fs/nfsd/nfs4proc.c

   1 /*
   2  *  Server-side procedures for NFSv4.
   3  *
   4  *  Copyright (c) 2002 The Regents of the University of Michigan.
   5  *  All rights reserved.
   6  *
   7  *  Kendrick Smith <kmsmith@umich.edu>
   8  *  Andy Adamson   <andros@umich.edu>
   9  *
  10  *  Redistribution and use in source and binary forms, with or without
  11  *  modification, are permitted provided that the following conditions
  12  *  are met:
  13  *
  14  *  1. Redistributions of source code must retain the above copyright
  15  *     notice, this list of conditions and the following disclaimer.
  16  *  2. Redistributions in binary form must reproduce the above copyright
  17  *     notice, this list of conditions and the following disclaimer in the
  18  *     documentation and/or other materials provided with the distribution.
  19  *  3. Neither the name of the University nor the names of its
  20  *     contributors may be used to endorse or promote products derived
  21  *     from this software without specific prior written permission.
  22  *
  23  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  24  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  25  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  26  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  27  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  30  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  31  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  32  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  33  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  34  */
  35 #include <linux/fs_struct.h>
  36 #include <linux/file.h>
  37 #include <linux/falloc.h>
  38 #include <linux/slab.h>
  39 #include <linux/kthread.h>
  40 #include <linux/namei.h>
  41
  42 #include <linux/sunrpc/addr.h>
  43 #include <linux/nfs_ssc.h>
  44
  45 #include "idmap.h"
  46 #include "cache.h"
  47 #include "xdr4.h"
  48 #include "vfs.h"
  49 #include "current_stateid.h"
  50 #include "netns.h"
  51 #include "acl.h"
  52 #include "pnfs.h"
  53 #include "trace.h"
  54
  55 static bool inter_copy_offload_enable;
  56 module_param(inter_copy_offload_enable, bool, 0644);
  57 MODULE_PARM_DESC(inter_copy_offload_enable,
  58                  "Enable inter server to server copy offload. Default: false");
  59
  60 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
  61 static int nfsd4_ssc_umount_timeout = 900000;           /* default to 15 mins */
  62 module_param(nfsd4_ssc_umount_timeout, int, 0644);
  63 MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
  64                 "idle msecs before unmount export from source server");
  65 #endif
  66
  67 #define NFSDDBG_FACILITY                NFSDDBG_PROC
  68
  69 static u32 nfsd_attrmask[] = {
  70         NFSD_WRITEABLE_ATTRS_WORD0,
  71         NFSD_WRITEABLE_ATTRS_WORD1,
  72         NFSD_WRITEABLE_ATTRS_WORD2
  73 };
  74
  75 static u32 nfsd41_ex_attrmask[] = {
  76         NFSD_SUPPATTR_EXCLCREAT_WORD0,
  77         NFSD_SUPPATTR_EXCLCREAT_WORD1,
  78         NFSD_SUPPATTR_EXCLCREAT_WORD2
  79 };
  80
  81 static __be32
  82 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
  83                    u32 *bmval, u32 *writable)
  84 {
  85         struct dentry *dentry = cstate->current_fh.fh_dentry;
  86         struct svc_export *exp = cstate->current_fh.fh_export;
  87
  88         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
  89                 return nfserr_attrnotsupp;
  90         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
  91                 return nfserr_attrnotsupp;
  92         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
  93                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
  94                 return nfserr_attrnotsupp;
  95         if (writable && !bmval_is_subset(bmval, writable))
  96                 return nfserr_inval;
  97         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
  98                         (bmval[1] & FATTR4_WORD1_MODE))
  99                 return nfserr_inval;
 100         return nfs_ok;
 101 }
 102
 103 static __be32
 104 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
 105         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 106 {
 107         __be32 status = nfs_ok;
 108
 109         if (open->op_create == NFS4_OPEN_CREATE) {
 110                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
 111                     || open->op_createmode == NFS4_CREATE_GUARDED)
 112                         status = check_attr_support(rqstp, cstate,
 113                                         open->op_bmval, nfsd_attrmask);
 114                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
 115                         status = check_attr_support(rqstp, cstate,
 116                                         open->op_bmval, nfsd41_ex_attrmask);
 117         }
 118
 119         return status;
 120 }
 121
 122 static int
 123 is_create_with_attrs(struct nfsd4_open *open)
 124 {
 125         return open->op_create == NFS4_OPEN_CREATE
 126                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
 127                     || open->op_createmode == NFS4_CREATE_GUARDED
 128                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
 129 }
 130
 131 static inline void
 132 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
 133 {
 134         fh_put(dst);
 135         dget(src->fh_dentry);
 136         if (src->fh_export)
 137                 exp_get(src->fh_export);
 138         *dst = *src;
 139 }
 140
 141 static __be32
 142 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
 143 {
 144
 145         if (open->op_truncate &&
 146                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
 147                 return nfserr_inval;
 148
 149         accmode |= NFSD_MAY_READ_IF_EXEC;
 150
 151         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
 152                 accmode |= NFSD_MAY_READ;
 153         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
 154                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
 155         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
 156                 accmode |= NFSD_MAY_WRITE;
 157
 158         return fh_verify(rqstp, current_fh, S_IFREG, accmode);
 159 }
 160
 161 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
 162 {
 163         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
 164
 165         if (S_ISREG(mode))
 166                 return nfs_ok;
 167         if (S_ISDIR(mode))
 168                 return nfserr_isdir;
 169         /*
 170          * Using err_symlink as our catch-all case may look odd; but
 171          * there's no other obvious error for this case in 4.0, and we
 172          * happen to know that it will cause the linux v4 client to do
 173          * the right thing on attempts to open something other than a
 174          * regular file.
 175          */
 176         return nfserr_symlink;
 177 }
 178
 179 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
 180 {
 181         if (nfsd4_has_session(cstate))
 182                 return;
 183         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
 184                         &resfh->fh_handle);
 185 }
 186
 187 static inline bool nfsd4_create_is_exclusive(int createmode)
 188 {
 189         return createmode == NFS4_CREATE_EXCLUSIVE ||
 190                 createmode == NFS4_CREATE_EXCLUSIVE4_1;
 191 }
 192
 193 static __be32
 194 nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
 195                  struct nfsd4_open *open)
 196 {
 197         struct file *filp;
 198         struct path path;
 199         int oflags;
 200
 201         oflags = O_CREAT | O_LARGEFILE;
 202         switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
 203         case NFS4_SHARE_ACCESS_WRITE:
 204                 oflags |= O_WRONLY;
 205                 break;
 206         case NFS4_SHARE_ACCESS_BOTH:
 207                 oflags |= O_RDWR;
 208                 break;
 209         default:
 210                 oflags |= O_RDONLY;
 211         }
 212
 213         path.mnt = fhp->fh_export->ex_path.mnt;
 214         path.dentry = child;
 215         filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
 216                              current_cred());
 217         if (IS_ERR(filp))
 218                 return nfserrno(PTR_ERR(filp));
 219
 220         open->op_filp = filp;
 221         return nfs_ok;
 222 }
 223
 224 /*
 225  * Implement NFSv4's unchecked, guarded, and exclusive create
 226  * semantics for regular files. Open state for this new file is
 227  * subsequently fabricated in nfsd4_process_open2().
 228  *
 229  * Upon return, caller must release @fhp and @resfhp.
 230  */
 231 static __be32
 232 nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
 233                   struct svc_fh *resfhp, struct nfsd4_open *open)
 234 {
 235         struct iattr *iap = &open->op_iattr;
 236         struct nfsd_attrs attrs = {
 237                 .na_iattr       = iap,
 238                 .na_seclabel    = &open->op_label,
 239         };
 240         struct dentry *parent, *child;
 241         __u32 v_mtime, v_atime;
 242         struct inode *inode;
 243         __be32 status;
 244         int host_err;
 245
 246         if (isdotent(open->op_fname, open->op_fnamelen))
 247                 return nfserr_exist;
 248         if (!(iap->ia_valid & ATTR_MODE))
 249                 iap->ia_mode = 0;
 250
 251         status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
 252         if (status != nfs_ok)
 253                 return status;
 254         parent = fhp->fh_dentry;
 255         inode = d_inode(parent);
 256
 257         host_err = fh_want_write(fhp);
 258         if (host_err)
 259                 return nfserrno(host_err);
 260
 261         if (is_create_with_attrs(open))
 262                 nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
 263
 264         inode_lock_nested(inode, I_MUTEX_PARENT);
 265
 266         child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
 267         if (IS_ERR(child)) {
 268                 status = nfserrno(PTR_ERR(child));
 269                 goto out;
 270         }
 271
 272         if (d_really_is_negative(child)) {
 273                 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
 274                 if (status != nfs_ok)
 275                         goto out;
 276         }
 277
 278         status = fh_compose(resfhp, fhp->fh_export, child, fhp);
 279         if (status != nfs_ok)
 280                 goto out;
 281
 282         v_mtime = 0;
 283         v_atime = 0;
 284         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 285                 u32 *verifier = (u32 *)open->op_verf.data;
 286
 287                 /*
 288                  * Solaris 7 gets confused (bugid 4218508) if these have
 289                  * the high bit set, as do xfs filesystems without the
 290                  * "bigtime" feature. So just clear the high bits. If this
 291                  * is ever changed to use different attrs for storing the
 292                  * verifier, then do_open_lookup() will also need to be
 293                  * fixed accordingly.
 294                  */
 295                 v_mtime = verifier[0] & 0x7fffffff;
 296                 v_atime = verifier[1] & 0x7fffffff;
 297         }
 298
 299         if (d_really_is_positive(child)) {
 300                 /* NFSv4 protocol requires change attributes even though
 301                  * no change happened.
 302                  */
 303                 status = fh_fill_both_attrs(fhp);
 304                 if (status != nfs_ok)
 305                         goto out;
 306
 307                 switch (open->op_createmode) {
 308                 case NFS4_CREATE_UNCHECKED:
 309                         if (!d_is_reg(child))
 310                                 break;
 311
 312                         /*
 313                          * In NFSv4, we don't want to truncate the file
 314                          * now. This would be wrong if the OPEN fails for
 315                          * some other reason. Furthermore, if the size is
 316                          * nonzero, we should ignore it according to spec!
 317                          */
 318                         open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
 319                                                 !iap->ia_size;
 320                         break;
 321                 case NFS4_CREATE_GUARDED:
 322                         status = nfserr_exist;
 323                         break;
 324                 case NFS4_CREATE_EXCLUSIVE:
 325                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
 326                             d_inode(child)->i_atime.tv_sec == v_atime &&
 327                             d_inode(child)->i_size == 0) {
 328                                 open->op_created = true;
 329                                 break;          /* subtle */
 330                         }
 331                         status = nfserr_exist;
 332                         break;
 333                 case NFS4_CREATE_EXCLUSIVE4_1:
 334                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
 335                             d_inode(child)->i_atime.tv_sec == v_atime &&
 336                             d_inode(child)->i_size == 0) {
 337                                 open->op_created = true;
 338                                 goto set_attr;  /* subtle */
 339                         }
 340                         status = nfserr_exist;
 341                 }
 342                 goto out;
 343         }
 344
 345         if (!IS_POSIXACL(inode))
 346                 iap->ia_mode &= ~current_umask();
 347
 348         status = fh_fill_pre_attrs(fhp);
 349         if (status != nfs_ok)
 350                 goto out;
 351         status = nfsd4_vfs_create(fhp, child, open);
 352         if (status != nfs_ok)
 353                 goto out;
 354         open->op_created = true;
 355         fh_fill_post_attrs(fhp);
 356
 357         /* A newly created file already has a file size of zero. */
 358         if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
 359                 iap->ia_valid &= ~ATTR_SIZE;
 360         if (nfsd4_create_is_exclusive(open->op_createmode)) {
 361                 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
 362                                 ATTR_MTIME_SET|ATTR_ATIME_SET;
 363                 iap->ia_mtime.tv_sec = v_mtime;
 364                 iap->ia_atime.tv_sec = v_atime;
 365                 iap->ia_mtime.tv_nsec = 0;
 366                 iap->ia_atime.tv_nsec = 0;
 367         }
 368
 369 set_attr:
 370         status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
 371
 372         if (attrs.na_labelerr)
 373                 open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 374         if (attrs.na_aclerr)
 375                 open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
 376 out:
 377         inode_unlock(inode);
 378         nfsd_attrs_free(&attrs);
 379         if (child && !IS_ERR(child))
 380                 dput(child);
 381         fh_drop_write(fhp);
 382         return status;
 383 }
 384
 385 /**
 386  * set_change_info - set up the change_info4 for a reply
 387  * @cinfo: pointer to nfsd4_change_info to be populated
 388  * @fhp: pointer to svc_fh to use as source
 389  *
 390  * Many operations in NFSv4 require change_info4 in the reply. This function
 391  * populates that from the info that we (should!) have already collected. In
 392  * the event that we didn't get any pre-attrs, just zero out both.
 393  */
 394 static void
 395 set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp)
 396 {
 397         cinfo->atomic = (u32)(fhp->fh_pre_saved && fhp->fh_post_saved && !fhp->fh_no_atomic_attr);
 398         cinfo->before_change = fhp->fh_pre_change;
 399         cinfo->after_change = fhp->fh_post_change;
 400
 401         /*
 402          * If fetching the pre-change attributes failed, then we should
 403          * have already failed the whole operation. We could have still
 404          * failed to fetch post-change attributes however.
 405          *
 406          * If we didn't get post-op attrs, just zero-out the after
 407          * field since we don't know what it should be. If the pre_saved
 408          * field isn't set for some reason, throw warning and just copy
 409          * whatever is in the after field.
