xref: /openbmc/linux/fs/nfsd/nfs4proc.c (revision 4949009e)
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/file.h>
36 #include <linux/falloc.h>
37 #include <linux/slab.h>
38 
39 #include "idmap.h"
40 #include "cache.h"
41 #include "xdr4.h"
42 #include "vfs.h"
43 #include "current_stateid.h"
44 #include "netns.h"
45 #include "acl.h"
46 
47 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
48 #include <linux/security.h>
49 
50 static inline void
51 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
52 {
53 	struct inode *inode = resfh->fh_dentry->d_inode;
54 	int status;
55 
56 	mutex_lock(&inode->i_mutex);
57 	status = security_inode_setsecctx(resfh->fh_dentry,
58 		label->data, label->len);
59 	mutex_unlock(&inode->i_mutex);
60 
61 	if (status)
62 		/*
63 		 * XXX: We should really fail the whole open, but we may
64 		 * already have created a new file, so it may be too
65 		 * late.  For now this seems the least of evils:
66 		 */
67 		bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
68 
69 	return;
70 }
71 #else
72 static inline void
73 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
74 { }
75 #endif
76 
77 #define NFSDDBG_FACILITY		NFSDDBG_PROC
78 
79 static u32 nfsd_attrmask[] = {
80 	NFSD_WRITEABLE_ATTRS_WORD0,
81 	NFSD_WRITEABLE_ATTRS_WORD1,
82 	NFSD_WRITEABLE_ATTRS_WORD2
83 };
84 
85 static u32 nfsd41_ex_attrmask[] = {
86 	NFSD_SUPPATTR_EXCLCREAT_WORD0,
87 	NFSD_SUPPATTR_EXCLCREAT_WORD1,
88 	NFSD_SUPPATTR_EXCLCREAT_WORD2
89 };
90 
91 static __be32
92 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
93 		   u32 *bmval, u32 *writable)
94 {
95 	struct dentry *dentry = cstate->current_fh.fh_dentry;
96 
97 	/*
98 	 * Check about attributes are supported by the NFSv4 server or not.
99 	 * According to spec, unsupported attributes return ERR_ATTRNOTSUPP.
100 	 */
101 	if ((bmval[0] & ~nfsd_suppattrs0(cstate->minorversion)) ||
102 	    (bmval[1] & ~nfsd_suppattrs1(cstate->minorversion)) ||
103 	    (bmval[2] & ~nfsd_suppattrs2(cstate->minorversion)))
104 		return nfserr_attrnotsupp;
105 
106 	/*
107 	 * Check FATTR4_WORD0_ACL can be supported
108 	 * in current environment or not.
109 	 */
110 	if (bmval[0] & FATTR4_WORD0_ACL) {
111 		if (!IS_POSIXACL(dentry->d_inode))
112 			return nfserr_attrnotsupp;
113 	}
114 
115 	/*
116 	 * According to spec, read-only attributes return ERR_INVAL.
117 	 */
118 	if (writable) {
119 		if ((bmval[0] & ~writable[0]) || (bmval[1] & ~writable[1]) ||
120 		    (bmval[2] & ~writable[2]))
121 			return nfserr_inval;
122 	}
123 
124 	return nfs_ok;
125 }
126 
127 static __be32
128 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
129 	struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
130 {
131 	__be32 status = nfs_ok;
132 
133 	if (open->op_create == NFS4_OPEN_CREATE) {
134 		if (open->op_createmode == NFS4_CREATE_UNCHECKED
135 		    || open->op_createmode == NFS4_CREATE_GUARDED)
136 			status = check_attr_support(rqstp, cstate,
137 					open->op_bmval, nfsd_attrmask);
138 		else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
139 			status = check_attr_support(rqstp, cstate,
140 					open->op_bmval, nfsd41_ex_attrmask);
141 	}
142 
143 	return status;
144 }
145 
146 static int
147 is_create_with_attrs(struct nfsd4_open *open)
148 {
149 	return open->op_create == NFS4_OPEN_CREATE
150 		&& (open->op_createmode == NFS4_CREATE_UNCHECKED
151 		    || open->op_createmode == NFS4_CREATE_GUARDED
152 		    || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
153 }
154 
155 /*
156  * if error occurs when setting the acl, just clear the acl bit
157  * in the returned attr bitmap.
158  */
159 static void
160 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
161 		struct nfs4_acl *acl, u32 *bmval)
162 {
163 	__be32 status;
164 
165 	status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
166 	if (status)
167 		/*
168 		 * We should probably fail the whole open at this point,
169 		 * but we've already created the file, so it's too late;
170 		 * So this seems the least of evils:
171 		 */
172 		bmval[0] &= ~FATTR4_WORD0_ACL;
173 }
174 
175 static inline void
176 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
177 {
178 	fh_put(dst);
179 	dget(src->fh_dentry);
180 	if (src->fh_export)
181 		exp_get(src->fh_export);
182 	*dst = *src;
183 }
184 
185 static __be32
186 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
187 {
188 	__be32 status;
189 
190 	if (open->op_truncate &&
191 		!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
192 		return nfserr_inval;
193 
194 	accmode |= NFSD_MAY_READ_IF_EXEC;
195 
196 	if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
197 		accmode |= NFSD_MAY_READ;
198 	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
199 		accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
200 	if (open->op_share_deny & NFS4_SHARE_DENY_READ)
201 		accmode |= NFSD_MAY_WRITE;
202 
203 	status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
204 
205 	return status;
206 }
207 
208 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
209 {
210 	umode_t mode = fh->fh_dentry->d_inode->i_mode;
211 
212 	if (S_ISREG(mode))
213 		return nfs_ok;
214 	if (S_ISDIR(mode))
215 		return nfserr_isdir;
216 	/*
217 	 * Using err_symlink as our catch-all case may look odd; but
218 	 * there's no other obvious error for this case in 4.0, and we
219 	 * happen to know that it will cause the linux v4 client to do
220 	 * the right thing on attempts to open something other than a
221 	 * regular file.
222 	 */
223 	return nfserr_symlink;
224 }
225 
226 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
227 {
228 	if (nfsd4_has_session(cstate))
229 		return;
230 	fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
231 			&resfh->fh_handle);
232 }
233 
234 static __be32
235 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
236 {
237 	struct svc_fh *current_fh = &cstate->current_fh;
238 	int accmode;
239 	__be32 status;
240 
241 	*resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
242 	if (!*resfh)
243 		return nfserr_jukebox;
244 	fh_init(*resfh, NFS4_FHSIZE);
245 	open->op_truncate = 0;
246 
247 	if (open->op_create) {
248 		/* FIXME: check session persistence and pnfs flags.
249 		 * The nfsv4.1 spec requires the following semantics:
250 		 *
251 		 * Persistent   | pNFS   | Server REQUIRED | Client Allowed
252 		 * Reply Cache  | server |                 |
253 		 * -------------+--------+-----------------+--------------------
254 		 * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
255 		 *              |        |                 | (SHOULD)
256 		 *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
257 		 *              |        |                 | (SHOULD NOT)
258 		 * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
259 		 * yes          | no     | GUARDED4        | GUARDED4
260 		 * yes          | yes    | GUARDED4        | GUARDED4
261 		 */
262 
263 		/*
264 		 * Note: create modes (UNCHECKED,GUARDED...) are the same
265 		 * in NFSv4 as in v3 except EXCLUSIVE4_1.
266 		 */
267 		status = do_nfsd_create(rqstp, current_fh, open->op_fname.data,
268 					open->op_fname.len, &open->op_iattr,
269 					*resfh, open->op_createmode,
270 					(u32 *)open->op_verf.data,
271 					&open->op_truncate, &open->op_created);
272 
273 		if (!status && open->op_label.len)
274 			nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
275 
276 		/*
277 		 * Following rfc 3530 14.2.16, use the returned bitmask
278 		 * to indicate which attributes we used to store the
279 		 * verifier:
280 		 */
281 		if (open->op_createmode == NFS4_CREATE_EXCLUSIVE && status == 0)
282 			open->op_bmval[1] = (FATTR4_WORD1_TIME_ACCESS |
283 							FATTR4_WORD1_TIME_MODIFY);
284 	} else
285 		/*
286 		 * Note this may exit with the parent still locked.
287 		 * We will hold the lock until nfsd4_open's final
288 		 * lookup, to prevent renames or unlinks until we've had
289 		 * a chance to an acquire a delegation if appropriate.
