Merge tag 'vfio-v5.10-rc1' of git://github.com/awilliam/linux-vfio
[linux-2.6-microblaze.git] / fs / nfsd / nfs4proc.c
1 /*
2  *  Server-side procedures for NFSv4.
3  *
4  *  Copyright (c) 2002 The Regents of the University of Michigan.
5  *  All rights reserved.
6  *
7  *  Kendrick Smith <kmsmith@umich.edu>
8  *  Andy Adamson   <andros@umich.edu>
9  *
10  *  Redistribution and use in source and binary forms, with or without
11  *  modification, are permitted provided that the following conditions
12  *  are met:
13  *
14  *  1. Redistributions of source code must retain the above copyright
15  *     notice, this list of conditions and the following disclaimer.
16  *  2. Redistributions in binary form must reproduce the above copyright
17  *     notice, this list of conditions and the following disclaimer in the
18  *     documentation and/or other materials provided with the distribution.
19  *  3. Neither the name of the University nor the names of its
20  *     contributors may be used to endorse or promote products derived
21  *     from this software without specific prior written permission.
22  *
23  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34  */
35 #include <linux/fs_struct.h>
36 #include <linux/file.h>
37 #include <linux/falloc.h>
38 #include <linux/slab.h>
39 #include <linux/kthread.h>
40 #include <linux/sunrpc/addr.h>
41 #include <linux/nfs_ssc.h>
42
43 #include "idmap.h"
44 #include "cache.h"
45 #include "xdr4.h"
46 #include "vfs.h"
47 #include "current_stateid.h"
48 #include "netns.h"
49 #include "acl.h"
50 #include "pnfs.h"
51 #include "trace.h"
52
53 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
54 #include <linux/security.h>
55
56 static inline void
57 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
58 {
59         struct inode *inode = d_inode(resfh->fh_dentry);
60         int status;
61
62         inode_lock(inode);
63         status = security_inode_setsecctx(resfh->fh_dentry,
64                 label->data, label->len);
65         inode_unlock(inode);
66
67         if (status)
68                 /*
69                  * XXX: We should really fail the whole open, but we may
70                  * already have created a new file, so it may be too
71                  * late.  For now this seems the least of evils:
72                  */
73                 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
74
75         return;
76 }
77 #else
78 static inline void
79 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
80 { }
81 #endif
82
83 #define NFSDDBG_FACILITY                NFSDDBG_PROC
84
85 static u32 nfsd_attrmask[] = {
86         NFSD_WRITEABLE_ATTRS_WORD0,
87         NFSD_WRITEABLE_ATTRS_WORD1,
88         NFSD_WRITEABLE_ATTRS_WORD2
89 };
90
91 static u32 nfsd41_ex_attrmask[] = {
92         NFSD_SUPPATTR_EXCLCREAT_WORD0,
93         NFSD_SUPPATTR_EXCLCREAT_WORD1,
94         NFSD_SUPPATTR_EXCLCREAT_WORD2
95 };
96
97 static __be32
98 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
99                    u32 *bmval, u32 *writable)
100 {
101         struct dentry *dentry = cstate->current_fh.fh_dentry;
102         struct svc_export *exp = cstate->current_fh.fh_export;
103
104         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
105                 return nfserr_attrnotsupp;
106         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
107                 return nfserr_attrnotsupp;
108         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
109                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
110                 return nfserr_attrnotsupp;
111         if (writable && !bmval_is_subset(bmval, writable))
112                 return nfserr_inval;
113         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
114                         (bmval[1] & FATTR4_WORD1_MODE))
115                 return nfserr_inval;
116         return nfs_ok;
117 }
118
119 static __be32
120 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
121         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
122 {
123         __be32 status = nfs_ok;
124
125         if (open->op_create == NFS4_OPEN_CREATE) {
126                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
127                     || open->op_createmode == NFS4_CREATE_GUARDED)
128                         status = check_attr_support(rqstp, cstate,
129                                         open->op_bmval, nfsd_attrmask);
130                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
131                         status = check_attr_support(rqstp, cstate,
132                                         open->op_bmval, nfsd41_ex_attrmask);
133         }
134
135         return status;
136 }
137
138 static int
139 is_create_with_attrs(struct nfsd4_open *open)
140 {
141         return open->op_create == NFS4_OPEN_CREATE
142                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
143                     || open->op_createmode == NFS4_CREATE_GUARDED
144                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
145 }
146
147 /*
148  * if error occurs when setting the acl, just clear the acl bit
149  * in the returned attr bitmap.
150  */
151 static void
152 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
153                 struct nfs4_acl *acl, u32 *bmval)
154 {
155         __be32 status;
156
157         status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
158         if (status)
159                 /*
160                  * We should probably fail the whole open at this point,
161                  * but we've already created the file, so it's too late;
162                  * So this seems the least of evils:
163                  */
164                 bmval[0] &= ~FATTR4_WORD0_ACL;
165 }
166
167 static inline void
168 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
169 {
170         fh_put(dst);
171         dget(src->fh_dentry);
172         if (src->fh_export)
173                 exp_get(src->fh_export);
174         *dst = *src;
175 }
176
177 static __be32
178 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
179 {
180         __be32 status;
181
182         if (open->op_truncate &&
183                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
184                 return nfserr_inval;
185
186         accmode |= NFSD_MAY_READ_IF_EXEC;
187
188         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
189                 accmode |= NFSD_MAY_READ;
190         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
191                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
192         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
193                 accmode |= NFSD_MAY_WRITE;
194
195         status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
196
197         return status;
198 }
199
200 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
201 {
202         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
203
204         if (S_ISREG(mode))
205                 return nfs_ok;
206         if (S_ISDIR(mode))
207                 return nfserr_isdir;
208         /*
209          * Using err_symlink as our catch-all case may look odd; but
210          * there's no other obvious error for this case in 4.0, and we
211          * happen to know that it will cause the linux v4 client to do
212          * the right thing on attempts to open something other than a
213          * regular file.
214          */
215         return nfserr_symlink;
216 }
217
218 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
219 {
220         if (nfsd4_has_session(cstate))
221                 return;
222         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
223                         &resfh->fh_handle);
224 }
225
226 static __be32
227 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
228 {
229         struct svc_fh *current_fh = &cstate->current_fh;
230         int accmode;
231         __be32 status;
232
233         *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
234         if (!*resfh)
235                 return nfserr_jukebox;
236         fh_init(*resfh, NFS4_FHSIZE);
237         open->op_truncate = false;
238
239         if (open->op_create) {
240                 /* FIXME: check session persistence and pnfs flags.
241                  * The nfsv4.1 spec requires the following semantics:
242                  *
243                  * Persistent   | pNFS   | Server REQUIRED | Client Allowed
244                  * Reply Cache  | server |                 |
245                  * -------------+--------+-----------------+--------------------
246                  * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
247                  *              |        |                 | (SHOULD)
248                  *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
249                  *              |        |                 | (SHOULD NOT)
250                  * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
251                  * yes          | no     | GUARDED4        | GUARDED4
252                  * yes          | yes    | GUARDED4        | GUARDED4
253                  */
254
255                 /*
256                  * Note: create modes (UNCHECKED,GUARDED...) are the same
257                  * in NFSv4 as in v3 except EXCLUSIVE4_1.
258                  */
259                 current->fs->umask = open->op_umask;
260                 status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
261                                         open->op_fname.len, &open->op_iattr,
262                                         *resfh, open->op_createmode,
263                                         (u32 *)open->op_verf.data,
264                                         &open->op_truncate, &open->op_created);
265                 current->fs->umask = 0;
266
267                 if (!status && open->op_label.len)
268                         nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
269
270                 /*
271                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
272                  * use the returned bitmask to indicate which attributes
273                  * we used to store the verifier:
274                  */
275                 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
276                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
277                                                 FATTR4_WORD1_TIME_MODIFY);
278         } else
279                 /*
280                  * Note this may exit with the parent still locked.
281                  * We will hold the lock until nfsd4_open's final
282                  * lookup, to prevent renames or unlinks until we've had
283                  * a chance to an acquire a delegation if appropriate.
284                  */
285                 status = nfsd_lookup(rqstp, current_fh,
286                                      open->op_fname.data, open->op_fname.len, *resfh);
287         if (status)
288                 goto out;
289         status = nfsd_check_obj_isreg(*resfh);
290         if (status)
291                 goto out;
292
293         if (is_create_with_attrs(open) && open->op_acl != NULL)
294                 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
295
296         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
297         accmode = NFSD_MAY_NOP;
298         if (open->op_created ||
299                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
300                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
301         status = do_open_permission(rqstp, *resfh, open, accmode);
302         set_change_info(&open->op_cinfo, current_fh);
303 out:
304         return status;
305 }
306
307 static __be32
308 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
309 {
310         struct svc_fh *current_fh = &cstate->current_fh;
311         __be32 status;
312         int accmode = 0;
313
314         /* We don't know the target directory, and therefore can not
315         * set the change info
316         */
317
318         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
319
320         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
321
322         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
323                 (open->op_iattr.ia_size == 0);
324         /*
325          * In the delegation case, the client is telling us about an
326          * open that it *already* performed locally, some time ago.  We
327          * should let it succeed now if possible.
328          *
329          * In the case of a CLAIM_FH open, on the other hand, the client
330          * may be counting on us to enforce permissions (the Linux 4.1
331          * client uses this for normal opens, for example).
332          */
333         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
334                 accmode = NFSD_MAY_OWNER_OVERRIDE;
335
336         status = do_open_permission(rqstp, current_fh, open, accmode);
337
338         return status;
339 }
340
341 static void
342 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
343 {
344         struct nfsd4_sessionid *sid =
345                         (struct nfsd4_sessionid *)session->se_sessionid.data;
346
347         clid->cl_boot = sid->clientid.cl_boot;
348         clid->cl_id = sid->clientid.cl_id;
349 }
350
351 static __be32
352 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
353            union nfsd4_op_u *u)
354 {
355         struct nfsd4_open *open = &u->open;
356         __be32 status;
357         struct svc_fh *resfh = NULL;
358         struct net *net = SVC_NET(rqstp);
359         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
360         bool reclaim = false;
361
362         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
363                 (int)open->op_fname.len, open->op_fname.data,
364                 open->op_openowner);
365
366         /* This check required by spec. */
367         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
368                 return nfserr_inval;
369
370         open->op_created = false;
371         /*
372          * RFC5661 18.51.3
373          * Before RECLAIM_COMPLETE done, server should deny new lock
374          */
375         if (nfsd4_has_session(cstate) &&
376             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
377                       &cstate->session->se_client->cl_flags) &&
378             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
379                 return nfserr_grace;
380
381         if (nfsd4_has_session(cstate))
382                 copy_clientid(&open->op_clientid, cstate->session);
383
384         /* check seqid for replay. set nfs4_owner */
385         status = nfsd4_process_open1(cstate, open, nn);
386         if (status == nfserr_replay_me) {
387                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
388                 fh_put(&cstate->current_fh);
389                 fh_copy_shallow(&cstate->current_fh.fh_handle,
390                                 &rp->rp_openfh);
391                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
392                 if (status)
393                         dprintk("nfsd4_open: replay failed"
394                                 " restoring previous filehandle\n");
395                 else
396                         status = nfserr_replay_me;
397         }
398         if (status)
399                 goto out;
400         if (open->op_xdr_error) {
401                 status = open->op_xdr_error;
402                 goto out;
403         }
404
405         status = nfsd4_check_open_attributes(rqstp, cstate, open);
406         if (status)
407                 goto out;
408
409         /* Openowner is now set, so sequence id will get bumped.  Now we need
410          * these checks before we do any creates: */
411         status = nfserr_grace;
412         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
413                 goto out;
414         status = nfserr_no_grace;
415         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
416                 goto out;
417
418         switch (open->op_claim_type) {
419                 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
420                 case NFS4_OPEN_CLAIM_NULL:
421                         status = do_open_lookup(rqstp, cstate, open, &resfh);
422                         if (status)
423                                 goto out;
424                         break;
425                 case NFS4_OPEN_CLAIM_PREVIOUS:
426                         status = nfs4_check_open_reclaim(&open->op_clientid,
427                                                          cstate, nn);
428                         if (status)
429                                 goto out;
430                         open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
431                         reclaim = true;
432                         fallthrough;
433                 case NFS4_OPEN_CLAIM_FH:
434                 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
435                         status = do_open_fhandle(rqstp, cstate, open);
436                         if (status)
437                                 goto out;
438                         resfh = &cstate->current_fh;
439                         break;
440                 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
441                 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
442                         dprintk("NFSD: unsupported OPEN claim type %d\n",
443                                 open->op_claim_type);
444                         status = nfserr_notsupp;
445                         goto out;
446                 default:
447                         dprintk("NFSD: Invalid OPEN claim type %d\n",
448                                 open->op_claim_type);
449                         status = nfserr_inval;
450                         goto out;
451         }
452         /*
453          * nfsd4_process_open2() does the actual opening of the file.  If
454          * successful, it (1) truncates the file if open->op_truncate was
455          * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
456          */
457         status = nfsd4_process_open2(rqstp, resfh, open);
458         WARN(status && open->op_created,
459              "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
460              be32_to_cpu(status));
461         if (reclaim && !status)
462                 nn->somebody_reclaimed = true;
463 out:
464         if (resfh && resfh != &cstate->current_fh) {
465                 fh_dup2(&cstate->current_fh, resfh);
466                 fh_put(resfh);
467                 kfree(resfh);
468         }
469         nfsd4_cleanup_open_state(cstate, open);
470         nfsd4_bump_seqid(cstate, status);
471         return status;
472 }
473
474 /*
475  * OPEN is the only seqid-mutating operation whose decoding can fail
476  * with a seqid-mutating error (specifically, decoding of user names in
477  * the attributes).  Therefore we have to do some processing to look up
478  * the stateowner so that we can bump the seqid.