 410          */
 411         if (WARN_ON_ONCE(!fhp->fh_pre_saved))
 412                 cinfo->before_change = 0;
 413         if (!fhp->fh_post_saved)
 414                 cinfo->after_change = cinfo->before_change + 1;
 415 }
 416
 417 static __be32
 418 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
 419 {
 420         struct svc_fh *current_fh = &cstate->current_fh;
 421         int accmode;
 422         __be32 status;
 423
 424         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
 425         if (!*resfh)
 426                 return nfserr_jukebox;
 427         fh_init(*resfh, NFS4_FHSIZE);
 428         open->op_truncate = false;
 429
 430         if (open->op_create) {
 431                 /* FIXME: check session persistence and pnfs flags.
 432                  * The nfsv4.1 spec requires the following semantics:
 433                  *
 434                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
 435                  * Reply Cache  | server |                 |
 436                  * -------------+--------+-----------------+--------------------
 437                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
 438                  *              |        |                 | (SHOULD)
 439                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
 440                  *              |        |                 | (SHOULD NOT)
 441                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
 442                  * yes          | no     | GUARDED4        | GUARDED4
 443                  * yes          | yes    | GUARDED4        | GUARDED4
 444                  */
 445
 446                 current->fs->umask = open->op_umask;
 447                 status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
 448                 current->fs->umask = 0;
 449
 450                 /*
 451                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
 452                  * use the returned bitmask to indicate which attributes
 453                  * we used to store the verifier:
 454                  */
 455                 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
 456                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
 457                                                 FATTR4_WORD1_TIME_MODIFY);
 458         } else {
 459                 status = nfsd_lookup(rqstp, current_fh,
 460                                      open->op_fname, open->op_fnamelen, *resfh);
 461                 if (status == nfs_ok)
 462                         /* NFSv4 protocol requires change attributes even though
 463                          * no change happened.
 464                          */
 465                         status = fh_fill_both_attrs(current_fh);
 466         }
 467         if (status)
 468                 goto out;
 469         status = nfsd_check_obj_isreg(*resfh);
 470         if (status)
 471                 goto out;
 472
 473         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
 474         accmode = NFSD_MAY_NOP;
 475         if (open->op_created ||
 476                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
 477                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
 478         status = do_open_permission(rqstp, *resfh, open, accmode);
 479         set_change_info(&open->op_cinfo, current_fh);
 480 out:
 481         return status;
 482 }
 483
 484 static __be32
 485 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
 486 {
 487         struct svc_fh *current_fh = &cstate->current_fh;
 488         int accmode = 0;
 489
 490         /* We don't know the target directory, and therefore can not
 491         * set the change info
 492         */
 493
 494         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
 495
 496         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
 497
 498         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
 499                 (open->op_iattr.ia_size == 0);
 500         /*
 501          * In the delegation case, the client is telling us about an
 502          * open that it *already* performed locally, some time ago.  We
 503          * should let it succeed now if possible.
 504          *
 505          * In the case of a CLAIM_FH open, on the other hand, the client
 506          * may be counting on us to enforce permissions (the Linux 4.1
 507          * client uses this for normal opens, for example).
 508          */
 509         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
 510                 accmode = NFSD_MAY_OWNER_OVERRIDE;
 511
 512         return do_open_permission(rqstp, current_fh, open, accmode);
 513 }
 514
 515 static void
 516 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
 517 {
 518         struct nfsd4_sessionid *sid =
 519                         (struct nfsd4_sessionid *)session->se_sessionid.data;
 520
 521         clid->cl_boot = sid->clientid.cl_boot;
 522         clid->cl_id = sid->clientid.cl_id;
 523 }
 524
 525 static __be32
 526 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 527            union nfsd4_op_u *u)
 528 {
 529         struct nfsd4_open *open = &u->open;
 530         __be32 status;
 531         struct svc_fh *resfh = NULL;
 532         struct net *net = SVC_NET(rqstp);
 533         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 534         bool reclaim = false;
 535
 536         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
 537                 (int)open->op_fnamelen, open->op_fname,
 538                 open->op_openowner);
 539
 540         open->op_filp = NULL;
 541         open->op_rqstp = rqstp;
 542
 543         /* This check required by spec. */
 544         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
 545                 return nfserr_inval;
 546
 547         open->op_created = false;
 548         /*
 549          * RFC5661 18.51.3
 550          * Before RECLAIM_COMPLETE done, server should deny new lock
 551          */
 552         if (nfsd4_has_session(cstate) &&
 553             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
 554             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 555                 return nfserr_grace;
 556
 557         if (nfsd4_has_session(cstate))
 558                 copy_clientid(&open->op_clientid, cstate->session);
 559
 560         /* check seqid for replay. set nfs4_owner */
 561         status = nfsd4_process_open1(cstate, open, nn);
 562         if (status == nfserr_replay_me) {
 563                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
 564                 fh_put(&cstate->current_fh);
 565                 fh_copy_shallow(&cstate->current_fh.fh_handle,
 566                                 &rp->rp_openfh);
 567                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 568                 if (status)
 569                         dprintk("nfsd4_open: replay failed"
 570                                 " restoring previous filehandle\n");
 571                 else
 572                         status = nfserr_replay_me;
 573         }
 574         if (status)
 575                 goto out;
 576         if (open->op_xdr_error) {
 577                 status = open->op_xdr_error;
 578                 goto out;
 579         }
 580
 581         status = nfsd4_check_open_attributes(rqstp, cstate, open);
 582         if (status)
 583                 goto out;
 584
 585         /* Openowner is now set, so sequence id will get bumped.  Now we need
 586          * these checks before we do any creates: */
 587         status = nfserr_grace;
 588         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
 589                 goto out;
 590         status = nfserr_no_grace;
 591         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
 592                 goto out;
 593
 594         switch (open->op_claim_type) {
 595         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
 596         case NFS4_OPEN_CLAIM_NULL:
 597                 status = do_open_lookup(rqstp, cstate, open, &resfh);
 598                 if (status)
 599                         goto out;
 600                 break;
 601         case NFS4_OPEN_CLAIM_PREVIOUS:
 602                 status = nfs4_check_open_reclaim(cstate->clp);
 603                 if (status)
 604                         goto out;
 605                 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
 606                 reclaim = true;
 607                 fallthrough;
 608         case NFS4_OPEN_CLAIM_FH:
 609         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
 610                 status = do_open_fhandle(rqstp, cstate, open);
 611                 if (status)
 612                         goto out;
 613                 resfh = &cstate->current_fh;
 614                 break;
 615         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
 616         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
 617                 status = nfserr_notsupp;
 618                 goto out;
 619         default:
 620                 status = nfserr_inval;
 621                 goto out;
 622         }
 623
 624         status = nfsd4_process_open2(rqstp, resfh, open);
 625         if (status && open->op_created)
 626                 pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
 627                         be32_to_cpu(status));
 628         if (reclaim && !status)
 629                 nn->somebody_reclaimed = true;
 630 out:
 631         if (open->op_filp) {
 632                 fput(open->op_filp);
 633                 open->op_filp = NULL;
 634         }
 635         if (resfh && resfh != &cstate->current_fh) {
 636                 fh_dup2(&cstate->current_fh, resfh);
 637                 fh_put(resfh);
 638                 kfree(resfh);
 639         }
 640         nfsd4_cleanup_open_state(cstate, open);
 641         nfsd4_bump_seqid(cstate, status);
 642         return status;
 643 }
 644
 645 /*
 646  * OPEN is the only seqid-mutating operation whose decoding can fail
 647  * with a seqid-mutating error (specifically, decoding of user names in
 648  * the attributes).  Therefore we have to do some processing to look up
 649  * the stateowner so that we can bump the seqid.
 650  */
 651 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
 652 {
 653         struct nfsd4_open *open = &op->u.open;
 654
 655         if (!seqid_mutating_err(ntohl(op->status)))
 656                 return op->status;
 657         if (nfsd4_has_session(cstate))
 658                 return op->status;
 659         open->op_xdr_error = op->status;
 660         return nfsd4_open(rqstp, cstate, &op->u);
 661 }
 662
 663 /*
 664  * filehandle-manipulating ops.
 665  */
 666 static __be32
 667 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 668             union nfsd4_op_u *u)
 669 {
 670         u->getfh = &cstate->current_fh;
 671         return nfs_ok;
 672 }
 673
 674 static __be32
 675 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 676             union nfsd4_op_u *u)
 677 {
 678         struct nfsd4_putfh *putfh = &u->putfh;
 679         __be32 ret;
 680
 681         fh_put(&cstate->current_fh);
 682         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
 683         memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
 684                putfh->pf_fhlen);
 685         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
 686 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
 687         if (ret == nfserr_stale && putfh->no_verify) {
 688                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
 689                 ret = 0;
 690         }
 691 #endif
 692         return ret;
 693 }
 694
 695 static __be32
 696 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 697                 union nfsd4_op_u *u)
 698 {
 699         fh_put(&cstate->current_fh);
 700
 701         return exp_pseudoroot(rqstp, &cstate->current_fh);
 702 }
 703
 704 static __be32
 705 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 706                 union nfsd4_op_u *u)
 707 {
 708         if (!cstate->save_fh.fh_dentry)
 709                 return nfserr_restorefh;
 710
 711         fh_dup2(&cstate->current_fh, &cstate->save_fh);
 712         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
 713                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
 714                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
 715         }
 716         return nfs_ok;
 717 }
 718
 719 static __be32
 720 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 721              union nfsd4_op_u *u)
 722 {
 723         fh_dup2(&cstate->save_fh, &cstate->current_fh);
 724         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
 725                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
 726                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
 727         }
 728         return nfs_ok;
 729 }
 730
 731 /*
 732  * misc nfsv4 ops
 733  */
 734 static __be32
 735 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 736              union nfsd4_op_u *u)
 737 {
 738         struct nfsd4_access *access = &u->access;
 739         u32 access_full;
 740
 741         access_full = NFS3_ACCESS_FULL;
 742         if (cstate->minorversion >= 2)
 743                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
 744                                NFS4_ACCESS_XAWRITE;
 745
 746         if (access->ac_req_access & ~access_full)
 747                 return nfserr_inval;
 748
 749         access->ac_resp_access = access->ac_req_access;
 750         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
 751                            &access->ac_supported);
 752 }
 753
 754 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
 755 {
 756         __be32 *verf = (__be32 *)verifier->data;
 757
 758         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
 759
 760         nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
 761 }
 762
 763 static __be32
 764 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 765              union nfsd4_op_u *u)
 766 {
 767         struct nfsd4_commit *commit = &u->commit;
 768         struct nfsd_file *nf;
 769         __be32 status;
 770
 771         status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
 772                                    NFSD_MAY_NOT_BREAK_LEASE, &nf);
 773         if (status != nfs_ok)
 774                 return status;
 775
 776         status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
 777                              commit->co_count,
 778                              (__be32 *)commit->co_verf.data);
 779         nfsd_file_put(nf);
 780         return status;
 781 }
 782
 783 static __be32
 784 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 785              union nfsd4_op_u *u)
 786 {
 787         struct nfsd4_create *create = &u->create;
 788         struct nfsd_attrs attrs = {
 789                 .na_iattr       = &create->cr_iattr,
 790                 .na_seclabel    = &create->cr_label,
 791         };
 792         struct svc_fh resfh;
 793         __be32 status;
 794         dev_t rdev;
 795
 796         fh_init(&resfh, NFS4_FHSIZE);
 797
 798         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
 799         if (status)
 800                 return status;
 801
 802         status = check_attr_support(rqstp, cstate, create->cr_bmval,
 803                                     nfsd_attrmask);
 804         if (status)
 805                 return status;
 806
 807         status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
 808         current->fs->umask = create->cr_umask;
 809         switch (create->cr_type) {
 810         case NF4LNK:
 811                 status = nfsd_symlink(rqstp, &cstate->current_fh,
 812                                       create->cr_name, create->cr_namelen,
 813                                       create->cr_data, &attrs, &resfh);
 814                 break;
 815
 816         case NF4BLK:
 817                 status = nfserr_inval;
 818                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 819                 if (MAJOR(rdev) != create->cr_specdata1 ||
 820                     MINOR(rdev) != create->cr_specdata2)
 821                         goto out_umask;
 822                 status = nfsd_create(rqstp, &cstate->current_fh,
 823                                      create->cr_name, create->cr_namelen,
 824                                      &attrs, S_IFBLK, rdev, &resfh);
 825                 break;
 826
 827         case NF4CHR:
 828                 status = nfserr_inval;
 829                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
 830                 if (MAJOR(rdev) != create->cr_specdata1 ||
 831                     MINOR(rdev) != create->cr_specdata2)
 832                         goto out_umask;
 833                 status = nfsd_create(rqstp, &cstate->current_fh,
 834                                      create->cr_name, create->cr_namelen,
 835                                      &attrs, S_IFCHR, rdev, &resfh);
 836                 break;
 837
 838         case NF4SOCK:
 839                 status = nfsd_create(rqstp, &cstate->current_fh,
 840                                      create->cr_name, create->cr_namelen,
 841                                      &attrs, S_IFSOCK, 0, &resfh);
 842                 break;
 843
 844         case NF4FIFO:
 845                 status = nfsd_create(rqstp, &cstate->current_fh,
 846                                      create->cr_name, create->cr_namelen,
 847                                      &attrs, S_IFIFO, 0, &resfh);
 848                 break;
 849
 850         case NF4DIR:
 851                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
 852                 status = nfsd_create(rqstp, &cstate->current_fh,
 853                                      create->cr_name, create->cr_namelen,
 854                                      &attrs, S_IFDIR, 0, &resfh);
 855                 break;
 856
 857         default:
 858                 status = nfserr_badtype;
 859         }
 860
 861         if (status)
 862                 goto out;
 863
 864         if (attrs.na_labelerr)
 865                 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
 866         if (attrs.na_aclerr)
 867                 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
 868         set_change_info(&create->cr_cinfo, &cstate->current_fh);
 869         fh_dup2(&cstate->current_fh, &resfh);
 870 out:
 871         fh_put(&resfh);
 872 out_umask:
 873         current->fs->umask = 0;
 874         nfsd_attrs_free(&attrs);
 875         return status;
 876 }
 877
 878 static __be32
 879 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 880               union nfsd4_op_u *u)
 881 {
 882         struct nfsd4_getattr *getattr = &u->getattr;
 883         __be32 status;
 884
 885         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
 886         if (status)
 887                 return status;
 888
 889         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 890                 return nfserr_inval;
 891
 892         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
 893         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
 894         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 895
 896         getattr->ga_fhp = &cstate->current_fh;
 897         return nfs_ok;
 898 }
 899
 900 static __be32
 901 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 902            union nfsd4_op_u *u)
 903 {
 904         struct nfsd4_link *link = &u->link;
 905         __be32 status;
 906
 907         status = nfsd_link(rqstp, &cstate->current_fh,
 908                            link->li_name, link->li_namelen, &cstate->save_fh);
 909         if (!status)
 910                 set_change_info(&link->li_cinfo, &cstate->current_fh);
 911         return status;
 912 }
 913
 914 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
 915 {
 916         struct svc_fh tmp_fh;
 917         __be32 ret;
 918
 919         fh_init(&tmp_fh, NFS4_FHSIZE);
 920         ret = exp_pseudoroot(rqstp, &tmp_fh);
 921         if (ret)
 922                 return ret;
 923         if (tmp_fh.fh_dentry == fh->fh_dentry) {
 924                 fh_put(&tmp_fh);
 925                 return nfserr_noent;
 926         }
 927         fh_put(&tmp_fh);
 928         return nfsd_lookup(rqstp, fh, "..", 2, fh);
 929 }
 930
 931 static __be32
 932 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 933               union nfsd4_op_u *u)
 934 {
 935         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
 936 }
 937
 938 static __be32
 939 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 940              union nfsd4_op_u *u)
 941 {
 942         return nfsd_lookup(rqstp, &cstate->current_fh,
 943                            u->lookup.lo_name, u->lookup.lo_len,
 944                            &cstate->current_fh);
 945 }
 946
 947 static __be32
 948 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 949            union nfsd4_op_u *u)
 950 {
 951         struct nfsd4_read *read = &u->read;
 952         __be32 status;
 953
 954         read->rd_nf = NULL;
 955
 956         trace_nfsd_read_start(rqstp, &cstate->current_fh,
 957                               read->rd_offset, read->rd_length);
 958
 959         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
 960         if (read->rd_offset > (u64)OFFSET_MAX)
 961                 read->rd_offset = (u64)OFFSET_MAX;
 962         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
 963                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
 964
 965         /*
 966          * If we do a zero copy read, then a client will see read data
 967          * that reflects the state of the file *after* performing the
 968          * following compound.
