_cb);
1826 }
1827
1828 /**
1829  * nfsd4_do_async_copy - kthread function for background server-side COPY
1830  * @data: arguments for COPY operation
1831  *
1832  * Return values:
1833  *   %0: Copy operation is done.
1834  */
1835 static int nfsd4_do_async_copy(void *data)
1836 {
1837         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1838
1839         trace_nfsd_copy_async(copy);
1840         if (nfsd4_ssc_is_inter(copy)) {
1841                 struct file *filp;
1842
1843                 filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1844                                       &copy->c_fh, &copy->stateid);
1845                 if (IS_ERR(filp)) {
1846                         switch (PTR_ERR(filp)) {
1847                         case -EBADF:
1848                                 copy->nfserr = nfserr_wrong_type;
1849                                 break;
1850                         default:
1851                                 copy->nfserr = nfserr_offload_denied;
1852                         }
1853                         /* ss_mnt will be unmounted by the laundromat */
1854                         goto do_callback;
1855                 }
1856                 copy->nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1857                                              false);
1858                 nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1859         } else {
1860                 copy->nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1861                                              copy->nf_dst->nf_file, false);
1862         }
1863
1864 do_callback:
1865         /* The kthread exits forthwith. Ensure that a subsequent
1866          * OFFLOAD_CANCEL won't try to kill it again. */
1867         set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags);
1868
1869         set_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags);
1870         trace_nfsd_copy_async_done(copy);
1871         nfsd4_send_cb_offload(copy);
1872         atomic_dec(&copy->cp_nn->pending_async_copies);
1873         return 0;
1874 }
1875
1876 static __be32
1877 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1878                 union nfsd4_op_u *u)
1879 {
1880         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1881         struct nfsd4_copy *async_copy = NULL;
1882         struct nfsd4_copy *copy = &u->copy;
1883         struct nfsd42_write_res *result;
1884         __be32 status;
1885
1886         /*
1887          * Currently, async COPY is not reliable. Force all COPY
1888          * requests to be synchronous to avoid client application
1889          * hangs waiting for COPY completion.
1890          */
1891         nfsd4_copy_set_sync(copy, true);
1892
1893         result = &copy->cp_res;
1894         nfsd_copy_write_verifier((__be32 *)&result->wr_verifier.data, nn);
1895
1896         copy->cp_clp = cstate->clp;
1897         if (nfsd4_ssc_is_inter(copy)) {
1898                 trace_nfsd_copy_inter(copy);
1899                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1900                         status = nfserr_notsupp;
1901                         goto out;
1902                 }
1903                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1904                 if (status) {
1905                         trace_nfsd_copy_done(copy, status);
1906                         return nfserr_offload_denied;
1907                 }
1908         } else {
1909                 trace_nfsd_copy_intra(copy);
1910                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1911                 if (status) {
1912                         trace_nfsd_copy_done(copy, status);
1913                         return status;
1914                 }
1915         }
1916
1917         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1918                 sizeof(struct knfsd_fh));
1919         if (nfsd4_copy_is_async(copy)) {
1920                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1921                 if (!async_copy)
1922                         goto out_err;
1923                 async_copy->cp_nn = nn;
1924                 INIT_LIST_HEAD(&async_copy->copies);
1925                 refcount_set(&async_copy->refcount, 1);
1926                 async_copy->cp_ttl = NFSD_COPY_INITIAL_TTL;
1927                 /* Arbitrary cap on number of pending async copy operations */
1928                 if (atomic_inc_return(&nn->pending_async_copies) >
1929                                 (int)rqstp->rq_pool->sp_nrthreads)
1930                         goto out_dec_async_copy_err;
1931                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1932                 if (!async_copy->cp_src)
1933                         goto out_dec_async_copy_err;
1934                 if (!nfs4_init_copy_state(nn, copy))
1935                         goto out_dec_async_copy_err;
1936                 memcpy(&result->cb_stateid, &copy->cp_stateid.cs_stid,
1937                         sizeof(result->cb_stateid));
1938                 dup_copy_fields(copy, async_copy);
1939                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1940                                 async_copy, "%s", "copy thread");
1941                 if (IS_ERR(async_copy->copy_task))
1942                         goto out_dec_async_copy_err;
1943                 spin_lock(&async_copy->cp_clp->async_lock);
1944                 list_add(&async_copy->copies,
1945                                 &async_copy->cp_clp->async_copies);
1946                 spin_unlock(&async_copy->cp_clp->async_lock);
1947                 wake_up_process(async_copy->copy_task);
1948                 status = nfs_ok;
1949         } else {
1950                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1951                                        copy->nf_dst->nf_file, true);
1952         }
1953 out:
1954         trace_nfsd_copy_done(copy, status);
1955         release_copy_files(copy);
1956         return status;
1957 out_dec_async_copy_err:
1958         if (async_copy)
1959                 atomic_dec(&nn->pending_async_copies);
1960 out_err:
1961         if (nfsd4_ssc_is_inter(copy)) {
1962                 /*
1963                  * Source's vfsmount of inter-copy will be unmounted
1964                  * by the laundromat. Use copy instead of async_copy
1965                  * since async_copy->ss_nsui might not be set yet.
1966                  */
1967                 refcount_dec(&copy->ss_nsui->nsui_refcnt);
1968         }
1969         if (async_copy)
1970                 cleanup_async_copy(async_copy);
1971         status = nfserr_jukebox;
1972         goto out;
1973 }
1974
1975 static struct nfsd4_copy *
1976 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1977 {
1978         struct nfsd4_copy *copy;
1979
1980         lockdep_assert_held(&clp->async_lock);
1981
1982         list_for_each_entry(copy, &clp->async_copies, copies) {
1983                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1984                         continue;
1985                 return copy;
1986         }
1987         return NULL;
1988 }
1989
1990 static struct nfsd4_copy *
1991 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1992 {
1993         struct nfsd4_copy *copy;
1994
1995         spin_lock(&clp->async_lock);
1996         copy = find_async_copy_locked(clp, stateid);
1997         if (copy)
1998                 refcount_inc(&copy->refcount);
1999         spin_unlock(&clp->async_lock);
2000         return copy;
2001 }
2002
2003 static __be32
2004 nfsd4_offload_cancel(struct svc_rqst *rqstp,
2005                      struct nfsd4_compound_state *cstate,
2006                      union nfsd4_op_u *u)
2007 {
2008         struct nfsd4_offload_status *os = &u->offload_status;
2009         struct nfsd4_copy *copy;
2010         struct nfs4_client *clp = cstate->clp;
2011
2012         copy = find_async_copy(clp, &os->stateid);
2013         if (!copy) {
2014                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2015
2016                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
2017         } else
2018                 nfsd4_stop_copy(copy);
2019
2020         return nfs_ok;
2021 }
2022
2023 static __be32
2024 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2025                   union nfsd4_op_u *u)
2026 {
2027         struct nfsd4_copy_notify *cn = &u->copy_notify;
2028         __be32 status;
2029         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2030         struct nfs4_stid *stid = NULL;
2031         struct nfs4_cpntf_state *cps;
2032         struct nfs4_client *clp = cstate->clp;
2033
2034         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2035                                         &cn->cpn_src_stateid, RD_STATE, NULL,
2036                                         &stid);
2037         if (status)
2038                 return status;
2039         if (!stid)
2040                 return nfserr_bad_stateid;
2041
2042         cn->cpn_lease_time.tv_sec = nn->nfsd4_lease;
2043         cn->cpn_lease_time.tv_nsec = 0;
2044
2045         status = nfserrno(-ENOMEM);
2046         cps = nfs4_alloc_init_cpntf_state(nn, stid);
2047         if (!cps)
2048                 goto out;
2049         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
2050         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
2051         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
2052
2053         /* For now, only return one server address in cpn_src, the
2054          * address used by the client to connect to this server.