479  */
480 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
481 {
482         struct nfsd4_open *open = &op->u.open;
483
484         if (!seqid_mutating_err(ntohl(op->status)))
485                 return op->status;
486         if (nfsd4_has_session(cstate))
487                 return op->status;
488         open->op_xdr_error = op->status;
489         return nfsd4_open(rqstp, cstate, &op->u);
490 }
491
492 /*
493  * filehandle-manipulating ops.
494  */
495 static __be32
496 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
497             union nfsd4_op_u *u)
498 {
499         u->getfh = &cstate->current_fh;
500         return nfs_ok;
501 }
502
503 static __be32
504 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
505             union nfsd4_op_u *u)
506 {
507         struct nfsd4_putfh *putfh = &u->putfh;
508         __be32 ret;
509
510         fh_put(&cstate->current_fh);
511         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
512         memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
513                putfh->pf_fhlen);
514         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
515 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
516         if (ret == nfserr_stale && putfh->no_verify) {
517                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
518                 ret = 0;
519         }
520 #endif
521         return ret;
522 }
523
524 static __be32
525 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
526                 union nfsd4_op_u *u)
527 {
528         __be32 status;
529
530         fh_put(&cstate->current_fh);
531         status = exp_pseudoroot(rqstp, &cstate->current_fh);
532         return status;
533 }
534
535 static __be32
536 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
537                 union nfsd4_op_u *u)
538 {
539         if (!cstate->save_fh.fh_dentry)
540                 return nfserr_restorefh;
541
542         fh_dup2(&cstate->current_fh, &cstate->save_fh);
543         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
544                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
545                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
546         }
547         return nfs_ok;
548 }
549
550 static __be32
551 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
552              union nfsd4_op_u *u)
553 {
554         fh_dup2(&cstate->save_fh, &cstate->current_fh);
555         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
556                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
557                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
558         }
559         return nfs_ok;
560 }
561
562 /*
563  * misc nfsv4 ops
564  */
565 static __be32
566 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
567              union nfsd4_op_u *u)
568 {
569         struct nfsd4_access *access = &u->access;
570         u32 access_full;
571
572         access_full = NFS3_ACCESS_FULL;
573         if (cstate->minorversion >= 2)
574                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
575                                NFS4_ACCESS_XAWRITE;
576
577         if (access->ac_req_access & ~access_full)
578                 return nfserr_inval;
579
580         access->ac_resp_access = access->ac_req_access;
581         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
582                            &access->ac_supported);
583 }
584
585 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
586 {
587         __be32 *verf = (__be32 *)verifier->data;
588
589         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
590
591         nfsd_copy_boot_verifier(verf, net_generic(net, nfsd_net_id));
592 }
593
594 static __be32
595 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
596              union nfsd4_op_u *u)
597 {
598         struct nfsd4_commit *commit = &u->commit;
599
600         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
601                              commit->co_count,
602                              (__be32 *)commit->co_verf.data);
603 }
604
605 static __be32
606 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
607              union nfsd4_op_u *u)
608 {
609         struct nfsd4_create *create = &u->create;
610         struct svc_fh resfh;
611         __be32 status;
612         dev_t rdev;
613
614         fh_init(&resfh, NFS4_FHSIZE);
615
616         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
617         if (status)
618                 return status;
619
620         status = check_attr_support(rqstp, cstate, create->cr_bmval,
621                                     nfsd_attrmask);
622         if (status)
623                 return status;
624
625         current->fs->umask = create->cr_umask;
626         switch (create->cr_type) {
627         case NF4LNK:
628                 status = nfsd_symlink(rqstp, &cstate->current_fh,
629                                       create->cr_name, create->cr_namelen,
630                                       create->cr_data, &resfh);
631                 break;
632
633         case NF4BLK:
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_IFBLK, rdev, &resfh);
642                 break;
643
644         case NF4CHR:
645                 status = nfserr_inval;
646                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
647                 if (MAJOR(rdev) != create->cr_specdata1 ||
648                     MINOR(rdev) != create->cr_specdata2)
649                         goto out_umask;
650                 status = nfsd_create(rqstp, &cstate->current_fh,
651                                      create->cr_name, create->cr_namelen,
652                                      &create->cr_iattr,S_IFCHR, rdev, &resfh);
653                 break;
654
655         case NF4SOCK:
656                 status = nfsd_create(rqstp, &cstate->current_fh,
657                                      create->cr_name, create->cr_namelen,
658                                      &create->cr_iattr, S_IFSOCK, 0, &resfh);
659                 break;
660
661         case NF4FIFO:
662                 status = nfsd_create(rqstp, &cstate->current_fh,
663                                      create->cr_name, create->cr_namelen,
664                                      &create->cr_iattr, S_IFIFO, 0, &resfh);
665                 break;
666
667         case NF4DIR:
668                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
669                 status = nfsd_create(rqstp, &cstate->current_fh,
670                                      create->cr_name, create->cr_namelen,
671                                      &create->cr_iattr, S_IFDIR, 0, &resfh);
672                 break;
673
674         default:
675                 status = nfserr_badtype;
676         }
677
678         if (status)
679                 goto out;
680
681         if (create->cr_label.len)
682                 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
683
684         if (create->cr_acl != NULL)
685                 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
686                                 create->cr_bmval);
687
688         fh_unlock(&cstate->current_fh);
689         set_change_info(&create->cr_cinfo, &cstate->current_fh);
690         fh_dup2(&cstate->current_fh, &resfh);
691 out:
692         fh_put(&resfh);
693 out_umask:
694         current->fs->umask = 0;
695         return status;
696 }
697
698 static __be32
699 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
700               union nfsd4_op_u *u)
701 {
702         struct nfsd4_getattr *getattr = &u->getattr;
703         __be32 status;
704
705         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
706         if (status)
707                 return status;
708
709         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
710                 return nfserr_inval;
711
712         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
713         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
714         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
715
716         getattr->ga_fhp = &cstate->current_fh;
717         return nfs_ok;
718 }
719
720 static __be32
721 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
722            union nfsd4_op_u *u)
723 {
724         struct nfsd4_link *link = &u->link;
725         __be32 status;
726
727         status = nfsd_link(rqstp, &cstate->current_fh,
728                            link->li_name, link->li_namelen, &cstate->save_fh);
729         if (!status)
730                 set_change_info(&link->li_cinfo, &cstate->current_fh);
731         return status;
732 }
733
734 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
735 {
736         struct svc_fh tmp_fh;
737         __be32 ret;
738
739         fh_init(&tmp_fh, NFS4_FHSIZE);
740         ret = exp_pseudoroot(rqstp, &tmp_fh);
741         if (ret)
742                 return ret;
743         if (tmp_fh.fh_dentry == fh->fh_dentry) {
744                 fh_put(&tmp_fh);
745                 return nfserr_noent;
746         }
747         fh_put(&tmp_fh);
748         return nfsd_lookup(rqstp, fh, "..", 2, fh);
749 }
750
751 static __be32
752 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
753               union nfsd4_op_u *u)
754 {
755         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
756 }
757
758 static __be32
759 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
760              union nfsd4_op_u *u)
761 {
762         return nfsd_lookup(rqstp, &cstate->current_fh,
763                            u->lookup.lo_name, u->lookup.lo_len,
764                            &cstate->current_fh);
765 }
766
767 static __be32
768 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
769            union nfsd4_op_u *u)
770 {
771         struct nfsd4_read *read = &u->read;
772         __be32 status;
773
774         read->rd_nf = NULL;
775         if (read->rd_offset >= OFFSET_MAX)
776                 return nfserr_inval;
777
778         trace_nfsd_read_start(rqstp, &cstate->current_fh,
779                               read->rd_offset, read->rd_length);
780
781         /*
782          * If we do a zero copy read, then a client will see read data
783          * that reflects the state of the file *after* performing the
784          * following compound.
