Merge tag '4.21-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
[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
41 #include "idmap.h"
42 #include "cache.h"
43 #include "xdr4.h"
44 #include "vfs.h"
45 #include "current_stateid.h"
46 #include "netns.h"
47 #include "acl.h"
48 #include "pnfs.h"
49 #include "trace.h"
50
51 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
52 #include <linux/security.h>
53
54 static inline void
55 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
56 {
57         struct inode *inode = d_inode(resfh->fh_dentry);
58         int status;
59
60         inode_lock(inode);
61         status = security_inode_setsecctx(resfh->fh_dentry,
62                 label->data, label->len);
63         inode_unlock(inode);
64
65         if (status)
66                 /*
67                  * XXX: We should really fail the whole open, but we may
68                  * already have created a new file, so it may be too
69                  * late.  For now this seems the least of evils:
70                  */
71                 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
72
73         return;
74 }
75 #else
76 static inline void
77 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
78 { }
79 #endif
80
81 #define NFSDDBG_FACILITY                NFSDDBG_PROC
82
83 static u32 nfsd_attrmask[] = {
84         NFSD_WRITEABLE_ATTRS_WORD0,
85         NFSD_WRITEABLE_ATTRS_WORD1,
86         NFSD_WRITEABLE_ATTRS_WORD2
87 };
88
89 static u32 nfsd41_ex_attrmask[] = {
90         NFSD_SUPPATTR_EXCLCREAT_WORD0,
91         NFSD_SUPPATTR_EXCLCREAT_WORD1,
92         NFSD_SUPPATTR_EXCLCREAT_WORD2
93 };
94
95 static __be32
96 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
97                    u32 *bmval, u32 *writable)
98 {
99         struct dentry *dentry = cstate->current_fh.fh_dentry;
100         struct svc_export *exp = cstate->current_fh.fh_export;
101
102         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
103                 return nfserr_attrnotsupp;
104         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
105                 return nfserr_attrnotsupp;
106         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
107                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
108                 return nfserr_attrnotsupp;
109         if (writable && !bmval_is_subset(bmval, writable))
110                 return nfserr_inval;
111         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
112                         (bmval[1] & FATTR4_WORD1_MODE))
113                 return nfserr_inval;
114         return nfs_ok;
115 }
116
117 static __be32
118 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
119         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
120 {
121         __be32 status = nfs_ok;
122
123         if (open->op_create == NFS4_OPEN_CREATE) {
124                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
125                     || open->op_createmode == NFS4_CREATE_GUARDED)
126                         status = check_attr_support(rqstp, cstate,
127                                         open->op_bmval, nfsd_attrmask);
128                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
129                         status = check_attr_support(rqstp, cstate,
130                                         open->op_bmval, nfsd41_ex_attrmask);
131         }
132
133         return status;
134 }
135
136 static int
137 is_create_with_attrs(struct nfsd4_open *open)
138 {
139         return open->op_create == NFS4_OPEN_CREATE
140                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
141                     || open->op_createmode == NFS4_CREATE_GUARDED
142                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
143 }
144
145 /*
146  * if error occurs when setting the acl, just clear the acl bit
147  * in the returned attr bitmap.
148  */
149 static void
150 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
151                 struct nfs4_acl *acl, u32 *bmval)
152 {
153         __be32 status;
154
155         status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
156         if (status)
157                 /*
158                  * We should probably fail the whole open at this point,
159                  * but we've already created the file, so it's too late;
160                  * So this seems the least of evils:
161                  */
162                 bmval[0] &= ~FATTR4_WORD0_ACL;
163 }
164
165 static inline void
166 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
167 {
168         fh_put(dst);
169         dget(src->fh_dentry);
170         if (src->fh_export)
171                 exp_get(src->fh_export);
172         *dst = *src;
173 }
174
175 static __be32
176 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
177 {
178         __be32 status;
179
180         if (open->op_truncate &&
181                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
182                 return nfserr_inval;
183
184         accmode |= NFSD_MAY_READ_IF_EXEC;
185
186         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
187                 accmode |= NFSD_MAY_READ;
188         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
189                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
190         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
191                 accmode |= NFSD_MAY_WRITE;
192
193         status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
194
195         return status;
196 }
197
198 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
199 {
200         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
201
202         if (S_ISREG(mode))
203                 return nfs_ok;
204         if (S_ISDIR(mode))
205                 return nfserr_isdir;
206         /*
207          * Using err_symlink as our catch-all case may look odd; but
208          * there's no other obvious error for this case in 4.0, and we
209          * happen to know that it will cause the linux v4 client to do
210          * the right thing on attempts to open something other than a
211          * regular file.
212          */
213         return nfserr_symlink;
214 }
215
216 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
217 {
218         if (nfsd4_has_session(cstate))
219                 return;
220         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
221                         &resfh->fh_handle);
222 }
223
224 static __be32
225 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
226 {
227         struct svc_fh *current_fh = &cstate->current_fh;
228         int accmode;
229         __be32 status;
230
231         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
232         if (!*resfh)
233                 return nfserr_jukebox;
234         fh_init(*resfh, NFS4_FHSIZE);
235         open->op_truncate = 0;
236
237         if (open->op_create) {
238                 /* FIXME: check session persistence and pnfs flags.
239                  * The nfsv4.1 spec requires the following semantics:
240                  *
241                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
242                  * Reply Cache  | server |                 |
243                  * -------------+--------+-----------------+--------------------
244                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
245                  *              |        |                 | (SHOULD)
246                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
247                  *              |        |                 | (SHOULD NOT)
248                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
249                  * yes          | no     | GUARDED4        | GUARDED4
250                  * yes          | yes    | GUARDED4        | GUARDED4
251                  */
252
253                 /*
254                  * Note: create modes (UNCHECKED,GUARDED...) are the same
255                  * in NFSv4 as in v3 except EXCLUSIVE4_1.
256                  */
257                 current->fs->umask = open->op_umask;
258                 status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
259                                         open->op_fname.len, &open->op_iattr,
260                                         *resfh, open->op_createmode,
261                                         (u32 *)open->op_verf.data,
262                                         &open->op_truncate, &open->op_created);
263                 current->fs->umask = 0;
264
265                 if (!status && open->op_label.len)
266                         nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
267
268                 /*
269                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
270                  * use the returned bitmask to indicate which attributes
271                  * we used to store the verifier:
272                  */
273                 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
274                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
275                                                 FATTR4_WORD1_TIME_MODIFY);
276         } else
277                 /*
278                  * Note this may exit with the parent still locked.
279                  * We will hold the lock until nfsd4_open's final
280                  * lookup, to prevent renames or unlinks until we've had
281                  * a chance to an acquire a delegation if appropriate.
282                  */
283                 status = nfsd_lookup(rqstp, current_fh,
284                                      open->op_fname.data, open->op_fname.len, *resfh);
285         if (status)
286                 goto out;
287         status = nfsd_check_obj_isreg(*resfh);
288         if (status)
289                 goto out;
290
291         if (is_create_with_attrs(open) && open->op_acl != NULL)
292                 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
293
294         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
295         accmode = NFSD_MAY_NOP;
296         if (open->op_created ||
297                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
298                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
299         status = do_open_permission(rqstp, *resfh, open, accmode);
300         set_change_info(&open->op_cinfo, current_fh);
301 out:
302         return status;
303 }
304
305 static __be32
306 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
307 {
308         struct svc_fh *current_fh = &cstate->current_fh;
309         __be32 status;
310         int accmode = 0;
311
312         /* We don't know the target directory, and therefore can not
313         * set the change info
314         */
315
316         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
317
318         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
319
320         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
321                 (open->op_iattr.ia_size == 0);
322         /*
323          * In the delegation case, the client is telling us about an
324          * open that it *already* performed locally, some time ago.  We
325          * should let it succeed now if possible.
326          *
327          * In the case of a CLAIM_FH open, on the other hand, the client
328          * may be counting on us to enforce permissions (the Linux 4.1
329          * client uses this for normal opens, for example).
330          */
331         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
332                 accmode = NFSD_MAY_OWNER_OVERRIDE;
333
334         status = do_open_permission(rqstp, current_fh, open, accmode);
335
336         return status;
337 }
338
339 static void
340 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
341 {
342         struct nfsd4_sessionid *sid =
343                         (struct nfsd4_sessionid *)session->se_sessionid.data;
344
345         clid->cl_boot = sid->clientid.cl_boot;
346         clid->cl_id = sid->clientid.cl_id;
347 }
348
349 static __be32
350 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
351            union nfsd4_op_u *u)
352 {
353         struct nfsd4_open *open = &u->open;
354         __be32 status;
355         struct svc_fh *resfh = NULL;
356         struct net *net = SVC_NET(rqstp);
357         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
358         bool reclaim = false;
359
360         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
361                 (int)open->op_fname.len, open->op_fname.data,
362                 open->op_openowner);
363
364         /* This check required by spec. */
365         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
366                 return nfserr_inval;
367
368         open->op_created = 0;
369         /*
370          * RFC5661 18.51.3
371          * Before RECLAIM_COMPLETE done, server should deny new lock
372          */
373         if (nfsd4_has_session(cstate) &&
374             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
375                       &cstate->session->se_client->cl_flags) &&
376             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
377                 return nfserr_grace;
378
379         if (nfsd4_has_session(cstate))
380                 copy_clientid(&open->op_clientid, cstate->session);
381
382         /* check seqid for replay. set nfs4_owner */
383         status = nfsd4_process_open1(cstate, open, nn);
384         if (status == nfserr_replay_me) {
385                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
386                 fh_put(&cstate->current_fh);
387                 fh_copy_shallow(&cstate->current_fh.fh_handle,
388                                 &rp->rp_openfh);
389                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
390                 if (status)
391                         dprintk("nfsd4_open: replay failed"
392                                 " restoring previous filehandle\n");
393                 else
394                         status = nfserr_replay_me;
395         }
396         if (status)
397                 goto out;
398         if (open->op_xdr_error) {
399                 status = open->op_xdr_error;
400                 goto out;
401         }
402
403         status = nfsd4_check_open_attributes(rqstp, cstate, open);
404         if (status)
405                 goto out;
406
407         /* Openowner is now set, so sequence id will get bumped.  Now we need
408          * these checks before we do any creates: */
409         status = nfserr_grace;
410         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
411                 goto out;
412         status = nfserr_no_grace;
413         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
414                 goto out;
415
416         switch (open->op_claim_type) {
417                 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
418                 case NFS4_OPEN_CLAIM_NULL:
419                         status = do_open_lookup(rqstp, cstate, open, &resfh);
420                         if (status)
421                                 goto out;
422                         break;
423                 case NFS4_OPEN_CLAIM_PREVIOUS:
424                         status = nfs4_check_open_reclaim(&open->op_clientid,
425                                                          cstate, nn);
426                         if (status)
427                                 goto out;
428                         open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
429                         reclaim = true;
430                 case NFS4_OPEN_CLAIM_FH:
431                 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
432                         status = do_open_fhandle(rqstp, cstate, open);
433                         if (status)
434                                 goto out;
435                         resfh = &cstate->current_fh;
436                         break;
437                 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
438                 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
439                         dprintk("NFSD: unsupported OPEN claim type %d\n",
440                                 open->op_claim_type);
441                         status = nfserr_notsupp;
442                         goto out;
443                 default:
444                         dprintk("NFSD: Invalid OPEN claim type %d\n",
445                                 open->op_claim_type);
446                         status = nfserr_inval;
447                         goto out;
448         }
449         /*
450          * nfsd4_process_open2() does the actual opening of the file.  If
451          * successful, it (1) truncates the file if open->op_truncate was
452          * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
453          */
454         status = nfsd4_process_open2(rqstp, resfh, open);
455         WARN(status && open->op_created,
456              "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
457              be32_to_cpu(status));
458         if (reclaim && !status)
459                 nn->somebody_reclaimed = true;
460 out:
461         if (resfh && resfh != &cstate->current_fh) {
462                 fh_dup2(&cstate->current_fh, resfh);
463                 fh_put(resfh);
464                 kfree(resfh);
465         }
466         nfsd4_cleanup_open_state(cstate, open);
467         nfsd4_bump_seqid(cstate, status);
468         return status;
469 }
470
471 /*
472  * OPEN is the only seqid-mutating operation whose decoding can fail
473  * with a seqid-mutating error (specifically, decoding of user names in
474  * the attributes).  Therefore we have to do some processing to look up
475  * the stateowner so that we can bump the seqid.
