xref: /openbmc/linux/fs/nfsd/nfs4proc.c (revision c7e1962a)
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 (d_inode(child)->i_mtime.tv_sec == v_mtime &&
326 			    d_inode(child)->i_atime.tv_sec == 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 (d_inode(child)->i_mtime.tv_sec == v_mtime &&
335 			    d_inode(child)->i_atime.tv_sec == 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 __be32
755 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
756 	     union nfsd4_op_u *u)
757 {
758 	struct nfsd4_commit *commit = &u->commit;
759 	struct nfsd_file *nf;
760 	__be32 status;
761 
762 	status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
763 				   NFSD_MAY_NOT_BREAK_LEASE, &nf);
764 	if (status != nfs_ok)
765 		return status;
766 
767 	status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
768 			     commit->co_count,
769 			     (__be32 *)commit->co_verf.data);
770 	nfsd_file_put(nf);
771 	return status;
772 }
773 
774 static __be32
775 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
776 	     union nfsd4_op_u *u)
777 {
778 	struct nfsd4_create *create = &u->create;
779 	struct nfsd_attrs attrs = {
780 		.na_iattr	= &create->cr_iattr,
781 		.na_seclabel	= &create->cr_label,
782 	};
783 	struct svc_fh resfh;
784 	__be32 status;
785 	dev_t rdev;
786 
787 	fh_init(&resfh, NFS4_FHSIZE);
788 
789 	status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
790 	if (status)
791 		return status;
792 
793 	status = check_attr_support(rqstp, cstate, create->cr_bmval,
794 				    nfsd_attrmask);
795 	if (status)
796 		return status;
797 
798 	status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
799 	current->fs->umask = create->cr_umask;
800 	switch (create->cr_type) {
801 	case NF4LNK:
802 		status = nfsd_symlink(rqstp, &cstate->current_fh,
803 				      create->cr_name, create->cr_namelen,
804 				      create->cr_data, &attrs, &resfh);
805 		break;
806 
807 	case NF4BLK:
808 		status = nfserr_inval;
809 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
810 		if (MAJOR(rdev) != create->cr_specdata1 ||
811 		    MINOR(rdev) != create->cr_specdata2)
812 			goto out_umask;
813 		status = nfsd_create(rqstp, &cstate->current_fh,
814 				     create->cr_name, create->cr_namelen,
815 				     &attrs, S_IFBLK, rdev, &resfh);
816 		break;
817 
818 	case NF4CHR:
819 		status = nfserr_inval;
820 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
821 		if (MAJOR(rdev) != create->cr_specdata1 ||
822 		    MINOR(rdev) != create->cr_specdata2)
823 			goto out_umask;
824 		status = nfsd_create(rqstp, &cstate->current_fh,
825 				     create->cr_name, create->cr_namelen,
826 				     &attrs, S_IFCHR, rdev, &resfh);
827 		break;
828 
829 	case NF4SOCK:
830 		status = nfsd_create(rqstp, &cstate->current_fh,
831 				     create->cr_name, create->cr_namelen,
832 				     &attrs, S_IFSOCK, 0, &resfh);
833 		break;
834 
835 	case NF4FIFO:
836 		status = nfsd_create(rqstp, &cstate->current_fh,
837 				     create->cr_name, create->cr_namelen,
838 				     &attrs, S_IFIFO, 0, &resfh);
839 		break;
840 
841 	case NF4DIR:
842 		create->cr_iattr.ia_valid &= ~ATTR_SIZE;
843 		status = nfsd_create(rqstp, &cstate->current_fh,
844 				     create->cr_name, create->cr_namelen,
845 				     &attrs, S_IFDIR, 0, &resfh);
846 		break;
847 
848 	default:
849 		status = nfserr_badtype;
850 	}
851 
852 	if (status)
853 		goto out;
854 
855 	if (attrs.na_labelerr)
856 		create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
857 	if (attrs.na_aclerr)
858 		create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
859 	set_change_info(&create->cr_cinfo, &cstate->current_fh);
860 	fh_dup2(&cstate->current_fh, &resfh);
861 out:
862 	fh_put(&resfh);
863 out_umask:
864 	current->fs->umask = 0;
865 	nfsd_attrs_free(&attrs);
866 	return status;
867 }
868 
869 static __be32
870 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
871 	      union nfsd4_op_u *u)
872 {
873 	struct nfsd4_getattr *getattr = &u->getattr;
874 	__be32 status;
875 
876 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
877 	if (status)
878 		return status;
879 
880 	if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
881 		return nfserr_inval;
882 
883 	getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
884 	getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
885 	getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
886 
887 	getattr->ga_fhp = &cstate->current_fh;
888 	return nfs_ok;
889 }
890 
891 static __be32
892 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
893 	   union nfsd4_op_u *u)
894 {
895 	struct nfsd4_link *link = &u->link;
896 	__be32 status;
897 
898 	status = nfsd_link(rqstp, &cstate->current_fh,
899 			   link->li_name, link->li_namelen, &cstate->save_fh);
900 	if (!status)
901 		set_change_info(&link->li_cinfo, &cstate->current_fh);
902 	return status;
903 }
904 
905 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
906 {
907 	struct svc_fh tmp_fh;
908 	__be32 ret;
909 
910 	fh_init(&tmp_fh, NFS4_FHSIZE);
911 	ret = exp_pseudoroot(rqstp, &tmp_fh);
912 	if (ret)
913 		return ret;
914 	if (tmp_fh.fh_dentry == fh->fh_dentry) {
915 		fh_put(&tmp_fh);
916 		return nfserr_noent;
917 	}
918 	fh_put(&tmp_fh);
919 	return nfsd_lookup(rqstp, fh, "..", 2, fh);
920 }
921 
922 static __be32
923 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
924 	      union nfsd4_op_u *u)
925 {
926 	return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
927 }
928 
929 static __be32
930 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
931 	     union nfsd4_op_u *u)
932 {
933 	return nfsd_lookup(rqstp, &cstate->current_fh,
934 			   u->lookup.lo_name, u->lookup.lo_len,
935 			   &cstate->current_fh);
936 }
937 
938 static __be32
939 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
940 	   union nfsd4_op_u *u)
941 {
942 	struct nfsd4_read *read = &u->read;
943 	__be32 status;
944 
945 	read->rd_nf = NULL;
946 
947 	trace_nfsd_read_start(rqstp, &cstate->current_fh,
948 			      read->rd_offset, read->rd_length);
949 
950 	read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
951 	if (read->rd_offset > (u64)OFFSET_MAX)
952 		read->rd_offset = (u64)OFFSET_MAX;
953 	if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
954 		read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
955 
956 	/*
957 	 * If we do a zero copy read, then a client will see read data
958 	 * that reflects the state of the file *after* performing the
959 	 * following compound.
