xref: /linux/fs/nfsd/nfs4proc.c (revision 8e07e0e3964ca4e23ce7b68e2096fe660a888942)
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 
55 static bool inter_copy_offload_enable;
56 module_param(inter_copy_offload_enable, bool, 0644);
57 MODULE_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
61 static int nfsd4_ssc_umount_timeout = 900000;		/* default to 15 mins */
62 module_param(nfsd4_ssc_umount_timeout, int, 0644);
63 MODULE_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 
69 static u32 nfsd_attrmask[] = {
70 	NFSD_WRITEABLE_ATTRS_WORD0,
71 	NFSD_WRITEABLE_ATTRS_WORD1,
72 	NFSD_WRITEABLE_ATTRS_WORD2
73 };
74 
75 static u32 nfsd41_ex_attrmask[] = {
76 	NFSD_SUPPATTR_EXCLCREAT_WORD0,
77 	NFSD_SUPPATTR_EXCLCREAT_WORD1,
78 	NFSD_SUPPATTR_EXCLCREAT_WORD2
79 };
80 
81 static __be32
82 check_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 
103 static __be32
104 nfsd4_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 
122 static int
123 is_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 
131 static inline void
132 fh_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 
141 static __be32
142 do_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 
161 static __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 
179 static 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 
187 static inline bool nfsd4_create_is_exclusive(int createmode)
188 {
189 	return createmode == NFS4_CREATE_EXCLUSIVE ||
190 		createmode == NFS4_CREATE_EXCLUSIVE4_1;
191 }
192 
193 static __be32
194 nfsd4_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  */
231 static __be32
232 nfsd4_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 
369 set_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;
376 out:
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  */
394 static void
395 set_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 
417 static __be32
418 do_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);
480 out:
481 	return status;
482 }
483 
484 static __be32
485 do_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 
515 static void
516 copy_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 
525 static __be32
526 nfsd4_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;
630 out:
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  */
651 static __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  */
666 static __be32
667 nfsd4_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 
674 static __be32
675 nfsd4_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 
695 static __be32
696 nfsd4_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 
704 static __be32
705 nfsd4_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 
719 static __be32
720 nfsd4_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  */
734 static __be32
735 nfsd4_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 
754 static 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 
763 static __be32
764 nfsd4_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 
783 static __be32
784 nfsd4_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);
870 out:
871 	fh_put(&resfh);
872 out_umask:
873 	current->fs->umask = 0;
874 	nfsd_attrs_free(&attrs);
875 	return status;
876 }
877 
878 static __be32
879 nfsd4_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 
900 static __be32
901 nfsd4_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 
914 static __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 
931 static __be32
932 nfsd4_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 
938 static __be32
939 nfsd4_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 
947 static __be32
948 nfsd4_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 		clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
975 
976 	/* check stateid */
977 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
978 					&read->rd_stateid, RD_STATE,
979 					&read->rd_nf, NULL);
980 
981 	read->rd_rqstp = rqstp;
982 	read->rd_fhp = &cstate->current_fh;
983 	return status;
984 }
985 
986 
987 static void
988 nfsd4_read_release(union nfsd4_op_u *u)
989 {
990 	if (u->read.rd_nf)
991 		nfsd_file_put(u->read.rd_nf);
992 	trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
993 			     u->read.rd_offset, u->read.rd_length);
994 }
995 
996 static __be32
997 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
998 	      union nfsd4_op_u *u)
999 {
1000 	struct nfsd4_readdir *readdir = &u->readdir;
1001 	u64 cookie = readdir->rd_cookie;
1002 	static const nfs4_verifier zeroverf;
1003 
1004 	/* no need to check permission - this will be done in nfsd_readdir() */
1005 
1006 	if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1007 		return nfserr_inval;
1008 
1009 	readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1010 	readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1011 	readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1012 
1013 	if ((cookie == 1) || (cookie == 2) ||
1014 	    (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1015 		return nfserr_bad_cookie;
1016 
1017 	readdir->rd_rqstp = rqstp;
1018 	readdir->rd_fhp = &cstate->current_fh;
1019 	return nfs_ok;
1020 }
1021 
1022 static __be32
1023 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1024 	       union nfsd4_op_u *u)
1025 {
1026 	u->readlink.rl_rqstp = rqstp;
1027 	u->readlink.rl_fhp = &cstate->current_fh;
1028 	return nfs_ok;
1029 }
1030 
1031 static __be32
1032 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1033 	     union nfsd4_op_u *u)
1034 {
1035 	struct nfsd4_remove *remove = &u->remove;
1036 	__be32 status;
1037 
1038 	if (opens_in_grace(SVC_NET(rqstp)))
1039 		return nfserr_grace;
1040 	status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1041 			     remove->rm_name, remove->rm_namelen);
1042 	if (!status)
1043 		set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1044 	return status;
1045 }
1046 
1047 static __be32
1048 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1049 	     union nfsd4_op_u *u)
1050 {
1051 	struct nfsd4_rename *rename = &u->rename;
1052 	__be32 status;
1053 
1054 	if (opens_in_grace(SVC_NET(rqstp)))
1055 		return nfserr_grace;
1056 	status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1057 			     rename->rn_snamelen, &cstate->current_fh,
1058 			     rename->rn_tname, rename->rn_tnamelen);
1059 	if (status)
1060 		return status;
1061 	set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1062 	set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1063 	return nfs_ok;
1064 }
1065 
1066 static __be32
1067 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1068 	      union nfsd4_op_u *u)
1069 {
1070 	struct nfsd4_secinfo *secinfo = &u->secinfo;
1071 	struct svc_export *exp;
1072 	struct dentry *dentry;
1073 	__be32 err;
1074 
1075 	err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1076 	if (err)
1077 		return err;
1078 	err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1079 				    secinfo->si_name, secinfo->si_namelen,
1080 				    &exp, &dentry);
1081 	if (err)
1082 		return err;
1083 	if (d_really_is_negative(dentry)) {
1084 		exp_put(exp);
1085 		err = nfserr_noent;
1086 	} else
1087 		secinfo->si_exp = exp;
1088 	dput(dentry);
1089 	if (cstate->minorversion)
1090 		/* See rfc 5661 section 2.6.3.1.1.8 */
1091 		fh_put(&cstate->current_fh);
1092 	return err;
1093 }
1094 
1095 static __be32
1096 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1097 		union nfsd4_op_u *u)
1098 {
1099 	__be32 err;
1100 
1101 	switch (u->secinfo_no_name.sin_style) {
1102 	case NFS4_SECINFO_STYLE4_CURRENT_FH:
1103 		break;
1104 	case NFS4_SECINFO_STYLE4_PARENT:
1105 		err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1106 		if (err)
1107 			return err;
1108 		break;
1109 	default:
1110 		return nfserr_inval;
1111 	}
1112 
1113 	u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1114 	fh_put(&cstate->current_fh);
1115 	return nfs_ok;
1116 }
1117 
1118 static void
1119 nfsd4_secinfo_release(union nfsd4_op_u *u)
1120 {
1121 	if (u->secinfo.si_exp)
1122 		exp_put(u->secinfo.si_exp);
1123 }
1124 
1125 static void
1126 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1127 {
1128 	if (u->secinfo_no_name.sin_exp)
1129 		exp_put(u->secinfo_no_name.sin_exp);
1130 }
1131 
1132 static __be32
1133 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1134 	      union nfsd4_op_u *u)
1135 {
1136 	struct nfsd4_setattr *setattr = &u->setattr;
1137 	struct nfsd_attrs attrs = {
1138 		.na_iattr	= &setattr->sa_iattr,
1139 		.na_seclabel	= &setattr->sa_label,
1140 	};
1141 	struct inode *inode;
1142 	__be32 status = nfs_ok;
1143 	int err;
1144 
1145 	if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1146 		status = nfs4_preprocess_stateid_op(rqstp, cstate,
1147 				&cstate->current_fh, &setattr->sa_stateid,
1148 				WR_STATE, NULL, NULL);
1149 		if (status)
1150 			return status;
1151 	}
1152 	err = fh_want_write(&cstate->current_fh);
1153 	if (err)
1154 		return nfserrno(err);
1155 	status = nfs_ok;
1156 
1157 	status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1158 				    nfsd_attrmask);
1159 	if (status)
1160 		goto out;
1161 
1162 	inode = cstate->current_fh.fh_dentry->d_inode;
1163 	status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1164 				   setattr->sa_acl, &attrs);
1165 
1166 	if (status)
1167 		goto out;
1168 	status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1169 				0, (time64_t)0);
1170 	if (!status)
1171 		status = nfserrno(attrs.na_labelerr);
1172 	if (!status)
1173 		status = nfserrno(attrs.na_aclerr);
1174 out:
1175 	nfsd_attrs_free(&attrs);
1176 	fh_drop_write(&cstate->current_fh);
1177 	return status;
1178 }
1179 
1180 static __be32
1181 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1182 	    union nfsd4_op_u *u)
1183 {
1184 	struct nfsd4_write *write = &u->write;
1185 	stateid_t *stateid = &write->wr_stateid;
1186 	struct nfsd_file *nf = NULL;
1187 	__be32 status = nfs_ok;
1188 	unsigned long cnt;
1189 	int nvecs;
1190 
1191 	if (write->wr_offset > (u64)OFFSET_MAX ||
1192 	    write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1193 		return nfserr_fbig;
1194 
1195 	cnt = write->wr_buflen;
1196 	trace_nfsd_write_start(rqstp, &cstate->current_fh,
1197 			       write->wr_offset, cnt);
1198 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1199 						stateid, WR_STATE, &nf, NULL);
1200 	if (status)
1201 		return status;
1202 
1203 	write->wr_how_written = write->wr_stable_how;
1204 
1205 	nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1206 	WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1207 
1208 	status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1209 				write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1210 				write->wr_how_written,
1211 				(__be32 *)write->wr_verifier.data);
1212 	nfsd_file_put(nf);
1213 
1214 	write->wr_bytes_written = cnt;
1215 	trace_nfsd_write_done(rqstp, &cstate->current_fh,
1216 			      write->wr_offset, cnt);
1217 	return status;
1218 }
1219 
1220 static __be32
1221 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1222 		  stateid_t *src_stateid, struct nfsd_file **src,
1223 		  stateid_t *dst_stateid, struct nfsd_file **dst)
1224 {
1225 	__be32 status;
1226 
1227 	if (!cstate->save_fh.fh_dentry)
1228 		return nfserr_nofilehandle;
1229 
1230 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1231 					    src_stateid, RD_STATE, src, NULL);
1232 	if (status)
1233 		goto out;
1234 
1235 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1236 					    dst_stateid, WR_STATE, dst, NULL);
1237 	if (status)
1238 		goto out_put_src;
1239 
1240 	/* fix up for NFS-specific error code */
1241 	if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1242 	    !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1243 		status = nfserr_wrong_type;
1244 		goto out_put_dst;
1245 	}
1246 
1247 out:
1248 	return status;
1249 out_put_dst:
1250 	nfsd_file_put(*dst);
1251 	*dst = NULL;
1252 out_put_src:
1253 	nfsd_file_put(*src);
1254 	*src = NULL;
1255 	goto out;
1256 }
1257 
1258 static __be32
1259 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1260 		union nfsd4_op_u *u)
1261 {
1262 	struct nfsd4_clone *clone = &u->clone;
1263 	struct nfsd_file *src, *dst;
1264 	__be32 status;
1265 
1266 	status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1267 				   &clone->cl_dst_stateid, &dst);
1268 	if (status)
1269 		goto out;
1270 
1271 	status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1272 			dst, clone->cl_dst_pos, clone->cl_count,
1273 			EX_ISSYNC(cstate->current_fh.fh_export));
1274 
1275 	nfsd_file_put(dst);
1276 	nfsd_file_put(src);
1277 out:
1278 	return status;
1279 }
1280 
1281 static void nfs4_put_copy(struct nfsd4_copy *copy)
1282 {
1283 	if (!refcount_dec_and_test(&copy->refcount))
1284 		return;
1285 	kfree(copy->cp_src);
1286 	kfree(copy);
1287 }
1288 
1289 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1290 {
1291 	if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1292 		kthread_stop(copy->copy_task);
1293 	nfs4_put_copy(copy);
1294 }
1295 
1296 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1297 {
1298 	struct nfsd4_copy *copy = NULL;
1299 
1300 	spin_lock(&clp->async_lock);
1301 	if (!list_empty(&clp->async_copies)) {
1302 		copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1303 					copies);
1304 		refcount_inc(&copy->refcount);
1305 	}
1306 	spin_unlock(&clp->async_lock);
1307 	return copy;
1308 }
1309 
1310 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1311 {
1312 	struct nfsd4_copy *copy;
1313 
1314 	while ((copy = nfsd4_get_copy(clp)) != NULL)
1315 		nfsd4_stop_copy(copy);
1316 }
1317 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1318 
1319 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1320 				   struct nfs_fh *src_fh,
1321 				   nfs4_stateid *stateid);
1322 extern void nfs42_ssc_close(struct file *filep);
1323 
1324 extern void nfs_sb_deactive(struct super_block *sb);
1325 
1326 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1327 
1328 /*
1329  * setup a work entry in the ssc delayed unmount list.