476  */
477 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
478 {
479         struct nfsd4_open *open = &op->u.open;
480
481         if (!seqid_mutating_err(ntohl(op->status)))
482                 return op->status;
483         if (nfsd4_has_session(cstate))
484                 return op->status;
485         open->op_xdr_error = op->status;
486         return nfsd4_open(rqstp, cstate, &op->u);
487 }
488
489 /*
490  * filehandle-manipulating ops.
491  */
492 static __be32
493 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
494             union nfsd4_op_u *u)
495 {
496         u->getfh = &cstate->current_fh;
497         return nfs_ok;
498 }
499
500 static __be32
501 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
502             union nfsd4_op_u *u)
503 {
504         struct nfsd4_putfh *putfh = &u->putfh;
505
506         fh_put(&cstate->current_fh);
507         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
508         memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
509                putfh->pf_fhlen);
510         return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
511 }
512
513 static __be32
514 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
515                 union nfsd4_op_u *u)
516 {
517         __be32 status;
518
519         fh_put(&cstate->current_fh);
520         status = exp_pseudoroot(rqstp, &cstate->current_fh);
521         return status;
522 }
523
524 static __be32
525 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
526                 union nfsd4_op_u *u)
527 {
528         if (!cstate->save_fh.fh_dentry)
529                 return nfserr_restorefh;
530
531         fh_dup2(&cstate->current_fh, &cstate->save_fh);
532         if (HAS_STATE_ID(cstate, SAVED_STATE_ID_FLAG)) {
533                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
534                 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
535         }
536         return nfs_ok;
537 }
538
539 static __be32
540 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
541              union nfsd4_op_u *u)
542 {
543         fh_dup2(&cstate->save_fh, &cstate->current_fh);
544         if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG)) {
545                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
546                 SET_STATE_ID(cstate, SAVED_STATE_ID_FLAG);
547         }
548         return nfs_ok;
549 }
550
551 /*
552  * misc nfsv4 ops
553  */
554 static __be32
555 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
556              union nfsd4_op_u *u)
557 {
558         struct nfsd4_access *access = &u->access;
559
560         if (access->ac_req_access & ~NFS3_ACCESS_FULL)
561                 return nfserr_inval;
562
563         access->ac_resp_access = access->ac_req_access;
564         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
565                            &access->ac_supported);
566 }
567
568 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
569 {
570         __be32 verf[2];
571         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
572
573         /*
574          * This is opaque to client, so no need to byte-swap. Use
575          * __force to keep sparse happy. y2038 time_t overflow is
576          * irrelevant in this usage.
577          */
578         verf[0] = (__force __be32)nn->nfssvc_boot.tv_sec;
579         verf[1] = (__force __be32)nn->nfssvc_boot.tv_nsec;
580         memcpy(verifier->data, verf, sizeof(verifier->data));
581 }
582
583 static __be32
584 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
585              union nfsd4_op_u *u)
586 {
587         struct nfsd4_commit *commit = &u->commit;
588
589         gen_boot_verifier(&commit->co_verf, SVC_NET(rqstp));
590         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
591                              commit->co_count);
592 }
593
594 static __be32
595 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
596              union nfsd4_op_u *u)
597 {
598         struct nfsd4_create *create = &u->create;
599         struct svc_fh resfh;
600         __be32 status;
601         dev_t rdev;
602
603         fh_init(&resfh, NFS4_FHSIZE);
604
605         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
606         if (status)
607                 return status;
608
609         status = check_attr_support(rqstp, cstate, create->cr_bmval,
610                                     nfsd_attrmask);
611         if (status)
612                 return status;
613
614         current->fs->umask = create->cr_umask;
615         switch (create->cr_type) {
616         case NF4LNK:
617                 status = nfsd_symlink(rqstp, &cstate->current_fh,
618                                       create->cr_name, create->cr_namelen,
619                                       create->cr_data, &resfh);
620                 break;
621
622         case NF4BLK:
623                 status = nfserr_inval;
624                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
625                 if (MAJOR(rdev) != create->cr_specdata1 ||
626                     MINOR(rdev) != create->cr_specdata2)
627                         goto out_umask;
628                 status = nfsd_create(rqstp, &cstate->current_fh,
629                                      create->cr_name, create->cr_namelen,
630                                      &create->cr_iattr, S_IFBLK, rdev, &resfh);
631                 break;
632
633         case NF4CHR:
634                 status = nfserr_inval;
635                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
636                 if (MAJOR(rdev) != create->cr_specdata1 ||
637                     MINOR(rdev) != create->cr_specdata2)
638                         goto out_umask;
639                 status = nfsd_create(rqstp, &cstate->current_fh,
640                                      create->cr_name, create->cr_namelen,
641                                      &create->cr_iattr,S_IFCHR, rdev, &resfh);
642                 break;
643
644         case NF4SOCK:
645                 status = nfsd_create(rqstp, &cstate->current_fh,
646                                      create->cr_name, create->cr_namelen,
647                                      &create->cr_iattr, S_IFSOCK, 0, &resfh);
648                 break;
649
650         case NF4FIFO:
651                 status = nfsd_create(rqstp, &cstate->current_fh,
652                                      create->cr_name, create->cr_namelen,
653                                      &create->cr_iattr, S_IFIFO, 0, &resfh);
654                 break;
655
656         case NF4DIR:
657                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
658                 status = nfsd_create(rqstp, &cstate->current_fh,
659                                      create->cr_name, create->cr_namelen,
660                                      &create->cr_iattr, S_IFDIR, 0, &resfh);
661                 break;
662
663         default:
664                 status = nfserr_badtype;
665         }
666
667         if (status)
668                 goto out;
669
670         if (create->cr_label.len)
671                 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
672
673         if (create->cr_acl != NULL)
674                 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
675                                 create->cr_bmval);
676
677         fh_unlock(&cstate->current_fh);
678         set_change_info(&create->cr_cinfo, &cstate->current_fh);
679         fh_dup2(&cstate->current_fh, &resfh);
680 out:
681         fh_put(&resfh);
682 out_umask:
683         current->fs->umask = 0;
684         return status;
685 }
686
687 static __be32
688 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
689               union nfsd4_op_u *u)
690 {
691         struct nfsd4_getattr *getattr = &u->getattr;
692         __be32 status;
693
694         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
695         if (status)
696                 return status;
697
698         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
699                 return nfserr_inval;
700
701         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
702         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
703         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
704
705         getattr->ga_fhp = &cstate->current_fh;
706         return nfs_ok;
707 }
708
709 static __be32
710 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
711            union nfsd4_op_u *u)
712 {
713         struct nfsd4_link *link = &u->link;
714         __be32 status;
715
716         status = nfsd_link(rqstp, &cstate->current_fh,
717                            link->li_name, link->li_namelen, &cstate->save_fh);
718         if (!status)
719                 set_change_info(&link->li_cinfo, &cstate->current_fh);
720         return status;
721 }
722
723 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
724 {
725         struct svc_fh tmp_fh;
726         __be32 ret;
727
728         fh_init(&tmp_fh, NFS4_FHSIZE);
729         ret = exp_pseudoroot(rqstp, &tmp_fh);
730         if (ret)
731                 return ret;
732         if (tmp_fh.fh_dentry == fh->fh_dentry) {
733                 fh_put(&tmp_fh);
734                 return nfserr_noent;
735         }
736         fh_put(&tmp_fh);
737         return nfsd_lookup(rqstp, fh, "..", 2, fh);
738 }
739
740 static __be32
741 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
742               union nfsd4_op_u *u)
743 {
744         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
745 }
746
747 static __be32
748 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
749              union nfsd4_op_u *u)
750 {
751         return nfsd_lookup(rqstp, &cstate->current_fh,
752                            u->lookup.lo_name, u->lookup.lo_len,
753                            &cstate->current_fh);
754 }
755
756 static __be32
757 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
758            union nfsd4_op_u *u)
759 {
760         struct nfsd4_read *read = &u->read;
761         __be32 status;
762
763         read->rd_filp = NULL;
764         if (read->rd_offset >= OFFSET_MAX)
765                 return nfserr_inval;
766
767         trace_nfsd_read_start(rqstp, &cstate->current_fh,
768                               read->rd_offset, read->rd_length);
769
770         /*
771          * If we do a zero copy read, then a client will see read data
772          * that reflects the state of the file *after* performing the
773          * following compound.