960 	 *
961 	 * To ensure proper ordering, we therefore turn off zero copy if
962 	 * the client wants us to do more in this compound:
963 	 */
964 	if (!nfsd4_last_compound_op(rqstp))
965 		clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
966 
967 	/* check stateid */
968 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
969 					&read->rd_stateid, RD_STATE,
970 					&read->rd_nf, NULL);
971 
972 	read->rd_rqstp = rqstp;
973 	read->rd_fhp = &cstate->current_fh;
974 	return status;
975 }
976 
977 
978 static void
979 nfsd4_read_release(union nfsd4_op_u *u)
980 {
981 	if (u->read.rd_nf)
982 		nfsd_file_put(u->read.rd_nf);
983 	trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
984 			     u->read.rd_offset, u->read.rd_length);
985 }
986 
987 static __be32
988 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
989 	      union nfsd4_op_u *u)
990 {
991 	struct nfsd4_readdir *readdir = &u->readdir;
992 	u64 cookie = readdir->rd_cookie;
993 	static const nfs4_verifier zeroverf;
994 
995 	/* no need to check permission - this will be done in nfsd_readdir() */
996 
997 	if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
998 		return nfserr_inval;
999 
1000 	readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1001 	readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1002 	readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1003 
1004 	if ((cookie == 1) || (cookie == 2) ||
1005 	    (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1006 		return nfserr_bad_cookie;
1007 
1008 	readdir->rd_rqstp = rqstp;
1009 	readdir->rd_fhp = &cstate->current_fh;
1010 	return nfs_ok;
1011 }
1012 
1013 static __be32
1014 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1015 	       union nfsd4_op_u *u)
1016 {
1017 	u->readlink.rl_rqstp = rqstp;
1018 	u->readlink.rl_fhp = &cstate->current_fh;
1019 	return nfs_ok;
1020 }
1021 
1022 static __be32
1023 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1024 	     union nfsd4_op_u *u)
1025 {
1026 	struct nfsd4_remove *remove = &u->remove;
1027 	__be32 status;
1028 
1029 	if (opens_in_grace(SVC_NET(rqstp)))
1030 		return nfserr_grace;
1031 	status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1032 			     remove->rm_name, remove->rm_namelen);
1033 	if (!status)
1034 		set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1035 	return status;
1036 }
1037 
1038 static __be32
1039 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1040 	     union nfsd4_op_u *u)
1041 {
1042 	struct nfsd4_rename *rename = &u->rename;
1043 	__be32 status;
1044 
1045 	if (opens_in_grace(SVC_NET(rqstp)))
1046 		return nfserr_grace;
1047 	status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1048 			     rename->rn_snamelen, &cstate->current_fh,
1049 			     rename->rn_tname, rename->rn_tnamelen);
1050 	if (status)
1051 		return status;
1052 	set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1053 	set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1054 	return nfs_ok;
1055 }
1056 
1057 static __be32
1058 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1059 	      union nfsd4_op_u *u)
1060 {
1061 	struct nfsd4_secinfo *secinfo = &u->secinfo;
1062 	struct svc_export *exp;
1063 	struct dentry *dentry;
1064 	__be32 err;
1065 
1066 	err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1067 	if (err)
1068 		return err;
1069 	err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1070 				    secinfo->si_name, secinfo->si_namelen,
1071 				    &exp, &dentry);
1072 	if (err)
1073 		return err;
1074 	if (d_really_is_negative(dentry)) {
1075 		exp_put(exp);
1076 		err = nfserr_noent;
1077 	} else
1078 		secinfo->si_exp = exp;
1079 	dput(dentry);
1080 	if (cstate->minorversion)
1081 		/* See rfc 5661 section 2.6.3.1.1.8 */
1082 		fh_put(&cstate->current_fh);
1083 	return err;
1084 }
1085 
1086 static __be32
1087 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1088 		union nfsd4_op_u *u)
1089 {
1090 	__be32 err;
1091 
1092 	switch (u->secinfo_no_name.sin_style) {
1093 	case NFS4_SECINFO_STYLE4_CURRENT_FH:
1094 		break;
1095 	case NFS4_SECINFO_STYLE4_PARENT:
1096 		err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1097 		if (err)
1098 			return err;
1099 		break;
1100 	default:
1101 		return nfserr_inval;
1102 	}
1103 
1104 	u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1105 	fh_put(&cstate->current_fh);
1106 	return nfs_ok;
1107 }
1108 
1109 static void
1110 nfsd4_secinfo_release(union nfsd4_op_u *u)
1111 {
1112 	if (u->secinfo.si_exp)
1113 		exp_put(u->secinfo.si_exp);
1114 }
1115 
1116 static void
1117 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1118 {
1119 	if (u->secinfo_no_name.sin_exp)
1120 		exp_put(u->secinfo_no_name.sin_exp);
1121 }
1122 
1123 static __be32
1124 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1125 	      union nfsd4_op_u *u)
1126 {
1127 	struct nfsd4_setattr *setattr = &u->setattr;
1128 	struct nfsd_attrs attrs = {
1129 		.na_iattr	= &setattr->sa_iattr,
1130 		.na_seclabel	= &setattr->sa_label,
1131 	};
1132 	struct inode *inode;
1133 	__be32 status = nfs_ok;
1134 	int err;
1135 
1136 	if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1137 		status = nfs4_preprocess_stateid_op(rqstp, cstate,
1138 				&cstate->current_fh, &setattr->sa_stateid,
1139 				WR_STATE, NULL, NULL);
1140 		if (status)
1141 			return status;
1142 	}
1143 	err = fh_want_write(&cstate->current_fh);
1144 	if (err)
1145 		return nfserrno(err);
1146 	status = nfs_ok;
1147 
1148 	status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1149 				    nfsd_attrmask);
1150 	if (status)
1151 		goto out;
1152 
1153 	inode = cstate->current_fh.fh_dentry->d_inode;
1154 	status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1155 				   setattr->sa_acl, &attrs);
1156 
1157 	if (status)
1158 		goto out;
1159 	status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
1160 				0, (time64_t)0);
1161 	if (!status)
1162 		status = nfserrno(attrs.na_labelerr);
1163 	if (!status)
1164 		status = nfserrno(attrs.na_aclerr);
1165 out:
1166 	nfsd_attrs_free(&attrs);
1167 	fh_drop_write(&cstate->current_fh);
1168 	return status;
1169 }
1170 
1171 static __be32
1172 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1173 	    union nfsd4_op_u *u)
1174 {
1175 	struct nfsd4_write *write = &u->write;
1176 	stateid_t *stateid = &write->wr_stateid;
1177 	struct nfsd_file *nf = NULL;
1178 	__be32 status = nfs_ok;
1179 	unsigned long cnt;
1180 	int nvecs;
1181 
1182 	if (write->wr_offset > (u64)OFFSET_MAX ||
1183 	    write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1184 		return nfserr_fbig;
1185 
1186 	cnt = write->wr_buflen;
1187 	trace_nfsd_write_start(rqstp, &cstate->current_fh,
1188 			       write->wr_offset, cnt);
1189 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1190 						stateid, WR_STATE, &nf, NULL);
1191 	if (status)
1192 		return status;
1193 
1194 	write->wr_how_written = write->wr_stable_how;
1195 
1196 	nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1197 	WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1198 
1199 	status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1200 				write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1201 				write->wr_how_written,
1202 				(__be32 *)write->wr_verifier.data);
1203 	nfsd_file_put(nf);
1204 
1205 	write->wr_bytes_written = cnt;
1206 	trace_nfsd_write_done(rqstp, &cstate->current_fh,
1207 			      write->wr_offset, cnt);
1208 	return status;
1209 }
1210 
1211 static __be32
1212 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1213 		  stateid_t *src_stateid, struct nfsd_file **src,
1214 		  stateid_t *dst_stateid, struct nfsd_file **dst)
1215 {
1216 	__be32 status;
1217 
1218 	if (!cstate->save_fh.fh_dentry)
1219 		return nfserr_nofilehandle;
1220 
1221 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1222 					    src_stateid, RD_STATE, src, NULL);
1223 	if (status)
1224 		goto out;
1225 
1226 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1227 					    dst_stateid, WR_STATE, dst, NULL);
1228 	if (status)
1229 		goto out_put_src;
1230 
1231 	/* fix up for NFS-specific error code */
1232 	if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1233 	    !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1234 		status = nfserr_wrong_type;
1235 		goto out_put_dst;
1236 	}
1237 
1238 out:
1239 	return status;
1240 out_put_dst:
1241 	nfsd_file_put(*dst);
1242 	*dst = NULL;
1243 out_put_src:
1244 	nfsd_file_put(*src);
1245 	*src = NULL;
1246 	goto out;
1247 }
1248 
1249 static __be32
1250 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1251 		union nfsd4_op_u *u)
1252 {
1253 	struct nfsd4_clone *clone = &u->clone;
1254 	struct nfsd_file *src, *dst;
1255 	__be32 status;
1256 
1257 	status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1258 				   &clone->cl_dst_stateid, &dst);
1259 	if (status)
1260 		goto out;
1261 
1262 	status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1263 			dst, clone->cl_dst_pos, clone->cl_count,
1264 			EX_ISSYNC(cstate->current_fh.fh_export));
1265 
1266 	nfsd_file_put(dst);
1267 	nfsd_file_put(src);
1268 out:
1269 	return status;
1270 }
1271 
1272 static void nfs4_put_copy(struct nfsd4_copy *copy)
1273 {
1274 	if (!refcount_dec_and_test(&copy->refcount))
1275 		return;
1276 	atomic_dec(&copy->cp_nn->pending_async_copies);
1277 	kfree(copy->cp_src);
1278 	kfree(copy);
1279 }
1280 
1281 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1282 {
1283 	if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1284 		kthread_stop(copy->copy_task);
1285 	nfs4_put_copy(copy);
1286 }
1287 
1288 static struct nfsd4_copy *nfsd4_unhash_copy(struct nfs4_client *clp)
1289 {
1290 	struct nfsd4_copy *copy = NULL;
1291 
1292 	spin_lock(&clp->async_lock);
1293 	if (!list_empty(&clp->async_copies)) {
1294 		copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1295 					copies);
1296 		refcount_inc(&copy->refcount);
1297 		copy->cp_clp = NULL;
1298 		if (!list_empty(&copy->copies))
1299 			list_del_init(&copy->copies);
1300 	}
1301 	spin_unlock(&clp->async_lock);
1302 	return copy;
1303 }
1304 
1305 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1306 {
1307 	struct nfsd4_copy *copy;
1308 
1309 	while ((copy = nfsd4_unhash_copy(clp)) != NULL)
1310 		nfsd4_stop_copy(copy);
1311 }
1312 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1313 
1314 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1315 				   struct nfs_fh *src_fh,
1316 				   nfs4_stateid *stateid);
1317 extern void nfs42_ssc_close(struct file *filep);
1318 
1319 extern void nfs_sb_deactive(struct super_block *sb);
1320 
1321 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1322 
1323 /*
1324  * setup a work entry in the ssc delayed unmount list.