2055          */
2056         cn->cpn_src->nl4_type = NL4_NETADDR;
2057         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
2058                                  &cn->cpn_src->u.nl4_addr);
2059         WARN_ON_ONCE(status);
2060         if (status) {
2061                 nfs4_put_cpntf_state(nn, cps);
2062                 goto out;
2063         }
2064 out:
2065         nfs4_put_stid(stid);
2066         return status;
2067 }
2068
2069 static __be32
2070 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2071                 struct nfsd4_fallocate *fallocate, int flags)
2072 {
2073         __be32 status;
2074         struct nfsd_file *nf;
2075
2076         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2077                                             &fallocate->falloc_stateid,
2078                                             WR_STATE, &nf, NULL);
2079         if (status != nfs_ok)
2080                 return status;
2081
2082         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
2083                                      fallocate->falloc_offset,
2084                                      fallocate->falloc_length,
2085                                      flags);
2086         nfsd_file_put(nf);
2087         return status;
2088 }
2089
2090 static __be32
2091 nfsd4_offload_status(struct svc_rqst *rqstp,
2092                      struct nfsd4_compound_state *cstate,
2093                      union nfsd4_op_u *u)
2094 {
2095         struct nfsd4_offload_status *os = &u->offload_status;
2096         __be32 status = nfs_ok;
2097         struct nfsd4_copy *copy;
2098         struct nfs4_client *clp = cstate->clp;
2099
2100         os->completed = false;
2101         spin_lock(&clp->async_lock);
2102         copy = find_async_copy_locked(clp, &os->stateid);
2103         if (copy) {
2104                 os->count = copy->cp_res.wr_bytes_written;
2105                 if (test_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags)) {
2106                         os->completed = true;
2107                         os->status = copy->nfserr;
2108                 }
2109         } else
2110                 status = nfserr_bad_stateid;
2111         spin_unlock(&clp->async_lock);
2112
2113         return status;
2114 }
2115
2116 static __be32
2117 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2118                union nfsd4_op_u *u)
2119 {
2120         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2121 }
2122
2123 static __be32
2124 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2125                  union nfsd4_op_u *u)
2126 {
2127         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2128                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2129 }
2130
2131 static __be32
2132 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2133            union nfsd4_op_u *u)
2134 {
2135         struct nfsd4_seek *seek = &u->seek;
2136         int whence;
2137         __be32 status;
2138         struct nfsd_file *nf;
2139
2140         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2141                                             &seek->seek_stateid,
2142                                             RD_STATE, &nf, NULL);
2143         if (status)
2144                 return status;
2145
2146         switch (seek->seek_whence) {
2147         case NFS4_CONTENT_DATA:
2148                 whence = SEEK_DATA;
2149                 break;
2150         case NFS4_CONTENT_HOLE:
2151                 whence = SEEK_HOLE;
2152                 break;
2153         default:
2154                 status = nfserr_union_notsupp;
2155                 goto out;
2156         }
2157
2158         /*
2159          * Note:  This call does change file->f_pos, but nothing in NFSD
2160          *        should ever file->f_pos.
2161          */
2162         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2163         if (seek->seek_pos < 0)
2164                 status = nfserrno(seek->seek_pos);
2165         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2166                 seek->seek_eof = true;
2167
2168 out:
2169         nfsd_file_put(nf);
2170         return status;
2171 }
2172
2173 /* This routine never returns NFS_OK!  If there are no other errors, it
2174  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2175  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2176  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2177  */
2178 static __be32
2179 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2180              struct nfsd4_verify *verify)
2181 {
2182         __be32 *buf, *p;
2183         int count;
2184         __be32 status;
2185
2186         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2187         if (status)
2188                 return status;
2189
2190         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2191         if (status)
2192                 return status;
2193
2194         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2195             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2196                 return nfserr_inval;
2197         if (verify->ve_attrlen & 3)
2198                 return nfserr_inval;
2199
2200         /* count in words:
2201          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2202          */
2203         count = 4 + (verify->ve_attrlen >> 2);
2204         buf = kmalloc(count << 2, GFP_KERNEL);
2205         if (!buf)
2206                 return nfserr_jukebox;
2207
2208         p = buf;
2209         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2210                                     cstate->current_fh.fh_export,
2211                                     cstate->current_fh.fh_dentry,
2212                                     verify->ve_bmval,
2213                                     rqstp, 0);
2214         /*
2215          * If nfsd4_encode_fattr() ran out of space, assume that's because
2216          * the attributes are longer (hence different) than those given:
2217          */
2218         if (status == nfserr_resource)
2219                 status = nfserr_not_same;
2220         if (status)
2221                 goto out_kfree;
2222
2223         /* skip bitmap */
2224         p = buf + 1 + ntohl(buf[0]);
2225         status = nfserr_not_same;
2226         if (ntohl(*p++) != verify->ve_attrlen)
2227                 goto out_kfree;
2228         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2229                 status = nfserr_same;
2230
2231 out_kfree:
2232         kfree(buf);
2233         return status;
2234 }
2235
2236 static __be32
2237 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2238               union nfsd4_op_u *u)
2239 {
2240         __be32 status;
2241
2242         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2243         return status == nfserr_not_same ? nfs_ok : status;
2244 }
2245
2246 static __be32
2247 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2248              union nfsd4_op_u *u)
2249 {
2250         __be32 status;
2251
2252         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2253         return status == nfserr_same ? nfs_ok : status;
2254 }
2255
2256 static __be32
2257 nfsd4_get_dir_delegation(struct svc_rqst *rqstp,
2258                          struct nfsd4_compound_state *cstate,
2259                          union nfsd4_op_u *u)
2260 {
2261         struct nfsd4_get_dir_delegation *gdd = &u->get_dir_delegation;
2262
2263         /*
2264          * RFC 8881, section 18.39.3 says:
2265          *
2266          * "The server may refuse to grant the delegation. In that case, the
2267          *  server will return NFS4ERR_DIRDELEG_UNAVAIL."
2268          *
2269          * This is sub-optimal, since it means that the server would need to
2270          * abort compound processing just because the delegation wasn't
2271          * available. RFC8881bis should change this to allow the server to
2272          * return NFS4_OK with a non-fatal status of GDD4_UNAVAIL in this
2273          * situation.
2274          */
2275         gdd->gddrnf_status = GDD4_UNAVAIL;
2276         return nfs_ok;
2277 }
2278
2279 #ifdef CONFIG_NFSD_PNFS
2280 static const struct nfsd4_layout_ops *
2281 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2282 {
2283         if (!exp->ex_layout_types) {
2284                 dprintk("%s: export does not support pNFS\n", __func__);
2285                 return NULL;
2286         }
2287
2288         if (layout_type >= LAYOUT_TYPE_MAX ||
2289             !(exp->ex_layout_types & (1 << layout_type))) {
2290                 dprintk("%s: layout type %d not supported\n",
2291                         __func__, layout_type);
2292                 return NULL;
2293         }
2294
2295         return nfsd4_layout_ops[layout_type];
2296 }
2297
2298 static __be32
2299 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2300                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2301 {
2302         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2303         const struct nfsd4_layout_ops *ops;
2304         struct nfsd4_deviceid_map *map;
2305         struct svc_export *exp;
2306         __be32 nfserr;
2307
2308         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2309                __func__,
2310                gdp->gd_layout_type,
2311                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2312                gdp->gd_maxcount);
2313
2314         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2315         if (!map) {
2316                 dprintk("%s: couldn't find device ID to export mapping!\n",
2317                         __func__);
2318                 return nfserr_noent;
2319         }
2320
2321         exp = rqst_exp_find(&rqstp->rq_chandle, SVC_NET(rqstp),
2322                             rqstp->rq_client, rqstp->rq_gssclient,
2323                             map->fsid_type, map->fsid);
2324         if (IS_ERR(exp)) {
2325                 dprintk("%s: could not find device id\n", __func__);
2326                 return nfserr_noent;
2327         }
2328
2329         nfserr = nfserr_layoutunavailable;
2330         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2331         if (!ops)
2332                 goto out;
2333
2334         nfserr = nfs_ok;
2335         if (gdp->gd_maxcount != 0) {
2336                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2337                                 rqstp, cstate->clp, gdp);
2338         }
2339
2340         gdp->gd_notify_types &= ops->notify_types;
2341 out:
2342         exp_put(exp);
2343         return nfserr;
2344 }
2345
2346 static void
2347 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2348 {
2349         kfree(u->getdeviceinfo.gd_device);
2350 }
2351
2352 static __be32
2353 nfsd4_layoutget(struct svc_rqst *rqstp,
2354                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2355 {
2356         struct nfsd4_layoutget *lgp = &u->layoutget;
2357         struct svc_fh *current_fh = &cstate->current_fh;
2358         const struct nfsd4_layout_ops *ops;
2359         struct nfs4_layout_stateid *ls;
2360         __be32 nfserr;
2361         int accmode = NFSD_MAY_READ_IF_EXEC | NFSD_MAY_OWNER_OVERRIDE;
2362
2363         switch (lgp->lg_seg.iomode) {
2364         case IOMODE_READ:
2365                 accmode |= NFSD_MAY_READ;
2366                 break;
2367         case IOMODE_RW:
2368                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2369                 break;
2370         default:
2371                 dprintk("%s: invalid iomode %d\n",
2372                         __func__, lgp->lg_seg.iomode);
2373                 nfserr = nfserr_badiomode;
2374                 goto out;
2375         }
2376
2377         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2378         if (nfserr)
2379                 goto out;
2380
2381         nfserr = nfserr_layoutunavailable;
2382         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2383         if (!ops)
2384                 goto out;
2385
2386         /*
2387          * Verify minlength and range as per RFC5661:
2388          *  o  If loga_length is less than loga_minlength,
2389          *     the metadata server MUST return NFS4ERR_INVAL.
