Merge drm/drm-next into drm-intel-gt-next
[linux-2.6-microblaze.git] / 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/sunrpc/addr.h>
41 #include <linux/nfs_ssc.h>
42
43 #include "idmap.h"
44 #include "cache.h"
45 #include "xdr4.h"
46 #include "vfs.h"
47 #include "current_stateid.h"
48 #include "netns.h"
49 #include "acl.h"
50 #include "pnfs.h"
51 #include "trace.h"
52
53 static bool inter_copy_offload_enable;
54 module_param(inter_copy_offload_enable, bool, 0644);
55 MODULE_PARM_DESC(inter_copy_offload_enable,
56                  "Enable inter server to server copy offload. Default: false");
57
58 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
59 #include <linux/security.h>
60
61 static inline void
62 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
63 {
64         struct inode *inode = d_inode(resfh->fh_dentry);
65         int status;
66
67         inode_lock(inode);
68         status = security_inode_setsecctx(resfh->fh_dentry,
69                 label->data, label->len);
70         inode_unlock(inode);
71
72         if (status)
73                 /*
74                  * XXX: We should really fail the whole open, but we may
75                  * already have created a new file, so it may be too
76                  * late.  For now this seems the least of evils:
77                  */
78                 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
79
80         return;
81 }
82 #else
83 static inline void
84 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
85 { }
86 #endif
87
88 #define NFSDDBG_FACILITY                NFSDDBG_PROC
89
90 static u32 nfsd_attrmask[] = {
91         NFSD_WRITEABLE_ATTRS_WORD0,
92         NFSD_WRITEABLE_ATTRS_WORD1,
93         NFSD_WRITEABLE_ATTRS_WORD2
94 };
95
96 static u32 nfsd41_ex_attrmask[] = {
97         NFSD_SUPPATTR_EXCLCREAT_WORD0,
98         NFSD_SUPPATTR_EXCLCREAT_WORD1,
99         NFSD_SUPPATTR_EXCLCREAT_WORD2
100 };
101
102 static __be32
103 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
104                    u32 *bmval, u32 *writable)
105 {
106         struct dentry *dentry = cstate->current_fh.fh_dentry;
107         struct svc_export *exp = cstate->current_fh.fh_export;
108
109         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
110                 return nfserr_attrnotsupp;
111         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
112                 return nfserr_attrnotsupp;
113         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
114                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
115                 return nfserr_attrnotsupp;
116         if (writable && !bmval_is_subset(bmval, writable))
117                 return nfserr_inval;
118         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
119                         (bmval[1] & FATTR4_WORD1_MODE))
120                 return nfserr_inval;
121         return nfs_ok;
122 }
123
124 static __be32
125 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
126         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
127 {
128         __be32 status = nfs_ok;
129
130         if (open->op_create == NFS4_OPEN_CREATE) {
131                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
132                     || open->op_createmode == NFS4_CREATE_GUARDED)
133                         status = check_attr_support(rqstp, cstate,
134                                         open->op_bmval, nfsd_attrmask);
135                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
136                         status = check_attr_support(rqstp, cstate,
137                                         open->op_bmval, nfsd41_ex_attrmask);
138         }
139
140         return status;
141 }
142
143 static int
144 is_create_with_attrs(struct nfsd4_open *open)
145 {
146         return open->op_create == NFS4_OPEN_CREATE
147                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
148                     || open->op_createmode == NFS4_CREATE_GUARDED
149                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
150 }
151
152 /*
153  * if error occurs when setting the acl, just clear the acl bit
154  * in the returned attr bitmap.
155  */
156 static void
157 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
158                 struct nfs4_acl *acl, u32 *bmval)
159 {
160         __be32 status;
161
162         status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
163         if (status)
164                 /*
165                  * We should probably fail the whole open at this point,
166                  * but we've already created the file, so it's too late;
167                  * So this seems the least of evils:
168                  */
169                 bmval[0] &= ~FATTR4_WORD0_ACL;
170 }
171
172 static inline void
173 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
174 {
175         fh_put(dst);
176         dget(src->fh_dentry);
177         if (src->fh_export)
178                 exp_get(src->fh_export);
179         *dst = *src;
180 }
181
182 static __be32
183 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
184 {
185         __be32 status;
186
187         if (open->op_truncate &&
188                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
189                 return nfserr_inval;
190
191         accmode |= NFSD_MAY_READ_IF_EXEC;
192
193         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
194                 accmode |= NFSD_MAY_READ;
195         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
196                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
197         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
198                 accmode |= NFSD_MAY_WRITE;
199
200         status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
201
202         return status;
203 }
204
205 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
206 {
207         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
208
209         if (S_ISREG(mode))
210                 return nfs_ok;
211         if (S_ISDIR(mode))
212                 return nfserr_isdir;
213         /*
214          * Using err_symlink as our catch-all case may look odd; but
215          * there's no other obvious error for this case in 4.0, and we
216          * happen to know that it will cause the linux v4 client to do
217          * the right thing on attempts to open something other than a
218          * regular file.
219          */
220         return nfserr_symlink;
221 }
222
223 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
224 {
225         if (nfsd4_has_session(cstate))
226                 return;
227         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
228                         &resfh->fh_handle);
229 }
230
231 static __be32
232 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
233 {
234         struct svc_fh *current_fh = &cstate->current_fh;
235         int accmode;
236         __be32 status;
237
238         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
239         if (!*resfh)
240                 return nfserr_jukebox;
241         fh_init(*resfh, NFS4_FHSIZE);
242         open->op_truncate = false;
243
244         if (open->op_create) {
245                 /* FIXME: check session persistence and pnfs flags.
246                  * The nfsv4.1 spec requires the following semantics:
247                  *
248                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
249                  * Reply Cache  | server |                 |
250                  * -------------+--------+-----------------+--------------------
251                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
252                  *              |        |                 | (SHOULD)
253                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
254                  *              |        |                 | (SHOULD NOT)
255                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
256                  * yes          | no     | GUARDED4        | GUARDED4
257                  * yes          | yes    | GUARDED4        | GUARDED4
258                  */
259
260                 /*
261                  * Note: create modes (UNCHECKED,GUARDED...) are the same
262                  * in NFSv4 as in v3 except EXCLUSIVE4_1.
263                  */
264                 current->fs->umask = open->op_umask;
265                 status = do_nfsd_create(rqstp, current_fh, open->op_fname,
266                                         open->op_fnamelen, &open->op_iattr,
267                                         *resfh, open->op_createmode,
268                                         (u32 *)open->op_verf.data,
269                                         &open->op_truncate, &open->op_created);
270                 current->fs->umask = 0;
271
272                 if (!status && open->op_label.len)
273                         nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
274
275                 /*
276                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
277                  * use the returned bitmask to indicate which attributes
278                  * we used to store the verifier:
279                  */
280                 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
281                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
282                                                 FATTR4_WORD1_TIME_MODIFY);
283         } else
284                 /*
285                  * Note this may exit with the parent still locked.
286                  * We will hold the lock until nfsd4_open's final
287                  * lookup, to prevent renames or unlinks until we've had
288                  * a chance to an acquire a delegation if appropriate.
289                  */
290                 status = nfsd_lookup(rqstp, current_fh,
291                                      open->op_fname, open->op_fnamelen, *resfh);
292         if (status)
293                 goto out;
294         status = nfsd_check_obj_isreg(*resfh);
295         if (status)
296                 goto out;
297
298         if (is_create_with_attrs(open) && open->op_acl != NULL)
299                 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
300
301         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
302         accmode = NFSD_MAY_NOP;
303         if (open->op_created ||
304                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
305                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
306         status = do_open_permission(rqstp, *resfh, open, accmode);
307         set_change_info(&open->op_cinfo, current_fh);
308 out:
309         return status;
310 }
311
312 static __be32
313 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
314 {
315         struct svc_fh *current_fh = &cstate->current_fh;
316         __be32 status;
317         int accmode = 0;
318
319         /* We don't know the target directory, and therefore can not
320         * set the change info
321         */
322
323         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
324
325         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
326
327         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
328                 (open->op_iattr.ia_size == 0);
329         /*
330          * In the delegation case, the client is telling us about an
331          * open that it *already* performed locally, some time ago.  We
332          * should let it succeed now if possible.
333          *
334          * In the case of a CLAIM_FH open, on the other hand, the client
335          * may be counting on us to enforce permissions (the Linux 4.1
336          * client uses this for normal opens, for example).
337          */
338         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
339                 accmode = NFSD_MAY_OWNER_OVERRIDE;
340
341         status = do_open_permission(rqstp, current_fh, open, accmode);
342
343         return status;
344 }
345
346 static void
347 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
348 {
349         struct nfsd4_sessionid *sid =
350                         (struct nfsd4_sessionid *)session->se_sessionid.data;
351
352         clid->cl_boot = sid->clientid.cl_boot;
353         clid->cl_id = sid->clientid.cl_id;
354 }
355
356 static __be32
357 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
358            union nfsd4_op_u *u)
359 {
360         struct nfsd4_open *open = &u->open;
361         __be32 status;
362         struct svc_fh *resfh = NULL;
363         struct net *net = SVC_NET(rqstp);
364         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
365         bool reclaim = false;
366
367         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
368                 (int)open->op_fnamelen, open->op_fname,
369                 open->op_openowner);
370
371         /* This check required by spec. */
372         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
373                 return nfserr_inval;
374
375         open->op_created = false;
376         /*
377          * RFC5661 18.51.3
378          * Before RECLAIM_COMPLETE done, server should deny new lock
379          */
380         if (nfsd4_has_session(cstate) &&
381             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
382             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
383                 return nfserr_grace;
384
385         if (nfsd4_has_session(cstate))
386                 copy_clientid(&open->op_clientid, cstate->session);
387
388         /* check seqid for replay. set nfs4_owner */
389         status = nfsd4_process_open1(cstate, open, nn);
390         if (status == nfserr_replay_me) {
391                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
392                 fh_put(&cstate->current_fh);
393                 fh_copy_shallow(&cstate->current_fh.fh_handle,
394                                 &rp->rp_openfh);
395                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
396                 if (status)
397                         dprintk("nfsd4_open: replay failed"
398                                 " restoring previous filehandle\n");
399                 else
400                         status = nfserr_replay_me;
401         }
402         if (status)
403                 goto out;
404         if (open->op_xdr_error) {
405                 status = open->op_xdr_error;
406                 goto out;
407         }
408
409         status = nfsd4_check_open_attributes(rqstp, cstate, open);
410         if (status)
411                 goto out;
412
413         /* Openowner is now set, so sequence id will get bumped.  Now we need
414          * these checks before we do any creates: */
415         status = nfserr_grace;
416         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
417                 goto out;
418         status = nfserr_no_grace;
419         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
420                 goto out;
421
422         switch (open->op_claim_type) {
423                 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
424                 case NFS4_OPEN_CLAIM_NULL:
425                         status = do_open_lookup(rqstp, cstate, open, &resfh);
426                         if (status)
427                                 goto out;
428                         break;
429                 case NFS4_OPEN_CLAIM_PREVIOUS:
430                         status = nfs4_check_open_reclaim(cstate->clp);
431                         if (status)
432                                 goto out;
433                         open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
434                         reclaim = true;
435                         fallthrough;
436                 case NFS4_OPEN_CLAIM_FH:
437                 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
438                         status = do_open_fhandle(rqstp, cstate, open);
439                         if (status)
440                                 goto out;
441                         resfh = &cstate->current_fh;
442                         break;
443                 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
444                 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
445                         dprintk("NFSD: unsupported OPEN claim type %d\n",
446                                 open->op_claim_type);
447                         status = nfserr_notsupp;
448                         goto out;
449                 default:
450                         dprintk("NFSD: Invalid OPEN claim type %d\n",
451                                 open->op_claim_type);
452                         status = nfserr_inval;
453                         goto out;
454         }
455         /*
456          * nfsd4_process_open2() does the actual opening of the file.  If
457          * successful, it (1) truncates the file if open->op_truncate was
458          * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
459          */
460         status = nfsd4_process_open2(rqstp, resfh, open);
461         WARN(status && open->op_created,
462              "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
463              be32_to_cpu(status));
464         if (reclaim && !status)
465                 nn->somebody_reclaimed = true;
466 out:
467         if (resfh && resfh != &cstate->current_fh) {
468                 fh_dup2(&cstate->current_fh, resfh);
469                 fh_put(resfh);
470                 kfree(resfh);
471         }
472         nfsd4_cleanup_open_state(cstate, open);
473         nfsd4_bump_seqid(cstate, status);
474         return status;
475 }
476
477 /*
478  * OPEN is the only seqid-mutating operation whose decoding can fail
479  * with a seqid-mutating error (specifically, decoding of user names in
480  * the attributes).  Therefore we have to do some processing to look up
481  * the stateowner so that we can bump the seqid.