785          *
786          * To ensure proper ordering, we therefore turn off zero copy if
787          * the client wants us to do more in this compound:
788          */
789         if (!nfsd4_last_compound_op(rqstp))
790                 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
791
792         /* check stateid */
793         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
794                                         &read->rd_stateid, RD_STATE,
795                                         &read->rd_nf, NULL);
796         if (status) {
797                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
798                 goto out;
799         }
800         status = nfs_ok;
801 out:
802         read->rd_rqstp = rqstp;
803         read->rd_fhp = &cstate->current_fh;
804         return status;
805 }
806
807
808 static void
809 nfsd4_read_release(union nfsd4_op_u *u)
810 {
811         if (u->read.rd_nf)
812                 nfsd_file_put(u->read.rd_nf);
813         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
814                              u->read.rd_offset, u->read.rd_length);
815 }
816
817 static __be32
818 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
819               union nfsd4_op_u *u)
820 {
821         struct nfsd4_readdir *readdir = &u->readdir;
822         u64 cookie = readdir->rd_cookie;
823         static const nfs4_verifier zeroverf;
824
825         /* no need to check permission - this will be done in nfsd_readdir() */
826
827         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
828                 return nfserr_inval;
829
830         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
831         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
832         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
833
834         if ((cookie == 1) || (cookie == 2) ||
835             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
836                 return nfserr_bad_cookie;
837
838         readdir->rd_rqstp = rqstp;
839         readdir->rd_fhp = &cstate->current_fh;
840         return nfs_ok;
841 }
842
843 static __be32
844 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
845                union nfsd4_op_u *u)
846 {
847         u->readlink.rl_rqstp = rqstp;
848         u->readlink.rl_fhp = &cstate->current_fh;
849         return nfs_ok;
850 }
851
852 static __be32
853 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
854              union nfsd4_op_u *u)
855 {
856         struct nfsd4_remove *remove = &u->remove;
857         __be32 status;
858
859         if (opens_in_grace(SVC_NET(rqstp)))
860                 return nfserr_grace;
861         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
862                              remove->rm_name, remove->rm_namelen);
863         if (!status) {
864                 fh_unlock(&cstate->current_fh);
865                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
866         }
867         return status;
868 }
869
870 static __be32
871 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
872              union nfsd4_op_u *u)
873 {
874         struct nfsd4_rename *rename = &u->rename;
875         __be32 status;
876
877         if (opens_in_grace(SVC_NET(rqstp)))
878                 return nfserr_grace;
879         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
880                              rename->rn_snamelen, &cstate->current_fh,
881                              rename->rn_tname, rename->rn_tnamelen);
882         if (status)
883                 return status;
884         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
885         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
886         return nfs_ok;
887 }
888
889 static __be32
890 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
891               union nfsd4_op_u *u)
892 {
893         struct nfsd4_secinfo *secinfo = &u->secinfo;
894         struct svc_export *exp;
895         struct dentry *dentry;
896         __be32 err;
897
898         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
899         if (err)
900                 return err;
901         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
902                                     secinfo->si_name, secinfo->si_namelen,
903                                     &exp, &dentry);
904         if (err)
905                 return err;
906         fh_unlock(&cstate->current_fh);
907         if (d_really_is_negative(dentry)) {
908                 exp_put(exp);
909                 err = nfserr_noent;
910         } else
911                 secinfo->si_exp = exp;
912         dput(dentry);
913         if (cstate->minorversion)
914                 /* See rfc 5661 section 2.6.3.1.1.8 */
915                 fh_put(&cstate->current_fh);
916         return err;
917 }
918
919 static __be32
920 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
921                 union nfsd4_op_u *u)
922 {
923         __be32 err;
924
925         switch (u->secinfo_no_name.sin_style) {
926         case NFS4_SECINFO_STYLE4_CURRENT_FH:
927                 break;
928         case NFS4_SECINFO_STYLE4_PARENT:
929                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
930                 if (err)
931                         return err;
932                 break;
933         default:
934                 return nfserr_inval;
935         }
936
937         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
938         fh_put(&cstate->current_fh);
939         return nfs_ok;
940 }
941
942 static void
943 nfsd4_secinfo_release(union nfsd4_op_u *u)
944 {
945         if (u->secinfo.si_exp)
946                 exp_put(u->secinfo.si_exp);
947 }
948
949 static void
950 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
951 {
952         if (u->secinfo_no_name.sin_exp)
953                 exp_put(u->secinfo_no_name.sin_exp);
954 }
955
956 static __be32
957 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
958               union nfsd4_op_u *u)
959 {
960         struct nfsd4_setattr *setattr = &u->setattr;
961         __be32 status = nfs_ok;
962         int err;
963
964         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
965                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
966                                 &cstate->current_fh, &setattr->sa_stateid,
967                                 WR_STATE, NULL, NULL);
968                 if (status) {
969                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
970                         return status;
971                 }
972         }
973         err = fh_want_write(&cstate->current_fh);
974         if (err)
975                 return nfserrno(err);
976         status = nfs_ok;
977
978         status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
979                                     nfsd_attrmask);
980         if (status)
981                 goto out;
982
983         if (setattr->sa_acl != NULL)
984                 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
985                                             setattr->sa_acl);
986         if (status)
987                 goto out;
988         if (setattr->sa_label.len)
989                 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
990                                 &setattr->sa_label);
991         if (status)
992                 goto out;
993         status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
994                                 0, (time64_t)0);
995 out:
996         fh_drop_write(&cstate->current_fh);
997         return status;
998 }
999
1000 static __be32
1001 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1002             union nfsd4_op_u *u)
1003 {
1004         struct nfsd4_write *write = &u->write;
1005         stateid_t *stateid = &write->wr_stateid;
1006         struct nfsd_file *nf = NULL;
1007         __be32 status = nfs_ok;
1008         unsigned long cnt;
1009         int nvecs;
1010
1011         if (write->wr_offset >= OFFSET_MAX)
1012                 return nfserr_inval;
1013
1014         cnt = write->wr_buflen;
1015         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1016                                write->wr_offset, cnt);
1017         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1018                                                 stateid, WR_STATE, &nf, NULL);
1019         if (status) {
1020                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1021                 return status;
1022         }
1023
1024         write->wr_how_written = write->wr_stable_how;
1025
1026         nvecs = svc_fill_write_vector(rqstp, write->wr_pagelist,
1027                                       &write->wr_head, write->wr_buflen);
1028         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1029
1030         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1031                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1032                                 write->wr_how_written,
1033                                 (__be32 *)write->wr_verifier.data);
1034         nfsd_file_put(nf);
1035
1036         write->wr_bytes_written = cnt;
1037         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1038                               write->wr_offset, cnt);
1039         return status;
1040 }
1041
1042 static __be32
1043 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1044                   stateid_t *src_stateid, struct nfsd_file **src,
1045                   stateid_t *dst_stateid, struct nfsd_file **dst)
1046 {
1047         __be32 status;
1048
1049         if (!cstate->save_fh.fh_dentry)
1050                 return nfserr_nofilehandle;
1051
1052         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1053                                             src_stateid, RD_STATE, src, NULL);
1054         if (status) {
1055                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1056                 goto out;
1057         }
1058
1059         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1060                                             dst_stateid, WR_STATE, dst, NULL);
1061         if (status) {
1062                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1063                 goto out_put_src;
1064         }
1065
1066         /* fix up for NFS-specific error code */
1067         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1068             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1069                 status = nfserr_wrong_type;
1070                 goto out_put_dst;
1071         }
1072
1073 out:
1074         return status;
1075 out_put_dst:
1076         nfsd_file_put(*dst);
1077 out_put_src:
1078         nfsd_file_put(*src);
1079         goto out;
1080 }
1081
1082 static __be32
1083 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1084                 union nfsd4_op_u *u)
1085 {
1086         struct nfsd4_clone *clone = &u->clone;
1087         struct nfsd_file *src, *dst;
1088         __be32 status;
1089
1090         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1091                                    &clone->cl_dst_stateid, &dst);
1092         if (status)
1093                 goto out;
1094
1095         status = nfsd4_clone_file_range(src, clone->cl_src_pos,
1096                         dst, clone->cl_dst_pos, clone->cl_count,
1097                         EX_ISSYNC(cstate->current_fh.fh_export));
1098
1099         nfsd_file_put(dst);
1100         nfsd_file_put(src);
1101 out:
1102         return status;
1103 }
1104
1105 void nfs4_put_copy(struct nfsd4_copy *copy)
1106 {
1107         if (!refcount_dec_and_test(&copy->refcount))
1108                 return;
1109         kfree(copy);
1110 }
1111
1112 static bool
1113 check_and_set_stop_copy(struct nfsd4_copy *copy)
1114 {
1115         bool value;
1116
1117         spin_lock(&copy->cp_clp->async_lock);
1118         value = copy->stopped;
1119         if (!copy->stopped)
1120                 copy->stopped = true;
1121         spin_unlock(&copy->cp_clp->async_lock);
1122         return value;
1123 }
1124
1125 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1126 {
1127         /* only 1 thread should stop the copy */
1128         if (!check_and_set_stop_copy(copy))
1129                 kthread_stop(copy->copy_task);
1130         nfs4_put_copy(copy);
1131 }
1132
1133 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1134 {
1135         struct nfsd4_copy *copy = NULL;
1136
1137         spin_lock(&clp->async_lock);
1138         if (!list_empty(&clp->async_copies)) {
1139                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1140                                         copies);
1141                 refcount_inc(&copy->refcount);
1142         }
1143         spin_unlock(&clp->async_lock);
1144         return copy;
1145 }
1146
1147 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1148 {
1149         struct nfsd4_copy *copy;
1150
1151         while ((copy = nfsd4_get_copy(clp)) != NULL)
1152                 nfsd4_stop_copy(copy);
1153 }
1154 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1155
1156 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1157                                    struct nfs_fh *src_fh,
1158                                    nfs4_stateid *stateid);
1159 extern void nfs42_ssc_close(struct file *filep);
1160
1161 extern void nfs_sb_deactive(struct super_block *sb);
1162
1163 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1164
1165 /*
1166  * Support one copy source server for now.
1167  */
1168 static __be32
1169 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1170                        struct vfsmount **mount)
1171 {
1172         struct file_system_type *type;
1173         struct vfsmount *ss_mnt;
1174         struct nfs42_netaddr *naddr;
1175         struct sockaddr_storage tmp_addr;
1176         size_t tmp_addrlen, match_netid_len = 3;
1177         char *startsep = "", *endsep = "", *match_netid = "tcp";
1178         char *ipaddr, *dev_name, *raw_data;
1179         int len, raw_len;
1180         __be32 status = nfserr_inval;
1181
1182         naddr = &nss->u.nl4_addr;
1183         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1184                                          naddr->addr_len,
1185                                          (struct sockaddr *)&tmp_addr,
1186                                          sizeof(tmp_addr));
1187         if (tmp_addrlen == 0)
1188                 goto out_err;
1189
1190         if (tmp_addr.ss_family == AF_INET6) {
1191                 startsep = "[";
1192                 endsep = "]";
1193                 match_netid = "tcp6";
1194                 match_netid_len = 4;
1195         }
1196
1197         if (naddr->netid_len != match_netid_len ||
1198                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1199                 goto out_err;
1200
1201         /* Construct the raw data for the vfs_kern_mount call */
1202         len = RPC_MAX_ADDRBUFLEN + 1;
1203         ipaddr = kzalloc(len, GFP_KERNEL);
1204         if (!ipaddr)
1205                 goto out_err;
1206
1207         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1208
1209         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1210
1211         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1212         raw_data = kzalloc(raw_len, GFP_KERNEL);
1213         if (!raw_data)
1214                 goto out_free_ipaddr;
1215
1216         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1217
1218         status = nfserr_nodev;
1219         type = get_fs_type("nfs");
1220         if (!type)
1221                 goto out_free_rawdata;
1222
1223         /* Set the server:<export> for the vfs_kern_mount call */
1224         dev_name = kzalloc(len + 5, GFP_KERNEL);
1225         if (!dev_name)
1226                 goto out_free_rawdata;
1227         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1228
1229         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1230         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1231         module_put(type->owner);
1232         if (IS_ERR(ss_mnt))
1233                 goto out_free_devname;
1234
1235         status = 0;
1236         *mount = ss_mnt;
1237
1238 out_free_devname:
1239         kfree(dev_name);
1240 out_free_rawdata:
1241         kfree(raw_data);
1242 out_free_ipaddr:
1243         kfree(ipaddr);
1244 out_err:
1245         return status;
1246 }
1247
1248 static void
1249 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1250 {
1251         nfs_do_sb_deactive(ss_mnt->mnt_sb);
1252         mntput(ss_mnt);
1253 }
1254
1255 /*
1256  * Verify COPY destination stateid.
1257  *
1258  * Connect to the source server with NFSv4.1.
1259  * Create the source struct file for nfsd_copy_range.