774          *
775          * To ensure proper ordering, we therefore turn off zero copy if
776          * the client wants us to do more in this compound:
777          */
778         if (!nfsd4_last_compound_op(rqstp))
779                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
780
781         /* check stateid */
782         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
783                                         &read->rd_stateid, RD_STATE,
784                                         &read->rd_filp, &read->rd_tmp_file);
785         if (status) {
786                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
787                 goto out;
788         }
789         status = nfs_ok;
790 out:
791         read->rd_rqstp = rqstp;
792         read->rd_fhp = &cstate->current_fh;
793         return status;
794 }
795
796
797 static void
798 nfsd4_read_release(union nfsd4_op_u *u)
799 {
800         if (u->read.rd_filp)
801                 fput(u->read.rd_filp);
802         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
803                              u->read.rd_offset, u->read.rd_length);
804 }
805
806 static __be32
807 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
808               union nfsd4_op_u *u)
809 {
810         struct nfsd4_readdir *readdir = &u->readdir;
811         u64 cookie = readdir->rd_cookie;
812         static const nfs4_verifier zeroverf;
813
814         /* no need to check permission - this will be done in nfsd_readdir() */
815
816         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
817                 return nfserr_inval;
818
819         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
820         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
821         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
822
823         if ((cookie == 1) || (cookie == 2) ||
824             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
825                 return nfserr_bad_cookie;
826
827         readdir->rd_rqstp = rqstp;
828         readdir->rd_fhp = &cstate->current_fh;
829         return nfs_ok;
830 }
831
832 static __be32
833 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
834                union nfsd4_op_u *u)
835 {
836         u->readlink.rl_rqstp = rqstp;
837         u->readlink.rl_fhp = &cstate->current_fh;
838         return nfs_ok;
839 }
840
841 static __be32
842 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
843              union nfsd4_op_u *u)
844 {
845         struct nfsd4_remove *remove = &u->remove;
846         __be32 status;
847
848         if (opens_in_grace(SVC_NET(rqstp)))
849                 return nfserr_grace;
850         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
851                              remove->rm_name, remove->rm_namelen);
852         if (!status) {
853                 fh_unlock(&cstate->current_fh);
854                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
855         }
856         return status;
857 }
858
859 static __be32
860 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
861              union nfsd4_op_u *u)
862 {
863         struct nfsd4_rename *rename = &u->rename;
864         __be32 status;
865
866         if (opens_in_grace(SVC_NET(rqstp)) &&
867                 !(cstate->save_fh.fh_export->ex_flags & NFSEXP_NOSUBTREECHECK))
868                 return nfserr_grace;
869         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
870                              rename->rn_snamelen, &cstate->current_fh,
871                              rename->rn_tname, rename->rn_tnamelen);
872         if (status)
873                 return status;
874         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
875         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
876         return nfs_ok;
877 }
878
879 static __be32
880 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
881               union nfsd4_op_u *u)
882 {
883         struct nfsd4_secinfo *secinfo = &u->secinfo;
884         struct svc_export *exp;
885         struct dentry *dentry;
886         __be32 err;
887
888         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
889         if (err)
890                 return err;
891         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
892                                     secinfo->si_name, secinfo->si_namelen,
893                                     &exp, &dentry);
894         if (err)
895                 return err;
896         fh_unlock(&cstate->current_fh);
897         if (d_really_is_negative(dentry)) {
898                 exp_put(exp);
899                 err = nfserr_noent;
900         } else
901                 secinfo->si_exp = exp;
902         dput(dentry);
903         if (cstate->minorversion)
904                 /* See rfc 5661 section 2.6.3.1.1.8 */
905                 fh_put(&cstate->current_fh);
906         return err;
907 }
908
909 static __be32
910 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
911                 union nfsd4_op_u *u)
912 {
913         __be32 err;
914
915         switch (u->secinfo_no_name.sin_style) {
916         case NFS4_SECINFO_STYLE4_CURRENT_FH:
917                 break;
918         case NFS4_SECINFO_STYLE4_PARENT:
919                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
920                 if (err)
921                         return err;
922                 break;
923         default:
924                 return nfserr_inval;
925         }
926
927         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
928         fh_put(&cstate->current_fh);
929         return nfs_ok;
930 }
931
932 static void
933 nfsd4_secinfo_release(union nfsd4_op_u *u)
934 {
935         if (u->secinfo.si_exp)
936                 exp_put(u->secinfo.si_exp);
937 }
938
939 static void
940 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
941 {
942         if (u->secinfo_no_name.sin_exp)
943                 exp_put(u->secinfo_no_name.sin_exp);
944 }
945
946 static __be32
947 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
948               union nfsd4_op_u *u)
949 {
950         struct nfsd4_setattr *setattr = &u->setattr;
951         __be32 status = nfs_ok;
952         int err;
953
954         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
955                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
956                                 &cstate->current_fh, &setattr->sa_stateid,
957                                 WR_STATE, NULL, NULL);
958                 if (status) {
959                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
960                         return status;
961                 }
962         }
963         err = fh_want_write(&cstate->current_fh);
964         if (err)
965                 return nfserrno(err);
966         status = nfs_ok;
967
968         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
969                                     nfsd_attrmask);
970         if (status)
971                 goto out;
972
973         if (setattr->sa_acl != NULL)
974                 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
975                                             setattr->sa_acl);
976         if (status)
977                 goto out;
978         if (setattr->sa_label.len)
979                 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
980                                 &setattr->sa_label);
981         if (status)
982                 goto out;
983         status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
984                                 0, (time_t)0);
985 out:
986         fh_drop_write(&cstate->current_fh);
987         return status;
988 }
989
990 static __be32
991 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
992             union nfsd4_op_u *u)
993 {
994         struct nfsd4_write *write = &u->write;
995         stateid_t *stateid = &write->wr_stateid;
996         struct file *filp = NULL;
997         __be32 status = nfs_ok;
998         unsigned long cnt;
999         int nvecs;
1000
1001         if (write->wr_offset >= OFFSET_MAX)
1002                 return nfserr_inval;
1003
1004         cnt = write->wr_buflen;
1005         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1006                                write->wr_offset, cnt);
1007         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1008                                                 stateid, WR_STATE, &filp, NULL);
1009         if (status) {
1010                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1011                 return status;
1012         }
1013
1014         write->wr_how_written = write->wr_stable_how;
1015         gen_boot_verifier(&write->wr_verifier, SVC_NET(rqstp));
1016
1017         nvecs = svc_fill_write_vector(rqstp, write->wr_pagelist,
1018                                       &write->wr_head, write->wr_buflen);
1019         if (!nvecs)
1020                 return nfserr_io;
1021         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1022
1023         status = nfsd_vfs_write(rqstp, &cstate->current_fh, filp,
1024                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1025                                 write->wr_how_written);
1026         fput(filp);
1027
1028         write->wr_bytes_written = cnt;
1029         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1030                               write->wr_offset, cnt);
1031         return status;
1032 }
1033
1034 static __be32
1035 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1036                   stateid_t *src_stateid, struct file **src,
1037                   stateid_t *dst_stateid, struct file **dst)
1038 {
1039         __be32 status;
1040
1041         if (!cstate->save_fh.fh_dentry)
1042                 return nfserr_nofilehandle;
1043
1044         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1045                                             src_stateid, RD_STATE, src, NULL);
1046         if (status) {
1047                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1048                 goto out;
1049         }
1050
1051         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1052                                             dst_stateid, WR_STATE, dst, NULL);
1053         if (status) {
1054                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1055                 goto out_put_src;
1056         }
1057
1058         /* fix up for NFS-specific error code */
1059         if (!S_ISREG(file_inode(*src)->i_mode) ||
1060             !S_ISREG(file_inode(*dst)->i_mode)) {
1061                 status = nfserr_wrong_type;
1062                 goto out_put_dst;
1063         }
1064
1065 out:
1066         return status;
1067 out_put_dst:
1068         fput(*dst);
1069 out_put_src:
1070         fput(*src);
1071         goto out;
1072 }
1073
1074 static __be32
1075 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1076                 union nfsd4_op_u *u)
1077 {
1078         struct nfsd4_clone *clone = &u->clone;
1079         struct file *src, *dst;
1080         __be32 status;
1081
1082         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1083                                    &clone->cl_dst_stateid, &dst);
1084         if (status)
1085                 goto out;
1086
1087         status = nfsd4_clone_file_range(src, clone->cl_src_pos,
1088                         dst, clone->cl_dst_pos, clone->cl_count);
1089
1090         fput(dst);
1091         fput(src);
1092 out:
1093         return status;
1094 }
1095
1096 void nfs4_put_copy(struct nfsd4_copy *copy)
1097 {
1098         if (!refcount_dec_and_test(&copy->refcount))
1099                 return;
1100         kfree(copy);
1101 }
1102
1103 static bool
1104 check_and_set_stop_copy(struct nfsd4_copy *copy)
1105 {
1106         bool value;
1107
1108         spin_lock(&copy->cp_clp->async_lock);
1109         value = copy->stopped;
1110         if (!copy->stopped)
1111                 copy->stopped = true;
1112         spin_unlock(&copy->cp_clp->async_lock);
1113         return value;
1114 }
1115
1116 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1117 {
1118         /* only 1 thread should stop the copy */
1119         if (!check_and_set_stop_copy(copy))
1120                 kthread_stop(copy->copy_task);
1121         nfs4_put_copy(copy);
1122 }
1123
1124 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1125 {
1126         struct nfsd4_copy *copy = NULL;
1127
1128         spin_lock(&clp->async_lock);
1129         if (!list_empty(&clp->async_copies)) {
1130                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1131                                         copies);
1132                 refcount_inc(&copy->refcount);
1133         }
1134         spin_unlock(&clp->async_lock);
1135         return copy;
1136 }
1137
1138 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1139 {
1140         struct nfsd4_copy *copy;
1141
1142         while ((copy = nfsd4_get_copy(clp)) != NULL)
1143                 nfsd4_stop_copy(copy);
1144 }
1145
1146 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1147 {
1148         struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1149
1150         nfs4_put_copy(copy);
1151 }
1152
1153 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1154                                  struct rpc_task *task)
1155 {
1156         return 1;
1157 }
1158
1159 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1160         .release = nfsd4_cb_offload_release,
1161         .done = nfsd4_cb_offload_done
1162 };
1163
1164 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1165 {
1166         copy->cp_res.wr_stable_how = NFS_UNSTABLE;
1167         copy->cp_synchronous = sync;
1168         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1169 }
1170
1171 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1172 {
1173         ssize_t bytes_copied = 0;
1174         size_t bytes_total = copy->cp_count;
1175         u64 src_pos = copy->cp_src_pos;
1176         u64 dst_pos = copy->cp_dst_pos;
1177
1178         do {
1179                 if (kthread_should_stop())
1180                         break;
1181                 bytes_copied = nfsd_copy_file_range(copy->file_src, src_pos,
1182                                 copy->file_dst, dst_pos, bytes_total);
1183                 if (bytes_copied <= 0)
1184                         break;
1185                 bytes_total -= bytes_copied;
1186                 copy->cp_res.wr_bytes_written += bytes_copied;
1187                 src_pos += bytes_copied;
1188                 dst_pos += bytes_copied;
1189         } while (bytes_total > 0 && !copy->cp_synchronous);
1190         return bytes_copied;
1191 }
1192
1193 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1194 {
1195         __be32 status;
1196         ssize_t bytes;
1197
1198         bytes = _nfsd_copy_file_range(copy);
1199         /* for async copy, we ignore the error, client can always retry
1200          * to get the error
1201          */
1202         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1203                 status = nfserrno(bytes);
1204         else {
1205                 nfsd4_init_copy_res(copy, sync);
1206                 status = nfs_ok;
1207         }
1208
1209         fput(copy->file_src);
1210         fput(copy->file_dst);
1211         return status;
1212 }
1213
1214 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1215 {
1216         dst->cp_src_pos = src->cp_src_pos;
1217         dst->cp_dst_pos = src->cp_dst_pos;
1218         dst->cp_count = src->cp_count;
1219         dst->cp_synchronous = src->cp_synchronous;