482  */
483 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
484 {
485         struct nfsd4_open *open = &op->u.open;
486
487         if (!seqid_mutating_err(ntohl(op->status)))
488                 return op->status;
489         if (nfsd4_has_session(cstate))
490                 return op->status;
491         open->op_xdr_error = op->status;
492         return nfsd4_open(rqstp, cstate, &op->u);
493 }
494
495 /*
496  * filehandle-manipulating ops.
497  */
498 static __be32
499 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
500             union nfsd4_op_u *u)
501 {
502         u->getfh = &cstate->current_fh;
503         return nfs_ok;
504 }
505
506 static __be32
507 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
508             union nfsd4_op_u *u)
509 {
510         struct nfsd4_putfh *putfh = &u->putfh;
511         __be32 ret;
512
513         fh_put(&cstate->current_fh);
514         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
515         memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
516                putfh->pf_fhlen);
517         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
518 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
519         if (ret == nfserr_stale && putfh->no_verify) {
520                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
521                 ret = 0;
522         }
523 #endif
524         return ret;
525 }
526
527 static __be32
528 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
529                 union nfsd4_op_u *u)
530 {
531         __be32 status;
532
533         fh_put(&cstate->current_fh);
534         status = exp_pseudoroot(rqstp, &cstate->current_fh);
535         return status;
536 }
537
538 static __be32
539 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
540                 union nfsd4_op_u *u)
541 {
542         if (!cstate->save_fh.fh_dentry)
543                 return nfserr_restorefh;
544
545         fh_dup2(&cstate->current_fh, &cstate->save_fh);
546         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
547                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
548                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
549         }
550         return nfs_ok;
551 }
552
553 static __be32
554 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
555              union nfsd4_op_u *u)
556 {
557         fh_dup2(&cstate->save_fh, &cstate->current_fh);
558         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
559                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
560                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
561         }
562         return nfs_ok;
563 }
564
565 /*
566  * misc nfsv4 ops
567  */
568 static __be32
569 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
570              union nfsd4_op_u *u)
571 {
572         struct nfsd4_access *access = &u->access;
573         u32 access_full;
574
575         access_full = NFS3_ACCESS_FULL;
576         if (cstate->minorversion >= 2)
577                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
578                                NFS4_ACCESS_XAWRITE;
579
580         if (access->ac_req_access & ~access_full)
581                 return nfserr_inval;
582
583         access->ac_resp_access = access->ac_req_access;
584         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
585                            &access->ac_supported);
586 }
587
588 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
589 {
590         __be32 *verf = (__be32 *)verifier->data;
591
592         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
593
594         nfsd_copy_boot_verifier(verf, net_generic(net, nfsd_net_id));
595 }
596
597 static __be32
598 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
599              union nfsd4_op_u *u)
600 {
601         struct nfsd4_commit *commit = &u->commit;
602
603         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
604                              commit->co_count,
605                              (__be32 *)commit->co_verf.data);
606 }
607
608 static __be32
609 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
610              union nfsd4_op_u *u)
611 {
612         struct nfsd4_create *create = &u->create;
613         struct svc_fh resfh;
614         __be32 status;
615         dev_t rdev;
616
617         fh_init(&resfh, NFS4_FHSIZE);
618
619         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
620         if (status)
621                 return status;
622
623         status = check_attr_support(rqstp, cstate, create->cr_bmval,
624                                     nfsd_attrmask);
625         if (status)
626                 return status;
627
628         current->fs->umask = create->cr_umask;
629         switch (create->cr_type) {
630         case NF4LNK:
631                 status = nfsd_symlink(rqstp, &cstate->current_fh,
632                                       create->cr_name, create->cr_namelen,
633                                       create->cr_data, &resfh);
634                 break;
635
636         case NF4BLK:
637                 status = nfserr_inval;
638                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
639                 if (MAJOR(rdev) != create->cr_specdata1 ||
640                     MINOR(rdev) != create->cr_specdata2)
641                         goto out_umask;
642                 status = nfsd_create(rqstp, &cstate->current_fh,
643                                      create->cr_name, create->cr_namelen,
644                                      &create->cr_iattr, S_IFBLK, rdev, &resfh);
645                 break;
646
647         case NF4CHR:
648                 status = nfserr_inval;
649                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
650                 if (MAJOR(rdev) != create->cr_specdata1 ||
651                     MINOR(rdev) != create->cr_specdata2)
652                         goto out_umask;
653                 status = nfsd_create(rqstp, &cstate->current_fh,
654                                      create->cr_name, create->cr_namelen,
655                                      &create->cr_iattr,S_IFCHR, rdev, &resfh);
656                 break;
657
658         case NF4SOCK:
659                 status = nfsd_create(rqstp, &cstate->current_fh,
660                                      create->cr_name, create->cr_namelen,
661                                      &create->cr_iattr, S_IFSOCK, 0, &resfh);
662                 break;
663
664         case NF4FIFO:
665                 status = nfsd_create(rqstp, &cstate->current_fh,
666                                      create->cr_name, create->cr_namelen,
667                                      &create->cr_iattr, S_IFIFO, 0, &resfh);
668                 break;
669
670         case NF4DIR:
671                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
672                 status = nfsd_create(rqstp, &cstate->current_fh,
673                                      create->cr_name, create->cr_namelen,
674                                      &create->cr_iattr, S_IFDIR, 0, &resfh);
675                 break;
676
677         default:
678                 status = nfserr_badtype;
679         }
680
681         if (status)
682                 goto out;
683
684         if (create->cr_label.len)
685                 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
686
687         if (create->cr_acl != NULL)
688                 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
689                                 create->cr_bmval);
690
691         fh_unlock(&cstate->current_fh);
692         set_change_info(&create->cr_cinfo, &cstate->current_fh);
693         fh_dup2(&cstate->current_fh, &resfh);
694 out:
695         fh_put(&resfh);
696 out_umask:
697         current->fs->umask = 0;
698         return status;
699 }
700
701 static __be32
702 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
703               union nfsd4_op_u *u)
704 {
705         struct nfsd4_getattr *getattr = &u->getattr;
706         __be32 status;
707
708         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
709         if (status)
710                 return status;
711
712         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
713                 return nfserr_inval;
714
715         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
716         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
717         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
718
719         getattr->ga_fhp = &cstate->current_fh;
720         return nfs_ok;
721 }
722
723 static __be32
724 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
725            union nfsd4_op_u *u)
726 {
727         struct nfsd4_link *link = &u->link;
728         __be32 status;
729
730         status = nfsd_link(rqstp, &cstate->current_fh,
731                            link->li_name, link->li_namelen, &cstate->save_fh);
732         if (!status)
733                 set_change_info(&link->li_cinfo, &cstate->current_fh);
734         return status;
735 }
736
737 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
738 {
739         struct svc_fh tmp_fh;
740         __be32 ret;
741
742         fh_init(&tmp_fh, NFS4_FHSIZE);
743         ret = exp_pseudoroot(rqstp, &tmp_fh);
744         if (ret)
745                 return ret;
746         if (tmp_fh.fh_dentry == fh->fh_dentry) {
747                 fh_put(&tmp_fh);
748                 return nfserr_noent;
749         }
750         fh_put(&tmp_fh);
751         return nfsd_lookup(rqstp, fh, "..", 2, fh);
752 }
753
754 static __be32
755 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
756               union nfsd4_op_u *u)
757 {
758         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
759 }
760
761 static __be32
762 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
763              union nfsd4_op_u *u)
764 {
765         return nfsd_lookup(rqstp, &cstate->current_fh,
766                            u->lookup.lo_name, u->lookup.lo_len,
767                            &cstate->current_fh);
768 }
769
770 static __be32
771 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
772            union nfsd4_op_u *u)
773 {
774         struct nfsd4_read *read = &u->read;
775         __be32 status;
776
777         read->rd_nf = NULL;
778         if (read->rd_offset >= OFFSET_MAX)
779                 return nfserr_inval;
780
781         trace_nfsd_read_start(rqstp, &cstate->current_fh,
782                               read->rd_offset, read->rd_length);
783
784         /*
785          * If we do a zero copy read, then a client will see read data
786          * that reflects the state of the file *after* performing the
787          * following compound.
788          *
789          * To ensure proper ordering, we therefore turn off zero copy if
790          * the client wants us to do more in this compound:
791          */
792         if (!nfsd4_last_compound_op(rqstp))
793                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
794
795         /* check stateid */
796         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
797                                         &read->rd_stateid, RD_STATE,
798                                         &read->rd_nf, NULL);
799         if (status) {
800                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
801                 goto out;
802         }
803         status = nfs_ok;
804 out:
805         read->rd_rqstp = rqstp;
806         read->rd_fhp = &cstate->current_fh;
807         return status;
808 }
809
810
811 static void
812 nfsd4_read_release(union nfsd4_op_u *u)
813 {
814         if (u->read.rd_nf)
815                 nfsd_file_put(u->read.rd_nf);
816         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
817                              u->read.rd_offset, u->read.rd_length);
818 }
819
820 static __be32
821 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
822               union nfsd4_op_u *u)
823 {
824         struct nfsd4_readdir *readdir = &u->readdir;
825         u64 cookie = readdir->rd_cookie;
826         static const nfs4_verifier zeroverf;
827
828         /* no need to check permission - this will be done in nfsd_readdir() */
829
830         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
831                 return nfserr_inval;
832
833         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
834         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
835         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
836
837         if ((cookie == 1) || (cookie == 2) ||
838             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
839                 return nfserr_bad_cookie;
840
841         readdir->rd_rqstp = rqstp;
842         readdir->rd_fhp = &cstate->current_fh;
843         return nfs_ok;
844 }
845
846 static __be32
847 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
848                union nfsd4_op_u *u)
849 {
850         u->readlink.rl_rqstp = rqstp;
851         u->readlink.rl_fhp = &cstate->current_fh;
852         return nfs_ok;
853 }
854
855 static __be32
856 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
857              union nfsd4_op_u *u)
858 {
859         struct nfsd4_remove *remove = &u->remove;
860         __be32 status;
861
862         if (opens_in_grace(SVC_NET(rqstp)))
863                 return nfserr_grace;
864         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
865                              remove->rm_name, remove->rm_namelen);
866         if (!status) {
867                 fh_unlock(&cstate->current_fh);
868                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
869         }
870         return status;
871 }
872
873 static __be32