 969          *
 970          * To ensure proper ordering, we therefore turn off zero copy if
 971          * the client wants us to do more in this compound:
 972          */
 973         if (!nfsd4_last_compound_op(rqstp))
 974                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 975
 976         /* check stateid */
 977         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
 978                                         &read->rd_stateid, RD_STATE,
 979                                         &read->rd_nf, NULL);
 980
 981         read->rd_rqstp = rqstp;
 982         read->rd_fhp = &cstate->current_fh;
 983         return status;
 984 }
 985
 986
 987 static void
 988 nfsd4_read_release(union nfsd4_op_u *u)
 989 {
 990         if (u->read.rd_nf)
 991                 nfsd_file_put(u->read.rd_nf);
 992         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
 993                              u->read.rd_offset, u->read.rd_length);
 994 }
 995
 996 static __be32
 997 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 998               union nfsd4_op_u *u)
 999 {
1000         struct nfsd4_readdir *readdir = &u->readdir;
1001         u64 cookie = readdir->rd_cookie;
1002         static const nfs4_verifier zeroverf;
1003
1004         /* no need to check permission - this will be done in nfsd_readdir() */
1005
1006         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1007                 return nfserr_inval;
1008
1009         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1010         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1011         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1012
1013         if ((cookie == 1) || (cookie == 2) ||
1014             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1015                 return nfserr_bad_cookie;
1016
1017         readdir->rd_rqstp = rqstp;
1018         readdir->rd_fhp = &cstate->current_fh;
1019         return nfs_ok;
1020 }
1021
1022 static __be32
1023 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1024                union nfsd4_op_u *u)
1025 {
1026         u->readlink.rl_rqstp = rqstp;
1027         u->readlink.rl_fhp = &cstate->current_fh;
1028         return nfs_ok;
1029 }
1030
1031 static __be32
1032 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1033              union nfsd4_op_u *u)
1034 {
1035         struct nfsd4_remove *remove = &u->remove;
1036         __be32 status;
1037
1038         if (opens_in_grace(SVC_NET(rqstp)))
1039                 return nfserr_grace;
1040         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1041                              remove->rm_name, remove->rm_namelen);
1042         if (!status)
1043                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1044         return status;
1045 }
1046
1047 static __be32
1048 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1049              union nfsd4_op_u *u)
1050 {
1051         struct nfsd4_rename *rename = &u->rename;
1052         __be32 status;
1053
1054         if (opens_in_grace(SVC_NET(rqstp)))
1055                 return nfserr_grace;
1056         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1057                              rename->rn_snamelen, &cstate->current_fh,
1058                              rename->rn_tname, rename->rn_tnamelen);
1059         if (status)
1060                 return status;
1061         set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1062         set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1063         return nfs_ok;
1064 }
1065
1066 static __be32
1067 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1068               union nfsd4_op_u *u)
1069 {
1070         struct nfsd4_secinfo *secinfo = &u->secinfo;
1071         struct svc_export *exp;
1072         struct dentry *dentry;
1073         __be32 err;
1074
1075         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1076         if (err)
1077                 return err;
1078         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1079                                     secinfo->si_name, secinfo->si_namelen,
1080                                     &exp, &dentry);
1081         if (err)
1082                 return err;
1083         if (d_really_is_negative(dentry)) {
1084                 exp_put(exp);
1085                 err = nfserr_noent;
1086         } else
1087                 secinfo->si_exp = exp;
1088         dput(dentry);
1089         if (cstate->minorversion)
1090                 /* See rfc 5661 section 2.6.3.1.1.8 */
1091                 fh_put(&cstate->current_fh);
1092         return err;
1093 }
1094
1095 static __be32
1096 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1097                 union nfsd4_op_u *u)
1098 {
1099         __be32 err;
1100
1101         switch (u->secinfo_no_name.sin_style) {
1102         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1103                 break;
1104         case NFS4_SECINFO_STYLE4_PARENT:
1105                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1106                 if (err)
1107                         return err;
1108                 break;
1109         default:
1110                 return nfserr_inval;
1111         }
1112
1113         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1114         fh_put(&cstate->current_fh);
1115         return nfs_ok;
1116 }
1117
1118 static void
1119 nfsd4_secinfo_release(union nfsd4_op_u *u)
1120 {
1121         if (u->secinfo.si_exp)
1122                 exp_put(u->secinfo.si_exp);
1123 }
1124
1125 static void
1126 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1127 {
1128         if (u->secinfo_no_name.sin_exp)
1129                 exp_put(u->secinfo_no_name.sin_exp);
1130 }
1131
1132 static __be32
1133 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1134               union nfsd4_op_u *u)
1135 {
1136         struct nfsd4_setattr *setattr = &u->setattr;
1137         struct nfsd_attrs attrs = {
1138                 .na_iattr       = &setattr->sa_iattr,
1139                 .na_seclabel    = &setattr->sa_label,
1140         };
1141         struct inode *inode;
1142         __be32 status = nfs_ok;
1143         int err;
1144
1145         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1146                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1147                                 &cstate->current_fh, &setattr->sa_stateid,
1148                                 WR_STATE, NULL, NULL);
1149                 if (status)
1150                         return status;
1151         }
1152         err = fh_want_write(&cstate->current_fh);
1153         if (err)
1154                 return nfserrno(err);
1155         status = nfs_ok;
1156
1157         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1158                                     nfsd_attrmask);
1159         if (status)
1160                 goto out;
1161
1162         inode = cstate->current_fh.fh_dentry->d_inode;
1163         status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1164                                    setattr->sa_acl, &attrs);
1165
1166         if (status)
1167                 goto out;
1168         status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1169                                 0, (time64_t)0);
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 static void nfs4_put_copy(struct nfsd4_copy *copy)
1282 {
1283         if (!refcount_dec_and_test(&copy->refcount))
1284                 return;
1285         kfree(copy->cp_src);
1286         kfree(copy);
1287 }
1288
1289 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1290 {
1291         if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1292                 kthread_stop(copy->copy_task);
1293         nfs4_put_copy(copy);
1294 }
1295
1296 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1297 {
1298         struct nfsd4_copy *copy = NULL;
1299
1300         spin_lock(&clp->async_lock);
1301         if (!list_empty(&clp->async_copies)) {
1302                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1303                                         copies);
1304                 refcount_inc(&copy->refcount);
1305         }
1306         spin_unlock(&clp->async_lock);
1307         return copy;
1308 }
1309
1310 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1311 {
1312         struct nfsd4_copy *copy;
1313
1314         while ((copy = nfsd4_get_copy(clp)) != NULL)
1315                 nfsd4_stop_copy(copy);
1316 }
1317 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1318
1319 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1320                                    struct nfs_fh *src_fh,
1321                                    nfs4_stateid *stateid);
1322 extern void nfs42_ssc_close(struct file *filep);
1323
1324 extern void nfs_sb_deactive(struct super_block *sb);
1325
1326 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1327
1328 /*
1329  * setup a work entry in the ssc delayed unmount list.
1330  */
1331 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1332                                   struct nfsd4_ssc_umount_item **nsui)
1333 {
1334         struct nfsd4_ssc_umount_item *ni = NULL;
1335         struct nfsd4_ssc_umount_item *work = NULL;
1336         struct nfsd4_ssc_umount_item *tmp;
1337         DEFINE_WAIT(wait);
1338         __be32 status = 0;
1339
1340         *nsui = NULL;
1341         work = kzalloc(sizeof(*work), GFP_KERNEL);
1342 try_again:
1343         spin_lock(&nn->nfsd_ssc_lock);
1344         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1345                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1346                         continue;
1347                 /* found a match */
1348                 if (ni->nsui_busy) {
1349                         /*  wait - and try again */
1350                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1351                         spin_unlock(&nn->nfsd_ssc_lock);
1352
1353                         /* allow 20secs for mount/unmount for now - revisit */
1354                         if (kthread_should_stop() ||
1355                                         (schedule_timeout(20*HZ) == 0)) {
1356                                 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1357                                 kfree(work);
1358                                 return nfserr_eagain;
1359                         }
1360                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1361                         goto try_again;
1362                 }
1363                 *nsui = ni;
1364                 refcount_inc(&ni->nsui_refcnt);
1365                 spin_unlock(&nn->nfsd_ssc_lock);
1366                 kfree(work);
1367
1368                 /* return vfsmount in (*nsui)->nsui_vfsmount */
1369                 return 0;
1370         }
1371         if (work) {
1372                 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1373                 refcount_set(&work->nsui_refcnt, 2);
1374                 work->nsui_busy = true;
1375                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1376                 *nsui = work;
1377         } else
1378                 status = nfserr_resource;
1379         spin_unlock(&nn->nfsd_ssc_lock);
1380         return status;
1381 }
1382
1383 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1384                                  struct nfsd4_ssc_umount_item *nsui,
1385                                  struct vfsmount *ss_mnt)
1386 {
1387         spin_lock(&nn->nfsd_ssc_lock);
1388         nsui->nsui_vfsmount = ss_mnt;
1389         nsui->nsui_busy = false;
1390         wake_up_all(&nn->nfsd_ssc_waitq);
1391         spin_unlock(&nn->nfsd_ssc_lock);
1392 }
1393
1394 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1395                                  struct nfsd4_ssc_umount_item *nsui)
1396 {
1397         spin_lock(&nn->nfsd_ssc_lock);
1398         list_del(&nsui->nsui_list);
1399         wake_up_all(&nn->nfsd_ssc_waitq);
1400         spin_unlock(&nn->nfsd_ssc_lock);
1401         kfree(nsui);
1402 }
1403
1404 /*
1405  * Support one copy source server for now.
