Loading...
1/*
2 * Server-side procedures for NFSv4.
3 *
4 * Copyright (c) 2002 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <andros@umich.edu>
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */
35#include <linux/fs_struct.h>
36#include <linux/file.h>
37#include <linux/falloc.h>
38#include <linux/slab.h>
39#include <linux/kthread.h>
40
41#include "idmap.h"
42#include "cache.h"
43#include "xdr4.h"
44#include "vfs.h"
45#include "current_stateid.h"
46#include "netns.h"
47#include "acl.h"
48#include "pnfs.h"
49#include "trace.h"
50
51#ifdef CONFIG_NFSD_V4_SECURITY_LABEL
52#include <linux/security.h>
53
54static inline void
55nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
56{
57 struct inode *inode = d_inode(resfh->fh_dentry);
58 int status;
59
60 inode_lock(inode);
61 status = security_inode_setsecctx(resfh->fh_dentry,
62 label->data, label->len);
63 inode_unlock(inode);
64
65 if (status)
66 /*
67 * XXX: We should really fail the whole open, but we may
68 * already have created a new file, so it may be too
69 * late. For now this seems the least of evils:
70 */
71 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
72
73 return;
74}
75#else
76static inline void
77nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
78{ }
79#endif
80
81#define NFSDDBG_FACILITY NFSDDBG_PROC
82
83static u32 nfsd_attrmask[] = {
84 NFSD_WRITEABLE_ATTRS_WORD0,
85 NFSD_WRITEABLE_ATTRS_WORD1,
86 NFSD_WRITEABLE_ATTRS_WORD2
87};
88
89static u32 nfsd41_ex_attrmask[] = {
90 NFSD_SUPPATTR_EXCLCREAT_WORD0,
91 NFSD_SUPPATTR_EXCLCREAT_WORD1,
92 NFSD_SUPPATTR_EXCLCREAT_WORD2
93};
94
95static __be32
96check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
97 u32 *bmval, u32 *writable)
98{
99 struct dentry *dentry = cstate->current_fh.fh_dentry;
100 struct svc_export *exp = cstate->current_fh.fh_export;
101
102 if (!nfsd_attrs_supported(cstate->minorversion, bmval))
103 return nfserr_attrnotsupp;
104 if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
105 return nfserr_attrnotsupp;
106 if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
107 !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
108 return nfserr_attrnotsupp;
109 if (writable && !bmval_is_subset(bmval, writable))
110 return nfserr_inval;
111 if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
112 (bmval[1] & FATTR4_WORD1_MODE))
113 return nfserr_inval;
114 return nfs_ok;
115}
116
117static __be32
118nfsd4_check_open_attributes(struct svc_rqst *rqstp,
119 struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
120{
121 __be32 status = nfs_ok;
122
123 if (open->op_create == NFS4_OPEN_CREATE) {
124 if (open->op_createmode == NFS4_CREATE_UNCHECKED
125 || open->op_createmode == NFS4_CREATE_GUARDED)
126 status = check_attr_support(rqstp, cstate,
127 open->op_bmval, nfsd_attrmask);
128 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
129 status = check_attr_support(rqstp, cstate,
130 open->op_bmval, nfsd41_ex_attrmask);
131 }
132
133 return status;
134}
135
136static int
137is_create_with_attrs(struct nfsd4_open *open)
138{
139 return open->op_create == NFS4_OPEN_CREATE
140 && (open->op_createmode == NFS4_CREATE_UNCHECKED
141 || open->op_createmode == NFS4_CREATE_GUARDED
142 || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
143}
144
145/*
146 * if error occurs when setting the acl, just clear the acl bit
147 * in the returned attr bitmap.
148 */
149static void
150do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
151 struct nfs4_acl *acl, u32 *bmval)
152{
153 __be32 status;
154
155 status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
156 if (status)
157 /*
158 * We should probably fail the whole open at this point,
159 * but we've already created the file, so it's too late;
160 * So this seems the least of evils:
161 */
162 bmval[0] &= ~FATTR4_WORD0_ACL;
163}
164
165static inline void
166fh_dup2(struct svc_fh *dst, struct svc_fh *src)
167{
168 fh_put(dst);
169 dget(src->fh_dentry);
170 if (src->fh_export)
171 exp_get(src->fh_export);
172 *dst = *src;
173}
174
175static __be32
176do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
177{
178 __be32 status;
179
180 if (open->op_truncate &&
181 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
182 return nfserr_inval;
183
184 accmode |= NFSD_MAY_READ_IF_EXEC;
185
186 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
187 accmode |= NFSD_MAY_READ;
188 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
189 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
190 if (open->op_share_deny & NFS4_SHARE_DENY_READ)
191 accmode |= NFSD_MAY_WRITE;
192
193 status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
194
195 return status;
196}
197
198static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
199{
200 umode_t mode = d_inode(fh->fh_dentry)->i_mode;
201
202 if (S_ISREG(mode))
203 return nfs_ok;
204 if (S_ISDIR(mode))
205 return nfserr_isdir;
206 /*
207 * Using err_symlink as our catch-all case may look odd; but
208 * there's no other obvious error for this case in 4.0, and we
209 * happen to know that it will cause the linux v4 client to do
210 * the right thing on attempts to open something other than a
211 * regular file.
212 */
213 return nfserr_symlink;
214}
215
216static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
217{
218 if (nfsd4_has_session(cstate))
219 return;
220 fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
221 &resfh->fh_handle);
222}
223
224static __be32
225do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
226{
227 struct svc_fh *current_fh = &cstate->current_fh;
228 int accmode;
229 __be32 status;
230
231 *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
232 if (!*resfh)
233 return nfserr_jukebox;
234 fh_init(*resfh, NFS4_FHSIZE);
235 open->op_truncate = 0;
236
237 if (open->op_create) {
238 /* FIXME: check session persistence and pnfs flags.
239 * The nfsv4.1 spec requires the following semantics:
240 *
241 * Persistent | pNFS | Server REQUIRED | Client Allowed
242 * Reply Cache | server | |
243 * -------------+--------+-----------------+--------------------
244 * no | no | EXCLUSIVE4_1 | EXCLUSIVE4_1
245 * | | | (SHOULD)
246 * | | and EXCLUSIVE4 | or EXCLUSIVE4
247 * | | | (SHOULD NOT)
248 * no | yes | EXCLUSIVE4_1 | EXCLUSIVE4_1
249 * yes | no | GUARDED4 | GUARDED4
250 * yes | yes | GUARDED4 | GUARDED4
251 */
252
253 /*
254 * Note: create modes (UNCHECKED,GUARDED...) are the same
255 * in NFSv4 as in v3 except EXCLUSIVE4_1.
256 */
257 current->fs->umask = open->op_umask;
258 status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
259 open->op_fname.len, &open->op_iattr,
260 *resfh, open->op_createmode,
261 (u32 *)open->op_verf.data,
262 &open->op_truncate, &open->op_created);
263 current->fs->umask = 0;
264
265 if (!status && open->op_label.len)
266 nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
267
268 /*
269 * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
270 * use the returned bitmask to indicate which attributes
271 * we used to store the verifier:
272 */
273 if (nfsd_create_is_exclusive(open->op_createmode) && status == 0)
274 open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
275 FATTR4_WORD1_TIME_MODIFY);
276 } else
277 /*
278 * Note this may exit with the parent still locked.
279 * We will hold the lock until nfsd4_open's final
280 * lookup, to prevent renames or unlinks until we've had
281 * a chance to an acquire a delegation if appropriate.
282 */
283 status = nfsd_lookup(rqstp, current_fh,
284 open->op_fname.data, open->op_fname.len, *resfh);
285 if (status)
286 goto out;
287 status = nfsd_check_obj_isreg(*resfh);
288 if (status)
289 goto out;
290
291 if (is_create_with_attrs(open) && open->op_acl != NULL)
292 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
293
294 nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
295 accmode = NFSD_MAY_NOP;
296 if (open->op_created ||
297 open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
298 accmode |= NFSD_MAY_OWNER_OVERRIDE;
299 status = do_open_permission(rqstp, *resfh, open, accmode);
300 set_change_info(&open->op_cinfo, current_fh);
301out:
302 return status;
303}
304
305static __be32
306do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
307{
308 struct svc_fh *current_fh = &cstate->current_fh;
309 __be32 status;
310 int accmode = 0;
311
312 /* We don't know the target directory, and therefore can not
313 * set the change info
314 */
315
316 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
317
318 nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
319
320 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
321 (open->op_iattr.ia_size == 0);
322 /*
323 * In the delegation case, the client is telling us about an
324 * open that it *already* performed locally, some time ago. We
325 * should let it succeed now if possible.
326 *
327 * In the case of a CLAIM_FH open, on the other hand, the client
328 * may be counting on us to enforce permissions (the Linux 4.1
329 * client uses this for normal opens, for example).
330 */
331 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
332 accmode = NFSD_MAY_OWNER_OVERRIDE;
333
334 status = do_open_permission(rqstp, current_fh, open, accmode);
335
336 return status;
337}
338
339static void
340copy_clientid(clientid_t *clid, struct nfsd4_session *session)
341{
342 struct nfsd4_sessionid *sid =
343 (struct nfsd4_sessionid *)session->se_sessionid.data;
344
345 clid->cl_boot = sid->clientid.cl_boot;
346 clid->cl_id = sid->clientid.cl_id;
347}
348
349static __be32
350nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
351 union nfsd4_op_u *u)
352{
353 struct nfsd4_open *open = &u->open;
354 __be32 status;
355 struct svc_fh *resfh = NULL;
356 struct net *net = SVC_NET(rqstp);
357 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
358 bool reclaim = false;
359
360 dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
361 (int)open->op_fname.len, open->op_fname.data,
362 open->op_openowner);
363
364 /* This check required by spec. */
365 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
366 return nfserr_inval;
367
368 open->op_created = 0;
369 /*
370 * RFC5661 18.51.3
371 * Before RECLAIM_COMPLETE done, server should deny new lock
372 */
373 if (nfsd4_has_session(cstate) &&
374 !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
375 &cstate->session->se_client->cl_flags) &&
376 open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
377 return nfserr_grace;
378
379 if (nfsd4_has_session(cstate))
380 copy_clientid(&open->op_clientid, cstate->session);
381
382 /* check seqid for replay. set nfs4_owner */
383 status = nfsd4_process_open1(cstate, open, nn);
384 if (status == nfserr_replay_me) {
385 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
386 fh_put(&cstate->current_fh);
387 fh_copy_shallow(&cstate->current_fh.fh_handle,
388 &rp->rp_openfh);
389 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
390 if (status)
391 dprintk("nfsd4_open: replay failed"
392 " restoring previous filehandle\n");
393 else
394 status = nfserr_replay_me;
395 }
396 if (status)
397 goto out;
398 if (open->op_xdr_error) {
399 status = open->op_xdr_error;
400 goto out;
401 }
402
403 status = nfsd4_check_open_attributes(rqstp, cstate, open);
404 if (status)
405 goto out;
406
407 /* Openowner is now set, so sequence id will get bumped. Now we need
408 * these checks before we do any creates: */
409 status = nfserr_grace;
410 if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
411 goto out;
412 status = nfserr_no_grace;
413 if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
414 goto out;
415
416 switch (open->op_claim_type) {
417 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
418 case NFS4_OPEN_CLAIM_NULL:
419 status = do_open_lookup(rqstp, cstate, open, &resfh);
420 if (status)
421 goto out;
422 break;
423 case NFS4_OPEN_CLAIM_PREVIOUS:
424 status = nfs4_check_open_reclaim(&open->op_clientid,
425 cstate, nn);
426 if (status)
427 goto out;
428 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
429 reclaim = true;
430 /* fall through */
431 case NFS4_OPEN_CLAIM_FH:
432 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
433 status = do_open_fhandle(rqstp, cstate, open);
434 if (status)
435 goto out;
436 resfh = &cstate->current_fh;
437 break;
438 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
439 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
440 dprintk("NFSD: unsupported OPEN claim type %d\n",
441 open->op_claim_type);
442 status = nfserr_notsupp;
443 goto out;
444 default:
445 dprintk("NFSD: Invalid OPEN claim type %d\n",
446 open->op_claim_type);
447 status = nfserr_inval;
448 goto out;
449 }
450 /*
451 * nfsd4_process_open2() does the actual opening of the file. If
452 * successful, it (1) truncates the file if open->op_truncate was
453 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
454 */
455 status = nfsd4_process_open2(rqstp, resfh, open);
456 WARN(status && open->op_created,
457 "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
458 be32_to_cpu(status));
459 if (reclaim && !status)
460 nn->somebody_reclaimed = true;
461out:
462 if (resfh && resfh != &cstate->current_fh) {
463 fh_dup2(&cstate->current_fh, resfh);
464 fh_put(resfh);
465 kfree(resfh);
466 }
467 nfsd4_cleanup_open_state(cstate, open);
468 nfsd4_bump_seqid(cstate, status);
469 return status;
470}
471
472/*
473 * OPEN is the only seqid-mutating operation whose decoding can fail
474 * with a seqid-mutating error (specifically, decoding of user names in
475 * the attributes). Therefore we have to do some processing to look up
476 * the stateowner so that we can bump the seqid.
477 */
478static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
479{
480 struct nfsd4_open *open = &op->u.open;
481
482 if (!seqid_mutating_err(ntohl(op->status)))
483 return op->status;
484 if (nfsd4_has_session(cstate))
485 return op->status;
486 open->op_xdr_error = op->status;
487 return nfsd4_open(rqstp, cstate, &op->u);
488}
489
490/*
491 * filehandle-manipulating ops.
492 */
493static __be32
494nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
495 union nfsd4_op_u *u)
496{
497 u->getfh = &cstate->current_fh;
498 return nfs_ok;
499}
500
501static __be32
502nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
503 union nfsd4_op_u *u)
504{
505 struct nfsd4_putfh *putfh = &u->putfh;
506
507 fh_put(&cstate->current_fh);
508 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
509 memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
510 putfh->pf_fhlen);
511 return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
512}
513
514static __be32
515nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
516 union nfsd4_op_u *u)
517{
518 __be32 status;
519
520 fh_put(&cstate->current_fh);
521 status = exp_pseudoroot(rqstp, &cstate->current_fh);
522 return status;
523}
524
525static __be32
526nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
527 union nfsd4_op_u *u)
528{
529 if (!cstate->save_fh.fh_dentry)
530 return nfserr_restorefh;
531
532 fh_dup2(&cstate->current_fh, &cstate->save_fh);
533 if (HAS_STATE_ID(cstate, SAVED_STATE_ID_FLAG)) {
534 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
535 SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
536 }
537 return nfs_ok;
538}
539
540static __be32
541nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
542 union nfsd4_op_u *u)
543{
544 fh_dup2(&cstate->save_fh, &cstate->current_fh);
545 if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG)) {
546 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
547 SET_STATE_ID(cstate, SAVED_STATE_ID_FLAG);
548 }
549 return nfs_ok;
550}
551
552/*
553 * misc nfsv4 ops
554 */
555static __be32
556nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
557 union nfsd4_op_u *u)
558{
559 struct nfsd4_access *access = &u->access;
560
561 if (access->ac_req_access & ~NFS3_ACCESS_FULL)
562 return nfserr_inval;
563
564 access->ac_resp_access = access->ac_req_access;
565 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
566 &access->ac_supported);
567}
568
569static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
570{
571 __be32 *verf = (__be32 *)verifier->data;
572
573 BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
574
575 nfsd_copy_boot_verifier(verf, net_generic(net, nfsd_net_id));
576}
577
578static __be32
579nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
580 union nfsd4_op_u *u)
581{
582 struct nfsd4_commit *commit = &u->commit;
583
584 gen_boot_verifier(&commit->co_verf, SVC_NET(rqstp));
585 return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
586 commit->co_count);
587}
588
589static __be32
590nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
591 union nfsd4_op_u *u)
592{
593 struct nfsd4_create *create = &u->create;
594 struct svc_fh resfh;
595 __be32 status;
596 dev_t rdev;
597
598 fh_init(&resfh, NFS4_FHSIZE);
599
600 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
601 if (status)
602 return status;
603
604 status = check_attr_support(rqstp, cstate, create->cr_bmval,
605 nfsd_attrmask);
606 if (status)
607 return status;
608
609 current->fs->umask = create->cr_umask;
610 switch (create->cr_type) {
611 case NF4LNK:
612 status = nfsd_symlink(rqstp, &cstate->current_fh,
613 create->cr_name, create->cr_namelen,
614 create->cr_data, &resfh);
615 break;
616
617 case NF4BLK:
618 status = nfserr_inval;
619 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
620 if (MAJOR(rdev) != create->cr_specdata1 ||
621 MINOR(rdev) != create->cr_specdata2)
622 goto out_umask;
623 status = nfsd_create(rqstp, &cstate->current_fh,
624 create->cr_name, create->cr_namelen,
625 &create->cr_iattr, S_IFBLK, rdev, &resfh);
626 break;
627
628 case NF4CHR:
629 status = nfserr_inval;
630 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
631 if (MAJOR(rdev) != create->cr_specdata1 ||
632 MINOR(rdev) != create->cr_specdata2)
633 goto out_umask;
634 status = nfsd_create(rqstp, &cstate->current_fh,
635 create->cr_name, create->cr_namelen,
636 &create->cr_iattr,S_IFCHR, rdev, &resfh);
637 break;
638
639 case NF4SOCK:
640 status = nfsd_create(rqstp, &cstate->current_fh,
641 create->cr_name, create->cr_namelen,
642 &create->cr_iattr, S_IFSOCK, 0, &resfh);
643 break;
644
645 case NF4FIFO:
646 status = nfsd_create(rqstp, &cstate->current_fh,
647 create->cr_name, create->cr_namelen,
648 &create->cr_iattr, S_IFIFO, 0, &resfh);
649 break;
650
651 case NF4DIR:
652 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
653 status = nfsd_create(rqstp, &cstate->current_fh,
654 create->cr_name, create->cr_namelen,
655 &create->cr_iattr, S_IFDIR, 0, &resfh);
656 break;
657
658 default:
659 status = nfserr_badtype;
660 }
661
662 if (status)
663 goto out;
664
665 if (create->cr_label.len)
666 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
667
668 if (create->cr_acl != NULL)
669 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
670 create->cr_bmval);
671
672 fh_unlock(&cstate->current_fh);
673 set_change_info(&create->cr_cinfo, &cstate->current_fh);
674 fh_dup2(&cstate->current_fh, &resfh);
675out:
676 fh_put(&resfh);
677out_umask:
678 current->fs->umask = 0;
679 return status;
680}
681
682static __be32
683nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
684 union nfsd4_op_u *u)
685{
686 struct nfsd4_getattr *getattr = &u->getattr;
687 __be32 status;
688
689 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
690 if (status)
691 return status;
692
693 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
694 return nfserr_inval;
695
696 getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
697 getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
698 getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
699
700 getattr->ga_fhp = &cstate->current_fh;
701 return nfs_ok;
702}
703
704static __be32
705nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
706 union nfsd4_op_u *u)
707{
708 struct nfsd4_link *link = &u->link;
709 __be32 status;
710
711 status = nfsd_link(rqstp, &cstate->current_fh,
712 link->li_name, link->li_namelen, &cstate->save_fh);
713 if (!status)
714 set_change_info(&link->li_cinfo, &cstate->current_fh);
715 return status;
716}
717
718static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
719{
720 struct svc_fh tmp_fh;
721 __be32 ret;
722
723 fh_init(&tmp_fh, NFS4_FHSIZE);
724 ret = exp_pseudoroot(rqstp, &tmp_fh);
725 if (ret)
726 return ret;
727 if (tmp_fh.fh_dentry == fh->fh_dentry) {
728 fh_put(&tmp_fh);
729 return nfserr_noent;
730 }
731 fh_put(&tmp_fh);
732 return nfsd_lookup(rqstp, fh, "..", 2, fh);
733}
734
735static __be32
736nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
737 union nfsd4_op_u *u)
738{
739 return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
740}
741
742static __be32
743nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
744 union nfsd4_op_u *u)
745{
746 return nfsd_lookup(rqstp, &cstate->current_fh,
747 u->lookup.lo_name, u->lookup.lo_len,
748 &cstate->current_fh);
749}
750
751static __be32
752nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
753 union nfsd4_op_u *u)
754{
755 struct nfsd4_read *read = &u->read;
756 __be32 status;
757
758 read->rd_nf = NULL;
759 if (read->rd_offset >= OFFSET_MAX)
760 return nfserr_inval;
761
762 trace_nfsd_read_start(rqstp, &cstate->current_fh,
763 read->rd_offset, read->rd_length);
764
765 /*
766 * If we do a zero copy read, then a client will see read data
767 * that reflects the state of the file *after* performing the
768 * following compound.