1220         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1221         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1222         dst->cp_clp = src->cp_clp;
1223         dst->file_dst = get_file(src->file_dst);
1224         dst->file_src = get_file(src->file_src);
1225         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1226 }
1227
1228 static void cleanup_async_copy(struct nfsd4_copy *copy)
1229 {
1230         nfs4_free_cp_state(copy);
1231         fput(copy->file_dst);
1232         fput(copy->file_src);
1233         spin_lock(&copy->cp_clp->async_lock);
1234         list_del(&copy->copies);
1235         spin_unlock(&copy->cp_clp->async_lock);
1236         nfs4_put_copy(copy);
1237 }
1238
1239 static int nfsd4_do_async_copy(void *data)
1240 {
1241         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1242         struct nfsd4_copy *cb_copy;
1243
1244         copy->nfserr = nfsd4_do_copy(copy, 0);
1245         cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1246         if (!cb_copy)
1247                 goto out;
1248         memcpy(&cb_copy->cp_res, &copy->cp_res, sizeof(copy->cp_res));
1249         cb_copy->cp_clp = copy->cp_clp;
1250         cb_copy->nfserr = copy->nfserr;
1251         memcpy(&cb_copy->fh, &copy->fh, sizeof(copy->fh));
1252         nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1253                         &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1254         nfsd4_run_cb(&cb_copy->cp_cb);
1255 out:
1256         cleanup_async_copy(copy);
1257         return 0;
1258 }
1259
1260 static __be32
1261 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1262                 union nfsd4_op_u *u)
1263 {
1264         struct nfsd4_copy *copy = &u->copy;
1265         __be32 status;
1266         struct nfsd4_copy *async_copy = NULL;
1267
1268         status = nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1269                                    &copy->file_src, &copy->cp_dst_stateid,
1270                                    &copy->file_dst);
1271         if (status)
1272                 goto out;
1273
1274         copy->cp_clp = cstate->clp;
1275         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1276                 sizeof(struct knfsd_fh));
1277         if (!copy->cp_synchronous) {
1278                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1279
1280                 status = nfserrno(-ENOMEM);
1281                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1282                 if (!async_copy)
1283                         goto out;
1284                 if (!nfs4_init_cp_state(nn, copy)) {
1285                         kfree(async_copy);
1286                         goto out;
1287                 }
1288                 refcount_set(&async_copy->refcount, 1);
1289                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid,
1290                         sizeof(copy->cp_stateid));
1291                 dup_copy_fields(copy, async_copy);
1292                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1293                                 async_copy, "%s", "copy thread");
1294                 if (IS_ERR(async_copy->copy_task))
1295                         goto out_err;
1296                 spin_lock(&async_copy->cp_clp->async_lock);
1297                 list_add(&async_copy->copies,
1298                                 &async_copy->cp_clp->async_copies);
1299                 spin_unlock(&async_copy->cp_clp->async_lock);
1300                 wake_up_process(async_copy->copy_task);
1301                 status = nfs_ok;
1302         } else
1303                 status = nfsd4_do_copy(copy, 1);
1304 out:
1305         return status;
1306 out_err:
1307         cleanup_async_copy(async_copy);
1308         goto out;
1309 }
1310
1311 struct nfsd4_copy *
1312 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1313 {
1314         struct nfsd4_copy *copy;
1315
1316         spin_lock(&clp->async_lock);
1317         list_for_each_entry(copy, &clp->async_copies, copies) {
1318                 if (memcmp(&copy->cp_stateid, stateid, NFS4_STATEID_SIZE))
1319                         continue;
1320                 refcount_inc(&copy->refcount);
1321                 spin_unlock(&clp->async_lock);
1322                 return copy;
1323         }
1324         spin_unlock(&clp->async_lock);
1325         return NULL;
1326 }
1327
1328 static __be32
1329 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1330                      struct nfsd4_compound_state *cstate,
1331                      union nfsd4_op_u *u)
1332 {
1333         struct nfsd4_offload_status *os = &u->offload_status;
1334         __be32 status = 0;
1335         struct nfsd4_copy *copy;
1336         struct nfs4_client *clp = cstate->clp;
1337
1338         copy = find_async_copy(clp, &os->stateid);
1339         if (copy)
1340                 nfsd4_stop_copy(copy);
1341         else
1342                 status = nfserr_bad_stateid;
1343
1344         return status;
1345 }
1346
1347 static __be32
1348 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1349                 struct nfsd4_fallocate *fallocate, int flags)
1350 {
1351         __be32 status = nfserr_notsupp;
1352         struct file *file;
1353
1354         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1355                                             &fallocate->falloc_stateid,
1356                                             WR_STATE, &file, NULL);
1357         if (status != nfs_ok) {
1358                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1359                 return status;
1360         }
1361
1362         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, file,
1363                                      fallocate->falloc_offset,
1364                                      fallocate->falloc_length,
1365                                      flags);
1366         fput(file);
1367         return status;
1368 }
1369 static __be32
1370 nfsd4_offload_status(struct svc_rqst *rqstp,
1371                      struct nfsd4_compound_state *cstate,
1372                      union nfsd4_op_u *u)
1373 {
1374         struct nfsd4_offload_status *os = &u->offload_status;
1375         __be32 status = 0;
1376         struct nfsd4_copy *copy;
1377         struct nfs4_client *clp = cstate->clp;
1378
1379         copy = find_async_copy(clp, &os->stateid);
1380         if (copy) {
1381                 os->count = copy->cp_res.wr_bytes_written;
1382                 nfs4_put_copy(copy);
1383         } else
1384                 status = nfserr_bad_stateid;
1385
1386         return status;
1387 }
1388
1389 static __be32
1390 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1391                union nfsd4_op_u *u)
1392 {
1393         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1394 }
1395
1396 static __be32
1397 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1398                  union nfsd4_op_u *u)
1399 {
1400         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1401                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1402 }
1403
1404 static __be32
1405 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1406            union nfsd4_op_u *u)
1407 {
1408         struct nfsd4_seek *seek = &u->seek;
1409         int whence;
1410         __be32 status;
1411         struct file *file;
1412
1413         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1414                                             &seek->seek_stateid,
1415                                             RD_STATE, &file, NULL);
1416         if (status) {
1417                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1418                 return status;
1419         }
1420
1421         switch (seek->seek_whence) {
1422         case NFS4_CONTENT_DATA:
1423                 whence = SEEK_DATA;
1424                 break;
1425         case NFS4_CONTENT_HOLE:
1426                 whence = SEEK_HOLE;
1427                 break;
1428         default:
1429                 status = nfserr_union_notsupp;
1430                 goto out;
1431         }
1432
1433         /*
1434          * Note:  This call does change file->f_pos, but nothing in NFSD
1435          *        should ever file->f_pos.
1436          */
1437         seek->seek_pos = vfs_llseek(file, seek->seek_offset, whence);
1438         if (seek->seek_pos < 0)
1439                 status = nfserrno(seek->seek_pos);
1440         else if (seek->seek_pos >= i_size_read(file_inode(file)))
1441                 seek->seek_eof = true;
1442
1443 out:
1444         fput(file);
1445         return status;
1446 }
1447
1448 /* This routine never returns NFS_OK!  If there are no other errors, it
1449  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1450  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
1451  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1452  */
1453 static __be32
1454 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1455              struct nfsd4_verify *verify)
1456 {
1457         __be32 *buf, *p;
1458         int count;
1459         __be32 status;
1460
1461         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1462         if (status)
1463                 return status;
1464
1465         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1466         if (status)
1467                 return status;
1468
1469         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1470             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1471                 return nfserr_inval;
1472         if (verify->ve_attrlen & 3)
1473                 return nfserr_inval;
1474
1475         /* count in words:
1476          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1477          */
1478         count = 4 + (verify->ve_attrlen >> 2);
1479         buf = kmalloc(count << 2, GFP_KERNEL);
1480         if (!buf)
1481                 return nfserr_jukebox;
1482
1483         p = buf;
1484         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1485                                     cstate->current_fh.fh_export,
1486                                     cstate->current_fh.fh_dentry,
1487                                     verify->ve_bmval,
1488                                     rqstp, 0);
1489         /*
1490          * If nfsd4_encode_fattr() ran out of space, assume that's because
1491          * the attributes are longer (hence different) than those given:
1492          */
1493         if (status == nfserr_resource)
1494                 status = nfserr_not_same;
1495         if (status)
1496                 goto out_kfree;
1497
1498         /* skip bitmap */
1499         p = buf + 1 + ntohl(buf[0]);
1500         status = nfserr_not_same;
1501         if (ntohl(*p++) != verify->ve_attrlen)
1502                 goto out_kfree;
1503         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1504                 status = nfserr_same;
1505
1506 out_kfree:
1507         kfree(buf);
1508         return status;
1509 }
1510
1511 static __be32
1512 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1513               union nfsd4_op_u *u)
1514 {
1515         __be32 status;
1516
1517         status = _nfsd4_verify(rqstp, cstate, &u->verify);
1518         return status == nfserr_not_same ? nfs_ok : status;
1519 }
1520
1521 static __be32
1522 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1523              union nfsd4_op_u *u)
1524 {
1525         __be32 status;
1526
1527         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1528         return status == nfserr_same ? nfs_ok : status;
1529 }
1530
1531 #ifdef CONFIG_NFSD_PNFS
1532 static const struct nfsd4_layout_ops *
1533 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1534 {
1535         if (!exp->ex_layout_types) {
1536                 dprintk("%s: export does not support pNFS\n", __func__);
1537                 return NULL;
1538         }
1539
1540         if (layout_type >= LAYOUT_TYPE_MAX ||
1541             !(exp->ex_layout_types & (1 << layout_type))) {
1542                 dprintk("%s: layout type %d not supported\n",
1543                         __func__, layout_type);
1544                 return NULL;
1545         }
1546
1547         return nfsd4_layout_ops[layout_type];
1548 }
1549
1550 static __be32
1551 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
1552                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1553 {
1554         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
1555         const struct nfsd4_layout_ops *ops;
1556         struct nfsd4_deviceid_map *map;
1557         struct svc_export *exp;
1558         __be32 nfserr;
1559
1560         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
1561                __func__,
1562                gdp->gd_layout_type,
1563                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
1564                gdp->gd_maxcount);
1565
1566         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
1567         if (!map) {
1568                 dprintk("%s: couldn't find device ID to export mapping!\n",
1569                         __func__);
1570                 return nfserr_noent;
1571         }
1572
1573         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
1574         if (IS_ERR(exp)) {
1575                 dprintk("%s: could not find device id\n", __func__);
1576                 return nfserr_noent;
1577         }
1578
1579         nfserr = nfserr_layoutunavailable;
1580         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
1581         if (!ops)
1582                 goto out;
1583
1584         nfserr = nfs_ok;
1585         if (gdp->gd_maxcount != 0) {
1586                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
1587                                 rqstp, cstate->session->se_client, gdp);
1588         }
1589
1590         gdp->gd_notify_types &= ops->notify_types;
1591 out:
1592         exp_put(exp);
1593         return nfserr;
1594 }
1595
1596 static void
1597 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
1598 {
1599         kfree(u->getdeviceinfo.gd_device);
1600 }
1601
1602 static __be32
1603 nfsd4_layoutget(struct svc_rqst *rqstp,
1604                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1605 {
1606         struct nfsd4_layoutget *lgp = &u->layoutget;
1607         struct svc_fh *current_fh = &cstate->current_fh;
1608         const struct nfsd4_layout_ops *ops;
1609         struct nfs4_layout_stateid *ls;
1610         __be32 nfserr;
1611         int accmode = NFSD_MAY_READ_IF_EXEC;
1612
1613         switch (lgp->lg_seg.iomode) {
1614         case IOMODE_READ:
1615                 accmode |= NFSD_MAY_READ;
1616                 break;
1617         case IOMODE_RW:
1618                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
1619                 break;
1620         default:
1621                 dprintk("%s: invalid iomode %d\n",
1622                         __func__, lgp->lg_seg.iomode);
1623                 nfserr = nfserr_badiomode;
1624                 goto out;
1625         }
1626
1627         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
1628         if (nfserr)
1629                 goto out;
1630
1631         nfserr = nfserr_layoutunavailable;
1632         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
1633         if (!ops)
1634                 goto out;
1635
1636         /*
1637          * Verify minlength and range as per RFC5661:
1638          *  o  If loga_length is less than loga_minlength,
1639          *     the metadata server MUST return NFS4ERR_INVAL.