1260  * Called with COPY cstate:
1261  *    SAVED_FH: source filehandle
1262  *    CURRENT_FH: destination filehandle
1263  */
1264 static __be32
1265 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1266                       struct nfsd4_compound_state *cstate,
1267                       struct nfsd4_copy *copy, struct vfsmount **mount)
1268 {
1269         struct svc_fh *s_fh = NULL;
1270         stateid_t *s_stid = &copy->cp_src_stateid;
1271         __be32 status = nfserr_inval;
1272
1273         /* Verify the destination stateid and set dst struct file*/
1274         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1275                                             &copy->cp_dst_stateid,
1276                                             WR_STATE, &copy->nf_dst, NULL);
1277         if (status)
1278                 goto out;
1279
1280         status = nfsd4_interssc_connect(&copy->cp_src, rqstp, mount);
1281         if (status)
1282                 goto out;
1283
1284         s_fh = &cstate->save_fh;
1285
1286         copy->c_fh.size = s_fh->fh_handle.fh_size;
1287         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_base, copy->c_fh.size);
1288         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1289         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1290                sizeof(stateid_opaque_t));
1291
1292         status = 0;
1293 out:
1294         return status;
1295 }
1296
1297 static void
1298 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1299                         struct nfsd_file *dst)
1300 {
1301         nfs42_ssc_close(src->nf_file);
1302         nfsd_file_put(src);
1303         nfsd_file_put(dst);
1304         mntput(ss_mnt);
1305 }
1306
1307 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1308
1309 static __be32
1310 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1311                       struct nfsd4_compound_state *cstate,
1312                       struct nfsd4_copy *copy,
1313                       struct vfsmount **mount)
1314 {
1315         *mount = NULL;
1316         return nfserr_inval;
1317 }
1318
1319 static void
1320 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1321                         struct nfsd_file *dst)
1322 {
1323 }
1324
1325 static void
1326 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1327 {
1328 }
1329
1330 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1331                                    struct nfs_fh *src_fh,
1332                                    nfs4_stateid *stateid)
1333 {
1334         return NULL;
1335 }
1336 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1337
1338 static __be32
1339 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1340                       struct nfsd4_compound_state *cstate,
1341                       struct nfsd4_copy *copy)
1342 {
1343         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1344                                  &copy->nf_src, &copy->cp_dst_stateid,
1345                                  &copy->nf_dst);
1346 }
1347
1348 static void
1349 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1350 {
1351         nfsd_file_put(src);
1352         nfsd_file_put(dst);
1353 }
1354
1355 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1356 {
1357         struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1358
1359         nfs4_put_copy(copy);
1360 }
1361
1362 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1363                                  struct rpc_task *task)
1364 {
1365         return 1;
1366 }
1367
1368 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1369         .release = nfsd4_cb_offload_release,
1370         .done = nfsd4_cb_offload_done
1371 };
1372
1373 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1374 {
1375         copy->cp_res.wr_stable_how = NFS_UNSTABLE;
1376         copy->cp_synchronous = sync;
1377         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1378 }
1379
1380 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1381 {
1382         ssize_t bytes_copied = 0;
1383         size_t bytes_total = copy->cp_count;
1384         u64 src_pos = copy->cp_src_pos;
1385         u64 dst_pos = copy->cp_dst_pos;
1386
1387         do {
1388                 if (kthread_should_stop())
1389                         break;
1390                 bytes_copied = nfsd_copy_file_range(copy->nf_src->nf_file,
1391                                 src_pos, copy->nf_dst->nf_file, dst_pos,
1392                                 bytes_total);
1393                 if (bytes_copied <= 0)
1394                         break;
1395                 bytes_total -= bytes_copied;
1396                 copy->cp_res.wr_bytes_written += bytes_copied;
1397                 src_pos += bytes_copied;
1398                 dst_pos += bytes_copied;
1399         } while (bytes_total > 0 && !copy->cp_synchronous);
1400         return bytes_copied;
1401 }
1402
1403 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1404 {
1405         __be32 status;
1406         ssize_t bytes;
1407
1408         bytes = _nfsd_copy_file_range(copy);
1409         /* for async copy, we ignore the error, client can always retry
1410          * to get the error
1411          */
1412         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1413                 status = nfserrno(bytes);
1414         else {
1415                 nfsd4_init_copy_res(copy, sync);
1416                 status = nfs_ok;
1417         }
1418
1419         if (!copy->cp_intra) /* Inter server SSC */
1420                 nfsd4_cleanup_inter_ssc(copy->ss_mnt, copy->nf_src,
1421                                         copy->nf_dst);
1422         else
1423                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1424
1425         return status;
1426 }
1427
1428 static int dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1429 {
1430         dst->cp_src_pos = src->cp_src_pos;
1431         dst->cp_dst_pos = src->cp_dst_pos;
1432         dst->cp_count = src->cp_count;
1433         dst->cp_synchronous = src->cp_synchronous;
1434         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1435         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1436         dst->cp_clp = src->cp_clp;
1437         dst->nf_dst = nfsd_file_get(src->nf_dst);
1438         dst->cp_intra = src->cp_intra;
1439         if (src->cp_intra) /* for inter, file_src doesn't exist yet */
1440                 dst->nf_src = nfsd_file_get(src->nf_src);
1441
1442         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1443         memcpy(&dst->cp_src, &src->cp_src, sizeof(struct nl4_server));
1444         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1445         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1446         dst->ss_mnt = src->ss_mnt;
1447
1448         return 0;
1449 }
1450
1451 static void cleanup_async_copy(struct nfsd4_copy *copy)
1452 {
1453         nfs4_free_copy_state(copy);
1454         nfsd_file_put(copy->nf_dst);
1455         if (copy->cp_intra)
1456                 nfsd_file_put(copy->nf_src);
1457         spin_lock(&copy->cp_clp->async_lock);
1458         list_del(&copy->copies);
1459         spin_unlock(&copy->cp_clp->async_lock);
1460         nfs4_put_copy(copy);
1461 }
1462
1463 static int nfsd4_do_async_copy(void *data)
1464 {
1465         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1466         struct nfsd4_copy *cb_copy;
1467
1468         if (!copy->cp_intra) { /* Inter server SSC */
1469                 copy->nf_src = kzalloc(sizeof(struct nfsd_file), GFP_KERNEL);
1470                 if (!copy->nf_src) {
1471                         copy->nfserr = nfserr_serverfault;
1472                         nfsd4_interssc_disconnect(copy->ss_mnt);
1473                         goto do_callback;
1474                 }
1475                 copy->nf_src->nf_file = nfs42_ssc_open(copy->ss_mnt, &copy->c_fh,
1476                                               &copy->stateid);
1477                 if (IS_ERR(copy->nf_src->nf_file)) {
1478                         copy->nfserr = nfserr_offload_denied;
1479                         nfsd4_interssc_disconnect(copy->ss_mnt);
1480                         goto do_callback;
1481                 }
1482         }
1483
1484         copy->nfserr = nfsd4_do_copy(copy, 0);
1485 do_callback:
1486         cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1487         if (!cb_copy)
1488                 goto out;
1489         memcpy(&cb_copy->cp_res, &copy->cp_res, sizeof(copy->cp_res));
1490         cb_copy->cp_clp = copy->cp_clp;
1491         cb_copy->nfserr = copy->nfserr;
1492         memcpy(&cb_copy->fh, &copy->fh, sizeof(copy->fh));
1493         nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1494                         &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1495         nfsd4_run_cb(&cb_copy->cp_cb);
1496 out:
1497         if (!copy->cp_intra)
1498                 kfree(copy->nf_src);
1499         cleanup_async_copy(copy);
1500         return 0;
1501 }
1502
1503 static __be32
1504 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1505                 union nfsd4_op_u *u)
1506 {
1507         struct nfsd4_copy *copy = &u->copy;
1508         __be32 status;
1509         struct nfsd4_copy *async_copy = NULL;
1510
1511         if (!copy->cp_intra) { /* Inter server SSC */
1512                 if (!inter_copy_offload_enable || copy->cp_synchronous) {
1513                         status = nfserr_notsupp;
1514                         goto out;
1515                 }
1516                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1517                                 &copy->ss_mnt);
1518                 if (status)
1519                         return nfserr_offload_denied;
1520         } else {
1521                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1522                 if (status)
1523                         return status;
1524         }
1525
1526         copy->cp_clp = cstate->clp;
1527         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1528                 sizeof(struct knfsd_fh));
1529         if (!copy->cp_synchronous) {
1530                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1531
1532                 status = nfserrno(-ENOMEM);
1533                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1534                 if (!async_copy)
1535                         goto out_err;
1536                 if (!nfs4_init_copy_state(nn, copy))
1537                         goto out_err;
1538                 refcount_set(&async_copy->refcount, 1);
1539                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid,
1540                         sizeof(copy->cp_stateid));
1541                 status = dup_copy_fields(copy, async_copy);
1542                 if (status)
1543                         goto out_err;
1544                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1545                                 async_copy, "%s", "copy thread");
1546                 if (IS_ERR(async_copy->copy_task))
1547                         goto out_err;
1548                 spin_lock(&async_copy->cp_clp->async_lock);
1549                 list_add(&async_copy->copies,
1550                                 &async_copy->cp_clp->async_copies);
1551                 spin_unlock(&async_copy->cp_clp->async_lock);
1552                 wake_up_process(async_copy->copy_task);
1553                 status = nfs_ok;
1554         } else {
1555                 status = nfsd4_do_copy(copy, 1);
1556         }
1557 out:
1558         return status;
1559 out_err:
1560         if (async_copy)
1561                 cleanup_async_copy(async_copy);
1562         status = nfserrno(-ENOMEM);
1563         if (!copy->cp_intra)
1564                 nfsd4_interssc_disconnect(copy->ss_mnt);
1565         goto out;
1566 }
1567
1568 struct nfsd4_copy *
1569 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1570 {
1571         struct nfsd4_copy *copy;
1572
1573         spin_lock(&clp->async_lock);
1574         list_for_each_entry(copy, &clp->async_copies, copies) {
1575                 if (memcmp(&copy->cp_stateid.stid, stateid, NFS4_STATEID_SIZE))
1576                         continue;
1577                 refcount_inc(&copy->refcount);
1578                 spin_unlock(&clp->async_lock);
1579                 return copy;
1580         }
1581         spin_unlock(&clp->async_lock);
1582         return NULL;
1583 }
1584
1585 static __be32
1586 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1587                      struct nfsd4_compound_state *cstate,
1588                      union nfsd4_op_u *u)
1589 {
1590         struct nfsd4_offload_status *os = &u->offload_status;
1591         struct nfsd4_copy *copy;
1592         struct nfs4_client *clp = cstate->clp;
1593
1594         copy = find_async_copy(clp, &os->stateid);
1595         if (!copy) {
1596                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1597
1598                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1599         } else
1600                 nfsd4_stop_copy(copy);
1601
1602         return nfs_ok;
1603 }
1604
1605 static __be32
1606 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1607                   union nfsd4_op_u *u)
1608 {
1609         struct nfsd4_copy_notify *cn = &u->copy_notify;
1610         __be32 status;
1611         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1612         struct nfs4_stid *stid;
1613         struct nfs4_cpntf_state *cps;
1614         struct nfs4_client *clp = cstate->clp;
1615
1616         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1617                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1618                                         &stid);
1619         if (status)
1620                 return status;
1621
1622         cn->cpn_sec = nn->nfsd4_lease;
1623         cn->cpn_nsec = 0;
1624
1625         status = nfserrno(-ENOMEM);
1626         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1627         if (!cps)
1628                 goto out;
1629         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.stid, sizeof(stateid_t));
1630         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1631         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1632
1633         /* For now, only return one server address in cpn_src, the
1634          * address used by the client to connect to this server.
1635          */
1636         cn->cpn_src.nl4_type = NL4_NETADDR;
1637         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1638                                  &cn->cpn_src.u.nl4_addr);
1639         WARN_ON_ONCE(status);
1640         if (status) {
1641                 nfs4_put_cpntf_state(nn, cps);
1642                 goto out;
1643         }
1644 out:
1645         nfs4_put_stid(stid);
1646         return status;
1647 }
1648
1649 static __be32
1650 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1651                 struct nfsd4_fallocate *fallocate, int flags)
1652 {
1653         __be32 status;
1654         struct nfsd_file *nf;
1655
1656         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1657                                             &fallocate->falloc_stateid,
1658                                             WR_STATE, &nf, NULL);
1659         if (status != nfs_ok) {
1660                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1661                 return status;
1662         }
1663
1664         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1665                                      fallocate->falloc_offset,
1666                                      fallocate->falloc_length,
1667                                      flags);
1668         nfsd_file_put(nf);
1669         return status;
1670 }
1671 static __be32
1672 nfsd4_offload_status(struct svc_rqst *rqstp,
1673                      struct nfsd4_compound_state *cstate,
1674                      union nfsd4_op_u *u)
1675 {
1676         struct nfsd4_offload_status *os = &u->offload_status;
1677         __be32 status = 0;
1678         struct nfsd4_copy *copy;
1679         struct nfs4_client *clp = cstate->clp;
1680
1681         copy = find_async_copy(clp, &os->stateid);
1682         if (copy) {
1683                 os->count = copy->cp_res.wr_bytes_written;
1684                 nfs4_put_copy(copy);
1685         } else
1686                 status = nfserr_bad_stateid;
1687
1688         return status;
1689 }
1690
1691 static __be32
1692 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1693                union nfsd4_op_u *u)
1694 {
1695         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1696 }
1697
1698 static __be32
1699 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1700                  union nfsd4_op_u *u)
1701 {
1702         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1703                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1704 }
1705
1706 static __be32
1707 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1708            union nfsd4_op_u *u)
1709 {
1710         struct nfsd4_seek *seek = &u->seek;
1711         int whence;
1712         __be32 status;
1713         struct nfsd_file *nf;
1714
1715         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1716                                             &seek->seek_stateid,
1717                                             RD_STATE, &nf, NULL);
1718         if (status) {
1719                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1720                 return status;
1721         }
1722
1723         switch (seek->seek_whence) {
1724         case NFS4_CONTENT_DATA:
1725                 whence = SEEK_DATA;
1726                 break;
1727         case NFS4_CONTENT_HOLE:
1728                 whence = SEEK_HOLE;
1729                 break;
1730         default:
1731                 status = nfserr_union_notsupp;
1732                 goto out;
1733         }
1734
1735         /*
1736          * Note:  This call does change file->f_pos, but nothing in NFSD
1737          *        should ever file->f_pos.