1330  */
1331 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1332 				  struct nfsd4_ssc_umount_item **nsui,
1333 				  struct svc_rqst *rqstp)
1334 {
1335 	struct nfsd4_ssc_umount_item *ni = NULL;
1336 	struct nfsd4_ssc_umount_item *work = NULL;
1337 	struct nfsd4_ssc_umount_item *tmp;
1338 	DEFINE_WAIT(wait);
1339 	__be32 status = 0;
1340 
1341 	*nsui = NULL;
1342 	work = kzalloc(sizeof(*work), GFP_KERNEL);
1343 try_again:
1344 	spin_lock(&nn->nfsd_ssc_lock);
1345 	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1346 		if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1347 			continue;
1348 		/* found a match */
1349 		if (ni->nsui_busy) {
1350 			/*  wait - and try again */
1351 			prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1352 			spin_unlock(&nn->nfsd_ssc_lock);
1353 
1354 			/* allow 20secs for mount/unmount for now - revisit */
1355 			if (svc_thread_should_stop(rqstp) ||
1356 					(schedule_timeout(20*HZ) == 0)) {
1357 				finish_wait(&nn->nfsd_ssc_waitq, &wait);
1358 				kfree(work);
1359 				return nfserr_eagain;
1360 			}
1361 			finish_wait(&nn->nfsd_ssc_waitq, &wait);
1362 			goto try_again;
1363 		}
1364 		*nsui = ni;
1365 		refcount_inc(&ni->nsui_refcnt);
1366 		spin_unlock(&nn->nfsd_ssc_lock);
1367 		kfree(work);
1368 
1369 		/* return vfsmount in (*nsui)->nsui_vfsmount */
1370 		return 0;
1371 	}
1372 	if (work) {
1373 		strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1374 		refcount_set(&work->nsui_refcnt, 2);
1375 		work->nsui_busy = true;
1376 		list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1377 		*nsui = work;
1378 	} else
1379 		status = nfserr_resource;
1380 	spin_unlock(&nn->nfsd_ssc_lock);
1381 	return status;
1382 }
1383 
1384 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1385 				 struct nfsd4_ssc_umount_item *nsui,
1386 				 struct vfsmount *ss_mnt)
1387 {
1388 	spin_lock(&nn->nfsd_ssc_lock);
1389 	nsui->nsui_vfsmount = ss_mnt;
1390 	nsui->nsui_busy = false;
1391 	wake_up_all(&nn->nfsd_ssc_waitq);
1392 	spin_unlock(&nn->nfsd_ssc_lock);
1393 }
1394 
1395 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1396 				 struct nfsd4_ssc_umount_item *nsui)
1397 {
1398 	spin_lock(&nn->nfsd_ssc_lock);
1399 	list_del(&nsui->nsui_list);
1400 	wake_up_all(&nn->nfsd_ssc_waitq);
1401 	spin_unlock(&nn->nfsd_ssc_lock);
1402 	kfree(nsui);
1403 }
1404 
1405 /*
1406  * Support one copy source server for now.
1407  */
1408 static __be32
1409 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1410 		       struct nfsd4_ssc_umount_item **nsui)
1411 {
1412 	struct file_system_type *type;
1413 	struct vfsmount *ss_mnt;
1414 	struct nfs42_netaddr *naddr;
1415 	struct sockaddr_storage tmp_addr;
1416 	size_t tmp_addrlen, match_netid_len = 3;
1417 	char *startsep = "", *endsep = "", *match_netid = "tcp";
1418 	char *ipaddr, *dev_name, *raw_data;
1419 	int len, raw_len;
1420 	__be32 status = nfserr_inval;
1421 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1422 
1423 	naddr = &nss->u.nl4_addr;
1424 	tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1425 					 naddr->addr_len,
1426 					 (struct sockaddr *)&tmp_addr,
1427 					 sizeof(tmp_addr));
1428 	*nsui = NULL;
1429 	if (tmp_addrlen == 0)
1430 		goto out_err;
1431 
1432 	if (tmp_addr.ss_family == AF_INET6) {
1433 		startsep = "[";
1434 		endsep = "]";
1435 		match_netid = "tcp6";
1436 		match_netid_len = 4;
1437 	}
1438 
1439 	if (naddr->netid_len != match_netid_len ||
1440 		strncmp(naddr->netid, match_netid, naddr->netid_len))
1441 		goto out_err;
1442 
1443 	/* Construct the raw data for the vfs_kern_mount call */
1444 	len = RPC_MAX_ADDRBUFLEN + 1;
1445 	ipaddr = kzalloc(len, GFP_KERNEL);
1446 	if (!ipaddr)
1447 		goto out_err;
1448 
1449 	rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1450 
1451 	/* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1452 
1453 	raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1454 	raw_data = kzalloc(raw_len, GFP_KERNEL);
1455 	if (!raw_data)
1456 		goto out_free_ipaddr;
1457 
1458 	snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1459 
1460 	status = nfserr_nodev;
1461 	type = get_fs_type("nfs");
1462 	if (!type)
1463 		goto out_free_rawdata;
1464 
1465 	/* Set the server:<export> for the vfs_kern_mount call */
1466 	dev_name = kzalloc(len + 5, GFP_KERNEL);
1467 	if (!dev_name)
1468 		goto out_free_rawdata;
1469 	snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1470 
1471 	status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui, rqstp);
1472 	if (status)
1473 		goto out_free_devname;
1474 	if ((*nsui)->nsui_vfsmount)
1475 		goto out_done;
1476 
1477 	/* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1478 	ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1479 	module_put(type->owner);
1480 	if (IS_ERR(ss_mnt)) {
1481 		status = nfserr_nodev;
1482 		nfsd4_ssc_cancel_dul(nn, *nsui);
1483 		goto out_free_devname;
1484 	}
1485 	nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1486 out_done:
1487 	status = 0;
1488 
1489 out_free_devname:
1490 	kfree(dev_name);
1491 out_free_rawdata:
1492 	kfree(raw_data);
1493 out_free_ipaddr:
1494 	kfree(ipaddr);
1495 out_err:
1496 	return status;
1497 }
1498 
1499 /*
1500  * Verify COPY destination stateid.
1501  *
1502  * Connect to the source server with NFSv4.1.
1503  * Create the source struct file for nfsd_copy_range.
1504  * Called with COPY cstate:
1505  *    SAVED_FH: source filehandle
1506  *    CURRENT_FH: destination filehandle
1507  */
1508 static __be32
1509 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1510 		      struct nfsd4_compound_state *cstate,
1511 		      struct nfsd4_copy *copy)
1512 {
1513 	struct svc_fh *s_fh = NULL;
1514 	stateid_t *s_stid = &copy->cp_src_stateid;
1515 	__be32 status = nfserr_inval;
1516 
1517 	/* Verify the destination stateid and set dst struct file*/
1518 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1519 					    &copy->cp_dst_stateid,
1520 					    WR_STATE, &copy->nf_dst, NULL);
1521 	if (status)
1522 		goto out;
1523 
1524 	status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1525 	if (status)
1526 		goto out;
1527 
1528 	s_fh = &cstate->save_fh;
1529 
1530 	copy->c_fh.size = s_fh->fh_handle.fh_size;
1531 	memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1532 	copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1533 	memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1534 	       sizeof(stateid_opaque_t));
1535 
1536 	status = 0;
1537 out:
1538 	return status;
1539 }
1540 
1541 static void
1542 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1543 			struct nfsd_file *dst)
1544 {
1545 	struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1546 	long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1547 
1548 	nfs42_ssc_close(filp);
1549 	fput(filp);
1550 
1551 	spin_lock(&nn->nfsd_ssc_lock);
1552 	list_del(&nsui->nsui_list);
1553 	/*
1554 	 * vfsmount can be shared by multiple exports,
1555 	 * decrement refcnt. If the count drops to 1 it
1556 	 * will be unmounted when nsui_expire expires.