290 		 */
291 		status = nfsd_lookup(rqstp, current_fh,
292 				     open->op_fname.data, open->op_fname.len, *resfh);
293 	if (status)
294 		goto out;
295 	status = nfsd_check_obj_isreg(*resfh);
296 	if (status)
297 		goto out;
298 
299 	if (is_create_with_attrs(open) && open->op_acl != NULL)
300 		do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
301 
302 	nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
303 	accmode = NFSD_MAY_NOP;
304 	if (open->op_created ||
305 			open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
306 		accmode |= NFSD_MAY_OWNER_OVERRIDE;
307 	status = do_open_permission(rqstp, *resfh, open, accmode);
308 	set_change_info(&open->op_cinfo, current_fh);
309 out:
310 	return status;
311 }
312 
313 static __be32
314 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
315 {
316 	struct svc_fh *current_fh = &cstate->current_fh;
317 	__be32 status;
318 	int accmode = 0;
319 
320 	/* We don't know the target directory, and therefore can not
321 	* set the change info
322 	*/
323 
324 	memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
325 
326 	nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
327 
328 	open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
329 		(open->op_iattr.ia_size == 0);
330 	/*
331 	 * In the delegation case, the client is telling us about an
332 	 * open that it *already* performed locally, some time ago.  We
333 	 * should let it succeed now if possible.
334 	 *
335 	 * In the case of a CLAIM_FH open, on the other hand, the client
336 	 * may be counting on us to enforce permissions (the Linux 4.1
337 	 * client uses this for normal opens, for example).
338 	 */
339 	if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
340 		accmode = NFSD_MAY_OWNER_OVERRIDE;
341 
342 	status = do_open_permission(rqstp, current_fh, open, accmode);
343 
344 	return status;
345 }
346 
347 static void
348 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
349 {
350 	struct nfsd4_sessionid *sid =
351 			(struct nfsd4_sessionid *)session->se_sessionid.data;
352 
353 	clid->cl_boot = sid->clientid.cl_boot;
354 	clid->cl_id = sid->clientid.cl_id;
355 }
356 
357 static __be32
358 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
359 	   struct nfsd4_open *open)
360 {
361 	__be32 status;
362 	struct svc_fh *resfh = NULL;
363 	struct nfsd4_compoundres *resp;
364 	struct net *net = SVC_NET(rqstp);
365 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
366 
367 	dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
368 		(int)open->op_fname.len, open->op_fname.data,
369 		open->op_openowner);
370 
371 	/* This check required by spec. */
372 	if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
373 		return nfserr_inval;
374 
375 	open->op_created = 0;
376 	/*
377 	 * RFC5661 18.51.3
378 	 * Before RECLAIM_COMPLETE done, server should deny new lock
379 	 */
380 	if (nfsd4_has_session(cstate) &&
381 	    !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
382 		      &cstate->session->se_client->cl_flags) &&
383 	    open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
384 		return nfserr_grace;
385 
386 	if (nfsd4_has_session(cstate))
387 		copy_clientid(&open->op_clientid, cstate->session);
388 
389 	/* check seqid for replay. set nfs4_owner */
390 	resp = rqstp->rq_resp;
391 	status = nfsd4_process_open1(&resp->cstate, open, nn);
392 	if (status == nfserr_replay_me) {
393 		struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
394 		fh_put(&cstate->current_fh);
395 		fh_copy_shallow(&cstate->current_fh.fh_handle,
396 				&rp->rp_openfh);
397 		status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
398 		if (status)
399 			dprintk("nfsd4_open: replay failed"
400 				" restoring previous filehandle\n");
401 		else
402 			status = nfserr_replay_me;
403 	}
404 	if (status)
405 		goto out;
406 	if (open->op_xdr_error) {
407 		status = open->op_xdr_error;
408 		goto out;
409 	}
410 
411 	status = nfsd4_check_open_attributes(rqstp, cstate, open);
412 	if (status)
413 		goto out;
414 
415 	/* Openowner is now set, so sequence id will get bumped.  Now we need
416 	 * these checks before we do any creates: */
417 	status = nfserr_grace;
418 	if (locks_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
419 		goto out;
420 	status = nfserr_no_grace;
421 	if (!locks_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
422 		goto out;
423 
424 	switch (open->op_claim_type) {
425 		case NFS4_OPEN_CLAIM_DELEGATE_CUR:
426 		case NFS4_OPEN_CLAIM_NULL:
427 			status = do_open_lookup(rqstp, cstate, open, &resfh);
428 			if (status)
429 				goto out;
430 			break;
431 		case NFS4_OPEN_CLAIM_PREVIOUS:
432 			status = nfs4_check_open_reclaim(&open->op_clientid,
433 							 cstate, nn);
434 			if (status)
435 				goto out;
436 			open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
437 		case NFS4_OPEN_CLAIM_FH:
438 		case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
439 			status = do_open_fhandle(rqstp, cstate, open);
440 			if (status)
441 				goto out;
442 			resfh = &cstate->current_fh;
443 			break;
444 		case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
445              	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
446 			dprintk("NFSD: unsupported OPEN claim type %d\n",
447 				open->op_claim_type);
448 			status = nfserr_notsupp;
449 			goto out;
450 		default:
451 			dprintk("NFSD: Invalid OPEN claim type %d\n",
452 				open->op_claim_type);
453 			status = nfserr_inval;
454 			goto out;
455 	}
456 	/*
457 	 * nfsd4_process_open2() does the actual opening of the file.  If
458 	 * successful, it (1) truncates the file if open->op_truncate was
459 	 * set, (2) sets open->op_stateid, (3) sets open->op_delegation.
460 	 */
461 	status = nfsd4_process_open2(rqstp, resfh, open);
462 	WARN(status && open->op_created,
463 	     "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
464 	     be32_to_cpu(status));
465 out:
466 	if (resfh && resfh != &cstate->current_fh) {
467 		fh_dup2(&cstate->current_fh, resfh);
468 		fh_put(resfh);
469 		kfree(resfh);
470 	}
471 	nfsd4_cleanup_open_state(cstate, open, status);
472 	nfsd4_bump_seqid(cstate, status);
473 	return status;
474 }
475 
476 /*
477  * OPEN is the only seqid-mutating operation whose decoding can fail
478  * with a seqid-mutating error (specifically, decoding of user names in
479  * the attributes).  Therefore we have to do some processing to look up
480  * the stateowner so that we can bump the seqid.
481  */
482 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
483 {
484 	struct nfsd4_open *open = (struct nfsd4_open *)&op->u;
485 
486 	if (!seqid_mutating_err(ntohl(op->status)))
487 		return op->status;
488 	if (nfsd4_has_session(cstate))
489 		return op->status;
490 	open->op_xdr_error = op->status;
491 	return nfsd4_open(rqstp, cstate, open);
492 }
493 
494 /*
495  * filehandle-manipulating ops.