769 *
770 * To ensure proper ordering, we therefore turn off zero copy if
771 * the client wants us to do more in this compound:
772 */
773 if (!nfsd4_last_compound_op(rqstp))
774 clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
775
776 /* check stateid */
777 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
778 &read->rd_stateid, RD_STATE,
779 &read->rd_nf);
780 if (status) {
781 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
782 goto out;
783 }
784 status = nfs_ok;
785out:
786 read->rd_rqstp = rqstp;
787 read->rd_fhp = &cstate->current_fh;
788 return status;
789}
790
791
792static void
793nfsd4_read_release(union nfsd4_op_u *u)
794{
795 if (u->read.rd_nf)
796 nfsd_file_put(u->read.rd_nf);
797 trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
798 u->read.rd_offset, u->read.rd_length);
799}
800
801static __be32
802nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
803 union nfsd4_op_u *u)
804{
805 struct nfsd4_readdir *readdir = &u->readdir;
806 u64 cookie = readdir->rd_cookie;
807 static const nfs4_verifier zeroverf;
808
809 /* no need to check permission - this will be done in nfsd_readdir() */
810
811 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
812 return nfserr_inval;
813
814 readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
815 readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
816 readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
817
818 if ((cookie == 1) || (cookie == 2) ||
819 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
820 return nfserr_bad_cookie;
821
822 readdir->rd_rqstp = rqstp;
823 readdir->rd_fhp = &cstate->current_fh;
824 return nfs_ok;
825}
826
827static __be32
828nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
829 union nfsd4_op_u *u)
830{
831 u->readlink.rl_rqstp = rqstp;
832 u->readlink.rl_fhp = &cstate->current_fh;
833 return nfs_ok;
834}
835
836static __be32
837nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
838 union nfsd4_op_u *u)
839{
840 struct nfsd4_remove *remove = &u->remove;
841 __be32 status;
842
843 if (opens_in_grace(SVC_NET(rqstp)))
844 return nfserr_grace;
845 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
846 remove->rm_name, remove->rm_namelen);
847 if (!status) {
848 fh_unlock(&cstate->current_fh);
849 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
850 }
851 return status;
852}
853
854static __be32
855nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
856 union nfsd4_op_u *u)
857{
858 struct nfsd4_rename *rename = &u->rename;
859 __be32 status;
860
861 if (opens_in_grace(SVC_NET(rqstp)))
862 return nfserr_grace;
863 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
864 rename->rn_snamelen, &cstate->current_fh,
865 rename->rn_tname, rename->rn_tnamelen);
866 if (status)
867 return status;
868 set_change_info(&rename->rn_sinfo, &cstate->current_fh);
869 set_change_info(&rename->rn_tinfo, &cstate->save_fh);
870 return nfs_ok;
871}
872
873static __be32
874nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
875 union nfsd4_op_u *u)
876{
877 struct nfsd4_secinfo *secinfo = &u->secinfo;
878 struct svc_export *exp;
879 struct dentry *dentry;
880 __be32 err;
881
882 err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
883 if (err)
884 return err;
885 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
886 secinfo->si_name, secinfo->si_namelen,
887 &exp, &dentry);
888 if (err)
889 return err;
890 fh_unlock(&cstate->current_fh);
891 if (d_really_is_negative(dentry)) {
892 exp_put(exp);
893 err = nfserr_noent;
894 } else
895 secinfo->si_exp = exp;
896 dput(dentry);
897 if (cstate->minorversion)
898 /* See rfc 5661 section 2.6.3.1.1.8 */
899 fh_put(&cstate->current_fh);
900 return err;
901}
902
903static __be32
904nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
905 union nfsd4_op_u *u)
906{
907 __be32 err;
908
909 switch (u->secinfo_no_name.sin_style) {
910 case NFS4_SECINFO_STYLE4_CURRENT_FH:
911 break;
912 case NFS4_SECINFO_STYLE4_PARENT:
913 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
914 if (err)
915 return err;
916 break;
917 default:
918 return nfserr_inval;
919 }
920
921 u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
922 fh_put(&cstate->current_fh);
923 return nfs_ok;
924}
925
926static void
927nfsd4_secinfo_release(union nfsd4_op_u *u)
928{
929 if (u->secinfo.si_exp)
930 exp_put(u->secinfo.si_exp);
931}
932
933static void
934nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
935{
936 if (u->secinfo_no_name.sin_exp)
937 exp_put(u->secinfo_no_name.sin_exp);
938}
939
940static __be32
941nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
942 union nfsd4_op_u *u)
943{
944 struct nfsd4_setattr *setattr = &u->setattr;
945 __be32 status = nfs_ok;
946 int err;
947
948 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
949 status = nfs4_preprocess_stateid_op(rqstp, cstate,
950 &cstate->current_fh, &setattr->sa_stateid,
951 WR_STATE, NULL);
952 if (status) {
953 dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
954 return status;
955 }
956 }
957 err = fh_want_write(&cstate->current_fh);
958 if (err)
959 return nfserrno(err);
960 status = nfs_ok;
961
962 status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
963 nfsd_attrmask);
964 if (status)
965 goto out;
966
967 if (setattr->sa_acl != NULL)
968 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
969 setattr->sa_acl);
970 if (status)
971 goto out;
972 if (setattr->sa_label.len)
973 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
974 &setattr->sa_label);
975 if (status)
976 goto out;
977 status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
978 0, (time_t)0);
979out:
980 fh_drop_write(&cstate->current_fh);
981 return status;
982}
983
984static __be32
985nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
986 union nfsd4_op_u *u)
987{
988 struct nfsd4_write *write = &u->write;
989 stateid_t *stateid = &write->wr_stateid;
990 struct nfsd_file *nf = NULL;
991 __be32 status = nfs_ok;
992 unsigned long cnt;
993 int nvecs;
994
995 if (write->wr_offset >= OFFSET_MAX)
996 return nfserr_inval;
997
998 cnt = write->wr_buflen;
999 trace_nfsd_write_start(rqstp, &cstate->current_fh,
1000 write->wr_offset, cnt);
1001 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1002 stateid, WR_STATE, &nf);
1003 if (status) {
1004 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1005 return status;
1006 }
1007
1008 write->wr_how_written = write->wr_stable_how;
1009 gen_boot_verifier(&write->wr_verifier, SVC_NET(rqstp));
1010
1011 nvecs = svc_fill_write_vector(rqstp, write->wr_pagelist,
1012 &write->wr_head, write->wr_buflen);
1013 WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1014
1015 status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf->nf_file,
1016 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1017 write->wr_how_written);
1018 nfsd_file_put(nf);
1019
1020 write->wr_bytes_written = cnt;
1021 trace_nfsd_write_done(rqstp, &cstate->current_fh,
1022 write->wr_offset, cnt);
1023 return status;
1024}
1025
1026static __be32
1027nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1028 stateid_t *src_stateid, struct nfsd_file **src,
1029 stateid_t *dst_stateid, struct nfsd_file **dst)
1030{
1031 __be32 status;
1032
1033 if (!cstate->save_fh.fh_dentry)
1034 return nfserr_nofilehandle;
1035
1036 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1037 src_stateid, RD_STATE, src);
1038 if (status) {
1039 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1040 goto out;
1041 }
1042
1043 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1044 dst_stateid, WR_STATE, dst);
1045 if (status) {
1046 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1047 goto out_put_src;
1048 }
1049
1050 /* fix up for NFS-specific error code */
1051 if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1052 !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1053 status = nfserr_wrong_type;
1054 goto out_put_dst;
1055 }
1056
1057out:
1058 return status;
1059out_put_dst:
1060 nfsd_file_put(*dst);
1061out_put_src:
1062 nfsd_file_put(*src);
1063 goto out;
1064}
1065
1066static __be32
1067nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1068 union nfsd4_op_u *u)
1069{
1070 struct nfsd4_clone *clone = &u->clone;
1071 struct nfsd_file *src, *dst;
1072 __be32 status;
1073
1074 status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1075 &clone->cl_dst_stateid, &dst);
1076 if (status)
1077 goto out;
1078
1079 status = nfsd4_clone_file_range(src->nf_file, clone->cl_src_pos,
1080 dst->nf_file, clone->cl_dst_pos, clone->cl_count);
1081
1082 nfsd_file_put(dst);
1083 nfsd_file_put(src);
1084out:
1085 return status;
1086}
1087
1088void nfs4_put_copy(struct nfsd4_copy *copy)
1089{
1090 if (!refcount_dec_and_test(©->refcount))
1091 return;
1092 kfree(copy);
1093}
1094
1095static bool
1096check_and_set_stop_copy(struct nfsd4_copy *copy)
1097{
1098 bool value;
1099
1100 spin_lock(©->cp_clp->async_lock);
1101 value = copy->stopped;
1102 if (!copy->stopped)
1103 copy->stopped = true;
1104 spin_unlock(©->cp_clp->async_lock);
1105 return value;
1106}
1107
1108static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1109{
1110 /* only 1 thread should stop the copy */
1111 if (!check_and_set_stop_copy(copy))
1112 kthread_stop(copy->copy_task);
1113 nfs4_put_copy(copy);
1114}
1115
1116static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1117{
1118 struct nfsd4_copy *copy = NULL;
1119
1120 spin_lock(&clp->async_lock);
1121 if (!list_empty(&clp->async_copies)) {
1122 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1123 copies);
1124 refcount_inc(©->refcount);
1125 }
1126 spin_unlock(&clp->async_lock);
1127 return copy;
1128}
1129
1130void nfsd4_shutdown_copy(struct nfs4_client *clp)
1131{
1132 struct nfsd4_copy *copy;
1133
1134 while ((copy = nfsd4_get_copy(clp)) != NULL)
1135 nfsd4_stop_copy(copy);
1136}
1137
1138static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1139{
1140 struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1141
1142 nfs4_put_copy(copy);
1143}
1144
1145static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1146 struct rpc_task *task)
1147{
1148 return 1;
1149}
1150
1151static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1152 .release = nfsd4_cb_offload_release,
1153 .done = nfsd4_cb_offload_done
1154};
1155
1156static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1157{
1158 copy->cp_res.wr_stable_how = NFS_UNSTABLE;
1159 copy->cp_synchronous = sync;
1160 gen_boot_verifier(©->cp_res.wr_verifier, copy->cp_clp->net);
1161}
1162
1163static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1164{
1165 ssize_t bytes_copied = 0;
1166 size_t bytes_total = copy->cp_count;
1167 u64 src_pos = copy->cp_src_pos;
1168 u64 dst_pos = copy->cp_dst_pos;
1169
1170 do {
1171 if (kthread_should_stop())
1172 break;
1173 bytes_copied = nfsd_copy_file_range(copy->nf_src->nf_file,
1174 src_pos, copy->nf_dst->nf_file, dst_pos,
1175 bytes_total);
1176 if (bytes_copied <= 0)
1177 break;
1178 bytes_total -= bytes_copied;
1179 copy->cp_res.wr_bytes_written += bytes_copied;
1180 src_pos += bytes_copied;
1181 dst_pos += bytes_copied;
1182 } while (bytes_total > 0 && !copy->cp_synchronous);
1183 return bytes_copied;
1184}
1185
1186static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1187{
1188 __be32 status;
1189 ssize_t bytes;
1190
1191 bytes = _nfsd_copy_file_range(copy);
1192 /* for async copy, we ignore the error, client can always retry
1193 * to get the error
1194 */
1195 if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1196 status = nfserrno(bytes);
1197 else {
1198 nfsd4_init_copy_res(copy, sync);
1199 status = nfs_ok;
1200 }
1201
1202 nfsd_file_put(copy->nf_src);
1203 nfsd_file_put(copy->nf_dst);
1204 return status;
1205}
1206
1207static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1208{
1209 dst->cp_src_pos = src->cp_src_pos;
1210 dst->cp_dst_pos = src->cp_dst_pos;
1211 dst->cp_count = src->cp_count;
1212 dst->cp_synchronous = src->cp_synchronous;
1213 memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1214 memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1215 dst->cp_clp = src->cp_clp;
1216 dst->nf_dst = nfsd_file_get(src->nf_dst);
1217 dst->nf_src = nfsd_file_get(src->nf_src);
1218 memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1219}
1220
1221static void cleanup_async_copy(struct nfsd4_copy *copy)
1222{
1223 nfs4_free_cp_state(copy);
1224 nfsd_file_put(copy->nf_dst);
1225 nfsd_file_put(copy->nf_src);
1226 spin_lock(©->cp_clp->async_lock);
1227 list_del(©->copies);
1228 spin_unlock(©->cp_clp->async_lock);
1229 nfs4_put_copy(copy);
1230}
1231
1232static int nfsd4_do_async_copy(void *data)
1233{
1234 struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1235 struct nfsd4_copy *cb_copy;
1236
1237 copy->nfserr = nfsd4_do_copy(copy, 0);
1238 cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1239 if (!cb_copy)
1240 goto out;
1241 memcpy(&cb_copy->cp_res, ©->cp_res, sizeof(copy->cp_res));
1242 cb_copy->cp_clp = copy->cp_clp;
1243 cb_copy->nfserr = copy->nfserr;
1244 memcpy(&cb_copy->fh, ©->fh, sizeof(copy->fh));
1245 nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1246 &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1247 nfsd4_run_cb(&cb_copy->cp_cb);
1248out:
1249 cleanup_async_copy(copy);
1250 return 0;
1251}
1252
1253static __be32
1254nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1255 union nfsd4_op_u *u)
1256{
1257 struct nfsd4_copy *copy = &u->copy;
1258 __be32 status;
1259 struct nfsd4_copy *async_copy = NULL;
1260
1261 status = nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid,
1262 ©->nf_src, ©->cp_dst_stateid,
1263 ©->nf_dst);
1264 if (status)
1265 goto out;
1266
1267 copy->cp_clp = cstate->clp;
1268 memcpy(©->fh, &cstate->current_fh.fh_handle,
1269 sizeof(struct knfsd_fh));
1270 if (!copy->cp_synchronous) {
1271 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1272
1273 status = nfserrno(-ENOMEM);
1274 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1275 if (!async_copy)
1276 goto out;
1277 if (!nfs4_init_cp_state(nn, copy)) {
1278 kfree(async_copy);
1279 goto out;
1280 }
1281 refcount_set(&async_copy->refcount, 1);
1282 memcpy(©->cp_res.cb_stateid, ©->cp_stateid,
1283 sizeof(copy->cp_stateid));
1284 dup_copy_fields(copy, async_copy);
1285 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1286 async_copy, "%s", "copy thread");
1287 if (IS_ERR(async_copy->copy_task))
1288 goto out_err;
1289 spin_lock(&async_copy->cp_clp->async_lock);
1290 list_add(&async_copy->copies,
1291 &async_copy->cp_clp->async_copies);
1292 spin_unlock(&async_copy->cp_clp->async_lock);
1293 wake_up_process(async_copy->copy_task);
1294 status = nfs_ok;
1295 } else
1296 status = nfsd4_do_copy(copy, 1);
1297out:
1298 return status;
1299out_err:
1300 cleanup_async_copy(async_copy);
1301 goto out;
1302}
1303
1304struct nfsd4_copy *
1305find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1306{
1307 struct nfsd4_copy *copy;
1308
1309 spin_lock(&clp->async_lock);
1310 list_for_each_entry(copy, &clp->async_copies, copies) {
1311 if (memcmp(©->cp_stateid, stateid, NFS4_STATEID_SIZE))
1312 continue;
1313 refcount_inc(©->refcount);
1314 spin_unlock(&clp->async_lock);
1315 return copy;
1316 }
1317 spin_unlock(&clp->async_lock);
1318 return NULL;
1319}
1320
1321static __be32
1322nfsd4_offload_cancel(struct svc_rqst *rqstp,
1323 struct nfsd4_compound_state *cstate,
1324 union nfsd4_op_u *u)
1325{
1326 struct nfsd4_offload_status *os = &u->offload_status;
1327 __be32 status = 0;
1328 struct nfsd4_copy *copy;
1329 struct nfs4_client *clp = cstate->clp;
1330
1331 copy = find_async_copy(clp, &os->stateid);
1332 if (copy)
1333 nfsd4_stop_copy(copy);
1334 else
1335 status = nfserr_bad_stateid;
1336
1337 return status;
1338}
1339
1340static __be32
1341nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1342 struct nfsd4_fallocate *fallocate, int flags)
1343{
1344 __be32 status;
1345 struct nfsd_file *nf;
1346
1347 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1348 &fallocate->falloc_stateid,
1349 WR_STATE, &nf);
1350 if (status != nfs_ok) {
1351 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1352 return status;
1353 }
1354
1355 status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1356 fallocate->falloc_offset,
1357 fallocate->falloc_length,
1358 flags);
1359 nfsd_file_put(nf);
1360 return status;
1361}
1362static __be32
1363nfsd4_offload_status(struct svc_rqst *rqstp,
1364 struct nfsd4_compound_state *cstate,
1365 union nfsd4_op_u *u)
1366{
1367 struct nfsd4_offload_status *os = &u->offload_status;
1368 __be32 status = 0;
1369 struct nfsd4_copy *copy;
1370 struct nfs4_client *clp = cstate->clp;
1371
1372 copy = find_async_copy(clp, &os->stateid);
1373 if (copy) {
1374 os->count = copy->cp_res.wr_bytes_written;
1375 nfs4_put_copy(copy);
1376 } else
1377 status = nfserr_bad_stateid;
1378
1379 return status;
1380}
1381
1382static __be32
1383nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1384 union nfsd4_op_u *u)
1385{
1386 return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
1387}
1388
1389static __be32
1390nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1391 union nfsd4_op_u *u)
1392{
1393 return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
1394 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1395}
1396
1397static __be32
1398nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1399 union nfsd4_op_u *u)
1400{
1401 struct nfsd4_seek *seek = &u->seek;
1402 int whence;
1403 __be32 status;
1404 struct nfsd_file *nf;
1405
1406 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1407 &seek->seek_stateid,
1408 RD_STATE, &nf);
1409 if (status) {
1410 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1411 return status;
1412 }
1413
1414 switch (seek->seek_whence) {
1415 case NFS4_CONTENT_DATA:
1416 whence = SEEK_DATA;
1417 break;
1418 case NFS4_CONTENT_HOLE:
1419 whence = SEEK_HOLE;
1420 break;
1421 default:
1422 status = nfserr_union_notsupp;
1423 goto out;
1424 }
1425
1426 /*
1427 * Note: This call does change file->f_pos, but nothing in NFSD
1428 * should ever file->f_pos.
1429 */
1430 seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
1431 if (seek->seek_pos < 0)
1432 status = nfserrno(seek->seek_pos);
1433 else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
1434 seek->seek_eof = true;
1435
1436out:
1437 nfsd_file_put(nf);
1438 return status;
1439}
1440
1441/* This routine never returns NFS_OK! If there are no other errors, it
1442 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1443 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
1444 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1445 */
1446static __be32
1447_nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1448 struct nfsd4_verify *verify)
1449{
1450 __be32 *buf, *p;
1451 int count;
1452 __be32 status;
1453
1454 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1455 if (status)
1456 return status;
1457
1458 status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1459 if (status)
1460 return status;
1461
1462 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1463 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1464 return nfserr_inval;
1465 if (verify->ve_attrlen & 3)
1466 return nfserr_inval;
1467
1468 /* count in words:
1469 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1470 */
1471 count = 4 + (verify->ve_attrlen >> 2);
1472 buf = kmalloc(count << 2, GFP_KERNEL);
1473 if (!buf)
1474 return nfserr_jukebox;
1475
1476 p = buf;
1477 status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1478 cstate->current_fh.fh_export,
1479 cstate->current_fh.fh_dentry,
1480 verify->ve_bmval,
1481 rqstp, 0);
1482 /*
1483 * If nfsd4_encode_fattr() ran out of space, assume that's because
1484 * the attributes are longer (hence different) than those given:
1485 */
1486 if (status == nfserr_resource)
1487 status = nfserr_not_same;
1488 if (status)
1489 goto out_kfree;
1490
1491 /* skip bitmap */
1492 p = buf + 1 + ntohl(buf[0]);
1493 status = nfserr_not_same;
1494 if (ntohl(*p++) != verify->ve_attrlen)
1495 goto out_kfree;
1496 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1497 status = nfserr_same;
1498
1499out_kfree:
1500 kfree(buf);
1501 return status;
1502}
1503
1504static __be32
1505nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1506 union nfsd4_op_u *u)
1507{
1508 __be32 status;
1509
1510 status = _nfsd4_verify(rqstp, cstate, &u->verify);
1511 return status == nfserr_not_same ? nfs_ok : status;
1512}
1513
1514static __be32
1515nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1516 union nfsd4_op_u *u)
1517{
1518 __be32 status;
1519
1520 status = _nfsd4_verify(rqstp, cstate, &u->nverify);
1521 return status == nfserr_same ? nfs_ok : status;
1522}
1523
1524#ifdef CONFIG_NFSD_PNFS
1525static const struct nfsd4_layout_ops *
1526nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
1527{
1528 if (!exp->ex_layout_types) {
1529 dprintk("%s: export does not support pNFS\n", __func__);
1530 return NULL;
1531 }
1532
1533 if (layout_type >= LAYOUT_TYPE_MAX ||
1534 !(exp->ex_layout_types & (1 << layout_type))) {
1535 dprintk("%s: layout type %d not supported\n",
1536 __func__, layout_type);
1537 return NULL;
1538 }
1539
1540 return nfsd4_layout_ops[layout_type];
1541}
1542
1543static __be32
1544nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
1545 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1546{
1547 struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
1548 const struct nfsd4_layout_ops *ops;
1549 struct nfsd4_deviceid_map *map;
1550 struct svc_export *exp;
1551 __be32 nfserr;
1552
1553 dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
1554 __func__,
1555 gdp->gd_layout_type,
1556 gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
1557 gdp->gd_maxcount);
1558
1559 map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
1560 if (!map) {
1561 dprintk("%s: couldn't find device ID to export mapping!\n",
1562 __func__);
1563 return nfserr_noent;
1564 }
1565
1566 exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
1567 if (IS_ERR(exp)) {
1568 dprintk("%s: could not find device id\n", __func__);
1569 return nfserr_noent;
1570 }
1571
1572 nfserr = nfserr_layoutunavailable;
1573 ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
1574 if (!ops)
1575 goto out;
1576
1577 nfserr = nfs_ok;
1578 if (gdp->gd_maxcount != 0) {
1579 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
1580 rqstp, cstate->session->se_client, gdp);
1581 }
1582
1583 gdp->gd_notify_types &= ops->notify_types;
1584out:
1585 exp_put(exp);
1586 return nfserr;
1587}
1588
1589static void
1590nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
1591{
1592 kfree(u->getdeviceinfo.gd_device);
1593}
1594
1595static __be32
1596nfsd4_layoutget(struct svc_rqst *rqstp,
1597 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1598{
1599 struct nfsd4_layoutget *lgp = &u->layoutget;
1600 struct svc_fh *current_fh = &cstate->current_fh;
1601 const struct nfsd4_layout_ops *ops;
1602 struct nfs4_layout_stateid *ls;
1603 __be32 nfserr;
1604 int accmode = NFSD_MAY_READ_IF_EXEC;
1605
1606 switch (lgp->lg_seg.iomode) {
1607 case IOMODE_READ:
1608 accmode |= NFSD_MAY_READ;
1609 break;
1610 case IOMODE_RW:
1611 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
1612 break;
1613 default:
1614 dprintk("%s: invalid iomode %d\n",
1615 __func__, lgp->lg_seg.iomode);
1616 nfserr = nfserr_badiomode;
1617 goto out;
1618 }
1619
1620 nfserr = fh_verify(rqstp, current_fh, 0, accmode);
1621 if (nfserr)
1622 goto out;
1623
1624 nfserr = nfserr_layoutunavailable;
1625 ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
1626 if (!ops)
1627 goto out;
1628
1629 /*
1630 * Verify minlength and range as per RFC5661:
1631 * o If loga_length is less than loga_minlength,
1632 * the metadata server MUST return NFS4ERR_INVAL.
1633 * o If the sum of loga_offset and loga_minlength exceeds
1634 * NFS4_UINT64_MAX, and loga_minlength is not
1635 * NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
1636 * o If the sum of loga_offset and loga_length exceeds
1637 * NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
1638 * the error NFS4ERR_INVAL MUST result.
1639 */
1640 nfserr = nfserr_inval;
1641 if (lgp->lg_seg.length < lgp->lg_minlength ||
1642 (lgp->lg_minlength != NFS4_MAX_UINT64 &&
1643 lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
1644 (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
1645 lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
1646 goto out;
1647 if (lgp->lg_seg.length == 0)
1648 goto out;
1649
1650 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
1651 true, lgp->lg_layout_type, &ls);
1652 if (nfserr) {
1653 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
1654 goto out;
1655 }
1656
1657 nfserr = nfserr_recallconflict;
1658 if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
1659 goto out_put_stid;
1660
1661 nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
1662 current_fh, lgp);
1663 if (nfserr)
1664 goto out_put_stid;
1665
1666 nfserr = nfsd4_insert_layout(lgp, ls);
1667
1668out_put_stid:
1669 mutex_unlock(&ls->ls_mutex);
1670 nfs4_put_stid(&ls->ls_stid);
1671out:
1672 return nfserr;
1673}
1674
1675static void
1676nfsd4_layoutget_release(union nfsd4_op_u *u)
1677{
1678 kfree(u->layoutget.lg_content);
1679}
1680
1681static __be32
1682nfsd4_layoutcommit(struct svc_rqst *rqstp,
1683 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1684{
1685 struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
1686 const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
1687 struct svc_fh *current_fh = &cstate->current_fh;
1688 const struct nfsd4_layout_ops *ops;
1689 loff_t new_size = lcp->lc_last_wr + 1;
1690 struct inode *inode;
1691 struct nfs4_layout_stateid *ls;
1692 __be32 nfserr;
1693
1694 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
1695 if (nfserr)
1696 goto out;
1697
1698 nfserr = nfserr_layoutunavailable;
1699 ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
1700 if (!ops)
1701 goto out;
1702 inode = d_inode(current_fh->fh_dentry);
1703
1704 nfserr = nfserr_inval;
1705 if (new_size <= seg->offset) {
1706 dprintk("pnfsd: last write before layout segment\n");
1707 goto out;
1708 }
1709 if (new_size > seg->offset + seg->length) {
1710 dprintk("pnfsd: last write beyond layout segment\n");
1711 goto out;
1712 }
1713 if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
1714 dprintk("pnfsd: layoutcommit beyond EOF\n");
1715 goto out;
1716 }
1717
1718 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
1719 false, lcp->lc_layout_type,
1720 &ls);
1721 if (nfserr) {
1722 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
1723 /* fixup error code as per RFC5661 */
1724 if (nfserr == nfserr_bad_stateid)
1725 nfserr = nfserr_badlayout;
1726 goto out;
1727 }
1728
1729 /* LAYOUTCOMMIT does not require any serialization */
1730 mutex_unlock(&ls->ls_mutex);
1731
1732 if (new_size > i_size_read(inode)) {
1733 lcp->lc_size_chg = 1;
1734 lcp->lc_newsize = new_size;
1735 } else {
1736 lcp->lc_size_chg = 0;
1737 }
1738
1739 nfserr = ops->proc_layoutcommit(inode, lcp);
1740 nfs4_put_stid(&ls->ls_stid);
1741out:
1742 return nfserr;
1743}
1744
1745static __be32
1746nfsd4_layoutreturn(struct svc_rqst *rqstp,
1747 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
1748{
1749 struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
1750 struct svc_fh *current_fh = &cstate->current_fh;
1751 __be32 nfserr;
1752
1753 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
1754 if (nfserr)
1755 goto out;
1756
1757 nfserr = nfserr_layoutunavailable;
1758 if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
1759 goto out;
1760
1761 switch (lrp->lr_seg.iomode) {
1762 case IOMODE_READ:
1763 case IOMODE_RW:
1764 case IOMODE_ANY:
1765 break;
1766 default:
1767 dprintk("%s: invalid iomode %d\n", __func__,
1768 lrp->lr_seg.iomode);
1769 nfserr = nfserr_inval;
1770 goto out;
1771 }
1772
1773 switch (lrp->lr_return_type) {
1774 case RETURN_FILE:
1775 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
1776 break;
1777 case RETURN_FSID:
1778 case RETURN_ALL:
1779 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
1780 break;
1781 default:
1782 dprintk("%s: invalid return_type %d\n", __func__,
1783 lrp->lr_return_type);
1784 nfserr = nfserr_inval;
1785 break;
1786 }
1787out:
1788 return nfserr;
1789}
1790#endif /* CONFIG_NFSD_PNFS */
1791
1792/*
1793 * NULL call.