1557 	 */
1558 	refcount_dec(&nsui->nsui_refcnt);
1559 	nsui->nsui_expire = jiffies + timeout;
1560 	list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1561 	spin_unlock(&nn->nfsd_ssc_lock);
1562 }
1563 
1564 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1565 
1566 static __be32
1567 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1568 		      struct nfsd4_compound_state *cstate,
1569 		      struct nfsd4_copy *copy)
1570 {
1571 	return nfserr_inval;
1572 }
1573 
1574 static void
1575 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1576 			struct nfsd_file *dst)
1577 {
1578 }
1579 
1580 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1581 				   struct nfs_fh *src_fh,
1582 				   nfs4_stateid *stateid)
1583 {
1584 	return NULL;
1585 }
1586 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1587 
1588 static __be32
1589 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1590 		      struct nfsd4_compound_state *cstate,
1591 		      struct nfsd4_copy *copy)
1592 {
1593 	return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1594 				 &copy->nf_src, &copy->cp_dst_stateid,
1595 				 &copy->nf_dst);
1596 }
1597 
1598 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1599 {
1600 	struct nfsd4_cb_offload *cbo =
1601 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1602 
1603 	kfree(cbo);
1604 }
1605 
1606 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1607 				 struct rpc_task *task)
1608 {
1609 	struct nfsd4_cb_offload *cbo =
1610 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1611 
1612 	trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1613 	return 1;
1614 }
1615 
1616 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1617 	.release = nfsd4_cb_offload_release,
1618 	.done = nfsd4_cb_offload_done
1619 };
1620 
1621 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1622 {
1623 	copy->cp_res.wr_stable_how =
1624 		test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1625 			NFS_FILE_SYNC : NFS_UNSTABLE;
1626 	nfsd4_copy_set_sync(copy, sync);
1627 	gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1628 }
1629 
1630 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1631 				     struct file *dst,
1632 				     struct file *src)
1633 {
1634 	errseq_t since;
1635 	ssize_t bytes_copied = 0;
1636 	u64 bytes_total = copy->cp_count;
1637 	u64 src_pos = copy->cp_src_pos;
1638 	u64 dst_pos = copy->cp_dst_pos;
1639 	int status;
1640 	loff_t end;
1641 
1642 	/* See RFC 7862 p.67: */
1643 	if (bytes_total == 0)
1644 		bytes_total = ULLONG_MAX;
1645 	do {
1646 		/* Only async copies can be stopped here */
1647 		if (kthread_should_stop())
1648 			break;
1649 		bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1650 						    bytes_total);
1651 		if (bytes_copied <= 0)
1652 			break;
1653 		bytes_total -= bytes_copied;
1654 		copy->cp_res.wr_bytes_written += bytes_copied;
1655 		src_pos += bytes_copied;
1656 		dst_pos += bytes_copied;
1657 	} while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1658 	/* for a non-zero asynchronous copy do a commit of data */
1659 	if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1660 		since = READ_ONCE(dst->f_wb_err);
1661 		end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1662 		status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1663 		if (!status)
1664 			status = filemap_check_wb_err(dst->f_mapping, since);
1665 		if (!status)
1666 			set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1667 	}
1668 	return bytes_copied;
1669 }
1670 
1671 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1672 			    struct file *src, struct file *dst,
1673 			    bool sync)
1674 {
1675 	__be32 status;
1676 	ssize_t bytes;
1677 
1678 	bytes = _nfsd_copy_file_range(copy, dst, src);
1679 
1680 	/* for async copy, we ignore the error, client can always retry
1681 	 * to get the error
1682 	 */
1683 	if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1684 		status = nfserrno(bytes);
1685 	else {
1686 		nfsd4_init_copy_res(copy, sync);
1687 		status = nfs_ok;
1688 	}
1689 	return status;
1690 }
1691 
1692 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1693 {
1694 	dst->cp_src_pos = src->cp_src_pos;
1695 	dst->cp_dst_pos = src->cp_dst_pos;
1696 	dst->cp_count = src->cp_count;
1697 	dst->cp_flags = src->cp_flags;
1698 	memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1699 	memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1700 	dst->cp_clp = src->cp_clp;
1701 	dst->nf_dst = nfsd_file_get(src->nf_dst);
1702 	/* for inter, nf_src doesn't exist yet */
1703 	if (!nfsd4_ssc_is_inter(src))
1704 		dst->nf_src = nfsd_file_get(src->nf_src);
1705 
1706 	memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1707 	memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1708 	memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1709 	memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1710 	dst->ss_nsui = src->ss_nsui;
1711 }
1712 
1713 static void release_copy_files(struct nfsd4_copy *copy)
1714 {
1715 	if (copy->nf_src)
1716 		nfsd_file_put(copy->nf_src);
1717 	if (copy->nf_dst)
1718 		nfsd_file_put(copy->nf_dst);
1719 }
1720 
1721 static void cleanup_async_copy(struct nfsd4_copy *copy)
1722 {
1723 	nfs4_free_copy_state(copy);
1724 	release_copy_files(copy);
1725 	if (copy->cp_clp) {
1726 		spin_lock(&copy->cp_clp->async_lock);
1727 		if (!list_empty(&copy->copies))
1728 			list_del_init(&copy->copies);
1729 		spin_unlock(&copy->cp_clp->async_lock);
1730 	}
1731 	nfs4_put_copy(copy);
1732 }
1733 
1734 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1735 {
1736 	struct nfsd4_cb_offload *cbo;
1737 
1738 	cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1739 	if (!cbo)
1740 		return;
1741 
1742 	memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1743 	memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1744 	cbo->co_nfserr = nfserr;
1745 
1746 	nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1747 		      NFSPROC4_CLNT_CB_OFFLOAD);
1748 	trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1749 			      &cbo->co_fh, copy->cp_count, nfserr);
1750 	nfsd4_run_cb(&cbo->co_cb);
1751 }
1752 
1753 /**
1754  * nfsd4_do_async_copy - kthread function for background server-side COPY
1755  * @data: arguments for COPY operation
1756  *
1757  * Return values:
1758  *   %0: Copy operation is done.
1759  */
1760 static int nfsd4_do_async_copy(void *data)
1761 {
1762 	struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1763 	__be32 nfserr;
1764 
1765 	trace_nfsd_copy_do_async(copy);
1766 	if (nfsd4_ssc_is_inter(copy)) {
1767 		struct file *filp;
1768 
1769 		filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1770 				      &copy->c_fh, &copy->stateid);
1771 		if (IS_ERR(filp)) {
1772 			switch (PTR_ERR(filp)) {
1773 			case -EBADF:
1774 				nfserr = nfserr_wrong_type;
1775 				break;
1776 			default:
1777 				nfserr = nfserr_offload_denied;
1778 			}
1779 			/* ss_mnt will be unmounted by the laundromat */
1780 			goto do_callback;
1781 		}
1782 		nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1783 				       false);
1784 		nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1785 	} else {
1786 		nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1787 				       copy->nf_dst->nf_file, false);
1788 	}
1789 
1790 do_callback:
1791 	nfsd4_send_cb_offload(copy, nfserr);
1792 	cleanup_async_copy(copy);
1793 	return 0;
1794 }
1795 
1796 static __be32
1797 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1798 		union nfsd4_op_u *u)
1799 {
1800 	struct nfsd4_copy *copy = &u->copy;
1801 	__be32 status;
1802 	struct nfsd4_copy *async_copy = NULL;
1803 
1804 	copy->cp_clp = cstate->clp;
1805 	if (nfsd4_ssc_is_inter(copy)) {
1806 		trace_nfsd_copy_inter(copy);
1807 		if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1808 			status = nfserr_notsupp;
1809 			goto out;
1810 		}
1811 		status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1812 		if (status) {
1813 			trace_nfsd_copy_done(copy, status);
1814 			return nfserr_offload_denied;
1815 		}
1816 	} else {
1817 		trace_nfsd_copy_intra(copy);
1818 		status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1819 		if (status) {
1820 			trace_nfsd_copy_done(copy, status);
1821 			return status;
1822 		}
1823 	}
1824 
1825 	memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1826 		sizeof(struct knfsd_fh));
1827 	if (nfsd4_copy_is_async(copy)) {
1828 		struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1829 
1830 		status = nfserrno(-ENOMEM);
1831 		async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1832 		if (!async_copy)
1833 			goto out_err;
1834 		INIT_LIST_HEAD(&async_copy->copies);
1835 		refcount_set(&async_copy->refcount, 1);
1836 		async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1837 		if (!async_copy->cp_src)
1838 			goto out_err;
1839 		if (!nfs4_init_copy_state(nn, copy))
1840 			goto out_err;
1841 		memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1842 			sizeof(copy->cp_res.cb_stateid));
1843 		dup_copy_fields(copy, async_copy);
1844 		async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1845 				async_copy, "%s", "copy thread");
1846 		if (IS_ERR(async_copy->copy_task))
1847 			goto out_err;
1848 		spin_lock(&async_copy->cp_clp->async_lock);
1849 		list_add(&async_copy->copies,
1850 				&async_copy->cp_clp->async_copies);
1851 		spin_unlock(&async_copy->cp_clp->async_lock);
1852 		wake_up_process(async_copy->copy_task);
1853 		status = nfs_ok;
1854 	} else {
1855 		status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1856 				       copy->nf_dst->nf_file, true);
1857 	}
1858 out:
1859 	trace_nfsd_copy_done(copy, status);
1860 	release_copy_files(copy);
1861 	return status;
1862 out_err:
1863 	if (nfsd4_ssc_is_inter(copy)) {
1864 		/*
1865 		 * Source's vfsmount of inter-copy will be unmounted
1866 		 * by the laundromat. Use copy instead of async_copy
1867 		 * since async_copy->ss_nsui might not be set yet.
1868 		 */
1869 		refcount_dec(&copy->ss_nsui->nsui_refcnt);
1870 	}
1871 	if (async_copy)
1872 		cleanup_async_copy(async_copy);
1873 	status = nfserrno(-ENOMEM);
1874 	goto out;
1875 }
1876 
1877 static struct nfsd4_copy *
1878 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1879 {
1880 	struct nfsd4_copy *copy;
1881 
1882 	lockdep_assert_held(&clp->async_lock);
1883 
1884 	list_for_each_entry(copy, &clp->async_copies, copies) {
1885 		if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1886 			continue;
1887 		return copy;
1888 	}
1889 	return NULL;
1890 }
1891 
1892 static struct nfsd4_copy *
1893 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1894 {
1895 	struct nfsd4_copy *copy;
1896 
1897 	spin_lock(&clp->async_lock);
1898 	copy = find_async_copy_locked(clp, stateid);
1899 	if (copy)
1900 		refcount_inc(&copy->refcount);
1901 	spin_unlock(&clp->async_lock);
1902 	return copy;
1903 }
1904 
1905 static __be32
1906 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1907 		     struct nfsd4_compound_state *cstate,
1908 		     union nfsd4_op_u *u)
1909 {
1910 	struct nfsd4_offload_status *os = &u->offload_status;
1911 	struct nfsd4_copy *copy;
1912 	struct nfs4_client *clp = cstate->clp;
1913 
1914 	copy = find_async_copy(clp, &os->stateid);
1915 	if (!copy) {
1916 		struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1917 
1918 		return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1919 	} else
1920 		nfsd4_stop_copy(copy);
1921 
1922 	return nfs_ok;
1923 }
1924 
1925 static __be32
1926 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1927 		  union nfsd4_op_u *u)
1928 {
1929 	struct nfsd4_copy_notify *cn = &u->copy_notify;
1930 	__be32 status;
1931 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1932 	struct nfs4_stid *stid;
1933 	struct nfs4_cpntf_state *cps;
1934 	struct nfs4_client *clp = cstate->clp;
1935 
1936 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1937 					&cn->cpn_src_stateid, RD_STATE, NULL,
1938 					&stid);
1939 	if (status)
1940 		return status;
1941 
1942 	cn->cpn_lease_time.tv_sec = nn->nfsd4_lease;
1943 	cn->cpn_lease_time.tv_nsec = 0;
1944 
1945 	status = nfserrno(-ENOMEM);
1946 	cps = nfs4_alloc_init_cpntf_state(nn, stid);
1947 	if (!cps)
1948 		goto out;
1949 	memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1950 	memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1951 	memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1952 
1953 	/* For now, only return one server address in cpn_src, the
1954 	 * address used by the client to connect to this server.