496  */
497 static __be32
498 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
499 	    struct svc_fh **getfh)
500 {
501 	if (!cstate->current_fh.fh_dentry)
502 		return nfserr_nofilehandle;
503 
504 	*getfh = &cstate->current_fh;
505 	return nfs_ok;
506 }
507 
508 static __be32
509 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
510 	    struct nfsd4_putfh *putfh)
511 {
512 	fh_put(&cstate->current_fh);
513 	cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
514 	memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
515 	       putfh->pf_fhlen);
516 	return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
517 }
518 
519 static __be32
520 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
521 		void *arg)
522 {
523 	__be32 status;
524 
525 	fh_put(&cstate->current_fh);
526 	status = exp_pseudoroot(rqstp, &cstate->current_fh);
527 	return status;
528 }
529 
530 static __be32
531 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
532 		void *arg)
533 {
534 	if (!cstate->save_fh.fh_dentry)
535 		return nfserr_restorefh;
536 
537 	fh_dup2(&cstate->current_fh, &cstate->save_fh);
538 	if (HAS_STATE_ID(cstate, SAVED_STATE_ID_FLAG)) {
539 		memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
540 		SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
541 	}
542 	return nfs_ok;
543 }
544 
545 static __be32
546 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
547 	     void *arg)
548 {
549 	if (!cstate->current_fh.fh_dentry)
550 		return nfserr_nofilehandle;
551 
552 	fh_dup2(&cstate->save_fh, &cstate->current_fh);
553 	if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG)) {
554 		memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
555 		SET_STATE_ID(cstate, SAVED_STATE_ID_FLAG);
556 	}
557 	return nfs_ok;
558 }
559 
560 /*
561  * misc nfsv4 ops
562  */
563 static __be32
564 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
565 	     struct nfsd4_access *access)
566 {
567 	if (access->ac_req_access & ~NFS3_ACCESS_FULL)
568 		return nfserr_inval;
569 
570 	access->ac_resp_access = access->ac_req_access;
571 	return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
572 			   &access->ac_supported);
573 }
574 
575 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
576 {
577 	__be32 verf[2];
578 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
579 
580 	/*
581 	 * This is opaque to client, so no need to byte-swap. Use
582 	 * __force to keep sparse happy
583 	 */
584 	verf[0] = (__force __be32)nn->nfssvc_boot.tv_sec;
585 	verf[1] = (__force __be32)nn->nfssvc_boot.tv_usec;
586 	memcpy(verifier->data, verf, sizeof(verifier->data));
587 }
588 
589 static __be32
590 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
591 	     struct nfsd4_commit *commit)
592 {
593 	gen_boot_verifier(&commit->co_verf, SVC_NET(rqstp));
594 	return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
595 			     commit->co_count);
596 }
597 
598 static __be32
599 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
600 	     struct nfsd4_create *create)
601 {
602 	struct svc_fh resfh;
603 	__be32 status;
604 	dev_t rdev;
605 
606 	fh_init(&resfh, NFS4_FHSIZE);
607 
608 	status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR,
609 			   NFSD_MAY_CREATE);
610 	if (status)
611 		return status;
612 
613 	status = check_attr_support(rqstp, cstate, create->cr_bmval,
614 				    nfsd_attrmask);
615 	if (status)
616 		return status;
617 
618 	switch (create->cr_type) {
619 	case NF4LNK:
620 		status = nfsd_symlink(rqstp, &cstate->current_fh,
621 				      create->cr_name, create->cr_namelen,
622 				      create->cr_data, &resfh);
623 		break;
624 
625 	case NF4BLK:
626 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
627 		if (MAJOR(rdev) != create->cr_specdata1 ||
628 		    MINOR(rdev) != create->cr_specdata2)
629 			return nfserr_inval;
630 		status = nfsd_create(rqstp, &cstate->current_fh,
631 				     create->cr_name, create->cr_namelen,
632 				     &create->cr_iattr, S_IFBLK, rdev, &resfh);
633 		break;
634 
635 	case NF4CHR:
636 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
637 		if (MAJOR(rdev) != create->cr_specdata1 ||
638 		    MINOR(rdev) != create->cr_specdata2)
639 			return nfserr_inval;
640 		status = nfsd_create(rqstp, &cstate->current_fh,
641 				     create->cr_name, create->cr_namelen,
642 				     &create->cr_iattr,S_IFCHR, rdev, &resfh);
643 		break;
644 
645 	case NF4SOCK:
646 		status = nfsd_create(rqstp, &cstate->current_fh,
647 				     create->cr_name, create->cr_namelen,
648 				     &create->cr_iattr, S_IFSOCK, 0, &resfh);
649 		break;
650 
651 	case NF4FIFO:
652 		status = nfsd_create(rqstp, &cstate->current_fh,
653 				     create->cr_name, create->cr_namelen,
654 				     &create->cr_iattr, S_IFIFO, 0, &resfh);
655 		break;
656 
657 	case NF4DIR:
658 		create->cr_iattr.ia_valid &= ~ATTR_SIZE;
659 		status = nfsd_create(rqstp, &cstate->current_fh,
660 				     create->cr_name, create->cr_namelen,
661 				     &create->cr_iattr, S_IFDIR, 0, &resfh);
662 		break;
663 
664 	default:
665 		status = nfserr_badtype;
666 	}
667 
668 	if (status)
669 		goto out;
670 
671 	if (create->cr_label.len)
672 		nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
673 
674 	if (create->cr_acl != NULL)
675 		do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
676 				create->cr_bmval);
677 
678 	fh_unlock(&cstate->current_fh);
679 	set_change_info(&create->cr_cinfo, &cstate->current_fh);
680 	fh_dup2(&cstate->current_fh, &resfh);
681 out:
682 	fh_put(&resfh);
683 	return status;
684 }
685 
686 static __be32
687 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
688 	      struct nfsd4_getattr *getattr)
689 {
690 	__be32 status;
691 
692 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
693 	if (status)
694 		return status;
695 
696 	if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
697 		return nfserr_inval;
698 
699 	getattr->ga_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
700 	getattr->ga_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
701 	getattr->ga_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
702 
703 	getattr->ga_fhp = &cstate->current_fh;
704 	return nfs_ok;
705 }
706 
707 static __be32
708 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
709 	   struct nfsd4_link *link)
710 {
711 	__be32 status = nfserr_nofilehandle;
712 
713 	if (!cstate->save_fh.fh_dentry)
714 		return status;
715 	status = nfsd_link(rqstp, &cstate->current_fh,
716 			   link->li_name, link->li_namelen, &cstate->save_fh);
717 	if (!status)
718 		set_change_info(&link->li_cinfo, &cstate->current_fh);
719 	return status;
720 }
721 
722 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
723 {
724 	struct svc_fh tmp_fh;
725 	__be32 ret;
726 
727 	fh_init(&tmp_fh, NFS4_FHSIZE);
728 	ret = exp_pseudoroot(rqstp, &tmp_fh);
729 	if (ret)
730 		return ret;
731 	if (tmp_fh.fh_dentry == fh->fh_dentry) {
732 		fh_put(&tmp_fh);
733 		return nfserr_noent;
734 	}
735 	fh_put(&tmp_fh);
736 	return nfsd_lookup(rqstp, fh, "..", 2, fh);
737 }
738 
739 static __be32
740 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
741 	      void *arg)
742 {
743 	return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
744 }
745 
746 static __be32
747 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
748 	     struct nfsd4_lookup *lookup)
749 {
750 	return nfsd_lookup(rqstp, &cstate->current_fh,
751 			   lookup->lo_name, lookup->lo_len,
752 			   &cstate->current_fh);
753 }
754 
755 static __be32
756 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
757 	   struct nfsd4_read *read)
758 {
759 	__be32 status;
760 
761 	/* no need to check permission - this will be done in nfsd_read() */
762 
763 	read->rd_filp = NULL;
764 	if (read->rd_offset >= OFFSET_MAX)
765 		return nfserr_inval;
766 
767 	/*
768 	 * If we do a zero copy read, then a client will see read data
769 	 * that reflects the state of the file *after* performing the
770 	 * following compound.
771 	 *
772 	 * To ensure proper ordering, we therefore turn off zero copy if
773 	 * the client wants us to do more in this compound:
774 	 */
775 	if (!nfsd4_last_compound_op(rqstp))
776 		clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
777 
778 	/* check stateid */
779 	if ((status = nfs4_preprocess_stateid_op(SVC_NET(rqstp),
780 						 cstate, &read->rd_stateid,
781 						 RD_STATE, &read->rd_filp))) {
782 		dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
783 		goto out;
784 	}
785 	status = nfs_ok;
786 out:
787 	read->rd_rqstp = rqstp;
788 	read->rd_fhp = &cstate->current_fh;
789 	return status;
790 }
791 
792 static __be32
793 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
794 	      struct nfsd4_readdir *readdir)
795 {
796 	u64 cookie = readdir->rd_cookie;
797 	static const nfs4_verifier zeroverf;
798 
799 	/* no need to check permission - this will be done in nfsd_readdir() */
800 
801 	if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
802 		return nfserr_inval;
803 
804 	readdir->rd_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
805 	readdir->rd_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
806 	readdir->rd_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
807 
808 	if ((cookie == 1) || (cookie == 2) ||
809 	    (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
810 		return nfserr_bad_cookie;
811 
812 	readdir->rd_rqstp = rqstp;
813 	readdir->rd_fhp = &cstate->current_fh;
814 	return nfs_ok;
815 }
816 
817 static __be32
818 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
819 	       struct nfsd4_readlink *readlink)
820 {
821 	readlink->rl_rqstp = rqstp;
822 	readlink->rl_fhp = &cstate->current_fh;
823 	return nfs_ok;
824 }
825 
826 static __be32
827 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
828 	     struct nfsd4_remove *remove)
829 {
830 	__be32 status;
831 
832 	if (locks_in_grace(SVC_NET(rqstp)))
833 		return nfserr_grace;
834 	status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
835 			     remove->rm_name, remove->rm_namelen);
836 	if (!status) {
837 		fh_unlock(&cstate->current_fh);