1325  */
1326 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1327 				  struct nfsd4_ssc_umount_item **nsui)
1328 {
1329 	struct nfsd4_ssc_umount_item *ni = NULL;
1330 	struct nfsd4_ssc_umount_item *work = NULL;
1331 	struct nfsd4_ssc_umount_item *tmp;
1332 	DEFINE_WAIT(wait);
1333 	__be32 status = 0;
1334 
1335 	*nsui = NULL;
1336 	work = kzalloc(sizeof(*work), GFP_KERNEL);
1337 try_again:
1338 	spin_lock(&nn->nfsd_ssc_lock);
1339 	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1340 		if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1341 			continue;
1342 		/* found a match */
1343 		if (ni->nsui_busy) {
1344 			/*  wait - and try again */
1345 			prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1346 			spin_unlock(&nn->nfsd_ssc_lock);
1347 
1348 			/* allow 20secs for mount/unmount for now - revisit */
1349 			if (kthread_should_stop() ||
1350 					(schedule_timeout(20*HZ) == 0)) {
1351 				finish_wait(&nn->nfsd_ssc_waitq, &wait);
1352 				kfree(work);
1353 				return nfserr_eagain;
1354 			}
1355 			finish_wait(&nn->nfsd_ssc_waitq, &wait);
1356 			goto try_again;
1357 		}
1358 		*nsui = ni;
1359 		refcount_inc(&ni->nsui_refcnt);
1360 		spin_unlock(&nn->nfsd_ssc_lock);
1361 		kfree(work);
1362 
1363 		/* return vfsmount in (*nsui)->nsui_vfsmount */
1364 		return 0;
1365 	}
1366 	if (work) {
1367 		strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1368 		refcount_set(&work->nsui_refcnt, 2);
1369 		work->nsui_busy = true;
1370 		list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1371 		*nsui = work;
1372 	} else
1373 		status = nfserr_resource;
1374 	spin_unlock(&nn->nfsd_ssc_lock);
1375 	return status;
1376 }
1377 
1378 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1379 				 struct nfsd4_ssc_umount_item *nsui,
1380 				 struct vfsmount *ss_mnt)
1381 {
1382 	spin_lock(&nn->nfsd_ssc_lock);
1383 	nsui->nsui_vfsmount = ss_mnt;
1384 	nsui->nsui_busy = false;
1385 	wake_up_all(&nn->nfsd_ssc_waitq);
1386 	spin_unlock(&nn->nfsd_ssc_lock);
1387 }
1388 
1389 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1390 				 struct nfsd4_ssc_umount_item *nsui)
1391 {
1392 	spin_lock(&nn->nfsd_ssc_lock);
1393 	list_del(&nsui->nsui_list);
1394 	wake_up_all(&nn->nfsd_ssc_waitq);
1395 	spin_unlock(&nn->nfsd_ssc_lock);
1396 	kfree(nsui);
1397 }
1398 
1399 /*
1400  * Support one copy source server for now.
1401  */
1402 static __be32
1403 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1404 		       struct nfsd4_ssc_umount_item **nsui)
1405 {
1406 	struct file_system_type *type;
1407 	struct vfsmount *ss_mnt;
1408 	struct nfs42_netaddr *naddr;
1409 	struct sockaddr_storage tmp_addr;
1410 	size_t tmp_addrlen, match_netid_len = 3;
1411 	char *startsep = "", *endsep = "", *match_netid = "tcp";
1412 	char *ipaddr, *dev_name, *raw_data;
1413 	int len, raw_len;
1414 	__be32 status = nfserr_inval;
1415 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1416 
1417 	naddr = &nss->u.nl4_addr;
1418 	tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1419 					 naddr->addr_len,
1420 					 (struct sockaddr *)&tmp_addr,
1421 					 sizeof(tmp_addr));
1422 	*nsui = NULL;
1423 	if (tmp_addrlen == 0)
1424 		goto out_err;
1425 
1426 	if (tmp_addr.ss_family == AF_INET6) {
1427 		startsep = "[";
1428 		endsep = "]";
1429 		match_netid = "tcp6";
1430 		match_netid_len = 4;
1431 	}
1432 
1433 	if (naddr->netid_len != match_netid_len ||
1434 		strncmp(naddr->netid, match_netid, naddr->netid_len))
1435 		goto out_err;
1436 
1437 	/* Construct the raw data for the vfs_kern_mount call */
1438 	len = RPC_MAX_ADDRBUFLEN + 1;
1439 	ipaddr = kzalloc(len, GFP_KERNEL);
1440 	if (!ipaddr)
1441 		goto out_err;
1442 
1443 	rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1444 
1445 	/* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1446 
1447 	raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1448 	raw_data = kzalloc(raw_len, GFP_KERNEL);
1449 	if (!raw_data)
1450 		goto out_free_ipaddr;
1451 
1452 	snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1453 
1454 	status = nfserr_nodev;
1455 	type = get_fs_type("nfs");
1456 	if (!type)
1457 		goto out_free_rawdata;
1458 
1459 	/* Set the server:<export> for the vfs_kern_mount call */
1460 	dev_name = kzalloc(len + 5, GFP_KERNEL);
1461 	if (!dev_name)
1462 		goto out_free_rawdata;
1463 	snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1464 
1465 	status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui);
1466 	if (status)
1467 		goto out_free_devname;
1468 	if ((*nsui)->nsui_vfsmount)
1469 		goto out_done;
1470 
1471 	/* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1472 	ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1473 	module_put(type->owner);
1474 	if (IS_ERR(ss_mnt)) {
1475 		status = nfserr_nodev;
1476 		nfsd4_ssc_cancel_dul(nn, *nsui);
1477 		goto out_free_devname;
1478 	}
1479 	nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1480 out_done:
1481 	status = 0;
1482 
1483 out_free_devname:
1484 	kfree(dev_name);
1485 out_free_rawdata:
1486 	kfree(raw_data);
1487 out_free_ipaddr:
1488 	kfree(ipaddr);
1489 out_err:
1490 	return status;
1491 }
1492 
1493 /*
1494  * Verify COPY destination stateid.
1495  *
1496  * Connect to the source server with NFSv4.1.
1497  * Create the source struct file for nfsd_copy_range.
1498  * Called with COPY cstate:
1499  *    SAVED_FH: source filehandle
1500  *    CURRENT_FH: destination filehandle
1501  */
1502 static __be32
1503 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1504 		      struct nfsd4_compound_state *cstate,
1505 		      struct nfsd4_copy *copy)
1506 {
1507 	struct svc_fh *s_fh = NULL;
1508 	stateid_t *s_stid = &copy->cp_src_stateid;
1509 	__be32 status = nfserr_inval;
1510 
1511 	/* Verify the destination stateid and set dst struct file*/
1512 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1513 					    &copy->cp_dst_stateid,
1514 					    WR_STATE, &copy->nf_dst, NULL);
1515 	if (status)
1516 		goto out;
1517 
1518 	status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1519 	if (status)
1520 		goto out;
1521 
1522 	s_fh = &cstate->save_fh;
1523 
1524 	copy->c_fh.size = s_fh->fh_handle.fh_size;
1525 	memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1526 	copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1527 	memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1528 	       sizeof(stateid_opaque_t));
1529 
1530 	status = 0;
1531 out:
1532 	return status;
1533 }
1534 
1535 static void
1536 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1537 			struct nfsd_file *dst)
1538 {
1539 	struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1540 	long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1541 
1542 	nfs42_ssc_close(filp);
1543 	fput(filp);
1544 
1545 	spin_lock(&nn->nfsd_ssc_lock);
1546 	list_del(&nsui->nsui_list);
1547 	/*
1548 	 * vfsmount can be shared by multiple exports,
1549 	 * decrement refcnt. If the count drops to 1 it
1550 	 * will be unmounted when nsui_expire expires.