1955 	 */
1956 	cn->cpn_src->nl4_type = NL4_NETADDR;
1957 	status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1958 				 &cn->cpn_src->u.nl4_addr);
1959 	WARN_ON_ONCE(status);
1960 	if (status) {
1961 		nfs4_put_cpntf_state(nn, cps);
1962 		goto out;
1963 	}
1964 out:
1965 	nfs4_put_stid(stid);
1966 	return status;
1967 }
1968 
1969 static __be32
1970 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1971 		struct nfsd4_fallocate *fallocate, int flags)
1972 {
1973 	__be32 status;
1974 	struct nfsd_file *nf;
1975 
1976 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1977 					    &fallocate->falloc_stateid,
1978 					    WR_STATE, &nf, NULL);
1979 	if (status != nfs_ok)
1980 		return status;
1981 
1982 	status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1983 				     fallocate->falloc_offset,
1984 				     fallocate->falloc_length,
1985 				     flags);
1986 	nfsd_file_put(nf);
1987 	return status;
1988 }
1989 
1990 static __be32
1991 nfsd4_offload_status(struct svc_rqst *rqstp,
1992 		     struct nfsd4_compound_state *cstate,
1993 		     union nfsd4_op_u *u)
1994 {
1995 	struct nfsd4_offload_status *os = &u->offload_status;
1996 	__be32 status = nfs_ok;
1997 	struct nfsd4_copy *copy;
1998 	struct nfs4_client *clp = cstate->clp;
1999 
2000 	spin_lock(&clp->async_lock);
2001 	copy = find_async_copy_locked(clp, &os->stateid);
2002 	if (copy)
2003 		os->count = copy->cp_res.wr_bytes_written;
2004 	else
2005 		status = nfserr_bad_stateid;
2006 	spin_unlock(&clp->async_lock);
2007 
2008 	return status;
2009 }
2010 
2011 static __be32
2012 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2013 	       union nfsd4_op_u *u)
2014 {
2015 	return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2016 }
2017 
2018 static __be32
2019 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2020 		 union nfsd4_op_u *u)
2021 {
2022 	return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2023 			       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2024 }
2025 
2026 static __be32
2027 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2028 	   union nfsd4_op_u *u)
2029 {
2030 	struct nfsd4_seek *seek = &u->seek;
2031 	int whence;
2032 	__be32 status;
2033 	struct nfsd_file *nf;
2034 
2035 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2036 					    &seek->seek_stateid,
2037 					    RD_STATE, &nf, NULL);
2038 	if (status)
2039 		return status;
2040 
2041 	switch (seek->seek_whence) {
2042 	case NFS4_CONTENT_DATA:
2043 		whence = SEEK_DATA;
2044 		break;
2045 	case NFS4_CONTENT_HOLE:
2046 		whence = SEEK_HOLE;
2047 		break;
2048 	default:
2049 		status = nfserr_union_notsupp;
2050 		goto out;
2051 	}
2052 
2053 	/*
2054 	 * Note:  This call does change file->f_pos, but nothing in NFSD
2055 	 *        should ever file->f_pos.
2056 	 */
2057 	seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2058 	if (seek->seek_pos < 0)
2059 		status = nfserrno(seek->seek_pos);
2060 	else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2061 		seek->seek_eof = true;
2062 
2063 out:
2064 	nfsd_file_put(nf);
2065 	return status;
2066 }
2067 
2068 /* This routine never returns NFS_OK!  If there are no other errors, it
2069  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2070  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2071  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2072  */
2073 static __be32
2074 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2075 	     struct nfsd4_verify *verify)
2076 {
2077 	__be32 *buf, *p;
2078 	int count;
2079 	__be32 status;
2080 
2081 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2082 	if (status)
2083 		return status;
2084 
2085 	status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2086 	if (status)
2087 		return status;
2088 
2089 	if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2090 	    || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2091 		return nfserr_inval;
2092 	if (verify->ve_attrlen & 3)
2093 		return nfserr_inval;
2094 
2095 	/* count in words:
2096 	 *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2097 	 */
2098 	count = 4 + (verify->ve_attrlen >> 2);
2099 	buf = kmalloc(count << 2, GFP_KERNEL);
2100 	if (!buf)
2101 		return nfserr_jukebox;
2102 
2103 	p = buf;
2104 	status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2105 				    cstate->current_fh.fh_export,
2106 				    cstate->current_fh.fh_dentry,
2107 				    verify->ve_bmval,
2108 				    rqstp, 0);
2109 	/*
2110 	 * If nfsd4_encode_fattr() ran out of space, assume that's because
2111 	 * the attributes are longer (hence different) than those given:
2112 	 */
2113 	if (status == nfserr_resource)
2114 		status = nfserr_not_same;
2115 	if (status)
2116 		goto out_kfree;
2117 
2118 	/* skip bitmap */
2119 	p = buf + 1 + ntohl(buf[0]);
2120 	status = nfserr_not_same;
2121 	if (ntohl(*p++) != verify->ve_attrlen)
2122 		goto out_kfree;
2123 	if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2124 		status = nfserr_same;
2125 
2126 out_kfree:
2127 	kfree(buf);
2128 	return status;
2129 }
2130 
2131 static __be32
2132 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2133 	      union nfsd4_op_u *u)
2134 {
2135 	__be32 status;
2136 
2137 	status = _nfsd4_verify(rqstp, cstate, &u->verify);
2138 	return status == nfserr_not_same ? nfs_ok : status;
2139 }
2140 
2141 static __be32
2142 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2143 	     union nfsd4_op_u *u)
2144 {
2145 	__be32 status;
2146 
2147 	status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2148 	return status == nfserr_same ? nfs_ok : status;
2149 }
2150 
2151 #ifdef CONFIG_NFSD_PNFS
2152 static const struct nfsd4_layout_ops *
2153 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2154 {
2155 	if (!exp->ex_layout_types) {
2156 		dprintk("%s: export does not support pNFS\n", __func__);
2157 		return NULL;
2158 	}
2159 
2160 	if (layout_type >= LAYOUT_TYPE_MAX ||
2161 	    !(exp->ex_layout_types & (1 << layout_type))) {
2162 		dprintk("%s: layout type %d not supported\n",
2163 			__func__, layout_type);
2164 		return NULL;
2165 	}
2166 
2167 	return nfsd4_layout_ops[layout_type];
2168 }
2169 
2170 static __be32
2171 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2172 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2173 {
2174 	struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2175 	const struct nfsd4_layout_ops *ops;
2176 	struct nfsd4_deviceid_map *map;
2177 	struct svc_export *exp;
2178 	__be32 nfserr;
2179 
2180 	dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2181 	       __func__,
2182 	       gdp->gd_layout_type,
2183 	       gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2184 	       gdp->gd_maxcount);
2185 
2186 	map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2187 	if (!map) {
2188 		dprintk("%s: couldn't find device ID to export mapping!\n",
2189 			__func__);
2190 		return nfserr_noent;
2191 	}
2192 
2193 	exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2194 	if (IS_ERR(exp)) {
2195 		dprintk("%s: could not find device id\n", __func__);
2196 		return nfserr_noent;
2197 	}
2198 
2199 	nfserr = nfserr_layoutunavailable;
2200 	ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2201 	if (!ops)
2202 		goto out;
2203 
2204 	nfserr = nfs_ok;
2205 	if (gdp->gd_maxcount != 0) {
2206 		nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2207 				rqstp, cstate->clp, gdp);
2208 	}
2209 
2210 	gdp->gd_notify_types &= ops->notify_types;
2211 out:
2212 	exp_put(exp);
2213 	return nfserr;
2214 }
2215 
2216 static void
2217 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2218 {
2219 	kfree(u->getdeviceinfo.gd_device);
2220 }
2221 
2222 static __be32
2223 nfsd4_layoutget(struct svc_rqst *rqstp,
2224 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2225 {
2226 	struct nfsd4_layoutget *lgp = &u->layoutget;
2227 	struct svc_fh *current_fh = &cstate->current_fh;
2228 	const struct nfsd4_layout_ops *ops;
2229 	struct nfs4_layout_stateid *ls;
2230 	__be32 nfserr;
2231 	int accmode = NFSD_MAY_READ_IF_EXEC;
2232 
2233 	switch (lgp->lg_seg.iomode) {
2234 	case IOMODE_READ:
2235 		accmode |= NFSD_MAY_READ;
2236 		break;
2237 	case IOMODE_RW:
2238 		accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2239 		break;
2240 	default:
2241 		dprintk("%s: invalid iomode %d\n",
2242 			__func__, lgp->lg_seg.iomode);
2243 		nfserr = nfserr_badiomode;
2244 		goto out;
2245 	}
2246 
2247 	nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2248 	if (nfserr)
2249 		goto out;
2250 
2251 	nfserr = nfserr_layoutunavailable;
2252 	ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2253 	if (!ops)
2254 		goto out;
2255 
2256 	/*
2257 	 * Verify minlength and range as per RFC5661:
2258 	 *  o  If loga_length is less than loga_minlength,
2259 	 *     the metadata server MUST return NFS4ERR_INVAL.
2260 	 *  o  If the sum of loga_offset and loga_minlength exceeds
2261 	 *     NFS4_UINT64_MAX, and loga_minlength is not
2262 	 *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2263 	 *  o  If the sum of loga_offset and loga_length exceeds
2264 	 *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2265 	 *     the error NFS4ERR_INVAL MUST result.
2266 	 */
2267 	nfserr = nfserr_inval;
2268 	if (lgp->lg_seg.length < lgp->lg_minlength ||
2269 	    (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2270 	     lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2271 	    (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2272 	     lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2273 		goto out;
2274 	if (lgp->lg_seg.length == 0)
2275 		goto out;
2276 
2277 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2278 						true, lgp->lg_layout_type, &ls);
2279 	if (nfserr) {
2280 		trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2281 		goto out;
2282 	}
2283 
2284 	nfserr = nfserr_recallconflict;
2285 	if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2286 		goto out_put_stid;
2287 
2288 	nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2289 				     current_fh, lgp);
2290 	if (nfserr)
2291 		goto out_put_stid;
2292 
2293 	nfserr = nfsd4_insert_layout(lgp, ls);
2294 
2295 out_put_stid:
2296 	mutex_unlock(&ls->ls_mutex);
2297 	nfs4_put_stid(&ls->ls_stid);
2298 out:
2299 	return nfserr;
2300 }
2301 
2302 static void
2303 nfsd4_layoutget_release(union nfsd4_op_u *u)
2304 {
2305 	kfree(u->layoutget.lg_content);
2306 }
2307 
2308 static __be32
2309 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2310 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2311 {
2312 	struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2313 	const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2314 	struct svc_fh *current_fh = &cstate->current_fh;
2315 	const struct nfsd4_layout_ops *ops;
2316 	loff_t new_size = lcp->lc_last_wr + 1;
2317 	struct inode *inode;
2318 	struct nfs4_layout_stateid *ls;
2319 	__be32 nfserr;
2320 
2321 	nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2322 	if (nfserr)
2323 		goto out;
2324 
2325 	nfserr = nfserr_layoutunavailable;
2326 	ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2327 	if (!ops)
2328 		goto out;
2329 	inode = d_inode(current_fh->fh_dentry);
2330 
2331 	nfserr = nfserr_inval;
2332 	if (new_size <= seg->offset) {
2333 		dprintk("pnfsd: last write before layout segment\n");
2334 		goto out;
2335 	}
2336 	if (new_size > seg->offset + seg->length) {
2337 		dprintk("pnfsd: last write beyond layout segment\n");
2338 		goto out;
2339 	}
2340 	if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2341 		dprintk("pnfsd: layoutcommit beyond EOF\n");
2342 		goto out;
2343 	}
2344 
2345 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2346 						false, lcp->lc_layout_type,
2347 						&ls);
2348 	if (nfserr) {
2349 		trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2350 		/* fixup error code as per RFC5661 */
2351 		if (nfserr == nfserr_bad_stateid)
2352 			nfserr = nfserr_badlayout;
2353 		goto out;
2354 	}
2355 
2356 	/* LAYOUTCOMMIT does not require any serialization */
2357 	mutex_unlock(&ls->ls_mutex);
2358 
2359 	if (new_size > i_size_read(inode)) {
2360 		lcp->lc_size_chg = true;
2361 		lcp->lc_newsize = new_size;
2362 	} else {
2363 		lcp->lc_size_chg = false;
2364 	}
2365 
2366 	nfserr = ops->proc_layoutcommit(inode, lcp);
2367 	nfs4_put_stid(&ls->ls_stid);
2368 out:
2369 	return nfserr;
2370 }
2371 
2372 static __be32
2373 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2374 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2375 {
2376 	struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2377 	struct svc_fh *current_fh = &cstate->current_fh;
2378 	__be32 nfserr;
2379 
2380 	nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2381 	if (nfserr)
2382 		goto out;
2383 
2384 	nfserr = nfserr_layoutunavailable;
2385 	if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2386 		goto out;
2387 
2388 	switch (lrp->lr_seg.iomode) {
2389 	case IOMODE_READ:
2390 	case IOMODE_RW:
2391 	case IOMODE_ANY:
2392 		break;
2393 	default:
2394 		dprintk("%s: invalid iomode %d\n", __func__,
2395 			lrp->lr_seg.iomode);
2396 		nfserr = nfserr_inval;
2397 		goto out;
2398 	}
2399 
2400 	switch (lrp->lr_return_type) {
2401 	case RETURN_FILE:
2402 		nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2403 		break;
2404 	case RETURN_FSID:
2405 	case RETURN_ALL:
2406 		nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2407 		break;
2408 	default:
2409 		dprintk("%s: invalid return_type %d\n", __func__,
2410 			lrp->lr_return_type);
2411 		nfserr = nfserr_inval;
2412 		break;
2413 	}
2414 out:
2415 	return nfserr;
2416 }
2417 #endif /* CONFIG_NFSD_PNFS */
2418 
2419 static __be32
2420 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2421 	       union nfsd4_op_u *u)
2422 {
2423 	struct nfsd4_getxattr *getxattr = &u->getxattr;
2424 
2425 	return nfsd_getxattr(rqstp, &cstate->current_fh,
2426 			     getxattr->getxa_name, &getxattr->getxa_buf,
2427 			     &getxattr->getxa_len);
2428 }
2429 
2430 static __be32
2431 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2432 	   union nfsd4_op_u *u)
2433 {
2434 	struct nfsd4_setxattr *setxattr = &u->setxattr;
2435 	__be32 ret;
2436 
2437 	if (opens_in_grace(SVC_NET(rqstp)))
2438 		return nfserr_grace;
2439 
2440 	ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2441 			    setxattr->setxa_buf, setxattr->setxa_len,
2442 			    setxattr->setxa_flags);
2443 
2444 	if (!ret)
2445 		set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2446 
2447 	return ret;
2448 }
2449 
2450 static __be32
2451 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2452 	   union nfsd4_op_u *u)
2453 {
2454 	/*
2455 	 * Get the entire list, then copy out only the user attributes
2456 	 * in the encode function.