838 		set_change_info(&remove->rm_cinfo, &cstate->current_fh);
839 	}
840 	return status;
841 }
842 
843 static __be32
844 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
845 	     struct nfsd4_rename *rename)
846 {
847 	__be32 status = nfserr_nofilehandle;
848 
849 	if (!cstate->save_fh.fh_dentry)
850 		return status;
851 	if (locks_in_grace(SVC_NET(rqstp)) &&
852 		!(cstate->save_fh.fh_export->ex_flags & NFSEXP_NOSUBTREECHECK))
853 		return nfserr_grace;
854 	status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
855 			     rename->rn_snamelen, &cstate->current_fh,
856 			     rename->rn_tname, rename->rn_tnamelen);
857 	if (status)
858 		return status;
859 	set_change_info(&rename->rn_sinfo, &cstate->current_fh);
860 	set_change_info(&rename->rn_tinfo, &cstate->save_fh);
861 	return nfs_ok;
862 }
863 
864 static __be32
865 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
866 	      struct nfsd4_secinfo *secinfo)
867 {
868 	struct svc_fh resfh;
869 	struct svc_export *exp;
870 	struct dentry *dentry;
871 	__be32 err;
872 
873 	fh_init(&resfh, NFS4_FHSIZE);
874 	err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
875 	if (err)
876 		return err;
877 	err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
878 				    secinfo->si_name, secinfo->si_namelen,
879 				    &exp, &dentry);
880 	if (err)
881 		return err;
882 	if (dentry->d_inode == NULL) {
883 		exp_put(exp);
884 		err = nfserr_noent;
885 	} else
886 		secinfo->si_exp = exp;
887 	dput(dentry);
888 	if (cstate->minorversion)
889 		/* See rfc 5661 section 2.6.3.1.1.8 */
890 		fh_put(&cstate->current_fh);
891 	return err;
892 }
893 
894 static __be32
895 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
896 	      struct nfsd4_secinfo_no_name *sin)
897 {
898 	__be32 err;
899 
900 	switch (sin->sin_style) {
901 	case NFS4_SECINFO_STYLE4_CURRENT_FH:
902 		break;
903 	case NFS4_SECINFO_STYLE4_PARENT:
904 		err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
905 		if (err)
906 			return err;
907 		break;
908 	default:
909 		return nfserr_inval;
910 	}
911 
912 	sin->sin_exp = exp_get(cstate->current_fh.fh_export);
913 	fh_put(&cstate->current_fh);
914 	return nfs_ok;
915 }
916 
917 static __be32
918 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
919 	      struct nfsd4_setattr *setattr)
920 {
921 	__be32 status = nfs_ok;
922 	int err;
923 
924 	if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
925 		status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), cstate,
926 			&setattr->sa_stateid, WR_STATE, NULL);
927 		if (status) {
928 			dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
929 			return status;
930 		}
931 	}
932 	err = fh_want_write(&cstate->current_fh);
933 	if (err)
934 		return nfserrno(err);
935 	status = nfs_ok;
936 
937 	status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
938 				    nfsd_attrmask);
939 	if (status)
940 		goto out;
941 
942 	if (setattr->sa_acl != NULL)
943 		status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
944 					    setattr->sa_acl);
945 	if (status)
946 		goto out;
947 	if (setattr->sa_label.len)
948 		status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
949 				&setattr->sa_label);
950 	if (status)
951 		goto out;
952 	status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
953 				0, (time_t)0);
954 out:
955 	fh_drop_write(&cstate->current_fh);
956 	return status;
957 }
958 
959 static int fill_in_write_vector(struct kvec *vec, struct nfsd4_write *write)
960 {
961         int i = 1;
962         int buflen = write->wr_buflen;
963 
964         vec[0].iov_base = write->wr_head.iov_base;
965         vec[0].iov_len = min_t(int, buflen, write->wr_head.iov_len);
966         buflen -= vec[0].iov_len;
967 
968         while (buflen) {
969                 vec[i].iov_base = page_address(write->wr_pagelist[i - 1]);
970                 vec[i].iov_len = min_t(int, PAGE_SIZE, buflen);
971                 buflen -= vec[i].iov_len;
972                 i++;
973         }
974         return i;
975 }
976 
977 static __be32
978 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
979 	    struct nfsd4_write *write)
980 {
981 	stateid_t *stateid = &write->wr_stateid;
982 	struct file *filp = NULL;
983 	__be32 status = nfs_ok;
984 	unsigned long cnt;
985 	int nvecs;
986 
987 	/* no need to check permission - this will be done in nfsd_write() */
988 
989 	if (write->wr_offset >= OFFSET_MAX)
990 		return nfserr_inval;
991 
992 	status = nfs4_preprocess_stateid_op(SVC_NET(rqstp),
993 					cstate, stateid, WR_STATE, &filp);
994 	if (status) {
995 		dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
996 		return status;
997 	}
998 
999 	cnt = write->wr_buflen;
1000 	write->wr_how_written = write->wr_stable_how;
1001 	gen_boot_verifier(&write->wr_verifier, SVC_NET(rqstp));
1002 
1003 	nvecs = fill_in_write_vector(rqstp->rq_vec, write);
1004 	WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1005 
1006 	status =  nfsd_write(rqstp, &cstate->current_fh, filp,
1007 			     write->wr_offset, rqstp->rq_vec, nvecs,
1008 			     &cnt, &write->wr_how_written);
1009 	if (filp)
1010 		fput(filp);
1011 
1012 	write->wr_bytes_written = cnt;
1013 
1014 	return status;
1015 }
1016 
1017 static __be32
1018 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1019 		struct nfsd4_fallocate *fallocate, int flags)
1020 {
1021 	__be32 status = nfserr_notsupp;
1022 	struct file *file;
1023 
1024 	status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), cstate,
1025 					    &fallocate->falloc_stateid,
1026 					    WR_STATE, &file);
1027 	if (status != nfs_ok) {
1028 		dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1029 		return status;
1030 	}
1031 
1032 	status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, file,
1033 				     fallocate->falloc_offset,
1034 				     fallocate->falloc_length,
1035 				     flags);
1036 	fput(file);
1037 	return status;
1038 }
1039 
1040 static __be32
1041 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1042 	       struct nfsd4_fallocate *fallocate)
1043 {
1044 	return nfsd4_fallocate(rqstp, cstate, fallocate, 0);
1045 }
1046 
1047 static __be32
1048 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1049 		 struct nfsd4_fallocate *fallocate)
1050 {
1051 	return nfsd4_fallocate(rqstp, cstate, fallocate,
1052 			       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
1053 }
1054 
1055 static __be32
1056 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1057 		struct nfsd4_seek *seek)
1058 {
1059 	int whence;
1060 	__be32 status;
1061 	struct file *file;
1062 
1063 	status = nfs4_preprocess_stateid_op(SVC_NET(rqstp), cstate,
1064 					    &seek->seek_stateid,
1065 					    RD_STATE, &file);
1066 	if (status) {
1067 		dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
1068 		return status;
1069 	}
1070 
1071 	switch (seek->seek_whence) {
1072 	case NFS4_CONTENT_DATA:
1073 		whence = SEEK_DATA;
1074 		break;
1075 	case NFS4_CONTENT_HOLE:
1076 		whence = SEEK_HOLE;
1077 		break;
1078 	default:
1079 		status = nfserr_union_notsupp;
1080 		goto out;
1081 	}
1082 
1083 	/*
1084 	 * Note:  This call does change file->f_pos, but nothing in NFSD
1085 	 *        should ever file->f_pos.
1086 	 */
1087 	seek->seek_pos = vfs_llseek(file, seek->seek_offset, whence);
1088 	if (seek->seek_pos < 0)
1089 		status = nfserrno(seek->seek_pos);
1090 	else if (seek->seek_pos >= i_size_read(file_inode(file)))
1091 		seek->seek_eof = true;
1092 
1093 out:
1094 	fput(file);
1095 	return status;
1096 }
1097 
1098 /* This routine never returns NFS_OK!  If there are no other errors, it
1099  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
1100  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
1101  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
1102  */
1103 static __be32
1104 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1105 	     struct nfsd4_verify *verify)
1106 {
1107 	__be32 *buf, *p;
1108 	int count;
1109 	__be32 status;
1110 
1111 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
1112 	if (status)
1113 		return status;
1114 
1115 	status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
1116 	if (status)
1117 		return status;
1118 
1119 	if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
1120 	    || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
1121 		return nfserr_inval;
1122 	if (verify->ve_attrlen & 3)
1123 		return nfserr_inval;
1124 
1125 	/* count in words:
1126 	 *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
1127 	 */
1128 	count = 4 + (verify->ve_attrlen >> 2);
1129 	buf = kmalloc(count << 2, GFP_KERNEL);
1130 	if (!buf)
1131 		return nfserr_jukebox;
1132 
1133 	p = buf;
1134 	status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
1135 				    cstate->current_fh.fh_export,
1136 				    cstate->current_fh.fh_dentry,
1137 				    verify->ve_bmval,
1138 				    rqstp, 0);
1139 	/*
1140 	 * If nfsd4_encode_fattr() ran out of space, assume that's because
1141 	 * the attributes are longer (hence different) than those given:
1142 	 */
1143 	if (status == nfserr_resource)
1144 		status = nfserr_not_same;
1145 	if (status)
1146 		goto out_kfree;
1147 
1148 	/* skip bitmap */
1149 	p = buf + 1 + ntohl(buf[0]);
1150 	status = nfserr_not_same;
1151 	if (ntohl(*p++) != verify->ve_attrlen)
1152 		goto out_kfree;
1153 	if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
1154 		status = nfserr_same;
1155 
1156 out_kfree:
1157 	kfree(buf);
1158 	return status;
1159 }
1160 
1161 static __be32
1162 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1163 	      struct nfsd4_verify *verify)
1164 {
1165 	__be32 status;
1166 
1167 	status = _nfsd4_verify(rqstp, cstate, verify);
1168 	return status == nfserr_not_same ? nfs_ok : status;
1169 }
1170 
1171 static __be32
1172 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1173 	     struct nfsd4_verify *verify)
1174 {
1175 	__be32 status;
1176 
1177 	status = _nfsd4_verify(rqstp, cstate, verify);
1178 	return status == nfserr_same ? nfs_ok : status;
1179 }
1180 
1181 /*
1182  * NULL call.