1551 	 */
1552 	refcount_dec(&nsui->nsui_refcnt);
1553 	nsui->nsui_expire = jiffies + timeout;
1554 	list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1555 	spin_unlock(&nn->nfsd_ssc_lock);
1556 }
1557 
1558 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1559 
1560 static __be32
1561 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1562 		      struct nfsd4_compound_state *cstate,
1563 		      struct nfsd4_copy *copy)
1564 {
1565 	return nfserr_inval;
1566 }
1567 
1568 static void
1569 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1570 			struct nfsd_file *dst)
1571 {
1572 }
1573 
1574 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1575 				   struct nfs_fh *src_fh,
1576 				   nfs4_stateid *stateid)
1577 {
1578 	return NULL;
1579 }
1580 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1581 
1582 static __be32
1583 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1584 		      struct nfsd4_compound_state *cstate,
1585 		      struct nfsd4_copy *copy)
1586 {
1587 	return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1588 				 &copy->nf_src, &copy->cp_dst_stateid,
1589 				 &copy->nf_dst);
1590 }
1591 
1592 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1593 {
1594 	struct nfsd4_cb_offload *cbo =
1595 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1596 
1597 	kfree(cbo);
1598 }
1599 
1600 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1601 				 struct rpc_task *task)
1602 {
1603 	struct nfsd4_cb_offload *cbo =
1604 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1605 
1606 	trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1607 	return 1;
1608 }
1609 
1610 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1611 	.release = nfsd4_cb_offload_release,
1612 	.done = nfsd4_cb_offload_done
1613 };
1614 
1615 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1616 {
1617 	copy->cp_res.wr_stable_how =
1618 		test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1619 			NFS_FILE_SYNC : NFS_UNSTABLE;
1620 	nfsd4_copy_set_sync(copy, sync);
1621 }
1622 
1623 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1624 				     struct file *dst,
1625 				     struct file *src)
1626 {
1627 	errseq_t since;
1628 	ssize_t bytes_copied = 0;
1629 	u64 bytes_total = copy->cp_count;
1630 	u64 src_pos = copy->cp_src_pos;
1631 	u64 dst_pos = copy->cp_dst_pos;
1632 	int status;
1633 	loff_t end;
1634 
1635 	/* See RFC 7862 p.67: */
1636 	if (bytes_total == 0)
1637 		bytes_total = ULLONG_MAX;
1638 	do {
1639 		if (kthread_should_stop())
1640 			break;
1641 		bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1642 						    bytes_total);
1643 		if (bytes_copied <= 0)
1644 			break;
1645 		bytes_total -= bytes_copied;
1646 		copy->cp_res.wr_bytes_written += bytes_copied;
1647 		src_pos += bytes_copied;
1648 		dst_pos += bytes_copied;
1649 	} while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1650 	/* for a non-zero asynchronous copy do a commit of data */
1651 	if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1652 		since = READ_ONCE(dst->f_wb_err);
1653 		end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1654 		status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1655 		if (!status)
1656 			status = filemap_check_wb_err(dst->f_mapping, since);
1657 		if (!status)
1658 			set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1659 	}
1660 	return bytes_copied;
1661 }
1662 
1663 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1664 			    struct file *src, struct file *dst,
1665 			    bool sync)
1666 {
1667 	__be32 status;
1668 	ssize_t bytes;
1669 
1670 	bytes = _nfsd_copy_file_range(copy, dst, src);
1671 
1672 	/* for async copy, we ignore the error, client can always retry
1673 	 * to get the error
1674 	 */
1675 	if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1676 		status = nfserrno(bytes);
1677 	else {
1678 		nfsd4_init_copy_res(copy, sync);
1679 		status = nfs_ok;
1680 	}
1681 	return status;
1682 }
1683 
1684 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1685 {
1686 	dst->cp_src_pos = src->cp_src_pos;
1687 	dst->cp_dst_pos = src->cp_dst_pos;
1688 	dst->cp_count = src->cp_count;
1689 	dst->cp_flags = src->cp_flags;
1690 	memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1691 	memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1692 	dst->cp_clp = src->cp_clp;
1693 	dst->nf_dst = nfsd_file_get(src->nf_dst);
1694 	/* for inter, nf_src doesn't exist yet */
1695 	if (!nfsd4_ssc_is_inter(src))
1696 		dst->nf_src = nfsd_file_get(src->nf_src);
1697 
1698 	memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1699 	memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1700 	memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1701 	memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1702 	dst->ss_nsui = src->ss_nsui;
1703 }
1704 
1705 static void release_copy_files(struct nfsd4_copy *copy)
1706 {
1707 	if (copy->nf_src)
1708 		nfsd_file_put(copy->nf_src);
1709 	if (copy->nf_dst)
1710 		nfsd_file_put(copy->nf_dst);
1711 }
1712 
1713 static void cleanup_async_copy(struct nfsd4_copy *copy)
1714 {
1715 	nfs4_free_copy_state(copy);
1716 	release_copy_files(copy);
1717 	if (copy->cp_clp) {
1718 		spin_lock(&copy->cp_clp->async_lock);
1719 		if (!list_empty(&copy->copies))
1720 			list_del_init(&copy->copies);
1721 		spin_unlock(&copy->cp_clp->async_lock);
1722 	}
1723 	nfs4_put_copy(copy);
1724 }
1725 
1726 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy, __be32 nfserr)
1727 {
1728 	struct nfsd4_cb_offload *cbo;
1729 
1730 	cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1731 	if (!cbo)
1732 		return;
1733 
1734 	memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1735 	memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1736 	cbo->co_nfserr = nfserr;
1737 
1738 	nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1739 		      NFSPROC4_CLNT_CB_OFFLOAD);
1740 	trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1741 			      &cbo->co_fh, copy->cp_count, nfserr);
1742 	nfsd4_run_cb(&cbo->co_cb);
1743 }
1744 
1745 /**
1746  * nfsd4_do_async_copy - kthread function for background server-side COPY
1747  * @data: arguments for COPY operation
1748  *
1749  * Return values:
1750  *   %0: Copy operation is done.
1751  */
1752 static int nfsd4_do_async_copy(void *data)
1753 {
1754 	struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1755 	__be32 nfserr;
1756 
1757 	if (nfsd4_ssc_is_inter(copy)) {
1758 		struct file *filp;
1759 
1760 		filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1761 				      &copy->c_fh, &copy->stateid);
1762 		if (IS_ERR(filp)) {
1763 			switch (PTR_ERR(filp)) {
1764 			case -EBADF:
1765 				nfserr = nfserr_wrong_type;
1766 				break;
1767 			default:
1768 				nfserr = nfserr_offload_denied;
1769 			}
1770 			/* ss_mnt will be unmounted by the laundromat */
1771 			goto do_callback;
1772 		}
1773 		nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1774 				       false);
1775 		nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1776 	} else {
1777 		nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1778 				       copy->nf_dst->nf_file, false);
1779 	}
1780 
1781 do_callback:
1782 	nfsd4_send_cb_offload(copy, nfserr);
1783 	cleanup_async_copy(copy);
1784 	return 0;
1785 }
1786 
1787 static __be32
1788 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1789 		union nfsd4_op_u *u)
1790 {
1791 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1792 	struct nfsd4_copy *async_copy = NULL;
1793 	struct nfsd4_copy *copy = &u->copy;
1794 	struct nfsd42_write_res *result;
1795 	__be32 status;
1796 
1797 	result = &copy->cp_res;
1798 	nfsd_copy_write_verifier((__be32 *)&result->wr_verifier.data, nn);
1799 
1800 	copy->cp_clp = cstate->clp;
1801 	if (nfsd4_ssc_is_inter(copy)) {
1802 		if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1803 			status = nfserr_notsupp;
1804 			goto out;
1805 		}
1806 		status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1807 		if (status)
1808 			return nfserr_offload_denied;
1809 	} else {
1810 		status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1811 		if (status)
1812 			return status;
1813 	}
1814 
1815 	memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1816 		sizeof(struct knfsd_fh));
1817 	if (nfsd4_copy_is_async(copy)) {
1818 		async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1819 		if (!async_copy)
1820 			goto out_err;
1821 		async_copy->cp_nn = nn;
1822 		INIT_LIST_HEAD(&async_copy->copies);
1823 		refcount_set(&async_copy->refcount, 1);
1824 		/* Arbitrary cap on number of pending async copy operations */
1825 		if (atomic_inc_return(&nn->pending_async_copies) >
1826 				(int)rqstp->rq_pool->sp_nrthreads)
1827 			goto out_err;
1828 		async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1829 		if (!async_copy->cp_src)
1830 			goto out_err;
1831 		if (!nfs4_init_copy_state(nn, copy))
1832 			goto out_err;
1833 		memcpy(&result->cb_stateid, &copy->cp_stateid.cs_stid,
1834 			sizeof(result->cb_stateid));
1835 		dup_copy_fields(copy, async_copy);
1836 		async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1837 				async_copy, "%s", "copy thread");
1838 		if (IS_ERR(async_copy->copy_task))
1839 			goto out_err;
1840 		spin_lock(&async_copy->cp_clp->async_lock);
1841 		list_add(&async_copy->copies,
1842 				&async_copy->cp_clp->async_copies);
1843 		spin_unlock(&async_copy->cp_clp->async_lock);
1844 		wake_up_process(async_copy->copy_task);
1845 		status = nfs_ok;
1846 	} else {
1847 		status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1848 				       copy->nf_dst->nf_file, true);
1849 	}
1850 out:
1851 	release_copy_files(copy);
1852 	return status;
1853 out_err:
1854 	if (nfsd4_ssc_is_inter(copy)) {
1855 		/*
1856 		 * Source's vfsmount of inter-copy will be unmounted
1857 		 * by the laundromat. Use copy instead of async_copy
1858 		 * since async_copy->ss_nsui might not be set yet.