1640          *  o  If the sum of loga_offset and loga_minlength exceeds
1641          *     NFS4_UINT64_MAX, and loga_minlength is not
1642          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
1643          *  o  If the sum of loga_offset and loga_length exceeds
1644          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
1645          *     the error NFS4ERR_INVAL MUST result.
1646          */
1647         nfserr = nfserr_inval;
1648         if (lgp->lg_seg.length < lgp->lg_minlength ||
1649             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
1650              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
1651             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
1652              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
1653                 goto out;
1654         if (lgp->lg_seg.length == 0)
1655                 goto out;
1656
1657         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
1658                                                 true, lgp->lg_layout_type, &ls);
1659         if (nfserr) {
1660                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
1661                 goto out;
1662         }
1663
1664         nfserr = nfserr_recallconflict;
1665         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
1666                 goto out_put_stid;
1667
1668         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
1669                                      current_fh, lgp);
1670         if (nfserr)
1671                 goto out_put_stid;
1672
1673         nfserr = nfsd4_insert_layout(lgp, ls);
1674
1675 out_put_stid:
1676         mutex_unlock(&ls->ls_mutex);
1677         nfs4_put_stid(&ls->ls_stid);
1678 out:
1679         return nfserr;
1680 }
1681
1682 static void
1683 nfsd4_layoutget_release(union nfsd4_op_u *u)
1684 {
1685         kfree(u->layoutget.lg_content);
1686 }
1687
1688 static __be32
1689 nfsd4_layoutcommit(struct svc_rqst *rqstp,
1690                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1691 {
1692         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
1693         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
1694         struct svc_fh *current_fh = &cstate->current_fh;
1695         const struct nfsd4_layout_ops *ops;
1696         loff_t new_size = lcp->lc_last_wr + 1;
1697         struct inode *inode;
1698         struct nfs4_layout_stateid *ls;
1699         __be32 nfserr;
1700
1701         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
1702         if (nfserr)
1703                 goto out;
1704
1705         nfserr = nfserr_layoutunavailable;
1706         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
1707         if (!ops)
1708                 goto out;
1709         inode = d_inode(current_fh->fh_dentry);
1710
1711         nfserr = nfserr_inval;
1712         if (new_size <= seg->offset) {
1713                 dprintk("pnfsd: last write before layout segment\n");
1714                 goto out;
1715         }
1716         if (new_size > seg->offset + seg->length) {
1717                 dprintk("pnfsd: last write beyond layout segment\n");
1718                 goto out;
1719         }
1720         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
1721                 dprintk("pnfsd: layoutcommit beyond EOF\n");
1722                 goto out;
1723         }
1724
1725         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
1726                                                 false, lcp->lc_layout_type,
1727                                                 &ls);
1728         if (nfserr) {
1729                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
1730                 /* fixup error code as per RFC5661 */
1731                 if (nfserr == nfserr_bad_stateid)
1732                         nfserr = nfserr_badlayout;
1733                 goto out;
1734         }
1735
1736         /* LAYOUTCOMMIT does not require any serialization */
1737         mutex_unlock(&ls->ls_mutex);
1738
1739         if (new_size > i_size_read(inode)) {
1740                 lcp->lc_size_chg = 1;
1741                 lcp->lc_newsize = new_size;
1742         } else {
1743                 lcp->lc_size_chg = 0;
1744         }
1745
1746         nfserr = ops->proc_layoutcommit(inode, lcp);
1747         nfs4_put_stid(&ls->ls_stid);
1748 out:
1749         return nfserr;
1750 }
1751
1752 static __be32
1753 nfsd4_layoutreturn(struct svc_rqst *rqstp,
1754                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1755 {
1756         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
1757         struct svc_fh *current_fh = &cstate->current_fh;
1758         __be32 nfserr;
1759
1760         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
1761         if (nfserr)
1762                 goto out;
1763
1764         nfserr = nfserr_layoutunavailable;
1765         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
1766                 goto out;
1767
1768         switch (lrp->lr_seg.iomode) {
1769         case IOMODE_READ:
1770         case IOMODE_RW:
1771         case IOMODE_ANY:
1772                 break;
1773         default:
1774                 dprintk("%s: invalid iomode %d\n", __func__,
1775                         lrp->lr_seg.iomode);
1776                 nfserr = nfserr_inval;
1777                 goto out;
1778         }
1779
1780         switch (lrp->lr_return_type) {
1781         case RETURN_FILE:
1782                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
1783                 break;
1784         case RETURN_FSID:
1785         case RETURN_ALL:
1786                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
1787                 break;
1788         default:
1789                 dprintk("%s: invalid return_type %d\n", __func__,
1790                         lrp->lr_return_type);
1791                 nfserr = nfserr_inval;
1792                 break;
1793         }
1794 out:
1795         return nfserr;
1796 }
1797 #endif /* CONFIG_NFSD_PNFS */
1798
1799 /*
1800  * NULL call.
1801  */
1802 static __be32
1803 nfsd4_proc_null(struct svc_rqst *rqstp)
1804 {
1805         return nfs_ok;
1806 }
1807
1808 static inline void nfsd4_increment_op_stats(u32 opnum)
1809 {
1810         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
1811                 nfsdstats.nfs4_opcount[opnum]++;
1812 }
1813
1814 static const struct nfsd4_operation nfsd4_ops[];
1815
1816 static const char *nfsd4_op_name(unsigned opnum);
1817
1818 /*
1819  * Enforce NFSv4.1 COMPOUND ordering rules:
1820  *
1821  * Also note, enforced elsewhere:
1822  *      - SEQUENCE other than as first op results in
1823  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
1824  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
1825  *        (Enforced in nfsd4_bind_conn_to_session().)
1826  *      - DESTROY_SESSION must be the final operation in a compound, if
1827  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
1828  *        (Enforced in nfsd4_destroy_session().)
1829  */
1830 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
1831 {
1832         struct nfsd4_op *first_op = &args->ops[0];
1833
1834         /* These ordering requirements don't apply to NFSv4.0: */
1835         if (args->minorversion == 0)
1836                 return nfs_ok;
1837         /* This is weird, but OK, not our problem: */
1838         if (args->opcnt == 0)
1839                 return nfs_ok;
1840         if (first_op->status == nfserr_op_illegal)
1841                 return nfs_ok;
1842         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
1843                 return nfserr_op_not_in_session;
1844         if (first_op->opnum == OP_SEQUENCE)
1845                 return nfs_ok;
1846         /*
1847          * So first_op is something allowed outside a session, like
1848          * EXCHANGE_ID; but then it has to be the only op in the
1849          * compound:
1850          */
1851         if (args->opcnt != 1)
1852                 return nfserr_not_only_op;
1853         return nfs_ok;
1854 }
1855
1856 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
1857 {
1858         return &nfsd4_ops[op->opnum];
1859 }
1860
1861 bool nfsd4_cache_this_op(struct nfsd4_op *op)
1862 {
1863         if (op->opnum == OP_ILLEGAL)
1864                 return false;
1865         return OPDESC(op)->op_flags & OP_CACHEME;
1866 }
1867
1868 static bool need_wrongsec_check(struct svc_rqst *rqstp)
1869 {
1870         struct nfsd4_compoundres *resp = rqstp->rq_resp;
1871         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1872         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
1873         struct nfsd4_op *next = &argp->ops[resp->opcnt];
1874         const struct nfsd4_operation *thisd = OPDESC(this);
1875         const struct nfsd4_operation *nextd;
1876
1877         /*
1878          * Most ops check wronsec on our own; only the putfh-like ops
1879          * have special rules.
1880          */
1881         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
1882                 return false;
1883         /*
1884          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
1885          * put-filehandle operation if we're not going to use the
1886          * result:
1887          */
1888         if (argp->opcnt == resp->opcnt)
1889                 return false;
1890         if (next->opnum == OP_ILLEGAL)
1891                 return false;
1892         nextd = OPDESC(next);
1893         /*
1894          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
1895          * errors themselves as necessary; others should check for them
1896          * now:
1897          */
1898         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
1899 }
1900
1901 static void svcxdr_init_encode(struct svc_rqst *rqstp,
1902                                struct nfsd4_compoundres *resp)
1903 {
1904         struct xdr_stream *xdr = &resp->xdr;
1905         struct xdr_buf *buf = &rqstp->rq_res;
1906         struct kvec *head = buf->head;
1907
1908         xdr->buf = buf;
1909         xdr->iov = head;
1910         xdr->p   = head->iov_base + head->iov_len;
1911         xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
1912         /* Tail and page_len should be zero at this point: */
1913         buf->len = buf->head[0].iov_len;
1914         xdr->scratch.iov_len = 0;
1915         xdr->page_ptr = buf->pages - 1;
1916         buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
1917                 - rqstp->rq_auth_slack;
1918 }
1919
1920 /*
1921  * COMPOUND call.