2457 	 */
2458 	return nfsd_listxattr(rqstp, &cstate->current_fh,
2459 			     &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2460 }
2461 
2462 static __be32
2463 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2464 	   union nfsd4_op_u *u)
2465 {
2466 	struct nfsd4_removexattr *removexattr = &u->removexattr;
2467 	__be32 ret;
2468 
2469 	if (opens_in_grace(SVC_NET(rqstp)))
2470 		return nfserr_grace;
2471 
2472 	ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2473 	    removexattr->rmxa_name);
2474 
2475 	if (!ret)
2476 		set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2477 
2478 	return ret;
2479 }
2480 
2481 /*
2482  * NULL call.
2483  */
2484 static __be32
2485 nfsd4_proc_null(struct svc_rqst *rqstp)
2486 {
2487 	return rpc_success;
2488 }
2489 
2490 static inline void nfsd4_increment_op_stats(u32 opnum)
2491 {
2492 	if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2493 		percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2494 }
2495 
2496 static const struct nfsd4_operation nfsd4_ops[];
2497 
2498 static const char *nfsd4_op_name(unsigned opnum);
2499 
2500 /*
2501  * Enforce NFSv4.1 COMPOUND ordering rules:
2502  *
2503  * Also note, enforced elsewhere:
2504  *	- SEQUENCE other than as first op results in
2505  *	  NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2506  *	- BIND_CONN_TO_SESSION must be the only op in its compound.
2507  *	  (Enforced in nfsd4_bind_conn_to_session().)
2508  *	- DESTROY_SESSION must be the final operation in a compound, if
2509  *	  sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2510  *	  (Enforced in nfsd4_destroy_session().)
2511  */
2512 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2513 {
2514 	struct nfsd4_op *first_op = &args->ops[0];
2515 
2516 	/* These ordering requirements don't apply to NFSv4.0: */
2517 	if (args->minorversion == 0)
2518 		return nfs_ok;
2519 	/* This is weird, but OK, not our problem: */
2520 	if (args->opcnt == 0)
2521 		return nfs_ok;
2522 	if (first_op->status == nfserr_op_illegal)
2523 		return nfs_ok;
2524 	if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2525 		return nfserr_op_not_in_session;
2526 	if (first_op->opnum == OP_SEQUENCE)
2527 		return nfs_ok;
2528 	/*
2529 	 * So first_op is something allowed outside a session, like
2530 	 * EXCHANGE_ID; but then it has to be the only op in the
2531 	 * compound:
2532 	 */
2533 	if (args->opcnt != 1)
2534 		return nfserr_not_only_op;
2535 	return nfs_ok;
2536 }
2537 
2538 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2539 {
2540 	return &nfsd4_ops[op->opnum];
2541 }
2542 
2543 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2544 {
2545 	if (op->opnum == OP_ILLEGAL)
2546 		return false;
2547 	return OPDESC(op)->op_flags & OP_CACHEME;
2548 }
2549 
2550 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2551 {
2552 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2553 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2554 	struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2555 	struct nfsd4_op *next = &argp->ops[resp->opcnt];
2556 	const struct nfsd4_operation *thisd = OPDESC(this);
2557 	const struct nfsd4_operation *nextd;
2558 
2559 	/*
2560 	 * Most ops check wronsec on our own; only the putfh-like ops
2561 	 * have special rules.
2562 	 */
2563 	if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2564 		return false;
2565 	/*
2566 	 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2567 	 * put-filehandle operation if we're not going to use the
2568 	 * result:
2569 	 */
2570 	if (argp->opcnt == resp->opcnt)
2571 		return false;
2572 	if (next->opnum == OP_ILLEGAL)
2573 		return false;
2574 	nextd = OPDESC(next);
2575 	/*
2576 	 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2577 	 * errors themselves as necessary; others should check for them
2578 	 * now:
2579 	 */
2580 	return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2581 }
2582 
2583 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2584 static void
2585 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2586 {
2587 	struct nfsd4_op	*op, *current_op = NULL, *saved_op = NULL;
2588 	struct nfsd4_copy *copy;
2589 	struct nfsd4_putfh *putfh;
2590 	int i;
2591 
2592 	/* traverse all operation and if it's a COPY compound, mark the
2593 	 * source filehandle to skip verification
2594 	 */
2595 	for (i = 0; i < args->opcnt; i++) {
2596 		op = &args->ops[i];
2597 		if (op->opnum == OP_PUTFH)
2598 			current_op = op;
2599 		else if (op->opnum == OP_SAVEFH)
2600 			saved_op = current_op;
2601 		else if (op->opnum == OP_RESTOREFH)
2602 			current_op = saved_op;
2603 		else if (op->opnum == OP_COPY) {
2604 			copy = (struct nfsd4_copy *)&op->u;
2605 			if (!saved_op) {
2606 				op->status = nfserr_nofilehandle;
2607 				return;
2608 			}
2609 			putfh = (struct nfsd4_putfh *)&saved_op->u;
2610 			if (nfsd4_ssc_is_inter(copy))
2611 				putfh->no_verify = true;
2612 		}
2613 	}
2614 }
2615 #else
2616 static void
2617 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2618 {
2619 }
2620 #endif
2621 
2622 /*
2623  * COMPOUND call.
2624  */
2625 static __be32
2626 nfsd4_proc_compound(struct svc_rqst *rqstp)
2627 {
2628 	struct nfsd4_compoundargs *args = rqstp->rq_argp;
2629 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2630 	struct nfsd4_op	*op;
2631 	struct nfsd4_compound_state *cstate = &resp->cstate;
2632 	struct svc_fh *current_fh = &cstate->current_fh;
2633 	struct svc_fh *save_fh = &cstate->save_fh;
2634 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2635 	__be32		status;
2636 
2637 	resp->xdr = &rqstp->rq_res_stream;
2638 	resp->statusp = resp->xdr->p;
2639 
2640 	/* reserve space for: NFS status code */
2641 	xdr_reserve_space(resp->xdr, XDR_UNIT);
2642 
2643 	/* reserve space for: taglen, tag, and opcnt */
2644 	xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2645 	resp->taglen = args->taglen;
2646 	resp->tag = args->tag;
2647 	resp->rqstp = rqstp;
2648 	cstate->minorversion = args->minorversion;
2649 	fh_init(current_fh, NFS4_FHSIZE);
2650 	fh_init(save_fh, NFS4_FHSIZE);
2651 	/*
2652 	 * Don't use the deferral mechanism for NFSv4; compounds make it
2653 	 * too hard to avoid non-idempotency problems.
2654 	 */
2655 	clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2656 
2657 	/*
2658 	 * According to RFC3010, this takes precedence over all other errors.
2659 	 */
2660 	status = nfserr_minor_vers_mismatch;
2661 	if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2662 		goto out;
2663 
2664 	status = nfs41_check_op_ordering(args);
2665 	if (status) {
2666 		op = &args->ops[0];
2667 		op->status = status;
2668 		resp->opcnt = 1;
2669 		goto encode_op;
2670 	}
2671 	check_if_stalefh_allowed(args);
2672 
2673 	rqstp->rq_lease_breaker = (void **)&cstate->clp;
2674 
2675 	trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2676 	while (!status && resp->opcnt < args->opcnt) {
2677 		op = &args->ops[resp->opcnt++];
2678 
2679 		if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2680 			/* If there are still more operations to process,
2681 			 * stop here and report NFS4ERR_RESOURCE. */
2682 			if (cstate->minorversion == 0 &&
2683 			    args->client_opcnt > resp->opcnt) {
2684 				op->status = nfserr_resource;
2685 				goto encode_op;
2686 			}
2687 		}
2688 
2689 		/*
2690 		 * The XDR decode routines may have pre-set op->status;
2691 		 * for example, if there is a miscellaneous XDR error
2692 		 * it will be set to nfserr_bad_xdr.