874 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
875              union nfsd4_op_u *u)
876 {
877         struct nfsd4_rename *rename = &u->rename;
878         __be32 status;
879
880         if (opens_in_grace(SVC_NET(rqstp)))
881                 return nfserr_grace;
882         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
883                              rename->rn_snamelen, &cstate->current_fh,
884                              rename->rn_tname, rename->rn_tnamelen);
885         if (status)
886                 return status;
887         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
888         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
889         return nfs_ok;
890 }
891
892 static __be32
893 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
894               union nfsd4_op_u *u)
895 {
896         struct nfsd4_secinfo *secinfo = &u->secinfo;
897         struct svc_export *exp;
898         struct dentry *dentry;
899         __be32 err;
900
901         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
902         if (err)
903                 return err;
904         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
905                                     secinfo->si_name, secinfo->si_namelen,
906                                     &exp, &dentry);
907         if (err)
908                 return err;
909         fh_unlock(&cstate->current_fh);
910         if (d_really_is_negative(dentry)) {
911                 exp_put(exp);
912                 err = nfserr_noent;
913         } else
914                 secinfo->si_exp = exp;
915         dput(dentry);
916         if (cstate->minorversion)
917                 /* See rfc 5661 section 2.6.3.1.1.8 */
918                 fh_put(&cstate->current_fh);
919         return err;
920 }
921
922 static __be32
923 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
924                 union nfsd4_op_u *u)
925 {
926         __be32 err;
927
928         switch (u->secinfo_no_name.sin_style) {
929         case NFS4_SECINFO_STYLE4_CURRENT_FH:
930                 break;
931         case NFS4_SECINFO_STYLE4_PARENT:
932                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
933                 if (err)
934                         return err;
935                 break;
936         default:
937                 return nfserr_inval;
938         }
939
940         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
941         fh_put(&cstate->current_fh);
942         return nfs_ok;
943 }
944
945 static void
946 nfsd4_secinfo_release(union nfsd4_op_u *u)
947 {
948         if (u->secinfo.si_exp)
949                 exp_put(u->secinfo.si_exp);
950 }
951
952 static void
953 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
954 {
955         if (u->secinfo_no_name.sin_exp)
956                 exp_put(u->secinfo_no_name.sin_exp);
957 }
958
959 static __be32
960 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
961               union nfsd4_op_u *u)
962 {
963         struct nfsd4_setattr *setattr = &u->setattr;
964         __be32 status = nfs_ok;
965         int err;
966
967         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
968                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
969                                 &cstate->current_fh, &setattr->sa_stateid,
970                                 WR_STATE, NULL, NULL);
971                 if (status) {
972                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
973                         return status;
974                 }
975         }
976         err = fh_want_write(&cstate->current_fh);
977         if (err)
978                 return nfserrno(err);
979         status = nfs_ok;
980
981         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
982                                     nfsd_attrmask);
983         if (status)
984                 goto out;
985
986         if (setattr->sa_acl != NULL)
987                 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
988                                             setattr->sa_acl);
989         if (status)
990                 goto out;
991         if (setattr->sa_label.len)
992                 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
993                                 &setattr->sa_label);
994         if (status)
995                 goto out;
996         status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
997                                 0, (time64_t)0);
998 out:
999         fh_drop_write(&cstate->current_fh);
1000         return status;
1001 }
1002
1003 static __be32
1004 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1005             union nfsd4_op_u *u)
1006 {
1007         struct nfsd4_write *write = &u->write;
1008         stateid_t *stateid = &write->wr_stateid;
1009         struct nfsd_file *nf = NULL;
1010         __be32 status = nfs_ok;
1011         unsigned long cnt;
1012         int nvecs;
1013
1014         if (write->wr_offset >= OFFSET_MAX)
1015                 return nfserr_inval;
1016
1017         cnt = write->wr_buflen;
1018         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1019                                write->wr_offset, cnt);
1020         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1021                                                 stateid, WR_STATE, &nf, NULL);
1022         if (status) {
1023                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1024                 return status;
1025         }
1026
1027         write->wr_how_written = write->wr_stable_how;
1028
1029         nvecs = svc_fill_write_vector(rqstp, write->wr_payload.pages,
1030                                       write->wr_payload.head, write->wr_buflen);
1031         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1032
1033         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1034                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1035                                 write->wr_how_written,
1036                                 (__be32 *)write->wr_verifier.data);
1037         nfsd_file_put(nf);
1038
1039         write->wr_bytes_written = cnt;
1040         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1041                               write->wr_offset, cnt);
1042         return status;
1043 }
1044
1045 static __be32
1046 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1047                   stateid_t *src_stateid, struct nfsd_file **src,
1048                   stateid_t *dst_stateid, struct nfsd_file **dst)
1049 {
1050         __be32 status;
1051
1052         if (!cstate->save_fh.fh_dentry)
1053                 return nfserr_nofilehandle;
1054
1055         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1056                                             src_stateid, RD_STATE, src, NULL);
1057         if (status) {
1058                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1059                 goto out;
1060         }
1061
1062         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1063                                             dst_stateid, WR_STATE, dst, NULL);
1064         if (status) {
1065                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1066                 goto out_put_src;
1067         }
1068
1069         /* fix up for NFS-specific error code */
1070         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1071             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1072                 status = nfserr_wrong_type;
1073                 goto out_put_dst;
1074         }
1075
1076 out:
1077         return status;
1078 out_put_dst:
1079         nfsd_file_put(*dst);
1080 out_put_src:
1081         nfsd_file_put(*src);
1082         goto out;
1083 }
1084
1085 static __be32
1086 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1087                 union nfsd4_op_u *u)
1088 {
1089         struct nfsd4_clone *clone = &u->clone;
1090         struct nfsd_file *src, *dst;
1091         __be32 status;
1092
1093         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1094                                    &clone->cl_dst_stateid, &dst);
1095         if (status)
1096                 goto out;
1097
1098         status = nfsd4_clone_file_range(src, clone->cl_src_pos,
1099                         dst, clone->cl_dst_pos, clone->cl_count,
1100                         EX_ISSYNC(cstate->current_fh.fh_export));
1101
1102         nfsd_file_put(dst);
1103         nfsd_file_put(src);
1104 out:
1105         return status;
1106 }
1107
1108 void nfs4_put_copy(struct nfsd4_copy *copy)
1109 {
1110         if (!refcount_dec_and_test(&copy->refcount))
1111                 return;
1112         kfree(copy);
1113 }
1114
1115 static bool
1116 check_and_set_stop_copy(struct nfsd4_copy *copy)
1117 {
1118         bool value;
1119
1120         spin_lock(&copy->cp_clp->async_lock);
1121         value = copy->stopped;
1122         if (!copy->stopped)
1123                 copy->stopped = true;
1124         spin_unlock(&copy->cp_clp->async_lock);
1125         return value;
1126 }
1127
1128 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1129 {
1130         /* only 1 thread should stop the copy */
1131         if (!check_and_set_stop_copy(copy))
1132                 kthread_stop(copy->copy_task);
1133         nfs4_put_copy(copy);
1134 }
1135
1136 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1137 {
1138         struct nfsd4_copy *copy = NULL;
1139
1140         spin_lock(&clp->async_lock);
1141         if (!list_empty(&clp->async_copies)) {
1142                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1143                                         copies);
1144                 refcount_inc(&copy->refcount);
1145         }
1146         spin_unlock(&clp->async_lock);
1147         return copy;
1148 }
1149
1150 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1151 {
1152         struct nfsd4_copy *copy;
1153
1154         while ((copy = nfsd4_get_copy(clp)) != NULL)
1155                 nfsd4_stop_copy(copy);
1156 }
1157 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1158
1159 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1160                                    struct nfs_fh *src_fh,
1161                                    nfs4_stateid *stateid);
1162 extern void nfs42_ssc_close(struct file *filep);
1163
1164 extern void nfs_sb_deactive(struct super_block *sb);
1165
1166 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1167
1168 /*
1169  * Support one copy source server for now.
1170  */
1171 static __be32
1172 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1173                        struct vfsmount **mount)
1174 {
1175         struct file_system_type *type;
1176         struct vfsmount *ss_mnt;
1177         struct nfs42_netaddr *naddr;
1178         struct sockaddr_storage tmp_addr;
1179         size_t tmp_addrlen, match_netid_len = 3;
1180         char *startsep = "", *endsep = "", *match_netid = "tcp";
1181         char *ipaddr, *dev_name, *raw_data;
1182         int len, raw_len;
1183         __be32 status = nfserr_inval;
1184
1185         naddr = &nss->u.nl4_addr;
1186         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1187                                          naddr->addr_len,
1188                                          (struct sockaddr *)&tmp_addr,
1189                                          sizeof(tmp_addr));
1190         if (tmp_addrlen == 0)
1191                 goto out_err;
1192
1193         if (tmp_addr.ss_family == AF_INET6) {
1194                 startsep = "[";
1195                 endsep = "]";
1196                 match_netid = "tcp6";
1197                 match_netid_len = 4;
1198         }
1199
1200         if (naddr->netid_len != match_netid_len ||
1201                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1202                 goto out_err;
1203
1204         /* Construct the raw data for the vfs_kern_mount call */
1205         len = RPC_MAX_ADDRBUFLEN + 1;
1206         ipaddr = kzalloc(len, GFP_KERNEL);
1207         if (!ipaddr)
1208                 goto out_err;
1209
1210         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1211
1212         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1213
1214         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1215         raw_data = kzalloc(raw_len, GFP_KERNEL);
1216         if (!raw_data)
1217                 goto out_free_ipaddr;
1218
1219         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1220
1221         status = nfserr_nodev;
1222         type = get_fs_type("nfs");
1223         if (!type)
1224                 goto out_free_rawdata;
1225
1226         /* Set the server:<export> for the vfs_kern_mount call */
1227         dev_name = kzalloc(len + 5, GFP_KERNEL);
1228         if (!dev_name)
1229                 goto out_free_rawdata;
1230         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1231
1232         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1233         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1234         module_put(type->owner);
1235         if (IS_ERR(ss_mnt))
1236                 goto out_free_devname;
1237
1238         status = 0;
1239         *mount = ss_mnt;
1240
1241 out_free_devname:
1242         kfree(dev_name);
1243 out_free_rawdata:
1244         kfree(raw_data);
1245 out_free_ipaddr:
1246         kfree(ipaddr);
1247 out_err:
1248         return status;
1249 }
1250
1251 static void
1252 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1253 {
1254         nfs_do_sb_deactive(ss_mnt->mnt_sb);
1255         mntput(ss_mnt);
1256 }
1257
1258 /*
1259  * Verify COPY destination stateid.
1260  *
1261  * Connect to the source server with NFSv4.1.
1262  * Create the source struct file for nfsd_copy_range.