1859 		 */
1860 		refcount_dec(&copy->ss_nsui->nsui_refcnt);
1861 	}
1862 	if (async_copy)
1863 		cleanup_async_copy(async_copy);
1864 	status = nfserr_jukebox;
1865 	goto out;
1866 }
1867 
1868 static struct nfsd4_copy *
1869 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1870 {
1871 	struct nfsd4_copy *copy;
1872 
1873 	lockdep_assert_held(&clp->async_lock);
1874 
1875 	list_for_each_entry(copy, &clp->async_copies, copies) {
1876 		if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1877 			continue;
1878 		return copy;
1879 	}
1880 	return NULL;
1881 }
1882 
1883 static struct nfsd4_copy *
1884 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1885 {
1886 	struct nfsd4_copy *copy;
1887 
1888 	spin_lock(&clp->async_lock);
1889 	copy = find_async_copy_locked(clp, stateid);
1890 	if (copy)
1891 		refcount_inc(&copy->refcount);
1892 	spin_unlock(&clp->async_lock);
1893 	return copy;
1894 }
1895 
1896 static __be32
1897 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1898 		     struct nfsd4_compound_state *cstate,
1899 		     union nfsd4_op_u *u)
1900 {
1901 	struct nfsd4_offload_status *os = &u->offload_status;
1902 	struct nfsd4_copy *copy;
1903 	struct nfs4_client *clp = cstate->clp;
1904 
1905 	copy = find_async_copy(clp, &os->stateid);
1906 	if (!copy) {
1907 		struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1908 
1909 		return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1910 	} else
1911 		nfsd4_stop_copy(copy);
1912 
1913 	return nfs_ok;
1914 }
1915 
1916 static __be32
1917 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1918 		  union nfsd4_op_u *u)
1919 {
1920 	struct nfsd4_copy_notify *cn = &u->copy_notify;
1921 	__be32 status;
1922 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1923 	struct nfs4_stid *stid;
1924 	struct nfs4_cpntf_state *cps;
1925 	struct nfs4_client *clp = cstate->clp;
1926 
1927 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1928 					&cn->cpn_src_stateid, RD_STATE, NULL,
1929 					&stid);
1930 	if (status)
1931 		return status;
1932 
1933 	cn->cpn_sec = nn->nfsd4_lease;
1934 	cn->cpn_nsec = 0;
1935 
1936 	status = nfserrno(-ENOMEM);
1937 	cps = nfs4_alloc_init_cpntf_state(nn, stid);
1938 	if (!cps)
1939 		goto out;
1940 	memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1941 	memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1942 	memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1943 
1944 	/* For now, only return one server address in cpn_src, the
1945 	 * address used by the client to connect to this server.
1946 	 */
1947 	cn->cpn_src->nl4_type = NL4_NETADDR;
1948 	status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1949 				 &cn->cpn_src->u.nl4_addr);
1950 	WARN_ON_ONCE(status);
1951 	if (status) {
1952 		nfs4_put_cpntf_state(nn, cps);
1953 		goto out;
1954 	}
1955 out:
1956 	nfs4_put_stid(stid);
1957 	return status;
1958 }
1959 
1960 static __be32
1961 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1962 		struct nfsd4_fallocate *fallocate, int flags)
1963 {
1964 	__be32 status;
1965 	struct nfsd_file *nf;
1966 
1967 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1968 					    &fallocate->falloc_stateid,
1969 					    WR_STATE, &nf, NULL);
1970 	if (status != nfs_ok)
1971 		return status;
1972 
1973 	status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1974 				     fallocate->falloc_offset,
1975 				     fallocate->falloc_length,
1976 				     flags);
1977 	nfsd_file_put(nf);
1978 	return status;
1979 }
1980 
1981 static __be32
1982 nfsd4_offload_status(struct svc_rqst *rqstp,
1983 		     struct nfsd4_compound_state *cstate,
1984 		     union nfsd4_op_u *u)
1985 {
1986 	struct nfsd4_offload_status *os = &u->offload_status;
1987 	__be32 status = nfs_ok;
1988 	struct nfsd4_copy *copy;
1989 	struct nfs4_client *clp = cstate->clp;
1990 
1991 	spin_lock(&clp->async_lock);
1992 	copy = find_async_copy_locked(clp, &os->stateid);
1993 	if (copy)
1994 		os->count = copy->cp_res.wr_bytes_written;
1995 	else
1996 		status = nfserr_bad_stateid;
1997 	spin_unlock(&clp->async_lock);
1998 
1999 	return status;
2000 }
2001 
2002 static __be32
2003 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2004 	       union nfsd4_op_u *u)
2005 {
2006 	return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2007 }
2008 
2009 static __be32
2010 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2011 		 union nfsd4_op_u *u)
2012 {
2013 	return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2014 			       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2015 }
2016 
2017 static __be32
2018 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2019 	   union nfsd4_op_u *u)
2020 {
2021 	struct nfsd4_seek *seek = &u->seek;
2022 	int whence;
2023 	__be32 status;
2024 	struct nfsd_file *nf;
2025 
2026 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2027 					    &seek->seek_stateid,
2028 					    RD_STATE, &nf, NULL);
2029 	if (status)
2030 		return status;
2031 
2032 	switch (seek->seek_whence) {
2033 	case NFS4_CONTENT_DATA:
2034 		whence = SEEK_DATA;
2035 		break;
2036 	case NFS4_CONTENT_HOLE:
2037 		whence = SEEK_HOLE;
2038 		break;
2039 	default:
2040 		status = nfserr_union_notsupp;
2041 		goto out;
2042 	}
2043 
2044 	/*
2045 	 * Note:  This call does change file->f_pos, but nothing in NFSD
2046 	 *        should ever file->f_pos.