1183  */
1184 static __be32
1185 nfsd4_proc_null(struct svc_rqst *rqstp, void *argp, void *resp)
1186 {
1187 	return nfs_ok;
1188 }
1189 
1190 static inline void nfsd4_increment_op_stats(u32 opnum)
1191 {
1192 	if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
1193 		nfsdstats.nfs4_opcount[opnum]++;
1194 }
1195 
1196 typedef __be32(*nfsd4op_func)(struct svc_rqst *, struct nfsd4_compound_state *,
1197 			      void *);
1198 typedef u32(*nfsd4op_rsize)(struct svc_rqst *, struct nfsd4_op *op);
1199 typedef void(*stateid_setter)(struct nfsd4_compound_state *, void *);
1200 typedef void(*stateid_getter)(struct nfsd4_compound_state *, void *);
1201 
1202 enum nfsd4_op_flags {
1203 	ALLOWED_WITHOUT_FH = 1 << 0,	/* No current filehandle required */
1204 	ALLOWED_ON_ABSENT_FS = 1 << 1,	/* ops processed on absent fs */
1205 	ALLOWED_AS_FIRST_OP = 1 << 2,	/* ops reqired first in compound */
1206 	/* For rfc 5661 section 2.6.3.1.1: */
1207 	OP_HANDLES_WRONGSEC = 1 << 3,
1208 	OP_IS_PUTFH_LIKE = 1 << 4,
1209 	/*
1210 	 * These are the ops whose result size we estimate before
1211 	 * encoding, to avoid performing an op then not being able to
1212 	 * respond or cache a response.  This includes writes and setattrs
1213 	 * as well as the operations usually called "nonidempotent":
1214 	 */
1215 	OP_MODIFIES_SOMETHING = 1 << 5,
1216 	/*
1217 	 * Cache compounds containing these ops in the xid-based drc:
1218 	 * We use the DRC for compounds containing non-idempotent
1219 	 * operations, *except* those that are 4.1-specific (since
1220 	 * sessions provide their own EOS), and except for stateful
1221 	 * operations other than setclientid and setclientid_confirm
1222 	 * (since sequence numbers provide EOS for open, lock, etc in
1223 	 * the v4.0 case).
1224 	 */
1225 	OP_CACHEME = 1 << 6,
1226 	/*
1227 	 * These are ops which clear current state id.
1228 	 */
1229 	OP_CLEAR_STATEID = 1 << 7,
1230 };
1231 
1232 struct nfsd4_operation {
1233 	nfsd4op_func op_func;
1234 	u32 op_flags;
1235 	char *op_name;
1236 	/* Try to get response size before operation */
1237 	nfsd4op_rsize op_rsize_bop;
1238 	stateid_getter op_get_currentstateid;
1239 	stateid_setter op_set_currentstateid;
1240 };
1241 
1242 static struct nfsd4_operation nfsd4_ops[];
1243 
1244 static const char *nfsd4_op_name(unsigned opnum);
1245 
1246 /*
1247  * Enforce NFSv4.1 COMPOUND ordering rules:
1248  *
1249  * Also note, enforced elsewhere:
1250  *	- SEQUENCE other than as first op results in
1251  *	  NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
1252  *	- BIND_CONN_TO_SESSION must be the only op in its compound.
1253  *	  (Enforced in nfsd4_bind_conn_to_session().)
1254  *	- DESTROY_SESSION must be the final operation in a compound, if
1255  *	  sessionid's in SEQUENCE and DESTROY_SESSION are the same.
1256  *	  (Enforced in nfsd4_destroy_session().)
1257  */
1258 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
1259 {
1260 	struct nfsd4_op *op = &args->ops[0];
1261 
1262 	/* These ordering requirements don't apply to NFSv4.0: */
1263 	if (args->minorversion == 0)
1264 		return nfs_ok;
1265 	/* This is weird, but OK, not our problem: */
1266 	if (args->opcnt == 0)
1267 		return nfs_ok;
1268 	if (op->status == nfserr_op_illegal)
1269 		return nfs_ok;
1270 	if (!(nfsd4_ops[op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
1271 		return nfserr_op_not_in_session;
1272 	if (op->opnum == OP_SEQUENCE)
1273 		return nfs_ok;
1274 	if (args->opcnt != 1)
1275 		return nfserr_not_only_op;
1276 	return nfs_ok;
1277 }
1278 
1279 static inline struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
1280 {
1281 	return &nfsd4_ops[op->opnum];
1282 }
1283 
1284 bool nfsd4_cache_this_op(struct nfsd4_op *op)
1285 {
1286 	if (op->opnum == OP_ILLEGAL)
1287 		return false;
1288 	return OPDESC(op)->op_flags & OP_CACHEME;
1289 }
1290 
1291 static bool need_wrongsec_check(struct svc_rqst *rqstp)
1292 {
1293 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
1294 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
1295 	struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
1296 	struct nfsd4_op *next = &argp->ops[resp->opcnt];
1297 	struct nfsd4_operation *thisd;
1298 	struct nfsd4_operation *nextd;
1299 
1300 	thisd = OPDESC(this);
1301 	/*
1302 	 * Most ops check wronsec on our own; only the putfh-like ops
1303 	 * have special rules.
1304 	 */
1305 	if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
1306 		return false;
1307 	/*
1308 	 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
1309 	 * put-filehandle operation if we're not going to use the
1310 	 * result:
1311 	 */
1312 	if (argp->opcnt == resp->opcnt)
1313 		return false;
1314 	if (next->opnum == OP_ILLEGAL)
1315 		return false;
1316 	nextd = OPDESC(next);
1317 	/*
1318 	 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
1319 	 * errors themselves as necessary; others should check for them
1320 	 * now:
1321 	 */
1322 	return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
1323 }
1324 
1325 static void svcxdr_init_encode(struct svc_rqst *rqstp,
1326 			       struct nfsd4_compoundres *resp)
1327 {
1328 	struct xdr_stream *xdr = &resp->xdr;
1329 	struct xdr_buf *buf = &rqstp->rq_res;
1330 	struct kvec *head = buf->head;
1331 
1332 	xdr->buf = buf;
1333 	xdr->iov = head;
1334 	xdr->p   = head->iov_base + head->iov_len;
1335 	xdr->end = head->iov_base + PAGE_SIZE - rqstp->rq_auth_slack;
1336 	/* Tail and page_len should be zero at this point: */
1337 	buf->len = buf->head[0].iov_len;
1338 	xdr->scratch.iov_len = 0;
1339 	xdr->page_ptr = buf->pages - 1;
1340 	buf->buflen = PAGE_SIZE * (1 + rqstp->rq_page_end - buf->pages)
1341 		- rqstp->rq_auth_slack;
1342 }
1343 
1344 /*
1345  * COMPOUND call.
1346  */
1347 static __be32
1348 nfsd4_proc_compound(struct svc_rqst *rqstp,
1349 		    struct nfsd4_compoundargs *args,
1350 		    struct nfsd4_compoundres *resp)
1351 {
1352 	struct nfsd4_op	*op;
1353 	struct nfsd4_operation *opdesc;
1354 	struct nfsd4_compound_state *cstate = &resp->cstate;
1355 	struct svc_fh *current_fh = &cstate->current_fh;
1356 	struct svc_fh *save_fh = &cstate->save_fh;
1357 	__be32		status;
1358 
1359 	svcxdr_init_encode(rqstp, resp);
1360 	resp->tagp = resp->xdr.p;
1361 	/* reserve space for: taglen, tag, and opcnt */
1362 	xdr_reserve_space(&resp->xdr, 8 + args->taglen);
1363 	resp->taglen = args->taglen;
1364 	resp->tag = args->tag;
1365 	resp->rqstp = rqstp;
1366 	cstate->minorversion = args->minorversion;
1367 	fh_init(current_fh, NFS4_FHSIZE);
1368 	fh_init(save_fh, NFS4_FHSIZE);
1369 	/*
1370 	 * Don't use the deferral mechanism for NFSv4; compounds make it
1371 	 * too hard to avoid non-idempotency problems.
1372 	 */
1373 	clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1374 
1375 	/*
1376 	 * According to RFC3010, this takes precedence over all other errors.