2693 		 */
2694 		if (op->status) {
2695 			if (op->opnum == OP_OPEN)
2696 				op->status = nfsd4_open_omfg(rqstp, cstate, op);
2697 			goto encode_op;
2698 		}
2699 		if (!current_fh->fh_dentry &&
2700 				!HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2701 			if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2702 				op->status = nfserr_nofilehandle;
2703 				goto encode_op;
2704 			}
2705 		} else if (current_fh->fh_export &&
2706 			   current_fh->fh_export->ex_fslocs.migrated &&
2707 			  !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2708 			op->status = nfserr_moved;
2709 			goto encode_op;
2710 		}
2711 
2712 		fh_clear_pre_post_attrs(current_fh);
2713 
2714 		/* If op is non-idempotent */
2715 		if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2716 			/*
2717 			 * Don't execute this op if we couldn't encode a
2718 			 * successful reply:
2719 			 */
2720 			u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2721 			/*
2722 			 * Plus if there's another operation, make sure
2723 			 * we'll have space to at least encode an error:
2724 			 */
2725 			if (resp->opcnt < args->opcnt)
2726 				plen += COMPOUND_ERR_SLACK_SPACE;
2727 			op->status = nfsd4_check_resp_size(resp, plen);
2728 		}
2729 
2730 		if (op->status)
2731 			goto encode_op;
2732 
2733 		if (op->opdesc->op_get_currentstateid)
2734 			op->opdesc->op_get_currentstateid(cstate, &op->u);
2735 		op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2736 
2737 		/* Only from SEQUENCE */
2738 		if (cstate->status == nfserr_replay_cache) {
2739 			dprintk("%s NFS4.1 replay from cache\n", __func__);
2740 			status = op->status;
2741 			goto out;
2742 		}
2743 		if (!op->status) {
2744 			if (op->opdesc->op_set_currentstateid)
2745 				op->opdesc->op_set_currentstateid(cstate, &op->u);
2746 
2747 			if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2748 				clear_current_stateid(cstate);
2749 
2750 			if (current_fh->fh_export &&
2751 					need_wrongsec_check(rqstp))
2752 				op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2753 		}
2754 encode_op:
2755 		if (op->status == nfserr_replay_me) {
2756 			op->replay = &cstate->replay_owner->so_replay;
2757 			nfsd4_encode_replay(resp->xdr, op);
2758 			status = op->status = op->replay->rp_status;
2759 		} else {
2760 			nfsd4_encode_operation(resp, op);
2761 			status = op->status;
2762 		}
2763 
2764 		trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2765 					   status, nfsd4_op_name(op->opnum));
2766 
2767 		nfsd4_cstate_clear_replay(cstate);
2768 		nfsd4_increment_op_stats(op->opnum);
2769 	}
2770 
2771 	fh_put(current_fh);
2772 	fh_put(save_fh);
2773 	BUG_ON(cstate->replay_owner);
2774 out:
2775 	cstate->status = status;
2776 	/* Reset deferral mechanism for RPC deferrals */
2777 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2778 	return rpc_success;
2779 }
2780 
2781 #define op_encode_hdr_size		(2)
2782 #define op_encode_stateid_maxsz		(XDR_QUADLEN(NFS4_STATEID_SIZE))
2783 #define op_encode_verifier_maxsz	(XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2784 #define op_encode_change_info_maxsz	(5)
2785 #define nfs4_fattr_bitmap_maxsz		(4)
2786 
2787 /* We'll fall back on returning no lockowner if run out of space: */
2788 #define op_encode_lockowner_maxsz	(0)
2789 #define op_encode_lock_denied_maxsz	(8 + op_encode_lockowner_maxsz)
2790 
2791 #define nfs4_owner_maxsz		(1 + XDR_QUADLEN(IDMAP_NAMESZ))
2792 
2793 #define op_encode_ace_maxsz		(3 + nfs4_owner_maxsz)
2794 #define op_encode_delegation_maxsz	(1 + op_encode_stateid_maxsz + 1 + \
2795 					 op_encode_ace_maxsz)
2796 
2797 #define op_encode_channel_attrs_maxsz	(6 + 1 + 1)
2798 
2799 /*
2800  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2801  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2802  * the maximum payload size here, based on transport limits and the size
2803  * of the remaining space in the rq_pages array.
2804  */
2805 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2806 {
2807 	u32 buflen;
2808 
2809 	buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2810 	buflen -= rqstp->rq_auth_slack;
2811 	buflen -= rqstp->rq_res.head[0].iov_len;
2812 	return min_t(u32, buflen, svc_max_payload(rqstp));
2813 }
2814 
2815 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2816 				   const struct nfsd4_op *op)
2817 {
2818 	return (op_encode_hdr_size) * sizeof(__be32);
2819 }
2820 
2821 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2822 				      const struct nfsd4_op *op)
2823 {
2824 	return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2825 }
2826 
2827 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2828 			      const struct nfsd4_op *op)
2829 {
2830 	/* ac_supported, ac_resp_access */
2831 	return (op_encode_hdr_size + 2)* sizeof(__be32);
2832 }
2833 
2834 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2835 			      const struct nfsd4_op *op)
2836 {
2837 	return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2838 }
2839 
2840 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2841 			      const struct nfsd4_op *op)
2842 {
2843 	return (op_encode_hdr_size + op_encode_change_info_maxsz
2844 		+ nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2845 }
2846 
2847 /*
2848  * Note since this is an idempotent operation we won't insist on failing
2849  * the op prematurely if the estimate is too large.  We may turn off splice
2850  * reads unnecessarily.
2851  */
2852 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2853 			       const struct nfsd4_op *op)
2854 {
2855 	const u32 *bmap = op->u.getattr.ga_bmval;
2856 	u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2857 	u32 ret = 0;
2858 
2859 	if (bmap0 & FATTR4_WORD0_ACL)
2860 		return nfsd4_max_payload(rqstp);
2861 	if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2862 		return nfsd4_max_payload(rqstp);
2863 
2864 	if (bmap1 & FATTR4_WORD1_OWNER) {
2865 		ret += IDMAP_NAMESZ + 4;
2866 		bmap1 &= ~FATTR4_WORD1_OWNER;
2867 	}
2868 	if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2869 		ret += IDMAP_NAMESZ + 4;
2870 		bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2871 	}
2872 	if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2873 		ret += NFS4_FHSIZE + 4;
2874 		bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2875 	}
2876 	if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2877 		ret += NFS4_MAXLABELLEN + 12;
2878 		bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2879 	}
2880 	/*
2881 	 * Largest of remaining attributes are 16 bytes (e.g.,
2882 	 * supported_attributes)
2883 	 */
2884 	ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2885 	/* bitmask, length */
2886 	ret += 20;
2887 	return ret;
2888 }
2889 
2890 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2891 			     const struct nfsd4_op *op)
2892 {
2893 	return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2894 }
2895 
2896 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2897 			    const struct nfsd4_op *op)
2898 {
2899 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
2900 		* sizeof(__be32);
2901 }
2902 
2903 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2904 			    const struct nfsd4_op *op)
2905 {
2906 	return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2907 		* sizeof(__be32);
2908 }
2909 
2910 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2911 			    const struct nfsd4_op *op)
2912 {
2913 	return (op_encode_hdr_size + op_encode_stateid_maxsz
2914 		+ op_encode_change_info_maxsz + 1
2915 		+ nfs4_fattr_bitmap_maxsz
2916 		+ op_encode_delegation_maxsz) * sizeof(__be32);
2917 }
2918 
2919 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2920 			    const struct nfsd4_op *op)
2921 {
2922 	u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2923 
2924 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2925 }
2926 
2927 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2928 				 const struct nfsd4_op *op)
2929 {
2930 	u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2931 	/*
2932 	 * If we detect that the file changed during hole encoding, then we
2933 	 * recover by encoding the remaining reply as data. This means we need
2934 	 * to set aside enough room to encode two data segments.
2935 	 */
2936 	u32 seg_len = 2 * (1 + 2 + 1);
2937 
2938 	return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2939 }
2940 
2941 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2942 			       const struct nfsd4_op *op)
2943 {
2944 	u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2945 
2946 	return (op_encode_hdr_size + op_encode_verifier_maxsz +
2947 		XDR_QUADLEN(rlen)) * sizeof(__be32);
2948 }
2949 
2950 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2951 				const struct nfsd4_op *op)
2952 {
2953 	return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2954 }
2955 
2956 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2957 			      const struct nfsd4_op *op)
2958 {
2959 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
2960 		* sizeof(__be32);
2961 }
2962 
2963 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2964 			      const struct nfsd4_op *op)
2965 {
2966 	return (op_encode_hdr_size + op_encode_change_info_maxsz
2967 		+ op_encode_change_info_maxsz) * sizeof(__be32);
2968 }
2969 
2970 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2971 				const struct nfsd4_op *op)
2972 {
2973 	return (op_encode_hdr_size
2974 		+ XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2975 }
2976 
2977 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2978 				    const struct nfsd4_op *op)
2979 {
2980 	return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2981 		* sizeof(__be32);
2982 }
2983 
2984 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2985 			       const struct nfsd4_op *op)
2986 {
2987 	return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2988 }
2989 
2990 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2991 			       const struct nfsd4_op *op)
2992 {
2993 	return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2994 		(4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2995 }
2996 
2997 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2998 				   const struct nfsd4_op *op)
2999 {
3000 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
3001 								sizeof(__be32);
3002 }
3003 
3004 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
3005 			     const struct nfsd4_op *op)
3006 {
3007 	return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3008 }
3009 
3010 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3011 				   const struct nfsd4_op *op)
3012 {
3013 	return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3014 		1 + 1 + /* eir_flags, spr_how */\
3015 		4 + /* spo_must_enforce & _allow with bitmap */\
3016 		2 + /*eir_server_owner.so_minor_id */\
3017 		/* eir_server_owner.so_major_id<> */\
3018 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3019 		/* eir_server_scope<> */\
3020 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3021 		1 + /* eir_server_impl_id array length */\
3022 		0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3023 }
3024 
3025 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3026 					    const struct nfsd4_op *op)
3027 {
3028 	return (op_encode_hdr_size + \
3029 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3030 		2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3031 }
3032 
3033 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3034 				      const struct nfsd4_op *op)
3035 {
3036 	return (op_encode_hdr_size + \
3037 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3038 		2 + /* csr_sequence, csr_flags */\
3039 		op_encode_channel_attrs_maxsz + \
3040 		op_encode_channel_attrs_maxsz) * sizeof(__be32);
3041 }
3042 
3043 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3044 			    const struct nfsd4_op *op)
3045 {
3046 	return (op_encode_hdr_size +
3047 		1 /* wr_callback */ +
3048 		op_encode_stateid_maxsz /* wr_callback */ +
3049 		2 /* wr_count */ +
3050 		1 /* wr_committed */ +
3051 		op_encode_verifier_maxsz +
3052 		1 /* cr_consecutive */ +
3053 		1 /* cr_synchronous */) * sizeof(__be32);
3054 }
3055 
3056 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3057 				      const struct nfsd4_op *op)
3058 {
3059 	return (op_encode_hdr_size +
3060 		2 /* osr_count */ +
3061 		1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3062 }
3063 
3064 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3065 				   const struct nfsd4_op *op)
3066 {
3067 	return (op_encode_hdr_size +
3068 		3 /* cnr_lease_time */ +
3069 		1 /* We support one cnr_source_server */ +
3070 		1 /* cnr_stateid seq */ +
3071 		op_encode_stateid_maxsz /* cnr_stateid */ +
3072 		1 /* num cnr_source_server*/ +
3073 		1 /* nl4_type */ +
3074 		1 /* nl4 size */ +
3075 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3076 		* sizeof(__be32);
3077 }
3078 
3079 #ifdef CONFIG_NFSD_PNFS
3080 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3081 				     const struct nfsd4_op *op)
3082 {
3083 	u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3084 
3085 	return (op_encode_hdr_size +
3086 		1 /* gd_layout_type*/ +
3087 		XDR_QUADLEN(rlen) +
3088 		2 /* gd_notify_types */) * sizeof(__be32);
3089 }
3090 
3091 /*
3092  * At this stage we don't really know what layout driver will handle the request,
3093  * so we need to define an arbitrary upper bound here.