1738          */
1739         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
1740         if (seek->seek_pos < 0)
1741                 status = nfserrno(seek->seek_pos);
1742         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
1743                 seek->seek_eof = true;
1744
1745 out:
1746         nfsd_file_put(nf);
1747         return status;
1748 }
1749
1750 /* This routine never returns NFS_OK!  If there are no other errors, it
1751  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1752  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
1753  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1754  */
1755 static __be32
1756 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1757              struct nfsd4_verify *verify)
1758 {
1759         __be32 *buf, *p;
1760         int count;
1761         __be32 status;
1762
1763         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1764         if (status)
1765                 return status;
1766
1767         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1768         if (status)
1769                 return status;
1770
1771         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1772             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1773                 return nfserr_inval;
1774         if (verify->ve_attrlen & 3)
1775                 return nfserr_inval;
1776
1777         /* count in words:
1778          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1779          */
1780         count = 4 + (verify->ve_attrlen >> 2);
1781         buf = kmalloc(count << 2, GFP_KERNEL);
1782         if (!buf)
1783                 return nfserr_jukebox;
1784
1785         p = buf;
1786         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1787                                     cstate->current_fh.fh_export,
1788                                     cstate->current_fh.fh_dentry,
1789                                     verify->ve_bmval,
1790                                     rqstp, 0);
1791         /*
1792          * If nfsd4_encode_fattr() ran out of space, assume that's because
1793          * the attributes are longer (hence different) than those given:
1794          */
1795         if (status == nfserr_resource)
1796                 status = nfserr_not_same;
1797         if (status)
1798                 goto out_kfree;
1799
1800         /* skip bitmap */
1801         p = buf + 1 + ntohl(buf[0]);
1802         status = nfserr_not_same;
1803         if (ntohl(*p++) != verify->ve_attrlen)
1804                 goto out_kfree;
1805         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1806                 status = nfserr_same;
1807
1808 out_kfree:
1809         kfree(buf);
1810         return status;
1811 }
1812
1813 static __be32
1814 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1815               union nfsd4_op_u *u)
1816 {
1817         __be32 status;
1818
1819         status = _nfsd4_verify(rqstp, cstate, &u->verify);
1820         return status == nfserr_not_same ? nfs_ok : status;
1821 }
1822
1823 static __be32
1824 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1825              union nfsd4_op_u *u)
1826 {
1827         __be32 status;
1828
1829         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1830         return status == nfserr_same ? nfs_ok : status;
1831 }
1832
1833 #ifdef CONFIG_NFSD_PNFS
1834 static const struct nfsd4_layout_ops *
1835 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1836 {
1837         if (!exp->ex_layout_types) {
1838                 dprintk("%s: export does not support pNFS\n", __func__);
1839                 return NULL;
1840         }
1841
1842         if (layout_type >= LAYOUT_TYPE_MAX ||
1843             !(exp->ex_layout_types & (1 << layout_type))) {
1844                 dprintk("%s: layout type %d not supported\n",
1845                         __func__, layout_type);
1846                 return NULL;
1847         }
1848
1849         return nfsd4_layout_ops[layout_type];
1850 }
1851
1852 static __be32
1853 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
1854                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1855 {
1856         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
1857         const struct nfsd4_layout_ops *ops;
1858         struct nfsd4_deviceid_map *map;
1859         struct svc_export *exp;
1860         __be32 nfserr;
1861
1862         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
1863                __func__,
1864                gdp->gd_layout_type,
1865                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
1866                gdp->gd_maxcount);
1867
1868         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
1869         if (!map) {
1870                 dprintk("%s: couldn't find device ID to export mapping!\n",
1871                         __func__);
1872                 return nfserr_noent;
1873         }
1874
1875         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
1876         if (IS_ERR(exp)) {
1877                 dprintk("%s: could not find device id\n", __func__);
1878                 return nfserr_noent;
1879         }
1880
1881         nfserr = nfserr_layoutunavailable;
1882         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
1883         if (!ops)
1884                 goto out;
1885
1886         nfserr = nfs_ok;
1887         if (gdp->gd_maxcount != 0) {
1888                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
1889                                 rqstp, cstate->session->se_client, gdp);
1890         }
1891
1892         gdp->gd_notify_types &= ops->notify_types;
1893 out:
1894         exp_put(exp);
1895         return nfserr;
1896 }
1897
1898 static void
1899 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
1900 {
1901         kfree(u->getdeviceinfo.gd_device);
1902 }
1903
1904 static __be32
1905 nfsd4_layoutget(struct svc_rqst *rqstp,
1906                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1907 {
1908         struct nfsd4_layoutget *lgp = &u->layoutget;
1909         struct svc_fh *current_fh = &cstate->current_fh;
1910         const struct nfsd4_layout_ops *ops;
1911         struct nfs4_layout_stateid *ls;
1912         __be32 nfserr;
1913         int accmode = NFSD_MAY_READ_IF_EXEC;
1914
1915         switch (lgp->lg_seg.iomode) {
1916         case IOMODE_READ:
1917                 accmode |= NFSD_MAY_READ;
1918                 break;
1919         case IOMODE_RW:
1920                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
1921                 break;
1922         default:
1923                 dprintk("%s: invalid iomode %d\n",
1924                         __func__, lgp->lg_seg.iomode);
1925                 nfserr = nfserr_badiomode;
1926                 goto out;
1927         }
1928
1929         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
1930         if (nfserr)
1931                 goto out;
1932
1933         nfserr = nfserr_layoutunavailable;
1934         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
1935         if (!ops)
1936                 goto out;
1937
1938         /*
1939          * Verify minlength and range as per RFC5661:
1940          *  o  If loga_length is less than loga_minlength,
1941          *     the metadata server MUST return NFS4ERR_INVAL.
1942          *  o  If the sum of loga_offset and loga_minlength exceeds
1943          *     NFS4_UINT64_MAX, and loga_minlength is not
1944          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
1945          *  o  If the sum of loga_offset and loga_length exceeds
1946          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
1947          *     the error NFS4ERR_INVAL MUST result.
1948          */
1949         nfserr = nfserr_inval;
1950         if (lgp->lg_seg.length < lgp->lg_minlength ||
1951             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
1952              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
1953             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
1954              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
1955                 goto out;
1956         if (lgp->lg_seg.length == 0)
1957                 goto out;
1958
1959         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
1960                                                 true, lgp->lg_layout_type, &ls);
1961         if (nfserr) {
1962                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
1963                 goto out;
1964         }
1965
1966         nfserr = nfserr_recallconflict;
1967         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
1968                 goto out_put_stid;
1969
1970         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
1971                                      current_fh, lgp);
1972         if (nfserr)
1973                 goto out_put_stid;
1974
1975         nfserr = nfsd4_insert_layout(lgp, ls);
1976
1977 out_put_stid:
1978         mutex_unlock(&ls->ls_mutex);
1979         nfs4_put_stid(&ls->ls_stid);
1980 out:
1981         return nfserr;
1982 }
1983
1984 static void
1985 nfsd4_layoutget_release(union nfsd4_op_u *u)
1986 {
1987         kfree(u->layoutget.lg_content);
1988 }
1989
1990 static __be32
1991 nfsd4_layoutcommit(struct svc_rqst *rqstp,
1992                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1993 {
1994         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
1995         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
1996         struct svc_fh *current_fh = &cstate->current_fh;
1997         const struct nfsd4_layout_ops *ops;
1998         loff_t new_size = lcp->lc_last_wr + 1;
1999         struct inode *inode;
2000         struct nfs4_layout_stateid *ls;
2001         __be32 nfserr;
2002
2003         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2004         if (nfserr)
2005                 goto out;
2006
2007         nfserr = nfserr_layoutunavailable;
2008         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2009         if (!ops)
2010                 goto out;
2011         inode = d_inode(current_fh->fh_dentry);
2012
2013         nfserr = nfserr_inval;
2014         if (new_size <= seg->offset) {
2015                 dprintk("pnfsd: last write before layout segment\n");
2016                 goto out;
2017         }
2018         if (new_size > seg->offset + seg->length) {
2019                 dprintk("pnfsd: last write beyond layout segment\n");
2020                 goto out;
2021         }
2022         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2023                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2024                 goto out;
2025         }
2026
2027         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2028                                                 false, lcp->lc_layout_type,
2029                                                 &ls);
2030         if (nfserr) {
2031                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2032                 /* fixup error code as per RFC5661 */
2033                 if (nfserr == nfserr_bad_stateid)
2034                         nfserr = nfserr_badlayout;
2035                 goto out;
2036         }
2037
2038         /* LAYOUTCOMMIT does not require any serialization */
2039         mutex_unlock(&ls->ls_mutex);
2040
2041         if (new_size > i_size_read(inode)) {
2042                 lcp->lc_size_chg = 1;
2043                 lcp->lc_newsize = new_size;
2044         } else {
2045                 lcp->lc_size_chg = 0;
2046         }
2047
2048         nfserr = ops->proc_layoutcommit(inode, lcp);
2049         nfs4_put_stid(&ls->ls_stid);
2050 out:
2051         return nfserr;
2052 }
2053
2054 static __be32
2055 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2056                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2057 {
2058         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2059         struct svc_fh *current_fh = &cstate->current_fh;
2060         __be32 nfserr;
2061
2062         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2063         if (nfserr)
2064                 goto out;
2065
2066         nfserr = nfserr_layoutunavailable;
2067         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2068                 goto out;
2069
2070         switch (lrp->lr_seg.iomode) {
2071         case IOMODE_READ:
2072         case IOMODE_RW:
2073         case IOMODE_ANY:
2074                 break;
2075         default:
2076                 dprintk("%s: invalid iomode %d\n", __func__,
2077                         lrp->lr_seg.iomode);
2078                 nfserr = nfserr_inval;
2079                 goto out;
2080         }
2081
2082         switch (lrp->lr_return_type) {
2083         case RETURN_FILE:
2084                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2085                 break;
2086         case RETURN_FSID:
2087         case RETURN_ALL:
2088                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2089                 break;
2090         default:
2091                 dprintk("%s: invalid return_type %d\n", __func__,
2092                         lrp->lr_return_type);
2093                 nfserr = nfserr_inval;
2094                 break;
2095         }
2096 out:
2097         return nfserr;
2098 }
2099 #endif /* CONFIG_NFSD_PNFS */
2100
2101 static __be32
2102 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2103                union nfsd4_op_u *u)
2104 {
2105         struct nfsd4_getxattr *getxattr = &u->getxattr;
2106
2107         return nfsd_getxattr(rqstp, &cstate->current_fh,
2108                              getxattr->getxa_name, &getxattr->getxa_buf,
2109                              &getxattr->getxa_len);
2110 }
2111
2112 static __be32
2113 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2114            union nfsd4_op_u *u)
2115 {
2116         struct nfsd4_setxattr *setxattr = &u->setxattr;
2117         __be32 ret;
2118
2119         if (opens_in_grace(SVC_NET(rqstp)))
2120                 return nfserr_grace;
2121
2122         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2123                             setxattr->setxa_buf, setxattr->setxa_len,
2124                             setxattr->setxa_flags);
2125
2126         if (!ret)
2127                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2128
2129         return ret;
2130 }
2131
2132 static __be32
2133 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2134            union nfsd4_op_u *u)
2135 {
2136         /*
2137          * Get the entire list, then copy out only the user attributes
2138          * in the encode function.
2139          */
2140         return nfsd_listxattr(rqstp, &cstate->current_fh,
2141                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2142 }
2143
2144 static __be32
2145 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2146            union nfsd4_op_u *u)
2147 {
2148         struct nfsd4_removexattr *removexattr = &u->removexattr;
2149         __be32 ret;
2150
2151         if (opens_in_grace(SVC_NET(rqstp)))
2152                 return nfserr_grace;
2153
2154         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2155             removexattr->rmxa_name);
2156
2157         if (!ret)
2158                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2159
2160         return ret;
2161 }
2162
2163 /*
2164  * NULL call.
2165  */
2166 static __be32
2167 nfsd4_proc_null(struct svc_rqst *rqstp)
2168 {
2169         return rpc_success;
2170 }
2171
2172 static inline void nfsd4_increment_op_stats(u32 opnum)
2173 {
2174         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2175                 nfsdstats.nfs4_opcount[opnum]++;
2176 }
2177
2178 static const struct nfsd4_operation nfsd4_ops[];
2179
2180 static const char *nfsd4_op_name(unsigned opnum);
2181
2182 /*
2183  * Enforce NFSv4.1 COMPOUND ordering rules:
2184  *
2185  * Also note, enforced elsewhere:
2186  *      - SEQUENCE other than as first op results in
2187  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2188  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2189  *        (Enforced in nfsd4_bind_conn_to_session().)
2190  *      - DESTROY_SESSION must be the final operation in a compound, if
2191  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2192  *        (Enforced in nfsd4_destroy_session().)
2193  */
2194 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2195 {
2196         struct nfsd4_op *first_op = &args->ops[0];
2197
2198         /* These ordering requirements don't apply to NFSv4.0: */
2199         if (args->minorversion == 0)
2200                 return nfs_ok;
2201         /* This is weird, but OK, not our problem: */
2202         if (args->opcnt == 0)
2203                 return nfs_ok;
2204         if (first_op->status == nfserr_op_illegal)
2205                 return nfs_ok;
2206         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2207                 return nfserr_op_not_in_session;
2208         if (first_op->opnum == OP_SEQUENCE)
2209                 return nfs_ok;
2210         /*
2211          * So first_op is something allowed outside a session, like
2212          * EXCHANGE_ID; but then it has to be the only op in the
2213          * compound:
2214          */
2215         if (args->opcnt != 1)
2216                 return nfserr_not_only_op;
2217         return nfs_ok;
2218 }
2219
2220 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2221 {
2222         return &nfsd4_ops[op->opnum];
2223 }
2224
2225 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2226 {
2227         if (op->opnum == OP_ILLEGAL)
2228                 return false;
2229         return OPDESC(op)->op_flags & OP_CACHEME;
2230 }
2231
2232 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2233 {
2234         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2235         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2236         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2237         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2238         const struct nfsd4_operation *thisd = OPDESC(this);
2239         const struct nfsd4_operation *nextd;
2240
2241         /*
2242          * Most ops check wronsec on our own; only the putfh-like ops
2243          * have special rules.