2390          *  o  If the sum of loga_offset and loga_minlength exceeds
2391          *     NFS4_UINT64_MAX, and loga_minlength is not
2392          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2393          *  o  If the sum of loga_offset and loga_length exceeds
2394          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2395          *     the error NFS4ERR_INVAL MUST result.
2396          */
2397         nfserr = nfserr_inval;
2398         if (lgp->lg_seg.length < lgp->lg_minlength ||
2399             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2400              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2401             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2402              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2403                 goto out;
2404         if (lgp->lg_seg.length == 0)
2405                 goto out;
2406
2407         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2408                                                 true, lgp->lg_layout_type, &ls);
2409         if (nfserr) {
2410                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2411                 goto out;
2412         }
2413
2414         nfserr = nfserr_recallconflict;
2415         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2416                 goto out_put_stid;
2417
2418         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2419                                      current_fh, lgp);
2420         if (nfserr)
2421                 goto out_put_stid;
2422
2423         nfserr = nfsd4_insert_layout(lgp, ls);
2424
2425 out_put_stid:
2426         mutex_unlock(&ls->ls_mutex);
2427         nfs4_put_stid(&ls->ls_stid);
2428 out:
2429         return nfserr;
2430 }
2431
2432 static void
2433 nfsd4_layoutget_release(union nfsd4_op_u *u)
2434 {
2435         kfree(u->layoutget.lg_content);
2436 }
2437
2438 static __be32
2439 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2440                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2441 {
2442         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2443         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2444         struct svc_fh *current_fh = &cstate->current_fh;
2445         const struct nfsd4_layout_ops *ops;
2446         loff_t new_size = lcp->lc_last_wr + 1;
2447         struct inode *inode;
2448         struct nfs4_layout_stateid *ls;
2449         __be32 nfserr;
2450
2451         nfserr = fh_verify(rqstp, current_fh, 0,
2452                            NFSD_MAY_WRITE | NFSD_MAY_OWNER_OVERRIDE);
2453         if (nfserr)
2454                 goto out;
2455
2456         nfserr = nfserr_layoutunavailable;
2457         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2458         if (!ops)
2459                 goto out;
2460         inode = d_inode(current_fh->fh_dentry);
2461
2462         nfserr = nfserr_inval;
2463         if (new_size <= seg->offset) {
2464                 dprintk("pnfsd: last write before layout segment\n");
2465                 goto out;
2466         }
2467         if (new_size > seg->offset + seg->length) {
2468                 dprintk("pnfsd: last write beyond layout segment\n");
2469                 goto out;
2470         }
2471         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2472                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2473                 goto out;
2474         }
2475
2476         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2477                                                 false, lcp->lc_layout_type,
2478                                                 &ls);
2479         if (nfserr) {
2480                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2481                 /* fixup error code as per RFC5661 */
2482                 if (nfserr == nfserr_bad_stateid)
2483                         nfserr = nfserr_badlayout;
2484                 goto out;
2485         }
2486
2487         /* LAYOUTCOMMIT does not require any serialization */
2488         mutex_unlock(&ls->ls_mutex);
2489
2490         if (new_size > i_size_read(inode)) {
2491                 lcp->lc_size_chg = true;
2492                 lcp->lc_newsize = new_size;
2493         } else {
2494                 lcp->lc_size_chg = false;
2495         }
2496
2497         nfserr = ops->proc_layoutcommit(inode, lcp);
2498         nfs4_put_stid(&ls->ls_stid);
2499 out:
2500         return nfserr;
2501 }
2502
2503 static __be32
2504 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2505                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2506 {
2507         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2508         struct svc_fh *current_fh = &cstate->current_fh;
2509         __be32 nfserr;
2510
2511         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2512         if (nfserr)
2513                 goto out;
2514
2515         nfserr = nfserr_layoutunavailable;
2516         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2517                 goto out;
2518
2519         switch (lrp->lr_seg.iomode) {
2520         case IOMODE_READ:
2521         case IOMODE_RW:
2522         case IOMODE_ANY:
2523                 break;
2524         default:
2525                 dprintk("%s: invalid iomode %d\n", __func__,
2526                         lrp->lr_seg.iomode);
2527                 nfserr = nfserr_inval;
2528                 goto out;
2529         }
2530
2531         switch (lrp->lr_return_type) {
2532         case RETURN_FILE:
2533                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2534                 break;
2535         case RETURN_FSID:
2536         case RETURN_ALL:
2537                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2538                 break;
2539         default:
2540                 dprintk("%s: invalid return_type %d\n", __func__,
2541                         lrp->lr_return_type);
2542                 nfserr = nfserr_inval;
2543                 break;
2544         }
2545 out:
2546         return nfserr;
2547 }
2548 #endif /* CONFIG_NFSD_PNFS */
2549
2550 static __be32
2551 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2552                union nfsd4_op_u *u)
2553 {
2554         struct nfsd4_getxattr *getxattr = &u->getxattr;
2555
2556         return nfsd_getxattr(rqstp, &cstate->current_fh,
2557                              getxattr->getxa_name, &getxattr->getxa_buf,
2558                              &getxattr->getxa_len);
2559 }
2560
2561 static __be32
2562 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2563            union nfsd4_op_u *u)
2564 {
2565         struct nfsd4_setxattr *setxattr = &u->setxattr;
2566         __be32 ret;
2567
2568         if (opens_in_grace(SVC_NET(rqstp)))
2569                 return nfserr_grace;
2570
2571         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2572                             setxattr->setxa_buf, setxattr->setxa_len,
2573                             setxattr->setxa_flags);
2574
2575         if (!ret)
2576                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2577
2578         return ret;
2579 }
2580
2581 static __be32
2582 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2583            union nfsd4_op_u *u)
2584 {
2585         /*
2586          * Get the entire list, then copy out only the user attributes
2587          * in the encode function.
2588          */
2589         return nfsd_listxattr(rqstp, &cstate->current_fh,
2590                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2591 }
2592
2593 static __be32
2594 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2595            union nfsd4_op_u *u)
2596 {
2597         struct nfsd4_removexattr *removexattr = &u->removexattr;
2598         __be32 ret;
2599
2600         if (opens_in_grace(SVC_NET(rqstp)))
2601                 return nfserr_grace;
2602
2603         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2604             removexattr->rmxa_name);
2605
2606         if (!ret)
2607                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2608
2609         return ret;
2610 }
2611
2612 /*
2613  * NULL call.
2614  */
2615 static __be32
2616 nfsd4_proc_null(struct svc_rqst *rqstp)
2617 {
2618         return rpc_success;
2619 }
2620
2621 static inline void nfsd4_increment_op_stats(struct nfsd_net *nn, u32 opnum)
2622 {
2623         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2624                 percpu_counter_inc(&nn->counter[NFSD_STATS_NFS4_OP(opnum)]);
2625 }
2626
2627 static const struct nfsd4_operation nfsd4_ops[];
2628
2629 static const char *nfsd4_op_name(unsigned opnum);
2630
2631 /*
2632  * Enforce NFSv4.1 COMPOUND ordering rules:
2633  *
2634  * Also note, enforced elsewhere:
2635  *      - SEQUENCE other than as first op results in
2636  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2637  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2638  *        (Enforced in nfsd4_bind_conn_to_session().)