2047 	 */
2048 	seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2049 	if (seek->seek_pos < 0)
2050 		status = nfserrno(seek->seek_pos);
2051 	else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2052 		seek->seek_eof = true;
2053 
2054 out:
2055 	nfsd_file_put(nf);
2056 	return status;
2057 }
2058 
2059 /* This routine never returns NFS_OK!  If there are no other errors, it
2060  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2061  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2062  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2063  */
2064 static __be32
2065 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2066 	     struct nfsd4_verify *verify)
2067 {
2068 	__be32 *buf, *p;
2069 	int count;
2070 	__be32 status;
2071 
2072 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2073 	if (status)
2074 		return status;
2075 
2076 	status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2077 	if (status)
2078 		return status;
2079 
2080 	if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2081 	    || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2082 		return nfserr_inval;
2083 	if (verify->ve_attrlen & 3)
2084 		return nfserr_inval;
2085 
2086 	/* count in words:
2087 	 *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2088 	 */
2089 	count = 4 + (verify->ve_attrlen >> 2);
2090 	buf = kmalloc(count << 2, GFP_KERNEL);
2091 	if (!buf)
2092 		return nfserr_jukebox;
2093 
2094 	p = buf;
2095 	status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2096 				    cstate->current_fh.fh_export,
2097 				    cstate->current_fh.fh_dentry,
2098 				    verify->ve_bmval,
2099 				    rqstp, 0);
2100 	/*
2101 	 * If nfsd4_encode_fattr() ran out of space, assume that's because
2102 	 * the attributes are longer (hence different) than those given:
2103 	 */
2104 	if (status == nfserr_resource)
2105 		status = nfserr_not_same;
2106 	if (status)
2107 		goto out_kfree;
2108 
2109 	/* skip bitmap */
2110 	p = buf + 1 + ntohl(buf[0]);
2111 	status = nfserr_not_same;
2112 	if (ntohl(*p++) != verify->ve_attrlen)
2113 		goto out_kfree;
2114 	if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2115 		status = nfserr_same;
2116 
2117 out_kfree:
2118 	kfree(buf);
2119 	return status;
2120 }
2121 
2122 static __be32
2123 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2124 	      union nfsd4_op_u *u)
2125 {
2126 	__be32 status;
2127 
2128 	status = _nfsd4_verify(rqstp, cstate, &u->verify);
2129 	return status == nfserr_not_same ? nfs_ok : status;
2130 }
2131 
2132 static __be32
2133 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2134 	     union nfsd4_op_u *u)
2135 {
2136 	__be32 status;
2137 
2138 	status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2139 	return status == nfserr_same ? nfs_ok : status;
2140 }
2141 
2142 #ifdef CONFIG_NFSD_PNFS
2143 static const struct nfsd4_layout_ops *
2144 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2145 {
2146 	if (!exp->ex_layout_types) {
2147 		dprintk("%s: export does not support pNFS\n", __func__);
2148 		return NULL;
2149 	}
2150 
2151 	if (layout_type >= LAYOUT_TYPE_MAX ||
2152 	    !(exp->ex_layout_types & (1 << layout_type))) {
2153 		dprintk("%s: layout type %d not supported\n",
2154 			__func__, layout_type);
2155 		return NULL;
2156 	}
2157 
2158 	return nfsd4_layout_ops[layout_type];
2159 }
2160 
2161 static __be32
2162 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2163 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2164 {
2165 	struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2166 	const struct nfsd4_layout_ops *ops;
2167 	struct nfsd4_deviceid_map *map;
2168 	struct svc_export *exp;
2169 	__be32 nfserr;
2170 
2171 	dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2172 	       __func__,
2173 	       gdp->gd_layout_type,
2174 	       gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2175 	       gdp->gd_maxcount);
2176 
2177 	map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2178 	if (!map) {
2179 		dprintk("%s: couldn't find device ID to export mapping!\n",
2180 			__func__);
2181 		return nfserr_noent;
2182 	}
2183 
2184 	exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2185 	if (IS_ERR(exp)) {
2186 		dprintk("%s: could not find device id\n", __func__);
2187 		return nfserr_noent;
2188 	}
2189 
2190 	nfserr = nfserr_layoutunavailable;
2191 	ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2192 	if (!ops)
2193 		goto out;
2194 
2195 	nfserr = nfs_ok;
2196 	if (gdp->gd_maxcount != 0) {
2197 		nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2198 				rqstp, cstate->clp, gdp);
2199 	}
2200 
2201 	gdp->gd_notify_types &= ops->notify_types;
2202 out:
2203 	exp_put(exp);
2204 	return nfserr;
2205 }
2206 
2207 static void
2208 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2209 {
2210 	kfree(u->getdeviceinfo.gd_device);
2211 }
2212 
2213 static __be32
2214 nfsd4_layoutget(struct svc_rqst *rqstp,
2215 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2216 {
2217 	struct nfsd4_layoutget *lgp = &u->layoutget;
2218 	struct svc_fh *current_fh = &cstate->current_fh;
2219 	const struct nfsd4_layout_ops *ops;
2220 	struct nfs4_layout_stateid *ls;
2221 	__be32 nfserr;
2222 	int accmode = NFSD_MAY_READ_IF_EXEC | NFSD_MAY_OWNER_OVERRIDE;
2223 
2224 	switch (lgp->lg_seg.iomode) {
2225 	case IOMODE_READ:
2226 		accmode |= NFSD_MAY_READ;
2227 		break;
2228 	case IOMODE_RW:
2229 		accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2230 		break;
2231 	default:
2232 		dprintk("%s: invalid iomode %d\n",
2233 			__func__, lgp->lg_seg.iomode);
2234 		nfserr = nfserr_badiomode;
2235 		goto out;
2236 	}
2237 
2238 	nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2239 	if (nfserr)
2240 		goto out;
2241 
2242 	nfserr = nfserr_layoutunavailable;
2243 	ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2244 	if (!ops)
2245 		goto out;
2246 
2247 	/*
2248 	 * Verify minlength and range as per RFC5661:
2249 	 *  o  If loga_length is less than loga_minlength,
2250 	 *     the metadata server MUST return NFS4ERR_INVAL.
2251 	 *  o  If the sum of loga_offset and loga_minlength exceeds
2252 	 *     NFS4_UINT64_MAX, and loga_minlength is not
2253 	 *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2254 	 *  o  If the sum of loga_offset and loga_length exceeds
2255 	 *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2256 	 *     the error NFS4ERR_INVAL MUST result.
2257 	 */
2258 	nfserr = nfserr_inval;
2259 	if (lgp->lg_seg.length < lgp->lg_minlength ||
2260 	    (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2261 	     lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2262 	    (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2263 	     lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2264 		goto out;
2265 	if (lgp->lg_seg.length == 0)
2266 		goto out;
2267 
2268 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2269 						true, lgp->lg_layout_type, &ls);
2270 	if (nfserr) {
2271 		trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2272 		goto out;
2273 	}
2274 
2275 	nfserr = nfserr_recallconflict;
2276 	if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2277 		goto out_put_stid;
2278 
2279 	nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2280 				     current_fh, lgp);
2281 	if (nfserr)
2282 		goto out_put_stid;
2283 
2284 	nfserr = nfsd4_insert_layout(lgp, ls);
2285 
2286 out_put_stid:
2287 	mutex_unlock(&ls->ls_mutex);
2288 	nfs4_put_stid(&ls->ls_stid);
2289 out:
2290 	return nfserr;
2291 }
2292 
2293 static void
2294 nfsd4_layoutget_release(union nfsd4_op_u *u)
2295 {
2296 	kfree(u->layoutget.lg_content);
2297 }
2298 
2299 static __be32
2300 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2301 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2302 {
2303 	struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2304 	const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2305 	struct svc_fh *current_fh = &cstate->current_fh;
2306 	const struct nfsd4_layout_ops *ops;
2307 	loff_t new_size = lcp->lc_last_wr + 1;
2308 	struct inode *inode;
2309 	struct nfs4_layout_stateid *ls;
2310 	__be32 nfserr;
2311 
2312 	nfserr = fh_verify(rqstp, current_fh, 0,
2313 			   NFSD_MAY_WRITE | NFSD_MAY_OWNER_OVERRIDE);
2314 	if (nfserr)
2315 		goto out;
2316 
2317 	nfserr = nfserr_layoutunavailable;
2318 	ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2319 	if (!ops)
2320 		goto out;
2321 	inode = d_inode(current_fh->fh_dentry);
2322 
2323 	nfserr = nfserr_inval;
2324 	if (new_size <= seg->offset) {
2325 		dprintk("pnfsd: last write before layout segment\n");
2326 		goto out;
2327 	}
2328 	if (new_size > seg->offset + seg->length) {
2329 		dprintk("pnfsd: last write beyond layout segment\n");
2330 		goto out;
2331 	}
2332 	if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2333 		dprintk("pnfsd: layoutcommit beyond EOF\n");
2334 		goto out;
2335 	}
2336 
2337 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2338 						false, lcp->lc_layout_type,
2339 						&ls);
2340 	if (nfserr) {
2341 		trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2342 		/* fixup error code as per RFC5661 */
2343 		if (nfserr == nfserr_bad_stateid)
2344 			nfserr = nfserr_badlayout;
2345 		goto out;
2346 	}
2347 
2348 	/* LAYOUTCOMMIT does not require any serialization */
2349 	mutex_unlock(&ls->ls_mutex);
2350 
2351 	if (new_size > i_size_read(inode)) {
2352 		lcp->lc_size_chg = 1;
2353 		lcp->lc_newsize = new_size;
2354 	} else {
2355 		lcp->lc_size_chg = 0;
2356 	}
2357 
2358 	nfserr = ops->proc_layoutcommit(inode, lcp);
2359 	nfs4_put_stid(&ls->ls_stid);
2360 out:
2361 	return nfserr;
2362 }
2363 
2364 static __be32
2365 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2366 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2367 {
2368 	struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2369 	struct svc_fh *current_fh = &cstate->current_fh;
2370 	__be32 nfserr;
2371 
2372 	nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2373 	if (nfserr)
2374 		goto out;
2375 
2376 	nfserr = nfserr_layoutunavailable;
2377 	if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2378 		goto out;
2379 
2380 	switch (lrp->lr_seg.iomode) {
2381 	case IOMODE_READ:
2382 	case IOMODE_RW:
2383 	case IOMODE_ANY:
2384 		break;
2385 	default:
2386 		dprintk("%s: invalid iomode %d\n", __func__,
2387 			lrp->lr_seg.iomode);
2388 		nfserr = nfserr_inval;
2389 		goto out;
2390 	}
2391 
2392 	switch (lrp->lr_return_type) {
2393 	case RETURN_FILE:
2394 		nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2395 		break;
2396 	case RETURN_FSID:
2397 	case RETURN_ALL:
2398 		nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2399 		break;
2400 	default:
2401 		dprintk("%s: invalid return_type %d\n", __func__,
2402 			lrp->lr_return_type);
2403 		nfserr = nfserr_inval;
2404 		break;
2405 	}
2406 out:
2407 	return nfserr;
2408 }
2409 #endif /* CONFIG_NFSD_PNFS */
2410 
2411 static __be32
2412 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2413 	       union nfsd4_op_u *u)
2414 {
2415 	struct nfsd4_getxattr *getxattr = &u->getxattr;
2416 
2417 	return nfsd_getxattr(rqstp, &cstate->current_fh,
2418 			     getxattr->getxa_name, &getxattr->getxa_buf,
2419 			     &getxattr->getxa_len);
2420 }
2421 
2422 static __be32
2423 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2424 	   union nfsd4_op_u *u)
2425 {
2426 	struct nfsd4_setxattr *setxattr = &u->setxattr;
2427 	__be32 ret;
2428 
2429 	if (opens_in_grace(SVC_NET(rqstp)))
2430 		return nfserr_grace;
2431 
2432 	ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2433 			    setxattr->setxa_buf, setxattr->setxa_len,
2434 			    setxattr->setxa_flags);
2435 
2436 	if (!ret)
2437 		set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2438 
2439 	return ret;
2440 }
2441 
2442 static __be32
2443 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2444 	   union nfsd4_op_u *u)
2445 {
2446 	/*
2447 	 * Get the entire list, then copy out only the user attributes
2448 	 * in the encode function.