1406  */
1407 static __be32
1408 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1409                        struct nfsd4_ssc_umount_item **nsui)
1410 {
1411         struct file_system_type *type;
1412         struct vfsmount *ss_mnt;
1413         struct nfs42_netaddr *naddr;
1414         struct sockaddr_storage tmp_addr;
1415         size_t tmp_addrlen, match_netid_len = 3;
1416         char *startsep = "", *endsep = "", *match_netid = "tcp";
1417         char *ipaddr, *dev_name, *raw_data;
1418         int len, raw_len;
1419         __be32 status = nfserr_inval;
1420         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1421
1422         naddr = &nss->u.nl4_addr;
1423         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1424                                          naddr->addr_len,
1425                                          (struct sockaddr *)&tmp_addr,
1426                                          sizeof(tmp_addr));
1427         *nsui = NULL;
1428         if (tmp_addrlen == 0)
1429                 goto out_err;
1430
1431         if (tmp_addr.ss_family == AF_INET6) {
1432                 startsep = "[";
1433                 endsep = "]";
1434                 match_netid = "tcp6";
1435                 match_netid_len = 4;
1436         }
1437
1438         if (naddr->netid_len != match_netid_len ||
1439                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1440                 goto out_err;
1441
1442         /* Construct the raw data for the vfs_kern_mount call */
1443         len = RPC_MAX_ADDRBUFLEN + 1;
1444         ipaddr = kzalloc(len, GFP_KERNEL);
1445         if (!ipaddr)
1446                 goto out_err;
1447
1448         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1449
1450         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1451
1452         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1453         raw_data = kzalloc(raw_len, GFP_KERNEL);
1454         if (!raw_data)
1455                 goto out_free_ipaddr;
1456
1457         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1458
1459         status = nfserr_nodev;
1460         type = get_fs_type("nfs");
1461         if (!type)
1462                 goto out_free_rawdata;
1463
1464         /* Set the server:<export> for the vfs_kern_mount call */
1465         dev_name = kzalloc(len + 5, GFP_KERNEL);
1466         if (!dev_name)
1467                 goto out_free_rawdata;
1468         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1469
1470         status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui);
1471         if (status)
1472                 goto out_free_devname;
1473         if ((*nsui)->nsui_vfsmount)
1474                 goto out_done;
1475
1476         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1477         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1478         module_put(type->owner);
1479         if (IS_ERR(ss_mnt)) {
1480                 status = nfserr_nodev;
1481                 nfsd4_ssc_cancel_dul(nn, *nsui);
1482                 goto out_free_devname;
1483         }
1484         nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1485 out_done:
1486         status = 0;
1487
1488 out_free_devname:
1489         kfree(dev_name);
1490 out_free_rawdata:
1491         kfree(raw_data);
1492 out_free_ipaddr:
1493         kfree(ipaddr);
1494 out_err:
1495         return status;
1496 }
1497
1498 /*
1499  * Verify COPY destination stateid.
1500  *
1501  * Connect to the source server with NFSv4.1.
1502  * Create the source struct file for nfsd_copy_range.
1503  * Called with COPY cstate:
1504  *    SAVED_FH: source filehandle
1505  *    CURRENT_FH: destination filehandle
1506  */
1507 static __be32
1508 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1509                       struct nfsd4_compound_state *cstate,
1510                       struct nfsd4_copy *copy)
1511 {
1512         struct svc_fh *s_fh = NULL;
1513         stateid_t *s_stid = &copy->cp_src_stateid;
1514         __be32 status = nfserr_inval;
1515
1516         /* Verify the destination stateid and set dst struct file*/
1517         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1518                                             &copy->cp_dst_stateid,
1519                                             WR_STATE, &copy->nf_dst, NULL);
1520         if (status)
1521                 goto out;
1522
1523         status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1524         if (status)
1525                 goto out;
1526
1527         s_fh = &cstate->save_fh;
1528
1529         copy->c_fh.size = s_fh->fh_handle.fh_size;
1530         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1531         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1532         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1533                sizeof(stateid_opaque_t));
1534
1535         status = 0;
1536 out:
1537         return status;
1538 }
1539
1540 static void
1541 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1542                         struct nfsd_file *dst)
1543 {
1544         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1545         long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1546
1547         nfs42_ssc_close(filp);
1548         fput(filp);
1549
1550         spin_lock(&nn->nfsd_ssc_lock);
1551         list_del(&nsui->nsui_list);
1552         /*
1553          * vfsmount can be shared by multiple exports,
1554          * decrement refcnt. If the count drops to 1 it
1555          * will be unmounted when nsui_expire expires.
1556          */
1557         refcount_dec(&nsui->nsui_refcnt);
1558         nsui->nsui_expire = jiffies + timeout;
1559         list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1560         spin_unlock(&nn->nfsd_ssc_lock);
1561 }
1562
1563 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1564
1565 static __be32
1566 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1567                       struct nfsd4_compound_state *cstate,
1568                       struct nfsd4_copy *copy)
1569 {
1570         return nfserr_inval;
1571 }
1572
1573 static void
1574 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1575                         struct nfsd_file *dst)
1576 {
1577 }
1578
1579 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1580                                    struct nfs_fh *src_fh,
1581                                    nfs4_stateid *stateid)
1582 {
1583         return NULL;
1584 }
1585 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1586
1587 static __be32
1588 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1589                       struct nfsd4_compound_state *cstate,
1590                       struct nfsd4_copy *copy)
1591 {
1592         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1593                                  &copy->nf_src, &copy->cp_dst_stateid,
1594                                  &copy->nf_dst);
1595 }
1596
1597 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1598 {
1599         struct nfsd4_cb_offload *cbo =
1600                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1601
1602         kfree(cbo);
1603 }
1604
1605 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1606                                  struct rpc_task *task)
1607 {
1608         struct nfsd4_cb_offload *cbo =
1609                 container_of(cb, struct nfsd4_cb_offload, co_cb);
1610
1611         trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1612         return 1;
1613 }
1614
1615 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1616         .release = nfsd4_cb_offload_release,
1617         .done = nfsd4_cb_offload_done
1618 };
1619
1620 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1621 {
1622         copy->cp_res.wr_stable_how =
1623                 test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1624                         NFS_FILE_SYNC : NFS_UNSTABLE;
1625         nfsd4_copy_set_sync(copy, sync);
1626         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1627 }
1628
1629 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1630                                      struct file *dst,
1631                                      struct file *src)
1632 {
1633         errseq_t since;
1634         ssize_t bytes_copied = 0;
1635         u64 bytes_total = copy->cp_count;
1636         u64 src_pos = copy->cp_src_pos;
1637         u64 dst_pos = copy->cp_dst_pos;
1638         int status;
1639         loff_t end;
1640
1641         /* See RFC 7862 p.67: */
1642         if (bytes_total == 0)
1643                 bytes_total = ULLONG_MAX;
1644         do {
1645                 if (kthread_should_stop())
1646                         break;
1647                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1648                                                     bytes_total);
1649                 if (bytes_copied <= 0)
1650                         break;
1651                 bytes_total -= bytes_copied;
1652                 copy->cp_res.wr_bytes_written += bytes_copied;
1653                 src_pos += bytes_copied;
1654                 dst_pos += bytes_copied;
1655         } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1656         /* for a non-zero asynchronous copy do a commit of data */
1657         if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1658                 since = READ_ONCE(dst->f_wb_err);
1659                 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1660                 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1661                 if (!status)
1662                         status = filemap_check_wb_err(dst->f_mapping, since);
1663                 if (!status)
1664                         set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1665         }
1666         return bytes_copied;
1667 }
1668
1669 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1670                             struct file *src, struct file *dst,
1671                             bool sync)
1672 {
1673         __be32 status;
1674         ssize_t bytes;
1675
1676         bytes = _nfsd_copy_file_range(copy, dst, src);
1677
1678         /* for async copy, we ignore the error, client can always retry
1679          * to get the error
1680          */
1681         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1682                 status = nfserrno(bytes);
1683         else {
1684                 nfsd4_init_copy_res(copy, sync);
1685                 status = nfs_ok;
1686         }
1687         return status;
1688 }
1689
1690 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1691 {
1692         dst->cp_src_pos = src->cp_src_pos;
1693         dst->cp_dst_pos = src->cp_dst_pos;
1694         dst->cp_count = src->cp_count;
1695         dst->cp_flags = src->cp_flags;
1696         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1697         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1698         dst->cp_clp = src->cp_clp;
1699         dst->nf_dst = nfsd_file_get(src->nf_dst);
1700         /* for inter, nf_src doesn't exist yet */
1701         if (!nfsd4_ssc_is_inter(src))
1702                 dst->nf_src = nfsd_file_get(src->nf_src);
1703
1704         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1705         memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1706         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1707         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1708         dst->ss_nsui = src->ss_nsui;
1709 }
1710
1711 static void release_copy_files(struct nfsd4_copy *copy)
1712 {
1713         if (copy->nf_src)
1714                 nfsd_file_put(copy->nf_src);
1715         if (copy->nf_dst)
1716                 nfsd_file_put(copy->nf_dst);
1717 }
1718
1719 static void cleanup_async_copy(struct nfsd4_copy *copy)
1720 {
1721         nfs4_free_copy_state(copy);
1722         release_copy_files(copy);
1723         if (copy->cp_clp) {
1724                 spin_lock(&copy->cp_clp->async_lock);
1725                 if (!list_empty(&copy->copies))
1726                         list_del_init(&copy->copies);
1727                 spin_unlock(&copy->cp_clp->async_lock);
1728         }
1729         nfs4_put_copy(copy);
1730 }
1731
1732 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1733 {
1734         struct nfsd4_cb_offload *cbo;
1735
1736         cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1737         if (!cbo)
1738                 return;
1739
1740         memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1741         memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1742         cbo->co_nfserr = nfserr;
1743
1744         nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1745                       NFSPROC4_CLNT_CB_OFFLOAD);
1746         trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1747                               &cbo->co_fh, copy->cp_count, nfserr);
1748         nfsd4_run_cb(&cbo->co_cb);
1749 }
1750
1751 /**
1752  * nfsd4_do_async_copy - kthread function for background server-side COPY
1753  * @data: arguments for COPY operation
1754  *
1755  * Return values:
1756  *   %0: Copy operation is done.
1757  */
1758 static int nfsd4_do_async_copy(void *data)
1759 {
1760         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1761         __be32 nfserr;
1762
1763         if (nfsd4_ssc_is_inter(copy)) {
1764                 struct file *filp;
1765
1766                 filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1767                                       &copy->c_fh, &copy->stateid);
1768                 if (IS_ERR(filp)) {
1769                         switch (PTR_ERR(filp)) {
1770                         case -EBADF:
1771                                 nfserr = nfserr_wrong_type;
1772                                 break;
1773                         default:
1774                                 nfserr = nfserr_offload_denied;
1775                         }
1776                         /* ss_mnt will be unmounted by the laundromat */
1777                         goto do_callback;
1778                 }
1779                 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1780                                        false);
1781                 nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1782         } else {
1783                 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1784                                        copy->nf_dst->nf_file, false);
1785         }
1786
1787 do_callback:
1788         nfsd4_send_cb_offload(copy, nfserr);
1789         cleanup_async_copy(copy);
1790         return 0;
1791 }
1792
1793 static __be32
1794 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1795                 union nfsd4_op_u *u)
1796 {
1797         struct nfsd4_copy *copy = &u->copy;
1798         __be32 status;
1799         struct nfsd4_copy *async_copy = NULL;
1800
1801         if (nfsd4_ssc_is_inter(copy)) {
1802                 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1803                         status = nfserr_notsupp;
1804                         goto out;
1805                 }
1806                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1807                 if (status)
1808                         return nfserr_offload_denied;
1809         } else {
1810                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1811                 if (status)
1812                         return status;
1813         }
1814
1815         copy->cp_clp = cstate->clp;
1816         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1817                 sizeof(struct knfsd_fh));
1818         if (nfsd4_copy_is_async(copy)) {
1819                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1820
1821                 status = nfserrno(-ENOMEM);
1822                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1823                 if (!async_copy)
1824                         goto out_err;
1825                 INIT_LIST_HEAD(&async_copy->copies);
1826                 refcount_set(&async_copy->refcount, 1);
1827                 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1828                 if (!async_copy->cp_src)
1829                         goto out_err;
1830                 if (!nfs4_init_copy_state(nn, copy))
1831                         goto out_err;
1832                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1833                         sizeof(copy->cp_res.cb_stateid));
1834                 dup_copy_fields(copy, async_copy);
1835                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1836                                 async_copy, "%s", "copy thread");
1837                 if (IS_ERR(async_copy->copy_task))
1838                         goto out_err;
1839                 spin_lock(&async_copy->cp_clp->async_lock);
1840                 list_add(&async_copy->copies,
1841                                 &async_copy->cp_clp->async_copies);
1842                 spin_unlock(&async_copy->cp_clp->async_lock);
1843                 wake_up_process(async_copy->copy_task);
1844                 status = nfs_ok;
1845         } else {
1846                 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1847                                        copy->nf_dst->nf_file, true);
1848         }
1849 out:
1850         release_copy_files(copy);
1851         return status;
1852 out_err:
1853         if (nfsd4_ssc_is_inter(copy)) {
1854                 /*
1855                  * Source's vfsmount of inter-copy will be unmounted
1856                  * by the laundromat. Use copy instead of async_copy
1857                  * since async_copy->ss_nsui might not be set yet.
1858                  */
1859                 refcount_dec(&copy->ss_nsui->nsui_refcnt);
1860         }
1861         if (async_copy)
1862                 cleanup_async_copy(async_copy);
1863         status = nfserrno(-ENOMEM);
1864         goto out;
1865 }
1866
1867 static struct nfsd4_copy *
1868 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1869 {
1870         struct nfsd4_copy *copy;
1871
1872         lockdep_assert_held(&clp->async_lock);
1873
1874         list_for_each_entry(copy, &clp->async_copies, copies) {
1875                 if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1876                         continue;
1877                 return copy;
1878         }
1879         return NULL;
1880 }
1881
1882 static struct nfsd4_copy *
1883 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1884 {
1885         struct nfsd4_copy *copy;
1886
1887         spin_lock(&clp->async_lock);
1888         copy = find_async_copy_locked(clp, stateid);
1889         if (copy)
1890                 refcount_inc(&copy->refcount);
1891         spin_unlock(&clp->async_lock);
1892         return copy;
1893 }
1894
1895 static __be32
1896 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1897                      struct nfsd4_compound_state *cstate,
1898                      union nfsd4_op_u *u)
1899 {
1900         struct nfsd4_offload_status *os = &u->offload_status;
1901         struct nfsd4_copy *copy;
1902         struct nfs4_client *clp = cstate->clp;
1903
1904         copy = find_async_copy(clp, &os->stateid);
1905         if (!copy) {
1906                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1907
1908                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1909         } else
1910                 nfsd4_stop_copy(copy);
1911
1912         return nfs_ok;
1913 }
1914
1915 static __be32
1916 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1917                   union nfsd4_op_u *u)
1918 {
1919         struct nfsd4_copy_notify *cn = &u->copy_notify;
1920         __be32 status;
1921         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1922         struct nfs4_stid *stid;
1923         struct nfs4_cpntf_state *cps;
1924         struct nfs4_client *clp = cstate->clp;
1925
1926         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1927                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1928                                         &stid);
1929         if (status)
1930                 return status;
1931
1932         cn->cpn_sec = nn->nfsd4_lease;
1933         cn->cpn_nsec = 0;
1934
1935         status = nfserrno(-ENOMEM);
1936         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1937         if (!cps)
1938                 goto out;
1939         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1940         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1941         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1942
1943         /* For now, only return one server address in cpn_src, the
1944          * address used by the client to connect to this server.