2639  *      - DESTROY_SESSION must be the final operation in a compound, if
2640  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2641  *        (Enforced in nfsd4_destroy_session().)
2642  */
2643 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2644 {
2645         struct nfsd4_op *first_op = &args->ops[0];
2646
2647         /* These ordering requirements don't apply to NFSv4.0: */
2648         if (args->minorversion == 0)
2649                 return nfs_ok;
2650         /* This is weird, but OK, not our problem: */
2651         if (args->opcnt == 0)
2652                 return nfs_ok;
2653         if (first_op->status == nfserr_op_illegal)
2654                 return nfs_ok;
2655         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2656                 return nfserr_op_not_in_session;
2657         if (first_op->opnum == OP_SEQUENCE)
2658                 return nfs_ok;
2659         /*
2660          * So first_op is something allowed outside a session, like
2661          * EXCHANGE_ID; but then it has to be the only op in the
2662          * compound:
2663          */
2664         if (args->opcnt != 1)
2665                 return nfserr_not_only_op;
2666         return nfs_ok;
2667 }
2668
2669 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2670 {
2671         return &nfsd4_ops[op->opnum];
2672 }
2673
2674 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2675 {
2676         if (op->opnum == OP_ILLEGAL)
2677                 return false;
2678         return OPDESC(op)->op_flags & OP_CACHEME;
2679 }
2680
2681 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2682 {
2683         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2684         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2685         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2686         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2687         const struct nfsd4_operation *thisd = OPDESC(this);
2688         const struct nfsd4_operation *nextd;
2689
2690         /*
2691          * Most ops check wronsec on our own; only the putfh-like ops
2692          * have special rules.
2693          */
2694         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2695                 return false;
2696         /*
2697          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2698          * put-filehandle operation if we're not going to use the
2699          * result:
2700          */
2701         if (argp->opcnt == resp->opcnt)
2702                 return false;
2703         if (next->opnum == OP_ILLEGAL)
2704                 return false;
2705         nextd = OPDESC(next);
2706         /*
2707          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2708          * errors themselves as necessary; others should check for them
2709          * now:
2710          */
2711         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2712 }
2713
2714 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2715 static void
2716 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2717 {
2718         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2719         struct nfsd4_copy *copy;
2720         struct nfsd4_putfh *putfh;
2721         int i;
2722
2723         /* traverse all operation and if it's a COPY compound, mark the
2724          * source filehandle to skip verification
2725          */
2726         for (i = 0; i < args->opcnt; i++) {
2727                 op = &args->ops[i];
2728                 if (op->opnum == OP_PUTFH)
2729                         current_op = op;
2730                 else if (op->opnum == OP_SAVEFH)
2731                         saved_op = current_op;
2732                 else if (op->opnum == OP_RESTOREFH)
2733                         current_op = saved_op;
2734                 else if (op->opnum == OP_COPY) {
2735                         copy = (struct nfsd4_copy *)&op->u;
2736                         if (!saved_op) {
2737                                 op->status = nfserr_nofilehandle;
2738                                 return;
2739                         }
2740                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2741                         if (nfsd4_ssc_is_inter(copy))
2742                                 putfh->no_verify = true;
2743                 }
2744         }
2745 }
2746 #else
2747 static void
2748 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2749 {
2750 }
2751 #endif
2752
2753 /*
2754  * COMPOUND call.
2755  */
2756 static __be32
2757 nfsd4_proc_compound(struct svc_rqst *rqstp)
2758 {
2759         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2760         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2761         struct nfsd4_op *op;
2762         struct nfsd4_compound_state *cstate = &resp->cstate;
2763         struct svc_fh *current_fh = &cstate->current_fh;
2764         struct svc_fh *save_fh = &cstate->save_fh;
2765         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2766         __be32          status;
2767
2768         resp->xdr = &rqstp->rq_res_stream;
2769         resp->statusp = resp->xdr->p;
2770
2771         /* reserve space for: NFS status code */
2772         xdr_reserve_space(resp->xdr, XDR_UNIT);
2773
2774         /* reserve space for: taglen, tag, and opcnt */
2775         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2776         resp->taglen = args->taglen;
2777         resp->tag = args->tag;
2778         resp->rqstp = rqstp;
2779         cstate->minorversion = args->minorversion;
2780         fh_init(current_fh, NFS4_FHSIZE);
2781         fh_init(save_fh, NFS4_FHSIZE);
2782         /*
2783          * Don't use the deferral mechanism for NFSv4; compounds make it
2784          * too hard to avoid non-idempotency problems.
2785          */
2786         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2787
2788         /*
2789          * According to RFC3010, this takes precedence over all other errors.
2790          */
2791         status = nfserr_minor_vers_mismatch;
2792         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2793                 goto out;
2794
2795         status = nfs41_check_op_ordering(args);
2796         if (status) {
2797                 op = &args->ops[0];
2798                 op->status = status;
2799                 resp->opcnt = 1;
2800                 goto encode_op;
2801         }
2802         check_if_stalefh_allowed(args);
2803
2804         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2805
2806         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2807         while (!status && resp->opcnt < args->opcnt) {
2808                 op = &args->ops[resp->opcnt++];
2809
2810                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2811                         /* If there are still more operations to process,
2812                          * stop here and report NFS4ERR_RESOURCE. */
2813                         if (cstate->minorversion == 0 &&
2814                             args->client_opcnt > resp->opcnt) {
2815                                 op->status = nfserr_resource;
2816                                 goto encode_op;
2817                         }
2818                 }
2819
2820                 /*
2821                  * The XDR decode routines may have pre-set op->status;
2822                  * for example, if there is a miscellaneous XDR error
2823                  * it will be set to nfserr_bad_xdr.
2824                  */
2825                 if (op->status) {
2826                         if (op->opnum == OP_OPEN)
2827                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2828                         goto encode_op;
2829                 }
2830                 if (!current_fh->fh_dentry &&
2831                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2832                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2833                                 op->status = nfserr_nofilehandle;
2834                                 goto encode_op;
2835                         }
2836                 } else if (current_fh->fh_export &&
2837                            current_fh->fh_export->ex_fslocs.migrated &&
2838                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2839                         op->status = nfserr_moved;
2840                         goto encode_op;
2841                 }
2842
2843                 fh_clear_pre_post_attrs(current_fh);
2844
2845                 /* If op is non-idempotent */
2846                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2847                         /*
2848                          * Don't execute this op if we couldn't encode a
2849                          * successful reply:
2850                          */
2851                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2852                         /*
2853                          * Plus if there's another operation, make sure
2854                          * we'll have space to at least encode an error:
2855                          */
2856                         if (resp->opcnt < args->opcnt)
2857                                 plen += COMPOUND_ERR_SLACK_SPACE;
2858                         op->status = nfsd4_check_resp_size(resp, plen);
2859                 }
2860
2861                 if (op->status)
2862                         goto encode_op;
2863
2864                 if (op->opdesc->op_get_currentstateid)
2865                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2866                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2867                 trace_nfsd_compound_op_err(rqstp, op->opnum, op->status);
2868
2869                 /* Only from SEQUENCE */
2870                 if (cstate->status == nfserr_replay_cache) {
2871                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2872                         status = op->status;
2873                         goto out;
2874                 }
2875                 if (!op->status) {
2876                         if (op->opdesc->op_set_currentstateid)
2877                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2878
2879                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2880                                 clear_current_stateid(cstate);
2881
2882                         if (current_fh->fh_export &&
2883                                         need_wrongsec_check(rqstp))
2884                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp, false);
2885                 }
2886 encode_op:
2887                 if (op->status == nfserr_replay_me) {
2888                         op->replay = &cstate->replay_owner->so_replay;
2889                         nfsd4_encode_replay(resp->xdr, op);
2890                         status = op->status = op->replay->rp_status;
2891                 } else {
2892                         nfsd4_encode_operation(resp, op);
2893                         status = op->status;
2894                 }
2895
2896                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2897                                            status, nfsd4_op_name(op->opnum));
2898
2899                 nfsd4_cstate_clear_replay(cstate);
2900                 nfsd4_increment_op_stats(nn, op->opnum);
2901         }
2902
2903         fh_put(current_fh);
2904         fh_put(save_fh);
2905         BUG_ON(cstate->replay_owner);
2906 out:
2907         cstate->status = status;
2908         /* Reset deferral mechanism for RPC deferrals */
2909         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2910         return rpc_success;
2911 }
2912
2913 #define op_encode_hdr_size              (2)
2914 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2915 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2916 #define op_encode_change_info_maxsz     (5)
2917 #define nfs4_fattr_bitmap_maxsz         (4)
2918
2919 /* We'll fall back on returning no lockowner if run out of space: */
2920 #define op_encode_lockowner_maxsz       (0)
2921 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2922
2923 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2924
2925 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2926 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2927                                          op_encode_ace_maxsz)
2928
2929 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2930
2931 /*
2932  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2933  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2934  * the maximum payload size here, based on transport limits and the size
2935  * of the remaining space in the rq_pages array.