1794 */
1795static __be32
1796nfsd4_proc_null(struct svc_rqst *rqstp)
1797{
1798 return nfs_ok;
1799}
1800
1801static inline void nfsd4_increment_op_stats(u32 opnum)
1802{
1803 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
1804 nfsdstats.nfs4_opcount[opnum]++;
1805}
1806
1807static const struct nfsd4_operation nfsd4_ops[];
1808
1809static const char *nfsd4_op_name(unsigned opnum);
1810
1811/*
1812 * Enforce NFSv4.1 COMPOUND ordering rules:
1813 *
1814 * Also note, enforced elsewhere:
1815 * - SEQUENCE other than as first op results in
1816 * NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
1817 * - BIND_CONN_TO_SESSION must be the only op in its compound.
1818 * (Enforced in nfsd4_bind_conn_to_session().)
1819 * - DESTROY_SESSION must be the final operation in a compound, if
1820 * sessionid's in SEQUENCE and DESTROY_SESSION are the same.
1821 * (Enforced in nfsd4_destroy_session().)
1822 */
1823static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
1824{
1825 struct nfsd4_op *first_op = &args->ops[0];
1826
1827 /* These ordering requirements don't apply to NFSv4.0: */
1828 if (args->minorversion == 0)
1829 return nfs_ok;
1830 /* This is weird, but OK, not our problem: */
1831 if (args->opcnt == 0)
1832 return nfs_ok;
1833 if (first_op->status == nfserr_op_illegal)
1834 return nfs_ok;
1835 if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
1836 return nfserr_op_not_in_session;
1837 if (first_op->opnum == OP_SEQUENCE)
1838 return nfs_ok;
1839 /*
1840 * So first_op is something allowed outside a session, like
1841 * EXCHANGE_ID; but then it has to be the only op in the
1842 * compound:
1843 */
1844 if (args->opcnt != 1)
1845 return nfserr_not_only_op;
1846 return nfs_ok;
1847}
1848
1849const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
1850{
1851 return &nfsd4_ops[op->opnum];
1852}
1853
1854bool nfsd4_cache_this_op(struct nfsd4_op *op)
1855{
1856 if (op->opnum == OP_ILLEGAL)
1857 return false;
1858 return OPDESC(op)->op_flags & OP_CACHEME;
1859}
1860
1861static bool need_wrongsec_check(struct svc_rqst *rqstp)
1862{
1863 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1864 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1865 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
1866 struct nfsd4_op *next = &argp->ops[resp->opcnt];
1867 const struct nfsd4_operation *thisd = OPDESC(this);
1868 const struct nfsd4_operation *nextd;
1869
1870 /*
1871 * Most ops check wronsec on our own; only the putfh-like ops
1872 * have special rules.
1873 */
1874 if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
1875 return false;
1876 /*
1877 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
1878 * put-filehandle operation if we're not going to use the
1879 * result:
1880 */
1881 if (argp->opcnt == resp->opcnt)
1882 return false;
1883 if (next->opnum == OP_ILLEGAL)
1884 return false;
1885 nextd = OPDESC(next);
1886 /*
1887 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
1888 * errors themselves as necessary; others should check for them
1889 * now:
1890 */
1891 return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
1892}
1893
1894static void svcxdr_init_encode(struct svc_rqst *rqstp,
1895 struct nfsd4_compoundres *resp)
1896{
1897 struct xdr_stream *xdr = &resp->xdr;
1898 struct xdr_buf *buf = &rqstp->rq_res;
1899 struct kvec *head = buf->head;
1900
1901 xdr->buf = buf;
1902 xdr->iov = head;
1903 xdr->p = head->iov_base + head->iov_len;
1904 xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
1905 /* Tail and page_len should be zero at this point: */
1906 buf->len = buf->head[0].iov_len;
1907 xdr->scratch.iov_len = 0;
1908 xdr->page_ptr = buf->pages - 1;
1909 buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
1910 - rqstp->rq_auth_slack;
1911}
1912
1913/*
1914 * COMPOUND call.
1915 */
1916static __be32
1917nfsd4_proc_compound(struct svc_rqst *rqstp)
1918{
1919 struct nfsd4_compoundargs *args = rqstp->rq_argp;
1920 struct nfsd4_compoundres *resp = rqstp->rq_resp;
1921 struct nfsd4_op *op;
1922 struct nfsd4_compound_state *cstate = &resp->cstate;
1923 struct svc_fh *current_fh = &cstate->current_fh;
1924 struct svc_fh *save_fh = &cstate->save_fh;
1925 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1926 __be32 status;
1927
1928 svcxdr_init_encode(rqstp, resp);
1929 resp->tagp = resp->xdr.p;
1930 /* reserve space for: taglen, tag, and opcnt */
1931 xdr_reserve_space(&resp->xdr, 8 + args->taglen);
1932 resp->taglen = args->taglen;
1933 resp->tag = args->tag;
1934 resp->rqstp = rqstp;
1935 cstate->minorversion = args->minorversion;
1936 fh_init(current_fh, NFS4_FHSIZE);
1937 fh_init(save_fh, NFS4_FHSIZE);
1938 /*
1939 * Don't use the deferral mechanism for NFSv4; compounds make it
1940 * too hard to avoid non-idempotency problems.
1941 */
1942 clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1943
1944 /*
1945 * According to RFC3010, this takes precedence over all other errors.
1946 */
1947 status = nfserr_minor_vers_mismatch;
1948 if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
1949 goto out;
1950 status = nfserr_resource;
1951 if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
1952 goto out;
1953
1954 status = nfs41_check_op_ordering(args);
1955 if (status) {
1956 op = &args->ops[0];
1957 op->status = status;
1958 resp->opcnt = 1;
1959 goto encode_op;
1960 }
1961
1962 trace_nfsd_compound(rqstp, args->opcnt);
1963 while (!status && resp->opcnt < args->opcnt) {
1964 op = &args->ops[resp->opcnt++];
1965
1966 /*
1967 * The XDR decode routines may have pre-set op->status;
1968 * for example, if there is a miscellaneous XDR error
1969 * it will be set to nfserr_bad_xdr.
1970 */
1971 if (op->status) {
1972 if (op->opnum == OP_OPEN)
1973 op->status = nfsd4_open_omfg(rqstp, cstate, op);
1974 goto encode_op;
1975 }
1976
1977 if (!current_fh->fh_dentry) {
1978 if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
1979 op->status = nfserr_nofilehandle;
1980 goto encode_op;
1981 }
1982 } else if (current_fh->fh_export->ex_fslocs.migrated &&
1983 !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
1984 op->status = nfserr_moved;
1985 goto encode_op;
1986 }
1987
1988 fh_clear_wcc(current_fh);
1989
1990 /* If op is non-idempotent */
1991 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
1992 /*
1993 * Don't execute this op if we couldn't encode a
1994 * succesful reply:
1995 */
1996 u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
1997 /*
1998 * Plus if there's another operation, make sure
1999 * we'll have space to at least encode an error:
2000 */
2001 if (resp->opcnt < args->opcnt)
2002 plen += COMPOUND_ERR_SLACK_SPACE;
2003 op->status = nfsd4_check_resp_size(resp, plen);
2004 }
2005
2006 if (op->status)
2007 goto encode_op;
2008
2009 if (op->opdesc->op_get_currentstateid)
2010 op->opdesc->op_get_currentstateid(cstate, &op->u);
2011 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2012
2013 /* Only from SEQUENCE */
2014 if (cstate->status == nfserr_replay_cache) {
2015 dprintk("%s NFS4.1 replay from cache\n", __func__);
2016 status = op->status;
2017 goto out;
2018 }
2019 if (!op->status) {
2020 if (op->opdesc->op_set_currentstateid)
2021 op->opdesc->op_set_currentstateid(cstate, &op->u);
2022
2023 if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2024 clear_current_stateid(cstate);
2025
2026 if (need_wrongsec_check(rqstp))
2027 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2028 }
2029encode_op:
2030 if (op->status == nfserr_replay_me) {
2031 op->replay = &cstate->replay_owner->so_replay;
2032 nfsd4_encode_replay(&resp->xdr, op);
2033 status = op->status = op->replay->rp_status;
2034 } else {
2035 nfsd4_encode_operation(resp, op);
2036 status = op->status;
2037 }
2038
2039 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2040 nfsd4_op_name(op->opnum));
2041
2042 nfsd4_cstate_clear_replay(cstate);
2043 nfsd4_increment_op_stats(op->opnum);
2044 }
2045
2046 cstate->status = status;
2047 fh_put(current_fh);
2048 fh_put(save_fh);
2049 BUG_ON(cstate->replay_owner);
2050out:
2051 /* Reset deferral mechanism for RPC deferrals */
2052 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2053 dprintk("nfsv4 compound returned %d\n", ntohl(status));
2054 return status;
2055}
2056
2057#define op_encode_hdr_size (2)
2058#define op_encode_stateid_maxsz (XDR_QUADLEN(NFS4_STATEID_SIZE))
2059#define op_encode_verifier_maxsz (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2060#define op_encode_change_info_maxsz (5)
2061#define nfs4_fattr_bitmap_maxsz (4)
2062
2063/* We'll fall back on returning no lockowner if run out of space: */
2064#define op_encode_lockowner_maxsz (0)
2065#define op_encode_lock_denied_maxsz (8 + op_encode_lockowner_maxsz)
2066
2067#define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2068
2069#define op_encode_ace_maxsz (3 + nfs4_owner_maxsz)
2070#define op_encode_delegation_maxsz (1 + op_encode_stateid_maxsz + 1 + \
2071 op_encode_ace_maxsz)
2072
2073#define op_encode_channel_attrs_maxsz (6 + 1 + 1)
2074
2075static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2076{
2077 return (op_encode_hdr_size) * sizeof(__be32);
2078}
2079
2080static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2081{
2082 return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2083}
2084
2085static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2086{
2087 /* ac_supported, ac_resp_access */
2088 return (op_encode_hdr_size + 2)* sizeof(__be32);
2089}
2090
2091static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2092{
2093 return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2094}
2095
2096static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2097{
2098 return (op_encode_hdr_size + op_encode_change_info_maxsz
2099 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2100}
2101
2102/*
2103 * Note since this is an idempotent operation we won't insist on failing
2104 * the op prematurely if the estimate is too large. We may turn off splice
2105 * reads unnecessarily.
2106 */
2107static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2108 struct nfsd4_op *op)
2109{
2110 u32 *bmap = op->u.getattr.ga_bmval;
2111 u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2112 u32 ret = 0;
2113
2114 if (bmap0 & FATTR4_WORD0_ACL)
2115 return svc_max_payload(rqstp);
2116 if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2117 return svc_max_payload(rqstp);
2118
2119 if (bmap1 & FATTR4_WORD1_OWNER) {
2120 ret += IDMAP_NAMESZ + 4;
2121 bmap1 &= ~FATTR4_WORD1_OWNER;
2122 }
2123 if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2124 ret += IDMAP_NAMESZ + 4;
2125 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2126 }
2127 if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2128 ret += NFS4_FHSIZE + 4;
2129 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2130 }
2131 if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2132 ret += NFS4_MAXLABELLEN + 12;
2133 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2134 }
2135 /*
2136 * Largest of remaining attributes are 16 bytes (e.g.,
2137 * supported_attributes)
2138 */
2139 ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2140 /* bitmask, length */
2141 ret += 20;
2142 return ret;
2143}
2144
2145static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2146{
2147 return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2148}
2149
2150static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2151{
2152 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2153 * sizeof(__be32);
2154}
2155
2156static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2157{
2158 return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2159 * sizeof(__be32);
2160}
2161
2162static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2163{
2164 return (op_encode_hdr_size + op_encode_stateid_maxsz
2165 + op_encode_change_info_maxsz + 1
2166 + nfs4_fattr_bitmap_maxsz
2167 + op_encode_delegation_maxsz) * sizeof(__be32);
2168}
2169
2170static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2171{
2172 u32 maxcount = 0, rlen = 0;
2173
2174 maxcount = svc_max_payload(rqstp);
2175 rlen = min(op->u.read.rd_length, maxcount);
2176
2177 return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2178}
2179
2180static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2181{
2182 u32 maxcount = 0, rlen = 0;
2183
2184 maxcount = svc_max_payload(rqstp);
2185 rlen = min(op->u.readdir.rd_maxcount, maxcount);
2186
2187 return (op_encode_hdr_size + op_encode_verifier_maxsz +
2188 XDR_QUADLEN(rlen)) * sizeof(__be32);
2189}
2190
2191static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2192{
2193 return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2194}
2195
2196static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2197{
2198 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2199 * sizeof(__be32);
2200}
2201
2202static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2203{
2204 return (op_encode_hdr_size + op_encode_change_info_maxsz
2205 + op_encode_change_info_maxsz) * sizeof(__be32);
2206}
2207
2208static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2209 struct nfsd4_op *op)
2210{
2211 return (op_encode_hdr_size
2212 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2213}
2214
2215static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2216{
2217 return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2218 * sizeof(__be32);
2219}
2220
2221static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2222{
2223 return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2224}
2225
2226static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2227{
2228 return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2229 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2230}
2231
2232static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2233{
2234 return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2235 sizeof(__be32);
2236}
2237
2238static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2239{
2240 return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2241}
2242
2243static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2244{
2245 return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2246 1 + 1 + /* eir_flags, spr_how */\
2247 4 + /* spo_must_enforce & _allow with bitmap */\
2248 2 + /*eir_server_owner.so_minor_id */\
2249 /* eir_server_owner.so_major_id<> */\
2250 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2251 /* eir_server_scope<> */\
2252 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2253 1 + /* eir_server_impl_id array length */\
2254 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2255}
2256
2257static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2258{
2259 return (op_encode_hdr_size + \
2260 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2261 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2262}
2263
2264static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2265{
2266 return (op_encode_hdr_size + \
2267 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
2268 2 + /* csr_sequence, csr_flags */\
2269 op_encode_channel_attrs_maxsz + \
2270 op_encode_channel_attrs_maxsz) * sizeof(__be32);
2271}
2272
2273static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2274{
2275 return (op_encode_hdr_size +
2276 1 /* wr_callback */ +
2277 op_encode_stateid_maxsz /* wr_callback */ +
2278 2 /* wr_count */ +
2279 1 /* wr_committed */ +
2280 op_encode_verifier_maxsz +
2281 1 /* cr_consecutive */ +
2282 1 /* cr_synchronous */) * sizeof(__be32);
2283}
2284
2285static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
2286 struct nfsd4_op *op)
2287{
2288 return (op_encode_hdr_size +
2289 2 /* osr_count */ +
2290 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
2291}
2292
2293#ifdef CONFIG_NFSD_PNFS
2294static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2295{
2296 u32 maxcount = 0, rlen = 0;
2297
2298 maxcount = svc_max_payload(rqstp);
2299 rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
2300
2301 return (op_encode_hdr_size +
2302 1 /* gd_layout_type*/ +
2303 XDR_QUADLEN(rlen) +
2304 2 /* gd_notify_types */) * sizeof(__be32);
2305}
2306
2307/*
2308 * At this stage we don't really know what layout driver will handle the request,
2309 * so we need to define an arbitrary upper bound here.
2310 */
2311#define MAX_LAYOUT_SIZE 128
2312static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2313{
2314 return (op_encode_hdr_size +
2315 1 /* logr_return_on_close */ +
2316 op_encode_stateid_maxsz +
2317 1 /* nr of layouts */ +
2318 MAX_LAYOUT_SIZE) * sizeof(__be32);
2319}
2320
2321static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2322{
2323 return (op_encode_hdr_size +
2324 1 /* locr_newsize */ +
2325 2 /* ns_size */) * sizeof(__be32);
2326}
2327
2328static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2329{
2330 return (op_encode_hdr_size +
2331 1 /* lrs_stateid */ +
2332 op_encode_stateid_maxsz) * sizeof(__be32);
2333}
2334#endif /* CONFIG_NFSD_PNFS */
2335
2336
2337static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2338{
2339 return (op_encode_hdr_size + 3) * sizeof(__be32);
2340}
2341
2342static const struct nfsd4_operation nfsd4_ops[] = {
2343 [OP_ACCESS] = {
2344 .op_func = nfsd4_access,
2345 .op_name = "OP_ACCESS",
2346 .op_rsize_bop = nfsd4_access_rsize,
2347 },
2348 [OP_CLOSE] = {
2349 .op_func = nfsd4_close,
2350 .op_flags = OP_MODIFIES_SOMETHING,
2351 .op_name = "OP_CLOSE",
2352 .op_rsize_bop = nfsd4_status_stateid_rsize,
2353 .op_get_currentstateid = nfsd4_get_closestateid,
2354 .op_set_currentstateid = nfsd4_set_closestateid,
2355 },
2356 [OP_COMMIT] = {
2357 .op_func = nfsd4_commit,
2358 .op_flags = OP_MODIFIES_SOMETHING,
2359 .op_name = "OP_COMMIT",
2360 .op_rsize_bop = nfsd4_commit_rsize,
2361 },
2362 [OP_CREATE] = {
2363 .op_func = nfsd4_create,
2364 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
2365 .op_name = "OP_CREATE",
2366 .op_rsize_bop = nfsd4_create_rsize,
2367 },
2368 [OP_DELEGRETURN] = {
2369 .op_func = nfsd4_delegreturn,
2370 .op_flags = OP_MODIFIES_SOMETHING,
2371 .op_name = "OP_DELEGRETURN",
2372 .op_rsize_bop = nfsd4_only_status_rsize,
2373 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
2374 },
2375 [OP_GETATTR] = {
2376 .op_func = nfsd4_getattr,
2377 .op_flags = ALLOWED_ON_ABSENT_FS,
2378 .op_rsize_bop = nfsd4_getattr_rsize,
2379 .op_name = "OP_GETATTR",
2380 },
2381 [OP_GETFH] = {
2382 .op_func = nfsd4_getfh,
2383 .op_name = "OP_GETFH",
2384 .op_rsize_bop = nfsd4_getfh_rsize,
2385 },
2386 [OP_LINK] = {
2387 .op_func = nfsd4_link,
2388 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
2389 | OP_CACHEME,
2390 .op_name = "OP_LINK",
2391 .op_rsize_bop = nfsd4_link_rsize,
2392 },
2393 [OP_LOCK] = {
2394 .op_func = nfsd4_lock,
2395 .op_flags = OP_MODIFIES_SOMETHING |
2396 OP_NONTRIVIAL_ERROR_ENCODE,
2397 .op_name = "OP_LOCK",
2398 .op_rsize_bop = nfsd4_lock_rsize,
2399 .op_set_currentstateid = nfsd4_set_lockstateid,
2400 },
2401 [OP_LOCKT] = {
2402 .op_func = nfsd4_lockt,
2403 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
2404 .op_name = "OP_LOCKT",
2405 .op_rsize_bop = nfsd4_lock_rsize,
2406 },
2407 [OP_LOCKU] = {
2408 .op_func = nfsd4_locku,
2409 .op_flags = OP_MODIFIES_SOMETHING,
2410 .op_name = "OP_LOCKU",
2411 .op_rsize_bop = nfsd4_status_stateid_rsize,
2412 .op_get_currentstateid = nfsd4_get_lockustateid,
2413 },
2414 [OP_LOOKUP] = {
2415 .op_func = nfsd4_lookup,
2416 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2417 .op_name = "OP_LOOKUP",
2418 .op_rsize_bop = nfsd4_only_status_rsize,
2419 },
2420 [OP_LOOKUPP] = {
2421 .op_func = nfsd4_lookupp,
2422 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
2423 .op_name = "OP_LOOKUPP",
2424 .op_rsize_bop = nfsd4_only_status_rsize,
2425 },
2426 [OP_NVERIFY] = {
2427 .op_func = nfsd4_nverify,
2428 .op_name = "OP_NVERIFY",
2429 .op_rsize_bop = nfsd4_only_status_rsize,
2430 },
2431 [OP_OPEN] = {
2432 .op_func = nfsd4_open,
2433 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2434 .op_name = "OP_OPEN",
2435 .op_rsize_bop = nfsd4_open_rsize,
2436 .op_set_currentstateid = nfsd4_set_openstateid,
2437 },
2438 [OP_OPEN_CONFIRM] = {
2439 .op_func = nfsd4_open_confirm,
2440 .op_flags = OP_MODIFIES_SOMETHING,
2441 .op_name = "OP_OPEN_CONFIRM",
2442 .op_rsize_bop = nfsd4_status_stateid_rsize,
2443 },
2444 [OP_OPEN_DOWNGRADE] = {
2445 .op_func = nfsd4_open_downgrade,
2446 .op_flags = OP_MODIFIES_SOMETHING,
2447 .op_name = "OP_OPEN_DOWNGRADE",
2448 .op_rsize_bop = nfsd4_status_stateid_rsize,
2449 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
2450 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
2451 },
2452 [OP_PUTFH] = {
2453 .op_func = nfsd4_putfh,
2454 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2455 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2456 .op_name = "OP_PUTFH",
2457 .op_rsize_bop = nfsd4_only_status_rsize,
2458 },
2459 [OP_PUTPUBFH] = {
2460 .op_func = nfsd4_putrootfh,
2461 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2462 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2463 .op_name = "OP_PUTPUBFH",
2464 .op_rsize_bop = nfsd4_only_status_rsize,
2465 },
2466 [OP_PUTROOTFH] = {
2467 .op_func = nfsd4_putrootfh,
2468 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2469 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
2470 .op_name = "OP_PUTROOTFH",
2471 .op_rsize_bop = nfsd4_only_status_rsize,
2472 },
2473 [OP_READ] = {
2474 .op_func = nfsd4_read,
2475 .op_release = nfsd4_read_release,
2476 .op_name = "OP_READ",
2477 .op_rsize_bop = nfsd4_read_rsize,
2478 .op_get_currentstateid = nfsd4_get_readstateid,
2479 },
2480 [OP_READDIR] = {
2481 .op_func = nfsd4_readdir,
2482 .op_name = "OP_READDIR",
2483 .op_rsize_bop = nfsd4_readdir_rsize,
2484 },
2485 [OP_READLINK] = {
2486 .op_func = nfsd4_readlink,
2487 .op_name = "OP_READLINK",
2488 .op_rsize_bop = nfsd4_readlink_rsize,
2489 },
2490 [OP_REMOVE] = {
2491 .op_func = nfsd4_remove,
2492 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2493 .op_name = "OP_REMOVE",
2494 .op_rsize_bop = nfsd4_remove_rsize,
2495 },
2496 [OP_RENAME] = {
2497 .op_func = nfsd4_rename,
2498 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2499 .op_name = "OP_RENAME",
2500 .op_rsize_bop = nfsd4_rename_rsize,
2501 },
2502 [OP_RENEW] = {
2503 .op_func = nfsd4_renew,
2504 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2505 | OP_MODIFIES_SOMETHING,
2506 .op_name = "OP_RENEW",
2507 .op_rsize_bop = nfsd4_only_status_rsize,
2508
2509 },
2510 [OP_RESTOREFH] = {
2511 .op_func = nfsd4_restorefh,
2512 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2513 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
2514 .op_name = "OP_RESTOREFH",
2515 .op_rsize_bop = nfsd4_only_status_rsize,
2516 },
2517 [OP_SAVEFH] = {
2518 .op_func = nfsd4_savefh,
2519 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
2520 .op_name = "OP_SAVEFH",