1922  */
1923 static __be32
1924 nfsd4_proc_compound(struct svc_rqst *rqstp)
1925 {
1926         struct nfsd4_compoundargs *args = rqstp->rq_argp;
1927         struct nfsd4_compoundres *resp = rqstp->rq_resp;
1928         struct nfsd4_op *op;
1929         struct nfsd4_compound_state *cstate = &resp->cstate;
1930         struct svc_fh *current_fh = &cstate->current_fh;
1931         struct svc_fh *save_fh = &cstate->save_fh;
1932         __be32          status;
1933
1934         svcxdr_init_encode(rqstp, resp);
1935         resp->tagp = resp->xdr.p;
1936         /* reserve space for: taglen, tag, and opcnt */
1937         xdr_reserve_space(&resp->xdr, 8 + args->taglen);
1938         resp->taglen = args->taglen;
1939         resp->tag = args->tag;
1940         resp->rqstp = rqstp;
1941         cstate->minorversion = args->minorversion;
1942         fh_init(current_fh, NFS4_FHSIZE);
1943         fh_init(save_fh, NFS4_FHSIZE);
1944         /*
1945          * Don't use the deferral mechanism for NFSv4; compounds make it
1946          * too hard to avoid non-idempotency problems.
1947          */
1948         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1949
1950         /*
1951          * According to RFC3010, this takes precedence over all other errors.
1952          */
1953         status = nfserr_minor_vers_mismatch;
1954         if (nfsd_minorversion(args->minorversion, NFSD_TEST) <= 0)
1955                 goto out;
1956         status = nfserr_resource;
1957         if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
1958                 goto out;
1959
1960         status = nfs41_check_op_ordering(args);
1961         if (status) {
1962                 op = &args->ops[0];
1963                 op->status = status;
1964                 resp->opcnt = 1;
1965                 goto encode_op;
1966         }
1967
1968         trace_nfsd_compound(rqstp, args->opcnt);
1969         while (!status && resp->opcnt < args->opcnt) {
1970                 op = &args->ops[resp->opcnt++];
1971
1972                 /*
1973                  * The XDR decode routines may have pre-set op->status;
1974                  * for example, if there is a miscellaneous XDR error
1975                  * it will be set to nfserr_bad_xdr.
1976                  */
1977                 if (op->status) {
1978                         if (op->opnum == OP_OPEN)
1979                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
1980                         goto encode_op;
1981                 }
1982
1983                 if (!current_fh->fh_dentry) {
1984                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
1985                                 op->status = nfserr_nofilehandle;
1986                                 goto encode_op;
1987                         }
1988                 } else if (current_fh->fh_export->ex_fslocs.migrated &&
1989                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
1990                         op->status = nfserr_moved;
1991                         goto encode_op;
1992                 }
1993
1994                 fh_clear_wcc(current_fh);
1995
1996                 /* If op is non-idempotent */
1997                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
1998                         /*
1999                          * Don't execute this op if we couldn't encode a
2000                          * succesful reply:
2001                          */
2002                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2003                         /*
2004                          * Plus if there's another operation, make sure
2005                          * we'll have space to at least encode an error:
2006                          */
2007                         if (resp->opcnt < args->opcnt)
2008                                 plen += COMPOUND_ERR_SLACK_SPACE;
2009                         op->status = nfsd4_check_resp_size(resp, plen);
2010                 }
2011
2012                 if (op->status)
2013                         goto encode_op;
2014
2015                 if (op->opdesc->op_get_currentstateid)
2016                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2017                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2018
2019                 /* Only from SEQUENCE */
2020                 if (cstate->status == nfserr_replay_cache) {
2021                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2022                         status = op->status;
2023                         goto out;
2024                 }
2025                 if (!op->status) {
2026                         if (op->opdesc->op_set_currentstateid)
2027                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2028
2029                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2030                                 clear_current_stateid(cstate);
2031
2032                         if (need_wrongsec_check(rqstp))
2033                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2034                 }
2035 encode_op:
2036                 if (op->status == nfserr_replay_me) {
2037                         op->replay = &cstate->replay_owner->so_replay;
2038                         nfsd4_encode_replay(&resp->xdr, op);
2039                         status = op->status = op->replay->rp_status;
2040                 } else {
2041                         nfsd4_encode_operation(resp, op);
2042                         status = op->status;
2043                 }
2044
2045                 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2046                                            nfsd4_op_name(op->opnum));
2047
2048                 nfsd4_cstate_clear_replay(cstate);
2049                 nfsd4_increment_op_stats(op->opnum);
2050         }
2051
2052         cstate->status = status;
2053         fh_put(current_fh);
2054         fh_put(save_fh);
2055         BUG_ON(cstate->replay_owner);
2056 out:
2057         /* Reset deferral mechanism for RPC deferrals */
2058         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2059         dprintk("nfsv4 compound returned %d\n", ntohl(status));
2060         return status;
2061 }
2062
2063 #define op_encode_hdr_size              (2)
2064 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2065 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2066 #define op_encode_change_info_maxsz     (5)
2067 #define nfs4_fattr_bitmap_maxsz         (4)
2068
2069 /* We'll fall back on returning no lockowner if run out of space: */
2070 #define op_encode_lockowner_maxsz       (0)
2071 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2072
2073 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2074
2075 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2076 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2077                                          op_encode_ace_maxsz)
2078
2079 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2080
2081 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2082 {
2083         return (op_encode_hdr_size) * sizeof(__be32);
2084 }
2085
2086 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2087 {
2088         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2089 }
2090
2091 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2092 {
2093         /* ac_supported, ac_resp_access */
2094         return (op_encode_hdr_size + 2)* sizeof(__be32);
2095 }
2096
2097 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2098 {
2099         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2100 }
2101
2102 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2103 {
2104         return (op_encode_hdr_size + op_encode_change_info_maxsz
2105                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2106 }
2107
2108 /*
2109  * Note since this is an idempotent operation we won't insist on failing
2110  * the op prematurely if the estimate is too large.  We may turn off splice
2111  * reads unnecessarily.
2112  */
2113 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2114                                       struct nfsd4_op *op)
2115 {
2116         u32 *bmap = op->u.getattr.ga_bmval;
2117         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2118         u32 ret = 0;
2119
2120         if (bmap0 & FATTR4_WORD0_ACL)
2121                 return svc_max_payload(rqstp);
2122         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2123                 return svc_max_payload(rqstp);
2124
2125         if (bmap1 & FATTR4_WORD1_OWNER) {
2126                 ret += IDMAP_NAMESZ + 4;
2127                 bmap1 &= ~FATTR4_WORD1_OWNER;
2128         }
2129         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2130                 ret += IDMAP_NAMESZ + 4;
2131                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2132         }
2133         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2134                 ret += NFS4_FHSIZE + 4;
2135                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2136         }
2137         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2138                 ret += NFS4_MAXLABELLEN + 12;
2139                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2140         }
2141         /*
2142          * Largest of remaining attributes are 16 bytes (e.g.,
2143          * supported_attributes)
2144          */
2145         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2146         /* bitmask, length */
2147         ret += 20;
2148         return ret;
2149 }
2150
2151 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2152 {
2153         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2154 }
2155
2156 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2157 {
2158         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2159                 * sizeof(__be32);
2160 }
2161
2162 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2163 {
2164         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2165                 * sizeof(__be32);
2166 }
2167
2168 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2169 {
2170         return (op_encode_hdr_size + op_encode_stateid_maxsz
2171                 + op_encode_change_info_maxsz + 1
2172                 + nfs4_fattr_bitmap_maxsz
2173                 + op_encode_delegation_maxsz) * sizeof(__be32);
2174 }
2175
2176 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2177 {
2178         u32 maxcount = 0, rlen = 0;
2179
2180         maxcount = svc_max_payload(rqstp);
2181         rlen = min(op->u.read.rd_length, maxcount);
2182
2183         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2184 }
2185
2186 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2187 {
2188         u32 maxcount = 0, rlen = 0;
2189
2190         maxcount = svc_max_payload(rqstp);
2191         rlen = min(op->u.readdir.rd_maxcount, maxcount);
2192
2193         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2194                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2195 }
2196
2197 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2198 {
2199         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2200 }
2201
2202 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2203 {
2204         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2205                 * sizeof(__be32);
2206 }
2207
2208 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2209 {
2210         return (op_encode_hdr_size + op_encode_change_info_maxsz
2211                 + op_encode_change_info_maxsz) * sizeof(__be32);
2212 }
2213
2214 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2215                                        struct nfsd4_op *op)
2216 {
2217         return (op_encode_hdr_size
2218                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2219 }
2220
2221 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2222 {
2223         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2224                 * sizeof(__be32);
2225 }
2226
2227 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2228 {
2229         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2230 }
2231
2232 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2233 {
2234         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2235                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2236 }
2237
2238 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2239 {
2240         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2241                                                                 sizeof(__be32);
2242 }
2243
2244 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2245 {
2246         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2247 }
2248
2249 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2250 {
2251         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2252                 1 + 1 + /* eir_flags, spr_how */\
2253                 4 + /* spo_must_enforce & _allow with bitmap */\
2254                 2 + /*eir_server_owner.so_minor_id */\
2255                 /* eir_server_owner.so_major_id<> */\
2256                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2257                 /* eir_server_scope<> */\
2258                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2259                 1 + /* eir_server_impl_id array length */\
2260                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2261 }
2262
2263 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2264 {
2265         return (op_encode_hdr_size + \
2266                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2267                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2268 }
2269
2270 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2271 {
2272         return (op_encode_hdr_size + \
2273                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2274                 2 + /* csr_sequence, csr_flags */\
2275                 op_encode_channel_attrs_maxsz + \
2276                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2277 }
2278
2279 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2280 {
2281         return (op_encode_hdr_size +
2282                 1 /* wr_callback */ +
2283                 op_encode_stateid_maxsz /* wr_callback */ +
2284                 2 /* wr_count */ +
2285                 1 /* wr_committed */ +
2286                 op_encode_verifier_maxsz +
2287                 1 /* cr_consecutive */ +
2288                 1 /* cr_synchronous */) * sizeof(__be32);
2289 }
2290
2291 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2292                                              struct nfsd4_op *op)
2293 {
2294         return (op_encode_hdr_size +
2295                 2 /* osr_count */ +
2296                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2297 }
2298
2299 #ifdef CONFIG_NFSD_PNFS
2300 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2301 {
2302         u32 maxcount = 0, rlen = 0;
2303
2304         maxcount = svc_max_payload(rqstp);
2305         rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2306
2307         return (op_encode_hdr_size +
2308                 1 /* gd_layout_type*/ +
2309                 XDR_QUADLEN(rlen) +
2310                 2 /* gd_notify_types */) * sizeof(__be32);
2311 }
2312
2313 /*
2314  * At this stage we don't really know what layout driver will handle the request,
2315  * so we need to define an arbitrary upper bound here.