2244          */
2245         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2246                 return false;
2247         /*
2248          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2249          * put-filehandle operation if we're not going to use the
2250          * result:
2251          */
2252         if (argp->opcnt == resp->opcnt)
2253                 return false;
2254         if (next->opnum == OP_ILLEGAL)
2255                 return false;
2256         nextd = OPDESC(next);
2257         /*
2258          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2259          * errors themselves as necessary; others should check for them
2260          * now:
2261          */
2262         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2263 }
2264
2265 static void svcxdr_init_encode(struct svc_rqst *rqstp,
2266                                struct nfsd4_compoundres *resp)
2267 {
2268         struct xdr_stream *xdr = &resp->xdr;
2269         struct xdr_buf *buf = &rqstp->rq_res;
2270         struct kvec *head = buf->head;
2271
2272         xdr->buf = buf;
2273         xdr->iov = head;
2274         xdr->p   = head->iov_base + head->iov_len;
2275         xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
2276         /* Tail and page_len should be zero at this point: */
2277         buf->len = buf->head[0].iov_len;
2278         xdr->scratch.iov_len = 0;
2279         xdr->page_ptr = buf->pages - 1;
2280         buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
2281                 - rqstp->rq_auth_slack;
2282 }
2283
2284 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2285 static void
2286 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2287 {
2288         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2289         struct nfsd4_copy *copy;
2290         struct nfsd4_putfh *putfh;
2291         int i;
2292
2293         /* traverse all operation and if it's a COPY compound, mark the
2294          * source filehandle to skip verification
2295          */
2296         for (i = 0; i < args->opcnt; i++) {
2297                 op = &args->ops[i];
2298                 if (op->opnum == OP_PUTFH)
2299                         current_op = op;
2300                 else if (op->opnum == OP_SAVEFH)
2301                         saved_op = current_op;
2302                 else if (op->opnum == OP_RESTOREFH)
2303                         current_op = saved_op;
2304                 else if (op->opnum == OP_COPY) {
2305                         copy = (struct nfsd4_copy *)&op->u;
2306                         if (!saved_op) {
2307                                 op->status = nfserr_nofilehandle;
2308                                 return;
2309                         }
2310                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2311                         if (!copy->cp_intra)
2312                                 putfh->no_verify = true;
2313                 }
2314         }
2315 }
2316 #else
2317 static void
2318 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2319 {
2320 }
2321 #endif
2322
2323 /*
2324  * COMPOUND call.
2325  */
2326 static __be32
2327 nfsd4_proc_compound(struct svc_rqst *rqstp)
2328 {
2329         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2330         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2331         struct nfsd4_op *op;
2332         struct nfsd4_compound_state *cstate = &resp->cstate;
2333         struct svc_fh *current_fh = &cstate->current_fh;
2334         struct svc_fh *save_fh = &cstate->save_fh;
2335         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2336         __be32          status;
2337
2338         svcxdr_init_encode(rqstp, resp);
2339         resp->tagp = resp->xdr.p;
2340         /* reserve space for: taglen, tag, and opcnt */
2341         xdr_reserve_space(&resp->xdr, 8 + args->taglen);
2342         resp->taglen = args->taglen;
2343         resp->tag = args->tag;
2344         resp->rqstp = rqstp;
2345         cstate->minorversion = args->minorversion;
2346         fh_init(current_fh, NFS4_FHSIZE);
2347         fh_init(save_fh, NFS4_FHSIZE);
2348         /*
2349          * Don't use the deferral mechanism for NFSv4; compounds make it
2350          * too hard to avoid non-idempotency problems.
2351          */
2352         clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2353
2354         /*
2355          * According to RFC3010, this takes precedence over all other errors.
2356          */
2357         status = nfserr_minor_vers_mismatch;
2358         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2359                 goto out;
2360         status = nfserr_resource;
2361         if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
2362                 goto out;
2363
2364         status = nfs41_check_op_ordering(args);
2365         if (status) {
2366                 op = &args->ops[0];
2367                 op->status = status;
2368                 resp->opcnt = 1;
2369                 goto encode_op;
2370         }
2371         check_if_stalefh_allowed(args);
2372
2373         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2374
2375         trace_nfsd_compound(rqstp, args->opcnt);
2376         while (!status && resp->opcnt < args->opcnt) {
2377                 op = &args->ops[resp->opcnt++];
2378
2379                 /*
2380                  * The XDR decode routines may have pre-set op->status;
2381                  * for example, if there is a miscellaneous XDR error
2382                  * it will be set to nfserr_bad_xdr.
2383                  */
2384                 if (op->status) {
2385                         if (op->opnum == OP_OPEN)
2386                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2387                         goto encode_op;
2388                 }
2389                 if (!current_fh->fh_dentry &&
2390                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2391                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2392                                 op->status = nfserr_nofilehandle;
2393                                 goto encode_op;
2394                         }
2395                 } else if (current_fh->fh_export &&
2396                            current_fh->fh_export->ex_fslocs.migrated &&
2397                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2398                         op->status = nfserr_moved;
2399                         goto encode_op;
2400                 }
2401
2402                 fh_clear_wcc(current_fh);
2403
2404                 /* If op is non-idempotent */
2405                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2406                         /*
2407                          * Don't execute this op if we couldn't encode a
2408                          * succesful reply:
2409                          */
2410                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2411                         /*
2412                          * Plus if there's another operation, make sure
2413                          * we'll have space to at least encode an error:
2414                          */
2415                         if (resp->opcnt < args->opcnt)
2416                                 plen += COMPOUND_ERR_SLACK_SPACE;
2417                         op->status = nfsd4_check_resp_size(resp, plen);
2418                 }
2419
2420                 if (op->status)
2421                         goto encode_op;
2422
2423                 if (op->opdesc->op_get_currentstateid)
2424                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2425                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2426
2427                 /* Only from SEQUENCE */
2428                 if (cstate->status == nfserr_replay_cache) {
2429                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2430                         status = op->status;
2431                         goto out;
2432                 }
2433                 if (!op->status) {
2434                         if (op->opdesc->op_set_currentstateid)
2435                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2436
2437                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2438                                 clear_current_stateid(cstate);
2439
2440                         if (current_fh->fh_export &&
2441                                         need_wrongsec_check(rqstp))
2442                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2443                 }
2444 encode_op:
2445                 if (op->status == nfserr_replay_me) {
2446                         op->replay = &cstate->replay_owner->so_replay;
2447                         nfsd4_encode_replay(&resp->xdr, op);
2448                         status = op->status = op->replay->rp_status;
2449                 } else {
2450                         nfsd4_encode_operation(resp, op);
2451                         status = op->status;
2452                 }
2453
2454                 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2455                                            nfsd4_op_name(op->opnum));
2456
2457                 nfsd4_cstate_clear_replay(cstate);
2458                 nfsd4_increment_op_stats(op->opnum);
2459         }
2460
2461         fh_put(current_fh);
2462         fh_put(save_fh);
2463         BUG_ON(cstate->replay_owner);
2464 out:
2465         cstate->status = status;
2466         /* Reset deferral mechanism for RPC deferrals */
2467         set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2468         return rpc_success;
2469 }
2470
2471 #define op_encode_hdr_size              (2)
2472 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2473 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2474 #define op_encode_change_info_maxsz     (5)
2475 #define nfs4_fattr_bitmap_maxsz         (4)
2476
2477 /* We'll fall back on returning no lockowner if run out of space: */
2478 #define op_encode_lockowner_maxsz       (0)
2479 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2480
2481 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2482
2483 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2484 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2485                                          op_encode_ace_maxsz)
2486
2487 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2488
2489 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2490 {
2491         return (op_encode_hdr_size) * sizeof(__be32);
2492 }
2493
2494 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2495 {
2496         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2497 }
2498
2499 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2500 {
2501         /* ac_supported, ac_resp_access */
2502         return (op_encode_hdr_size + 2)* sizeof(__be32);
2503 }
2504
2505 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2506 {
2507         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2508 }
2509
2510 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2511 {
2512         return (op_encode_hdr_size + op_encode_change_info_maxsz
2513                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2514 }
2515
2516 /*
2517  * Note since this is an idempotent operation we won't insist on failing
2518  * the op prematurely if the estimate is too large.  We may turn off splice
2519  * reads unnecessarily.
2520  */
2521 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2522                                       struct nfsd4_op *op)
2523 {
2524         u32 *bmap = op->u.getattr.ga_bmval;
2525         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2526         u32 ret = 0;
2527
2528         if (bmap0 & FATTR4_WORD0_ACL)
2529                 return svc_max_payload(rqstp);
2530         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2531                 return svc_max_payload(rqstp);
2532
2533         if (bmap1 & FATTR4_WORD1_OWNER) {
2534                 ret += IDMAP_NAMESZ + 4;
2535                 bmap1 &= ~FATTR4_WORD1_OWNER;
2536         }
2537         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2538                 ret += IDMAP_NAMESZ + 4;
2539                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2540         }
2541         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2542                 ret += NFS4_FHSIZE + 4;
2543                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2544         }
2545         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2546                 ret += NFS4_MAXLABELLEN + 12;
2547                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2548         }
2549         /*
2550          * Largest of remaining attributes are 16 bytes (e.g.,
2551          * supported_attributes)
2552          */
2553         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2554         /* bitmask, length */
2555         ret += 20;
2556         return ret;
2557 }
2558
2559 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2560 {
2561         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2562 }
2563
2564 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2565 {
2566         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2567                 * sizeof(__be32);
2568 }
2569
2570 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2571 {
2572         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2573                 * sizeof(__be32);
2574 }
2575
2576 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2577 {
2578         return (op_encode_hdr_size + op_encode_stateid_maxsz
2579                 + op_encode_change_info_maxsz + 1
2580                 + nfs4_fattr_bitmap_maxsz
2581                 + op_encode_delegation_maxsz) * sizeof(__be32);
2582 }
2583
2584 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2585 {
2586         u32 maxcount = 0, rlen = 0;
2587
2588         maxcount = svc_max_payload(rqstp);
2589         rlen = min(op->u.read.rd_length, maxcount);
2590
2591         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2592 }
2593
2594 static inline u32 nfsd4_read_plus_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2595 {
2596         u32 maxcount = svc_max_payload(rqstp);
2597         u32 rlen = min(op->u.read.rd_length, maxcount);
2598         /*
2599          * If we detect that the file changed during hole encoding, then we
2600          * recover by encoding the remaining reply as data. This means we need
2601          * to set aside enough room to encode two data segments.
2602          */
2603         u32 seg_len = 2 * (1 + 2 + 1);
2604
2605         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2606 }
2607
2608 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2609 {
2610         u32 maxcount = 0, rlen = 0;
2611
2612         maxcount = svc_max_payload(rqstp);
2613         rlen = min(op->u.readdir.rd_maxcount, maxcount);
2614
2615         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2616                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2617 }
2618
2619 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2620 {
2621         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2622 }
2623
2624 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2625 {
2626         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2627                 * sizeof(__be32);
2628 }
2629
2630 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2631 {
2632         return (op_encode_hdr_size + op_encode_change_info_maxsz
2633                 + op_encode_change_info_maxsz) * sizeof(__be32);
2634 }
2635
2636 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2637                                        struct nfsd4_op *op)
2638 {
2639         return (op_encode_hdr_size
2640                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2641 }
2642
2643 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2644 {
2645         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2646                 * sizeof(__be32);
2647 }
2648
2649 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2650 {
2651         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2652 }
2653
2654 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2655 {
2656         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2657                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2658 }
2659
2660 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2661 {
2662         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2663                                                                 sizeof(__be32);
2664 }
2665
2666 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2667 {
2668         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2669 }
2670
2671 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2672 {
2673         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2674                 1 + 1 + /* eir_flags, spr_how */\
2675                 4 + /* spo_must_enforce & _allow with bitmap */\
2676                 2 + /*eir_server_owner.so_minor_id */\
2677                 /* eir_server_owner.so_major_id<> */\
2678                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2679                 /* eir_server_scope<> */\
2680                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2681                 1 + /* eir_server_impl_id array length */\
2682                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2683 }
2684
2685 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2686 {
2687         return (op_encode_hdr_size + \
2688                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2689                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2690 }
2691
2692 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2693 {
2694         return (op_encode_hdr_size + \
2695                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2696                 2 + /* csr_sequence, csr_flags */\
2697                 op_encode_channel_attrs_maxsz + \
2698                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2699 }
2700
2701 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2702 {
2703         return (op_encode_hdr_size +
2704                 1 /* wr_callback */ +
2705                 op_encode_stateid_maxsz /* wr_callback */ +
2706                 2 /* wr_count */ +
2707                 1 /* wr_committed */ +
2708                 op_encode_verifier_maxsz +
2709                 1 /* cr_consecutive */ +
2710                 1 /* cr_synchronous */) * sizeof(__be32);
2711 }
2712
2713 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2714                                              struct nfsd4_op *op)
2715 {
2716         return (op_encode_hdr_size +
2717                 2 /* osr_count */ +
2718                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2719 }
2720
2721 static inline u32 nfsd4_copy_notify_rsize(struct svc_rqst *rqstp,
2722                                         struct nfsd4_op *op)
2723 {
2724         return (op_encode_hdr_size +
2725                 3 /* cnr_lease_time */ +
2726                 1 /* We support one cnr_source_server */ +
2727                 1 /* cnr_stateid seq */ +
2728                 op_encode_stateid_maxsz /* cnr_stateid */ +
2729                 1 /* num cnr_source_server*/ +
2730                 1 /* nl4_type */ +
2731                 1 /* nl4 size */ +
2732                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
2733                 * sizeof(__be32);
2734 }
2735
2736 #ifdef CONFIG_NFSD_PNFS
2737 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2738 {
2739         u32 maxcount = 0, rlen = 0;
2740
2741         maxcount = svc_max_payload(rqstp);
2742         rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2743
2744         return (op_encode_hdr_size +
2745                 1 /* gd_layout_type*/ +
2746                 XDR_QUADLEN(rlen) +
2747                 2 /* gd_notify_types */) * sizeof(__be32);
2748 }
2749
2750 /*
2751  * At this stage we don't really know what layout driver will handle the request,
2752  * so we need to define an arbitrary upper bound here.