2521 .op_rsize_bop = nfsd4_only_status_rsize,
2522 },
2523 [OP_SECINFO] = {
2524 .op_func = nfsd4_secinfo,
2525 .op_release = nfsd4_secinfo_release,
2526 .op_flags = OP_HANDLES_WRONGSEC,
2527 .op_name = "OP_SECINFO",
2528 .op_rsize_bop = nfsd4_secinfo_rsize,
2529 },
2530 [OP_SETATTR] = {
2531 .op_func = nfsd4_setattr,
2532 .op_name = "OP_SETATTR",
2533 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
2534 | OP_NONTRIVIAL_ERROR_ENCODE,
2535 .op_rsize_bop = nfsd4_setattr_rsize,
2536 .op_get_currentstateid = nfsd4_get_setattrstateid,
2537 },
2538 [OP_SETCLIENTID] = {
2539 .op_func = nfsd4_setclientid,
2540 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2541 | OP_MODIFIES_SOMETHING | OP_CACHEME
2542 | OP_NONTRIVIAL_ERROR_ENCODE,
2543 .op_name = "OP_SETCLIENTID",
2544 .op_rsize_bop = nfsd4_setclientid_rsize,
2545 },
2546 [OP_SETCLIENTID_CONFIRM] = {
2547 .op_func = nfsd4_setclientid_confirm,
2548 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2549 | OP_MODIFIES_SOMETHING | OP_CACHEME,
2550 .op_name = "OP_SETCLIENTID_CONFIRM",
2551 .op_rsize_bop = nfsd4_only_status_rsize,
2552 },
2553 [OP_VERIFY] = {
2554 .op_func = nfsd4_verify,
2555 .op_name = "OP_VERIFY",
2556 .op_rsize_bop = nfsd4_only_status_rsize,
2557 },
2558 [OP_WRITE] = {
2559 .op_func = nfsd4_write,
2560 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
2561 .op_name = "OP_WRITE",
2562 .op_rsize_bop = nfsd4_write_rsize,
2563 .op_get_currentstateid = nfsd4_get_writestateid,
2564 },
2565 [OP_RELEASE_LOCKOWNER] = {
2566 .op_func = nfsd4_release_lockowner,
2567 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
2568 | OP_MODIFIES_SOMETHING,
2569 .op_name = "OP_RELEASE_LOCKOWNER",
2570 .op_rsize_bop = nfsd4_only_status_rsize,
2571 },
2572
2573 /* NFSv4.1 operations */
2574 [OP_EXCHANGE_ID] = {
2575 .op_func = nfsd4_exchange_id,
2576 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2577 | OP_MODIFIES_SOMETHING,
2578 .op_name = "OP_EXCHANGE_ID",
2579 .op_rsize_bop = nfsd4_exchange_id_rsize,
2580 },
2581 [OP_BACKCHANNEL_CTL] = {
2582 .op_func = nfsd4_backchannel_ctl,
2583 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2584 .op_name = "OP_BACKCHANNEL_CTL",
2585 .op_rsize_bop = nfsd4_only_status_rsize,
2586 },
2587 [OP_BIND_CONN_TO_SESSION] = {
2588 .op_func = nfsd4_bind_conn_to_session,
2589 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2590 | OP_MODIFIES_SOMETHING,
2591 .op_name = "OP_BIND_CONN_TO_SESSION",
2592 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
2593 },
2594 [OP_CREATE_SESSION] = {
2595 .op_func = nfsd4_create_session,
2596 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2597 | OP_MODIFIES_SOMETHING,
2598 .op_name = "OP_CREATE_SESSION",
2599 .op_rsize_bop = nfsd4_create_session_rsize,
2600 },
2601 [OP_DESTROY_SESSION] = {
2602 .op_func = nfsd4_destroy_session,
2603 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2604 | OP_MODIFIES_SOMETHING,
2605 .op_name = "OP_DESTROY_SESSION",
2606 .op_rsize_bop = nfsd4_only_status_rsize,
2607 },
2608 [OP_SEQUENCE] = {
2609 .op_func = nfsd4_sequence,
2610 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
2611 .op_name = "OP_SEQUENCE",
2612 .op_rsize_bop = nfsd4_sequence_rsize,
2613 },
2614 [OP_DESTROY_CLIENTID] = {
2615 .op_func = nfsd4_destroy_clientid,
2616 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
2617 | OP_MODIFIES_SOMETHING,
2618 .op_name = "OP_DESTROY_CLIENTID",
2619 .op_rsize_bop = nfsd4_only_status_rsize,
2620 },
2621 [OP_RECLAIM_COMPLETE] = {
2622 .op_func = nfsd4_reclaim_complete,
2623 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2624 .op_name = "OP_RECLAIM_COMPLETE",
2625 .op_rsize_bop = nfsd4_only_status_rsize,
2626 },
2627 [OP_SECINFO_NO_NAME] = {
2628 .op_func = nfsd4_secinfo_no_name,
2629 .op_release = nfsd4_secinfo_no_name_release,
2630 .op_flags = OP_HANDLES_WRONGSEC,
2631 .op_name = "OP_SECINFO_NO_NAME",
2632 .op_rsize_bop = nfsd4_secinfo_rsize,
2633 },
2634 [OP_TEST_STATEID] = {
2635 .op_func = nfsd4_test_stateid,
2636 .op_flags = ALLOWED_WITHOUT_FH,
2637 .op_name = "OP_TEST_STATEID",
2638 .op_rsize_bop = nfsd4_test_stateid_rsize,
2639 },
2640 [OP_FREE_STATEID] = {
2641 .op_func = nfsd4_free_stateid,
2642 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
2643 .op_name = "OP_FREE_STATEID",
2644 .op_get_currentstateid = nfsd4_get_freestateid,
2645 .op_rsize_bop = nfsd4_only_status_rsize,
2646 },
2647#ifdef CONFIG_NFSD_PNFS
2648 [OP_GETDEVICEINFO] = {
2649 .op_func = nfsd4_getdeviceinfo,
2650 .op_release = nfsd4_getdeviceinfo_release,
2651 .op_flags = ALLOWED_WITHOUT_FH,
2652 .op_name = "OP_GETDEVICEINFO",
2653 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
2654 },
2655 [OP_LAYOUTGET] = {
2656 .op_func = nfsd4_layoutget,
2657 .op_release = nfsd4_layoutget_release,
2658 .op_flags = OP_MODIFIES_SOMETHING,
2659 .op_name = "OP_LAYOUTGET",
2660 .op_rsize_bop = nfsd4_layoutget_rsize,
2661 },
2662 [OP_LAYOUTCOMMIT] = {
2663 .op_func = nfsd4_layoutcommit,
2664 .op_flags = OP_MODIFIES_SOMETHING,
2665 .op_name = "OP_LAYOUTCOMMIT",
2666 .op_rsize_bop = nfsd4_layoutcommit_rsize,
2667 },
2668 [OP_LAYOUTRETURN] = {
2669 .op_func = nfsd4_layoutreturn,
2670 .op_flags = OP_MODIFIES_SOMETHING,
2671 .op_name = "OP_LAYOUTRETURN",
2672 .op_rsize_bop = nfsd4_layoutreturn_rsize,
2673 },
2674#endif /* CONFIG_NFSD_PNFS */
2675
2676 /* NFSv4.2 operations */
2677 [OP_ALLOCATE] = {
2678 .op_func = nfsd4_allocate,
2679 .op_flags = OP_MODIFIES_SOMETHING,
2680 .op_name = "OP_ALLOCATE",
2681 .op_rsize_bop = nfsd4_only_status_rsize,
2682 },
2683 [OP_DEALLOCATE] = {
2684 .op_func = nfsd4_deallocate,
2685 .op_flags = OP_MODIFIES_SOMETHING,
2686 .op_name = "OP_DEALLOCATE",
2687 .op_rsize_bop = nfsd4_only_status_rsize,
2688 },
2689 [OP_CLONE] = {
2690 .op_func = nfsd4_clone,
2691 .op_flags = OP_MODIFIES_SOMETHING,
2692 .op_name = "OP_CLONE",
2693 .op_rsize_bop = nfsd4_only_status_rsize,
2694 },
2695 [OP_COPY] = {
2696 .op_func = nfsd4_copy,
2697 .op_flags = OP_MODIFIES_SOMETHING,
2698 .op_name = "OP_COPY",
2699 .op_rsize_bop = nfsd4_copy_rsize,
2700 },
2701 [OP_SEEK] = {
2702 .op_func = nfsd4_seek,
2703 .op_name = "OP_SEEK",
2704 .op_rsize_bop = nfsd4_seek_rsize,
2705 },
2706 [OP_OFFLOAD_STATUS] = {
2707 .op_func = nfsd4_offload_status,
2708 .op_name = "OP_OFFLOAD_STATUS",
2709 .op_rsize_bop = nfsd4_offload_status_rsize,
2710 },
2711 [OP_OFFLOAD_CANCEL] = {
2712 .op_func = nfsd4_offload_cancel,
2713 .op_flags = OP_MODIFIES_SOMETHING,
2714 .op_name = "OP_OFFLOAD_CANCEL",
2715 .op_rsize_bop = nfsd4_only_status_rsize,
2716 },
2717};
2718
2719/**
2720 * nfsd4_spo_must_allow - Determine if the compound op contains an
2721 * operation that is allowed to be sent with machine credentials
2722 *
2723 * @rqstp: a pointer to the struct svc_rqst
2724 *
2725 * Checks to see if the compound contains a spo_must_allow op
2726 * and confirms that it was sent with the proper machine creds.
2727 */
2728
2729bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
2730{
2731 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2732 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2733 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2734 struct nfsd4_compound_state *cstate = &resp->cstate;
2735 struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
2736 u32 opiter;
2737
2738 if (!cstate->minorversion)
2739 return false;
2740
2741 if (cstate->spo_must_allowed == true)
2742 return true;
2743
2744 opiter = resp->opcnt;
2745 while (opiter < argp->opcnt) {
2746 this = &argp->ops[opiter++];
2747 if (test_bit(this->opnum, allow->u.longs) &&
2748 cstate->clp->cl_mach_cred &&
2749 nfsd4_mach_creds_match(cstate->clp, rqstp)) {
2750 cstate->spo_must_allowed = true;
2751 return true;
2752 }
2753 }
2754 cstate->spo_must_allowed = false;
2755 return false;
2756}
2757
2758int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
2759{
2760 if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
2761 return op_encode_hdr_size * sizeof(__be32);
2762
2763 BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
2764 return OPDESC(op)->op_rsize_bop(rqstp, op);
2765}
2766
2767void warn_on_nonidempotent_op(struct nfsd4_op *op)
2768{
2769 if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
2770 pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
2771 op->opnum, nfsd4_op_name(op->opnum));
2772 WARN_ON_ONCE(1);
2773 }
2774}
2775
2776static const char *nfsd4_op_name(unsigned opnum)
2777{
2778 if (opnum < ARRAY_SIZE(nfsd4_ops))
2779 return nfsd4_ops[opnum].op_name;
2780 return "unknown_operation";
2781}
2782
2783#define nfsd4_voidres nfsd4_voidargs
2784struct nfsd4_voidargs { int dummy; };
2785
2786static const struct svc_procedure nfsd_procedures4[2] = {
2787 [NFSPROC4_NULL] = {
2788 .pc_func = nfsd4_proc_null,
2789 .pc_encode = nfs4svc_encode_voidres,
2790 .pc_argsize = sizeof(struct nfsd4_voidargs),
2791 .pc_ressize = sizeof(struct nfsd4_voidres),
2792 .pc_cachetype = RC_NOCACHE,
2793 .pc_xdrressize = 1,
2794 },
2795 [NFSPROC4_COMPOUND] = {
2796 .pc_func = nfsd4_proc_compound,
2797 .pc_decode = nfs4svc_decode_compoundargs,
2798 .pc_encode = nfs4svc_encode_compoundres,
2799 .pc_argsize = sizeof(struct nfsd4_compoundargs),
2800 .pc_ressize = sizeof(struct nfsd4_compoundres),
2801 .pc_release = nfsd4_release_compoundargs,
2802 .pc_cachetype = RC_NOCACHE,
2803 .pc_xdrressize = NFSD_BUFSIZE/4,
2804 },
2805};
2806
2807static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
2808const struct svc_version nfsd_version4 = {
2809 .vs_vers = 4,
2810 .vs_nproc = 2,
2811 .vs_proc = nfsd_procedures4,
2812 .vs_count = nfsd_count3,
2813 .vs_dispatch = nfsd_dispatch,
2814 .vs_xdrsize = NFS4_SVC_XDRSIZE,
2815 .vs_rpcb_optnl = true,
2816 .vs_need_cong_ctrl = true,
2817};
2818
2819/*
2820 * Local variables:
2821 * c-basic-offset: 8
2822 * End:
2823 */
1/*
2 * Server-side procedures for NFSv4.
3 *
4 * Copyright (c) 2002 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <andros@umich.edu>
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */
35#include <linux/fs_struct.h>
36#include <linux/file.h>
37#include <linux/falloc.h>
38#include <linux/slab.h>
39#include <linux/kthread.h>
40#include <linux/namei.h>
41
42#include <linux/sunrpc/addr.h>
43#include <linux/nfs_ssc.h>
44
45#include "idmap.h"
46#include "cache.h"
47#include "xdr4.h"
48#include "vfs.h"
49#include "current_stateid.h"
50#include "netns.h"
51#include "acl.h"
52#include "pnfs.h"
53#include "trace.h"
54
55static bool inter_copy_offload_enable;
56module_param(inter_copy_offload_enable, bool, 0644);
57MODULE_PARM_DESC(inter_copy_offload_enable,
58 "Enable inter server to server copy offload. Default: false");
59
60#ifdef CONFIG_NFSD_V4_2_INTER_SSC
61static int nfsd4_ssc_umount_timeout = 900000; /* default to 15 mins */
62module_param(nfsd4_ssc_umount_timeout, int, 0644);
63MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
64 "idle msecs before unmount export from source server");
65#endif
66
67#define NFSDDBG_FACILITY NFSDDBG_PROC
68
69static u32 nfsd_attrmask[] = {
70 NFSD_WRITEABLE_ATTRS_WORD0,
71 NFSD_WRITEABLE_ATTRS_WORD1,
72 NFSD_WRITEABLE_ATTRS_WORD2
73};
74
75static u32 nfsd41_ex_attrmask[] = {
76 NFSD_SUPPATTR_EXCLCREAT_WORD0,
77 NFSD_SUPPATTR_EXCLCREAT_WORD1,
78 NFSD_SUPPATTR_EXCLCREAT_WORD2
79};
80
81static __be32
82check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
83 u32 *bmval, u32 *writable)
84{
85 struct dentry *dentry = cstate->current_fh.fh_dentry;
86 struct svc_export *exp = cstate->current_fh.fh_export;
87
88 if (!nfsd_attrs_supported(cstate->minorversion, bmval))
89 return nfserr_attrnotsupp;
90 if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
91 return nfserr_attrnotsupp;
92 if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
93 !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
94 return nfserr_attrnotsupp;
95 if (writable && !bmval_is_subset(bmval, writable))
96 return nfserr_inval;
97 if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
98 (bmval[1] & FATTR4_WORD1_MODE))
99 return nfserr_inval;
100 return nfs_ok;
101}
102
103static __be32
104nfsd4_check_open_attributes(struct svc_rqst *rqstp,
105 struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
106{
107 __be32 status = nfs_ok;
108
109 if (open->op_create == NFS4_OPEN_CREATE) {
110 if (open->op_createmode == NFS4_CREATE_UNCHECKED
111 || open->op_createmode == NFS4_CREATE_GUARDED)
112 status = check_attr_support(rqstp, cstate,
113 open->op_bmval, nfsd_attrmask);
114 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
115 status = check_attr_support(rqstp, cstate,
116 open->op_bmval, nfsd41_ex_attrmask);
117 }
118
119 return status;
120}
121
122static int
123is_create_with_attrs(struct nfsd4_open *open)
124{
125 return open->op_create == NFS4_OPEN_CREATE
126 && (open->op_createmode == NFS4_CREATE_UNCHECKED
127 || open->op_createmode == NFS4_CREATE_GUARDED
128 || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
129}
130
131static inline void
132fh_dup2(struct svc_fh *dst, struct svc_fh *src)
133{
134 fh_put(dst);
135 dget(src->fh_dentry);
136 if (src->fh_export)
137 exp_get(src->fh_export);
138 *dst = *src;
139}
140
141static __be32
142do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
143{
144
145 if (open->op_truncate &&
146 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
147 return nfserr_inval;
148
149 accmode |= NFSD_MAY_READ_IF_EXEC;
150
151 if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
152 accmode |= NFSD_MAY_READ;
153 if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
154 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
155 if (open->op_share_deny & NFS4_SHARE_DENY_READ)
156 accmode |= NFSD_MAY_WRITE;
157
158 return fh_verify(rqstp, current_fh, S_IFREG, accmode);
159}
160
161static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
162{
163 umode_t mode = d_inode(fh->fh_dentry)->i_mode;
164
165 if (S_ISREG(mode))
166 return nfs_ok;
167 if (S_ISDIR(mode))
168 return nfserr_isdir;
169 /*
170 * Using err_symlink as our catch-all case may look odd; but
171 * there's no other obvious error for this case in 4.0, and we
172 * happen to know that it will cause the linux v4 client to do
173 * the right thing on attempts to open something other than a
174 * regular file.
175 */
176 return nfserr_symlink;
177}
178
179static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
180{
181 if (nfsd4_has_session(cstate))
182 return;
183 fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
184 &resfh->fh_handle);
185}
186
187static inline bool nfsd4_create_is_exclusive(int createmode)
188{
189 return createmode == NFS4_CREATE_EXCLUSIVE ||
190 createmode == NFS4_CREATE_EXCLUSIVE4_1;
191}
192
193static __be32
194nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
195 struct nfsd4_open *open)
196{
197 struct file *filp;
198 struct path path;
199 int oflags;
200
201 oflags = O_CREAT | O_LARGEFILE;
202 switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
203 case NFS4_SHARE_ACCESS_WRITE:
204 oflags |= O_WRONLY;
205 break;
206 case NFS4_SHARE_ACCESS_BOTH:
207 oflags |= O_RDWR;
208 break;
209 default:
210 oflags |= O_RDONLY;
211 }
212
213 path.mnt = fhp->fh_export->ex_path.mnt;
214 path.dentry = child;
215 filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
216 current_cred());
217 if (IS_ERR(filp))
218 return nfserrno(PTR_ERR(filp));
219
220 open->op_filp = filp;
221 return nfs_ok;
222}
223
224/*
225 * Implement NFSv4's unchecked, guarded, and exclusive create
226 * semantics for regular files. Open state for this new file is
227 * subsequently fabricated in nfsd4_process_open2().
228 *
229 * Upon return, caller must release @fhp and @resfhp.
230 */
231static __be32
232nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
233 struct svc_fh *resfhp, struct nfsd4_open *open)
234{
235 struct iattr *iap = &open->op_iattr;
236 struct nfsd_attrs attrs = {
237 .na_iattr = iap,
238 .na_seclabel = &open->op_label,
239 };
240 struct dentry *parent, *child;
241 __u32 v_mtime, v_atime;
242 struct inode *inode;
243 __be32 status;
244 int host_err;
245
246 if (isdotent(open->op_fname, open->op_fnamelen))
247 return nfserr_exist;
248 if (!(iap->ia_valid & ATTR_MODE))
249 iap->ia_mode = 0;
250
251 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
252 if (status != nfs_ok)
253 return status;
254 parent = fhp->fh_dentry;
255 inode = d_inode(parent);
256
257 host_err = fh_want_write(fhp);
258 if (host_err)
259 return nfserrno(host_err);
260
261 if (is_create_with_attrs(open))
262 nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
263
264 inode_lock_nested(inode, I_MUTEX_PARENT);
265
266 child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
267 if (IS_ERR(child)) {
268 status = nfserrno(PTR_ERR(child));
269 goto out;
270 }
271
272 if (d_really_is_negative(child)) {
273 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
274 if (status != nfs_ok)
275 goto out;
276 }
277
278 status = fh_compose(resfhp, fhp->fh_export, child, fhp);
279 if (status != nfs_ok)
280 goto out;
281
282 v_mtime = 0;
283 v_atime = 0;
284 if (nfsd4_create_is_exclusive(open->op_createmode)) {
285 u32 *verifier = (u32 *)open->op_verf.data;
286
287 /*
288 * Solaris 7 gets confused (bugid 4218508) if these have
289 * the high bit set, as do xfs filesystems without the
290 * "bigtime" feature. So just clear the high bits. If this
291 * is ever changed to use different attrs for storing the
292 * verifier, then do_open_lookup() will also need to be
293 * fixed accordingly.
294 */
295 v_mtime = verifier[0] & 0x7fffffff;
296 v_atime = verifier[1] & 0x7fffffff;
297 }
298
299 if (d_really_is_positive(child)) {
300 /* NFSv4 protocol requires change attributes even though
301 * no change happened.
302 */
303 status = fh_fill_both_attrs(fhp);
304 if (status != nfs_ok)
305 goto out;
306
307 switch (open->op_createmode) {
308 case NFS4_CREATE_UNCHECKED:
309 if (!d_is_reg(child))
310 break;
311
312 /*
313 * In NFSv4, we don't want to truncate the file
314 * now. This would be wrong if the OPEN fails for
315 * some other reason. Furthermore, if the size is
316 * nonzero, we should ignore it according to spec!
317 */
318 open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
319 !iap->ia_size;
320 break;
321 case NFS4_CREATE_GUARDED:
322 status = nfserr_exist;
323 break;
324 case NFS4_CREATE_EXCLUSIVE:
325 if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
326 inode_get_atime_sec(d_inode(child)) == v_atime &&
327 d_inode(child)->i_size == 0) {
328 open->op_created = true;
329 break; /* subtle */
330 }
331 status = nfserr_exist;
332 break;
333 case NFS4_CREATE_EXCLUSIVE4_1:
334 if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
335 inode_get_atime_sec(d_inode(child)) == v_atime &&
336 d_inode(child)->i_size == 0) {
337 open->op_created = true;
338 goto set_attr; /* subtle */
339 }
340 status = nfserr_exist;
341 }
342 goto out;
343 }
344
345 if (!IS_POSIXACL(inode))
346 iap->ia_mode &= ~current_umask();
347
348 status = fh_fill_pre_attrs(fhp);
349 if (status != nfs_ok)
350 goto out;
351 status = nfsd4_vfs_create(fhp, child, open);
352 if (status != nfs_ok)
353 goto out;
354 open->op_created = true;
355 fh_fill_post_attrs(fhp);
356
357 /* A newly created file already has a file size of zero. */
358 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
359 iap->ia_valid &= ~ATTR_SIZE;
360 if (nfsd4_create_is_exclusive(open->op_createmode)) {
361 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
362 ATTR_MTIME_SET|ATTR_ATIME_SET;
363 iap->ia_mtime.tv_sec = v_mtime;
364 iap->ia_atime.tv_sec = v_atime;
365 iap->ia_mtime.tv_nsec = 0;
366 iap->ia_atime.tv_nsec = 0;
367 }
368
369set_attr:
370 status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
371
372 if (attrs.na_labelerr)
373 open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
374 if (attrs.na_aclerr)
375 open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
376out:
377 inode_unlock(inode);
378 nfsd_attrs_free(&attrs);
379 if (child && !IS_ERR(child))
380 dput(child);
381 fh_drop_write(fhp);
382 return status;
383}
384
385/**
386 * set_change_info - set up the change_info4 for a reply
387 * @cinfo: pointer to nfsd4_change_info to be populated
388 * @fhp: pointer to svc_fh to use as source
389 *
390 * Many operations in NFSv4 require change_info4 in the reply. This function
391 * populates that from the info that we (should!) have already collected. In
392 * the event that we didn't get any pre-attrs, just zero out both.
393 */
394static void
395set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp)
396{
397 cinfo->atomic = (u32)(fhp->fh_pre_saved && fhp->fh_post_saved && !fhp->fh_no_atomic_attr);
398 cinfo->before_change = fhp->fh_pre_change;
399 cinfo->after_change = fhp->fh_post_change;
400
401 /*
402 * If fetching the pre-change attributes failed, then we should
403 * have already failed the whole operation. We could have still
404 * failed to fetch post-change attributes however.
405 *
406 * If we didn't get post-op attrs, just zero-out the after
407 * field since we don't know what it should be. If the pre_saved
408 * field isn't set for some reason, throw warning and just copy
409 * whatever is in the after field.
410 */
411 if (WARN_ON_ONCE(!fhp->fh_pre_saved))
412 cinfo->before_change = 0;
413 if (!fhp->fh_post_saved)
414 cinfo->after_change = cinfo->before_change + 1;
415}
416
417static __be32
418do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
419{
420 struct svc_fh *current_fh = &cstate->current_fh;
421 int accmode;
422 __be32 status;
423
424 *resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
425 if (!*resfh)
426 return nfserr_jukebox;
427 fh_init(*resfh, NFS4_FHSIZE);
428 open->op_truncate = false;
429
430 if (open->op_create) {
431 /* FIXME: check session persistence and pnfs flags.
432 * The nfsv4.1 spec requires the following semantics:
433 *
434 * Persistent | pNFS | Server REQUIRED | Client Allowed
435 * Reply Cache | server | |
436 * -------------+--------+-----------------+--------------------
437 * no | no | EXCLUSIVE4_1 | EXCLUSIVE4_1
438 * | | | (SHOULD)
439 * | | and EXCLUSIVE4 | or EXCLUSIVE4
440 * | | | (SHOULD NOT)
441 * no | yes | EXCLUSIVE4_1 | EXCLUSIVE4_1
442 * yes | no | GUARDED4 | GUARDED4
443 * yes | yes | GUARDED4 | GUARDED4
444 */
445
446 current->fs->umask = open->op_umask;
447 status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
448 current->fs->umask = 0;
449
450 /*
451 * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
452 * use the returned bitmask to indicate which attributes
453 * we used to store the verifier:
454 */
455 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
456 open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
457 FATTR4_WORD1_TIME_MODIFY);
458 } else {
459 status = nfsd_lookup(rqstp, current_fh,
460 open->op_fname, open->op_fnamelen, *resfh);
461 if (status == nfs_ok)
462 /* NFSv4 protocol requires change attributes even though
463 * no change happened.