1263  * Called with COPY cstate:
1264  *    SAVED_FH: source filehandle
1265  *    CURRENT_FH: destination filehandle
1266  */
1267 static __be32
1268 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1269                       struct nfsd4_compound_state *cstate,
1270                       struct nfsd4_copy *copy, struct vfsmount **mount)
1271 {
1272         struct svc_fh *s_fh = NULL;
1273         stateid_t *s_stid = &copy->cp_src_stateid;
1274         __be32 status = nfserr_inval;
1275
1276         /* Verify the destination stateid and set dst struct file*/
1277         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1278                                             &copy->cp_dst_stateid,
1279                                             WR_STATE, &copy->nf_dst, NULL);
1280         if (status)
1281                 goto out;
1282
1283         status = nfsd4_interssc_connect(&copy->cp_src, rqstp, mount);
1284         if (status)
1285                 goto out;
1286
1287         s_fh = &cstate->save_fh;
1288
1289         copy->c_fh.size = s_fh->fh_handle.fh_size;
1290         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_base, copy->c_fh.size);
1291         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1292         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1293                sizeof(stateid_opaque_t));
1294
1295         status = 0;
1296 out:
1297         return status;
1298 }
1299
1300 static void
1301 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1302                         struct nfsd_file *dst)
1303 {
1304         nfs42_ssc_close(src->nf_file);
1305         fput(src->nf_file);
1306         nfsd_file_put(dst);
1307         mntput(ss_mnt);
1308 }
1309
1310 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1311
1312 static __be32
1313 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1314                       struct nfsd4_compound_state *cstate,
1315                       struct nfsd4_copy *copy,
1316                       struct vfsmount **mount)
1317 {
1318         *mount = NULL;
1319         return nfserr_inval;
1320 }
1321
1322 static void
1323 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1324                         struct nfsd_file *dst)
1325 {
1326 }
1327
1328 static void
1329 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1330 {
1331 }
1332
1333 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1334                                    struct nfs_fh *src_fh,
1335                                    nfs4_stateid *stateid)
1336 {
1337         return NULL;
1338 }
1339 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1340
1341 static __be32
1342 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1343                       struct nfsd4_compound_state *cstate,
1344                       struct nfsd4_copy *copy)
1345 {
1346         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1347                                  &copy->nf_src, &copy->cp_dst_stateid,
1348                                  &copy->nf_dst);
1349 }
1350
1351 static void
1352 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1353 {
1354         nfsd_file_put(src);
1355         nfsd_file_put(dst);
1356 }
1357
1358 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1359 {
1360         struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1361
1362         nfs4_put_copy(copy);
1363 }
1364
1365 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1366                                  struct rpc_task *task)
1367 {
1368         return 1;
1369 }
1370
1371 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1372         .release = nfsd4_cb_offload_release,
1373         .done = nfsd4_cb_offload_done
1374 };
1375
1376 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1377 {
1378         copy->cp_res.wr_stable_how = NFS_UNSTABLE;
1379         copy->cp_synchronous = sync;
1380         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1381 }
1382
1383 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1384 {
1385         ssize_t bytes_copied = 0;
1386         u64 bytes_total = copy->cp_count;
1387         u64 src_pos = copy->cp_src_pos;
1388         u64 dst_pos = copy->cp_dst_pos;
1389
1390         /* See RFC 7862 p.67: */
1391         if (bytes_total == 0)
1392                 bytes_total = ULLONG_MAX;
1393         do {
1394                 if (kthread_should_stop())
1395                         break;
1396                 bytes_copied = nfsd_copy_file_range(copy->nf_src->nf_file,
1397                                 src_pos, copy->nf_dst->nf_file, dst_pos,
1398                                 bytes_total);
1399                 if (bytes_copied <= 0)
1400                         break;
1401                 bytes_total -= bytes_copied;
1402                 copy->cp_res.wr_bytes_written += bytes_copied;
1403                 src_pos += bytes_copied;
1404                 dst_pos += bytes_copied;
1405         } while (bytes_total > 0 && !copy->cp_synchronous);
1406         return bytes_copied;
1407 }
1408
1409 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1410 {
1411         __be32 status;
1412         ssize_t bytes;
1413
1414         bytes = _nfsd_copy_file_range(copy);
1415         /* for async copy, we ignore the error, client can always retry
1416          * to get the error
1417          */
1418         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1419                 status = nfserrno(bytes);
1420         else {
1421                 nfsd4_init_copy_res(copy, sync);
1422                 status = nfs_ok;
1423         }
1424
1425         if (!copy->cp_intra) /* Inter server SSC */
1426                 nfsd4_cleanup_inter_ssc(copy->ss_mnt, copy->nf_src,
1427                                         copy->nf_dst);
1428         else
1429                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1430
1431         return status;
1432 }
1433
1434 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1435 {
1436         dst->cp_src_pos = src->cp_src_pos;
1437         dst->cp_dst_pos = src->cp_dst_pos;
1438         dst->cp_count = src->cp_count;
1439         dst->cp_synchronous = src->cp_synchronous;
1440         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1441         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1442         dst->cp_clp = src->cp_clp;
1443         dst->nf_dst = nfsd_file_get(src->nf_dst);
1444         dst->cp_intra = src->cp_intra;
1445         if (src->cp_intra) /* for inter, file_src doesn't exist yet */
1446                 dst->nf_src = nfsd_file_get(src->nf_src);
1447
1448         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1449         memcpy(&dst->cp_src, &src->cp_src, sizeof(struct nl4_server));
1450         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1451         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1452         dst->ss_mnt = src->ss_mnt;
1453 }
1454
1455 static void cleanup_async_copy(struct nfsd4_copy *copy)
1456 {
1457         nfs4_free_copy_state(copy);
1458         nfsd_file_put(copy->nf_dst);
1459         if (copy->cp_intra)
1460                 nfsd_file_put(copy->nf_src);
1461         spin_lock(&copy->cp_clp->async_lock);
1462         list_del(&copy->copies);
1463         spin_unlock(&copy->cp_clp->async_lock);
1464         nfs4_put_copy(copy);
1465 }
1466
1467 static int nfsd4_do_async_copy(void *data)
1468 {
1469         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1470         struct nfsd4_copy *cb_copy;
1471
1472         if (!copy->cp_intra) { /* Inter server SSC */
1473                 copy->nf_src = kzalloc(sizeof(struct nfsd_file), GFP_KERNEL);
1474                 if (!copy->nf_src) {
1475                         copy->nfserr = nfserr_serverfault;
1476                         nfsd4_interssc_disconnect(copy->ss_mnt);
1477                         goto do_callback;
1478                 }
1479                 copy->nf_src->nf_file = nfs42_ssc_open(copy->ss_mnt, &copy->c_fh,
1480                                               &copy->stateid);
1481                 if (IS_ERR(copy->nf_src->nf_file)) {
1482                         copy->nfserr = nfserr_offload_denied;
1483                         nfsd4_interssc_disconnect(copy->ss_mnt);
1484                         goto do_callback;
1485                 }
1486         }
1487
1488         copy->nfserr = nfsd4_do_copy(copy, 0);
1489 do_callback:
1490         cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1491         if (!cb_copy)
1492                 goto out;
1493         refcount_set(&cb_copy->refcount, 1);
1494         memcpy(&cb_copy->cp_res, &copy->cp_res, sizeof(copy->cp_res));
1495         cb_copy->cp_clp = copy->cp_clp;
1496         cb_copy->nfserr = copy->nfserr;
1497         memcpy(&cb_copy->fh, &copy->fh, sizeof(copy->fh));
1498         nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1499                         &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1500         nfsd4_run_cb(&cb_copy->cp_cb);
1501 out:
1502         if (!copy->cp_intra)
1503                 kfree(copy->nf_src);
1504         cleanup_async_copy(copy);
1505         return 0;
1506 }
1507
1508 static __be32
1509 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1510                 union nfsd4_op_u *u)
1511 {
1512         struct nfsd4_copy *copy = &u->copy;
1513         __be32 status;
1514         struct nfsd4_copy *async_copy = NULL;
1515
1516         if (!copy->cp_intra) { /* Inter server SSC */
1517                 if (!inter_copy_offload_enable || copy->cp_synchronous) {
1518                         status = nfserr_notsupp;
1519                         goto out;
1520                 }
1521                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1522                                 &copy->ss_mnt);
1523                 if (status)
1524                         return nfserr_offload_denied;
1525         } else {
1526                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1527                 if (status)
1528                         return status;
1529         }
1530
1531         copy->cp_clp = cstate->clp;
1532         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1533                 sizeof(struct knfsd_fh));
1534         if (!copy->cp_synchronous) {
1535                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1536
1537                 status = nfserrno(-ENOMEM);
1538                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1539                 if (!async_copy)
1540                         goto out_err;
1541                 if (!nfs4_init_copy_state(nn, copy))
1542                         goto out_err;
1543                 refcount_set(&async_copy->refcount, 1);
1544                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.stid,
1545                         sizeof(copy->cp_res.cb_stateid));
1546                 dup_copy_fields(copy, async_copy);
1547                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1548                                 async_copy, "%s", "copy thread");
1549                 if (IS_ERR(async_copy->copy_task))
1550                         goto out_err;
1551                 spin_lock(&async_copy->cp_clp->async_lock);
1552                 list_add(&async_copy->copies,
1553                                 &async_copy->cp_clp->async_copies);
1554                 spin_unlock(&async_copy->cp_clp->async_lock);
1555                 wake_up_process(async_copy->copy_task);
1556                 status = nfs_ok;
1557         } else {
1558                 status = nfsd4_do_copy(copy, 1);
1559         }
1560 out:
1561         return status;
1562 out_err:
1563         if (async_copy)
1564                 cleanup_async_copy(async_copy);
1565         status = nfserrno(-ENOMEM);
1566         if (!copy->cp_intra)
1567                 nfsd4_interssc_disconnect(copy->ss_mnt);
1568         goto out;
1569 }
1570
1571 struct nfsd4_copy *
1572 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1573 {
1574         struct nfsd4_copy *copy;
1575
1576         spin_lock(&clp->async_lock);
1577         list_for_each_entry(copy, &clp->async_copies, copies) {
1578                 if (memcmp(&copy->cp_stateid.stid, stateid, NFS4_STATEID_SIZE))
1579                         continue;
1580                 refcount_inc(&copy->refcount);
1581                 spin_unlock(&clp->async_lock);
1582                 return copy;
1583         }
1584         spin_unlock(&clp->async_lock);
1585         return NULL;
1586 }
1587
1588 static __be32
1589 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1590                      struct nfsd4_compound_state *cstate,
1591                      union nfsd4_op_u *u)
1592 {
1593         struct nfsd4_offload_status *os = &u->offload_status;
1594         struct nfsd4_copy *copy;
1595         struct nfs4_client *clp = cstate->clp;
1596
1597         copy = find_async_copy(clp, &os->stateid);
1598         if (!copy) {
1599                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1600
1601                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1602         } else
1603                 nfsd4_stop_copy(copy);
1604
1605         return nfs_ok;
1606 }
1607
1608 static __be32
1609 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1610                   union nfsd4_op_u *u)
1611 {
1612         struct nfsd4_copy_notify *cn = &u->copy_notify;
1613         __be32 status;
1614         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1615         struct nfs4_stid *stid;
1616         struct nfs4_cpntf_state *cps;
1617         struct nfs4_client *clp = cstate->clp;
1618
1619         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1620                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1621                                         &stid);
1622         if (status)
1623                 return status;
1624
1625         cn->cpn_sec = nn->nfsd4_lease;
1626         cn->cpn_nsec = 0;
1627
1628         status = nfserrno(-ENOMEM);
1629         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1630         if (!cps)
1631                 goto out;
1632         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.stid, sizeof(stateid_t));
1633         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1634         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1635
1636         /* For now, only return one server address in cpn_src, the
1637          * address used by the client to connect to this server.
1638          */
1639         cn->cpn_src.nl4_type = NL4_NETADDR;
1640         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1641                                  &cn->cpn_src.u.nl4_addr);
1642         WARN_ON_ONCE(status);
1643         if (status) {
1644                 nfs4_put_cpntf_state(nn, cps);
1645                 goto out;
1646         }
1647 out:
1648         nfs4_put_stid(stid);
1649         return status;
1650 }
1651
1652 static __be32
1653 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1654                 struct nfsd4_fallocate *fallocate, int flags)
1655 {
1656         __be32 status;
1657         struct nfsd_file *nf;
1658
1659         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1660                                             &fallocate->falloc_stateid,
1661                                             WR_STATE, &nf, NULL);
1662         if (status != nfs_ok) {
1663                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1664                 return status;
1665         }
1666
1667         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1668                                      fallocate->falloc_offset,
1669                                      fallocate->falloc_length,
1670                                      flags);
1671         nfsd_file_put(nf);
1672         return status;
1673 }
1674 static __be32
1675 nfsd4_offload_status(struct svc_rqst *rqstp,
1676                      struct nfsd4_compound_state *cstate,
1677                      union nfsd4_op_u *u)
1678 {
1679         struct nfsd4_offload_status *os = &u->offload_status;
1680         __be32 status = 0;
1681         struct nfsd4_copy *copy;
1682         struct nfs4_client *clp = cstate->clp;
1683
1684         copy = find_async_copy(clp, &os->stateid);
1685         if (copy) {
1686                 os->count = copy->cp_res.wr_bytes_written;
1687                 nfs4_put_copy(copy);
1688         } else
1689                 status = nfserr_bad_stateid;
1690
1691         return status;
1692 }
1693
1694 static __be32
1695 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1696                union nfsd4_op_u *u)
1697 {
1698         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1699 }
1700
1701 static __be32
1702 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1703                  union nfsd4_op_u *u)
1704 {
1705         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1706                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1707 }
1708
1709 static __be32
1710 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1711            union nfsd4_op_u *u)
1712 {
1713         struct nfsd4_seek *seek = &u->seek;
1714         int whence;
1715         __be32 status;
1716         struct nfsd_file *nf;
1717
1718         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1719                                             &seek->seek_stateid,
1720                                             RD_STATE, &nf, NULL);
1721         if (status) {
1722                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1723                 return status;
1724         }
1725
1726         switch (seek->seek_whence) {
1727         case NFS4_CONTENT_DATA:
1728                 whence = SEEK_DATA;
1729                 break;
1730         case NFS4_CONTENT_HOLE:
1731                 whence = SEEK_HOLE;
1732                 break;
1733         default:
1734                 status = nfserr_union_notsupp;
1735                 goto out;
1736         }
1737
1738         /*
1739          * Note:  This call does change file->f_pos, but nothing in NFSD
1740          *        should ever file->f_pos.