1945          */
1946         cn->cpn_src->nl4_type = NL4_NETADDR;
1947         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1948                                  &cn->cpn_src->u.nl4_addr);
1949         WARN_ON_ONCE(status);
1950         if (status) {
1951                 nfs4_put_cpntf_state(nn, cps);
1952                 goto out;
1953         }
1954 out:
1955         nfs4_put_stid(stid);
1956         return status;
1957 }
1958
1959 static __be32
1960 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1961                 struct nfsd4_fallocate *fallocate, int flags)
1962 {
1963         __be32 status;
1964         struct nfsd_file *nf;
1965
1966         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1967                                             &fallocate->falloc_stateid,
1968                                             WR_STATE, &nf, NULL);
1969         if (status != nfs_ok)
1970                 return status;
1971
1972         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1973                                      fallocate->falloc_offset,
1974                                      fallocate->falloc_length,
1975                                      flags);
1976         nfsd_file_put(nf);
1977         return status;
1978 }
1979
1980 static __be32
1981 nfsd4_offload_status(struct svc_rqst *rqstp,
1982                      struct nfsd4_compound_state *cstate,
1983                      union nfsd4_op_u *u)
1984 {
1985         struct nfsd4_offload_status *os = &u->offload_status;
1986         __be32 status = nfs_ok;
1987         struct nfsd4_copy *copy;
1988         struct nfs4_client *clp = cstate->clp;
1989
1990         spin_lock(&clp->async_lock);
1991         copy = find_async_copy_locked(clp, &os->stateid);
1992         if (copy)
1993                 os->count = copy->cp_res.wr_bytes_written;
1994         else
1995                 status = nfserr_bad_stateid;
1996         spin_unlock(&clp->async_lock);
1997
1998         return status;
1999 }
2000
2001 static __be32
2002 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2003                union nfsd4_op_u *u)
2004 {
2005         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2006 }
2007
2008 static __be32
2009 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2010                  union nfsd4_op_u *u)
2011 {
2012         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2013                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2014 }
2015
2016 static __be32
2017 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2018            union nfsd4_op_u *u)
2019 {
2020         struct nfsd4_seek *seek = &u->seek;
2021         int whence;
2022         __be32 status;
2023         struct nfsd_file *nf;
2024
2025         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2026                                             &seek->seek_stateid,
2027                                             RD_STATE, &nf, NULL);
2028         if (status)
2029                 return status;
2030
2031         switch (seek->seek_whence) {
2032         case NFS4_CONTENT_DATA:
2033                 whence = SEEK_DATA;
2034                 break;
2035         case NFS4_CONTENT_HOLE:
2036                 whence = SEEK_HOLE;
2037                 break;
2038         default:
2039                 status = nfserr_union_notsupp;
2040                 goto out;
2041         }
2042
2043         /*
2044          * Note:  This call does change file->f_pos, but nothing in NFSD
2045          *        should ever file->f_pos.
2046          */
2047         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2048         if (seek->seek_pos < 0)
2049                 status = nfserrno(seek->seek_pos);
2050         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2051                 seek->seek_eof = true;
2052
2053 out:
2054         nfsd_file_put(nf);
2055         return status;
2056 }
2057
2058 /* This routine never returns NFS_OK!  If there are no other errors, it
2059  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2060  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2061  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2062  */
2063 static __be32
2064 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2065              struct nfsd4_verify *verify)
2066 {
2067         __be32 *buf, *p;
2068         int count;
2069         __be32 status;
2070
2071         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2072         if (status)
2073                 return status;
2074
2075         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2076         if (status)
2077                 return status;
2078
2079         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2080             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2081                 return nfserr_inval;
2082         if (verify->ve_attrlen & 3)
2083                 return nfserr_inval;
2084
2085         /* count in words:
2086          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2087          */
2088         count = 4 + (verify->ve_attrlen >> 2);
2089         buf = kmalloc(count << 2, GFP_KERNEL);
2090         if (!buf)
2091                 return nfserr_jukebox;
2092
2093         p = buf;
2094         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2095                                     cstate->current_fh.fh_export,
2096                                     cstate->current_fh.fh_dentry,
2097                                     verify->ve_bmval,
2098                                     rqstp, 0);
2099         /*
2100          * If nfsd4_encode_fattr() ran out of space, assume that's because
2101          * the attributes are longer (hence different) than those given:
2102          */
2103         if (status == nfserr_resource)
2104                 status = nfserr_not_same;
2105         if (status)
2106                 goto out_kfree;
2107
2108         /* skip bitmap */
2109         p = buf + 1 + ntohl(buf[0]);
2110         status = nfserr_not_same;
2111         if (ntohl(*p++) != verify->ve_attrlen)
2112                 goto out_kfree;
2113         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2114                 status = nfserr_same;
2115
2116 out_kfree:
2117         kfree(buf);
2118         return status;
2119 }
2120
2121 static __be32
2122 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2123               union nfsd4_op_u *u)
2124 {
2125         __be32 status;
2126
2127         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2128         return status == nfserr_not_same ? nfs_ok : status;
2129 }
2130
2131 static __be32
2132 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2133              union nfsd4_op_u *u)
2134 {
2135         __be32 status;
2136
2137         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2138         return status == nfserr_same ? nfs_ok : status;
2139 }
2140
2141 #ifdef CONFIG_NFSD_PNFS
2142 static const struct nfsd4_layout_ops *
2143 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2144 {
2145         if (!exp->ex_layout_types) {
2146                 dprintk("%s: export does not support pNFS\n", __func__);
2147                 return NULL;
2148         }
2149
2150         if (layout_type >= LAYOUT_TYPE_MAX ||
2151             !(exp->ex_layout_types & (1 << layout_type))) {
2152                 dprintk("%s: layout type %d not supported\n",
2153                         __func__, layout_type);
2154                 return NULL;
2155         }
2156
2157         return nfsd4_layout_ops[layout_type];
2158 }
2159
2160 static __be32
2161 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2162                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2163 {
2164         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2165         const struct nfsd4_layout_ops *ops;
2166         struct nfsd4_deviceid_map *map;
2167         struct svc_export *exp;
2168         __be32 nfserr;
2169
2170         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2171                __func__,
2172                gdp->gd_layout_type,
2173                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2174                gdp->gd_maxcount);
2175
2176         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2177         if (!map) {
2178                 dprintk("%s: couldn't find device ID to export mapping!\n",
2179                         __func__);
2180                 return nfserr_noent;
2181         }
2182
2183         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2184         if (IS_ERR(exp)) {
2185                 dprintk("%s: could not find device id\n", __func__);
2186                 return nfserr_noent;
2187         }
2188
2189         nfserr = nfserr_layoutunavailable;
2190         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2191         if (!ops)
2192                 goto out;
2193
2194         nfserr = nfs_ok;
2195         if (gdp->gd_maxcount != 0) {
2196                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2197                                 rqstp, cstate->clp, gdp);
2198         }
2199
2200         gdp->gd_notify_types &= ops->notify_types;
2201 out:
2202         exp_put(exp);
2203         return nfserr;
2204 }
2205
2206 static void
2207 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2208 {
2209         kfree(u->getdeviceinfo.gd_device);
2210 }
2211
2212 static __be32
2213 nfsd4_layoutget(struct svc_rqst *rqstp,
2214                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2215 {
2216         struct nfsd4_layoutget *lgp = &u->layoutget;
2217         struct svc_fh *current_fh = &cstate->current_fh;
2218         const struct nfsd4_layout_ops *ops;
2219         struct nfs4_layout_stateid *ls;
2220         __be32 nfserr;
2221         int accmode = NFSD_MAY_READ_IF_EXEC;
2222
2223         switch (lgp->lg_seg.iomode) {
2224         case IOMODE_READ:
2225                 accmode |= NFSD_MAY_READ;
2226                 break;
2227         case IOMODE_RW:
2228                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2229                 break;
2230         default:
2231                 dprintk("%s: invalid iomode %d\n",
2232                         __func__, lgp->lg_seg.iomode);
2233                 nfserr = nfserr_badiomode;
2234                 goto out;
2235         }
2236
2237         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2238         if (nfserr)
2239                 goto out;
2240
2241         nfserr = nfserr_layoutunavailable;
2242         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2243         if (!ops)
2244                 goto out;
2245
2246         /*
2247          * Verify minlength and range as per RFC5661:
2248          *  o  If loga_length is less than loga_minlength,
2249          *     the metadata server MUST return NFS4ERR_INVAL.
2250          *  o  If the sum of loga_offset and loga_minlength exceeds
2251          *     NFS4_UINT64_MAX, and loga_minlength is not
2252          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2253          *  o  If the sum of loga_offset and loga_length exceeds
2254          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2255          *     the error NFS4ERR_INVAL MUST result.
2256          */
2257         nfserr = nfserr_inval;
2258         if (lgp->lg_seg.length < lgp->lg_minlength ||
2259             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2260              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2261             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2262              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2263                 goto out;
2264         if (lgp->lg_seg.length == 0)
2265                 goto out;
2266
2267         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2268                                                 true, lgp->lg_layout_type, &ls);
2269         if (nfserr) {
2270                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2271                 goto out;
2272         }
2273
2274         nfserr = nfserr_recallconflict;
2275         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2276                 goto out_put_stid;
2277
2278         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2279                                      current_fh, lgp);
2280         if (nfserr)
2281                 goto out_put_stid;
2282
2283         nfserr = nfsd4_insert_layout(lgp, ls);
2284
2285 out_put_stid:
2286         mutex_unlock(&ls->ls_mutex);
2287         nfs4_put_stid(&ls->ls_stid);
2288 out:
2289         return nfserr;
2290 }
2291
2292 static void
2293 nfsd4_layoutget_release(union nfsd4_op_u *u)
2294 {
2295         kfree(u->layoutget.lg_content);
2296 }
2297
2298 static __be32
2299 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2300                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2301 {
2302         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2303         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2304         struct svc_fh *current_fh = &cstate->current_fh;
2305         const struct nfsd4_layout_ops *ops;
2306         loff_t new_size = lcp->lc_last_wr + 1;
2307         struct inode *inode;
2308         struct nfs4_layout_stateid *ls;
2309         __be32 nfserr;
2310
2311         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2312         if (nfserr)
2313                 goto out;
2314
2315         nfserr = nfserr_layoutunavailable;
2316         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2317         if (!ops)
2318                 goto out;
2319         inode = d_inode(current_fh->fh_dentry);
2320
2321         nfserr = nfserr_inval;
2322         if (new_size <= seg->offset) {
2323                 dprintk("pnfsd: last write before layout segment\n");
2324                 goto out;
2325         }
2326         if (new_size > seg->offset + seg->length) {
2327                 dprintk("pnfsd: last write beyond layout segment\n");
2328                 goto out;
2329         }
2330         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2331                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2332                 goto out;
2333         }
2334
2335         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2336                                                 false, lcp->lc_layout_type,
2337                                                 &ls);
2338         if (nfserr) {
2339                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2340                 /* fixup error code as per RFC5661 */
2341                 if (nfserr == nfserr_bad_stateid)
2342                         nfserr = nfserr_badlayout;
2343                 goto out;
2344         }
2345
2346         /* LAYOUTCOMMIT does not require any serialization */
2347         mutex_unlock(&ls->ls_mutex);
2348
2349         if (new_size > i_size_read(inode)) {
2350                 lcp->lc_size_chg = 1;
2351                 lcp->lc_newsize = new_size;
2352         } else {
2353                 lcp->lc_size_chg = 0;
2354         }
2355
2356         nfserr = ops->proc_layoutcommit(inode, lcp);
2357         nfs4_put_stid(&ls->ls_stid);
2358 out:
2359         return nfserr;
2360 }
2361
2362 static __be32
2363 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2364                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2365 {
2366         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2367         struct svc_fh *current_fh = &cstate->current_fh;
2368         __be32 nfserr;
2369
2370         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2371         if (nfserr)
2372                 goto out;
2373
2374         nfserr = nfserr_layoutunavailable;
2375         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2376                 goto out;
2377
2378         switch (lrp->lr_seg.iomode) {
2379         case IOMODE_READ:
2380         case IOMODE_RW:
2381         case IOMODE_ANY:
2382                 break;
2383         default:
2384                 dprintk("%s: invalid iomode %d\n", __func__,
2385                         lrp->lr_seg.iomode);
2386                 nfserr = nfserr_inval;
2387                 goto out;
2388         }
2389
2390         switch (lrp->lr_return_type) {
2391         case RETURN_FILE:
2392                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2393                 break;
2394         case RETURN_FSID:
2395         case RETURN_ALL:
2396                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2397                 break;
2398         default:
2399                 dprintk("%s: invalid return_type %d\n", __func__,
2400                         lrp->lr_return_type);
2401                 nfserr = nfserr_inval;
2402                 break;
2403         }
2404 out:
2405         return nfserr;
2406 }
2407 #endif /* CONFIG_NFSD_PNFS */
2408
2409 static __be32
2410 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2411                union nfsd4_op_u *u)
2412 {
2413         struct nfsd4_getxattr *getxattr = &u->getxattr;
2414
2415         return nfsd_getxattr(rqstp, &cstate->current_fh,
2416                              getxattr->getxa_name, &getxattr->getxa_buf,
2417                              &getxattr->getxa_len);
2418 }
2419
2420 static __be32
2421 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2422            union nfsd4_op_u *u)
2423 {
2424         struct nfsd4_setxattr *setxattr = &u->setxattr;
2425         __be32 ret;
2426
2427         if (opens_in_grace(SVC_NET(rqstp)))
2428                 return nfserr_grace;
2429
2430         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2431                             setxattr->setxa_buf, setxattr->setxa_len,
2432                             setxattr->setxa_flags);
2433
2434         if (!ret)
2435                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2436
2437         return ret;
2438 }
2439
2440 static __be32
2441 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2442            union nfsd4_op_u *u)
2443 {
2444         /*
2445          * Get the entire list, then copy out only the user attributes
2446          * in the encode function.