2936  */
2937 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2938 {
2939         u32 buflen;
2940
2941         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2942         buflen -= rqstp->rq_auth_slack;
2943         buflen -= rqstp->rq_res.head[0].iov_len;
2944         return min_t(u32, buflen, svc_max_payload(rqstp));
2945 }
2946
2947 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2948                                    const struct nfsd4_op *op)
2949 {
2950         return (op_encode_hdr_size) * sizeof(__be32);
2951 }
2952
2953 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2954                                       const struct nfsd4_op *op)
2955 {
2956         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2957 }
2958
2959 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2960                               const struct nfsd4_op *op)
2961 {
2962         /* ac_supported, ac_resp_access */
2963         return (op_encode_hdr_size + 2)* sizeof(__be32);
2964 }
2965
2966 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2967                               const struct nfsd4_op *op)
2968 {
2969         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2970 }
2971
2972 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2973                               const struct nfsd4_op *op)
2974 {
2975         return (op_encode_hdr_size + op_encode_change_info_maxsz
2976                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2977 }
2978
2979 /*
2980  * Note since this is an idempotent operation we won't insist on failing
2981  * the op prematurely if the estimate is too large.  We may turn off splice
2982  * reads unnecessarily.
2983  */
2984 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2985                                const struct nfsd4_op *op)
2986 {
2987         const u32 *bmap = op->u.getattr.ga_bmval;
2988         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2989         u32 ret = 0;
2990
2991         if (bmap0 & FATTR4_WORD0_ACL)
2992                 return nfsd4_max_payload(rqstp);
2993         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2994                 return nfsd4_max_payload(rqstp);
2995
2996         if (bmap1 & FATTR4_WORD1_OWNER) {
2997                 ret += IDMAP_NAMESZ + 4;
2998                 bmap1 &= ~FATTR4_WORD1_OWNER;
2999         }
3000         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
3001                 ret += IDMAP_NAMESZ + 4;
3002                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
3003         }
3004         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
3005                 ret += NFS4_FHSIZE + 4;
3006                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
3007         }
3008         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
3009                 ret += NFS4_MAXLABELLEN + 12;
3010                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
3011         }
3012         /*
3013          * Largest of remaining attributes are 16 bytes (e.g.,
3014          * supported_attributes)
3015          */
3016         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
3017         /* bitmask, length */
3018         ret += 20;
3019         return ret;
3020 }
3021
3022 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
3023                              const struct nfsd4_op *op)
3024 {
3025         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
3026 }
3027
3028 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
3029                             const struct nfsd4_op *op)
3030 {
3031         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3032                 * sizeof(__be32);
3033 }
3034
3035 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
3036                             const struct nfsd4_op *op)
3037 {
3038         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
3039                 * sizeof(__be32);
3040 }
3041
3042 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
3043                             const struct nfsd4_op *op)
3044 {
3045         return (op_encode_hdr_size + op_encode_stateid_maxsz
3046                 + op_encode_change_info_maxsz + 1
3047                 + nfs4_fattr_bitmap_maxsz
3048                 + op_encode_delegation_maxsz) * sizeof(__be32);
3049 }
3050
3051 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
3052                             const struct nfsd4_op *op)
3053 {
3054         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
3055
3056         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3057 }
3058
3059 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
3060                                  const struct nfsd4_op *op)
3061 {
3062         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
3063         /*
3064          * If we detect that the file changed during hole encoding, then we
3065          * recover by encoding the remaining reply as data. This means we need
3066          * to set aside enough room to encode two data segments.
3067          */
3068         u32 seg_len = 2 * (1 + 2 + 1);
3069
3070         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
3071 }
3072
3073 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
3074                                const struct nfsd4_op *op)
3075 {
3076         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
3077
3078         return (op_encode_hdr_size + op_encode_verifier_maxsz +
3079                 XDR_QUADLEN(rlen)) * sizeof(__be32);
3080 }
3081
3082 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
3083                                 const struct nfsd4_op *op)
3084 {
3085         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
3086 }
3087
3088 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
3089                               const struct nfsd4_op *op)
3090 {
3091         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3092                 * sizeof(__be32);
3093 }
3094
3095 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
3096                               const struct nfsd4_op *op)
3097 {
3098         return (op_encode_hdr_size + op_encode_change_info_maxsz
3099                 + op_encode_change_info_maxsz) * sizeof(__be32);
3100 }
3101
3102 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
3103                                 const struct nfsd4_op *op)
3104 {
3105         return (op_encode_hdr_size
3106                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
3107 }
3108
3109 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
3110                                     const struct nfsd4_op *op)
3111 {
3112         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
3113                 * sizeof(__be32);
3114 }
3115
3116 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
3117                                const struct nfsd4_op *op)
3118 {
3119         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
3120 }
3121
3122 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
3123                                const struct nfsd4_op *op)
3124 {
3125         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
3126                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
3127 }
3128
3129 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
3130                                    const struct nfsd4_op *op)
3131 {
3132         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
3133                                                                 sizeof(__be32);
3134 }
3135
3136 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
3137                              const struct nfsd4_op *op)
3138 {
3139         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3140 }
3141
3142 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3143                                    const struct nfsd4_op *op)
3144 {
3145         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3146                 1 + 1 + /* eir_flags, spr_how */\
3147                 4 + /* spo_must_enforce & _allow with bitmap */\
3148                 2 + /*eir_server_owner.so_minor_id */\
3149                 /* eir_server_owner.so_major_id<> */\
3150                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3151                 /* eir_server_scope<> */\
3152                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3153                 1 + /* eir_server_impl_id array length */\
3154                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3155 }
3156
3157 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3158                                             const struct nfsd4_op *op)
3159 {
3160         return (op_encode_hdr_size + \
3161                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3162                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3163 }
3164
3165 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3166                                       const struct nfsd4_op *op)
3167 {
3168         return (op_encode_hdr_size + \
3169                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3170                 2 + /* csr_sequence, csr_flags */\
3171                 op_encode_channel_attrs_maxsz + \
3172                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3173 }
3174
3175 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3176                             const struct nfsd4_op *op)
3177 {
3178         return (op_encode_hdr_size +
3179                 1 /* wr_callback */ +
3180                 op_encode_stateid_maxsz /* wr_callback */ +
3181                 2 /* wr_count */ +
3182                 1 /* wr_committed */ +
3183                 op_encode_verifier_maxsz +
3184                 1 /* cr_consecutive */ +
3185                 1 /* cr_synchronous */) * sizeof(__be32);
3186 }
3187
3188 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3189                                       const struct nfsd4_op *op)
3190 {
3191         return (op_encode_hdr_size +
3192                 2 /* osr_count */ +
3193                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3194 }
3195
3196 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3197                                    const struct nfsd4_op *op)
3198 {
3199         return (op_encode_hdr_size +
3200                 3 /* cnr_lease_time */ +
3201                 1 /* We support one cnr_source_server */ +
3202                 1 /* cnr_stateid seq */ +
3203                 op_encode_stateid_maxsz /* cnr_stateid */ +
3204                 1 /* num cnr_source_server*/ +
3205                 1 /* nl4_type */ +
3206                 1 /* nl4 size */ +
3207                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3208                 * sizeof(__be32);
3209 }
3210
3211 static u32 nfsd4_get_dir_delegation_rsize(const struct svc_rqst *rqstp,
3212                                           const struct nfsd4_op *op)
3213 {
3214         return (op_encode_hdr_size +
3215                 1 /* gddr_status */ +
3216                 op_encode_verifier_maxsz +
3217                 op_encode_stateid_maxsz +
3218                 2 /* gddr_notification */ +
3219                 2 /* gddr_child_attributes */ +
3220                 2 /* gddr_dir_attributes */);
3221 }
3222
3223 #ifdef CONFIG_NFSD_PNFS
3224 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3225                                      const struct nfsd4_op *op)
3226 {
3227         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3228
3229         return (op_encode_hdr_size +
3230                 1 /* gd_layout_type*/ +
3231                 XDR_QUADLEN(rlen) +
3232                 2 /* gd_notify_types */) * sizeof(__be32);
3233 }
3234
3235 /*
3236  * At this stage we don't really know what layout driver will handle the request,
3237  * so we need to define an arbitrary upper bound here.