1741          */
1742         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
1743         if (seek->seek_pos < 0)
1744                 status = nfserrno(seek->seek_pos);
1745         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
1746                 seek->seek_eof = true;
1747
1748 out:
1749         nfsd_file_put(nf);
1750         return status;
1751 }
1752
1753 /* This routine never returns NFS_OK!  If there are no other errors, it
1754  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1755  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
1756  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1757  */
1758 static __be32
1759 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1760              struct nfsd4_verify *verify)
1761 {
1762         __be32 *buf, *p;
1763         int count;
1764         __be32 status;
1765
1766         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1767         if (status)
1768                 return status;
1769
1770         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1771         if (status)
1772                 return status;
1773
1774         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1775             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1776                 return nfserr_inval;
1777         if (verify->ve_attrlen & 3)
1778                 return nfserr_inval;
1779
1780         /* count in words:
1781          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1782          */
1783         count = 4 + (verify->ve_attrlen >> 2);
1784         buf = kmalloc(count << 2, GFP_KERNEL);
1785         if (!buf)
1786                 return nfserr_jukebox;
1787
1788         p = buf;
1789         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1790                                     cstate->current_fh.fh_export,
1791                                     cstate->current_fh.fh_dentry,
1792                                     verify->ve_bmval,
1793                                     rqstp, 0);
1794         /*
1795          * If nfsd4_encode_fattr() ran out of space, assume that's because
1796          * the attributes are longer (hence different) than those given:
1797          */
1798         if (status == nfserr_resource)
1799                 status = nfserr_not_same;
1800         if (status)
1801                 goto out_kfree;
1802
1803         /* skip bitmap */
1804         p = buf + 1 + ntohl(buf[0]);
1805         status = nfserr_not_same;
1806         if (ntohl(*p++) != verify->ve_attrlen)
1807                 goto out_kfree;
1808         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1809                 status = nfserr_same;
1810
1811 out_kfree:
1812         kfree(buf);
1813         return status;
1814 }
1815
1816 static __be32
1817 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1818               union nfsd4_op_u *u)
1819 {
1820         __be32 status;
1821
1822         status = _nfsd4_verify(rqstp, cstate, &u->verify);
1823         return status == nfserr_not_same ? nfs_ok : status;
1824 }
1825
1826 static __be32
1827 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1828              union nfsd4_op_u *u)
1829 {
1830         __be32 status;
1831
1832         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1833         return status == nfserr_same ? nfs_ok : status;
1834 }
1835
1836 #ifdef CONFIG_NFSD_PNFS
1837 static const struct nfsd4_layout_ops *
1838 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1839 {
1840         if (!exp->ex_layout_types) {
1841                 dprintk("%s: export does not support pNFS\n", __func__);
1842                 return NULL;
1843         }
1844
1845         if (layout_type >= LAYOUT_TYPE_MAX ||
1846             !(exp->ex_layout_types & (1 << layout_type))) {
1847                 dprintk("%s: layout type %d not supported\n",
1848                         __func__, layout_type);
1849                 return NULL;
1850         }
1851
1852         return nfsd4_layout_ops[layout_type];
1853 }
1854
1855 static __be32
1856 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
1857                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1858 {
1859         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
1860         const struct nfsd4_layout_ops *ops;
1861         struct nfsd4_deviceid_map *map;
1862         struct svc_export *exp;
1863         __be32 nfserr;
1864
1865         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
1866                __func__,
1867                gdp->gd_layout_type,
1868                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
1869                gdp->gd_maxcount);
1870
1871         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
1872         if (!map) {
1873                 dprintk("%s: couldn't find device ID to export mapping!\n",
1874                         __func__);
1875                 return nfserr_noent;
1876         }
1877
1878         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
1879         if (IS_ERR(exp)) {
1880                 dprintk("%s: could not find device id\n", __func__);
1881                 return nfserr_noent;
1882         }
1883
1884         nfserr = nfserr_layoutunavailable;
1885         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
1886         if (!ops)
1887                 goto out;
1888
1889         nfserr = nfs_ok;
1890         if (gdp->gd_maxcount != 0) {
1891                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
1892                                 rqstp, cstate->clp, gdp);
1893         }
1894
1895         gdp->gd_notify_types &= ops->notify_types;
1896 out:
1897         exp_put(exp);
1898         return nfserr;
1899 }
1900
1901 static void
1902 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
1903 {
1904         kfree(u->getdeviceinfo.gd_device);
1905 }
1906
1907 static __be32
1908 nfsd4_layoutget(struct svc_rqst *rqstp,
1909                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1910 {
1911         struct nfsd4_layoutget *lgp = &u->layoutget;
1912         struct svc_fh *current_fh = &cstate->current_fh;
1913         const struct nfsd4_layout_ops *ops;
1914         struct nfs4_layout_stateid *ls;
1915         __be32 nfserr;
1916         int accmode = NFSD_MAY_READ_IF_EXEC;
1917
1918         switch (lgp->lg_seg.iomode) {
1919         case IOMODE_READ:
1920                 accmode |= NFSD_MAY_READ;
1921                 break;
1922         case IOMODE_RW:
1923                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
1924                 break;
1925         default:
1926                 dprintk("%s: invalid iomode %d\n",
1927                         __func__, lgp->lg_seg.iomode);
1928                 nfserr = nfserr_badiomode;
1929                 goto out;
1930         }
1931
1932         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
1933         if (nfserr)
1934                 goto out;
1935
1936         nfserr = nfserr_layoutunavailable;
1937         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
1938         if (!ops)
1939                 goto out;
1940
1941         /*
1942          * Verify minlength and range as per RFC5661:
1943          *  o  If loga_length is less than loga_minlength,
1944          *     the metadata server MUST return NFS4ERR_INVAL.
1945          *  o  If the sum of loga_offset and loga_minlength exceeds
1946          *     NFS4_UINT64_MAX, and loga_minlength is not
1947          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
1948          *  o  If the sum of loga_offset and loga_length exceeds
1949          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
1950          *     the error NFS4ERR_INVAL MUST result.
1951          */
1952         nfserr = nfserr_inval;
1953         if (lgp->lg_seg.length < lgp->lg_minlength ||
1954             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
1955              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
1956             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
1957              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
1958                 goto out;
1959         if (lgp->lg_seg.length == 0)
1960                 goto out;
1961
1962         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
1963                                                 true, lgp->lg_layout_type, &ls);
1964         if (nfserr) {
1965                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
1966                 goto out;
1967         }
1968
1969         nfserr = nfserr_recallconflict;
1970         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
1971                 goto out_put_stid;
1972
1973         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
1974                                      current_fh, lgp);
1975         if (nfserr)
1976                 goto out_put_stid;
1977
1978         nfserr = nfsd4_insert_layout(lgp, ls);
1979
1980 out_put_stid:
1981         mutex_unlock(&ls->ls_mutex);
1982         nfs4_put_stid(&ls->ls_stid);
1983 out:
1984         return nfserr;
1985 }
1986
1987 static void
1988 nfsd4_layoutget_release(union nfsd4_op_u *u)
1989 {
1990         kfree(u->layoutget.lg_content);
1991 }
1992
1993 static __be32
1994 nfsd4_layoutcommit(struct svc_rqst *rqstp,
1995                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1996 {
1997         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
1998         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
1999         struct svc_fh *current_fh = &cstate->current_fh;
2000         const struct nfsd4_layout_ops *ops;
2001         loff_t new_size = lcp->lc_last_wr + 1;
2002         struct inode *inode;
2003         struct nfs4_layout_stateid *ls;
2004         __be32 nfserr;
2005
2006         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2007         if (nfserr)
2008                 goto out;
2009
2010         nfserr = nfserr_layoutunavailable;
2011         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2012         if (!ops)
2013                 goto out;
2014         inode = d_inode(current_fh->fh_dentry);
2015
2016         nfserr = nfserr_inval;
2017         if (new_size <= seg->offset) {
2018                 dprintk("pnfsd: last write before layout segment\n");
2019                 goto out;
2020         }
2021         if (new_size > seg->offset + seg->length) {
2022                 dprintk("pnfsd: last write beyond layout segment\n");
2023                 goto out;
2024         }
2025         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2026                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2027                 goto out;
2028         }
2029
2030         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2031                                                 false, lcp->lc_layout_type,
2032                                                 &ls);
2033         if (nfserr) {
2034                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2035                 /* fixup error code as per RFC5661 */
2036                 if (nfserr == nfserr_bad_stateid)
2037                         nfserr = nfserr_badlayout;
2038                 goto out;
2039         }
2040
2041         /* LAYOUTCOMMIT does not require any serialization */
2042         mutex_unlock(&ls->ls_mutex);
2043
2044         if (new_size > i_size_read(inode)) {
2045                 lcp->lc_size_chg = 1;
2046                 lcp->lc_newsize = new_size;
2047         } else {
2048                 lcp->lc_size_chg = 0;
2049         }
2050
2051         nfserr = ops->proc_layoutcommit(inode, lcp);
2052         nfs4_put_stid(&ls->ls_stid);
2053 out:
2054         return nfserr;
2055 }
2056
2057 static __be32
2058 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2059                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2060 {
2061         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2062         struct svc_fh *current_fh = &cstate->current_fh;
2063         __be32 nfserr;
2064
2065         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2066         if (nfserr)
2067                 goto out;
2068
2069         nfserr = nfserr_layoutunavailable;
2070         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2071                 goto out;
2072
2073         switch (lrp->lr_seg.iomode) {
2074         case IOMODE_READ:
2075         case IOMODE_RW:
2076         case IOMODE_ANY:
2077                 break;
2078         default:
2079                 dprintk("%s: invalid iomode %d\n", __func__,
2080                         lrp->lr_seg.iomode);
2081                 nfserr = nfserr_inval;
2082                 goto out;
2083         }
2084
2085         switch (lrp->lr_return_type) {
2086         case RETURN_FILE:
2087                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2088                 break;
2089         case RETURN_FSID:
2090         case RETURN_ALL:
2091                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2092                 break;
2093         default:
2094                 dprintk("%s: invalid return_type %d\n", __func__,
2095                         lrp->lr_return_type);
2096                 nfserr = nfserr_inval;
2097                 break;
2098         }
2099 out:
2100         return nfserr;
2101 }
2102 #endif /* CONFIG_NFSD_PNFS */
2103
2104 static __be32
2105 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2106                union nfsd4_op_u *u)
2107 {
2108         struct nfsd4_getxattr *getxattr = &u->getxattr;
2109
2110         return nfsd_getxattr(rqstp, &cstate->current_fh,
2111                              getxattr->getxa_name, &getxattr->getxa_buf,
2112                              &getxattr->getxa_len);
2113 }
2114
2115 static __be32
2116 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2117            union nfsd4_op_u *u)
2118 {
2119         struct nfsd4_setxattr *setxattr = &u->setxattr;
2120         __be32 ret;
2121
2122         if (opens_in_grace(SVC_NET(rqstp)))
2123                 return nfserr_grace;
2124
2125         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2126                             setxattr->setxa_buf, setxattr->setxa_len,
2127                             setxattr->setxa_flags);
2128
2129         if (!ret)
2130                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2131
2132         return ret;
2133 }
2134
2135 static __be32
2136 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2137            union nfsd4_op_u *u)
2138 {
2139         /*
2140          * Get the entire list, then copy out only the user attributes
2141          * in the encode function.
2142          */
2143         return nfsd_listxattr(rqstp, &cstate->current_fh,
2144                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2145 }
2146
2147 static __be32
2148 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2149            union nfsd4_op_u *u)
2150 {
2151         struct nfsd4_removexattr *removexattr = &u->removexattr;
2152         __be32 ret;
2153
2154         if (opens_in_grace(SVC_NET(rqstp)))
2155                 return nfserr_grace;
2156
2157         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2158             removexattr->rmxa_name);
2159
2160         if (!ret)
2161                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2162
2163         return ret;
2164 }
2165
2166 /*
2167  * NULL call.
2168  */
2169 static __be32
2170 nfsd4_proc_null(struct svc_rqst *rqstp)
2171 {
2172         return rpc_success;
2173 }
2174
2175 static inline void nfsd4_increment_op_stats(u32 opnum)
2176 {
2177         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2178                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2179 }
2180
2181 static const struct nfsd4_operation nfsd4_ops[];
2182
2183 static const char *nfsd4_op_name(unsigned opnum);
2184
2185 /*
2186  * Enforce NFSv4.1 COMPOUND ordering rules:
2187  *
2188  * Also note, enforced elsewhere:
2189  *      - SEQUENCE other than as first op results in
2190  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2191  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2192  *        (Enforced in nfsd4_bind_conn_to_session().)
2193  *      - DESTROY_SESSION must be the final operation in a compound, if
2194  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2195  *        (Enforced in nfsd4_destroy_session().)
2196  */
2197 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2198 {
2199         struct nfsd4_op *first_op = &args->ops[0];
2200
2201         /* These ordering requirements don't apply to NFSv4.0: */
2202         if (args->minorversion == 0)
2203                 return nfs_ok;
2204         /* This is weird, but OK, not our problem: */
2205         if (args->opcnt == 0)
2206                 return nfs_ok;
2207         if (first_op->status == nfserr_op_illegal)
2208                 return nfs_ok;
2209         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2210                 return nfserr_op_not_in_session;
2211         if (first_op->opnum == OP_SEQUENCE)
2212                 return nfs_ok;
2213         /*
2214          * So first_op is something allowed outside a session, like
2215          * EXCHANGE_ID; but then it has to be the only op in the
2216          * compound:
2217          */
2218         if (args->opcnt != 1)
2219                 return nfserr_not_only_op;
2220         return nfs_ok;
2221 }
2222
2223 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2224 {
2225         return &nfsd4_ops[op->opnum];
2226 }
2227
2228 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2229 {
2230         if (op->opnum == OP_ILLEGAL)
2231                 return false;
2232         return OPDESC(op)->op_flags & OP_CACHEME;
2233 }
2234
2235 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2236 {
2237         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2238         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2239         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2240         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2241         const struct nfsd4_operation *thisd = OPDESC(this);
2242         const struct nfsd4_operation *nextd;
2243
2244         /*
2245          * Most ops check wronsec on our own; only the putfh-like ops
2246          * have special rules.