464 */
465 status = fh_fill_both_attrs(current_fh);
466 }
467 if (status)
468 goto out;
469 status = nfsd_check_obj_isreg(*resfh);
470 if (status)
471 goto out;
472
473 nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
474 accmode = NFSD_MAY_NOP;
475 if (open->op_created ||
476 open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
477 accmode |= NFSD_MAY_OWNER_OVERRIDE;
478 status = do_open_permission(rqstp, *resfh, open, accmode);
479 set_change_info(&open->op_cinfo, current_fh);
480out:
481 return status;
482}
483
484static __be32
485do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
486{
487 struct svc_fh *current_fh = &cstate->current_fh;
488 int accmode = 0;
489
490 /* We don't know the target directory, and therefore can not
491 * set the change info
492 */
493
494 memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
495
496 nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
497
498 open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
499 (open->op_iattr.ia_size == 0);
500 /*
501 * In the delegation case, the client is telling us about an
502 * open that it *already* performed locally, some time ago. We
503 * should let it succeed now if possible.
504 *
505 * In the case of a CLAIM_FH open, on the other hand, the client
506 * may be counting on us to enforce permissions (the Linux 4.1
507 * client uses this for normal opens, for example).
508 */
509 if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
510 accmode = NFSD_MAY_OWNER_OVERRIDE;
511
512 return do_open_permission(rqstp, current_fh, open, accmode);
513}
514
515static void
516copy_clientid(clientid_t *clid, struct nfsd4_session *session)
517{
518 struct nfsd4_sessionid *sid =
519 (struct nfsd4_sessionid *)session->se_sessionid.data;
520
521 clid->cl_boot = sid->clientid.cl_boot;
522 clid->cl_id = sid->clientid.cl_id;
523}
524
525static __be32
526nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
527 union nfsd4_op_u *u)
528{
529 struct nfsd4_open *open = &u->open;
530 __be32 status;
531 struct svc_fh *resfh = NULL;
532 struct net *net = SVC_NET(rqstp);
533 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
534 bool reclaim = false;
535
536 dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
537 (int)open->op_fnamelen, open->op_fname,
538 open->op_openowner);
539
540 open->op_filp = NULL;
541 open->op_rqstp = rqstp;
542
543 /* This check required by spec. */
544 if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
545 return nfserr_inval;
546
547 open->op_created = false;
548 /*
549 * RFC5661 18.51.3
550 * Before RECLAIM_COMPLETE done, server should deny new lock
551 */
552 if (nfsd4_has_session(cstate) &&
553 !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
554 open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
555 return nfserr_grace;
556
557 if (nfsd4_has_session(cstate))
558 copy_clientid(&open->op_clientid, cstate->session);
559
560 /* check seqid for replay. set nfs4_owner */
561 status = nfsd4_process_open1(cstate, open, nn);
562 if (status == nfserr_replay_me) {
563 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
564 fh_put(&cstate->current_fh);
565 fh_copy_shallow(&cstate->current_fh.fh_handle,
566 &rp->rp_openfh);
567 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
568 if (status)
569 dprintk("nfsd4_open: replay failed"
570 " restoring previous filehandle\n");
571 else
572 status = nfserr_replay_me;
573 }
574 if (status)
575 goto out;
576 if (open->op_xdr_error) {
577 status = open->op_xdr_error;
578 goto out;
579 }
580
581 status = nfsd4_check_open_attributes(rqstp, cstate, open);
582 if (status)
583 goto out;
584
585 /* Openowner is now set, so sequence id will get bumped. Now we need
586 * these checks before we do any creates: */
587 status = nfserr_grace;
588 if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
589 goto out;
590 status = nfserr_no_grace;
591 if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
592 goto out;
593
594 switch (open->op_claim_type) {
595 case NFS4_OPEN_CLAIM_DELEGATE_CUR:
596 case NFS4_OPEN_CLAIM_NULL:
597 status = do_open_lookup(rqstp, cstate, open, &resfh);
598 if (status)
599 goto out;
600 break;
601 case NFS4_OPEN_CLAIM_PREVIOUS:
602 status = nfs4_check_open_reclaim(cstate->clp);
603 if (status)
604 goto out;
605 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
606 reclaim = true;
607 fallthrough;
608 case NFS4_OPEN_CLAIM_FH:
609 case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
610 status = do_open_fhandle(rqstp, cstate, open);
611 if (status)
612 goto out;
613 resfh = &cstate->current_fh;
614 break;
615 case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
616 case NFS4_OPEN_CLAIM_DELEGATE_PREV:
617 status = nfserr_notsupp;
618 goto out;
619 default:
620 status = nfserr_inval;
621 goto out;
622 }
623
624 status = nfsd4_process_open2(rqstp, resfh, open);
625 if (status && open->op_created)
626 pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
627 be32_to_cpu(status));
628 if (reclaim && !status)
629 nn->somebody_reclaimed = true;
630out:
631 if (open->op_filp) {
632 fput(open->op_filp);
633 open->op_filp = NULL;
634 }
635 if (resfh && resfh != &cstate->current_fh) {
636 fh_dup2(&cstate->current_fh, resfh);
637 fh_put(resfh);
638 kfree(resfh);
639 }
640 nfsd4_cleanup_open_state(cstate, open);
641 nfsd4_bump_seqid(cstate, status);
642 return status;
643}
644
645/*
646 * OPEN is the only seqid-mutating operation whose decoding can fail
647 * with a seqid-mutating error (specifically, decoding of user names in
648 * the attributes). Therefore we have to do some processing to look up
649 * the stateowner so that we can bump the seqid.
650 */
651static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
652{
653 struct nfsd4_open *open = &op->u.open;
654
655 if (!seqid_mutating_err(ntohl(op->status)))
656 return op->status;
657 if (nfsd4_has_session(cstate))
658 return op->status;
659 open->op_xdr_error = op->status;
660 return nfsd4_open(rqstp, cstate, &op->u);
661}
662
663/*
664 * filehandle-manipulating ops.
665 */
666static __be32
667nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
668 union nfsd4_op_u *u)
669{
670 u->getfh = &cstate->current_fh;
671 return nfs_ok;
672}
673
674static __be32
675nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
676 union nfsd4_op_u *u)
677{
678 struct nfsd4_putfh *putfh = &u->putfh;
679 __be32 ret;
680
681 fh_put(&cstate->current_fh);
682 cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
683 memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
684 putfh->pf_fhlen);
685 ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
686#ifdef CONFIG_NFSD_V4_2_INTER_SSC
687 if (ret == nfserr_stale && putfh->no_verify) {
688 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
689 ret = 0;
690 }
691#endif
692 return ret;
693}
694
695static __be32
696nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
697 union nfsd4_op_u *u)
698{
699 fh_put(&cstate->current_fh);
700
701 return exp_pseudoroot(rqstp, &cstate->current_fh);
702}
703
704static __be32
705nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
706 union nfsd4_op_u *u)
707{
708 if (!cstate->save_fh.fh_dentry)
709 return nfserr_restorefh;
710
711 fh_dup2(&cstate->current_fh, &cstate->save_fh);
712 if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
713 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
714 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
715 }
716 return nfs_ok;
717}
718
719static __be32
720nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
721 union nfsd4_op_u *u)
722{
723 fh_dup2(&cstate->save_fh, &cstate->current_fh);
724 if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
725 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
726 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
727 }
728 return nfs_ok;
729}
730
731/*
732 * misc nfsv4 ops
733 */
734static __be32
735nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
736 union nfsd4_op_u *u)
737{
738 struct nfsd4_access *access = &u->access;
739 u32 access_full;
740
741 access_full = NFS3_ACCESS_FULL;
742 if (cstate->minorversion >= 2)
743 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
744 NFS4_ACCESS_XAWRITE;
745
746 if (access->ac_req_access & ~access_full)
747 return nfserr_inval;
748
749 access->ac_resp_access = access->ac_req_access;
750 return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
751 &access->ac_supported);
752}
753
754static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
755{
756 __be32 *verf = (__be32 *)verifier->data;
757
758 BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
759
760 nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
761}
762
763static __be32
764nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
765 union nfsd4_op_u *u)
766{
767 struct nfsd4_commit *commit = &u->commit;
768 struct nfsd_file *nf;
769 __be32 status;
770
771 status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
772 NFSD_MAY_NOT_BREAK_LEASE, &nf);
773 if (status != nfs_ok)
774 return status;
775
776 status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
777 commit->co_count,
778 (__be32 *)commit->co_verf.data);
779 nfsd_file_put(nf);
780 return status;
781}
782
783static __be32
784nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
785 union nfsd4_op_u *u)
786{
787 struct nfsd4_create *create = &u->create;
788 struct nfsd_attrs attrs = {
789 .na_iattr = &create->cr_iattr,
790 .na_seclabel = &create->cr_label,
791 };
792 struct svc_fh resfh;
793 __be32 status;
794 dev_t rdev;
795
796 fh_init(&resfh, NFS4_FHSIZE);
797
798 status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
799 if (status)
800 return status;
801
802 status = check_attr_support(rqstp, cstate, create->cr_bmval,
803 nfsd_attrmask);
804 if (status)
805 return status;
806
807 status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
808 current->fs->umask = create->cr_umask;
809 switch (create->cr_type) {
810 case NF4LNK:
811 status = nfsd_symlink(rqstp, &cstate->current_fh,
812 create->cr_name, create->cr_namelen,
813 create->cr_data, &attrs, &resfh);
814 break;
815
816 case NF4BLK:
817 status = nfserr_inval;
818 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
819 if (MAJOR(rdev) != create->cr_specdata1 ||
820 MINOR(rdev) != create->cr_specdata2)
821 goto out_umask;
822 status = nfsd_create(rqstp, &cstate->current_fh,
823 create->cr_name, create->cr_namelen,
824 &attrs, S_IFBLK, rdev, &resfh);
825 break;
826
827 case NF4CHR:
828 status = nfserr_inval;
829 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
830 if (MAJOR(rdev) != create->cr_specdata1 ||
831 MINOR(rdev) != create->cr_specdata2)
832 goto out_umask;
833 status = nfsd_create(rqstp, &cstate->current_fh,
834 create->cr_name, create->cr_namelen,
835 &attrs, S_IFCHR, rdev, &resfh);
836 break;
837
838 case NF4SOCK:
839 status = nfsd_create(rqstp, &cstate->current_fh,
840 create->cr_name, create->cr_namelen,
841 &attrs, S_IFSOCK, 0, &resfh);
842 break;
843
844 case NF4FIFO:
845 status = nfsd_create(rqstp, &cstate->current_fh,
846 create->cr_name, create->cr_namelen,
847 &attrs, S_IFIFO, 0, &resfh);
848 break;
849
850 case NF4DIR:
851 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
852 status = nfsd_create(rqstp, &cstate->current_fh,
853 create->cr_name, create->cr_namelen,
854 &attrs, S_IFDIR, 0, &resfh);
855 break;
856
857 default:
858 status = nfserr_badtype;
859 }
860
861 if (status)
862 goto out;
863
864 if (attrs.na_labelerr)
865 create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
866 if (attrs.na_aclerr)
867 create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
868 set_change_info(&create->cr_cinfo, &cstate->current_fh);
869 fh_dup2(&cstate->current_fh, &resfh);
870out:
871 fh_put(&resfh);
872out_umask:
873 current->fs->umask = 0;
874 nfsd_attrs_free(&attrs);
875 return status;
876}
877
878static __be32
879nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
880 union nfsd4_op_u *u)
881{
882 struct nfsd4_getattr *getattr = &u->getattr;
883 __be32 status;
884
885 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
886 if (status)
887 return status;
888
889 if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
890 return nfserr_inval;
891
892 getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
893 getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
894 getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
895
896 getattr->ga_fhp = &cstate->current_fh;
897 return nfs_ok;
898}
899
900static __be32
901nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
902 union nfsd4_op_u *u)
903{
904 struct nfsd4_link *link = &u->link;
905 __be32 status;
906
907 status = nfsd_link(rqstp, &cstate->current_fh,
908 link->li_name, link->li_namelen, &cstate->save_fh);
909 if (!status)
910 set_change_info(&link->li_cinfo, &cstate->current_fh);
911 return status;
912}
913
914static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
915{
916 struct svc_fh tmp_fh;
917 __be32 ret;
918
919 fh_init(&tmp_fh, NFS4_FHSIZE);
920 ret = exp_pseudoroot(rqstp, &tmp_fh);
921 if (ret)
922 return ret;
923 if (tmp_fh.fh_dentry == fh->fh_dentry) {
924 fh_put(&tmp_fh);
925 return nfserr_noent;
926 }
927 fh_put(&tmp_fh);
928 return nfsd_lookup(rqstp, fh, "..", 2, fh);
929}
930
931static __be32
932nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
933 union nfsd4_op_u *u)
934{
935 return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
936}
937
938static __be32
939nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
940 union nfsd4_op_u *u)
941{
942 return nfsd_lookup(rqstp, &cstate->current_fh,
943 u->lookup.lo_name, u->lookup.lo_len,
944 &cstate->current_fh);
945}
946
947static __be32
948nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
949 union nfsd4_op_u *u)
950{
951 struct nfsd4_read *read = &u->read;
952 __be32 status;
953
954 read->rd_nf = NULL;
955
956 trace_nfsd_read_start(rqstp, &cstate->current_fh,
957 read->rd_offset, read->rd_length);
958
959 read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
960 if (read->rd_offset > (u64)OFFSET_MAX)
961 read->rd_offset = (u64)OFFSET_MAX;
962 if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
963 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
964
965 /*
966 * If we do a zero copy read, then a client will see read data
967 * that reflects the state of the file *after* performing the
968 * following compound.
969 *
970 * To ensure proper ordering, we therefore turn off zero copy if
971 * the client wants us to do more in this compound:
972 */
973 if (!nfsd4_last_compound_op(rqstp)) {
974 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
975
976 argp->splice_ok = false;
977 }
978
979 /* check stateid */
980 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
981 &read->rd_stateid, RD_STATE,
982 &read->rd_nf, NULL);
983
984 read->rd_rqstp = rqstp;
985 read->rd_fhp = &cstate->current_fh;
986 return status;
987}
988
989
990static void
991nfsd4_read_release(union nfsd4_op_u *u)
992{
993 if (u->read.rd_nf)
994 nfsd_file_put(u->read.rd_nf);
995 trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
996 u->read.rd_offset, u->read.rd_length);
997}
998
999static __be32
1000nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1001 union nfsd4_op_u *u)
1002{
1003 struct nfsd4_readdir *readdir = &u->readdir;
1004 u64 cookie = readdir->rd_cookie;
1005 static const nfs4_verifier zeroverf;
1006
1007 /* no need to check permission - this will be done in nfsd_readdir() */
1008
1009 if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1010 return nfserr_inval;
1011
1012 readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1013 readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1014 readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1015
1016 if ((cookie == 1) || (cookie == 2) ||
1017 (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1018 return nfserr_bad_cookie;
1019
1020 readdir->rd_rqstp = rqstp;
1021 readdir->rd_fhp = &cstate->current_fh;
1022 return nfs_ok;
1023}
1024
1025static __be32
1026nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1027 union nfsd4_op_u *u)
1028{
1029 u->readlink.rl_rqstp = rqstp;
1030 u->readlink.rl_fhp = &cstate->current_fh;
1031 return nfs_ok;
1032}
1033
1034static __be32
1035nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1036 union nfsd4_op_u *u)
1037{
1038 struct nfsd4_remove *remove = &u->remove;
1039 __be32 status;
1040
1041 if (opens_in_grace(SVC_NET(rqstp)))
1042 return nfserr_grace;
1043 status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1044 remove->rm_name, remove->rm_namelen);
1045 if (!status)
1046 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1047 return status;
1048}
1049
1050static __be32
1051nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1052 union nfsd4_op_u *u)
1053{
1054 struct nfsd4_rename *rename = &u->rename;
1055 __be32 status;
1056
1057 if (opens_in_grace(SVC_NET(rqstp)))
1058 return nfserr_grace;
1059 status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1060 rename->rn_snamelen, &cstate->current_fh,
1061 rename->rn_tname, rename->rn_tnamelen);
1062 if (status)
1063 return status;
1064 set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1065 set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1066 return nfs_ok;
1067}
1068
1069static __be32
1070nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1071 union nfsd4_op_u *u)
1072{
1073 struct nfsd4_secinfo *secinfo = &u->secinfo;
1074 struct svc_export *exp;
1075 struct dentry *dentry;
1076 __be32 err;
1077
1078 err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1079 if (err)
1080 return err;
1081 err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1082 secinfo->si_name, secinfo->si_namelen,
1083 &exp, &dentry);
1084 if (err)
1085 return err;
1086 if (d_really_is_negative(dentry)) {
1087 exp_put(exp);
1088 err = nfserr_noent;
1089 } else
1090 secinfo->si_exp = exp;
1091 dput(dentry);
1092 if (cstate->minorversion)
1093 /* See rfc 5661 section 2.6.3.1.1.8 */
1094 fh_put(&cstate->current_fh);
1095 return err;
1096}
1097
1098static __be32
1099nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1100 union nfsd4_op_u *u)
1101{
1102 __be32 err;
1103
1104 switch (u->secinfo_no_name.sin_style) {
1105 case NFS4_SECINFO_STYLE4_CURRENT_FH:
1106 break;
1107 case NFS4_SECINFO_STYLE4_PARENT:
1108 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1109 if (err)
1110 return err;
1111 break;
1112 default:
1113 return nfserr_inval;
1114 }
1115
1116 u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1117 fh_put(&cstate->current_fh);
1118 return nfs_ok;
1119}
1120
1121static void
1122nfsd4_secinfo_release(union nfsd4_op_u *u)
1123{
1124 if (u->secinfo.si_exp)
1125 exp_put(u->secinfo.si_exp);
1126}
1127
1128static void
1129nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1130{
1131 if (u->secinfo_no_name.sin_exp)
1132 exp_put(u->secinfo_no_name.sin_exp);
1133}
1134
1135static __be32
1136nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1137 union nfsd4_op_u *u)
1138{
1139 struct nfsd4_setattr *setattr = &u->setattr;
1140 struct nfsd_attrs attrs = {
1141 .na_iattr = &setattr->sa_iattr,
1142 .na_seclabel = &setattr->sa_label,
1143 };
1144 struct inode *inode;
1145 __be32 status = nfs_ok;
1146 bool save_no_wcc;
1147 int err;
1148
1149 if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1150 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1151 &cstate->current_fh, &setattr->sa_stateid,
1152 WR_STATE, NULL, NULL);
1153 if (status)
1154 return status;
1155 }
1156 err = fh_want_write(&cstate->current_fh);
1157 if (err)
1158 return nfserrno(err);
1159 status = nfs_ok;
1160
1161 status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1162 nfsd_attrmask);
1163 if (status)
1164 goto out;
1165
1166 inode = cstate->current_fh.fh_dentry->d_inode;
1167 status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1168 setattr->sa_acl, &attrs);
1169
1170 if (status)
1171 goto out;
1172 save_no_wcc = cstate->current_fh.fh_no_wcc;
1173 cstate->current_fh.fh_no_wcc = true;
1174 status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs, NULL);
1175 cstate->current_fh.fh_no_wcc = save_no_wcc;
1176 if (!status)
1177 status = nfserrno(attrs.na_labelerr);
1178 if (!status)
1179 status = nfserrno(attrs.na_aclerr);
1180out:
1181 nfsd_attrs_free(&attrs);
1182 fh_drop_write(&cstate->current_fh);
1183 return status;
1184}
1185
1186static __be32
1187nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1188 union nfsd4_op_u *u)
1189{
1190 struct nfsd4_write *write = &u->write;
1191 stateid_t *stateid = &write->wr_stateid;
1192 struct nfsd_file *nf = NULL;
1193 __be32 status = nfs_ok;
1194 unsigned long cnt;
1195 int nvecs;
1196
1197 if (write->wr_offset > (u64)OFFSET_MAX ||
1198 write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1199 return nfserr_fbig;
1200
1201 cnt = write->wr_buflen;
1202 trace_nfsd_write_start(rqstp, &cstate->current_fh,
1203 write->wr_offset, cnt);
1204 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1205 stateid, WR_STATE, &nf, NULL);
1206 if (status)
1207 return status;
1208
1209 write->wr_how_written = write->wr_stable_how;
1210
1211 nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1212 WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1213
1214 status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1215 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1216 write->wr_how_written,
1217 (__be32 *)write->wr_verifier.data);
1218 nfsd_file_put(nf);
1219
1220 write->wr_bytes_written = cnt;
1221 trace_nfsd_write_done(rqstp, &cstate->current_fh,
1222 write->wr_offset, cnt);
1223 return status;
1224}
1225
1226static __be32
1227nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1228 stateid_t *src_stateid, struct nfsd_file **src,
1229 stateid_t *dst_stateid, struct nfsd_file **dst)
1230{
1231 __be32 status;
1232
1233 if (!cstate->save_fh.fh_dentry)
1234 return nfserr_nofilehandle;
1235
1236 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1237 src_stateid, RD_STATE, src, NULL);
1238 if (status)
1239 goto out;
1240
1241 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1242 dst_stateid, WR_STATE, dst, NULL);
1243 if (status)
1244 goto out_put_src;
1245
1246 /* fix up for NFS-specific error code */
1247 if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1248 !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1249 status = nfserr_wrong_type;
1250 goto out_put_dst;
1251 }
1252
1253out:
1254 return status;
1255out_put_dst:
1256 nfsd_file_put(*dst);
1257 *dst = NULL;
1258out_put_src:
1259 nfsd_file_put(*src);
1260 *src = NULL;
1261 goto out;
1262}
1263
1264static __be32
1265nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1266 union nfsd4_op_u *u)
1267{
1268 struct nfsd4_clone *clone = &u->clone;
1269 struct nfsd_file *src, *dst;
1270 __be32 status;
1271
1272 status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1273 &clone->cl_dst_stateid, &dst);
1274 if (status)
1275 goto out;
1276
1277 status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1278 dst, clone->cl_dst_pos, clone->cl_count,
1279 EX_ISSYNC(cstate->current_fh.fh_export));
1280
1281 nfsd_file_put(dst);
1282 nfsd_file_put(src);
1283out:
1284 return status;
1285}
1286
1287static void nfs4_put_copy(struct nfsd4_copy *copy)
1288{
1289 if (!refcount_dec_and_test(©->refcount))
1290 return;
1291 kfree(copy->cp_src);
1292 kfree(copy);
1293}
1294
1295static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1296{
1297 if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, ©->cp_flags))
1298 kthread_stop(copy->copy_task);
1299 nfs4_put_copy(copy);
1300}
1301
1302static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1303{
1304 struct nfsd4_copy *copy = NULL;
1305
1306 spin_lock(&clp->async_lock);
1307 if (!list_empty(&clp->async_copies)) {
1308 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1309 copies);
1310 refcount_inc(©->refcount);
1311 }
1312 spin_unlock(&clp->async_lock);
1313 return copy;
1314}
1315
1316void nfsd4_shutdown_copy(struct nfs4_client *clp)
1317{
1318 struct nfsd4_copy *copy;
1319
1320 while ((copy = nfsd4_get_copy(clp)) != NULL)
1321 nfsd4_stop_copy(copy);
1322}
1323#ifdef CONFIG_NFSD_V4_2_INTER_SSC
1324
1325extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1326 struct nfs_fh *src_fh,
1327 nfs4_stateid *stateid);
1328extern void nfs42_ssc_close(struct file *filep);
1329
1330extern void nfs_sb_deactive(struct super_block *sb);
1331
1332#define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1333
1334/*
1335 * setup a work entry in the ssc delayed unmount list.