3238  */
3239 #define MAX_LAYOUT_SIZE         128
3240 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3241                                  const struct nfsd4_op *op)
3242 {
3243         return (op_encode_hdr_size +
3244                 1 /* logr_return_on_close */ +
3245                 op_encode_stateid_maxsz +
3246                 1 /* nr of layouts */ +
3247                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3248 }
3249
3250 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3251                                     const struct nfsd4_op *op)
3252 {
3253         return (op_encode_hdr_size +
3254                 1 /* locr_newsize */ +
3255                 2 /* ns_size */) * sizeof(__be32);
3256 }
3257
3258 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3259                                     const struct nfsd4_op *op)
3260 {
3261         return (op_encode_hdr_size +
3262                 1 /* lrs_stateid */ +
3263                 op_encode_stateid_maxsz) * sizeof(__be32);
3264 }
3265 #endif /* CONFIG_NFSD_PNFS */
3266
3267
3268 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3269                             const struct nfsd4_op *op)
3270 {
3271         return (op_encode_hdr_size + 3) * sizeof(__be32);
3272 }
3273
3274 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3275                                 const struct nfsd4_op *op)
3276 {
3277         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3278
3279         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3280 }
3281
3282 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3283                                 const struct nfsd4_op *op)
3284 {
3285         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3286                 * sizeof(__be32);
3287 }
3288 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3289                                   const struct nfsd4_op *op)
3290 {
3291         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3292
3293         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3294 }
3295
3296 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3297                                    const struct nfsd4_op *op)
3298 {
3299         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3300                 * sizeof(__be32);
3301 }
3302
3303
3304 static const struct nfsd4_operation nfsd4_ops[] = {
3305         [OP_ACCESS] = {
3306                 .op_func = nfsd4_access,
3307                 .op_name = "OP_ACCESS",
3308                 .op_rsize_bop = nfsd4_access_rsize,
3309         },
3310         [OP_CLOSE] = {
3311                 .op_func = nfsd4_close,
3312                 .op_flags = OP_MODIFIES_SOMETHING,
3313                 .op_name = "OP_CLOSE",
3314                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3315                 .op_get_currentstateid = nfsd4_get_closestateid,
3316                 .op_set_currentstateid = nfsd4_set_closestateid,
3317         },
3318         [OP_COMMIT] = {
3319                 .op_func = nfsd4_commit,
3320                 .op_flags = OP_MODIFIES_SOMETHING,
3321                 .op_name = "OP_COMMIT",
3322                 .op_rsize_bop = nfsd4_commit_rsize,
3323         },
3324         [OP_CREATE] = {
3325                 .op_func = nfsd4_create,
3326                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3327                 .op_name = "OP_CREATE",
3328                 .op_rsize_bop = nfsd4_create_rsize,
3329         },
3330         [OP_DELEGRETURN] = {
3331                 .op_func = nfsd4_delegreturn,
3332                 .op_flags = OP_MODIFIES_SOMETHING,
3333                 .op_name = "OP_DELEGRETURN",
3334                 .op_rsize_bop = nfsd4_only_status_rsize,
3335                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3336         },
3337         [OP_GETATTR] = {
3338                 .op_func = nfsd4_getattr,
3339                 .op_flags = ALLOWED_ON_ABSENT_FS,
3340                 .op_rsize_bop = nfsd4_getattr_rsize,
3341                 .op_name = "OP_GETATTR",
3342         },
3343         [OP_GETFH] = {
3344                 .op_func = nfsd4_getfh,
3345                 .op_name = "OP_GETFH",
3346                 .op_rsize_bop = nfsd4_getfh_rsize,
3347         },
3348         [OP_LINK] = {
3349                 .op_func = nfsd4_link,
3350                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3351                                 | OP_CACHEME,
3352                 .op_name = "OP_LINK",
3353                 .op_rsize_bop = nfsd4_link_rsize,
3354         },
3355         [OP_LOCK] = {
3356                 .op_func = nfsd4_lock,
3357                 .op_release = nfsd4_lock_release,
3358                 .op_flags = OP_MODIFIES_SOMETHING |
3359                                 OP_NONTRIVIAL_ERROR_ENCODE,
3360                 .op_name = "OP_LOCK",
3361                 .op_rsize_bop = nfsd4_lock_rsize,
3362                 .op_set_currentstateid = nfsd4_set_lockstateid,
3363         },
3364         [OP_LOCKT] = {
3365                 .op_func = nfsd4_lockt,
3366                 .op_release = nfsd4_lockt_release,
3367                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3368                 .op_name = "OP_LOCKT",
3369                 .op_rsize_bop = nfsd4_lock_rsize,
3370         },
3371         [OP_LOCKU] = {
3372                 .op_func = nfsd4_locku,
3373                 .op_flags = OP_MODIFIES_SOMETHING,
3374                 .op_name = "OP_LOCKU",
3375                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3376                 .op_get_currentstateid = nfsd4_get_lockustateid,
3377         },
3378         [OP_LOOKUP] = {
3379                 .op_func = nfsd4_lookup,
3380                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3381                 .op_name = "OP_LOOKUP",
3382                 .op_rsize_bop = nfsd4_only_status_rsize,
3383         },
3384         [OP_LOOKUPP] = {
3385                 .op_func = nfsd4_lookupp,
3386                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3387                 .op_name = "OP_LOOKUPP",
3388                 .op_rsize_bop = nfsd4_only_status_rsize,
3389         },
3390         [OP_NVERIFY] = {
3391                 .op_func = nfsd4_nverify,
3392                 .op_name = "OP_NVERIFY",
3393                 .op_rsize_bop = nfsd4_only_status_rsize,
3394         },
3395         [OP_OPEN] = {
3396                 .op_func = nfsd4_open,
3397                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3398                 .op_name = "OP_OPEN",
3399                 .op_rsize_bop = nfsd4_open_rsize,
3400                 .op_set_currentstateid = nfsd4_set_openstateid,
3401         },
3402         [OP_OPEN_CONFIRM] = {
3403                 .op_func = nfsd4_open_confirm,
3404                 .op_flags = OP_MODIFIES_SOMETHING,
3405                 .op_name = "OP_OPEN_CONFIRM",
3406                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3407         },
3408         [OP_OPEN_DOWNGRADE] = {
3409                 .op_func = nfsd4_open_downgrade,
3410                 .op_flags = OP_MODIFIES_SOMETHING,
3411                 .op_name = "OP_OPEN_DOWNGRADE",
3412                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3413                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3414                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3415         },
3416         [OP_PUTFH] = {
3417                 .op_func = nfsd4_putfh,
3418                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3419                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3420                 .op_name = "OP_PUTFH",
3421                 .op_rsize_bop = nfsd4_only_status_rsize,
3422         },
3423         [OP_PUTPUBFH] = {
3424                 .op_func = nfsd4_putrootfh,
3425                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3426                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3427                 .op_name = "OP_PUTPUBFH",
3428                 .op_rsize_bop = nfsd4_only_status_rsize,
3429         },
3430         [OP_PUTROOTFH] = {
3431                 .op_func = nfsd4_putrootfh,
3432                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3433                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3434                 .op_name = "OP_PUTROOTFH",
3435                 .op_rsize_bop = nfsd4_only_status_rsize,
3436         },
3437         [OP_READ] = {
3438                 .op_func = nfsd4_read,
3439                 .op_release = nfsd4_read_release,
3440                 .op_name = "OP_READ",
3441                 .op_rsize_bop = nfsd4_read_rsize,