2447          */
2448         return nfsd_listxattr(rqstp, &cstate->current_fh,
2449                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2450 }
2451
2452 static __be32
2453 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2454            union nfsd4_op_u *u)
2455 {
2456         struct nfsd4_removexattr *removexattr = &u->removexattr;
2457         __be32 ret;
2458
2459         if (opens_in_grace(SVC_NET(rqstp)))
2460                 return nfserr_grace;
2461
2462         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2463             removexattr->rmxa_name);
2464
2465         if (!ret)
2466                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2467
2468         return ret;
2469 }
2470
2471 /*
2472  * NULL call.
2473  */
2474 static __be32
2475 nfsd4_proc_null(struct svc_rqst *rqstp)
2476 {
2477         return rpc_success;
2478 }
2479
2480 static inline void nfsd4_increment_op_stats(u32 opnum)
2481 {
2482         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2483                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2484 }
2485
2486 static const struct nfsd4_operation nfsd4_ops[];
2487
2488 static const char *nfsd4_op_name(unsigned opnum);
2489
2490 /*
2491  * Enforce NFSv4.1 COMPOUND ordering rules:
2492  *
2493  * Also note, enforced elsewhere:
2494  *      - SEQUENCE other than as first op results in
2495  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2496  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2497  *        (Enforced in nfsd4_bind_conn_to_session().)
2498  *      - DESTROY_SESSION must be the final operation in a compound, if
2499  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2500  *        (Enforced in nfsd4_destroy_session().)
2501  */
2502 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2503 {
2504         struct nfsd4_op *first_op = &args->ops[0];
2505
2506         /* These ordering requirements don't apply to NFSv4.0: */
2507         if (args->minorversion == 0)
2508                 return nfs_ok;
2509         /* This is weird, but OK, not our problem: */
2510         if (args->opcnt == 0)
2511                 return nfs_ok;
2512         if (first_op->status == nfserr_op_illegal)
2513                 return nfs_ok;
2514         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2515                 return nfserr_op_not_in_session;
2516         if (first_op->opnum == OP_SEQUENCE)
2517                 return nfs_ok;
2518         /*
2519          * So first_op is something allowed outside a session, like
2520          * EXCHANGE_ID; but then it has to be the only op in the
2521          * compound:
2522          */
2523         if (args->opcnt != 1)
2524                 return nfserr_not_only_op;
2525         return nfs_ok;
2526 }
2527
2528 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2529 {
2530         return &nfsd4_ops[op->opnum];
2531 }
2532
2533 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2534 {
2535         if (op->opnum == OP_ILLEGAL)
2536                 return false;
2537         return OPDESC(op)->op_flags & OP_CACHEME;
2538 }
2539
2540 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2541 {
2542         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2543         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2544         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2545         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2546         const struct nfsd4_operation *thisd = OPDESC(this);
2547         const struct nfsd4_operation *nextd;
2548
2549         /*
2550          * Most ops check wronsec on our own; only the putfh-like ops
2551          * have special rules.
2552          */
2553         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2554                 return false;
2555         /*
2556          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2557          * put-filehandle operation if we're not going to use the
2558          * result:
2559          */
2560         if (argp->opcnt == resp->opcnt)
2561                 return false;
2562         if (next->opnum == OP_ILLEGAL)
2563                 return false;
2564         nextd = OPDESC(next);
2565         /*
2566          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2567          * errors themselves as necessary; others should check for them
2568          * now:
2569          */
2570         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2571 }
2572
2573 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2574 static void
2575 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2576 {
2577         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2578         struct nfsd4_copy *copy;
2579         struct nfsd4_putfh *putfh;
2580         int i;
2581
2582         /* traverse all operation and if it's a COPY compound, mark the
2583          * source filehandle to skip verification
2584          */
2585         for (i = 0; i < args->opcnt; i++) {
2586                 op = &args->ops[i];
2587                 if (op->opnum == OP_PUTFH)
2588                         current_op = op;
2589                 else if (op->opnum == OP_SAVEFH)
2590                         saved_op = current_op;
2591                 else if (op->opnum == OP_RESTOREFH)
2592                         current_op = saved_op;
2593                 else if (op->opnum == OP_COPY) {
2594                         copy = (struct nfsd4_copy *)&op->u;
2595                         if (!saved_op) {
2596                                 op->status = nfserr_nofilehandle;
2597                                 return;
2598                         }
2599                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2600                         if (nfsd4_ssc_is_inter(copy))
2601                                 putfh->no_verify = true;
2602                 }
2603         }
2604 }
2605 #else
2606 static void
2607 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2608 {
2609 }
2610 #endif
2611
2612 /*
2613  * COMPOUND call.
2614  */
2615 static __be32
2616 nfsd4_proc_compound(struct svc_rqst *rqstp)
2617 {
2618         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2619         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2620         struct nfsd4_op *op;
2621         struct nfsd4_compound_state *cstate = &resp->cstate;
2622         struct svc_fh *current_fh = &cstate->current_fh;
2623         struct svc_fh *save_fh = &cstate->save_fh;
2624         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2625         __be32          status;
2626
2627         resp->xdr = &rqstp->rq_res_stream;
2628         resp->statusp = resp->xdr->p;
2629
2630         /* reserve space for: NFS status code */
2631         xdr_reserve_space(resp->xdr, XDR_UNIT);
2632
2633         /* reserve space for: taglen, tag, and opcnt */
2634         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2635         resp->taglen = args->taglen;
2636         resp->tag = args->tag;
2637         resp->rqstp = rqstp;
2638         cstate->minorversion = args->minorversion;
2639         fh_init(current_fh, NFS4_FHSIZE);
2640         fh_init(save_fh, NFS4_FHSIZE);
2641         /*
2642          * Don't use the deferral mechanism for NFSv4; compounds make it
2643          * too hard to avoid non-idempotency problems.
2644          */
2645         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2646
2647         /*
2648          * According to RFC3010, this takes precedence over all other errors.
2649          */
2650         status = nfserr_minor_vers_mismatch;
2651         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2652                 goto out;
2653
2654         status = nfs41_check_op_ordering(args);
2655         if (status) {
2656                 op = &args->ops[0];
2657                 op->status = status;
2658                 resp->opcnt = 1;
2659                 goto encode_op;
2660         }
2661         check_if_stalefh_allowed(args);
2662
2663         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2664
2665         trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2666         while (!status && resp->opcnt < args->opcnt) {
2667                 op = &args->ops[resp->opcnt++];
2668
2669                 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2670                         /* If there are still more operations to process,
2671                          * stop here and report NFS4ERR_RESOURCE. */
2672                         if (cstate->minorversion == 0 &&
2673                             args->client_opcnt > resp->opcnt) {
2674                                 op->status = nfserr_resource;
2675                                 goto encode_op;
2676                         }
2677                 }
2678
2679                 /*
2680                  * The XDR decode routines may have pre-set op->status;
2681                  * for example, if there is a miscellaneous XDR error
2682                  * it will be set to nfserr_bad_xdr.
2683                  */
2684                 if (op->status) {
2685                         if (op->opnum == OP_OPEN)
2686                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2687                         goto encode_op;
2688                 }
2689                 if (!current_fh->fh_dentry &&
2690                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2691                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2692                                 op->status = nfserr_nofilehandle;
2693                                 goto encode_op;
2694                         }
2695                 } else if (current_fh->fh_export &&
2696                            current_fh->fh_export->ex_fslocs.migrated &&
2697                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2698                         op->status = nfserr_moved;
2699                         goto encode_op;
2700                 }
2701
2702                 fh_clear_pre_post_attrs(current_fh);
2703
2704                 /* If op is non-idempotent */
2705                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2706                         /*
2707                          * Don't execute this op if we couldn't encode a
2708                          * successful reply:
2709                          */
2710                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2711                         /*
2712                          * Plus if there's another operation, make sure
2713                          * we'll have space to at least encode an error:
2714                          */
2715                         if (resp->opcnt < args->opcnt)
2716                                 plen += COMPOUND_ERR_SLACK_SPACE;
2717                         op->status = nfsd4_check_resp_size(resp, plen);
2718                 }
2719
2720                 if (op->status)
2721                         goto encode_op;
2722
2723                 if (op->opdesc->op_get_currentstateid)
2724                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2725                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2726
2727                 /* Only from SEQUENCE */
2728                 if (cstate->status == nfserr_replay_cache) {
2729                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2730                         status = op->status;
2731                         goto out;
2732                 }
2733                 if (!op->status) {
2734                         if (op->opdesc->op_set_currentstateid)
2735                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2736
2737                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2738                                 clear_current_stateid(cstate);
2739
2740                         if (current_fh->fh_export &&
2741                                         need_wrongsec_check(rqstp))
2742                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2743                 }
2744 encode_op:
2745                 if (op->status == nfserr_replay_me) {
2746                         op->replay = &cstate->replay_owner->so_replay;
2747                         nfsd4_encode_replay(resp->xdr, op);
2748                         status = op->status = op->replay->rp_status;
2749                 } else {
2750                         nfsd4_encode_operation(resp, op);
2751                         status = op->status;
2752                 }
2753
2754                 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2755                                            status, nfsd4_op_name(op->opnum));
2756
2757                 nfsd4_cstate_clear_replay(cstate);
2758                 nfsd4_increment_op_stats(op->opnum);
2759         }
2760
2761         fh_put(current_fh);
2762         fh_put(save_fh);
2763         BUG_ON(cstate->replay_owner);
2764 out:
2765         cstate->status = status;
2766         /* Reset deferral mechanism for RPC deferrals */
2767         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2768         return rpc_success;
2769 }
2770
2771 #define op_encode_hdr_size              (2)
2772 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2773 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2774 #define op_encode_change_info_maxsz     (5)
2775 #define nfs4_fattr_bitmap_maxsz         (4)
2776
2777 /* We'll fall back on returning no lockowner if run out of space: */
2778 #define op_encode_lockowner_maxsz       (0)
2779 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2780
2781 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2782
2783 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2784 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2785                                          op_encode_ace_maxsz)
2786
2787 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2788
2789 /*
2790  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2791  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2792  * the maximum payload size here, based on transport limits and the size
2793  * of the remaining space in the rq_pages array.
2794  */
2795 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2796 {
2797         u32 buflen;
2798
2799         buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2800         buflen -= rqstp->rq_auth_slack;
2801         buflen -= rqstp->rq_res.head[0].iov_len;
2802         return min_t(u32, buflen, svc_max_payload(rqstp));
2803 }
2804
2805 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2806                                    const struct nfsd4_op *op)
2807 {
2808         return (op_encode_hdr_size) * sizeof(__be32);
2809 }
2810
2811 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2812                                       const struct nfsd4_op *op)
2813 {
2814         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2815 }
2816
2817 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2818                               const struct nfsd4_op *op)
2819 {
2820         /* ac_supported, ac_resp_access */
2821         return (op_encode_hdr_size + 2)* sizeof(__be32);
2822 }
2823
2824 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2825                               const struct nfsd4_op *op)
2826 {
2827         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2828 }
2829
2830 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2831                               const struct nfsd4_op *op)
2832 {
2833         return (op_encode_hdr_size + op_encode_change_info_maxsz
2834                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2835 }
2836
2837 /*
2838  * Note since this is an idempotent operation we won't insist on failing
2839  * the op prematurely if the estimate is too large.  We may turn off splice
2840  * reads unnecessarily.
2841  */
2842 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2843                                const struct nfsd4_op *op)
2844 {
2845         const u32 *bmap = op->u.getattr.ga_bmval;
2846         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2847         u32 ret = 0;
2848
2849         if (bmap0 & FATTR4_WORD0_ACL)
2850                 return nfsd4_max_payload(rqstp);
2851         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2852                 return nfsd4_max_payload(rqstp);
2853
2854         if (bmap1 & FATTR4_WORD1_OWNER) {
2855                 ret += IDMAP_NAMESZ + 4;
2856                 bmap1 &= ~FATTR4_WORD1_OWNER;
2857         }
2858         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2859                 ret += IDMAP_NAMESZ + 4;
2860                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2861         }
2862         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2863                 ret += NFS4_FHSIZE + 4;
2864                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2865         }
2866         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2867                 ret += NFS4_MAXLABELLEN + 12;
2868                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2869         }
2870         /*
2871          * Largest of remaining attributes are 16 bytes (e.g.,
2872          * supported_attributes)
2873          */
2874         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2875         /* bitmask, length */
2876         ret += 20;
2877         return ret;
2878 }
2879
2880 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2881                              const struct nfsd4_op *op)
2882 {
2883         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2884 }
2885
2886 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2887                             const struct nfsd4_op *op)
2888 {
2889         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2890                 * sizeof(__be32);
2891 }
2892
2893 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2894                             const struct nfsd4_op *op)
2895 {
2896         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2897                 * sizeof(__be32);
2898 }
2899
2900 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2901                             const struct nfsd4_op *op)
2902 {
2903         return (op_encode_hdr_size + op_encode_stateid_maxsz
2904                 + op_encode_change_info_maxsz + 1
2905                 + nfs4_fattr_bitmap_maxsz
2906                 + op_encode_delegation_maxsz) * sizeof(__be32);
2907 }
2908
2909 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2910                             const struct nfsd4_op *op)
2911 {
2912         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2913
2914         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2915 }
2916
2917 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2918                                  const struct nfsd4_op *op)
2919 {
2920         u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2921         /*
2922          * If we detect that the file changed during hole encoding, then we
2923          * recover by encoding the remaining reply as data. This means we need
2924          * to set aside enough room to encode two data segments.