1377 	 */
1378 	status = nfserr_minor_vers_mismatch;
1379 	if (nfsd_minorversion(args->minorversion, NFSD_TEST) <= 0)
1380 		goto out;
1381 
1382 	status = nfs41_check_op_ordering(args);
1383 	if (status) {
1384 		op = &args->ops[0];
1385 		op->status = status;
1386 		goto encode_op;
1387 	}
1388 
1389 	while (!status && resp->opcnt < args->opcnt) {
1390 		op = &args->ops[resp->opcnt++];
1391 
1392 		dprintk("nfsv4 compound op #%d/%d: %d (%s)\n",
1393 			resp->opcnt, args->opcnt, op->opnum,
1394 			nfsd4_op_name(op->opnum));
1395 		/*
1396 		 * The XDR decode routines may have pre-set op->status;
1397 		 * for example, if there is a miscellaneous XDR error
1398 		 * it will be set to nfserr_bad_xdr.
1399 		 */
1400 		if (op->status) {
1401 			if (op->opnum == OP_OPEN)
1402 				op->status = nfsd4_open_omfg(rqstp, cstate, op);
1403 			goto encode_op;
1404 		}
1405 
1406 		opdesc = OPDESC(op);
1407 
1408 		if (!current_fh->fh_dentry) {
1409 			if (!(opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
1410 				op->status = nfserr_nofilehandle;
1411 				goto encode_op;
1412 			}
1413 		} else if (current_fh->fh_export->ex_fslocs.migrated &&
1414 			  !(opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
1415 			op->status = nfserr_moved;
1416 			goto encode_op;
1417 		}
1418 
1419 		fh_clear_wcc(current_fh);
1420 
1421 		/* If op is non-idempotent */
1422 		if (opdesc->op_flags & OP_MODIFIES_SOMETHING) {
1423 			/*
1424 			 * Don't execute this op if we couldn't encode a
1425 			 * succesful reply:
1426 			 */
1427 			u32 plen = opdesc->op_rsize_bop(rqstp, op);
1428 			/*
1429 			 * Plus if there's another operation, make sure
1430 			 * we'll have space to at least encode an error:
1431 			 */
1432 			if (resp->opcnt < args->opcnt)
1433 				plen += COMPOUND_ERR_SLACK_SPACE;
1434 			op->status = nfsd4_check_resp_size(resp, plen);
1435 		}
1436 
1437 		if (op->status)
1438 			goto encode_op;
1439 
1440 		if (opdesc->op_get_currentstateid)
1441 			opdesc->op_get_currentstateid(cstate, &op->u);
1442 		op->status = opdesc->op_func(rqstp, cstate, &op->u);
1443 
1444 		if (!op->status) {
1445 			if (opdesc->op_set_currentstateid)
1446 				opdesc->op_set_currentstateid(cstate, &op->u);
1447 
1448 			if (opdesc->op_flags & OP_CLEAR_STATEID)
1449 				clear_current_stateid(cstate);
1450 
1451 			if (need_wrongsec_check(rqstp))
1452 				op->status = check_nfsd_access(current_fh->fh_export, rqstp);
1453 		}
1454 
1455 encode_op:
1456 		/* Only from SEQUENCE */
1457 		if (cstate->status == nfserr_replay_cache) {
1458 			dprintk("%s NFS4.1 replay from cache\n", __func__);
1459 			status = op->status;
1460 			goto out;
1461 		}
1462 		if (op->status == nfserr_replay_me) {
1463 			op->replay = &cstate->replay_owner->so_replay;
1464 			nfsd4_encode_replay(&resp->xdr, op);
1465 			status = op->status = op->replay->rp_status;
1466 		} else {
1467 			nfsd4_encode_operation(resp, op);
1468 			status = op->status;
1469 		}
1470 
1471 		dprintk("nfsv4 compound op %p opcnt %d #%d: %d: status %d\n",
1472 			args->ops, args->opcnt, resp->opcnt, op->opnum,
1473 			be32_to_cpu(status));
1474 
1475 		nfsd4_cstate_clear_replay(cstate);
1476 		/* XXX Ugh, we need to get rid of this kind of special case: */
1477 		if (op->opnum == OP_READ && op->u.read.rd_filp)
1478 			fput(op->u.read.rd_filp);
1479 
1480 		nfsd4_increment_op_stats(op->opnum);
1481 	}
1482 
1483 	cstate->status = status;
1484 	fh_put(current_fh);
1485 	fh_put(save_fh);
1486 	BUG_ON(cstate->replay_owner);
1487 out:
1488 	/* Reset deferral mechanism for RPC deferrals */
1489 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
1490 	dprintk("nfsv4 compound returned %d\n", ntohl(status));
1491 	return status;
1492 }
1493 
1494 #define op_encode_hdr_size		(2)
1495 #define op_encode_stateid_maxsz		(XDR_QUADLEN(NFS4_STATEID_SIZE))
1496 #define op_encode_verifier_maxsz	(XDR_QUADLEN(NFS4_VERIFIER_SIZE))
1497 #define op_encode_change_info_maxsz	(5)
1498 #define nfs4_fattr_bitmap_maxsz		(4)
1499 
1500 /* We'll fall back on returning no lockowner if run out of space: */
1501 #define op_encode_lockowner_maxsz	(0)
1502 #define op_encode_lock_denied_maxsz	(8 + op_encode_lockowner_maxsz)
1503 
1504 #define nfs4_owner_maxsz		(1 + XDR_QUADLEN(IDMAP_NAMESZ))
1505 
1506 #define op_encode_ace_maxsz		(3 + nfs4_owner_maxsz)
1507 #define op_encode_delegation_maxsz	(1 + op_encode_stateid_maxsz + 1 + \
1508 					 op_encode_ace_maxsz)
1509 
1510 #define op_encode_channel_attrs_maxsz	(6 + 1 + 1)
1511 
1512 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1513 {
1514 	return (op_encode_hdr_size) * sizeof(__be32);
1515 }
1516 
1517 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1518 {
1519 	return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
1520 }
1521 
1522 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1523 {
1524 	return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
1525 }
1526 
1527 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1528 {
1529 	return (op_encode_hdr_size + op_encode_change_info_maxsz
1530 		+ nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
1531 }
1532 
1533 /*
1534  * Note since this is an idempotent operation we won't insist on failing
1535  * the op prematurely if the estimate is too large.  We may turn off splice
1536  * reads unnecessarily.
1537  */
1538 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
1539 				      struct nfsd4_op *op)
1540 {
1541 	u32 *bmap = op->u.getattr.ga_bmval;
1542 	u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
1543 	u32 ret = 0;
1544 
1545 	if (bmap0 & FATTR4_WORD0_ACL)
1546 		return svc_max_payload(rqstp);
1547 	if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
1548 		return svc_max_payload(rqstp);
1549 
1550 	if (bmap1 & FATTR4_WORD1_OWNER) {
1551 		ret += IDMAP_NAMESZ + 4;
1552 		bmap1 &= ~FATTR4_WORD1_OWNER;
1553 	}
1554 	if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
1555 		ret += IDMAP_NAMESZ + 4;
1556 		bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
1557 	}
1558 	if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
1559 		ret += NFS4_FHSIZE + 4;
1560 		bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
1561 	}
1562 	if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
1563 		ret += NFSD4_MAX_SEC_LABEL_LEN + 12;
1564 		bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
1565 	}
1566 	/*
1567 	 * Largest of remaining attributes are 16 bytes (e.g.,
1568 	 * supported_attributes)
1569 	 */
1570 	ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
1571 	/* bitmask, length */
1572 	ret += 20;
1573 	return ret;
1574 }
1575 
1576 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1577 {
1578 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
1579 		* sizeof(__be32);
1580 }
1581 
1582 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1583 {
1584 	return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
1585 		* sizeof(__be32);
1586 }
1587 
1588 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1589 {
1590 	return (op_encode_hdr_size + op_encode_stateid_maxsz
1591 		+ op_encode_change_info_maxsz + 1
1592 		+ nfs4_fattr_bitmap_maxsz
1593 		+ op_encode_delegation_maxsz) * sizeof(__be32);
1594 }
1595 
1596 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1597 {
1598 	u32 maxcount = 0, rlen = 0;
1599 
1600 	maxcount = svc_max_payload(rqstp);
1601 	rlen = min(op->u.read.rd_length, maxcount);
1602 
1603 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
1604 }
1605 
1606 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1607 {
1608 	u32 maxcount = 0, rlen = 0;
1609 
1610 	maxcount = svc_max_payload(rqstp);
1611 	rlen = min(op->u.readdir.rd_maxcount, maxcount);
1612 
1613 	return (op_encode_hdr_size + op_encode_verifier_maxsz +
1614 		XDR_QUADLEN(rlen)) * sizeof(__be32);
1615 }
1616 
1617 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1618 {
1619 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
1620 		* sizeof(__be32);
1621 }
1622 