2753  */
2754 #define MAX_LAYOUT_SIZE         128
2755 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2756 {
2757         return (op_encode_hdr_size +
2758                 1 /* logr_return_on_close */ +
2759                 op_encode_stateid_maxsz +
2760                 1 /* nr of layouts */ +
2761                 MAX_LAYOUT_SIZE) * sizeof(__be32);
2762 }
2763
2764 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2765 {
2766         return (op_encode_hdr_size +
2767                 1 /* locr_newsize */ +
2768                 2 /* ns_size */) * sizeof(__be32);
2769 }
2770
2771 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2772 {
2773         return (op_encode_hdr_size +
2774                 1 /* lrs_stateid */ +
2775                 op_encode_stateid_maxsz) * sizeof(__be32);
2776 }
2777 #endif /* CONFIG_NFSD_PNFS */
2778
2779
2780 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2781 {
2782         return (op_encode_hdr_size + 3) * sizeof(__be32);
2783 }
2784
2785 static inline u32 nfsd4_getxattr_rsize(struct svc_rqst *rqstp,
2786                                        struct nfsd4_op *op)
2787 {
2788         u32 maxcount, rlen;
2789
2790         maxcount = svc_max_payload(rqstp);
2791         rlen = min_t(u32, XATTR_SIZE_MAX, maxcount);
2792
2793         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2794 }
2795
2796 static inline u32 nfsd4_setxattr_rsize(struct svc_rqst *rqstp,
2797                                        struct nfsd4_op *op)
2798 {
2799         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2800                 * sizeof(__be32);
2801 }
2802 static inline u32 nfsd4_listxattrs_rsize(struct svc_rqst *rqstp,
2803                                          struct nfsd4_op *op)
2804 {
2805         u32 maxcount, rlen;
2806
2807         maxcount = svc_max_payload(rqstp);
2808         rlen = min(op->u.listxattrs.lsxa_maxcount, maxcount);
2809
2810         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2811 }
2812
2813 static inline u32 nfsd4_removexattr_rsize(struct svc_rqst *rqstp,
2814                                           struct nfsd4_op *op)
2815 {
2816         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2817                 * sizeof(__be32);
2818 }
2819
2820
2821 static const struct nfsd4_operation nfsd4_ops[] = {
2822         [OP_ACCESS] = {
2823                 .op_func = nfsd4_access,
2824                 .op_name = "OP_ACCESS",
2825                 .op_rsize_bop = nfsd4_access_rsize,
2826         },
2827         [OP_CLOSE] = {
2828                 .op_func = nfsd4_close,
2829                 .op_flags = OP_MODIFIES_SOMETHING,
2830                 .op_name = "OP_CLOSE",
2831                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2832                 .op_get_currentstateid = nfsd4_get_closestateid,
2833                 .op_set_currentstateid = nfsd4_set_closestateid,
2834         },
2835         [OP_COMMIT] = {
2836                 .op_func = nfsd4_commit,
2837                 .op_flags = OP_MODIFIES_SOMETHING,
2838                 .op_name = "OP_COMMIT",
2839                 .op_rsize_bop = nfsd4_commit_rsize,
2840         },
2841         [OP_CREATE] = {
2842                 .op_func = nfsd4_create,
2843                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2844                 .op_name = "OP_CREATE",
2845                 .op_rsize_bop = nfsd4_create_rsize,
2846         },
2847         [OP_DELEGRETURN] = {
2848                 .op_func = nfsd4_delegreturn,
2849                 .op_flags = OP_MODIFIES_SOMETHING,
2850                 .op_name = "OP_DELEGRETURN",
2851                 .op_rsize_bop = nfsd4_only_status_rsize,
2852                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2853         },
2854         [OP_GETATTR] = {
2855                 .op_func = nfsd4_getattr,
2856                 .op_flags = ALLOWED_ON_ABSENT_FS,
2857                 .op_rsize_bop = nfsd4_getattr_rsize,
2858                 .op_name = "OP_GETATTR",
2859         },
2860         [OP_GETFH] = {
2861                 .op_func = nfsd4_getfh,
2862                 .op_name = "OP_GETFH",
2863                 .op_rsize_bop = nfsd4_getfh_rsize,
2864         },
2865         [OP_LINK] = {
2866                 .op_func = nfsd4_link,
2867                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
2868                                 | OP_CACHEME,
2869                 .op_name = "OP_LINK",
2870                 .op_rsize_bop = nfsd4_link_rsize,
2871         },
2872         [OP_LOCK] = {
2873                 .op_func = nfsd4_lock,
2874                 .op_flags = OP_MODIFIES_SOMETHING |
2875                                 OP_NONTRIVIAL_ERROR_ENCODE,
2876                 .op_name = "OP_LOCK",
2877                 .op_rsize_bop = nfsd4_lock_rsize,
2878                 .op_set_currentstateid = nfsd4_set_lockstateid,
2879         },
2880         [OP_LOCKT] = {
2881                 .op_func = nfsd4_lockt,
2882                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
2883                 .op_name = "OP_LOCKT",
2884                 .op_rsize_bop = nfsd4_lock_rsize,
2885         },
2886         [OP_LOCKU] = {
2887                 .op_func = nfsd4_locku,
2888                 .op_flags = OP_MODIFIES_SOMETHING,
2889                 .op_name = "OP_LOCKU",
2890                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2891                 .op_get_currentstateid = nfsd4_get_lockustateid,
2892         },
2893         [OP_LOOKUP] = {
2894                 .op_func = nfsd4_lookup,
2895                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2896                 .op_name = "OP_LOOKUP",
2897                 .op_rsize_bop = nfsd4_only_status_rsize,
2898         },
2899         [OP_LOOKUPP] = {
2900                 .op_func = nfsd4_lookupp,
2901                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2902                 .op_name = "OP_LOOKUPP",
2903                 .op_rsize_bop = nfsd4_only_status_rsize,
2904         },
2905         [OP_NVERIFY] = {
2906                 .op_func = nfsd4_nverify,
2907                 .op_name = "OP_NVERIFY",
2908                 .op_rsize_bop = nfsd4_only_status_rsize,
2909         },
2910         [OP_OPEN] = {
2911                 .op_func = nfsd4_open,
2912                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2913                 .op_name = "OP_OPEN",
2914                 .op_rsize_bop = nfsd4_open_rsize,
2915                 .op_set_currentstateid = nfsd4_set_openstateid,
2916         },
2917         [OP_OPEN_CONFIRM] = {
2918                 .op_func = nfsd4_open_confirm,
2919                 .op_flags = OP_MODIFIES_SOMETHING,
2920                 .op_name = "OP_OPEN_CONFIRM",
2921                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2922         },
2923         [OP_OPEN_DOWNGRADE] = {
2924                 .op_func = nfsd4_open_downgrade,
2925                 .op_flags = OP_MODIFIES_SOMETHING,
2926                 .op_name = "OP_OPEN_DOWNGRADE",
2927                 .op_rsize_bop = nfsd4_status_stateid_rsize,
2928                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
2929                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
2930         },
2931         [OP_PUTFH] = {
2932                 .op_func = nfsd4_putfh,
2933                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2934                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2935                 .op_name = "OP_PUTFH",
2936                 .op_rsize_bop = nfsd4_only_status_rsize,
2937         },
2938         [OP_PUTPUBFH] = {
2939                 .op_func = nfsd4_putrootfh,
2940                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2941                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2942                 .op_name = "OP_PUTPUBFH",
2943                 .op_rsize_bop = nfsd4_only_status_rsize,
2944         },
2945         [OP_PUTROOTFH] = {
2946                 .op_func = nfsd4_putrootfh,
2947                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2948                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2949                 .op_name = "OP_PUTROOTFH",
2950                 .op_rsize_bop = nfsd4_only_status_rsize,
2951         },
2952         [OP_READ] = {
2953                 .op_func = nfsd4_read,
2954                 .op_release = nfsd4_read_release,
2955                 .op_name = "OP_READ",
2956                 .op_rsize_bop = nfsd4_read_rsize,
2957                 .op_get_currentstateid = nfsd4_get_readstateid,
2958         },
2959         [OP_READDIR] = {
2960                 .op_func = nfsd4_readdir,
2961                 .op_name = "OP_READDIR",
2962                 .op_rsize_bop = nfsd4_readdir_rsize,
2963         },
2964         [OP_READLINK] = {
2965                 .op_func = nfsd4_readlink,
2966                 .op_name = "OP_READLINK",
2967                 .op_rsize_bop = nfsd4_readlink_rsize,
2968         },
2969         [OP_REMOVE] = {
2970                 .op_func = nfsd4_remove,
2971                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2972                 .op_name = "OP_REMOVE",
2973                 .op_rsize_bop = nfsd4_remove_rsize,
2974         },
2975         [OP_RENAME] = {
2976                 .op_func = nfsd4_rename,
2977                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2978                 .op_name = "OP_RENAME",
2979                 .op_rsize_bop = nfsd4_rename_rsize,
2980         },
2981         [OP_RENEW] = {
2982                 .op_func = nfsd4_renew,
2983                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2984                                 | OP_MODIFIES_SOMETHING,
2985                 .op_name = "OP_RENEW",
2986                 .op_rsize_bop = nfsd4_only_status_rsize,
2987
2988         },
2989         [OP_RESTOREFH] = {
2990                 .op_func = nfsd4_restorefh,
2991                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2992                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
2993                 .op_name = "OP_RESTOREFH",
2994                 .op_rsize_bop = nfsd4_only_status_rsize,
2995         },
2996         [OP_SAVEFH] = {
2997                 .op_func = nfsd4_savefh,
2998                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2999                 .op_name = "OP_SAVEFH",
3000                 .op_rsize_bop = nfsd4_only_status_rsize,
3001         },
3002         [OP_SECINFO] = {
3003                 .op_func = nfsd4_secinfo,
3004                 .op_release = nfsd4_secinfo_release,
3005                 .op_flags = OP_HANDLES_WRONGSEC,
3006                 .op_name = "OP_SECINFO",
3007                 .op_rsize_bop = nfsd4_secinfo_rsize,
3008         },
3009         [OP_SETATTR] = {
3010                 .op_func = nfsd4_setattr,
3011                 .op_name = "OP_SETATTR",
3012                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3013                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3014                 .op_rsize_bop = nfsd4_setattr_rsize,
3015                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3016         },
3017         [OP_SETCLIENTID] = {
3018                 .op_func = nfsd4_setclientid,
3019                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3020                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3021                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3022                 .op_name = "OP_SETCLIENTID",
3023                 .op_rsize_bop = nfsd4_setclientid_rsize,
3024         },
3025         [OP_SETCLIENTID_CONFIRM] = {
3026                 .op_func = nfsd4_setclientid_confirm,
3027                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3028                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3029                 .op_name = "OP_SETCLIENTID_CONFIRM",
3030                 .op_rsize_bop = nfsd4_only_status_rsize,
3031         },
3032         [OP_VERIFY] = {
3033                 .op_func = nfsd4_verify,
3034                 .op_name = "OP_VERIFY",
3035                 .op_rsize_bop = nfsd4_only_status_rsize,
3036         },
3037         [OP_WRITE] = {
3038                 .op_func = nfsd4_write,
3039                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3040                 .op_name = "OP_WRITE",
3041                 .op_rsize_bop = nfsd4_write_rsize,
3042                 .op_get_currentstateid = nfsd4_get_writestateid,
3043         },
3044         [OP_RELEASE_LOCKOWNER] = {
3045                 .op_func = nfsd4_release_lockowner,