2925          */
2926         u32 seg_len = 2 * (1 + 2 + 1);
2927
2928         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2929 }
2930
2931 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2932                                const struct nfsd4_op *op)
2933 {
2934         u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2935
2936         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2937                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2938 }
2939
2940 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2941                                 const struct nfsd4_op *op)
2942 {
2943         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2944 }
2945
2946 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2947                               const struct nfsd4_op *op)
2948 {
2949         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2950                 * sizeof(__be32);
2951 }
2952
2953 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2954                               const struct nfsd4_op *op)
2955 {
2956         return (op_encode_hdr_size + op_encode_change_info_maxsz
2957                 + op_encode_change_info_maxsz) * sizeof(__be32);
2958 }
2959
2960 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2961                                 const struct nfsd4_op *op)
2962 {
2963         return (op_encode_hdr_size
2964                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2965 }
2966
2967 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2968                                     const struct nfsd4_op *op)
2969 {
2970         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2971                 * sizeof(__be32);
2972 }
2973
2974 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2975                                const struct nfsd4_op *op)
2976 {
2977         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2978 }
2979
2980 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2981                                const struct nfsd4_op *op)
2982 {
2983         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2984                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2985 }
2986
2987 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2988                                    const struct nfsd4_op *op)
2989 {
2990         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2991                                                                 sizeof(__be32);
2992 }
2993
2994 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
2995                              const struct nfsd4_op *op)
2996 {
2997         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2998 }
2999
3000 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3001                                    const struct nfsd4_op *op)
3002 {
3003         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3004                 1 + 1 + /* eir_flags, spr_how */\
3005                 4 + /* spo_must_enforce & _allow with bitmap */\
3006                 2 + /*eir_server_owner.so_minor_id */\
3007                 /* eir_server_owner.so_major_id<> */\
3008                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3009                 /* eir_server_scope<> */\
3010                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3011                 1 + /* eir_server_impl_id array length */\
3012                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3013 }
3014
3015 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3016                                             const struct nfsd4_op *op)
3017 {
3018         return (op_encode_hdr_size + \
3019                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3020                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3021 }
3022
3023 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3024                                       const struct nfsd4_op *op)
3025 {
3026         return (op_encode_hdr_size + \
3027                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3028                 2 + /* csr_sequence, csr_flags */\
3029                 op_encode_channel_attrs_maxsz + \
3030                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3031 }
3032
3033 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3034                             const struct nfsd4_op *op)
3035 {
3036         return (op_encode_hdr_size +
3037                 1 /* wr_callback */ +
3038                 op_encode_stateid_maxsz /* wr_callback */ +
3039                 2 /* wr_count */ +
3040                 1 /* wr_committed */ +
3041                 op_encode_verifier_maxsz +
3042                 1 /* cr_consecutive */ +
3043                 1 /* cr_synchronous */) * sizeof(__be32);
3044 }
3045
3046 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3047                                       const struct nfsd4_op *op)
3048 {
3049         return (op_encode_hdr_size +
3050                 2 /* osr_count */ +
3051                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3052 }
3053
3054 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3055                                    const struct nfsd4_op *op)
3056 {
3057         return (op_encode_hdr_size +
3058                 3 /* cnr_lease_time */ +
3059                 1 /* We support one cnr_source_server */ +
3060                 1 /* cnr_stateid seq */ +
3061                 op_encode_stateid_maxsz /* cnr_stateid */ +
3062                 1 /* num cnr_source_server*/ +
3063                 1 /* nl4_type */ +
3064                 1 /* nl4 size */ +
3065                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3066                 * sizeof(__be32);
3067 }
3068
3069 #ifdef CONFIG_NFSD_PNFS
3070 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3071                                      const struct nfsd4_op *op)
3072 {
3073         u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3074
3075         return (op_encode_hdr_size +
3076                 1 /* gd_layout_type*/ +
3077                 XDR_QUADLEN(rlen) +
3078                 2 /* gd_notify_types */) * sizeof(__be32);
3079 }
3080
3081 /*
3082  * At this stage we don't really know what layout driver will handle the request,
3083  * so we need to define an arbitrary upper bound here.
3084  */
3085 #define MAX_LAYOUT_SIZE         128
3086 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3087                                  const struct nfsd4_op *op)
3088 {
3089         return (op_encode_hdr_size +
3090                 1 /* logr_return_on_close */ +
3091                 op_encode_stateid_maxsz +
3092                 1 /* nr of layouts */ +
3093                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3094 }
3095
3096 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3097                                     const struct nfsd4_op *op)
3098 {
3099         return (op_encode_hdr_size +
3100                 1 /* locr_newsize */ +
3101                 2 /* ns_size */) * sizeof(__be32);
3102 }
3103
3104 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3105                                     const struct nfsd4_op *op)
3106 {
3107         return (op_encode_hdr_size +
3108                 1 /* lrs_stateid */ +
3109                 op_encode_stateid_maxsz) * sizeof(__be32);
3110 }
3111 #endif /* CONFIG_NFSD_PNFS */
3112
3113
3114 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3115                             const struct nfsd4_op *op)
3116 {
3117         return (op_encode_hdr_size + 3) * sizeof(__be32);
3118 }
3119
3120 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3121                                 const struct nfsd4_op *op)
3122 {
3123         u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3124
3125         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3126 }
3127
3128 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3129                                 const struct nfsd4_op *op)
3130 {
3131         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3132                 * sizeof(__be32);
3133 }
3134 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3135                                   const struct nfsd4_op *op)
3136 {
3137         u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3138
3139         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3140 }
3141
3142 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3143                                    const struct nfsd4_op *op)
3144 {
3145         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3146                 * sizeof(__be32);
3147 }
3148
3149
3150 static const struct nfsd4_operation nfsd4_ops[] = {
3151         [OP_ACCESS] = {
3152                 .op_func = nfsd4_access,
3153                 .op_name = "OP_ACCESS",
3154                 .op_rsize_bop = nfsd4_access_rsize,
3155         },
3156         [OP_CLOSE] = {
3157                 .op_func = nfsd4_close,
3158                 .op_flags = OP_MODIFIES_SOMETHING,
3159                 .op_name = "OP_CLOSE",
3160                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3161                 .op_get_currentstateid = nfsd4_get_closestateid,
3162                 .op_set_currentstateid = nfsd4_set_closestateid,
3163         },
3164         [OP_COMMIT] = {
3165                 .op_func = nfsd4_commit,
3166                 .op_flags = OP_MODIFIES_SOMETHING,
3167                 .op_name = "OP_COMMIT",
3168                 .op_rsize_bop = nfsd4_commit_rsize,
3169         },
3170         [OP_CREATE] = {
3171                 .op_func = nfsd4_create,
3172                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3173                 .op_name = "OP_CREATE",
3174                 .op_rsize_bop = nfsd4_create_rsize,
3175         },
3176         [OP_DELEGRETURN] = {
3177                 .op_func = nfsd4_delegreturn,
3178                 .op_flags = OP_MODIFIES_SOMETHING,
3179                 .op_name = "OP_DELEGRETURN",
3180                 .op_rsize_bop = nfsd4_only_status_rsize,
3181                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3182         },
3183         [OP_GETATTR] = {
3184                 .op_func = nfsd4_getattr,
3185                 .op_flags = ALLOWED_ON_ABSENT_FS,
3186                 .op_rsize_bop = nfsd4_getattr_rsize,
3187                 .op_name = "OP_GETATTR",
3188         },
3189         [OP_GETFH] = {
3190                 .op_func = nfsd4_getfh,
3191                 .op_name = "OP_GETFH",
3192                 .op_rsize_bop = nfsd4_getfh_rsize,
3193         },
3194         [OP_LINK] = {
3195                 .op_func = nfsd4_link,
3196                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3197                                 | OP_CACHEME,
3198                 .op_name = "OP_LINK",
3199                 .op_rsize_bop = nfsd4_link_rsize,
3200         },
3201         [OP_LOCK] = {
3202                 .op_func = nfsd4_lock,
3203                 .op_flags = OP_MODIFIES_SOMETHING |
3204                                 OP_NONTRIVIAL_ERROR_ENCODE,
3205                 .op_name = "OP_LOCK",
3206                 .op_rsize_bop = nfsd4_lock_rsize,
3207                 .op_set_currentstateid = nfsd4_set_lockstateid,
3208         },
3209         [OP_LOCKT] = {
3210                 .op_func = nfsd4_lockt,
3211                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3212                 .op_name = "OP_LOCKT",
3213                 .op_rsize_bop = nfsd4_lock_rsize,
3214         },
3215         [OP_LOCKU] = {
3216                 .op_func = nfsd4_locku,
3217                 .op_flags = OP_MODIFIES_SOMETHING,
3218                 .op_name = "OP_LOCKU",
3219                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3220                 .op_get_currentstateid = nfsd4_get_lockustateid,
3221         },
3222         [OP_LOOKUP] = {
3223                 .op_func = nfsd4_lookup,
3224                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3225                 .op_name = "OP_LOOKUP",
3226                 .op_rsize_bop = nfsd4_only_status_rsize,
3227         },
3228         [OP_LOOKUPP] = {
3229                 .op_func = nfsd4_lookupp,
3230                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3231                 .op_name = "OP_LOOKUPP",
3232                 .op_rsize_bop = nfsd4_only_status_rsize,
3233         },
3234         [OP_NVERIFY] = {
3235                 .op_func = nfsd4_nverify,
3236                 .op_name = "OP_NVERIFY",
3237                 .op_rsize_bop = nfsd4_only_status_rsize,
3238         },
3239         [OP_OPEN] = {
3240                 .op_func = nfsd4_open,
3241                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3242                 .op_name = "OP_OPEN",
3243                 .op_rsize_bop = nfsd4_open_rsize,
3244                 .op_set_currentstateid = nfsd4_set_openstateid,
3245         },
3246         [OP_OPEN_CONFIRM] = {
3247                 .op_func = nfsd4_open_confirm,
3248                 .op_flags = OP_MODIFIES_SOMETHING,
3249                 .op_name = "OP_OPEN_CONFIRM",
3250                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3251         },
3252         [OP_OPEN_DOWNGRADE] = {
3253                 .op_func = nfsd4_open_downgrade,
3254                 .op_flags = OP_MODIFIES_SOMETHING,
3255                 .op_name = "OP_OPEN_DOWNGRADE",
3256                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3257                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3258                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3259         },
3260         [OP_PUTFH] = {
3261                 .op_func = nfsd4_putfh,
3262                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3263                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3264                 .op_name = "OP_PUTFH",
3265                 .op_rsize_bop = nfsd4_only_status_rsize,
3266         },
3267         [OP_PUTPUBFH] = {
3268                 .op_func = nfsd4_putrootfh,
3269                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3270                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3271                 .op_name = "OP_PUTPUBFH",
3272                 .op_rsize_bop = nfsd4_only_status_rsize,
3273         },
3274         [OP_PUTROOTFH] = {
3275                 .op_func = nfsd4_putrootfh,
3276                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3277                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3278                 .op_name = "OP_PUTROOTFH",
3279                 .op_rsize_bop = nfsd4_only_status_rsize,
3280         },
3281         [OP_READ] = {
3282                 .op_func = nfsd4_read,
3283                 .op_release = nfsd4_read_release,
3284                 .op_name = "OP_READ",
3285                 .op_rsize_bop = nfsd4_read_rsize,
3286                 .op_get_currentstateid = nfsd4_get_readstateid,
3287         },
3288         [OP_READDIR] = {
3289                 .op_func = nfsd4_readdir,
3290                 .op_name = "OP_READDIR",
3291                 .op_rsize_bop = nfsd4_readdir_rsize,
3292         },
3293         [OP_READLINK] = {
3294                 .op_func = nfsd4_readlink,
3295                 .op_name = "OP_READLINK",
3296                 .op_rsize_bop = nfsd4_readlink_rsize,
3297         },
3298         [OP_REMOVE] = {
3299                 .op_func = nfsd4_remove,
3300                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3301                 .op_name = "OP_REMOVE",
3302                 .op_rsize_bop = nfsd4_remove_rsize,
3303         },
3304         [OP_RENAME] = {
3305                 .op_func = nfsd4_rename,
3306                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3307                 .op_name = "OP_RENAME",
3308                 .op_rsize_bop = nfsd4_rename_rsize,
3309         },
3310         [OP_RENEW] = {