2247          */
2248         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2249                 return false;
2250         /*
2251          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2252          * put-filehandle operation if we're not going to use the
2253          * result:
2254          */
2255         if (argp->opcnt == resp->opcnt)
2256                 return false;
2257         if (next->opnum == OP_ILLEGAL)
2258                 return false;
2259         nextd = OPDESC(next);
2260         /*
2261          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2262          * errors themselves as necessary; others should check for them
2263          * now:
2264          */
2265         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2266 }
2267
2268 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2269 static void
2270 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2271 {
2272         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2273         struct nfsd4_copy *copy;
2274         struct nfsd4_putfh *putfh;
2275         int i;
2276
2277         /* traverse all operation and if it's a COPY compound, mark the
2278          * source filehandle to skip verification
2279          */
2280         for (i = 0; i < args->opcnt; i++) {
2281                 op = &args->ops[i];
2282                 if (op->opnum == OP_PUTFH)
2283                         current_op = op;
2284                 else if (op->opnum == OP_SAVEFH)
2285                         saved_op = current_op;
2286                 else if (op->opnum == OP_RESTOREFH)
2287                         current_op = saved_op;
2288                 else if (op->opnum == OP_COPY) {
2289                         copy = (struct nfsd4_copy *)&op->u;
2290                         if (!saved_op) {
2291                                 op->status = nfserr_nofilehandle;
2292                                 return;
2293                         }
2294                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2295                         if (!copy->cp_intra)
2296                                 putfh->no_verify = true;
2297                 }
2298         }
2299 }
2300 #else
2301 static void
2302 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2303 {
2304 }
2305 #endif
2306
2307 /*
2308  * COMPOUND call.
2309  */
2310 static __be32
2311 nfsd4_proc_compound(struct svc_rqst *rqstp)
2312 {
2313         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2314         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2315         struct nfsd4_op *op;
2316         struct nfsd4_compound_state *cstate = &resp->cstate;
2317         struct svc_fh *current_fh = &cstate->current_fh;
2318         struct svc_fh *save_fh = &cstate->save_fh;
2319         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2320         __be32          status;
2321
2322         resp->xdr = &rqstp->rq_res_stream;
2323
2324         /* reserve space for: NFS status code */
2325         xdr_reserve_space(resp->xdr, XDR_UNIT);
2326
2327         resp->tagp = resp->xdr->p;
2328         /* reserve space for: taglen, tag, and opcnt */
2329         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2330         resp->taglen = args->taglen;
2331         resp->tag = args->tag;
2332         resp->rqstp = rqstp;
2333         cstate->minorversion = args->minorversion;
2334         fh_init(current_fh, NFS4_FHSIZE);
2335         fh_init(save_fh, NFS4_FHSIZE);
2336         /*
2337          * Don't use the deferral mechanism for NFSv4; compounds make it
2338          * too hard to avoid non-idempotency problems.
2339          */
2340         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2341
2342         /*
2343          * According to RFC3010, this takes precedence over all other errors.
2344          */
2345         status = nfserr_minor_vers_mismatch;
2346         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2347                 goto out;
2348         status = nfserr_resource;
2349         if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
2350                 goto out;
2351
2352         status = nfs41_check_op_ordering(args);
2353         if (status) {
2354                 op = &args->ops[0];
2355                 op->status = status;
2356                 resp->opcnt = 1;
2357                 goto encode_op;
2358         }
2359         check_if_stalefh_allowed(args);
2360
2361         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2362
2363         trace_nfsd_compound(rqstp, args->opcnt);
2364         while (!status && resp->opcnt < args->opcnt) {
2365                 op = &args->ops[resp->opcnt++];
2366
2367                 /*
2368                  * The XDR decode routines may have pre-set op->status;
2369                  * for example, if there is a miscellaneous XDR error
2370                  * it will be set to nfserr_bad_xdr.
2371                  */
2372                 if (op->status) {
2373                         if (op->opnum == OP_OPEN)
2374                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2375                         goto encode_op;
2376                 }
2377                 if (!current_fh->fh_dentry &&
2378                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2379                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2380                                 op->status = nfserr_nofilehandle;
2381                                 goto encode_op;
2382                         }
2383                 } else if (current_fh->fh_export &&
2384                            current_fh->fh_export->ex_fslocs.migrated &&
2385                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2386                         op->status = nfserr_moved;
2387                         goto encode_op;
2388                 }
2389
2390                 fh_clear_wcc(current_fh);
2391
2392                 /* If op is non-idempotent */
2393                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2394                         /*
2395                          * Don't execute this op if we couldn't encode a
2396                          * succesful reply:
2397                          */
2398                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2399                         /*
2400                          * Plus if there's another operation, make sure
2401                          * we'll have space to at least encode an error:
2402                          */
2403                         if (resp->opcnt < args->opcnt)
2404                                 plen += COMPOUND_ERR_SLACK_SPACE;
2405                         op->status = nfsd4_check_resp_size(resp, plen);
2406                 }
2407
2408                 if (op->status)
2409                         goto encode_op;
2410
2411                 if (op->opdesc->op_get_currentstateid)
2412                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2413                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2414
2415                 /* Only from SEQUENCE */
2416                 if (cstate->status == nfserr_replay_cache) {
2417                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2418                         status = op->status;
2419                         goto out;
2420                 }
2421                 if (!op->status) {
2422                         if (op->opdesc->op_set_currentstateid)
2423                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2424
2425                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2426                                 clear_current_stateid(cstate);
2427
2428                         if (current_fh->fh_export &&
2429                                         need_wrongsec_check(rqstp))
2430                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2431                 }
2432 encode_op:
2433                 if (op->status == nfserr_replay_me) {
2434                         op->replay = &cstate->replay_owner->so_replay;
2435                         nfsd4_encode_replay(resp->xdr, op);
2436                         status = op->status = op->replay->rp_status;
2437                 } else {
2438                         nfsd4_encode_operation(resp, op);
2439                         status = op->status;
2440                 }
2441
2442                 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2443                                            nfsd4_op_name(op->opnum));
2444
2445                 nfsd4_cstate_clear_replay(cstate);
2446                 nfsd4_increment_op_stats(op->opnum);
2447         }
2448
2449         fh_put(current_fh);
2450         fh_put(save_fh);
2451         BUG_ON(cstate->replay_owner);
2452 out:
2453         cstate->status = status;
2454         /* Reset deferral mechanism for RPC deferrals */
2455         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2456         return rpc_success;
2457 }
2458
2459 #define op_encode_hdr_size              (2)
2460 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2461 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2462 #define op_encode_change_info_maxsz     (5)
2463 #define nfs4_fattr_bitmap_maxsz         (4)
2464
2465 /* We'll fall back on returning no lockowner if run out of space: */
2466 #define op_encode_lockowner_maxsz       (0)
2467 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2468
2469 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2470
2471 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2472 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2473                                          op_encode_ace_maxsz)
2474
2475 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2476
2477 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2478 {
2479         return (op_encode_hdr_size) * sizeof(__be32);
2480 }
2481
2482 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2483 {
2484         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2485 }
2486
2487 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2488 {
2489         /* ac_supported, ac_resp_access */
2490         return (op_encode_hdr_size + 2)* sizeof(__be32);
2491 }
2492
2493 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2494 {
2495         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2496 }
2497
2498 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2499 {
2500         return (op_encode_hdr_size + op_encode_change_info_maxsz
2501                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2502 }
2503
2504 /*
2505  * Note since this is an idempotent operation we won't insist on failing
2506  * the op prematurely if the estimate is too large.  We may turn off splice
2507  * reads unnecessarily.
2508  */
2509 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2510                                       struct nfsd4_op *op)
2511 {
2512         u32 *bmap = op->u.getattr.ga_bmval;
2513         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2514         u32 ret = 0;
2515
2516         if (bmap0 & FATTR4_WORD0_ACL)
2517                 return svc_max_payload(rqstp);
2518         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2519                 return svc_max_payload(rqstp);
2520
2521         if (bmap1 & FATTR4_WORD1_OWNER) {
2522                 ret += IDMAP_NAMESZ + 4;
2523                 bmap1 &= ~FATTR4_WORD1_OWNER;
2524         }
2525         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2526                 ret += IDMAP_NAMESZ + 4;
2527                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2528         }
2529         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2530                 ret += NFS4_FHSIZE + 4;
2531                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2532         }
2533         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2534                 ret += NFS4_MAXLABELLEN + 12;
2535                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2536         }
2537         /*
2538          * Largest of remaining attributes are 16 bytes (e.g.,
2539          * supported_attributes)
2540          */
2541         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2542         /* bitmask, length */
2543         ret += 20;
2544         return ret;
2545 }
2546
2547 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2548 {
2549         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2550 }
2551
2552 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2553 {
2554         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2555                 * sizeof(__be32);
2556 }
2557
2558 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2559 {
2560         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2561                 * sizeof(__be32);
2562 }
2563
2564 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2565 {
2566         return (op_encode_hdr_size + op_encode_stateid_maxsz
2567                 + op_encode_change_info_maxsz + 1
2568                 + nfs4_fattr_bitmap_maxsz
2569                 + op_encode_delegation_maxsz) * sizeof(__be32);
2570 }
2571
2572 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2573 {
2574         u32 maxcount = 0, rlen = 0;
2575
2576         maxcount = svc_max_payload(rqstp);
2577         rlen = min(op->u.read.rd_length, maxcount);
2578
2579         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2580 }
2581
2582 static inline u32 nfsd4_read_plus_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2583 {
2584         u32 maxcount = svc_max_payload(rqstp);
2585         u32 rlen = min(op->u.read.rd_length, maxcount);
2586         /*
2587          * If we detect that the file changed during hole encoding, then we
2588          * recover by encoding the remaining reply as data. This means we need
2589          * to set aside enough room to encode two data segments.
2590          */
2591         u32 seg_len = 2 * (1 + 2 + 1);
2592
2593         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2594 }
2595
2596 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2597 {
2598         u32 maxcount = 0, rlen = 0;
2599
2600         maxcount = svc_max_payload(rqstp);
2601         rlen = min(op->u.readdir.rd_maxcount, maxcount);
2602
2603         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2604                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2605 }
2606
2607 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2608 {
2609         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2610 }
2611
2612 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2613 {
2614         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2615                 * sizeof(__be32);
2616 }
2617
2618 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2619 {
2620         return (op_encode_hdr_size + op_encode_change_info_maxsz
2621                 + op_encode_change_info_maxsz) * sizeof(__be32);
2622 }
2623
2624 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2625                                        struct nfsd4_op *op)
2626 {
2627         return (op_encode_hdr_size
2628                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2629 }
2630
2631 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2632 {
2633         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2634                 * sizeof(__be32);
2635 }
2636
2637 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2638 {
2639         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2640 }
2641
2642 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2643 {
2644         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2645                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2646 }
2647
2648 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2649 {
2650         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2651                                                                 sizeof(__be32);
2652 }
2653
2654 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2655 {
2656         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2657 }
2658
2659 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2660 {
2661         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2662                 1 + 1 + /* eir_flags, spr_how */\
2663                 4 + /* spo_must_enforce & _allow with bitmap */\
2664                 2 + /*eir_server_owner.so_minor_id */\
2665                 /* eir_server_owner.so_major_id<> */\
2666                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2667                 /* eir_server_scope<> */\
2668                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2669                 1 + /* eir_server_impl_id array length */\
2670                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2671 }
2672
2673 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2674 {
2675         return (op_encode_hdr_size + \
2676                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2677                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2678 }
2679
2680 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2681 {
2682         return (op_encode_hdr_size + \
2683                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2684                 2 + /* csr_sequence, csr_flags */\
2685                 op_encode_channel_attrs_maxsz + \
2686                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2687 }
2688
2689 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2690 {
2691         return (op_encode_hdr_size +
2692                 1 /* wr_callback */ +
2693                 op_encode_stateid_maxsz /* wr_callback */ +
2694                 2 /* wr_count */ +
2695                 1 /* wr_committed */ +
2696                 op_encode_verifier_maxsz +
2697                 1 /* cr_consecutive */ +
2698                 1 /* cr_synchronous */) * sizeof(__be32);
2699 }
2700
2701 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2702                                              struct nfsd4_op *op)
2703 {
2704         return (op_encode_hdr_size +
2705                 2 /* osr_count */ +
2706                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2707 }
2708
2709 static inline u32 nfsd4_copy_notify_rsize(struct svc_rqst *rqstp,
2710                                         struct nfsd4_op *op)
2711 {
2712         return (op_encode_hdr_size +
2713                 3 /* cnr_lease_time */ +
2714                 1 /* We support one cnr_source_server */ +
2715                 1 /* cnr_stateid seq */ +
2716                 op_encode_stateid_maxsz /* cnr_stateid */ +
2717                 1 /* num cnr_source_server*/ +
2718                 1 /* nl4_type */ +
2719                 1 /* nl4 size */ +
2720                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
2721                 * sizeof(__be32);
2722 }
2723
2724 #ifdef CONFIG_NFSD_PNFS
2725 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2726 {
2727         u32 maxcount = 0, rlen = 0;
2728
2729         maxcount = svc_max_payload(rqstp);
2730         rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2731
2732         return (op_encode_hdr_size +
2733                 1 /* gd_layout_type*/ +
2734                 XDR_QUADLEN(rlen) +
2735                 2 /* gd_notify_types */) * sizeof(__be32);
2736 }
2737
2738 /*
2739  * At this stage we don't really know what layout driver will handle the request,
2740  * so we need to define an arbitrary upper bound here.