3442                 .op_get_currentstateid = nfsd4_get_readstateid,
3443         },
3444         [OP_READDIR] = {
3445                 .op_func = nfsd4_readdir,
3446                 .op_name = "OP_READDIR",
3447                 .op_rsize_bop = nfsd4_readdir_rsize,
3448         },
3449         [OP_READLINK] = {
3450                 .op_func = nfsd4_readlink,
3451                 .op_name = "OP_READLINK",
3452                 .op_rsize_bop = nfsd4_readlink_rsize,
3453         },
3454         [OP_REMOVE] = {
3455                 .op_func = nfsd4_remove,
3456                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3457                 .op_name = "OP_REMOVE",
3458                 .op_rsize_bop = nfsd4_remove_rsize,
3459         },
3460         [OP_RENAME] = {
3461                 .op_func = nfsd4_rename,
3462                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3463                 .op_name = "OP_RENAME",
3464                 .op_rsize_bop = nfsd4_rename_rsize,
3465         },
3466         [OP_RENEW] = {
3467                 .op_func = nfsd4_renew,
3468                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3469                                 | OP_MODIFIES_SOMETHING,
3470                 .op_name = "OP_RENEW",
3471                 .op_rsize_bop = nfsd4_only_status_rsize,
3472
3473         },
3474         [OP_RESTOREFH] = {
3475                 .op_func = nfsd4_restorefh,
3476                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3477                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3478                 .op_name = "OP_RESTOREFH",
3479                 .op_rsize_bop = nfsd4_only_status_rsize,
3480         },
3481         [OP_SAVEFH] = {
3482                 .op_func = nfsd4_savefh,
3483                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3484                 .op_name = "OP_SAVEFH",
3485                 .op_rsize_bop = nfsd4_only_status_rsize,
3486         },
3487         [OP_SECINFO] = {
3488                 .op_func = nfsd4_secinfo,
3489                 .op_release = nfsd4_secinfo_release,
3490                 .op_flags = OP_HANDLES_WRONGSEC,
3491                 .op_name = "OP_SECINFO",
3492                 .op_rsize_bop = nfsd4_secinfo_rsize,
3493         },
3494         [OP_SETATTR] = {
3495                 .op_func = nfsd4_setattr,
3496                 .op_name = "OP_SETATTR",
3497                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3498                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3499                 .op_rsize_bop = nfsd4_setattr_rsize,
3500                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3501         },
3502         [OP_SETCLIENTID] = {
3503                 .op_func = nfsd4_setclientid,
3504                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3505                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3506                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3507                 .op_name = "OP_SETCLIENTID",
3508                 .op_rsize_bop = nfsd4_setclientid_rsize,
3509         },
3510         [OP_SETCLIENTID_CONFIRM] = {
3511                 .op_func = nfsd4_setclientid_confirm,
3512                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3513                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3514                 .op_name = "OP_SETCLIENTID_CONFIRM",
3515                 .op_rsize_bop = nfsd4_only_status_rsize,
3516         },
3517         [OP_VERIFY] = {
3518                 .op_func = nfsd4_verify,
3519                 .op_name = "OP_VERIFY",
3520                 .op_rsize_bop = nfsd4_only_status_rsize,
3521         },
3522         [OP_WRITE] = {
3523                 .op_func = nfsd4_write,
3524                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3525                 .op_name = "OP_WRITE",
3526                 .op_rsize_bop = nfsd4_write_rsize,
3527                 .op_get_currentstateid = nfsd4_get_writestateid,
3528         },
3529         [OP_RELEASE_LOCKOWNER] = {
3530                 .op_func = nfsd4_release_lockowner,
3531                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3532                                 | OP_MODIFIES_SOMETHING,
3533                 .op_name = "OP_RELEASE_LOCKOWNER",
3534                 .op_rsize_bop = nfsd4_only_status_rsize,
3535         },
3536
3537         /* NFSv4.1 operations */
3538         [OP_EXCHANGE_ID] = {
3539                 .op_func = nfsd4_exchange_id,
3540                 .op_release = nfsd4_exchange_id_release,
3541                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3542                                 | OP_MODIFIES_SOMETHING,
3543                 .op_name = "OP_EXCHANGE_ID",
3544                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3545         },
3546         [OP_BACKCHANNEL_CTL] = {
3547                 .op_func = nfsd4_backchannel_ctl,
3548                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3549                 .op_name = "OP_BACKCHANNEL_CTL",
3550                 .op_rsize_bop = nfsd4_only_status_rsize,
3551         },
3552         [OP_BIND_CONN_TO_SESSION] = {
3553                 .op_func = nfsd4_bind_conn_to_session,
3554                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3555                                 | OP_MODIFIES_SOMETHING,
3556                 .op_name = "OP_BIND_CONN_TO_SESSION",
3557                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3558         },
3559         [OP_CREATE_SESSION] = {
3560                 .op_func = nfsd4_create_session,
3561                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3562                                 | OP_MODIFIES_SOMETHING,
3563                 .op_name = "OP_CREATE_SESSION",
3564                 .op_rsize_bop = nfsd4_create_session_rsize,
3565         },
3566         [OP_DESTROY_SESSION] = {
3567                 .op_func = nfsd4_destroy_session,
3568                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3569                                 | OP_MODIFIES_SOMETHING,
3570                 .op_name = "OP_DESTROY_SESSION",
3571                 .op_rsize_bop = nfsd4_only_status_rsize,
3572         },
3573         [OP_SEQUENCE] = {
3574                 .op_func = nfsd4_sequence,
3575                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3576                 .op_name = "OP_SEQUENCE",
3577                 .op_rsize_bop = nfsd4_sequence_rsize,
3578         },
3579         [OP_DESTROY_CLIENTID] = {
3580                 .op_func = nfsd4_destroy_clientid,
3581                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3582                                 | OP_MODIFIES_SOMETHING,
3583                 .op_name = "OP_DESTROY_CLIENTID",
3584                 .op_rsize_bop = nfsd4_only_status_rsize,
3585         },
3586         [OP_RECLAIM_COMPLETE] = {
3587                 .op_func = nfsd4_reclaim_complete,
3588                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3589                 .op_name = "OP_RECLAIM_COMPLETE",
3590                 .op_rsize_bop = nfsd4_only_status_rsize,
3591         },
3592         [OP_SECINFO_NO_NAME] = {
3593                 .op_func = nfsd4_secinfo_no_name,
3594                 .op_release = nfsd4_secinfo_no_name_release,
3595                 .op_flags = OP_HANDLES_WRONGSEC,
3596                 .op_name = "OP_SECINFO_NO_NAME",
3597                 .op_rsize_bop = nfsd4_secinfo_rsize,
3598         },
3599         [OP_TEST_STATEID] = {
3600                 .op_func = nfsd4_test_stateid,
3601                 .op_flags = ALLOWED_WITHOUT_FH,
3602                 .op_name = "OP_TEST_STATEID",
3603                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3604         },
3605         [OP_FREE_STATEID] = {
3606                 .op_func = nfsd4_free_stateid,
3607                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3608                 .op_name = "OP_FREE_STATEID",
3609                 .op_get_currentstateid = nfsd4_get_freestateid,
3610                 .op_rsize_bop = nfsd4_only_status_rsize,
3611         },
3612         [OP_GET_DIR_DELEGATION] = {
3613                 .op_func = nfsd4_get_dir_delegation,
3614                 .op_flags = OP_MODIFIES_SOMETHING,
3615                 .op_name = "OP_GET_DIR_DELEGATION",
3616                 .op_rsize_bop = nfsd4_get_dir_delegation_rsize,
3617         },
3618 #ifdef CONFIG_NFSD_PNFS
3619         [OP_GETDEVICEINFO] = {
3620                 .op_func = nfsd4_getdeviceinfo,
3621                 .op_release = nfsd4_getdeviceinfo_release,