2449 	 */
2450 	return nfsd_listxattr(rqstp, &cstate->current_fh,
2451 			     &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2452 }
2453 
2454 static __be32
2455 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2456 	   union nfsd4_op_u *u)
2457 {
2458 	struct nfsd4_removexattr *removexattr = &u->removexattr;
2459 	__be32 ret;
2460 
2461 	if (opens_in_grace(SVC_NET(rqstp)))
2462 		return nfserr_grace;
2463 
2464 	ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2465 	    removexattr->rmxa_name);
2466 
2467 	if (!ret)
2468 		set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2469 
2470 	return ret;
2471 }
2472 
2473 /*
2474  * NULL call.
2475  */
2476 static __be32
2477 nfsd4_proc_null(struct svc_rqst *rqstp)
2478 {
2479 	return rpc_success;
2480 }
2481 
2482 static inline void nfsd4_increment_op_stats(struct nfsd_net *nn, u32 opnum)
2483 {
2484 	if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2485 		percpu_counter_inc(&nn->counter[NFSD_STATS_NFS4_OP(opnum)]);
2486 }
2487 
2488 static const struct nfsd4_operation nfsd4_ops[];
2489 
2490 static const char *nfsd4_op_name(unsigned opnum);
2491 
2492 /*
2493  * Enforce NFSv4.1 COMPOUND ordering rules:
2494  *
2495  * Also note, enforced elsewhere:
2496  *	- SEQUENCE other than as first op results in
2497  *	  NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2498  *	- BIND_CONN_TO_SESSION must be the only op in its compound.
2499  *	  (Enforced in nfsd4_bind_conn_to_session().)
2500  *	- DESTROY_SESSION must be the final operation in a compound, if
2501  *	  sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2502  *	  (Enforced in nfsd4_destroy_session().)
2503  */
2504 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2505 {
2506 	struct nfsd4_op *first_op = &args->ops[0];
2507 
2508 	/* These ordering requirements don't apply to NFSv4.0: */
2509 	if (args->minorversion == 0)
2510 		return nfs_ok;
2511 	/* This is weird, but OK, not our problem: */
2512 	if (args->opcnt == 0)
2513 		return nfs_ok;
2514 	if (first_op->status == nfserr_op_illegal)
2515 		return nfs_ok;
2516 	if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2517 		return nfserr_op_not_in_session;
2518 	if (first_op->opnum == OP_SEQUENCE)
2519 		return nfs_ok;
2520 	/*
2521 	 * So first_op is something allowed outside a session, like
2522 	 * EXCHANGE_ID; but then it has to be the only op in the
2523 	 * compound:
2524 	 */
2525 	if (args->opcnt != 1)
2526 		return nfserr_not_only_op;
2527 	return nfs_ok;
2528 }
2529 
2530 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2531 {
2532 	return &nfsd4_ops[op->opnum];
2533 }
2534 
2535 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2536 {
2537 	if (op->opnum == OP_ILLEGAL)
2538 		return false;
2539 	return OPDESC(op)->op_flags & OP_CACHEME;
2540 }
2541 
2542 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2543 {
2544 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2545 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2546 	struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2547 	struct nfsd4_op *next = &argp->ops[resp->opcnt];
2548 	const struct nfsd4_operation *thisd = OPDESC(this);
2549 	const struct nfsd4_operation *nextd;
2550 
2551 	/*
2552 	 * Most ops check wronsec on our own; only the putfh-like ops
2553 	 * have special rules.
2554 	 */
2555 	if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2556 		return false;
2557 	/*
2558 	 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2559 	 * put-filehandle operation if we're not going to use the
2560 	 * result:
2561 	 */
2562 	if (argp->opcnt == resp->opcnt)
2563 		return false;
2564 	if (next->opnum == OP_ILLEGAL)
2565 		return false;
2566 	nextd = OPDESC(next);
2567 	/*
2568 	 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2569 	 * errors themselves as necessary; others should check for them
2570 	 * now:
2571 	 */
2572 	return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2573 }
2574 
2575 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2576 static void
2577 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2578 {
2579 	struct nfsd4_op	*op, *current_op = NULL, *saved_op = NULL;
2580 	struct nfsd4_copy *copy;
2581 	struct nfsd4_putfh *putfh;
2582 	int i;
2583 
2584 	/* traverse all operation and if it's a COPY compound, mark the
2585 	 * source filehandle to skip verification
2586 	 */
2587 	for (i = 0; i < args->opcnt; i++) {
2588 		op = &args->ops[i];
2589 		if (op->opnum == OP_PUTFH)
2590 			current_op = op;
2591 		else if (op->opnum == OP_SAVEFH)
2592 			saved_op = current_op;
2593 		else if (op->opnum == OP_RESTOREFH)
2594 			current_op = saved_op;
2595 		else if (op->opnum == OP_COPY) {
2596 			copy = (struct nfsd4_copy *)&op->u;
2597 			if (!saved_op) {
2598 				op->status = nfserr_nofilehandle;
2599 				return;
2600 			}
2601 			putfh = (struct nfsd4_putfh *)&saved_op->u;
2602 			if (nfsd4_ssc_is_inter(copy))
2603 				putfh->no_verify = true;
2604 		}
2605 	}
2606 }
2607 #else
2608 static void
2609 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2610 {
2611 }
2612 #endif
2613 
2614 /*
2615  * COMPOUND call.
2616  */
2617 static __be32
2618 nfsd4_proc_compound(struct svc_rqst *rqstp)
2619 {
2620 	struct nfsd4_compoundargs *args = rqstp->rq_argp;
2621 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2622 	struct nfsd4_op	*op;
2623 	struct nfsd4_compound_state *cstate = &resp->cstate;
2624 	struct svc_fh *current_fh = &cstate->current_fh;
2625 	struct svc_fh *save_fh = &cstate->save_fh;
2626 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2627 	__be32		status;
2628 
2629 	resp->xdr = &rqstp->rq_res_stream;
2630 	resp->statusp = resp->xdr->p;
2631 
2632 	/* reserve space for: NFS status code */
2633 	xdr_reserve_space(resp->xdr, XDR_UNIT);
2634 
2635 	/* reserve space for: taglen, tag, and opcnt */
2636 	xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2637 	resp->taglen = args->taglen;
2638 	resp->tag = args->tag;
2639 	resp->rqstp = rqstp;
2640 	cstate->minorversion = args->minorversion;
2641 	fh_init(current_fh, NFS4_FHSIZE);
2642 	fh_init(save_fh, NFS4_FHSIZE);
2643 	/*
2644 	 * Don't use the deferral mechanism for NFSv4; compounds make it
2645 	 * too hard to avoid non-idempotency problems.
2646 	 */
2647 	clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2648 
2649 	/*
2650 	 * According to RFC3010, this takes precedence over all other errors.
2651 	 */
2652 	status = nfserr_minor_vers_mismatch;
2653 	if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2654 		goto out;
2655 
2656 	status = nfs41_check_op_ordering(args);
2657 	if (status) {
2658 		op = &args->ops[0];
2659 		op->status = status;
2660 		resp->opcnt = 1;
2661 		goto encode_op;
2662 	}
2663 	check_if_stalefh_allowed(args);
2664 
2665 	rqstp->rq_lease_breaker = (void **)&cstate->clp;
2666 
2667 	trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2668 	while (!status && resp->opcnt < args->opcnt) {
2669 		op = &args->ops[resp->opcnt++];
2670 
2671 		if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2672 			/* If there are still more operations to process,
2673 			 * stop here and report NFS4ERR_RESOURCE. */
2674 			if (cstate->minorversion == 0 &&
2675 			    args->client_opcnt > resp->opcnt) {
2676 				op->status = nfserr_resource;
2677 				goto encode_op;
2678 			}
2679 		}
2680 
2681 		/*
2682 		 * The XDR decode routines may have pre-set op->status;
2683 		 * for example, if there is a miscellaneous XDR error
2684 		 * it will be set to nfserr_bad_xdr.