1623 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1624 {
1625 	return (op_encode_hdr_size + op_encode_change_info_maxsz
1626 		+ op_encode_change_info_maxsz) * sizeof(__be32);
1627 }
1628 
1629 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
1630 				       struct nfsd4_op *op)
1631 {
1632 	return (op_encode_hdr_size
1633 		+ XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
1634 }
1635 
1636 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1637 {
1638 	return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
1639 }
1640 
1641 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1642 {
1643 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
1644 								sizeof(__be32);
1645 }
1646 
1647 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1648 {
1649 	return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
1650 }
1651 
1652 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1653 {
1654 	return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
1655 		1 + 1 + /* eir_flags, spr_how */\
1656 		4 + /* spo_must_enforce & _allow with bitmap */\
1657 		2 + /*eir_server_owner.so_minor_id */\
1658 		/* eir_server_owner.so_major_id<> */\
1659 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
1660 		/* eir_server_scope<> */\
1661 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
1662 		1 + /* eir_server_impl_id array length */\
1663 		0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
1664 }
1665 
1666 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1667 {
1668 	return (op_encode_hdr_size + \
1669 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
1670 		2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
1671 }
1672 
1673 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
1674 {
1675 	return (op_encode_hdr_size + \
1676 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
1677 		2 + /* csr_sequence, csr_flags */\
1678 		op_encode_channel_attrs_maxsz + \
1679 		op_encode_channel_attrs_maxsz) * sizeof(__be32);
1680 }
1681 
1682 static struct nfsd4_operation nfsd4_ops[] = {
1683 	[OP_ACCESS] = {
1684 		.op_func = (nfsd4op_func)nfsd4_access,
1685 		.op_name = "OP_ACCESS",
1686 	},
1687 	[OP_CLOSE] = {
1688 		.op_func = (nfsd4op_func)nfsd4_close,
1689 		.op_flags = OP_MODIFIES_SOMETHING,
1690 		.op_name = "OP_CLOSE",
1691 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1692 		.op_get_currentstateid = (stateid_getter)nfsd4_get_closestateid,
1693 		.op_set_currentstateid = (stateid_setter)nfsd4_set_closestateid,
1694 	},
1695 	[OP_COMMIT] = {
1696 		.op_func = (nfsd4op_func)nfsd4_commit,
1697 		.op_flags = OP_MODIFIES_SOMETHING,
1698 		.op_name = "OP_COMMIT",
1699 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_commit_rsize,
1700 	},
1701 	[OP_CREATE] = {
1702 		.op_func = (nfsd4op_func)nfsd4_create,
1703 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
1704 		.op_name = "OP_CREATE",
1705 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_create_rsize,
1706 	},
1707 	[OP_DELEGRETURN] = {
1708 		.op_func = (nfsd4op_func)nfsd4_delegreturn,
1709 		.op_flags = OP_MODIFIES_SOMETHING,
1710 		.op_name = "OP_DELEGRETURN",
1711 		.op_rsize_bop = nfsd4_only_status_rsize,
1712 		.op_get_currentstateid = (stateid_getter)nfsd4_get_delegreturnstateid,
1713 	},
1714 	[OP_GETATTR] = {
1715 		.op_func = (nfsd4op_func)nfsd4_getattr,
1716 		.op_flags = ALLOWED_ON_ABSENT_FS,
1717 		.op_rsize_bop = nfsd4_getattr_rsize,
1718 		.op_name = "OP_GETATTR",
1719 	},
1720 	[OP_GETFH] = {
1721 		.op_func = (nfsd4op_func)nfsd4_getfh,
1722 		.op_name = "OP_GETFH",
1723 	},
1724 	[OP_LINK] = {
1725 		.op_func = (nfsd4op_func)nfsd4_link,
1726 		.op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
1727 				| OP_CACHEME,
1728 		.op_name = "OP_LINK",
1729 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_link_rsize,
1730 	},
1731 	[OP_LOCK] = {
1732 		.op_func = (nfsd4op_func)nfsd4_lock,
1733 		.op_flags = OP_MODIFIES_SOMETHING,
1734 		.op_name = "OP_LOCK",
1735 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_lock_rsize,
1736 		.op_set_currentstateid = (stateid_setter)nfsd4_set_lockstateid,
1737 	},
1738 	[OP_LOCKT] = {
1739 		.op_func = (nfsd4op_func)nfsd4_lockt,
1740 		.op_name = "OP_LOCKT",
1741 	},
1742 	[OP_LOCKU] = {
1743 		.op_func = (nfsd4op_func)nfsd4_locku,
1744 		.op_flags = OP_MODIFIES_SOMETHING,
1745 		.op_name = "OP_LOCKU",
1746 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1747 		.op_get_currentstateid = (stateid_getter)nfsd4_get_lockustateid,
1748 	},
1749 	[OP_LOOKUP] = {
1750 		.op_func = (nfsd4op_func)nfsd4_lookup,
1751 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
1752 		.op_name = "OP_LOOKUP",
1753 	},
1754 	[OP_LOOKUPP] = {
1755 		.op_func = (nfsd4op_func)nfsd4_lookupp,
1756 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
1757 		.op_name = "OP_LOOKUPP",
1758 	},
1759 	[OP_NVERIFY] = {
1760 		.op_func = (nfsd4op_func)nfsd4_nverify,
1761 		.op_name = "OP_NVERIFY",
1762 	},
1763 	[OP_OPEN] = {
1764 		.op_func = (nfsd4op_func)nfsd4_open,
1765 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
1766 		.op_name = "OP_OPEN",
1767 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_open_rsize,
1768 		.op_set_currentstateid = (stateid_setter)nfsd4_set_openstateid,
1769 	},
1770 	[OP_OPEN_CONFIRM] = {
1771 		.op_func = (nfsd4op_func)nfsd4_open_confirm,
1772 		.op_flags = OP_MODIFIES_SOMETHING,
1773 		.op_name = "OP_OPEN_CONFIRM",
1774 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1775 	},
1776 	[OP_OPEN_DOWNGRADE] = {
1777 		.op_func = (nfsd4op_func)nfsd4_open_downgrade,
1778 		.op_flags = OP_MODIFIES_SOMETHING,
1779 		.op_name = "OP_OPEN_DOWNGRADE",
1780 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_status_stateid_rsize,
1781 		.op_get_currentstateid = (stateid_getter)nfsd4_get_opendowngradestateid,
1782 		.op_set_currentstateid = (stateid_setter)nfsd4_set_opendowngradestateid,
1783 	},
1784 	[OP_PUTFH] = {
1785 		.op_func = (nfsd4op_func)nfsd4_putfh,
1786 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1787 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1788 		.op_name = "OP_PUTFH",
1789 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1790 	},
1791 	[OP_PUTPUBFH] = {
1792 		.op_func = (nfsd4op_func)nfsd4_putrootfh,
1793 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1794 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1795 		.op_name = "OP_PUTPUBFH",
1796 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1797 	},
1798 	[OP_PUTROOTFH] = {
1799 		.op_func = (nfsd4op_func)nfsd4_putrootfh,
1800 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1801 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
1802 		.op_name = "OP_PUTROOTFH",
1803 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1804 	},
1805 	[OP_READ] = {
1806 		.op_func = (nfsd4op_func)nfsd4_read,
1807 		.op_name = "OP_READ",
1808 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_read_rsize,
1809 		.op_get_currentstateid = (stateid_getter)nfsd4_get_readstateid,
1810 	},
1811 	[OP_READDIR] = {
1812 		.op_func = (nfsd4op_func)nfsd4_readdir,
1813 		.op_name = "OP_READDIR",
1814 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_readdir_rsize,
1815 	},
1816 	[OP_READLINK] = {
1817 		.op_func = (nfsd4op_func)nfsd4_readlink,
1818 		.op_name = "OP_READLINK",
1819 	},
1820 	[OP_REMOVE] = {
1821 		.op_func = (nfsd4op_func)nfsd4_remove,
1822 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1823 		.op_name = "OP_REMOVE",
1824 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_remove_rsize,
1825 	},
1826 	[OP_RENAME] = {
1827 		.op_func = (nfsd4op_func)nfsd4_rename,
1828 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1829 		.op_name = "OP_RENAME",
1830 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_rename_rsize,
1831 	},
1832 	[OP_RENEW] = {
1833 		.op_func = (nfsd4op_func)nfsd4_renew,
1834 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1835 				| OP_MODIFIES_SOMETHING,
1836 		.op_name = "OP_RENEW",
1837 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1838 