2316  */
2317 #define MAX_LAYOUT_SIZE         128
2318 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2319 {
2320         return (op_encode_hdr_size +
2321                 1 /* logr_return_on_close */ +
2322                 op_encode_stateid_maxsz +
2323                 1 /* nr of layouts */ +
2324                 MAX_LAYOUT_SIZE) * sizeof(__be32);
2325 }
2326
2327 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2328 {
2329         return (op_encode_hdr_size +
2330                 1 /* locr_newsize */ +
2331                 2 /* ns_size */) * sizeof(__be32);
2332 }
2333
2334 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2335 {
2336         return (op_encode_hdr_size +
2337                 1 /* lrs_stateid */ +
2338                 op_encode_stateid_maxsz) * sizeof(__be32);
2339 }
2340 #endif /* CONFIG_NFSD_PNFS */
2341
2342
2343 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2344 {
2345         return (op_encode_hdr_size + 3) * sizeof(__be32);
2346 }
2347
2348 static const struct nfsd4_operation nfsd4_ops[] = {
2349         [OP_ACCESS] = {
2350                 .op_func = nfsd4_access,
2351                 .op_name = "OP_ACCESS",
2352                 .op_rsize_bop = nfsd4_access_rsize,
2353         },
2354         [OP_CLOSE] = {
2355                 .op_func = nfsd4_close,
2356                 .op_flags = OP_MODIFIES_SOMETHING,
2357                 .op_name = "OP_CLOSE",
2358                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2359                 .op_get_currentstateid = nfsd4_get_closestateid,
2360                 .op_set_currentstateid = nfsd4_set_closestateid,
2361         },
2362         [OP_COMMIT] = {
2363                 .op_func = nfsd4_commit,
2364                 .op_flags = OP_MODIFIES_SOMETHING,
2365                 .op_name = "OP_COMMIT",
2366                 .op_rsize_bop = nfsd4_commit_rsize,
2367         },
2368         [OP_CREATE] = {
2369                 .op_func = nfsd4_create,
2370                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2371                 .op_name = "OP_CREATE",
2372                 .op_rsize_bop = nfsd4_create_rsize,
2373         },
2374         [OP_DELEGRETURN] = {
2375                 .op_func = nfsd4_delegreturn,
2376                 .op_flags = OP_MODIFIES_SOMETHING,
2377                 .op_name = "OP_DELEGRETURN",
2378                 .op_rsize_bop = nfsd4_only_status_rsize,
2379                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2380         },
2381         [OP_GETATTR] = {
2382                 .op_func = nfsd4_getattr,
2383                 .op_flags = ALLOWED_ON_ABSENT_FS,
2384                 .op_rsize_bop = nfsd4_getattr_rsize,
2385                 .op_name = "OP_GETATTR",
2386         },
2387         [OP_GETFH] = {
2388                 .op_func = nfsd4_getfh,
2389                 .op_name = "OP_GETFH",
2390                 .op_rsize_bop = nfsd4_getfh_rsize,
2391         },
2392         [OP_LINK] = {
2393                 .op_func = nfsd4_link,
2394                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
2395                                 | OP_CACHEME,
2396                 .op_name = "OP_LINK",
2397                 .op_rsize_bop = nfsd4_link_rsize,
2398         },
2399         [OP_LOCK] = {
2400                 .op_func = nfsd4_lock,
2401                 .op_flags = OP_MODIFIES_SOMETHING |
2402                                 OP_NONTRIVIAL_ERROR_ENCODE,
2403                 .op_name = "OP_LOCK",
2404                 .op_rsize_bop = nfsd4_lock_rsize,
2405                 .op_set_currentstateid = nfsd4_set_lockstateid,
2406         },
2407         [OP_LOCKT] = {
2408                 .op_func = nfsd4_lockt,
2409                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
2410                 .op_name = "OP_LOCKT",
2411                 .op_rsize_bop = nfsd4_lock_rsize,
2412         },
2413         [OP_LOCKU] = {
2414                 .op_func = nfsd4_locku,
2415                 .op_flags = OP_MODIFIES_SOMETHING,
2416                 .op_name = "OP_LOCKU",
2417                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2418                 .op_get_currentstateid = nfsd4_get_lockustateid,
2419         },
2420         [OP_LOOKUP] = {
2421                 .op_func = nfsd4_lookup,
2422                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2423                 .op_name = "OP_LOOKUP",
2424                 .op_rsize_bop = nfsd4_only_status_rsize,
2425         },
2426         [OP_LOOKUPP] = {
2427                 .op_func = nfsd4_lookupp,
2428                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2429                 .op_name = "OP_LOOKUPP",
2430                 .op_rsize_bop = nfsd4_only_status_rsize,
2431         },
2432         [OP_NVERIFY] = {
2433                 .op_func = nfsd4_nverify,
2434                 .op_name = "OP_NVERIFY",
2435                 .op_rsize_bop = nfsd4_only_status_rsize,
2436         },
2437         [OP_OPEN] = {
2438                 .op_func = nfsd4_open,
2439                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2440                 .op_name = "OP_OPEN",
2441                 .op_rsize_bop = nfsd4_open_rsize,
2442                 .op_set_currentstateid = nfsd4_set_openstateid,
2443         },
2444         [OP_OPEN_CONFIRM] = {
2445                 .op_func = nfsd4_open_confirm,
2446                 .op_flags = OP_MODIFIES_SOMETHING,
2447                 .op_name = "OP_OPEN_CONFIRM",
2448                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2449         },
2450         [OP_OPEN_DOWNGRADE] = {
2451                 .op_func = nfsd4_open_downgrade,
2452                 .op_flags = OP_MODIFIES_SOMETHING,
2453                 .op_name = "OP_OPEN_DOWNGRADE",
2454                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2455                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
2456                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
2457         },
2458         [OP_PUTFH] = {
2459                 .op_func = nfsd4_putfh,
2460                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2461                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2462                 .op_name = "OP_PUTFH",
2463                 .op_rsize_bop = nfsd4_only_status_rsize,
2464         },
2465         [OP_PUTPUBFH] = {
2466                 .op_func = nfsd4_putrootfh,
2467                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2468                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2469                 .op_name = "OP_PUTPUBFH",
2470                 .op_rsize_bop = nfsd4_only_status_rsize,
2471         },
2472         [OP_PUTROOTFH] = {
2473                 .op_func = nfsd4_putrootfh,
2474                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2475                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2476                 .op_name = "OP_PUTROOTFH",
2477                 .op_rsize_bop = nfsd4_only_status_rsize,
2478         },
2479         [OP_READ] = {
2480                 .op_func = nfsd4_read,
2481                 .op_release = nfsd4_read_release,
2482                 .op_name = "OP_READ",
2483                 .op_rsize_bop = nfsd4_read_rsize,
2484                 .op_get_currentstateid = nfsd4_get_readstateid,
2485         },
2486         [OP_READDIR] = {
2487                 .op_func = nfsd4_readdir,
2488                 .op_name = "OP_READDIR",
2489                 .op_rsize_bop = nfsd4_readdir_rsize,
2490         },
2491         [OP_READLINK] = {
2492                 .op_func = nfsd4_readlink,
2493                 .op_name = "OP_READLINK",
2494                 .op_rsize_bop = nfsd4_readlink_rsize,
2495         },
2496         [OP_REMOVE] = {
2497                 .op_func = nfsd4_remove,
2498                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2499                 .op_name = "OP_REMOVE",
2500                 .op_rsize_bop = nfsd4_remove_rsize,
2501         },
2502         [OP_RENAME] = {
2503                 .op_func = nfsd4_rename,
2504                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2505                 .op_name = "OP_RENAME",
2506                 .op_rsize_bop = nfsd4_rename_rsize,
2507         },
2508         [OP_RENEW] = {
2509                 .op_func = nfsd4_renew,
2510                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2511                                 | OP_MODIFIES_SOMETHING,
2512                 .op_name = "OP_RENEW",
2513                 .op_rsize_bop = nfsd4_only_status_rsize,
2514
2515         },
2516         [OP_RESTOREFH] = {
2517                 .op_func = nfsd4_restorefh,
2518                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2519                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
2520                 .op_name = "OP_RESTOREFH",
2521                 .op_rsize_bop = nfsd4_only_status_rsize,
2522         },
2523         [OP_SAVEFH] = {
2524                 .op_func = nfsd4_savefh,
2525                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2526                 .op_name = "OP_SAVEFH",
2527                 .op_rsize_bop = nfsd4_only_status_rsize,
2528         },
2529         [OP_SECINFO] = {
2530                 .op_func = nfsd4_secinfo,
2531                 .op_release = nfsd4_secinfo_release,
2532                 .op_flags = OP_HANDLES_WRONGSEC,
2533                 .op_name = "OP_SECINFO",
2534                 .op_rsize_bop = nfsd4_secinfo_rsize,
2535         },
2536         [OP_SETATTR] = {
2537                 .op_func = nfsd4_setattr,
2538                 .op_name = "OP_SETATTR",
2539                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
2540                                 | OP_NONTRIVIAL_ERROR_ENCODE,
2541                 .op_rsize_bop = nfsd4_setattr_rsize,
2542                 .op_get_currentstateid = nfsd4_get_setattrstateid,
2543         },
2544         [OP_SETCLIENTID] = {
2545                 .op_func = nfsd4_setclientid,
2546                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2547                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
2548                                 | OP_NONTRIVIAL_ERROR_ENCODE,
2549                 .op_name = "OP_SETCLIENTID",
2550                 .op_rsize_bop = nfsd4_setclientid_rsize,
2551         },
2552         [OP_SETCLIENTID_CONFIRM] = {
2553                 .op_func = nfsd4_setclientid_confirm,
2554                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2555                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
2556                 .op_name = "OP_SETCLIENTID_CONFIRM",
2557                 .op_rsize_bop = nfsd4_only_status_rsize,
2558         },
2559         [OP_VERIFY] = {
2560                 .op_func = nfsd4_verify,
2561                 .op_name = "OP_VERIFY",
2562                 .op_rsize_bop = nfsd4_only_status_rsize,
2563         },
2564         [OP_WRITE] = {
2565                 .op_func = nfsd4_write,
2566                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2567                 .op_name = "OP_WRITE",
2568                 .op_rsize_bop = nfsd4_write_rsize,
2569                 .op_get_currentstateid = nfsd4_get_writestateid,
2570         },