3094  */
3095 #define MAX_LAYOUT_SIZE		128
3096 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3097 				 const struct nfsd4_op *op)
3098 {
3099 	return (op_encode_hdr_size +
3100 		1 /* logr_return_on_close */ +
3101 		op_encode_stateid_maxsz +
3102 		1 /* nr of layouts */ +
3103 		MAX_LAYOUT_SIZE) * sizeof(__be32);
3104 }
3105 
3106 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3107 				    const struct nfsd4_op *op)
3108 {
3109 	return (op_encode_hdr_size +
3110 		1 /* locr_newsize */ +
3111 		2 /* ns_size */) * sizeof(__be32);
3112 }
3113 
3114 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3115 				    const struct nfsd4_op *op)
3116 {
3117 	return (op_encode_hdr_size +
3118 		1 /* lrs_stateid */ +
3119 		op_encode_stateid_maxsz) * sizeof(__be32);
3120 }
3121 #endif /* CONFIG_NFSD_PNFS */
3122 
3123 
3124 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3125 			    const struct nfsd4_op *op)
3126 {
3127 	return (op_encode_hdr_size + 3) * sizeof(__be32);
3128 }
3129 
3130 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3131 				const struct nfsd4_op *op)
3132 {
3133 	u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3134 
3135 	return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3136 }
3137 
3138 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3139 				const struct nfsd4_op *op)
3140 {
3141 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3142 		* sizeof(__be32);
3143 }
3144 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3145 				  const struct nfsd4_op *op)
3146 {
3147 	u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3148 
3149 	return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3150 }
3151 
3152 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3153 				   const struct nfsd4_op *op)
3154 {
3155 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3156 		* sizeof(__be32);
3157 }
3158 
3159 
3160 static const struct nfsd4_operation nfsd4_ops[] = {
3161 	[OP_ACCESS] = {
3162 		.op_func = nfsd4_access,
3163 		.op_name = "OP_ACCESS",
3164 		.op_rsize_bop = nfsd4_access_rsize,
3165 	},
3166 	[OP_CLOSE] = {
3167 		.op_func = nfsd4_close,
3168 		.op_flags = OP_MODIFIES_SOMETHING,
3169 		.op_name = "OP_CLOSE",
3170 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3171 		.op_get_currentstateid = nfsd4_get_closestateid,
3172 		.op_set_currentstateid = nfsd4_set_closestateid,
3173 	},
3174 	[OP_COMMIT] = {
3175 		.op_func = nfsd4_commit,
3176 		.op_flags = OP_MODIFIES_SOMETHING,
3177 		.op_name = "OP_COMMIT",
3178 		.op_rsize_bop = nfsd4_commit_rsize,
3179 	},
3180 	[OP_CREATE] = {
3181 		.op_func = nfsd4_create,
3182 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3183 		.op_name = "OP_CREATE",
3184 		.op_rsize_bop = nfsd4_create_rsize,
3185 	},
3186 	[OP_DELEGRETURN] = {
3187 		.op_func = nfsd4_delegreturn,
3188 		.op_flags = OP_MODIFIES_SOMETHING,
3189 		.op_name = "OP_DELEGRETURN",
3190 		.op_rsize_bop = nfsd4_only_status_rsize,
3191 		.op_get_currentstateid = nfsd4_get_delegreturnstateid,
3192 	},
3193 	[OP_GETATTR] = {
3194 		.op_func = nfsd4_getattr,
3195 		.op_flags = ALLOWED_ON_ABSENT_FS,
3196 		.op_rsize_bop = nfsd4_getattr_rsize,
3197 		.op_name = "OP_GETATTR",
3198 	},
3199 	[OP_GETFH] = {
3200 		.op_func = nfsd4_getfh,
3201 		.op_name = "OP_GETFH",
3202 		.op_rsize_bop = nfsd4_getfh_rsize,
3203 	},
3204 	[OP_LINK] = {
3205 		.op_func = nfsd4_link,
3206 		.op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3207 				| OP_CACHEME,
3208 		.op_name = "OP_LINK",
3209 		.op_rsize_bop = nfsd4_link_rsize,
3210 	},
3211 	[OP_LOCK] = {
3212 		.op_func = nfsd4_lock,
3213 		.op_release = nfsd4_lock_release,
3214 		.op_flags = OP_MODIFIES_SOMETHING |
3215 				OP_NONTRIVIAL_ERROR_ENCODE,
3216 		.op_name = "OP_LOCK",
3217 		.op_rsize_bop = nfsd4_lock_rsize,
3218 		.op_set_currentstateid = nfsd4_set_lockstateid,
3219 	},
3220 	[OP_LOCKT] = {
3221 		.op_func = nfsd4_lockt,
3222 		.op_release = nfsd4_lockt_release,
3223 		.op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3224 		.op_name = "OP_LOCKT",
3225 		.op_rsize_bop = nfsd4_lock_rsize,
3226 	},
3227 	[OP_LOCKU] = {
3228 		.op_func = nfsd4_locku,
3229 		.op_flags = OP_MODIFIES_SOMETHING,
3230 		.op_name = "OP_LOCKU",
3231 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3232 		.op_get_currentstateid = nfsd4_get_lockustateid,
3233 	},
3234 	[OP_LOOKUP] = {
3235 		.op_func = nfsd4_lookup,
3236 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3237 		.op_name = "OP_LOOKUP",
3238 		.op_rsize_bop = nfsd4_only_status_rsize,
3239 	},
3240 	[OP_LOOKUPP] = {
3241 		.op_func = nfsd4_lookupp,
3242 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3243 		.op_name = "OP_LOOKUPP",
3244 		.op_rsize_bop = nfsd4_only_status_rsize,
3245 	},
3246 	[OP_NVERIFY] = {
3247 		.op_func = nfsd4_nverify,
3248 		.op_name = "OP_NVERIFY",
3249 		.op_rsize_bop = nfsd4_only_status_rsize,
3250 	},
3251 	[OP_OPEN] = {
3252 		.op_func = nfsd4_open,
3253 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3254 		.op_name = "OP_OPEN",
3255 		.op_rsize_bop = nfsd4_open_rsize,
3256 		.op_set_currentstateid = nfsd4_set_openstateid,
3257 	},
3258 	[OP_OPEN_CONFIRM] = {
3259 		.op_func = nfsd4_open_confirm,
3260 		.op_flags = OP_MODIFIES_SOMETHING,
3261 		.op_name = "OP_OPEN_CONFIRM",
3262 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3263 	},
3264 	[OP_OPEN_DOWNGRADE] = {
3265 		.op_func = nfsd4_open_downgrade,
3266 		.op_flags = OP_MODIFIES_SOMETHING,
3267 		.op_name = "OP_OPEN_DOWNGRADE",
3268 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3269 		.op_get_currentstateid = nfsd4_get_opendowngradestateid,
3270 		.op_set_currentstateid = nfsd4_set_opendowngradestateid,
3271 	},
3272 	[OP_PUTFH] = {
3273 		.op_func = nfsd4_putfh,
3274 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3275 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3276 		.op_name = "OP_PUTFH",
3277 		.op_rsize_bop = nfsd4_only_status_rsize,
3278 	},
3279 	[OP_PUTPUBFH] = {
3280 		.op_func = nfsd4_putrootfh,
3281 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3282 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3283 		.op_name = "OP_PUTPUBFH",
3284 		.op_rsize_bop = nfsd4_only_status_rsize,
3285 	},
3286 	[OP_PUTROOTFH] = {
3287 		.op_func = nfsd4_putrootfh,
3288 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3289 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3290 		.op_name = "OP_PUTROOTFH",
3291 		.op_rsize_bop = nfsd4_only_status_rsize,
3292 	},
3293 	[OP_READ] = {
3294 		.op_func = nfsd4_read,
3295 		.op_release = nfsd4_read_release,
3296 		.op_name = "OP_READ",
3297 		.op_rsize_bop = nfsd4_read_rsize,
3298 		.op_get_currentstateid = nfsd4_get_readstateid,
3299 	},
3300 	[OP_READDIR] = {
3301 		.op_func = nfsd4_readdir,
3302 		.op_name = "OP_READDIR",
3303 		.op_rsize_bop = nfsd4_readdir_rsize,
3304 	},
3305 	[OP_READLINK] = {
3306 		.op_func = nfsd4_readlink,
3307 		.op_name = "OP_READLINK",
3308 		.op_rsize_bop = nfsd4_readlink_rsize,
3309 	},
3310 	[OP_REMOVE] = {
3311 		.op_func = nfsd4_remove,
3312 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3313 		.op_name = "OP_REMOVE",
3314 		.op_rsize_bop = nfsd4_remove_rsize,
3315 	},
3316 	[OP_RENAME] = {
3317 		.op_func = nfsd4_rename,
3318 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3319 		.op_name = "OP_RENAME",
3320 		.op_rsize_bop = nfsd4_rename_rsize,
3321 	},
3322 	[OP_RENEW] = {
3323 		.op_func = nfsd4_renew,
3324 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3325 				| OP_MODIFIES_SOMETHING,
3326 		.op_name = "OP_RENEW",
3327 		.op_rsize_bop = nfsd4_only_status_rsize,
3328 
3329 	},
3330 	[OP_RESTOREFH] = {
3331 		.op_func = nfsd4_restorefh,
3332 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3333 				| OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3334 		.op_name = "OP_RESTOREFH",
3335 		.op_rsize_bop = nfsd4_only_status_rsize,
3336 	},
3337 	[OP_SAVEFH] = {
3338 		.op_func = nfsd4_savefh,
3339 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3340 		.op_name = "OP_SAVEFH",
3341 		.op_rsize_bop = nfsd4_only_status_rsize,
3342 	},
3343 	[OP_SECINFO] = {
3344 		.op_func = nfsd4_secinfo,
3345 		.op_release = nfsd4_secinfo_release,
3346 		.op_flags = OP_HANDLES_WRONGSEC,
3347 		.op_name = "OP_SECINFO",
3348 		.op_rsize_bop = nfsd4_secinfo_rsize,
3349 	},
3350 	[OP_SETATTR] = {
3351 		.op_func = nfsd4_setattr,
3352 		.op_name = "OP_SETATTR",
3353 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3354 				| OP_NONTRIVIAL_ERROR_ENCODE,
3355 		.op_rsize_bop = nfsd4_setattr_rsize,
3356 		.op_get_currentstateid = nfsd4_get_setattrstateid,
3357 	},
3358 	[OP_SETCLIENTID] = {
3359 		.op_func = nfsd4_setclientid,
3360 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3361 				| OP_MODIFIES_SOMETHING | OP_CACHEME
3362 				| OP_NONTRIVIAL_ERROR_ENCODE,
3363 		.op_name = "OP_SETCLIENTID",
3364 		.op_rsize_bop = nfsd4_setclientid_rsize,
3365 	},
3366 	[OP_SETCLIENTID_CONFIRM] = {
3367 		.op_func = nfsd4_setclientid_confirm,
3368 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3369 				| OP_MODIFIES_SOMETHING | OP_CACHEME,
3370 		.op_name = "OP_SETCLIENTID_CONFIRM",
3371 		.op_rsize_bop = nfsd4_only_status_rsize,
3372 	},
3373 	[OP_VERIFY] = {
3374 		.op_func = nfsd4_verify,
3375 		.op_name = "OP_VERIFY",
3376 		.op_rsize_bop = nfsd4_only_status_rsize,
3377 	},
3378 	[OP_WRITE] = {
3379 		.op_func = nfsd4_write,
3380 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3381 		.op_name = "OP_WRITE",
3382 		.op_rsize_bop = nfsd4_write_rsize,
3383 		.op_get_currentstateid = nfsd4_get_writestateid,
3384 	},
3385 	[OP_RELEASE_LOCKOWNER] = {
3386 		.op_func = nfsd4_release_lockowner,
3387 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3388 				| OP_MODIFIES_SOMETHING,