3622                 .op_flags = ALLOWED_WITHOUT_FH,
3623                 .op_name = "OP_GETDEVICEINFO",
3624                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3625         },
3626         [OP_LAYOUTGET] = {
3627                 .op_func = nfsd4_layoutget,
3628                 .op_release = nfsd4_layoutget_release,
3629                 .op_flags = OP_MODIFIES_SOMETHING,
3630                 .op_name = "OP_LAYOUTGET",
3631                 .op_rsize_bop = nfsd4_layoutget_rsize,
3632         },
3633         [OP_LAYOUTCOMMIT] = {
3634                 .op_func = nfsd4_layoutcommit,
3635                 .op_flags = OP_MODIFIES_SOMETHING,
3636                 .op_name = "OP_LAYOUTCOMMIT",
3637                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3638         },
3639         [OP_LAYOUTRETURN] = {
3640                 .op_func = nfsd4_layoutreturn,
3641                 .op_flags = OP_MODIFIES_SOMETHING,
3642                 .op_name = "OP_LAYOUTRETURN",
3643                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3644         },
3645 #endif /* CONFIG_NFSD_PNFS */
3646
3647         /* NFSv4.2 operations */
3648         [OP_ALLOCATE] = {
3649                 .op_func = nfsd4_allocate,
3650                 .op_flags = OP_MODIFIES_SOMETHING,
3651                 .op_name = "OP_ALLOCATE",
3652                 .op_rsize_bop = nfsd4_only_status_rsize,
3653         },
3654         [OP_DEALLOCATE] = {
3655                 .op_func = nfsd4_deallocate,
3656                 .op_flags = OP_MODIFIES_SOMETHING,
3657                 .op_name = "OP_DEALLOCATE",
3658                 .op_rsize_bop = nfsd4_only_status_rsize,
3659         },
3660         [OP_CLONE] = {
3661                 .op_func = nfsd4_clone,
3662                 .op_flags = OP_MODIFIES_SOMETHING,
3663                 .op_name = "OP_CLONE",
3664                 .op_rsize_bop = nfsd4_only_status_rsize,
3665         },
3666         [OP_COPY] = {
3667                 .op_func = nfsd4_copy,
3668                 .op_flags = OP_MODIFIES_SOMETHING,
3669                 .op_name = "OP_COPY",
3670                 .op_rsize_bop = nfsd4_copy_rsize,
3671         },
3672         [OP_READ_PLUS] = {
3673                 .op_func = nfsd4_read,
3674                 .op_release = nfsd4_read_release,
3675                 .op_name = "OP_READ_PLUS",
3676                 .op_rsize_bop = nfsd4_read_plus_rsize,
3677                 .op_get_currentstateid = nfsd4_get_readstateid,
3678         },
3679         [OP_SEEK] = {
3680                 .op_func = nfsd4_seek,
3681                 .op_name = "OP_SEEK",
3682                 .op_rsize_bop = nfsd4_seek_rsize,
3683         },
3684         [OP_OFFLOAD_STATUS] = {
3685                 .op_func = nfsd4_offload_status,
3686                 .op_name = "OP_OFFLOAD_STATUS",
3687                 .op_rsize_bop = nfsd4_offload_status_rsize,
3688         },
3689         [OP_OFFLOAD_CANCEL] = {
3690                 .op_func = nfsd4_offload_cancel,
3691                 .op_flags = OP_MODIFIES_SOMETHING,
3692                 .op_name = "OP_OFFLOAD_CANCEL",
3693                 .op_rsize_bop = nfsd4_only_status_rsize,
3694         },
3695         [OP_COPY_NOTIFY] = {
3696                 .op_func = nfsd4_copy_notify,
3697                 .op_flags = OP_MODIFIES_SOMETHING,
3698                 .op_name = "OP_COPY_NOTIFY",
3699                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3700         },
3701         [OP_GETXATTR] = {
3702                 .op_func = nfsd4_getxattr,
3703                 .op_name = "OP_GETXATTR",
3704                 .op_rsize_bop = nfsd4_getxattr_rsize,
3705         },
3706         [OP_SETXATTR] = {
3707                 .op_func = nfsd4_setxattr,
3708                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3709                 .op_name = "OP_SETXATTR",
3710                 .op_rsize_bop = nfsd4_setxattr_rsize,
3711         },
3712         [OP_LISTXATTRS] = {
3713                 .op_func = nfsd4_listxattrs,
3714                 .op_name = "OP_LISTXATTRS",
3715                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3716         },
3717         [OP_REMOVEXATTR] = {
3718                 .op_func = nfsd4_removexattr,
3719                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3720                 .op_name = "OP_REMOVEXATTR",
3721                 .op_rsize_bop = nfsd4_removexattr_rsize,
3722         },
3723 };
3724
3725 /**
3726  * nfsd4_spo_must_allow - Determine if the compound op contains an
3727  * operation that is allowed to be sent with machine credentials
3728  *
3729  * @rqstp: a pointer to the struct svc_rqst
3730  *
3731  * Checks to see if the compound contains a spo_must_allow op
3732  * and confirms that it was sent with the proper machine creds.
3733  */
3734
3735 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3736 {
3737         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3738         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3739         struct nfsd4_op *this;
3740         struct nfsd4_compound_state *cstate = &resp->cstate;
3741         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3742         u32 opiter;
3743
3744         if (!cstate->minorversion)
3745                 return false;
3746
3747         if (cstate->spo_must_allowed)
3748                 return true;
3749
3750         opiter = resp->opcnt;
3751         while (opiter < argp->opcnt) {
3752                 this = &argp->ops[opiter++];
3753                 if (test_bit(this->opnum, allow->u.longs) &&
3754                         cstate->clp->cl_mach_cred &&
3755                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3756                         cstate->spo_must_allowed = true;
3757                         return true;
3758                 }
3759         }
3760         cstate->spo_must_allowed = false;
3761         return false;
3762 }
3763
3764 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3765 {
3766         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3767                 return op_encode_hdr_size * sizeof(__be32);
3768
3769         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3770         return OPDESC(op)->op_rsize_bop(rqstp, op);
3771 }
3772
3773 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3774 {
3775         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3776                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3777                         op->opnum, nfsd4_op_name(op->opnum));
3778                 WARN_ON_ONCE(1);
3779         }
3780 }
3781
3782 static const char *nfsd4_op_name(unsigned opnum)
3783 {
3784         if (opnum < ARRAY_SIZE(nfsd4_ops))
3785                 return nfsd4_ops[opnum].op_name;
3786         return "unknown_operation";
3787 }
3788
3789 static const struct svc_procedure nfsd_procedures4[2] = {
3790         [NFSPROC4_NULL] = {
3791                 .pc_func = nfsd4_proc_null,
3792                 .pc_decode = nfssvc_decode_voidarg,
3793                 .pc_encode = nfssvc_encode_voidres,
3794                 .pc_argsize = sizeof(struct nfsd_voidargs),
3795                 .pc_argzero = sizeof(struct nfsd_voidargs),
3796                 .pc_ressize = sizeof(struct nfsd_voidres),
3797                 .pc_cachetype = RC_NOCACHE,
3798                 .pc_xdrressize = 1,
3799                 .pc_name = "NULL",
3800         },
3801         [NFSPROC4_COMPOUND] = {
3802                 .pc_func = nfsd4_proc_compound,
3803                 .pc_decode = nfs4svc_decode_compoundargs,
3804                 .pc_encode = nfs4svc_encode_compoundres,
3805                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3806                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3807                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3808                 .pc_release = nfsd4_release_compoundargs,
3809                 .pc_cachetype = RC_NOCACHE,
3810                 .pc_xdrressize = NFSD_BUFSIZE/4,
3811                 .pc_name = "COMPOUND",
3812         },
3813 };
3814
3815 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3816                               nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3817 const struct svc_version nfsd_version4 = {
3818         .vs_vers                = 4,
3819         .vs_nproc               = ARRAY_SIZE(nfsd_procedures4),
3820         .vs_proc                = nfsd_procedures4,
3821         .vs_count               = nfsd_count4,
3822         .vs_dispatch            = nfsd_dispatch,
3823         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3824         .vs_rpcb_optnl          = true,
3825         .vs_need_cong_ctrl      = true,
3826 };