2571         [OP_RELEASE_LOCKOWNER] = {
2572                 .op_func = nfsd4_release_lockowner,
2573                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2574                                 | OP_MODIFIES_SOMETHING,
2575                 .op_name = "OP_RELEASE_LOCKOWNER",
2576                 .op_rsize_bop = nfsd4_only_status_rsize,
2577         },
2578
2579         /* NFSv4.1 operations */
2580         [OP_EXCHANGE_ID] = {
2581                 .op_func = nfsd4_exchange_id,
2582                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2583                                 | OP_MODIFIES_SOMETHING,
2584                 .op_name = "OP_EXCHANGE_ID",
2585                 .op_rsize_bop = nfsd4_exchange_id_rsize,
2586         },
2587         [OP_BACKCHANNEL_CTL] = {
2588                 .op_func = nfsd4_backchannel_ctl,
2589                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2590                 .op_name = "OP_BACKCHANNEL_CTL",
2591                 .op_rsize_bop = nfsd4_only_status_rsize,
2592         },
2593         [OP_BIND_CONN_TO_SESSION] = {
2594                 .op_func = nfsd4_bind_conn_to_session,
2595                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2596                                 | OP_MODIFIES_SOMETHING,
2597                 .op_name = "OP_BIND_CONN_TO_SESSION",
2598                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
2599         },
2600         [OP_CREATE_SESSION] = {
2601                 .op_func = nfsd4_create_session,
2602                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2603                                 | OP_MODIFIES_SOMETHING,
2604                 .op_name = "OP_CREATE_SESSION",
2605                 .op_rsize_bop = nfsd4_create_session_rsize,
2606         },
2607         [OP_DESTROY_SESSION] = {
2608                 .op_func = nfsd4_destroy_session,
2609                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2610                                 | OP_MODIFIES_SOMETHING,
2611                 .op_name = "OP_DESTROY_SESSION",
2612                 .op_rsize_bop = nfsd4_only_status_rsize,
2613         },
2614         [OP_SEQUENCE] = {
2615                 .op_func = nfsd4_sequence,
2616                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
2617                 .op_name = "OP_SEQUENCE",
2618                 .op_rsize_bop = nfsd4_sequence_rsize,
2619         },
2620         [OP_DESTROY_CLIENTID] = {
2621                 .op_func = nfsd4_destroy_clientid,
2622                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2623                                 | OP_MODIFIES_SOMETHING,
2624                 .op_name = "OP_DESTROY_CLIENTID",
2625                 .op_rsize_bop = nfsd4_only_status_rsize,
2626         },
2627         [OP_RECLAIM_COMPLETE] = {
2628                 .op_func = nfsd4_reclaim_complete,
2629                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2630                 .op_name = "OP_RECLAIM_COMPLETE",
2631                 .op_rsize_bop = nfsd4_only_status_rsize,
2632         },
2633         [OP_SECINFO_NO_NAME] = {
2634                 .op_func = nfsd4_secinfo_no_name,
2635                 .op_release = nfsd4_secinfo_no_name_release,
2636                 .op_flags = OP_HANDLES_WRONGSEC,
2637                 .op_name = "OP_SECINFO_NO_NAME",
2638                 .op_rsize_bop = nfsd4_secinfo_rsize,
2639         },
2640         [OP_TEST_STATEID] = {
2641                 .op_func = nfsd4_test_stateid,
2642                 .op_flags = ALLOWED_WITHOUT_FH,
2643                 .op_name = "OP_TEST_STATEID",
2644                 .op_rsize_bop = nfsd4_test_stateid_rsize,
2645         },
2646         [OP_FREE_STATEID] = {
2647                 .op_func = nfsd4_free_stateid,
2648                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2649                 .op_name = "OP_FREE_STATEID",
2650                 .op_get_currentstateid = nfsd4_get_freestateid,
2651                 .op_rsize_bop = nfsd4_only_status_rsize,
2652         },
2653 #ifdef CONFIG_NFSD_PNFS
2654         [OP_GETDEVICEINFO] = {
2655                 .op_func = nfsd4_getdeviceinfo,
2656                 .op_release = nfsd4_getdeviceinfo_release,
2657                 .op_flags = ALLOWED_WITHOUT_FH,
2658                 .op_name = "OP_GETDEVICEINFO",
2659                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
2660         },
2661         [OP_LAYOUTGET] = {
2662                 .op_func = nfsd4_layoutget,
2663                 .op_release = nfsd4_layoutget_release,
2664                 .op_flags = OP_MODIFIES_SOMETHING,
2665                 .op_name = "OP_LAYOUTGET",
2666                 .op_rsize_bop = nfsd4_layoutget_rsize,
2667         },
2668         [OP_LAYOUTCOMMIT] = {
2669                 .op_func = nfsd4_layoutcommit,
2670                 .op_flags = OP_MODIFIES_SOMETHING,
2671                 .op_name = "OP_LAYOUTCOMMIT",
2672                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
2673         },
2674         [OP_LAYOUTRETURN] = {
2675                 .op_func = nfsd4_layoutreturn,
2676                 .op_flags = OP_MODIFIES_SOMETHING,
2677                 .op_name = "OP_LAYOUTRETURN",
2678                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
2679         },
2680 #endif /* CONFIG_NFSD_PNFS */
2681
2682         /* NFSv4.2 operations */
2683         [OP_ALLOCATE] = {
2684                 .op_func = nfsd4_allocate,
2685                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2686                 .op_name = "OP_ALLOCATE",
2687                 .op_rsize_bop = nfsd4_only_status_rsize,
2688         },
2689         [OP_DEALLOCATE] = {
2690                 .op_func = nfsd4_deallocate,
2691                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2692                 .op_name = "OP_DEALLOCATE",
2693                 .op_rsize_bop = nfsd4_only_status_rsize,
2694         },
2695         [OP_CLONE] = {
2696                 .op_func = nfsd4_clone,
2697                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2698                 .op_name = "OP_CLONE",
2699                 .op_rsize_bop = nfsd4_only_status_rsize,
2700         },
2701         [OP_COPY] = {
2702                 .op_func = nfsd4_copy,
2703                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2704                 .op_name = "OP_COPY",
2705                 .op_rsize_bop = nfsd4_copy_rsize,
2706         },
2707         [OP_SEEK] = {
2708                 .op_func = nfsd4_seek,
2709                 .op_name = "OP_SEEK",
2710                 .op_rsize_bop = nfsd4_seek_rsize,
2711         },
2712         [OP_OFFLOAD_STATUS] = {
2713                 .op_func = nfsd4_offload_status,
2714                 .op_name = "OP_OFFLOAD_STATUS",
2715                 .op_rsize_bop = nfsd4_offload_status_rsize,
2716         },
2717         [OP_OFFLOAD_CANCEL] = {
2718                 .op_func = nfsd4_offload_cancel,
2719                 .op_flags = OP_MODIFIES_SOMETHING,
2720                 .op_name = "OP_OFFLOAD_CANCEL",
2721                 .op_rsize_bop = nfsd4_only_status_rsize,
2722         },
2723 };
2724
2725 /**
2726  * nfsd4_spo_must_allow - Determine if the compound op contains an
2727  * operation that is allowed to be sent with machine credentials
2728  *
2729  * @rqstp: a pointer to the struct svc_rqst
2730  *
2731  * Checks to see if the compound contains a spo_must_allow op
2732  * and confirms that it was sent with the proper machine creds.
2733  */
2734
2735 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
2736 {
2737         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2738         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2739         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2740         struct nfsd4_compound_state *cstate = &resp->cstate;
2741         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
2742         u32 opiter;
2743
2744         if (!cstate->minorversion)
2745                 return false;
2746
2747         if (cstate->spo_must_allowed == true)
2748                 return true;
2749
2750         opiter = resp->opcnt;
2751         while (opiter < argp->opcnt) {
2752                 this = &argp->ops[opiter++];
2753                 if (test_bit(this->opnum, allow->u.longs) &&
2754                         cstate->clp->cl_mach_cred &&
2755                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
2756                         cstate->spo_must_allowed = true;
2757                         return true;
2758                 }
2759         }
2760         cstate->spo_must_allowed = false;
2761         return false;
2762 }
2763
2764 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
2765 {
2766         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
2767                 return op_encode_hdr_size * sizeof(__be32);
2768
2769         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
2770         return OPDESC(op)->op_rsize_bop(rqstp, op);
2771 }
2772
2773 void warn_on_nonidempotent_op(struct nfsd4_op *op)
2774 {
2775         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
2776                 pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
2777                         op->opnum, nfsd4_op_name(op->opnum));
2778                 WARN_ON_ONCE(1);
2779         }
2780 }
2781
2782 static const char *nfsd4_op_name(unsigned opnum)
2783 {
2784         if (opnum < ARRAY_SIZE(nfsd4_ops))
2785                 return nfsd4_ops[opnum].op_name;
2786         return "unknown_operation";
2787 }
2788
2789 #define nfsd4_voidres                   nfsd4_voidargs
2790 struct nfsd4_voidargs { int dummy; };
2791
2792 static const struct svc_procedure nfsd_procedures4[2] = {
2793         [NFSPROC4_NULL] = {
2794                 .pc_func = nfsd4_proc_null,
2795                 .pc_encode = nfs4svc_encode_voidres,
2796                 .pc_argsize = sizeof(struct nfsd4_voidargs),
2797                 .pc_ressize = sizeof(struct nfsd4_voidres),
2798                 .pc_cachetype = RC_NOCACHE,
2799                 .pc_xdrressize = 1,
2800         },
2801         [NFSPROC4_COMPOUND] = {
2802                 .pc_func = nfsd4_proc_compound,
2803                 .pc_decode = nfs4svc_decode_compoundargs,
2804                 .pc_encode = nfs4svc_encode_compoundres,
2805                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
2806                 .pc_ressize = sizeof(struct nfsd4_compoundres),
2807                 .pc_release = nfsd4_release_compoundargs,
2808                 .pc_cachetype = RC_NOCACHE,
2809                 .pc_xdrressize = NFSD_BUFSIZE/4,
2810         },
2811 };
2812
2813 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
2814 const struct svc_version nfsd_version4 = {
2815         .vs_vers                = 4,
2816         .vs_nproc               = 2,
2817         .vs_proc                = nfsd_procedures4,
2818         .vs_count               = nfsd_count3,
2819         .vs_dispatch            = nfsd_dispatch,
2820         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
2821         .vs_rpcb_optnl          = true,
2822         .vs_need_cong_ctrl      = true,
2823 };
2824
2825 /*
2826  * Local variables:
2827  *  c-basic-offset: 8
2828  * End:
2829  */