3311                 .op_func = nfsd4_renew,
3312                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3313                                 | OP_MODIFIES_SOMETHING,
3314                 .op_name = "OP_RENEW",
3315                 .op_rsize_bop = nfsd4_only_status_rsize,
3316
3317         },
3318         [OP_RESTOREFH] = {
3319                 .op_func = nfsd4_restorefh,
3320                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3321                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3322                 .op_name = "OP_RESTOREFH",
3323                 .op_rsize_bop = nfsd4_only_status_rsize,
3324         },
3325         [OP_SAVEFH] = {
3326                 .op_func = nfsd4_savefh,
3327                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3328                 .op_name = "OP_SAVEFH",
3329                 .op_rsize_bop = nfsd4_only_status_rsize,
3330         },
3331         [OP_SECINFO] = {
3332                 .op_func = nfsd4_secinfo,
3333                 .op_release = nfsd4_secinfo_release,
3334                 .op_flags = OP_HANDLES_WRONGSEC,
3335                 .op_name = "OP_SECINFO",
3336                 .op_rsize_bop = nfsd4_secinfo_rsize,
3337         },
3338         [OP_SETATTR] = {
3339                 .op_func = nfsd4_setattr,
3340                 .op_name = "OP_SETATTR",
3341                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3342                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3343                 .op_rsize_bop = nfsd4_setattr_rsize,
3344                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3345         },
3346         [OP_SETCLIENTID] = {
3347                 .op_func = nfsd4_setclientid,
3348                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3349                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3350                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3351                 .op_name = "OP_SETCLIENTID",
3352                 .op_rsize_bop = nfsd4_setclientid_rsize,
3353         },
3354         [OP_SETCLIENTID_CONFIRM] = {
3355                 .op_func = nfsd4_setclientid_confirm,
3356                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3357                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3358                 .op_name = "OP_SETCLIENTID_CONFIRM",
3359                 .op_rsize_bop = nfsd4_only_status_rsize,
3360         },
3361         [OP_VERIFY] = {
3362                 .op_func = nfsd4_verify,
3363                 .op_name = "OP_VERIFY",
3364                 .op_rsize_bop = nfsd4_only_status_rsize,
3365         },
3366         [OP_WRITE] = {
3367                 .op_func = nfsd4_write,
3368                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3369                 .op_name = "OP_WRITE",
3370                 .op_rsize_bop = nfsd4_write_rsize,
3371                 .op_get_currentstateid = nfsd4_get_writestateid,
3372         },
3373         [OP_RELEASE_LOCKOWNER] = {
3374                 .op_func = nfsd4_release_lockowner,
3375                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3376                                 | OP_MODIFIES_SOMETHING,
3377                 .op_name = "OP_RELEASE_LOCKOWNER",
3378                 .op_rsize_bop = nfsd4_only_status_rsize,
3379         },
3380
3381         /* NFSv4.1 operations */
3382         [OP_EXCHANGE_ID] = {
3383                 .op_func = nfsd4_exchange_id,
3384                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3385                                 | OP_MODIFIES_SOMETHING,
3386                 .op_name = "OP_EXCHANGE_ID",
3387                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3388         },
3389         [OP_BACKCHANNEL_CTL] = {
3390                 .op_func = nfsd4_backchannel_ctl,
3391                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3392                 .op_name = "OP_BACKCHANNEL_CTL",
3393                 .op_rsize_bop = nfsd4_only_status_rsize,
3394         },
3395         [OP_BIND_CONN_TO_SESSION] = {
3396                 .op_func = nfsd4_bind_conn_to_session,
3397                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3398                                 | OP_MODIFIES_SOMETHING,
3399                 .op_name = "OP_BIND_CONN_TO_SESSION",
3400                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3401         },
3402         [OP_CREATE_SESSION] = {
3403                 .op_func = nfsd4_create_session,
3404                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3405                                 | OP_MODIFIES_SOMETHING,
3406                 .op_name = "OP_CREATE_SESSION",
3407                 .op_rsize_bop = nfsd4_create_session_rsize,
3408         },
3409         [OP_DESTROY_SESSION] = {
3410                 .op_func = nfsd4_destroy_session,
3411                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3412                                 | OP_MODIFIES_SOMETHING,
3413                 .op_name = "OP_DESTROY_SESSION",
3414                 .op_rsize_bop = nfsd4_only_status_rsize,
3415         },
3416         [OP_SEQUENCE] = {
3417                 .op_func = nfsd4_sequence,
3418                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3419                 .op_name = "OP_SEQUENCE",
3420                 .op_rsize_bop = nfsd4_sequence_rsize,
3421         },
3422         [OP_DESTROY_CLIENTID] = {
3423                 .op_func = nfsd4_destroy_clientid,
3424                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3425                                 | OP_MODIFIES_SOMETHING,
3426                 .op_name = "OP_DESTROY_CLIENTID",
3427                 .op_rsize_bop = nfsd4_only_status_rsize,
3428         },
3429         [OP_RECLAIM_COMPLETE] = {
3430                 .op_func = nfsd4_reclaim_complete,
3431                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3432                 .op_name = "OP_RECLAIM_COMPLETE",
3433                 .op_rsize_bop = nfsd4_only_status_rsize,
3434         },
3435         [OP_SECINFO_NO_NAME] = {
3436                 .op_func = nfsd4_secinfo_no_name,
3437                 .op_release = nfsd4_secinfo_no_name_release,
3438                 .op_flags = OP_HANDLES_WRONGSEC,
3439                 .op_name = "OP_SECINFO_NO_NAME",
3440                 .op_rsize_bop = nfsd4_secinfo_rsize,
3441         },
3442         [OP_TEST_STATEID] = {
3443                 .op_func = nfsd4_test_stateid,
3444                 .op_flags = ALLOWED_WITHOUT_FH,
3445                 .op_name = "OP_TEST_STATEID",
3446                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3447         },
3448         [OP_FREE_STATEID] = {
3449                 .op_func = nfsd4_free_stateid,
3450                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3451                 .op_name = "OP_FREE_STATEID",
3452                 .op_get_currentstateid = nfsd4_get_freestateid,
3453                 .op_rsize_bop = nfsd4_only_status_rsize,
3454         },
3455 #ifdef CONFIG_NFSD_PNFS
3456         [OP_GETDEVICEINFO] = {
3457                 .op_func = nfsd4_getdeviceinfo,
3458                 .op_release = nfsd4_getdeviceinfo_release,
3459                 .op_flags = ALLOWED_WITHOUT_FH,
3460                 .op_name = "OP_GETDEVICEINFO",
3461                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3462         },
3463         [OP_LAYOUTGET] = {
3464                 .op_func = nfsd4_layoutget,
3465                 .op_release = nfsd4_layoutget_release,
3466                 .op_flags = OP_MODIFIES_SOMETHING,
3467                 .op_name = "OP_LAYOUTGET",
3468                 .op_rsize_bop = nfsd4_layoutget_rsize,
3469         },
3470         [OP_LAYOUTCOMMIT] = {
3471                 .op_func = nfsd4_layoutcommit,
3472                 .op_flags = OP_MODIFIES_SOMETHING,
3473                 .op_name = "OP_LAYOUTCOMMIT",
3474                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3475         },
3476         [OP_LAYOUTRETURN] = {
3477                 .op_func = nfsd4_layoutreturn,
3478                 .op_flags = OP_MODIFIES_SOMETHING,
3479                 .op_name = "OP_LAYOUTRETURN",
3480                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3481         },
3482 #endif /* CONFIG_NFSD_PNFS */
3483
3484         /* NFSv4.2 operations */
3485         [OP_ALLOCATE] = {
3486                 .op_func = nfsd4_allocate,
3487                 .op_flags = OP_MODIFIES_SOMETHING,
3488                 .op_name = "OP_ALLOCATE",
3489                 .op_rsize_bop = nfsd4_only_status_rsize,
3490         },
3491         [OP_DEALLOCATE] = {
3492                 .op_func = nfsd4_deallocate,
3493                 .op_flags = OP_MODIFIES_SOMETHING,
3494                 .op_name = "OP_DEALLOCATE",
3495                 .op_rsize_bop = nfsd4_only_status_rsize,
3496         },
3497         [OP_CLONE] = {
3498                 .op_func = nfsd4_clone,
3499                 .op_flags = OP_MODIFIES_SOMETHING,
3500                 .op_name = "OP_CLONE",
3501                 .op_rsize_bop = nfsd4_only_status_rsize,
3502         },
3503         [OP_COPY] = {
3504                 .op_func = nfsd4_copy,
3505                 .op_flags = OP_MODIFIES_SOMETHING,
3506                 .op_name = "OP_COPY",
3507                 .op_rsize_bop = nfsd4_copy_rsize,
3508         },
3509         [OP_READ_PLUS] = {
3510                 .op_func = nfsd4_read,
3511                 .op_release = nfsd4_read_release,
3512                 .op_name = "OP_READ_PLUS",
3513                 .op_rsize_bop = nfsd4_read_plus_rsize,
3514                 .op_get_currentstateid = nfsd4_get_readstateid,
3515         },
3516         [OP_SEEK] = {
3517                 .op_func = nfsd4_seek,
3518                 .op_name = "OP_SEEK",
3519                 .op_rsize_bop = nfsd4_seek_rsize,
3520         },
3521         [OP_OFFLOAD_STATUS] = {
3522                 .op_func = nfsd4_offload_status,
3523                 .op_name = "OP_OFFLOAD_STATUS",
3524                 .op_rsize_bop = nfsd4_offload_status_rsize,
3525         },
3526         [OP_OFFLOAD_CANCEL] = {
3527                 .op_func = nfsd4_offload_cancel,
3528                 .op_flags = OP_MODIFIES_SOMETHING,
3529                 .op_name = "OP_OFFLOAD_CANCEL",
3530                 .op_rsize_bop = nfsd4_only_status_rsize,
3531         },
3532         [OP_COPY_NOTIFY] = {
3533                 .op_func = nfsd4_copy_notify,
3534                 .op_flags = OP_MODIFIES_SOMETHING,
3535                 .op_name = "OP_COPY_NOTIFY",
3536                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3537         },
3538         [OP_GETXATTR] = {
3539                 .op_func = nfsd4_getxattr,
3540                 .op_name = "OP_GETXATTR",
3541                 .op_rsize_bop = nfsd4_getxattr_rsize,
3542         },
3543         [OP_SETXATTR] = {
3544                 .op_func = nfsd4_setxattr,
3545                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3546                 .op_name = "OP_SETXATTR",
3547                 .op_rsize_bop = nfsd4_setxattr_rsize,
3548         },
3549         [OP_LISTXATTRS] = {
3550                 .op_func = nfsd4_listxattrs,
3551                 .op_name = "OP_LISTXATTRS",
3552                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3553         },
3554         [OP_REMOVEXATTR] = {
3555                 .op_func = nfsd4_removexattr,
3556                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3557                 .op_name = "OP_REMOVEXATTR",
3558                 .op_rsize_bop = nfsd4_removexattr_rsize,
3559         },
3560 };
3561
3562 /**
3563  * nfsd4_spo_must_allow - Determine if the compound op contains an
3564  * operation that is allowed to be sent with machine credentials
3565  *
3566  * @rqstp: a pointer to the struct svc_rqst
3567  *
3568  * Checks to see if the compound contains a spo_must_allow op
3569  * and confirms that it was sent with the proper machine creds.
3570  */
3571
3572 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3573 {
3574         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3575         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3576         struct nfsd4_op *this;
3577         struct nfsd4_compound_state *cstate = &resp->cstate;
3578         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3579         u32 opiter;
3580
3581         if (!cstate->minorversion)
3582                 return false;
3583
3584         if (cstate->spo_must_allowed)
3585                 return true;
3586
3587         opiter = resp->opcnt;
3588         while (opiter < argp->opcnt) {
3589                 this = &argp->ops[opiter++];
3590                 if (test_bit(this->opnum, allow->u.longs) &&
3591                         cstate->clp->cl_mach_cred &&
3592                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3593                         cstate->spo_must_allowed = true;
3594                         return true;
3595                 }
3596         }
3597         cstate->spo_must_allowed = false;
3598         return false;
3599 }
3600
3601 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3602 {
3603         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3604                 return op_encode_hdr_size * sizeof(__be32);
3605
3606         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3607         return OPDESC(op)->op_rsize_bop(rqstp, op);
3608 }
3609
3610 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3611 {
3612         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3613                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3614                         op->opnum, nfsd4_op_name(op->opnum));
3615                 WARN_ON_ONCE(1);
3616         }
3617 }
3618
3619 static const char *nfsd4_op_name(unsigned opnum)
3620 {
3621         if (opnum < ARRAY_SIZE(nfsd4_ops))
3622                 return nfsd4_ops[opnum].op_name;
3623         return "unknown_operation";
3624 }
3625
3626 static const struct svc_procedure nfsd_procedures4[2] = {
3627         [NFSPROC4_NULL] = {
3628                 .pc_func = nfsd4_proc_null,
3629                 .pc_decode = nfssvc_decode_voidarg,
3630                 .pc_encode = nfssvc_encode_voidres,
3631                 .pc_argsize = sizeof(struct nfsd_voidargs),
3632                 .pc_argzero = sizeof(struct nfsd_voidargs),
3633                 .pc_ressize = sizeof(struct nfsd_voidres),
3634                 .pc_cachetype = RC_NOCACHE,
3635                 .pc_xdrressize = 1,
3636                 .pc_name = "NULL",
3637         },
3638         [NFSPROC4_COMPOUND] = {
3639                 .pc_func = nfsd4_proc_compound,
3640                 .pc_decode = nfs4svc_decode_compoundargs,
3641                 .pc_encode = nfs4svc_encode_compoundres,
3642                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3643                 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3644                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3645                 .pc_release = nfsd4_release_compoundargs,
3646                 .pc_cachetype = RC_NOCACHE,
3647                 .pc_xdrressize = NFSD_BUFSIZE/4,
3648                 .pc_name = "COMPOUND",
3649         },
3650 };
3651
3652 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3653                               nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3654 const struct svc_version nfsd_version4 = {
3655         .vs_vers                = 4,
3656         .vs_nproc               = ARRAY_SIZE(nfsd_procedures4),
3657         .vs_proc                = nfsd_procedures4,
3658         .vs_count               = nfsd_count4,
3659         .vs_dispatch            = nfsd_dispatch,
3660         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3661         .vs_rpcb_optnl          = true,
3662         .vs_need_cong_ctrl      = true,
3663 };