1336 */
1337static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1338 struct nfsd4_ssc_umount_item **nsui,
1339 struct svc_rqst *rqstp)
1340{
1341 struct nfsd4_ssc_umount_item *ni = NULL;
1342 struct nfsd4_ssc_umount_item *work = NULL;
1343 struct nfsd4_ssc_umount_item *tmp;
1344 DEFINE_WAIT(wait);
1345 __be32 status = 0;
1346
1347 *nsui = NULL;
1348 work = kzalloc(sizeof(*work), GFP_KERNEL);
1349try_again:
1350 spin_lock(&nn->nfsd_ssc_lock);
1351 list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1352 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1353 continue;
1354 /* found a match */
1355 if (ni->nsui_busy) {
1356 /* wait - and try again */
1357 prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1358 spin_unlock(&nn->nfsd_ssc_lock);
1359
1360 /* allow 20secs for mount/unmount for now - revisit */
1361 if (svc_thread_should_stop(rqstp) ||
1362 (schedule_timeout(20*HZ) == 0)) {
1363 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1364 kfree(work);
1365 return nfserr_eagain;
1366 }
1367 finish_wait(&nn->nfsd_ssc_waitq, &wait);
1368 goto try_again;
1369 }
1370 *nsui = ni;
1371 refcount_inc(&ni->nsui_refcnt);
1372 spin_unlock(&nn->nfsd_ssc_lock);
1373 kfree(work);
1374
1375 /* return vfsmount in (*nsui)->nsui_vfsmount */
1376 return 0;
1377 }
1378 if (work) {
1379 strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1380 refcount_set(&work->nsui_refcnt, 2);
1381 work->nsui_busy = true;
1382 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1383 *nsui = work;
1384 } else
1385 status = nfserr_resource;
1386 spin_unlock(&nn->nfsd_ssc_lock);
1387 return status;
1388}
1389
1390static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1391 struct nfsd4_ssc_umount_item *nsui,
1392 struct vfsmount *ss_mnt)
1393{
1394 spin_lock(&nn->nfsd_ssc_lock);
1395 nsui->nsui_vfsmount = ss_mnt;
1396 nsui->nsui_busy = false;
1397 wake_up_all(&nn->nfsd_ssc_waitq);
1398 spin_unlock(&nn->nfsd_ssc_lock);
1399}
1400
1401static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1402 struct nfsd4_ssc_umount_item *nsui)
1403{
1404 spin_lock(&nn->nfsd_ssc_lock);
1405 list_del(&nsui->nsui_list);
1406 wake_up_all(&nn->nfsd_ssc_waitq);
1407 spin_unlock(&nn->nfsd_ssc_lock);
1408 kfree(nsui);
1409}
1410
1411/*
1412 * Support one copy source server for now.
1413 */
1414static __be32
1415nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1416 struct nfsd4_ssc_umount_item **nsui)
1417{
1418 struct file_system_type *type;
1419 struct vfsmount *ss_mnt;
1420 struct nfs42_netaddr *naddr;
1421 struct sockaddr_storage tmp_addr;
1422 size_t tmp_addrlen, match_netid_len = 3;
1423 char *startsep = "", *endsep = "", *match_netid = "tcp";
1424 char *ipaddr, *dev_name, *raw_data;
1425 int len, raw_len;
1426 __be32 status = nfserr_inval;
1427 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1428
1429 naddr = &nss->u.nl4_addr;
1430 tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1431 naddr->addr_len,
1432 (struct sockaddr *)&tmp_addr,
1433 sizeof(tmp_addr));
1434 *nsui = NULL;
1435 if (tmp_addrlen == 0)
1436 goto out_err;
1437
1438 if (tmp_addr.ss_family == AF_INET6) {
1439 startsep = "[";
1440 endsep = "]";
1441 match_netid = "tcp6";
1442 match_netid_len = 4;
1443 }
1444
1445 if (naddr->netid_len != match_netid_len ||
1446 strncmp(naddr->netid, match_netid, naddr->netid_len))
1447 goto out_err;
1448
1449 /* Construct the raw data for the vfs_kern_mount call */
1450 len = RPC_MAX_ADDRBUFLEN + 1;
1451 ipaddr = kzalloc(len, GFP_KERNEL);
1452 if (!ipaddr)
1453 goto out_err;
1454
1455 rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1456
1457 /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1458
1459 raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1460 raw_data = kzalloc(raw_len, GFP_KERNEL);
1461 if (!raw_data)
1462 goto out_free_ipaddr;
1463
1464 snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1465
1466 status = nfserr_nodev;
1467 type = get_fs_type("nfs");
1468 if (!type)
1469 goto out_free_rawdata;
1470
1471 /* Set the server:<export> for the vfs_kern_mount call */
1472 dev_name = kzalloc(len + 5, GFP_KERNEL);
1473 if (!dev_name)
1474 goto out_free_rawdata;
1475 snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1476
1477 status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui, rqstp);
1478 if (status)
1479 goto out_free_devname;
1480 if ((*nsui)->nsui_vfsmount)
1481 goto out_done;
1482
1483 /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1484 ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1485 module_put(type->owner);
1486 if (IS_ERR(ss_mnt)) {
1487 status = nfserr_nodev;
1488 nfsd4_ssc_cancel_dul(nn, *nsui);
1489 goto out_free_devname;
1490 }
1491 nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1492out_done:
1493 status = 0;
1494
1495out_free_devname:
1496 kfree(dev_name);
1497out_free_rawdata:
1498 kfree(raw_data);
1499out_free_ipaddr:
1500 kfree(ipaddr);
1501out_err:
1502 return status;
1503}
1504
1505/*
1506 * Verify COPY destination stateid.
1507 *
1508 * Connect to the source server with NFSv4.1.
1509 * Create the source struct file for nfsd_copy_range.
1510 * Called with COPY cstate:
1511 * SAVED_FH: source filehandle
1512 * CURRENT_FH: destination filehandle
1513 */
1514static __be32
1515nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1516 struct nfsd4_compound_state *cstate,
1517 struct nfsd4_copy *copy)
1518{
1519 struct svc_fh *s_fh = NULL;
1520 stateid_t *s_stid = ©->cp_src_stateid;
1521 __be32 status = nfserr_inval;
1522
1523 /* Verify the destination stateid and set dst struct file*/
1524 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1525 ©->cp_dst_stateid,
1526 WR_STATE, ©->nf_dst, NULL);
1527 if (status)
1528 goto out;
1529
1530 status = nfsd4_interssc_connect(copy->cp_src, rqstp, ©->ss_nsui);
1531 if (status)
1532 goto out;
1533
1534 s_fh = &cstate->save_fh;
1535
1536 copy->c_fh.size = s_fh->fh_handle.fh_size;
1537 memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1538 copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1539 memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1540 sizeof(stateid_opaque_t));
1541
1542 status = 0;
1543out:
1544 return status;
1545}
1546
1547static void
1548nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1549 struct nfsd_file *dst)
1550{
1551 struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1552 long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1553
1554 nfs42_ssc_close(filp);
1555 fput(filp);
1556
1557 spin_lock(&nn->nfsd_ssc_lock);
1558 list_del(&nsui->nsui_list);
1559 /*
1560 * vfsmount can be shared by multiple exports,
1561 * decrement refcnt. If the count drops to 1 it
1562 * will be unmounted when nsui_expire expires.
1563 */
1564 refcount_dec(&nsui->nsui_refcnt);
1565 nsui->nsui_expire = jiffies + timeout;
1566 list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1567 spin_unlock(&nn->nfsd_ssc_lock);
1568}
1569
1570#else /* CONFIG_NFSD_V4_2_INTER_SSC */
1571
1572static __be32
1573nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1574 struct nfsd4_compound_state *cstate,
1575 struct nfsd4_copy *copy)
1576{
1577 return nfserr_inval;
1578}
1579
1580static void
1581nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1582 struct nfsd_file *dst)
1583{
1584}
1585
1586static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1587 struct nfs_fh *src_fh,
1588 nfs4_stateid *stateid)
1589{
1590 return NULL;
1591}
1592#endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1593
1594static __be32
1595nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1596 struct nfsd4_compound_state *cstate,
1597 struct nfsd4_copy *copy)
1598{
1599 return nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid,
1600 ©->nf_src, ©->cp_dst_stateid,
1601 ©->nf_dst);
1602}
1603
1604static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1605{
1606 struct nfsd4_cb_offload *cbo =
1607 container_of(cb, struct nfsd4_cb_offload, co_cb);
1608
1609 kfree(cbo);
1610}
1611
1612static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1613 struct rpc_task *task)
1614{
1615 struct nfsd4_cb_offload *cbo =
1616 container_of(cb, struct nfsd4_cb_offload, co_cb);
1617
1618 trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1619 return 1;
1620}
1621
1622static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1623 .release = nfsd4_cb_offload_release,
1624 .done = nfsd4_cb_offload_done
1625};
1626
1627static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1628{
1629 copy->cp_res.wr_stable_how =
1630 test_bit(NFSD4_COPY_F_COMMITTED, ©->cp_flags) ?
1631 NFS_FILE_SYNC : NFS_UNSTABLE;
1632 nfsd4_copy_set_sync(copy, sync);
1633 gen_boot_verifier(©->cp_res.wr_verifier, copy->cp_clp->net);
1634}
1635
1636static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1637 struct file *dst,
1638 struct file *src)
1639{
1640 errseq_t since;
1641 ssize_t bytes_copied = 0;
1642 u64 bytes_total = copy->cp_count;
1643 u64 src_pos = copy->cp_src_pos;
1644 u64 dst_pos = copy->cp_dst_pos;
1645 int status;
1646 loff_t end;
1647
1648 /* See RFC 7862 p.67: */
1649 if (bytes_total == 0)
1650 bytes_total = ULLONG_MAX;
1651 do {
1652 /* Only async copies can be stopped here */
1653 if (kthread_should_stop())
1654 break;
1655 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1656 bytes_total);
1657 if (bytes_copied <= 0)
1658 break;
1659 bytes_total -= bytes_copied;
1660 copy->cp_res.wr_bytes_written += bytes_copied;
1661 src_pos += bytes_copied;
1662 dst_pos += bytes_copied;
1663 } while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1664 /* for a non-zero asynchronous copy do a commit of data */
1665 if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1666 since = READ_ONCE(dst->f_wb_err);
1667 end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1668 status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1669 if (!status)
1670 status = filemap_check_wb_err(dst->f_mapping, since);
1671 if (!status)
1672 set_bit(NFSD4_COPY_F_COMMITTED, ©->cp_flags);
1673 }
1674 return bytes_copied;
1675}
1676
1677static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1678 struct file *src, struct file *dst,
1679 bool sync)
1680{
1681 __be32 status;
1682 ssize_t bytes;
1683
1684 bytes = _nfsd_copy_file_range(copy, dst, src);
1685
1686 /* for async copy, we ignore the error, client can always retry
1687 * to get the error
1688 */
1689 if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1690 status = nfserrno(bytes);
1691 else {
1692 nfsd4_init_copy_res(copy, sync);
1693 status = nfs_ok;
1694 }
1695 return status;
1696}
1697
1698static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1699{
1700 dst->cp_src_pos = src->cp_src_pos;
1701 dst->cp_dst_pos = src->cp_dst_pos;
1702 dst->cp_count = src->cp_count;
1703 dst->cp_flags = src->cp_flags;
1704 memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1705 memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1706 dst->cp_clp = src->cp_clp;
1707 dst->nf_dst = nfsd_file_get(src->nf_dst);
1708 /* for inter, nf_src doesn't exist yet */
1709 if (!nfsd4_ssc_is_inter(src))
1710 dst->nf_src = nfsd_file_get(src->nf_src);
1711
1712 memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1713 memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1714 memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1715 memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1716 dst->ss_nsui = src->ss_nsui;
1717}
1718
1719static void release_copy_files(struct nfsd4_copy *copy)
1720{
1721 if (copy->nf_src)
1722 nfsd_file_put(copy->nf_src);
1723 if (copy->nf_dst)
1724 nfsd_file_put(copy->nf_dst);
1725}
1726
1727static void cleanup_async_copy(struct nfsd4_copy *copy)
1728{
1729 nfs4_free_copy_state(copy);
1730 release_copy_files(copy);
1731 if (copy->cp_clp) {
1732 spin_lock(©->cp_clp->async_lock);
1733 if (!list_empty(©->copies))
1734 list_del_init(©->copies);
1735 spin_unlock(©->cp_clp->async_lock);
1736 }
1737 nfs4_put_copy(copy);
1738}
1739
1740static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1741{
1742 struct nfsd4_cb_offload *cbo;
1743
1744 cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1745 if (!cbo)
1746 return;
1747
1748 memcpy(&cbo->co_res, ©->cp_res, sizeof(copy->cp_res));
1749 memcpy(&cbo->co_fh, ©->fh, sizeof(copy->fh));
1750 cbo->co_nfserr = nfserr;
1751
1752 nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1753 NFSPROC4_CLNT_CB_OFFLOAD);
1754 trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1755 &cbo->co_fh, copy->cp_count, nfserr);
1756 nfsd4_run_cb(&cbo->co_cb);
1757}
1758
1759/**
1760 * nfsd4_do_async_copy - kthread function for background server-side COPY
1761 * @data: arguments for COPY operation
1762 *
1763 * Return values:
1764 * %0: Copy operation is done.
1765 */
1766static int nfsd4_do_async_copy(void *data)
1767{
1768 struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1769 __be32 nfserr;
1770
1771 trace_nfsd_copy_do_async(copy);
1772 if (nfsd4_ssc_is_inter(copy)) {
1773 struct file *filp;
1774
1775 filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1776 ©->c_fh, ©->stateid);
1777 if (IS_ERR(filp)) {
1778 switch (PTR_ERR(filp)) {
1779 case -EBADF:
1780 nfserr = nfserr_wrong_type;
1781 break;
1782 default:
1783 nfserr = nfserr_offload_denied;
1784 }
1785 /* ss_mnt will be unmounted by the laundromat */
1786 goto do_callback;
1787 }
1788 nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1789 false);
1790 nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1791 } else {
1792 nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1793 copy->nf_dst->nf_file, false);
1794 }
1795
1796do_callback:
1797 nfsd4_send_cb_offload(copy, nfserr);
1798 cleanup_async_copy(copy);
1799 return 0;
1800}
1801
1802static __be32
1803nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1804 union nfsd4_op_u *u)
1805{
1806 struct nfsd4_copy *copy = &u->copy;
1807 __be32 status;
1808 struct nfsd4_copy *async_copy = NULL;
1809
1810 copy->cp_clp = cstate->clp;
1811 if (nfsd4_ssc_is_inter(copy)) {
1812 trace_nfsd_copy_inter(copy);
1813 if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1814 status = nfserr_notsupp;
1815 goto out;
1816 }
1817 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1818 if (status) {
1819 trace_nfsd_copy_done(copy, status);
1820 return nfserr_offload_denied;
1821 }
1822 } else {
1823 trace_nfsd_copy_intra(copy);
1824 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1825 if (status) {
1826 trace_nfsd_copy_done(copy, status);
1827 return status;
1828 }
1829 }
1830
1831 memcpy(©->fh, &cstate->current_fh.fh_handle,
1832 sizeof(struct knfsd_fh));
1833 if (nfsd4_copy_is_async(copy)) {
1834 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1835
1836 status = nfserrno(-ENOMEM);
1837 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1838 if (!async_copy)
1839 goto out_err;
1840 INIT_LIST_HEAD(&async_copy->copies);
1841 refcount_set(&async_copy->refcount, 1);
1842 async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1843 if (!async_copy->cp_src)
1844 goto out_err;
1845 if (!nfs4_init_copy_state(nn, copy))
1846 goto out_err;
1847 memcpy(©->cp_res.cb_stateid, ©->cp_stateid.cs_stid,
1848 sizeof(copy->cp_res.cb_stateid));
1849 dup_copy_fields(copy, async_copy);
1850 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1851 async_copy, "%s", "copy thread");
1852 if (IS_ERR(async_copy->copy_task))
1853 goto out_err;
1854 spin_lock(&async_copy->cp_clp->async_lock);
1855 list_add(&async_copy->copies,
1856 &async_copy->cp_clp->async_copies);
1857 spin_unlock(&async_copy->cp_clp->async_lock);
1858 wake_up_process(async_copy->copy_task);
1859 status = nfs_ok;
1860 } else {
1861 status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1862 copy->nf_dst->nf_file, true);
1863 }
1864out:
1865 trace_nfsd_copy_done(copy, status);
1866 release_copy_files(copy);
1867 return status;
1868out_err:
1869 if (nfsd4_ssc_is_inter(copy)) {
1870 /*
1871 * Source's vfsmount of inter-copy will be unmounted
1872 * by the laundromat. Use copy instead of async_copy
1873 * since async_copy->ss_nsui might not be set yet.
1874 */
1875 refcount_dec(©->ss_nsui->nsui_refcnt);
1876 }
1877 if (async_copy)
1878 cleanup_async_copy(async_copy);
1879 status = nfserrno(-ENOMEM);
1880 goto out;
1881}
1882
1883static struct nfsd4_copy *
1884find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1885{
1886 struct nfsd4_copy *copy;
1887
1888 lockdep_assert_held(&clp->async_lock);
1889
1890 list_for_each_entry(copy, &clp->async_copies, copies) {
1891 if (memcmp(©->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1892 continue;
1893 return copy;
1894 }
1895 return NULL;
1896}
1897
1898static struct nfsd4_copy *
1899find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1900{
1901 struct nfsd4_copy *copy;
1902
1903 spin_lock(&clp->async_lock);
1904 copy = find_async_copy_locked(clp, stateid);
1905 if (copy)
1906 refcount_inc(©->refcount);
1907 spin_unlock(&clp->async_lock);
1908 return copy;
1909}
1910
1911static __be32
1912nfsd4_offload_cancel(struct svc_rqst *rqstp,
1913 struct nfsd4_compound_state *cstate,
1914 union nfsd4_op_u *u)
1915{
1916 struct nfsd4_offload_status *os = &u->offload_status;
1917 struct nfsd4_copy *copy;
1918 struct nfs4_client *clp = cstate->clp;
1919
1920 copy = find_async_copy(clp, &os->stateid);
1921 if (!copy) {
1922 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1923
1924 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1925 } else
1926 nfsd4_stop_copy(copy);
1927
1928 return nfs_ok;
1929}
1930
1931static __be32
1932nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1933 union nfsd4_op_u *u)
1934{
1935 struct nfsd4_copy_notify *cn = &u->copy_notify;
1936 __be32 status;
1937 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1938 struct nfs4_stid *stid;
1939 struct nfs4_cpntf_state *cps;
1940 struct nfs4_client *clp = cstate->clp;
1941
1942 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1943 &cn->cpn_src_stateid, RD_STATE, NULL,
1944 &stid);
1945 if (status)
1946 return status;
1947
1948 cn->cpn_lease_time.tv_sec = nn->nfsd4_lease;
1949 cn->cpn_lease_time.tv_nsec = 0;
1950
1951 status = nfserrno(-ENOMEM);
1952 cps = nfs4_alloc_init_cpntf_state(nn, stid);
1953 if (!cps)
1954 goto out;
1955 memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1956 memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1957 memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1958
1959 /* For now, only return one server address in cpn_src, the
1960 * address used by the client to connect to this server.
1961 */
1962 cn->cpn_src->nl4_type = NL4_NETADDR;
1963 status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1964 &cn->cpn_src->u.nl4_addr);
1965 WARN_ON_ONCE(status);
1966 if (status) {
1967 nfs4_put_cpntf_state(nn, cps);
1968 goto out;
1969 }
1970out:
1971 nfs4_put_stid(stid);
1972 return status;
1973}
1974
1975static __be32
1976nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1977 struct nfsd4_fallocate *fallocate, int flags)
1978{
1979 __be32 status;
1980 struct nfsd_file *nf;
1981
1982 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1983 &fallocate->falloc_stateid,
1984 WR_STATE, &nf, NULL);
1985 if (status != nfs_ok)
1986 return status;
1987
1988 status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1989 fallocate->falloc_offset,
1990 fallocate->falloc_length,
1991 flags);
1992 nfsd_file_put(nf);
1993 return status;
1994}
1995
1996static __be32
1997nfsd4_offload_status(struct svc_rqst *rqstp,
1998 struct nfsd4_compound_state *cstate,
1999 union nfsd4_op_u *u)
2000{
2001 struct nfsd4_offload_status *os = &u->offload_status;
2002 __be32 status = nfs_ok;
2003 struct nfsd4_copy *copy;
2004 struct nfs4_client *clp = cstate->clp;
2005
2006 spin_lock(&clp->async_lock);
2007 copy = find_async_copy_locked(clp, &os->stateid);
2008 if (copy)
2009 os->count = copy->cp_res.wr_bytes_written;
2010 else
2011 status = nfserr_bad_stateid;
2012 spin_unlock(&clp->async_lock);
2013
2014 return status;
2015}
2016
2017static __be32
2018nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2019 union nfsd4_op_u *u)
2020{
2021 return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2022}
2023
2024static __be32
2025nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2026 union nfsd4_op_u *u)
2027{
2028 return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2029 FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2030}
2031
2032static __be32
2033nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2034 union nfsd4_op_u *u)
2035{
2036 struct nfsd4_seek *seek = &u->seek;
2037 int whence;
2038 __be32 status;
2039 struct nfsd_file *nf;
2040
2041 status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2042 &seek->seek_stateid,
2043 RD_STATE, &nf, NULL);
2044 if (status)
2045 return status;
2046
2047 switch (seek->seek_whence) {
2048 case NFS4_CONTENT_DATA:
2049 whence = SEEK_DATA;
2050 break;
2051 case NFS4_CONTENT_HOLE:
2052 whence = SEEK_HOLE;
2053 break;
2054 default:
2055 status = nfserr_union_notsupp;
2056 goto out;
2057 }
2058
2059 /*
2060 * Note: This call does change file->f_pos, but nothing in NFSD
2061 * should ever file->f_pos.
2062 */
2063 seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2064 if (seek->seek_pos < 0)
2065 status = nfserrno(seek->seek_pos);
2066 else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2067 seek->seek_eof = true;
2068
2069out:
2070 nfsd_file_put(nf);
2071 return status;
2072}
2073
2074/* This routine never returns NFS_OK! If there are no other errors, it
2075 * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2076 * attributes matched. VERIFY is implemented by mapping NFSERR_SAME
2077 * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2078 */
2079static __be32
2080_nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2081 struct nfsd4_verify *verify)
2082{
2083 __be32 *buf, *p;
2084 int count;
2085 __be32 status;
2086
2087 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2088 if (status)
2089 return status;
2090
2091 status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2092 if (status)
2093 return status;
2094
2095 if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2096 || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2097 return nfserr_inval;
2098 if (verify->ve_attrlen & 3)
2099 return nfserr_inval;
2100
2101 /* count in words:
2102 * bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2103 */
2104 count = 4 + (verify->ve_attrlen >> 2);
2105 buf = kmalloc(count << 2, GFP_KERNEL);
2106 if (!buf)
2107 return nfserr_jukebox;
2108
2109 p = buf;
2110 status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2111 cstate->current_fh.fh_export,
2112 cstate->current_fh.fh_dentry,
2113 verify->ve_bmval,
2114 rqstp, 0);
2115 /*
2116 * If nfsd4_encode_fattr() ran out of space, assume that's because
2117 * the attributes are longer (hence different) than those given:
2118 */
2119 if (status == nfserr_resource)
2120 status = nfserr_not_same;
2121 if (status)
2122 goto out_kfree;
2123
2124 /* skip bitmap */
2125 p = buf + 1 + ntohl(buf[0]);
2126 status = nfserr_not_same;
2127 if (ntohl(*p++) != verify->ve_attrlen)
2128 goto out_kfree;
2129 if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2130 status = nfserr_same;
2131
2132out_kfree:
2133 kfree(buf);
2134 return status;
2135}
2136
2137static __be32
2138nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2139 union nfsd4_op_u *u)
2140{
2141 __be32 status;
2142
2143 status = _nfsd4_verify(rqstp, cstate, &u->verify);
2144 return status == nfserr_not_same ? nfs_ok : status;
2145}
2146
2147static __be32
2148nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2149 union nfsd4_op_u *u)
2150{
2151 __be32 status;
2152
2153 status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2154 return status == nfserr_same ? nfs_ok : status;
2155}
2156
2157#ifdef CONFIG_NFSD_PNFS
2158static const struct nfsd4_layout_ops *
2159nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2160{
2161 if (!exp->ex_layout_types) {
2162 dprintk("%s: export does not support pNFS\n", __func__);
2163 return NULL;
2164 }
2165
2166 if (layout_type >= LAYOUT_TYPE_MAX ||
2167 !(exp->ex_layout_types & (1 << layout_type))) {
2168 dprintk("%s: layout type %d not supported\n",
2169 __func__, layout_type);
2170 return NULL;
2171 }
2172
2173 return nfsd4_layout_ops[layout_type];
2174}
2175
2176static __be32
2177nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2178 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2179{
2180 struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2181 const struct nfsd4_layout_ops *ops;
2182 struct nfsd4_deviceid_map *map;
2183 struct svc_export *exp;
2184 __be32 nfserr;
2185
2186 dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2187 __func__,
2188 gdp->gd_layout_type,
2189 gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2190 gdp->gd_maxcount);
2191
2192 map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2193 if (!map) {
2194 dprintk("%s: couldn't find device ID to export mapping!\n",
2195 __func__);
2196 return nfserr_noent;
2197 }
2198
2199 exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2200 if (IS_ERR(exp)) {
2201 dprintk("%s: could not find device id\n", __func__);
2202 return nfserr_noent;
2203 }
2204
2205 nfserr = nfserr_layoutunavailable;
2206 ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2207 if (!ops)
2208 goto out;
2209
2210 nfserr = nfs_ok;
2211 if (gdp->gd_maxcount != 0) {
2212 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2213 rqstp, cstate->clp, gdp);
2214 }
2215
2216 gdp->gd_notify_types &= ops->notify_types;
2217out:
2218 exp_put(exp);
2219 return nfserr;
2220}
2221
2222static void
2223nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2224{
2225 kfree(u->getdeviceinfo.gd_device);
2226}
2227
2228static __be32
2229nfsd4_layoutget(struct svc_rqst *rqstp,
2230 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2231{
2232 struct nfsd4_layoutget *lgp = &u->layoutget;
2233 struct svc_fh *current_fh = &cstate->current_fh;
2234 const struct nfsd4_layout_ops *ops;
2235 struct nfs4_layout_stateid *ls;
2236 __be32 nfserr;
2237 int accmode = NFSD_MAY_READ_IF_EXEC;
2238
2239 switch (lgp->lg_seg.iomode) {
2240 case IOMODE_READ:
2241 accmode |= NFSD_MAY_READ;
2242 break;
2243 case IOMODE_RW:
2244 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2245 break;
2246 default:
2247 dprintk("%s: invalid iomode %d\n",
2248 __func__, lgp->lg_seg.iomode);
2249 nfserr = nfserr_badiomode;
2250 goto out;
2251 }
2252
2253 nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2254 if (nfserr)
2255 goto out;
2256
2257 nfserr = nfserr_layoutunavailable;
2258 ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2259 if (!ops)
2260 goto out;
2261
2262 /*
2263 * Verify minlength and range as per RFC5661:
2264 * o If loga_length is less than loga_minlength,
2265 * the metadata server MUST return NFS4ERR_INVAL.