2741  */
2742 #define MAX_LAYOUT_SIZE         128
2743 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2744 {
2745         return (op_encode_hdr_size +
2746                 1 /* logr_return_on_close */ +
2747                 op_encode_stateid_maxsz +
2748                 1 /* nr of layouts */ +
2749                 MAX_LAYOUT_SIZE) * sizeof(__be32);
2750 }
2751
2752 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2753 {
2754         return (op_encode_hdr_size +
2755                 1 /* locr_newsize */ +
2756                 2 /* ns_size */) * sizeof(__be32);
2757 }
2758
2759 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2760 {
2761         return (op_encode_hdr_size +
2762                 1 /* lrs_stateid */ +
2763                 op_encode_stateid_maxsz) * sizeof(__be32);
2764 }
2765 #endif /* CONFIG_NFSD_PNFS */
2766
2767
2768 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2769 {
2770         return (op_encode_hdr_size + 3) * sizeof(__be32);
2771 }
2772
2773 static inline u32 nfsd4_getxattr_rsize(struct svc_rqst *rqstp,
2774                                        struct nfsd4_op *op)
2775 {
2776         u32 maxcount, rlen;
2777
2778         maxcount = svc_max_payload(rqstp);
2779         rlen = min_t(u32, XATTR_SIZE_MAX, maxcount);
2780
2781         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2782 }
2783
2784 static inline u32 nfsd4_setxattr_rsize(struct svc_rqst *rqstp,
2785                                        struct nfsd4_op *op)
2786 {
2787         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2788                 * sizeof(__be32);
2789 }
2790 static inline u32 nfsd4_listxattrs_rsize(struct svc_rqst *rqstp,
2791                                          struct nfsd4_op *op)
2792 {
2793         u32 maxcount, rlen;
2794
2795         maxcount = svc_max_payload(rqstp);
2796         rlen = min(op->u.listxattrs.lsxa_maxcount, maxcount);
2797
2798         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2799 }
2800
2801 static inline u32 nfsd4_removexattr_rsize(struct svc_rqst *rqstp,
2802                                           struct nfsd4_op *op)
2803 {
2804         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2805                 * sizeof(__be32);
2806 }
2807
2808
2809 static const struct nfsd4_operation nfsd4_ops[] = {
2810         [OP_ACCESS] = {
2811                 .op_func = nfsd4_access,
2812                 .op_name = "OP_ACCESS",
2813                 .op_rsize_bop = nfsd4_access_rsize,
2814         },
2815         [OP_CLOSE] = {
2816                 .op_func = nfsd4_close,
2817                 .op_flags = OP_MODIFIES_SOMETHING,
2818                 .op_name = "OP_CLOSE",
2819                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2820                 .op_get_currentstateid = nfsd4_get_closestateid,
2821                 .op_set_currentstateid = nfsd4_set_closestateid,
2822         },
2823         [OP_COMMIT] = {
2824                 .op_func = nfsd4_commit,
2825                 .op_flags = OP_MODIFIES_SOMETHING,
2826                 .op_name = "OP_COMMIT",
2827                 .op_rsize_bop = nfsd4_commit_rsize,
2828         },
2829         [OP_CREATE] = {
2830                 .op_func = nfsd4_create,
2831                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2832                 .op_name = "OP_CREATE",
2833                 .op_rsize_bop = nfsd4_create_rsize,
2834         },
2835         [OP_DELEGRETURN] = {
2836                 .op_func = nfsd4_delegreturn,
2837                 .op_flags = OP_MODIFIES_SOMETHING,
2838                 .op_name = "OP_DELEGRETURN",
2839                 .op_rsize_bop = nfsd4_only_status_rsize,
2840                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2841         },
2842         [OP_GETATTR] = {
2843                 .op_func = nfsd4_getattr,
2844                 .op_flags = ALLOWED_ON_ABSENT_FS,
2845                 .op_rsize_bop = nfsd4_getattr_rsize,
2846                 .op_name = "OP_GETATTR",
2847         },
2848         [OP_GETFH] = {
2849                 .op_func = nfsd4_getfh,
2850                 .op_name = "OP_GETFH",
2851                 .op_rsize_bop = nfsd4_getfh_rsize,
2852         },
2853         [OP_LINK] = {
2854                 .op_func = nfsd4_link,
2855                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
2856                                 | OP_CACHEME,
2857                 .op_name = "OP_LINK",
2858                 .op_rsize_bop = nfsd4_link_rsize,
2859         },
2860         [OP_LOCK] = {
2861                 .op_func = nfsd4_lock,
2862                 .op_flags = OP_MODIFIES_SOMETHING |
2863                                 OP_NONTRIVIAL_ERROR_ENCODE,
2864                 .op_name = "OP_LOCK",
2865                 .op_rsize_bop = nfsd4_lock_rsize,
2866                 .op_set_currentstateid = nfsd4_set_lockstateid,
2867         },
2868         [OP_LOCKT] = {
2869                 .op_func = nfsd4_lockt,
2870                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
2871                 .op_name = "OP_LOCKT",
2872                 .op_rsize_bop = nfsd4_lock_rsize,
2873         },
2874         [OP_LOCKU] = {
2875                 .op_func = nfsd4_locku,
2876                 .op_flags = OP_MODIFIES_SOMETHING,
2877                 .op_name = "OP_LOCKU",
2878                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2879                 .op_get_currentstateid = nfsd4_get_lockustateid,
2880         },
2881         [OP_LOOKUP] = {
2882                 .op_func = nfsd4_lookup,
2883                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2884                 .op_name = "OP_LOOKUP",
2885                 .op_rsize_bop = nfsd4_only_status_rsize,
2886         },
2887         [OP_LOOKUPP] = {
2888                 .op_func = nfsd4_lookupp,
2889                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2890                 .op_name = "OP_LOOKUPP",
2891                 .op_rsize_bop = nfsd4_only_status_rsize,
2892         },
2893         [OP_NVERIFY] = {
2894                 .op_func = nfsd4_nverify,
2895                 .op_name = "OP_NVERIFY",
2896                 .op_rsize_bop = nfsd4_only_status_rsize,
2897         },
2898         [OP_OPEN] = {
2899                 .op_func = nfsd4_open,
2900                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2901                 .op_name = "OP_OPEN",
2902                 .op_rsize_bop = nfsd4_open_rsize,
2903                 .op_set_currentstateid = nfsd4_set_openstateid,
2904         },
2905         [OP_OPEN_CONFIRM] = {
2906                 .op_func = nfsd4_open_confirm,
2907                 .op_flags = OP_MODIFIES_SOMETHING,
2908                 .op_name = "OP_OPEN_CONFIRM",
2909                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2910         },
2911         [OP_OPEN_DOWNGRADE] = {
2912                 .op_func = nfsd4_open_downgrade,
2913                 .op_flags = OP_MODIFIES_SOMETHING,
2914                 .op_name = "OP_OPEN_DOWNGRADE",
2915                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2916                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
2917                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
2918         },
2919         [OP_PUTFH] = {
2920                 .op_func = nfsd4_putfh,
2921                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2922                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2923                 .op_name = "OP_PUTFH",
2924                 .op_rsize_bop = nfsd4_only_status_rsize,
2925         },
2926         [OP_PUTPUBFH] = {
2927                 .op_func = nfsd4_putrootfh,
2928                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2929                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2930                 .op_name = "OP_PUTPUBFH",
2931                 .op_rsize_bop = nfsd4_only_status_rsize,
2932         },
2933         [OP_PUTROOTFH] = {
2934                 .op_func = nfsd4_putrootfh,
2935                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2936                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2937                 .op_name = "OP_PUTROOTFH",
2938                 .op_rsize_bop = nfsd4_only_status_rsize,
2939         },
2940         [OP_READ] = {
2941                 .op_func = nfsd4_read,
2942                 .op_release = nfsd4_read_release,
2943                 .op_name = "OP_READ",
2944                 .op_rsize_bop = nfsd4_read_rsize,
2945                 .op_get_currentstateid = nfsd4_get_readstateid,
2946         },
2947         [OP_READDIR] = {
2948                 .op_func = nfsd4_readdir,
2949                 .op_name = "OP_READDIR",
2950                 .op_rsize_bop = nfsd4_readdir_rsize,
2951         },
2952         [OP_READLINK] = {
2953                 .op_func = nfsd4_readlink,
2954                 .op_name = "OP_READLINK",
2955                 .op_rsize_bop = nfsd4_readlink_rsize,
2956         },
2957         [OP_REMOVE] = {
2958                 .op_func = nfsd4_remove,
2959                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2960                 .op_name = "OP_REMOVE",
2961                 .op_rsize_bop = nfsd4_remove_rsize,
2962         },
2963         [OP_RENAME] = {
2964                 .op_func = nfsd4_rename,
2965                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2966                 .op_name = "OP_RENAME",
2967                 .op_rsize_bop = nfsd4_rename_rsize,
2968         },
2969         [OP_RENEW] = {
2970                 .op_func = nfsd4_renew,
2971                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2972                                 | OP_MODIFIES_SOMETHING,
2973                 .op_name = "OP_RENEW",
2974                 .op_rsize_bop = nfsd4_only_status_rsize,
2975
2976         },
2977         [OP_RESTOREFH] = {
2978                 .op_func = nfsd4_restorefh,
2979                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2980                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
2981                 .op_name = "OP_RESTOREFH",
2982                 .op_rsize_bop = nfsd4_only_status_rsize,
2983         },
2984         [OP_SAVEFH] = {
2985                 .op_func = nfsd4_savefh,
2986                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2987                 .op_name = "OP_SAVEFH",
2988                 .op_rsize_bop = nfsd4_only_status_rsize,
2989         },
2990         [OP_SECINFO] = {
2991                 .op_func = nfsd4_secinfo,
2992                 .op_release = nfsd4_secinfo_release,
2993                 .op_flags = OP_HANDLES_WRONGSEC,
2994                 .op_name = "OP_SECINFO",
2995                 .op_rsize_bop = nfsd4_secinfo_rsize,
2996         },
2997         [OP_SETATTR] = {
2998                 .op_func = nfsd4_setattr,
2999                 .op_name = "OP_SETATTR",
3000                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3001                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3002                 .op_rsize_bop = nfsd4_setattr_rsize,
3003                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3004         },
3005         [OP_SETCLIENTID] = {
3006                 .op_func = nfsd4_setclientid,
3007                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3008                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3009                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3010                 .op_name = "OP_SETCLIENTID",
3011                 .op_rsize_bop = nfsd4_setclientid_rsize,
3012         },
3013         [OP_SETCLIENTID_CONFIRM] = {
3014                 .op_func = nfsd4_setclientid_confirm,
3015                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3016                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3017                 .op_name = "OP_SETCLIENTID_CONFIRM",
3018                 .op_rsize_bop = nfsd4_only_status_rsize,
3019         },
3020         [OP_VERIFY] = {
3021                 .op_func = nfsd4_verify,
3022                 .op_name = "OP_VERIFY",
3023                 .op_rsize_bop = nfsd4_only_status_rsize,
3024         },
3025         [OP_WRITE] = {
3026                 .op_func = nfsd4_write,
3027                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3028                 .op_name = "OP_WRITE",
3029                 .op_rsize_bop = nfsd4_write_rsize,
3030                 .op_get_currentstateid = nfsd4_get_writestateid,
3031         },
3032         [OP_RELEASE_LOCKOWNER] = {
3033                 .op_func = nfsd4_release_lockowner,