3389 		.op_name = "OP_RELEASE_LOCKOWNER",
3390 		.op_rsize_bop = nfsd4_only_status_rsize,
3391 	},
3392 
3393 	/* NFSv4.1 operations */
3394 	[OP_EXCHANGE_ID] = {
3395 		.op_func = nfsd4_exchange_id,
3396 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3397 				| OP_MODIFIES_SOMETHING,
3398 		.op_name = "OP_EXCHANGE_ID",
3399 		.op_rsize_bop = nfsd4_exchange_id_rsize,
3400 	},
3401 	[OP_BACKCHANNEL_CTL] = {
3402 		.op_func = nfsd4_backchannel_ctl,
3403 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3404 		.op_name = "OP_BACKCHANNEL_CTL",
3405 		.op_rsize_bop = nfsd4_only_status_rsize,
3406 	},
3407 	[OP_BIND_CONN_TO_SESSION] = {
3408 		.op_func = nfsd4_bind_conn_to_session,
3409 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3410 				| OP_MODIFIES_SOMETHING,
3411 		.op_name = "OP_BIND_CONN_TO_SESSION",
3412 		.op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3413 	},
3414 	[OP_CREATE_SESSION] = {
3415 		.op_func = nfsd4_create_session,
3416 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3417 				| OP_MODIFIES_SOMETHING,
3418 		.op_name = "OP_CREATE_SESSION",
3419 		.op_rsize_bop = nfsd4_create_session_rsize,
3420 	},
3421 	[OP_DESTROY_SESSION] = {
3422 		.op_func = nfsd4_destroy_session,
3423 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3424 				| OP_MODIFIES_SOMETHING,
3425 		.op_name = "OP_DESTROY_SESSION",
3426 		.op_rsize_bop = nfsd4_only_status_rsize,
3427 	},
3428 	[OP_SEQUENCE] = {
3429 		.op_func = nfsd4_sequence,
3430 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3431 		.op_name = "OP_SEQUENCE",
3432 		.op_rsize_bop = nfsd4_sequence_rsize,
3433 	},
3434 	[OP_DESTROY_CLIENTID] = {
3435 		.op_func = nfsd4_destroy_clientid,
3436 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3437 				| OP_MODIFIES_SOMETHING,
3438 		.op_name = "OP_DESTROY_CLIENTID",
3439 		.op_rsize_bop = nfsd4_only_status_rsize,
3440 	},
3441 	[OP_RECLAIM_COMPLETE] = {
3442 		.op_func = nfsd4_reclaim_complete,
3443 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3444 		.op_name = "OP_RECLAIM_COMPLETE",
3445 		.op_rsize_bop = nfsd4_only_status_rsize,
3446 	},
3447 	[OP_SECINFO_NO_NAME] = {
3448 		.op_func = nfsd4_secinfo_no_name,
3449 		.op_release = nfsd4_secinfo_no_name_release,
3450 		.op_flags = OP_HANDLES_WRONGSEC,
3451 		.op_name = "OP_SECINFO_NO_NAME",
3452 		.op_rsize_bop = nfsd4_secinfo_rsize,
3453 	},
3454 	[OP_TEST_STATEID] = {
3455 		.op_func = nfsd4_test_stateid,
3456 		.op_flags = ALLOWED_WITHOUT_FH,
3457 		.op_name = "OP_TEST_STATEID",
3458 		.op_rsize_bop = nfsd4_test_stateid_rsize,
3459 	},
3460 	[OP_FREE_STATEID] = {
3461 		.op_func = nfsd4_free_stateid,
3462 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3463 		.op_name = "OP_FREE_STATEID",
3464 		.op_get_currentstateid = nfsd4_get_freestateid,
3465 		.op_rsize_bop = nfsd4_only_status_rsize,
3466 	},
3467 #ifdef CONFIG_NFSD_PNFS
3468 	[OP_GETDEVICEINFO] = {
3469 		.op_func = nfsd4_getdeviceinfo,
3470 		.op_release = nfsd4_getdeviceinfo_release,
3471 		.op_flags = ALLOWED_WITHOUT_FH,
3472 		.op_name = "OP_GETDEVICEINFO",
3473 		.op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3474 	},
3475 	[OP_LAYOUTGET] = {
3476 		.op_func = nfsd4_layoutget,
3477 		.op_release = nfsd4_layoutget_release,
3478 		.op_flags = OP_MODIFIES_SOMETHING,
3479 		.op_name = "OP_LAYOUTGET",
3480 		.op_rsize_bop = nfsd4_layoutget_rsize,
3481 	},
3482 	[OP_LAYOUTCOMMIT] = {
3483 		.op_func = nfsd4_layoutcommit,
3484 		.op_flags = OP_MODIFIES_SOMETHING,
3485 		.op_name = "OP_LAYOUTCOMMIT",
3486 		.op_rsize_bop = nfsd4_layoutcommit_rsize,
3487 	},
3488 	[OP_LAYOUTRETURN] = {
3489 		.op_func = nfsd4_layoutreturn,
3490 		.op_flags = OP_MODIFIES_SOMETHING,
3491 		.op_name = "OP_LAYOUTRETURN",
3492 		.op_rsize_bop = nfsd4_layoutreturn_rsize,
3493 	},
3494 #endif /* CONFIG_NFSD_PNFS */
3495 
3496 	/* NFSv4.2 operations */
3497 	[OP_ALLOCATE] = {
3498 		.op_func = nfsd4_allocate,
3499 		.op_flags = OP_MODIFIES_SOMETHING,
3500 		.op_name = "OP_ALLOCATE",
3501 		.op_rsize_bop = nfsd4_only_status_rsize,
3502 	},
3503 	[OP_DEALLOCATE] = {
3504 		.op_func = nfsd4_deallocate,
3505 		.op_flags = OP_MODIFIES_SOMETHING,
3506 		.op_name = "OP_DEALLOCATE",
3507 		.op_rsize_bop = nfsd4_only_status_rsize,
3508 	},
3509 	[OP_CLONE] = {
3510 		.op_func = nfsd4_clone,
3511 		.op_flags = OP_MODIFIES_SOMETHING,
3512 		.op_name = "OP_CLONE",
3513 		.op_rsize_bop = nfsd4_only_status_rsize,
3514 	},
3515 	[OP_COPY] = {
3516 		.op_func = nfsd4_copy,
3517 		.op_flags = OP_MODIFIES_SOMETHING,
3518 		.op_name = "OP_COPY",
3519 		.op_rsize_bop = nfsd4_copy_rsize,
3520 	},
3521 	[OP_READ_PLUS] = {
3522 		.op_func = nfsd4_read,
3523 		.op_release = nfsd4_read_release,
3524 		.op_name = "OP_READ_PLUS",
3525 		.op_rsize_bop = nfsd4_read_plus_rsize,
3526 		.op_get_currentstateid = nfsd4_get_readstateid,
3527 	},
3528 	[OP_SEEK] = {
3529 		.op_func = nfsd4_seek,
3530 		.op_name = "OP_SEEK",
3531 		.op_rsize_bop = nfsd4_seek_rsize,
3532 	},
3533 	[OP_OFFLOAD_STATUS] = {
3534 		.op_func = nfsd4_offload_status,
3535 		.op_name = "OP_OFFLOAD_STATUS",
3536 		.op_rsize_bop = nfsd4_offload_status_rsize,
3537 	},
3538 	[OP_OFFLOAD_CANCEL] = {
3539 		.op_func = nfsd4_offload_cancel,
3540 		.op_flags = OP_MODIFIES_SOMETHING,
3541 		.op_name = "OP_OFFLOAD_CANCEL",
3542 		.op_rsize_bop = nfsd4_only_status_rsize,
3543 	},
3544 	[OP_COPY_NOTIFY] = {
3545 		.op_func = nfsd4_copy_notify,
3546 		.op_flags = OP_MODIFIES_SOMETHING,
3547 		.op_name = "OP_COPY_NOTIFY",
3548 		.op_rsize_bop = nfsd4_copy_notify_rsize,
3549 	},
3550 	[OP_GETXATTR] = {
3551 		.op_func = nfsd4_getxattr,
3552 		.op_name = "OP_GETXATTR",
3553 		.op_rsize_bop = nfsd4_getxattr_rsize,
3554 	},
3555 	[OP_SETXATTR] = {
3556 		.op_func = nfsd4_setxattr,
3557 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3558 		.op_name = "OP_SETXATTR",
3559 		.op_rsize_bop = nfsd4_setxattr_rsize,
3560 	},
3561 	[OP_LISTXATTRS] = {
3562 		.op_func = nfsd4_listxattrs,
3563 		.op_name = "OP_LISTXATTRS",
3564 		.op_rsize_bop = nfsd4_listxattrs_rsize,
3565 	},
3566 	[OP_REMOVEXATTR] = {
3567 		.op_func = nfsd4_removexattr,
3568 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3569 		.op_name = "OP_REMOVEXATTR",
3570 		.op_rsize_bop = nfsd4_removexattr_rsize,
3571 	},
3572 };
3573 
3574 /**
3575  * nfsd4_spo_must_allow - Determine if the compound op contains an
3576  * operation that is allowed to be sent with machine credentials
3577  *
3578  * @rqstp: a pointer to the struct svc_rqst
3579  *
3580  * Checks to see if the compound contains a spo_must_allow op
3581  * and confirms that it was sent with the proper machine creds.
3582  */
3583 
3584 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3585 {
3586 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
3587 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3588 	struct nfsd4_op *this;
3589 	struct nfsd4_compound_state *cstate = &resp->cstate;
3590 	struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3591 	u32 opiter;
3592 
3593 	if (!cstate->minorversion)
3594 		return false;
3595 
3596 	if (cstate->spo_must_allowed)
3597 		return true;
3598 
3599 	opiter = resp->opcnt;
3600 	while (opiter < argp->opcnt) {
3601 		this = &argp->ops[opiter++];
3602 		if (test_bit(this->opnum, allow->u.longs) &&
3603 			cstate->clp->cl_mach_cred &&
3604 			nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3605 			cstate->spo_must_allowed = true;
3606 			return true;
3607 		}
3608 	}
3609 	cstate->spo_must_allowed = false;
3610 	return false;
3611 }
3612 
3613 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3614 {
3615 	if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3616 		return op_encode_hdr_size * sizeof(__be32);
3617 
3618 	BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3619 	return OPDESC(op)->op_rsize_bop(rqstp, op);
3620 }
3621 
3622 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3623 {
3624 	if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3625 		pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3626 			op->opnum, nfsd4_op_name(op->opnum));
3627 		WARN_ON_ONCE(1);
3628 	}
3629 }
3630 
3631 static const char *nfsd4_op_name(unsigned opnum)
3632 {
3633 	if (opnum < ARRAY_SIZE(nfsd4_ops))
3634 		return nfsd4_ops[opnum].op_name;
3635 	return "unknown_operation";
3636 }
3637 
3638 static const struct svc_procedure nfsd_procedures4[2] = {
3639 	[NFSPROC4_NULL] = {
3640 		.pc_func = nfsd4_proc_null,
3641 		.pc_decode = nfssvc_decode_voidarg,
3642 		.pc_encode = nfssvc_encode_voidres,
3643 		.pc_argsize = sizeof(struct nfsd_voidargs),
3644 		.pc_argzero = sizeof(struct nfsd_voidargs),
3645 		.pc_ressize = sizeof(struct nfsd_voidres),
3646 		.pc_cachetype = RC_NOCACHE,
3647 		.pc_xdrressize = 1,
3648 		.pc_name = "NULL",
3649 	},
3650 	[NFSPROC4_COMPOUND] = {
3651 		.pc_func = nfsd4_proc_compound,
3652 		.pc_decode = nfs4svc_decode_compoundargs,
3653 		.pc_encode = nfs4svc_encode_compoundres,
3654 		.pc_argsize = sizeof(struct nfsd4_compoundargs),
3655 		.pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3656 		.pc_ressize = sizeof(struct nfsd4_compoundres),
3657 		.pc_release = nfsd4_release_compoundargs,
3658 		.pc_cachetype = RC_NOCACHE,
3659 		.pc_xdrressize = NFSD_BUFSIZE/4,
3660 		.pc_name = "COMPOUND",
3661 	},
3662 };
3663 
3664 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3665 			      nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3666 const struct svc_version nfsd_version4 = {
3667 	.vs_vers		= 4,
3668 	.vs_nproc		= ARRAY_SIZE(nfsd_procedures4),
3669 	.vs_proc		= nfsd_procedures4,
3670 	.vs_count		= nfsd_count4,
3671 	.vs_dispatch		= nfsd_dispatch,
3672 	.vs_xdrsize		= NFS4_SVC_XDRSIZE,
3673 	.vs_rpcb_optnl		= true,
3674 	.vs_need_cong_ctrl	= true,
3675 };
3676