2266 * o If the sum of loga_offset and loga_minlength exceeds
2267 * NFS4_UINT64_MAX, and loga_minlength is not
2268 * NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2269 * o If the sum of loga_offset and loga_length exceeds
2270 * NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2271 * the error NFS4ERR_INVAL MUST result.
2272 */
2273 nfserr = nfserr_inval;
2274 if (lgp->lg_seg.length < lgp->lg_minlength ||
2275 (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2276 lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2277 (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2278 lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2279 goto out;
2280 if (lgp->lg_seg.length == 0)
2281 goto out;
2282
2283 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2284 true, lgp->lg_layout_type, &ls);
2285 if (nfserr) {
2286 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2287 goto out;
2288 }
2289
2290 nfserr = nfserr_recallconflict;
2291 if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2292 goto out_put_stid;
2293
2294 nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2295 current_fh, lgp);
2296 if (nfserr)
2297 goto out_put_stid;
2298
2299 nfserr = nfsd4_insert_layout(lgp, ls);
2300
2301out_put_stid:
2302 mutex_unlock(&ls->ls_mutex);
2303 nfs4_put_stid(&ls->ls_stid);
2304out:
2305 return nfserr;
2306}
2307
2308static void
2309nfsd4_layoutget_release(union nfsd4_op_u *u)
2310{
2311 kfree(u->layoutget.lg_content);
2312}
2313
2314static __be32
2315nfsd4_layoutcommit(struct svc_rqst *rqstp,
2316 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2317{
2318 struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2319 const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2320 struct svc_fh *current_fh = &cstate->current_fh;
2321 const struct nfsd4_layout_ops *ops;
2322 loff_t new_size = lcp->lc_last_wr + 1;
2323 struct inode *inode;
2324 struct nfs4_layout_stateid *ls;
2325 __be32 nfserr;
2326
2327 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2328 if (nfserr)
2329 goto out;
2330
2331 nfserr = nfserr_layoutunavailable;
2332 ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2333 if (!ops)
2334 goto out;
2335 inode = d_inode(current_fh->fh_dentry);
2336
2337 nfserr = nfserr_inval;
2338 if (new_size <= seg->offset) {
2339 dprintk("pnfsd: last write before layout segment\n");
2340 goto out;
2341 }
2342 if (new_size > seg->offset + seg->length) {
2343 dprintk("pnfsd: last write beyond layout segment\n");
2344 goto out;
2345 }
2346 if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2347 dprintk("pnfsd: layoutcommit beyond EOF\n");
2348 goto out;
2349 }
2350
2351 nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2352 false, lcp->lc_layout_type,
2353 &ls);
2354 if (nfserr) {
2355 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2356 /* fixup error code as per RFC5661 */
2357 if (nfserr == nfserr_bad_stateid)
2358 nfserr = nfserr_badlayout;
2359 goto out;
2360 }
2361
2362 /* LAYOUTCOMMIT does not require any serialization */
2363 mutex_unlock(&ls->ls_mutex);
2364
2365 if (new_size > i_size_read(inode)) {
2366 lcp->lc_size_chg = true;
2367 lcp->lc_newsize = new_size;
2368 } else {
2369 lcp->lc_size_chg = false;
2370 }
2371
2372 nfserr = ops->proc_layoutcommit(inode, lcp);
2373 nfs4_put_stid(&ls->ls_stid);
2374out:
2375 return nfserr;
2376}
2377
2378static __be32
2379nfsd4_layoutreturn(struct svc_rqst *rqstp,
2380 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2381{
2382 struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2383 struct svc_fh *current_fh = &cstate->current_fh;
2384 __be32 nfserr;
2385
2386 nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2387 if (nfserr)
2388 goto out;
2389
2390 nfserr = nfserr_layoutunavailable;
2391 if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2392 goto out;
2393
2394 switch (lrp->lr_seg.iomode) {
2395 case IOMODE_READ:
2396 case IOMODE_RW:
2397 case IOMODE_ANY:
2398 break;
2399 default:
2400 dprintk("%s: invalid iomode %d\n", __func__,
2401 lrp->lr_seg.iomode);
2402 nfserr = nfserr_inval;
2403 goto out;
2404 }
2405
2406 switch (lrp->lr_return_type) {
2407 case RETURN_FILE:
2408 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2409 break;
2410 case RETURN_FSID:
2411 case RETURN_ALL:
2412 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2413 break;
2414 default:
2415 dprintk("%s: invalid return_type %d\n", __func__,
2416 lrp->lr_return_type);
2417 nfserr = nfserr_inval;
2418 break;
2419 }
2420out:
2421 return nfserr;
2422}
2423#endif /* CONFIG_NFSD_PNFS */
2424
2425static __be32
2426nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2427 union nfsd4_op_u *u)
2428{
2429 struct nfsd4_getxattr *getxattr = &u->getxattr;
2430
2431 return nfsd_getxattr(rqstp, &cstate->current_fh,
2432 getxattr->getxa_name, &getxattr->getxa_buf,
2433 &getxattr->getxa_len);
2434}
2435
2436static __be32
2437nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2438 union nfsd4_op_u *u)
2439{
2440 struct nfsd4_setxattr *setxattr = &u->setxattr;
2441 __be32 ret;
2442
2443 if (opens_in_grace(SVC_NET(rqstp)))
2444 return nfserr_grace;
2445
2446 ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2447 setxattr->setxa_buf, setxattr->setxa_len,
2448 setxattr->setxa_flags);
2449
2450 if (!ret)
2451 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2452
2453 return ret;
2454}
2455
2456static __be32
2457nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2458 union nfsd4_op_u *u)
2459{
2460 /*
2461 * Get the entire list, then copy out only the user attributes
2462 * in the encode function.
2463 */
2464 return nfsd_listxattr(rqstp, &cstate->current_fh,
2465 &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2466}
2467
2468static __be32
2469nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2470 union nfsd4_op_u *u)
2471{
2472 struct nfsd4_removexattr *removexattr = &u->removexattr;
2473 __be32 ret;
2474
2475 if (opens_in_grace(SVC_NET(rqstp)))
2476 return nfserr_grace;
2477
2478 ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2479 removexattr->rmxa_name);
2480
2481 if (!ret)
2482 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2483
2484 return ret;
2485}
2486
2487/*
2488 * NULL call.
2489 */
2490static __be32
2491nfsd4_proc_null(struct svc_rqst *rqstp)
2492{
2493 return rpc_success;
2494}
2495
2496static inline void nfsd4_increment_op_stats(struct nfsd_net *nn, u32 opnum)
2497{
2498 if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2499 percpu_counter_inc(&nn->counter[NFSD_STATS_NFS4_OP(opnum)]);
2500}
2501
2502static const struct nfsd4_operation nfsd4_ops[];
2503
2504static const char *nfsd4_op_name(unsigned opnum);
2505
2506/*
2507 * Enforce NFSv4.1 COMPOUND ordering rules:
2508 *
2509 * Also note, enforced elsewhere:
2510 * - SEQUENCE other than as first op results in
2511 * NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2512 * - BIND_CONN_TO_SESSION must be the only op in its compound.
2513 * (Enforced in nfsd4_bind_conn_to_session().)
2514 * - DESTROY_SESSION must be the final operation in a compound, if
2515 * sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2516 * (Enforced in nfsd4_destroy_session().)
2517 */
2518static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2519{
2520 struct nfsd4_op *first_op = &args->ops[0];
2521
2522 /* These ordering requirements don't apply to NFSv4.0: */
2523 if (args->minorversion == 0)
2524 return nfs_ok;
2525 /* This is weird, but OK, not our problem: */
2526 if (args->opcnt == 0)
2527 return nfs_ok;
2528 if (first_op->status == nfserr_op_illegal)
2529 return nfs_ok;
2530 if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2531 return nfserr_op_not_in_session;
2532 if (first_op->opnum == OP_SEQUENCE)
2533 return nfs_ok;
2534 /*
2535 * So first_op is something allowed outside a session, like
2536 * EXCHANGE_ID; but then it has to be the only op in the
2537 * compound:
2538 */
2539 if (args->opcnt != 1)
2540 return nfserr_not_only_op;
2541 return nfs_ok;
2542}
2543
2544const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2545{
2546 return &nfsd4_ops[op->opnum];
2547}
2548
2549bool nfsd4_cache_this_op(struct nfsd4_op *op)
2550{
2551 if (op->opnum == OP_ILLEGAL)
2552 return false;
2553 return OPDESC(op)->op_flags & OP_CACHEME;
2554}
2555
2556static bool need_wrongsec_check(struct svc_rqst *rqstp)
2557{
2558 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2559 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2560 struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2561 struct nfsd4_op *next = &argp->ops[resp->opcnt];
2562 const struct nfsd4_operation *thisd = OPDESC(this);
2563 const struct nfsd4_operation *nextd;
2564
2565 /*
2566 * Most ops check wronsec on our own; only the putfh-like ops
2567 * have special rules.
2568 */
2569 if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2570 return false;
2571 /*
2572 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2573 * put-filehandle operation if we're not going to use the
2574 * result:
2575 */
2576 if (argp->opcnt == resp->opcnt)
2577 return false;
2578 if (next->opnum == OP_ILLEGAL)
2579 return false;
2580 nextd = OPDESC(next);
2581 /*
2582 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2583 * errors themselves as necessary; others should check for them
2584 * now:
2585 */
2586 return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2587}
2588
2589#ifdef CONFIG_NFSD_V4_2_INTER_SSC
2590static void
2591check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2592{
2593 struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2594 struct nfsd4_copy *copy;
2595 struct nfsd4_putfh *putfh;
2596 int i;
2597
2598 /* traverse all operation and if it's a COPY compound, mark the
2599 * source filehandle to skip verification
2600 */
2601 for (i = 0; i < args->opcnt; i++) {
2602 op = &args->ops[i];
2603 if (op->opnum == OP_PUTFH)
2604 current_op = op;
2605 else if (op->opnum == OP_SAVEFH)
2606 saved_op = current_op;
2607 else if (op->opnum == OP_RESTOREFH)
2608 current_op = saved_op;
2609 else if (op->opnum == OP_COPY) {
2610 copy = (struct nfsd4_copy *)&op->u;
2611 if (!saved_op) {
2612 op->status = nfserr_nofilehandle;
2613 return;
2614 }
2615 putfh = (struct nfsd4_putfh *)&saved_op->u;
2616 if (nfsd4_ssc_is_inter(copy))
2617 putfh->no_verify = true;
2618 }
2619 }
2620}
2621#else
2622static void
2623check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2624{
2625}
2626#endif
2627
2628/*
2629 * COMPOUND call.
2630 */
2631static __be32
2632nfsd4_proc_compound(struct svc_rqst *rqstp)
2633{
2634 struct nfsd4_compoundargs *args = rqstp->rq_argp;
2635 struct nfsd4_compoundres *resp = rqstp->rq_resp;
2636 struct nfsd4_op *op;
2637 struct nfsd4_compound_state *cstate = &resp->cstate;
2638 struct svc_fh *current_fh = &cstate->current_fh;
2639 struct svc_fh *save_fh = &cstate->save_fh;
2640 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2641 __be32 status;
2642
2643 resp->xdr = &rqstp->rq_res_stream;
2644 resp->statusp = resp->xdr->p;
2645
2646 /* reserve space for: NFS status code */
2647 xdr_reserve_space(resp->xdr, XDR_UNIT);
2648
2649 /* reserve space for: taglen, tag, and opcnt */
2650 xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2651 resp->taglen = args->taglen;
2652 resp->tag = args->tag;
2653 resp->rqstp = rqstp;
2654 cstate->minorversion = args->minorversion;
2655 fh_init(current_fh, NFS4_FHSIZE);
2656 fh_init(save_fh, NFS4_FHSIZE);
2657 /*
2658 * Don't use the deferral mechanism for NFSv4; compounds make it
2659 * too hard to avoid non-idempotency problems.
2660 */
2661 clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2662
2663 /*
2664 * According to RFC3010, this takes precedence over all other errors.
2665 */
2666 status = nfserr_minor_vers_mismatch;
2667 if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2668 goto out;
2669
2670 status = nfs41_check_op_ordering(args);
2671 if (status) {
2672 op = &args->ops[0];
2673 op->status = status;
2674 resp->opcnt = 1;
2675 goto encode_op;
2676 }
2677 check_if_stalefh_allowed(args);
2678
2679 rqstp->rq_lease_breaker = (void **)&cstate->clp;
2680
2681 trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2682 while (!status && resp->opcnt < args->opcnt) {
2683 op = &args->ops[resp->opcnt++];
2684
2685 if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2686 /* If there are still more operations to process,
2687 * stop here and report NFS4ERR_RESOURCE. */
2688 if (cstate->minorversion == 0 &&
2689 args->client_opcnt > resp->opcnt) {
2690 op->status = nfserr_resource;
2691 goto encode_op;
2692 }
2693 }
2694
2695 /*
2696 * The XDR decode routines may have pre-set op->status;
2697 * for example, if there is a miscellaneous XDR error
2698 * it will be set to nfserr_bad_xdr.
2699 */
2700 if (op->status) {
2701 if (op->opnum == OP_OPEN)
2702 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2703 goto encode_op;
2704 }
2705 if (!current_fh->fh_dentry &&
2706 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2707 if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2708 op->status = nfserr_nofilehandle;
2709 goto encode_op;
2710 }
2711 } else if (current_fh->fh_export &&
2712 current_fh->fh_export->ex_fslocs.migrated &&
2713 !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2714 op->status = nfserr_moved;
2715 goto encode_op;
2716 }
2717
2718 fh_clear_pre_post_attrs(current_fh);
2719
2720 /* If op is non-idempotent */
2721 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2722 /*
2723 * Don't execute this op if we couldn't encode a
2724 * successful reply:
2725 */
2726 u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2727 /*
2728 * Plus if there's another operation, make sure
2729 * we'll have space to at least encode an error:
2730 */
2731 if (resp->opcnt < args->opcnt)
2732 plen += COMPOUND_ERR_SLACK_SPACE;
2733 op->status = nfsd4_check_resp_size(resp, plen);
2734 }
2735
2736 if (op->status)
2737 goto encode_op;
2738
2739 if (op->opdesc->op_get_currentstateid)
2740 op->opdesc->op_get_currentstateid(cstate, &op->u);
2741 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2742
2743 /* Only from SEQUENCE */
2744 if (cstate->status == nfserr_replay_cache) {
2745 dprintk("%s NFS4.1 replay from cache\n", __func__);
2746 status = op->status;
2747 goto out;
2748 }
2749 if (!op->status) {
2750 if (op->opdesc->op_set_currentstateid)
2751 op->opdesc->op_set_currentstateid(cstate, &op->u);
2752
2753 if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2754 clear_current_stateid(cstate);
2755
2756 if (current_fh->fh_export &&
2757 need_wrongsec_check(rqstp))
2758 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2759 }
2760encode_op:
2761 if (op->status == nfserr_replay_me) {
2762 op->replay = &cstate->replay_owner->so_replay;
2763 nfsd4_encode_replay(resp->xdr, op);
2764 status = op->status = op->replay->rp_status;
2765 } else {
2766 nfsd4_encode_operation(resp, op);
2767 status = op->status;
2768 }
2769
2770 trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2771 status, nfsd4_op_name(op->opnum));
2772
2773 nfsd4_cstate_clear_replay(cstate);
2774 nfsd4_increment_op_stats(nn, op->opnum);
2775 }
2776
2777 fh_put(current_fh);
2778 fh_put(save_fh);
2779 BUG_ON(cstate->replay_owner);
2780out:
2781 cstate->status = status;
2782 /* Reset deferral mechanism for RPC deferrals */
2783 set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2784 return rpc_success;
2785}
2786
2787#define op_encode_hdr_size (2)
2788#define op_encode_stateid_maxsz (XDR_QUADLEN(NFS4_STATEID_SIZE))
2789#define op_encode_verifier_maxsz (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2790#define op_encode_change_info_maxsz (5)
2791#define nfs4_fattr_bitmap_maxsz (4)
2792
2793/* We'll fall back on returning no lockowner if run out of space: */
2794#define op_encode_lockowner_maxsz (0)
2795#define op_encode_lock_denied_maxsz (8 + op_encode_lockowner_maxsz)
2796
2797#define nfs4_owner_maxsz (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2798
2799#define op_encode_ace_maxsz (3 + nfs4_owner_maxsz)
2800#define op_encode_delegation_maxsz (1 + op_encode_stateid_maxsz + 1 + \
2801 op_encode_ace_maxsz)
2802
2803#define op_encode_channel_attrs_maxsz (6 + 1 + 1)
2804
2805/*
2806 * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2807 * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2808 * the maximum payload size here, based on transport limits and the size
2809 * of the remaining space in the rq_pages array.
2810 */
2811static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2812{
2813 u32 buflen;
2814
2815 buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2816 buflen -= rqstp->rq_auth_slack;
2817 buflen -= rqstp->rq_res.head[0].iov_len;
2818 return min_t(u32, buflen, svc_max_payload(rqstp));
2819}
2820
2821static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2822 const struct nfsd4_op *op)
2823{
2824 return (op_encode_hdr_size) * sizeof(__be32);
2825}
2826
2827static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2828 const struct nfsd4_op *op)
2829{
2830 return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2831}
2832
2833static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2834 const struct nfsd4_op *op)
2835{
2836 /* ac_supported, ac_resp_access */
2837 return (op_encode_hdr_size + 2)* sizeof(__be32);
2838}
2839
2840static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2841 const struct nfsd4_op *op)
2842{
2843 return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2844}
2845
2846static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2847 const struct nfsd4_op *op)
2848{
2849 return (op_encode_hdr_size + op_encode_change_info_maxsz
2850 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2851}
2852
2853/*
2854 * Note since this is an idempotent operation we won't insist on failing
2855 * the op prematurely if the estimate is too large. We may turn off splice
2856 * reads unnecessarily.
2857 */
2858static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2859 const struct nfsd4_op *op)
2860{
2861 const u32 *bmap = op->u.getattr.ga_bmval;
2862 u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2863 u32 ret = 0;
2864
2865 if (bmap0 & FATTR4_WORD0_ACL)
2866 return nfsd4_max_payload(rqstp);
2867 if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2868 return nfsd4_max_payload(rqstp);
2869
2870 if (bmap1 & FATTR4_WORD1_OWNER) {
2871 ret += IDMAP_NAMESZ + 4;
2872 bmap1 &= ~FATTR4_WORD1_OWNER;
2873 }
2874 if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2875 ret += IDMAP_NAMESZ + 4;
2876 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2877 }
2878 if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2879 ret += NFS4_FHSIZE + 4;
2880 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2881 }
2882 if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2883 ret += NFS4_MAXLABELLEN + 12;
2884 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2885 }
2886 /*
2887 * Largest of remaining attributes are 16 bytes (e.g.,
2888 * supported_attributes)
2889 */
2890 ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2891 /* bitmask, length */
2892 ret += 20;
2893 return ret;
2894}
2895
2896static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2897 const struct nfsd4_op *op)
2898{
2899 return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2900}
2901
2902static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2903 const struct nfsd4_op *op)
2904{
2905 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2906 * sizeof(__be32);
2907}
2908
2909static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2910 const struct nfsd4_op *op)
2911{
2912 return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2913 * sizeof(__be32);
2914}
2915
2916static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2917 const struct nfsd4_op *op)
2918{
2919 return (op_encode_hdr_size + op_encode_stateid_maxsz
2920 + op_encode_change_info_maxsz + 1
2921 + nfs4_fattr_bitmap_maxsz
2922 + op_encode_delegation_maxsz) * sizeof(__be32);
2923}
2924
2925static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2926 const struct nfsd4_op *op)
2927{
2928 u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2929
2930 return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2931}
2932
2933static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2934 const struct nfsd4_op *op)
2935{
2936 u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2937 /*
2938 * If we detect that the file changed during hole encoding, then we
2939 * recover by encoding the remaining reply as data. This means we need
2940 * to set aside enough room to encode two data segments.
2941 */
2942 u32 seg_len = 2 * (1 + 2 + 1);
2943
2944 return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2945}
2946
2947static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2948 const struct nfsd4_op *op)
2949{
2950 u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2951
2952 return (op_encode_hdr_size + op_encode_verifier_maxsz +
2953 XDR_QUADLEN(rlen)) * sizeof(__be32);
2954}
2955
2956static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2957 const struct nfsd4_op *op)
2958{
2959 return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2960}
2961
2962static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2963 const struct nfsd4_op *op)
2964{
2965 return (op_encode_hdr_size + op_encode_change_info_maxsz)
2966 * sizeof(__be32);
2967}
2968
2969static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2970 const struct nfsd4_op *op)
2971{
2972 return (op_encode_hdr_size + op_encode_change_info_maxsz
2973 + op_encode_change_info_maxsz) * sizeof(__be32);
2974}
2975
2976static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2977 const struct nfsd4_op *op)
2978{
2979 return (op_encode_hdr_size
2980 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2981}
2982
2983static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2984 const struct nfsd4_op *op)
2985{
2986 return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2987 * sizeof(__be32);
2988}
2989
2990static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2991 const struct nfsd4_op *op)
2992{
2993 return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2994}
2995
2996static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2997 const struct nfsd4_op *op)
2998{
2999 return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
3000 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
3001}
3002
3003static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
3004 const struct nfsd4_op *op)
3005{
3006 return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
3007 sizeof(__be32);
3008}
3009
3010static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
3011 const struct nfsd4_op *op)
3012{
3013 return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3014}
3015
3016static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3017 const struct nfsd4_op *op)
3018{
3019 return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3020 1 + 1 + /* eir_flags, spr_how */\
3021 4 + /* spo_must_enforce & _allow with bitmap */\
3022 2 + /*eir_server_owner.so_minor_id */\
3023 /* eir_server_owner.so_major_id<> */\
3024 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3025 /* eir_server_scope<> */\
3026 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3027 1 + /* eir_server_impl_id array length */\
3028 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3029}
3030
3031static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3032 const struct nfsd4_op *op)
3033{
3034 return (op_encode_hdr_size + \
3035 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3036 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3037}
3038
3039static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3040 const struct nfsd4_op *op)
3041{
3042 return (op_encode_hdr_size + \
3043 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3044 2 + /* csr_sequence, csr_flags */\
3045 op_encode_channel_attrs_maxsz + \
3046 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3047}
3048
3049static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3050 const struct nfsd4_op *op)
3051{
3052 return (op_encode_hdr_size +
3053 1 /* wr_callback */ +
3054 op_encode_stateid_maxsz /* wr_callback */ +
3055 2 /* wr_count */ +
3056 1 /* wr_committed */ +
3057 op_encode_verifier_maxsz +
3058 1 /* cr_consecutive */ +
3059 1 /* cr_synchronous */) * sizeof(__be32);
3060}
3061
3062static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3063 const struct nfsd4_op *op)
3064{
3065 return (op_encode_hdr_size +
3066 2 /* osr_count */ +
3067 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3068}
3069
3070static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3071 const struct nfsd4_op *op)
3072{
3073 return (op_encode_hdr_size +
3074 3 /* cnr_lease_time */ +
3075 1 /* We support one cnr_source_server */ +
3076 1 /* cnr_stateid seq */ +
3077 op_encode_stateid_maxsz /* cnr_stateid */ +
3078 1 /* num cnr_source_server*/ +
3079 1 /* nl4_type */ +
3080 1 /* nl4 size */ +
3081 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3082 * sizeof(__be32);
3083}
3084
3085#ifdef CONFIG_NFSD_PNFS
3086static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3087 const struct nfsd4_op *op)
3088{
3089 u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3090
3091 return (op_encode_hdr_size +
3092 1 /* gd_layout_type*/ +
3093 XDR_QUADLEN(rlen) +
3094 2 /* gd_notify_types */) * sizeof(__be32);
3095}
3096
3097/*
3098 * At this stage we don't really know what layout driver will handle the request,
3099 * so we need to define an arbitrary upper bound here.