3034                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3035                                 | OP_MODIFIES_SOMETHING,
3036                 .op_name = "OP_RELEASE_LOCKOWNER",
3037                 .op_rsize_bop = nfsd4_only_status_rsize,
3038         },
3039
3040         /* NFSv4.1 operations */
3041         [OP_EXCHANGE_ID] = {
3042                 .op_func = nfsd4_exchange_id,
3043                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3044                                 | OP_MODIFIES_SOMETHING,
3045                 .op_name = "OP_EXCHANGE_ID",
3046                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3047         },
3048         [OP_BACKCHANNEL_CTL] = {
3049                 .op_func = nfsd4_backchannel_ctl,
3050                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3051                 .op_name = "OP_BACKCHANNEL_CTL",
3052                 .op_rsize_bop = nfsd4_only_status_rsize,
3053         },
3054         [OP_BIND_CONN_TO_SESSION] = {
3055                 .op_func = nfsd4_bind_conn_to_session,
3056                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3057                                 | OP_MODIFIES_SOMETHING,
3058                 .op_name = "OP_BIND_CONN_TO_SESSION",
3059                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3060         },
3061         [OP_CREATE_SESSION] = {
3062                 .op_func = nfsd4_create_session,
3063                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3064                                 | OP_MODIFIES_SOMETHING,
3065                 .op_name = "OP_CREATE_SESSION",
3066                 .op_rsize_bop = nfsd4_create_session_rsize,
3067         },
3068         [OP_DESTROY_SESSION] = {
3069                 .op_func = nfsd4_destroy_session,
3070                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3071                                 | OP_MODIFIES_SOMETHING,
3072                 .op_name = "OP_DESTROY_SESSION",
3073                 .op_rsize_bop = nfsd4_only_status_rsize,
3074         },
3075         [OP_SEQUENCE] = {
3076                 .op_func = nfsd4_sequence,
3077                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3078                 .op_name = "OP_SEQUENCE",
3079                 .op_rsize_bop = nfsd4_sequence_rsize,
3080         },
3081         [OP_DESTROY_CLIENTID] = {
3082                 .op_func = nfsd4_destroy_clientid,
3083                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3084                                 | OP_MODIFIES_SOMETHING,
3085                 .op_name = "OP_DESTROY_CLIENTID",
3086                 .op_rsize_bop = nfsd4_only_status_rsize,
3087         },
3088         [OP_RECLAIM_COMPLETE] = {
3089                 .op_func = nfsd4_reclaim_complete,
3090                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3091                 .op_name = "OP_RECLAIM_COMPLETE",
3092                 .op_rsize_bop = nfsd4_only_status_rsize,
3093         },
3094         [OP_SECINFO_NO_NAME] = {
3095                 .op_func = nfsd4_secinfo_no_name,
3096                 .op_release = nfsd4_secinfo_no_name_release,
3097                 .op_flags = OP_HANDLES_WRONGSEC,
3098                 .op_name = "OP_SECINFO_NO_NAME",
3099                 .op_rsize_bop = nfsd4_secinfo_rsize,
3100         },
3101         [OP_TEST_STATEID] = {
3102                 .op_func = nfsd4_test_stateid,
3103                 .op_flags = ALLOWED_WITHOUT_FH,
3104                 .op_name = "OP_TEST_STATEID",
3105                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3106         },
3107         [OP_FREE_STATEID] = {
3108                 .op_func = nfsd4_free_stateid,
3109                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3110                 .op_name = "OP_FREE_STATEID",
3111                 .op_get_currentstateid = nfsd4_get_freestateid,
3112                 .op_rsize_bop = nfsd4_only_status_rsize,
3113         },
3114 #ifdef CONFIG_NFSD_PNFS
3115         [OP_GETDEVICEINFO] = {
3116                 .op_func = nfsd4_getdeviceinfo,
3117                 .op_release = nfsd4_getdeviceinfo_release,
3118                 .op_flags = ALLOWED_WITHOUT_FH,
3119                 .op_name = "OP_GETDEVICEINFO",
3120                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3121         },
3122         [OP_LAYOUTGET] = {
3123                 .op_func = nfsd4_layoutget,
3124                 .op_release = nfsd4_layoutget_release,
3125                 .op_flags = OP_MODIFIES_SOMETHING,
3126                 .op_name = "OP_LAYOUTGET",
3127                 .op_rsize_bop = nfsd4_layoutget_rsize,
3128         },
3129         [OP_LAYOUTCOMMIT] = {
3130                 .op_func = nfsd4_layoutcommit,
3131                 .op_flags = OP_MODIFIES_SOMETHING,
3132                 .op_name = "OP_LAYOUTCOMMIT",
3133                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3134         },
3135         [OP_LAYOUTRETURN] = {
3136                 .op_func = nfsd4_layoutreturn,
3137                 .op_flags = OP_MODIFIES_SOMETHING,
3138                 .op_name = "OP_LAYOUTRETURN",
3139                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3140         },
3141 #endif /* CONFIG_NFSD_PNFS */
3142
3143         /* NFSv4.2 operations */
3144         [OP_ALLOCATE] = {
3145                 .op_func = nfsd4_allocate,
3146                 .op_flags = OP_MODIFIES_SOMETHING,
3147                 .op_name = "OP_ALLOCATE",
3148                 .op_rsize_bop = nfsd4_only_status_rsize,
3149         },
3150         [OP_DEALLOCATE] = {
3151                 .op_func = nfsd4_deallocate,
3152                 .op_flags = OP_MODIFIES_SOMETHING,
3153                 .op_name = "OP_DEALLOCATE",
3154                 .op_rsize_bop = nfsd4_only_status_rsize,
3155         },
3156         [OP_CLONE] = {
3157                 .op_func = nfsd4_clone,
3158                 .op_flags = OP_MODIFIES_SOMETHING,
3159                 .op_name = "OP_CLONE",
3160                 .op_rsize_bop = nfsd4_only_status_rsize,
3161         },
3162         [OP_COPY] = {
3163                 .op_func = nfsd4_copy,
3164                 .op_flags = OP_MODIFIES_SOMETHING,
3165                 .op_name = "OP_COPY",
3166                 .op_rsize_bop = nfsd4_copy_rsize,
3167         },
3168         [OP_READ_PLUS] = {
3169                 .op_func = nfsd4_read,
3170                 .op_release = nfsd4_read_release,
3171                 .op_name = "OP_READ_PLUS",
3172                 .op_rsize_bop = nfsd4_read_plus_rsize,
3173                 .op_get_currentstateid = nfsd4_get_readstateid,
3174         },
3175         [OP_SEEK] = {
3176                 .op_func = nfsd4_seek,
3177                 .op_name = "OP_SEEK",
3178                 .op_rsize_bop = nfsd4_seek_rsize,
3179         },
3180         [OP_OFFLOAD_STATUS] = {
3181                 .op_func = nfsd4_offload_status,
3182                 .op_name = "OP_OFFLOAD_STATUS",
3183                 .op_rsize_bop = nfsd4_offload_status_rsize,
3184         },
3185         [OP_OFFLOAD_CANCEL] = {
3186                 .op_func = nfsd4_offload_cancel,
3187                 .op_flags = OP_MODIFIES_SOMETHING,
3188                 .op_name = "OP_OFFLOAD_CANCEL",
3189                 .op_rsize_bop = nfsd4_only_status_rsize,
3190         },
3191         [OP_COPY_NOTIFY] = {
3192                 .op_func = nfsd4_copy_notify,
3193                 .op_flags = OP_MODIFIES_SOMETHING,
3194                 .op_name = "OP_COPY_NOTIFY",
3195                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3196         },
3197         [OP_GETXATTR] = {
3198                 .op_func = nfsd4_getxattr,
3199                 .op_name = "OP_GETXATTR",
3200                 .op_rsize_bop = nfsd4_getxattr_rsize,
3201         },
3202         [OP_SETXATTR] = {
3203                 .op_func = nfsd4_setxattr,
3204                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3205                 .op_name = "OP_SETXATTR",
3206                 .op_rsize_bop = nfsd4_setxattr_rsize,
3207         },
3208         [OP_LISTXATTRS] = {
3209                 .op_func = nfsd4_listxattrs,
3210                 .op_name = "OP_LISTXATTRS",
3211                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3212         },
3213         [OP_REMOVEXATTR] = {
3214                 .op_func = nfsd4_removexattr,
3215                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3216                 .op_name = "OP_REMOVEXATTR",
3217                 .op_rsize_bop = nfsd4_removexattr_rsize,
3218         },
3219 };
3220
3221 /**
3222  * nfsd4_spo_must_allow - Determine if the compound op contains an
3223  * operation that is allowed to be sent with machine credentials
3224  *
3225  * @rqstp: a pointer to the struct svc_rqst
3226  *
3227  * Checks to see if the compound contains a spo_must_allow op
3228  * and confirms that it was sent with the proper machine creds.
3229  */
3230
3231 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3232 {
3233         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3234         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3235         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
3236         struct nfsd4_compound_state *cstate = &resp->cstate;
3237         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3238         u32 opiter;
3239
3240         if (!cstate->minorversion)
3241                 return false;
3242
3243         if (cstate->spo_must_allowed)
3244                 return true;
3245
3246         opiter = resp->opcnt;
3247         while (opiter < argp->opcnt) {
3248                 this = &argp->ops[opiter++];
3249                 if (test_bit(this->opnum, allow->u.longs) &&
3250                         cstate->clp->cl_mach_cred &&
3251                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3252                         cstate->spo_must_allowed = true;
3253                         return true;
3254                 }
3255         }
3256         cstate->spo_must_allowed = false;
3257         return false;
3258 }
3259
3260 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3261 {
3262         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3263                 return op_encode_hdr_size * sizeof(__be32);
3264
3265         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3266         return OPDESC(op)->op_rsize_bop(rqstp, op);
3267 }
3268
3269 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3270 {
3271         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3272                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3273                         op->opnum, nfsd4_op_name(op->opnum));
3274                 WARN_ON_ONCE(1);
3275         }
3276 }
3277
3278 static const char *nfsd4_op_name(unsigned opnum)
3279 {
3280         if (opnum < ARRAY_SIZE(nfsd4_ops))
3281                 return nfsd4_ops[opnum].op_name;
3282         return "unknown_operation";
3283 }
3284
3285 static const struct svc_procedure nfsd_procedures4[2] = {
3286         [NFSPROC4_NULL] = {
3287                 .pc_func = nfsd4_proc_null,
3288                 .pc_decode = nfssvc_decode_voidarg,
3289                 .pc_encode = nfssvc_encode_voidres,
3290                 .pc_argsize = sizeof(struct nfsd_voidargs),
3291                 .pc_ressize = sizeof(struct nfsd_voidres),
3292                 .pc_cachetype = RC_NOCACHE,
3293                 .pc_xdrressize = 1,
3294                 .pc_name = "NULL",
3295         },
3296         [NFSPROC4_COMPOUND] = {
3297                 .pc_func = nfsd4_proc_compound,
3298                 .pc_decode = nfs4svc_decode_compoundargs,
3299                 .pc_encode = nfs4svc_encode_compoundres,
3300                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3301                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3302                 .pc_release = nfsd4_release_compoundargs,
3303                 .pc_cachetype = RC_NOCACHE,
3304                 .pc_xdrressize = NFSD_BUFSIZE/4,
3305                 .pc_name = "COMPOUND",
3306         },
3307 };
3308
3309 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3310 const struct svc_version nfsd_version4 = {
3311         .vs_vers                = 4,
3312         .vs_nproc               = 2,
3313         .vs_proc                = nfsd_procedures4,
3314         .vs_count               = nfsd_count3,
3315         .vs_dispatch            = nfsd_dispatch,
3316         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3317         .vs_rpcb_optnl          = true,
3318         .vs_need_cong_ctrl      = true,
3319 };