1839 	},
1840 	[OP_RESTOREFH] = {
1841 		.op_func = (nfsd4op_func)nfsd4_restorefh,
1842 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1843 				| OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
1844 		.op_name = "OP_RESTOREFH",
1845 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1846 	},
1847 	[OP_SAVEFH] = {
1848 		.op_func = (nfsd4op_func)nfsd4_savefh,
1849 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
1850 		.op_name = "OP_SAVEFH",
1851 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1852 	},
1853 	[OP_SECINFO] = {
1854 		.op_func = (nfsd4op_func)nfsd4_secinfo,
1855 		.op_flags = OP_HANDLES_WRONGSEC,
1856 		.op_name = "OP_SECINFO",
1857 	},
1858 	[OP_SETATTR] = {
1859 		.op_func = (nfsd4op_func)nfsd4_setattr,
1860 		.op_name = "OP_SETATTR",
1861 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1862 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_setattr_rsize,
1863 		.op_get_currentstateid = (stateid_getter)nfsd4_get_setattrstateid,
1864 	},
1865 	[OP_SETCLIENTID] = {
1866 		.op_func = (nfsd4op_func)nfsd4_setclientid,
1867 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1868 				| OP_MODIFIES_SOMETHING | OP_CACHEME,
1869 		.op_name = "OP_SETCLIENTID",
1870 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_setclientid_rsize,
1871 	},
1872 	[OP_SETCLIENTID_CONFIRM] = {
1873 		.op_func = (nfsd4op_func)nfsd4_setclientid_confirm,
1874 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1875 				| OP_MODIFIES_SOMETHING | OP_CACHEME,
1876 		.op_name = "OP_SETCLIENTID_CONFIRM",
1877 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1878 	},
1879 	[OP_VERIFY] = {
1880 		.op_func = (nfsd4op_func)nfsd4_verify,
1881 		.op_name = "OP_VERIFY",
1882 	},
1883 	[OP_WRITE] = {
1884 		.op_func = (nfsd4op_func)nfsd4_write,
1885 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1886 		.op_name = "OP_WRITE",
1887 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1888 		.op_get_currentstateid = (stateid_getter)nfsd4_get_writestateid,
1889 	},
1890 	[OP_RELEASE_LOCKOWNER] = {
1891 		.op_func = (nfsd4op_func)nfsd4_release_lockowner,
1892 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
1893 				| OP_MODIFIES_SOMETHING,
1894 		.op_name = "OP_RELEASE_LOCKOWNER",
1895 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1896 	},
1897 
1898 	/* NFSv4.1 operations */
1899 	[OP_EXCHANGE_ID] = {
1900 		.op_func = (nfsd4op_func)nfsd4_exchange_id,
1901 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1902 				| OP_MODIFIES_SOMETHING,
1903 		.op_name = "OP_EXCHANGE_ID",
1904 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_exchange_id_rsize,
1905 	},
1906 	[OP_BACKCHANNEL_CTL] = {
1907 		.op_func = (nfsd4op_func)nfsd4_backchannel_ctl,
1908 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1909 		.op_name = "OP_BACKCHANNEL_CTL",
1910 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1911 	},
1912 	[OP_BIND_CONN_TO_SESSION] = {
1913 		.op_func = (nfsd4op_func)nfsd4_bind_conn_to_session,
1914 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1915 				| OP_MODIFIES_SOMETHING,
1916 		.op_name = "OP_BIND_CONN_TO_SESSION",
1917 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_bind_conn_to_session_rsize,
1918 	},
1919 	[OP_CREATE_SESSION] = {
1920 		.op_func = (nfsd4op_func)nfsd4_create_session,
1921 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1922 				| OP_MODIFIES_SOMETHING,
1923 		.op_name = "OP_CREATE_SESSION",
1924 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_create_session_rsize,
1925 	},
1926 	[OP_DESTROY_SESSION] = {
1927 		.op_func = (nfsd4op_func)nfsd4_destroy_session,
1928 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1929 				| OP_MODIFIES_SOMETHING,
1930 		.op_name = "OP_DESTROY_SESSION",
1931 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1932 	},
1933 	[OP_SEQUENCE] = {
1934 		.op_func = (nfsd4op_func)nfsd4_sequence,
1935 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
1936 		.op_name = "OP_SEQUENCE",
1937 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_sequence_rsize,
1938 	},
1939 	[OP_DESTROY_CLIENTID] = {
1940 		.op_func = (nfsd4op_func)nfsd4_destroy_clientid,
1941 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
1942 				| OP_MODIFIES_SOMETHING,
1943 		.op_name = "OP_DESTROY_CLIENTID",
1944 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1945 	},
1946 	[OP_RECLAIM_COMPLETE] = {
1947 		.op_func = (nfsd4op_func)nfsd4_reclaim_complete,
1948 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1949 		.op_name = "OP_RECLAIM_COMPLETE",
1950 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1951 	},
1952 	[OP_SECINFO_NO_NAME] = {
1953 		.op_func = (nfsd4op_func)nfsd4_secinfo_no_name,
1954 		.op_flags = OP_HANDLES_WRONGSEC,
1955 		.op_name = "OP_SECINFO_NO_NAME",
1956 	},
1957 	[OP_TEST_STATEID] = {
1958 		.op_func = (nfsd4op_func)nfsd4_test_stateid,
1959 		.op_flags = ALLOWED_WITHOUT_FH,
1960 		.op_name = "OP_TEST_STATEID",
1961 	},
1962 	[OP_FREE_STATEID] = {
1963 		.op_func = (nfsd4op_func)nfsd4_free_stateid,
1964 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
1965 		.op_name = "OP_FREE_STATEID",
1966 		.op_get_currentstateid = (stateid_getter)nfsd4_get_freestateid,
1967 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_only_status_rsize,
1968 	},
1969 
1970 	/* NFSv4.2 operations */
1971 	[OP_ALLOCATE] = {
1972 		.op_func = (nfsd4op_func)nfsd4_allocate,
1973 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1974 		.op_name = "OP_ALLOCATE",
1975 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1976 	},
1977 	[OP_DEALLOCATE] = {
1978 		.op_func = (nfsd4op_func)nfsd4_deallocate,
1979 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
1980 		.op_name = "OP_DEALLOCATE",
1981 		.op_rsize_bop = (nfsd4op_rsize)nfsd4_write_rsize,
1982 	},
1983 	[OP_SEEK] = {
1984 		.op_func = (nfsd4op_func)nfsd4_seek,
1985 		.op_name = "OP_SEEK",
1986 	},
1987 };
1988 
1989 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
1990 {
1991 	struct nfsd4_operation *opdesc;
1992 	nfsd4op_rsize estimator;
1993 
1994 	if (op->opnum == OP_ILLEGAL)
1995 		return op_encode_hdr_size * sizeof(__be32);
1996 	opdesc = OPDESC(op);
1997 	estimator = opdesc->op_rsize_bop;
1998 	return estimator ? estimator(rqstp, op) : PAGE_SIZE;
1999 }
2000 
2001 void warn_on_nonidempotent_op(struct nfsd4_op *op)
2002 {
2003 	if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
2004 		pr_err("unable to encode reply to nonidempotent op %d (%s)\n",
2005 			op->opnum, nfsd4_op_name(op->opnum));
2006 		WARN_ON_ONCE(1);
2007 	}
2008 }
2009 
2010 static const char *nfsd4_op_name(unsigned opnum)
2011 {
2012 	if (opnum < ARRAY_SIZE(nfsd4_ops))
2013 		return nfsd4_ops[opnum].op_name;
2014 	return "unknown_operation";
2015 }
2016 
2017 #define nfsd4_voidres			nfsd4_voidargs
2018 struct nfsd4_voidargs { int dummy; };
2019 
2020 static struct svc_procedure		nfsd_procedures4[2] = {
2021 	[NFSPROC4_NULL] = {
2022 		.pc_func = (svc_procfunc) nfsd4_proc_null,
2023 		.pc_encode = (kxdrproc_t) nfs4svc_encode_voidres,
2024 		.pc_argsize = sizeof(struct nfsd4_voidargs),
2025 		.pc_ressize = sizeof(struct nfsd4_voidres),
2026 		.pc_cachetype = RC_NOCACHE,
2027 		.pc_xdrressize = 1,
2028 	},
2029 	[NFSPROC4_COMPOUND] = {
2030 		.pc_func = (svc_procfunc) nfsd4_proc_compound,
2031 		.pc_decode = (kxdrproc_t) nfs4svc_decode_compoundargs,
2032 		.pc_encode = (kxdrproc_t) nfs4svc_encode_compoundres,
2033 		.pc_argsize = sizeof(struct nfsd4_compoundargs),
2034 		.pc_ressize = sizeof(struct nfsd4_compoundres),
2035 		.pc_release = nfsd4_release_compoundargs,
2036 		.pc_cachetype = RC_NOCACHE,
2037 		.pc_xdrressize = NFSD_BUFSIZE/4,
2038 	},
2039 };
2040 
2041 struct svc_version	nfsd_version4 = {
2042 		.vs_vers	= 4,
2043 		.vs_nproc	= 2,
2044 		.vs_proc	= nfsd_procedures4,
2045 		.vs_dispatch	= nfsd_dispatch,
2046 		.vs_xdrsize	= NFS4_SVC_XDRSIZE,
2047 		.vs_rpcb_optnl	= 1,
2048 };
2049 
2050 /*
2051  * Local variables:
2052  *  c-basic-offset: 8
2053  * End:
2054  */
2055