3100 */
3101#define MAX_LAYOUT_SIZE 128
3102static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3103 const struct nfsd4_op *op)
3104{
3105 return (op_encode_hdr_size +
3106 1 /* logr_return_on_close */ +
3107 op_encode_stateid_maxsz +
3108 1 /* nr of layouts */ +
3109 MAX_LAYOUT_SIZE) * sizeof(__be32);
3110}
3111
3112static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3113 const struct nfsd4_op *op)
3114{
3115 return (op_encode_hdr_size +
3116 1 /* locr_newsize */ +
3117 2 /* ns_size */) * sizeof(__be32);
3118}
3119
3120static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3121 const struct nfsd4_op *op)
3122{
3123 return (op_encode_hdr_size +
3124 1 /* lrs_stateid */ +
3125 op_encode_stateid_maxsz) * sizeof(__be32);
3126}
3127#endif /* CONFIG_NFSD_PNFS */
3128
3129
3130static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3131 const struct nfsd4_op *op)
3132{
3133 return (op_encode_hdr_size + 3) * sizeof(__be32);
3134}
3135
3136static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3137 const struct nfsd4_op *op)
3138{
3139 u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3140
3141 return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3142}
3143
3144static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3145 const struct nfsd4_op *op)
3146{
3147 return (op_encode_hdr_size + op_encode_change_info_maxsz)
3148 * sizeof(__be32);
3149}
3150static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3151 const struct nfsd4_op *op)
3152{
3153 u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3154
3155 return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3156}
3157
3158static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3159 const struct nfsd4_op *op)
3160{
3161 return (op_encode_hdr_size + op_encode_change_info_maxsz)
3162 * sizeof(__be32);
3163}
3164
3165
3166static const struct nfsd4_operation nfsd4_ops[] = {
3167 [OP_ACCESS] = {
3168 .op_func = nfsd4_access,
3169 .op_name = "OP_ACCESS",
3170 .op_rsize_bop = nfsd4_access_rsize,
3171 },
3172 [OP_CLOSE] = {
3173 .op_func = nfsd4_close,
3174 .op_flags = OP_MODIFIES_SOMETHING,
3175 .op_name = "OP_CLOSE",
3176 .op_rsize_bop = nfsd4_status_stateid_rsize,
3177 .op_get_currentstateid = nfsd4_get_closestateid,
3178 .op_set_currentstateid = nfsd4_set_closestateid,
3179 },
3180 [OP_COMMIT] = {
3181 .op_func = nfsd4_commit,
3182 .op_flags = OP_MODIFIES_SOMETHING,
3183 .op_name = "OP_COMMIT",
3184 .op_rsize_bop = nfsd4_commit_rsize,
3185 },
3186 [OP_CREATE] = {
3187 .op_func = nfsd4_create,
3188 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3189 .op_name = "OP_CREATE",
3190 .op_rsize_bop = nfsd4_create_rsize,
3191 },
3192 [OP_DELEGRETURN] = {
3193 .op_func = nfsd4_delegreturn,
3194 .op_flags = OP_MODIFIES_SOMETHING,
3195 .op_name = "OP_DELEGRETURN",
3196 .op_rsize_bop = nfsd4_only_status_rsize,
3197 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3198 },
3199 [OP_GETATTR] = {
3200 .op_func = nfsd4_getattr,
3201 .op_flags = ALLOWED_ON_ABSENT_FS,
3202 .op_rsize_bop = nfsd4_getattr_rsize,
3203 .op_name = "OP_GETATTR",
3204 },
3205 [OP_GETFH] = {
3206 .op_func = nfsd4_getfh,
3207 .op_name = "OP_GETFH",
3208 .op_rsize_bop = nfsd4_getfh_rsize,
3209 },
3210 [OP_LINK] = {
3211 .op_func = nfsd4_link,
3212 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3213 | OP_CACHEME,
3214 .op_name = "OP_LINK",
3215 .op_rsize_bop = nfsd4_link_rsize,
3216 },
3217 [OP_LOCK] = {
3218 .op_func = nfsd4_lock,
3219 .op_release = nfsd4_lock_release,
3220 .op_flags = OP_MODIFIES_SOMETHING |
3221 OP_NONTRIVIAL_ERROR_ENCODE,
3222 .op_name = "OP_LOCK",
3223 .op_rsize_bop = nfsd4_lock_rsize,
3224 .op_set_currentstateid = nfsd4_set_lockstateid,
3225 },
3226 [OP_LOCKT] = {
3227 .op_func = nfsd4_lockt,
3228 .op_release = nfsd4_lockt_release,
3229 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3230 .op_name = "OP_LOCKT",
3231 .op_rsize_bop = nfsd4_lock_rsize,
3232 },
3233 [OP_LOCKU] = {
3234 .op_func = nfsd4_locku,
3235 .op_flags = OP_MODIFIES_SOMETHING,
3236 .op_name = "OP_LOCKU",
3237 .op_rsize_bop = nfsd4_status_stateid_rsize,
3238 .op_get_currentstateid = nfsd4_get_lockustateid,
3239 },
3240 [OP_LOOKUP] = {
3241 .op_func = nfsd4_lookup,
3242 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3243 .op_name = "OP_LOOKUP",
3244 .op_rsize_bop = nfsd4_only_status_rsize,
3245 },
3246 [OP_LOOKUPP] = {
3247 .op_func = nfsd4_lookupp,
3248 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3249 .op_name = "OP_LOOKUPP",
3250 .op_rsize_bop = nfsd4_only_status_rsize,
3251 },
3252 [OP_NVERIFY] = {
3253 .op_func = nfsd4_nverify,
3254 .op_name = "OP_NVERIFY",
3255 .op_rsize_bop = nfsd4_only_status_rsize,
3256 },
3257 [OP_OPEN] = {
3258 .op_func = nfsd4_open,
3259 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3260 .op_name = "OP_OPEN",
3261 .op_rsize_bop = nfsd4_open_rsize,
3262 .op_set_currentstateid = nfsd4_set_openstateid,
3263 },
3264 [OP_OPEN_CONFIRM] = {
3265 .op_func = nfsd4_open_confirm,
3266 .op_flags = OP_MODIFIES_SOMETHING,
3267 .op_name = "OP_OPEN_CONFIRM",
3268 .op_rsize_bop = nfsd4_status_stateid_rsize,
3269 },
3270 [OP_OPEN_DOWNGRADE] = {
3271 .op_func = nfsd4_open_downgrade,
3272 .op_flags = OP_MODIFIES_SOMETHING,
3273 .op_name = "OP_OPEN_DOWNGRADE",
3274 .op_rsize_bop = nfsd4_status_stateid_rsize,
3275 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3276 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3277 },
3278 [OP_PUTFH] = {
3279 .op_func = nfsd4_putfh,
3280 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3281 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3282 .op_name = "OP_PUTFH",
3283 .op_rsize_bop = nfsd4_only_status_rsize,
3284 },
3285 [OP_PUTPUBFH] = {
3286 .op_func = nfsd4_putrootfh,
3287 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3288 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3289 .op_name = "OP_PUTPUBFH",
3290 .op_rsize_bop = nfsd4_only_status_rsize,
3291 },
3292 [OP_PUTROOTFH] = {
3293 .op_func = nfsd4_putrootfh,
3294 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3295 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3296 .op_name = "OP_PUTROOTFH",
3297 .op_rsize_bop = nfsd4_only_status_rsize,
3298 },
3299 [OP_READ] = {
3300 .op_func = nfsd4_read,
3301 .op_release = nfsd4_read_release,
3302 .op_name = "OP_READ",
3303 .op_rsize_bop = nfsd4_read_rsize,
3304 .op_get_currentstateid = nfsd4_get_readstateid,
3305 },
3306 [OP_READDIR] = {
3307 .op_func = nfsd4_readdir,
3308 .op_name = "OP_READDIR",
3309 .op_rsize_bop = nfsd4_readdir_rsize,
3310 },
3311 [OP_READLINK] = {
3312 .op_func = nfsd4_readlink,
3313 .op_name = "OP_READLINK",
3314 .op_rsize_bop = nfsd4_readlink_rsize,
3315 },
3316 [OP_REMOVE] = {
3317 .op_func = nfsd4_remove,
3318 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3319 .op_name = "OP_REMOVE",
3320 .op_rsize_bop = nfsd4_remove_rsize,
3321 },
3322 [OP_RENAME] = {
3323 .op_func = nfsd4_rename,
3324 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3325 .op_name = "OP_RENAME",
3326 .op_rsize_bop = nfsd4_rename_rsize,
3327 },
3328 [OP_RENEW] = {
3329 .op_func = nfsd4_renew,
3330 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3331 | OP_MODIFIES_SOMETHING,
3332 .op_name = "OP_RENEW",
3333 .op_rsize_bop = nfsd4_only_status_rsize,
3334
3335 },
3336 [OP_RESTOREFH] = {
3337 .op_func = nfsd4_restorefh,
3338 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3339 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3340 .op_name = "OP_RESTOREFH",
3341 .op_rsize_bop = nfsd4_only_status_rsize,
3342 },
3343 [OP_SAVEFH] = {
3344 .op_func = nfsd4_savefh,
3345 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3346 .op_name = "OP_SAVEFH",
3347 .op_rsize_bop = nfsd4_only_status_rsize,
3348 },
3349 [OP_SECINFO] = {
3350 .op_func = nfsd4_secinfo,
3351 .op_release = nfsd4_secinfo_release,
3352 .op_flags = OP_HANDLES_WRONGSEC,
3353 .op_name = "OP_SECINFO",
3354 .op_rsize_bop = nfsd4_secinfo_rsize,
3355 },
3356 [OP_SETATTR] = {
3357 .op_func = nfsd4_setattr,
3358 .op_name = "OP_SETATTR",
3359 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3360 | OP_NONTRIVIAL_ERROR_ENCODE,
3361 .op_rsize_bop = nfsd4_setattr_rsize,
3362 .op_get_currentstateid = nfsd4_get_setattrstateid,
3363 },
3364 [OP_SETCLIENTID] = {
3365 .op_func = nfsd4_setclientid,
3366 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3367 | OP_MODIFIES_SOMETHING | OP_CACHEME
3368 | OP_NONTRIVIAL_ERROR_ENCODE,
3369 .op_name = "OP_SETCLIENTID",
3370 .op_rsize_bop = nfsd4_setclientid_rsize,
3371 },
3372 [OP_SETCLIENTID_CONFIRM] = {
3373 .op_func = nfsd4_setclientid_confirm,
3374 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3375 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3376 .op_name = "OP_SETCLIENTID_CONFIRM",
3377 .op_rsize_bop = nfsd4_only_status_rsize,
3378 },
3379 [OP_VERIFY] = {
3380 .op_func = nfsd4_verify,
3381 .op_name = "OP_VERIFY",
3382 .op_rsize_bop = nfsd4_only_status_rsize,
3383 },
3384 [OP_WRITE] = {
3385 .op_func = nfsd4_write,
3386 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3387 .op_name = "OP_WRITE",
3388 .op_rsize_bop = nfsd4_write_rsize,
3389 .op_get_currentstateid = nfsd4_get_writestateid,
3390 },
3391 [OP_RELEASE_LOCKOWNER] = {
3392 .op_func = nfsd4_release_lockowner,
3393 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3394 | OP_MODIFIES_SOMETHING,
3395 .op_name = "OP_RELEASE_LOCKOWNER",
3396 .op_rsize_bop = nfsd4_only_status_rsize,
3397 },
3398
3399 /* NFSv4.1 operations */
3400 [OP_EXCHANGE_ID] = {
3401 .op_func = nfsd4_exchange_id,
3402 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3403 | OP_MODIFIES_SOMETHING,
3404 .op_name = "OP_EXCHANGE_ID",
3405 .op_rsize_bop = nfsd4_exchange_id_rsize,
3406 },
3407 [OP_BACKCHANNEL_CTL] = {
3408 .op_func = nfsd4_backchannel_ctl,
3409 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3410 .op_name = "OP_BACKCHANNEL_CTL",
3411 .op_rsize_bop = nfsd4_only_status_rsize,
3412 },
3413 [OP_BIND_CONN_TO_SESSION] = {
3414 .op_func = nfsd4_bind_conn_to_session,
3415 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3416 | OP_MODIFIES_SOMETHING,
3417 .op_name = "OP_BIND_CONN_TO_SESSION",
3418 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3419 },
3420 [OP_CREATE_SESSION] = {
3421 .op_func = nfsd4_create_session,
3422 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3423 | OP_MODIFIES_SOMETHING,
3424 .op_name = "OP_CREATE_SESSION",
3425 .op_rsize_bop = nfsd4_create_session_rsize,
3426 },
3427 [OP_DESTROY_SESSION] = {
3428 .op_func = nfsd4_destroy_session,
3429 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3430 | OP_MODIFIES_SOMETHING,
3431 .op_name = "OP_DESTROY_SESSION",
3432 .op_rsize_bop = nfsd4_only_status_rsize,
3433 },
3434 [OP_SEQUENCE] = {
3435 .op_func = nfsd4_sequence,
3436 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3437 .op_name = "OP_SEQUENCE",
3438 .op_rsize_bop = nfsd4_sequence_rsize,
3439 },
3440 [OP_DESTROY_CLIENTID] = {
3441 .op_func = nfsd4_destroy_clientid,
3442 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3443 | OP_MODIFIES_SOMETHING,
3444 .op_name = "OP_DESTROY_CLIENTID",
3445 .op_rsize_bop = nfsd4_only_status_rsize,
3446 },
3447 [OP_RECLAIM_COMPLETE] = {
3448 .op_func = nfsd4_reclaim_complete,
3449 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3450 .op_name = "OP_RECLAIM_COMPLETE",
3451 .op_rsize_bop = nfsd4_only_status_rsize,
3452 },
3453 [OP_SECINFO_NO_NAME] = {
3454 .op_func = nfsd4_secinfo_no_name,
3455 .op_release = nfsd4_secinfo_no_name_release,
3456 .op_flags = OP_HANDLES_WRONGSEC,
3457 .op_name = "OP_SECINFO_NO_NAME",
3458 .op_rsize_bop = nfsd4_secinfo_rsize,
3459 },
3460 [OP_TEST_STATEID] = {
3461 .op_func = nfsd4_test_stateid,
3462 .op_flags = ALLOWED_WITHOUT_FH,
3463 .op_name = "OP_TEST_STATEID",
3464 .op_rsize_bop = nfsd4_test_stateid_rsize,
3465 },
3466 [OP_FREE_STATEID] = {
3467 .op_func = nfsd4_free_stateid,
3468 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3469 .op_name = "OP_FREE_STATEID",
3470 .op_get_currentstateid = nfsd4_get_freestateid,
3471 .op_rsize_bop = nfsd4_only_status_rsize,
3472 },
3473#ifdef CONFIG_NFSD_PNFS
3474 [OP_GETDEVICEINFO] = {
3475 .op_func = nfsd4_getdeviceinfo,
3476 .op_release = nfsd4_getdeviceinfo_release,
3477 .op_flags = ALLOWED_WITHOUT_FH,
3478 .op_name = "OP_GETDEVICEINFO",
3479 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3480 },
3481 [OP_LAYOUTGET] = {
3482 .op_func = nfsd4_layoutget,
3483 .op_release = nfsd4_layoutget_release,
3484 .op_flags = OP_MODIFIES_SOMETHING,
3485 .op_name = "OP_LAYOUTGET",
3486 .op_rsize_bop = nfsd4_layoutget_rsize,
3487 },
3488 [OP_LAYOUTCOMMIT] = {
3489 .op_func = nfsd4_layoutcommit,
3490 .op_flags = OP_MODIFIES_SOMETHING,
3491 .op_name = "OP_LAYOUTCOMMIT",
3492 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3493 },
3494 [OP_LAYOUTRETURN] = {
3495 .op_func = nfsd4_layoutreturn,
3496 .op_flags = OP_MODIFIES_SOMETHING,
3497 .op_name = "OP_LAYOUTRETURN",
3498 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3499 },
3500#endif /* CONFIG_NFSD_PNFS */
3501
3502 /* NFSv4.2 operations */
3503 [OP_ALLOCATE] = {
3504 .op_func = nfsd4_allocate,
3505 .op_flags = OP_MODIFIES_SOMETHING,
3506 .op_name = "OP_ALLOCATE",
3507 .op_rsize_bop = nfsd4_only_status_rsize,
3508 },
3509 [OP_DEALLOCATE] = {
3510 .op_func = nfsd4_deallocate,
3511 .op_flags = OP_MODIFIES_SOMETHING,
3512 .op_name = "OP_DEALLOCATE",
3513 .op_rsize_bop = nfsd4_only_status_rsize,
3514 },
3515 [OP_CLONE] = {
3516 .op_func = nfsd4_clone,
3517 .op_flags = OP_MODIFIES_SOMETHING,
3518 .op_name = "OP_CLONE",
3519 .op_rsize_bop = nfsd4_only_status_rsize,
3520 },
3521 [OP_COPY] = {
3522 .op_func = nfsd4_copy,
3523 .op_flags = OP_MODIFIES_SOMETHING,
3524 .op_name = "OP_COPY",
3525 .op_rsize_bop = nfsd4_copy_rsize,
3526 },
3527 [OP_READ_PLUS] = {
3528 .op_func = nfsd4_read,
3529 .op_release = nfsd4_read_release,
3530 .op_name = "OP_READ_PLUS",
3531 .op_rsize_bop = nfsd4_read_plus_rsize,
3532 .op_get_currentstateid = nfsd4_get_readstateid,
3533 },
3534 [OP_SEEK] = {
3535 .op_func = nfsd4_seek,
3536 .op_name = "OP_SEEK",
3537 .op_rsize_bop = nfsd4_seek_rsize,
3538 },
3539 [OP_OFFLOAD_STATUS] = {
3540 .op_func = nfsd4_offload_status,
3541 .op_name = "OP_OFFLOAD_STATUS",
3542 .op_rsize_bop = nfsd4_offload_status_rsize,
3543 },
3544 [OP_OFFLOAD_CANCEL] = {
3545 .op_func = nfsd4_offload_cancel,
3546 .op_flags = OP_MODIFIES_SOMETHING,
3547 .op_name = "OP_OFFLOAD_CANCEL",
3548 .op_rsize_bop = nfsd4_only_status_rsize,
3549 },
3550 [OP_COPY_NOTIFY] = {
3551 .op_func = nfsd4_copy_notify,
3552 .op_flags = OP_MODIFIES_SOMETHING,
3553 .op_name = "OP_COPY_NOTIFY",
3554 .op_rsize_bop = nfsd4_copy_notify_rsize,
3555 },
3556 [OP_GETXATTR] = {
3557 .op_func = nfsd4_getxattr,
3558 .op_name = "OP_GETXATTR",
3559 .op_rsize_bop = nfsd4_getxattr_rsize,
3560 },
3561 [OP_SETXATTR] = {
3562 .op_func = nfsd4_setxattr,
3563 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3564 .op_name = "OP_SETXATTR",
3565 .op_rsize_bop = nfsd4_setxattr_rsize,
3566 },
3567 [OP_LISTXATTRS] = {
3568 .op_func = nfsd4_listxattrs,
3569 .op_name = "OP_LISTXATTRS",
3570 .op_rsize_bop = nfsd4_listxattrs_rsize,
3571 },
3572 [OP_REMOVEXATTR] = {
3573 .op_func = nfsd4_removexattr,
3574 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3575 .op_name = "OP_REMOVEXATTR",
3576 .op_rsize_bop = nfsd4_removexattr_rsize,
3577 },
3578};
3579
3580/**
3581 * nfsd4_spo_must_allow - Determine if the compound op contains an
3582 * operation that is allowed to be sent with machine credentials
3583 *
3584 * @rqstp: a pointer to the struct svc_rqst
3585 *
3586 * Checks to see if the compound contains a spo_must_allow op
3587 * and confirms that it was sent with the proper machine creds.
3588 */
3589
3590bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3591{
3592 struct nfsd4_compoundres *resp = rqstp->rq_resp;
3593 struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3594 struct nfsd4_op *this;
3595 struct nfsd4_compound_state *cstate = &resp->cstate;
3596 struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3597 u32 opiter;
3598
3599 if (!cstate->minorversion)
3600 return false;
3601
3602 if (cstate->spo_must_allowed)
3603 return true;
3604
3605 opiter = resp->opcnt;
3606 while (opiter < argp->opcnt) {
3607 this = &argp->ops[opiter++];
3608 if (test_bit(this->opnum, allow->u.longs) &&
3609 cstate->clp->cl_mach_cred &&
3610 nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3611 cstate->spo_must_allowed = true;
3612 return true;
3613 }
3614 }
3615 cstate->spo_must_allowed = false;
3616 return false;
3617}
3618
3619int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3620{
3621 if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3622 return op_encode_hdr_size * sizeof(__be32);
3623
3624 BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3625 return OPDESC(op)->op_rsize_bop(rqstp, op);
3626}
3627
3628void warn_on_nonidempotent_op(struct nfsd4_op *op)
3629{
3630 if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3631 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3632 op->opnum, nfsd4_op_name(op->opnum));
3633 WARN_ON_ONCE(1);
3634 }
3635}
3636
3637static const char *nfsd4_op_name(unsigned opnum)
3638{
3639 if (opnum < ARRAY_SIZE(nfsd4_ops))
3640 return nfsd4_ops[opnum].op_name;
3641 return "unknown_operation";
3642}
3643
3644static const struct svc_procedure nfsd_procedures4[2] = {
3645 [NFSPROC4_NULL] = {
3646 .pc_func = nfsd4_proc_null,
3647 .pc_decode = nfssvc_decode_voidarg,
3648 .pc_encode = nfssvc_encode_voidres,
3649 .pc_argsize = sizeof(struct nfsd_voidargs),
3650 .pc_argzero = sizeof(struct nfsd_voidargs),
3651 .pc_ressize = sizeof(struct nfsd_voidres),
3652 .pc_cachetype = RC_NOCACHE,
3653 .pc_xdrressize = 1,
3654 .pc_name = "NULL",
3655 },
3656 [NFSPROC4_COMPOUND] = {
3657 .pc_func = nfsd4_proc_compound,
3658 .pc_decode = nfs4svc_decode_compoundargs,
3659 .pc_encode = nfs4svc_encode_compoundres,
3660 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3661 .pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3662 .pc_ressize = sizeof(struct nfsd4_compoundres),
3663 .pc_release = nfsd4_release_compoundargs,
3664 .pc_cachetype = RC_NOCACHE,
3665 .pc_xdrressize = NFSD_BUFSIZE/4,
3666 .pc_name = "COMPOUND",
3667 },
3668};
3669
3670static DEFINE_PER_CPU_ALIGNED(unsigned long,
3671 nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3672const struct svc_version nfsd_version4 = {
3673 .vs_vers = 4,
3674 .vs_nproc = ARRAY_SIZE(nfsd_procedures4),
3675 .vs_proc = nfsd_procedures4,
3676 .vs_count = nfsd_count4,
3677 .vs_dispatch = nfsd_dispatch,
3678 .vs_xdrsize = NFS4_SVC_XDRSIZE,
3679 .vs_rpcb_optnl = true,
3680 .vs_need_cong_ctrl = true,
3681};