3046                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3047                                 | OP_MODIFIES_SOMETHING,
3048                 .op_name = "OP_RELEASE_LOCKOWNER",
3049                 .op_rsize_bop = nfsd4_only_status_rsize,
3050         },
3051
3052         /* NFSv4.1 operations */
3053         [OP_EXCHANGE_ID] = {
3054                 .op_func = nfsd4_exchange_id,
3055                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3056                                 | OP_MODIFIES_SOMETHING,
3057                 .op_name = "OP_EXCHANGE_ID",
3058                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3059         },
3060         [OP_BACKCHANNEL_CTL] = {
3061                 .op_func = nfsd4_backchannel_ctl,
3062                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3063                 .op_name = "OP_BACKCHANNEL_CTL",
3064                 .op_rsize_bop = nfsd4_only_status_rsize,
3065         },
3066         [OP_BIND_CONN_TO_SESSION] = {
3067                 .op_func = nfsd4_bind_conn_to_session,
3068                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3069                                 | OP_MODIFIES_SOMETHING,
3070                 .op_name = "OP_BIND_CONN_TO_SESSION",
3071                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3072         },
3073         [OP_CREATE_SESSION] = {
3074                 .op_func = nfsd4_create_session,
3075                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3076                                 | OP_MODIFIES_SOMETHING,
3077                 .op_name = "OP_CREATE_SESSION",
3078                 .op_rsize_bop = nfsd4_create_session_rsize,
3079         },
3080         [OP_DESTROY_SESSION] = {
3081                 .op_func = nfsd4_destroy_session,
3082                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3083                                 | OP_MODIFIES_SOMETHING,
3084                 .op_name = "OP_DESTROY_SESSION",
3085                 .op_rsize_bop = nfsd4_only_status_rsize,
3086         },
3087         [OP_SEQUENCE] = {
3088                 .op_func = nfsd4_sequence,
3089                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3090                 .op_name = "OP_SEQUENCE",
3091                 .op_rsize_bop = nfsd4_sequence_rsize,
3092         },
3093         [OP_DESTROY_CLIENTID] = {
3094                 .op_func = nfsd4_destroy_clientid,
3095                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3096                                 | OP_MODIFIES_SOMETHING,
3097                 .op_name = "OP_DESTROY_CLIENTID",
3098                 .op_rsize_bop = nfsd4_only_status_rsize,
3099         },
3100         [OP_RECLAIM_COMPLETE] = {
3101                 .op_func = nfsd4_reclaim_complete,
3102                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3103                 .op_name = "OP_RECLAIM_COMPLETE",
3104                 .op_rsize_bop = nfsd4_only_status_rsize,
3105         },
3106         [OP_SECINFO_NO_NAME] = {
3107                 .op_func = nfsd4_secinfo_no_name,
3108                 .op_release = nfsd4_secinfo_no_name_release,
3109                 .op_flags = OP_HANDLES_WRONGSEC,
3110                 .op_name = "OP_SECINFO_NO_NAME",
3111                 .op_rsize_bop = nfsd4_secinfo_rsize,
3112         },
3113         [OP_TEST_STATEID] = {
3114                 .op_func = nfsd4_test_stateid,
3115                 .op_flags = ALLOWED_WITHOUT_FH,
3116                 .op_name = "OP_TEST_STATEID",
3117                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3118         },
3119         [OP_FREE_STATEID] = {
3120                 .op_func = nfsd4_free_stateid,
3121                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3122                 .op_name = "OP_FREE_STATEID",
3123                 .op_get_currentstateid = nfsd4_get_freestateid,
3124                 .op_rsize_bop = nfsd4_only_status_rsize,
3125         },
3126 #ifdef CONFIG_NFSD_PNFS
3127         [OP_GETDEVICEINFO] = {
3128                 .op_func = nfsd4_getdeviceinfo,
3129                 .op_release = nfsd4_getdeviceinfo_release,
3130                 .op_flags = ALLOWED_WITHOUT_FH,
3131                 .op_name = "OP_GETDEVICEINFO",
3132                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3133         },
3134         [OP_LAYOUTGET] = {
3135                 .op_func = nfsd4_layoutget,
3136                 .op_release = nfsd4_layoutget_release,
3137                 .op_flags = OP_MODIFIES_SOMETHING,
3138                 .op_name = "OP_LAYOUTGET",
3139                 .op_rsize_bop = nfsd4_layoutget_rsize,
3140         },
3141         [OP_LAYOUTCOMMIT] = {
3142                 .op_func = nfsd4_layoutcommit,
3143                 .op_flags = OP_MODIFIES_SOMETHING,
3144                 .op_name = "OP_LAYOUTCOMMIT",
3145                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3146         },
3147         [OP_LAYOUTRETURN] = {
3148                 .op_func = nfsd4_layoutreturn,
3149                 .op_flags = OP_MODIFIES_SOMETHING,
3150                 .op_name = "OP_LAYOUTRETURN",
3151                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3152         },
3153 #endif /* CONFIG_NFSD_PNFS */
3154
3155         /* NFSv4.2 operations */
3156         [OP_ALLOCATE] = {
3157                 .op_func = nfsd4_allocate,
3158                 .op_flags = OP_MODIFIES_SOMETHING,
3159                 .op_name = "OP_ALLOCATE",
3160                 .op_rsize_bop = nfsd4_only_status_rsize,
3161         },
3162         [OP_DEALLOCATE] = {
3163                 .op_func = nfsd4_deallocate,
3164                 .op_flags = OP_MODIFIES_SOMETHING,
3165                 .op_name = "OP_DEALLOCATE",
3166                 .op_rsize_bop = nfsd4_only_status_rsize,
3167         },
3168         [OP_CLONE] = {
3169                 .op_func = nfsd4_clone,
3170                 .op_flags = OP_MODIFIES_SOMETHING,
3171                 .op_name = "OP_CLONE",
3172                 .op_rsize_bop = nfsd4_only_status_rsize,
3173         },
3174         [OP_COPY] = {
3175                 .op_func = nfsd4_copy,
3176                 .op_flags = OP_MODIFIES_SOMETHING,
3177                 .op_name = "OP_COPY",
3178                 .op_rsize_bop = nfsd4_copy_rsize,
3179         },
3180         [OP_READ_PLUS] = {
3181                 .op_func = nfsd4_read,
3182                 .op_release = nfsd4_read_release,
3183                 .op_name = "OP_READ_PLUS",
3184                 .op_rsize_bop = nfsd4_read_plus_rsize,
3185                 .op_get_currentstateid = nfsd4_get_readstateid,
3186         },
3187         [OP_SEEK] = {
3188                 .op_func = nfsd4_seek,
3189                 .op_name = "OP_SEEK",
3190                 .op_rsize_bop = nfsd4_seek_rsize,
3191         },
3192         [OP_OFFLOAD_STATUS] = {
3193                 .op_func = nfsd4_offload_status,
3194                 .op_name = "OP_OFFLOAD_STATUS",
3195                 .op_rsize_bop = nfsd4_offload_status_rsize,
3196         },
3197         [OP_OFFLOAD_CANCEL] = {
3198                 .op_func = nfsd4_offload_cancel,
3199                 .op_flags = OP_MODIFIES_SOMETHING,
3200                 .op_name = "OP_OFFLOAD_CANCEL",
3201                 .op_rsize_bop = nfsd4_only_status_rsize,
3202         },
3203         [OP_COPY_NOTIFY] = {
3204                 .op_func = nfsd4_copy_notify,
3205                 .op_flags = OP_MODIFIES_SOMETHING,
3206                 .op_name = "OP_COPY_NOTIFY",
3207                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3208         },
3209         [OP_GETXATTR] = {
3210                 .op_func = nfsd4_getxattr,
3211                 .op_name = "OP_GETXATTR",
3212                 .op_rsize_bop = nfsd4_getxattr_rsize,
3213         },
3214         [OP_SETXATTR] = {
3215                 .op_func = nfsd4_setxattr,
3216                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3217                 .op_name = "OP_SETXATTR",
3218                 .op_rsize_bop = nfsd4_setxattr_rsize,
3219         },
3220         [OP_LISTXATTRS] = {
3221                 .op_func = nfsd4_listxattrs,
3222                 .op_name = "OP_LISTXATTRS",
3223                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3224         },
3225         [OP_REMOVEXATTR] = {
3226                 .op_func = nfsd4_removexattr,
3227                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3228                 .op_name = "OP_REMOVEXATTR",
3229                 .op_rsize_bop = nfsd4_removexattr_rsize,
3230         },
3231 };
3232
3233 /**
3234  * nfsd4_spo_must_allow - Determine if the compound op contains an
3235  * operation that is allowed to be sent with machine credentials
3236  *
3237  * @rqstp: a pointer to the struct svc_rqst
3238  *
3239  * Checks to see if the compound contains a spo_must_allow op
3240  * and confirms that it was sent with the proper machine creds.
3241  */
3242
3243 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3244 {
3245         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3246         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3247         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
3248         struct nfsd4_compound_state *cstate = &resp->cstate;
3249         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3250         u32 opiter;
3251
3252         if (!cstate->minorversion)
3253                 return false;
3254
3255         if (cstate->spo_must_allowed)
3256                 return true;
3257
3258         opiter = resp->opcnt;
3259         while (opiter < argp->opcnt) {
3260                 this = &argp->ops[opiter++];
3261                 if (test_bit(this->opnum, allow->u.longs) &&
3262                         cstate->clp->cl_mach_cred &&
3263                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3264                         cstate->spo_must_allowed = true;
3265                         return true;
3266                 }
3267         }
3268         cstate->spo_must_allowed = false;
3269         return false;
3270 }
3271
3272 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3273 {
3274         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3275                 return op_encode_hdr_size * sizeof(__be32);
3276
3277         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3278         return OPDESC(op)->op_rsize_bop(rqstp, op);
3279 }
3280
3281 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3282 {
3283         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3284                 pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
3285                         op->opnum, nfsd4_op_name(op->opnum));
3286                 WARN_ON_ONCE(1);
3287         }
3288 }
3289
3290 static const char *nfsd4_op_name(unsigned opnum)
3291 {
3292         if (opnum < ARRAY_SIZE(nfsd4_ops))
3293                 return nfsd4_ops[opnum].op_name;
3294         return "unknown_operation";
3295 }
3296
3297 #define nfsd4_voidres                   nfsd4_voidargs
3298 struct nfsd4_voidargs { int dummy; };
3299
3300 static const struct svc_procedure nfsd_procedures4[2] = {
3301         [NFSPROC4_NULL] = {
3302                 .pc_func = nfsd4_proc_null,
3303                 .pc_decode = nfs4svc_decode_voidarg,
3304                 .pc_encode = nfs4svc_encode_voidres,
3305                 .pc_argsize = sizeof(struct nfsd4_voidargs),
3306                 .pc_ressize = sizeof(struct nfsd4_voidres),
3307                 .pc_cachetype = RC_NOCACHE,
3308                 .pc_xdrressize = 1,
3309         },
3310         [NFSPROC4_COMPOUND] = {
3311                 .pc_func = nfsd4_proc_compound,
3312                 .pc_decode = nfs4svc_decode_compoundargs,
3313                 .pc_encode = nfs4svc_encode_compoundres,
3314                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3315                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3316                 .pc_release = nfsd4_release_compoundargs,
3317                 .pc_cachetype = RC_NOCACHE,
3318                 .pc_xdrressize = NFSD_BUFSIZE/4,
3319         },
3320 };
3321
3322 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3323 const struct svc_version nfsd_version4 = {
3324         .vs_vers                = 4,
3325         .vs_nproc               = 2,
3326         .vs_proc                = nfsd_procedures4,
3327         .vs_count               = nfsd_count3,
3328         .vs_dispatch            = nfsd_dispatch,
3329         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3330         .vs_rpcb_optnl          = true,
3331         .vs_need_cong_ctrl      = true,
3332 };
3333
3334 /*
3335  * Local variables:
3336  *  c-basic-offset: 8
3337  * End:
3338  */