2685 		 */
2686 		if (op->status) {
2687 			if (op->opnum == OP_OPEN)
2688 				op->status = nfsd4_open_omfg(rqstp, cstate, op);
2689 			goto encode_op;
2690 		}
2691 		if (!current_fh->fh_dentry &&
2692 				!HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2693 			if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2694 				op->status = nfserr_nofilehandle;
2695 				goto encode_op;
2696 			}
2697 		} else if (current_fh->fh_export &&
2698 			   current_fh->fh_export->ex_fslocs.migrated &&
2699 			  !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2700 			op->status = nfserr_moved;
2701 			goto encode_op;
2702 		}
2703 
2704 		fh_clear_pre_post_attrs(current_fh);
2705 
2706 		/* If op is non-idempotent */
2707 		if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2708 			/*
2709 			 * Don't execute this op if we couldn't encode a
2710 			 * successful reply:
2711 			 */
2712 			u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2713 			/*
2714 			 * Plus if there's another operation, make sure
2715 			 * we'll have space to at least encode an error:
2716 			 */
2717 			if (resp->opcnt < args->opcnt)
2718 				plen += COMPOUND_ERR_SLACK_SPACE;
2719 			op->status = nfsd4_check_resp_size(resp, plen);
2720 		}
2721 
2722 		if (op->status)
2723 			goto encode_op;
2724 
2725 		if (op->opdesc->op_get_currentstateid)
2726 			op->opdesc->op_get_currentstateid(cstate, &op->u);
2727 		op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2728 
2729 		/* Only from SEQUENCE */
2730 		if (cstate->status == nfserr_replay_cache) {
2731 			dprintk("%s NFS4.1 replay from cache\n", __func__);
2732 			status = op->status;
2733 			goto out;
2734 		}
2735 		if (!op->status) {
2736 			if (op->opdesc->op_set_currentstateid)
2737 				op->opdesc->op_set_currentstateid(cstate, &op->u);
2738 
2739 			if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2740 				clear_current_stateid(cstate);
2741 
2742 			if (current_fh->fh_export &&
2743 					need_wrongsec_check(rqstp))
2744 				op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2745 		}
2746 encode_op:
2747 		if (op->status == nfserr_replay_me) {
2748 			op->replay = &cstate->replay_owner->so_replay;
2749 			nfsd4_encode_replay(resp->xdr, op);
2750 			status = op->status = op->replay->rp_status;
2751 		} else {
2752 			nfsd4_encode_operation(resp, op);
2753 			status = op->status;
2754 		}
2755 
2756 		trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2757 					   status, nfsd4_op_name(op->opnum));
2758 
2759 		nfsd4_cstate_clear_replay(cstate);
2760 		nfsd4_increment_op_stats(nn, op->opnum);
2761 	}
2762 
2763 	fh_put(current_fh);
2764 	fh_put(save_fh);
2765 	BUG_ON(cstate->replay_owner);
2766 out:
2767 	cstate->status = status;
2768 	/* Reset deferral mechanism for RPC deferrals */
2769 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2770 	return rpc_success;
2771 }
2772 
2773 #define op_encode_hdr_size		(2)
2774 #define op_encode_stateid_maxsz		(XDR_QUADLEN(NFS4_STATEID_SIZE))
2775 #define op_encode_verifier_maxsz	(XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2776 #define op_encode_change_info_maxsz	(5)
2777 #define nfs4_fattr_bitmap_maxsz		(4)
2778 
2779 /* We'll fall back on returning no lockowner if run out of space: */
2780 #define op_encode_lockowner_maxsz	(0)
2781 #define op_encode_lock_denied_maxsz	(8 + op_encode_lockowner_maxsz)
2782 
2783 #define nfs4_owner_maxsz		(1 + XDR_QUADLEN(IDMAP_NAMESZ))
2784 
2785 #define op_encode_ace_maxsz		(3 + nfs4_owner_maxsz)
2786 #define op_encode_delegation_maxsz	(1 + op_encode_stateid_maxsz + 1 + \
2787 					 op_encode_ace_maxsz)
2788 
2789 #define op_encode_channel_attrs_maxsz	(6 + 1 + 1)
2790 
2791 /*
2792  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2793  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2794  * the maximum payload size here, based on transport limits and the size
2795  * of the remaining space in the rq_pages array.
2796  */
2797 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2798 {
2799 	u32 buflen;
2800 
2801 	buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2802 	buflen -= rqstp->rq_auth_slack;
2803 	buflen -= rqstp->rq_res.head[0].iov_len;
2804 	return min_t(u32, buflen, svc_max_payload(rqstp));
2805 }
2806 
2807 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2808 				   const struct nfsd4_op *op)
2809 {
2810 	return (op_encode_hdr_size) * sizeof(__be32);
2811 }
2812 
2813 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2814 				      const struct nfsd4_op *op)
2815 {
2816 	return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2817 }
2818 
2819 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2820 			      const struct nfsd4_op *op)
2821 {
2822 	/* ac_supported, ac_resp_access */
2823 	return (op_encode_hdr_size + 2)* sizeof(__be32);
2824 }
2825 
2826 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2827 			      const struct nfsd4_op *op)
2828 {
2829 	return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2830 }
2831 
2832 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2833 			      const struct nfsd4_op *op)
2834 {
2835 	return (op_encode_hdr_size + op_encode_change_info_maxsz
2836 		+ nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2837 }
2838 
2839 /*
2840  * Note since this is an idempotent operation we won't insist on failing
2841  * the op prematurely if the estimate is too large.  We may turn off splice
2842  * reads unnecessarily.
2843  */
2844 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2845 			       const struct nfsd4_op *op)
2846 {
2847 	const u32 *bmap = op->u.getattr.ga_bmval;
2848 	u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2849 	u32 ret = 0;
2850 
2851 	if (bmap0 & FATTR4_WORD0_ACL)
2852 		return nfsd4_max_payload(rqstp);
2853 	if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2854 		return nfsd4_max_payload(rqstp);
2855 
2856 	if (bmap1 & FATTR4_WORD1_OWNER) {
2857 		ret += IDMAP_NAMESZ + 4;
2858 		bmap1 &= ~FATTR4_WORD1_OWNER;
2859 	}
2860 	if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2861 		ret += IDMAP_NAMESZ + 4;
2862 		bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2863 	}
2864 	if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2865 		ret += NFS4_FHSIZE + 4;
2866 		bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2867 	}
2868 	if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2869 		ret += NFS4_MAXLABELLEN + 12;
2870 		bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2871 	}
2872 	/*
2873 	 * Largest of remaining attributes are 16 bytes (e.g.,
2874 	 * supported_attributes)
2875 	 */
2876 	ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2877 	/* bitmask, length */
2878 	ret += 20;
2879 	return ret;
2880 }
2881 
2882 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2883 			     const struct nfsd4_op *op)
2884 {
2885 	return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2886 }
2887 
2888 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2889 			    const struct nfsd4_op *op)
2890 {
2891 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
2892 		* sizeof(__be32);
2893 }
2894 
2895 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2896 			    const struct nfsd4_op *op)
2897 {
2898 	return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2899 		* sizeof(__be32);
2900 }
2901 
2902 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2903 			    const struct nfsd4_op *op)
2904 {
2905 	return (op_encode_hdr_size + op_encode_stateid_maxsz
2906 		+ op_encode_change_info_maxsz + 1
2907 		+ nfs4_fattr_bitmap_maxsz
2908 		+ op_encode_delegation_maxsz) * sizeof(__be32);
2909 }
2910 
2911 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2912 			    const struct nfsd4_op *op)
2913 {
2914 	u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2915 
2916 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2917 }
2918 
2919 static u32 nfsd4_read_plus_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 	 * If we detect that the file changed during hole encoding, then we
2925 	 * recover by encoding the remaining reply as data. This means we need
2926 	 * to set aside enough room to encode two data segments.
2927 	 */
2928 	u32 seg_len = 2 * (1 + 2 + 1);
2929 
2930 	return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2931 }
2932 
2933 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2934 			       const struct nfsd4_op *op)
2935 {
2936 	u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2937 
2938 	return (op_encode_hdr_size + op_encode_verifier_maxsz +
2939 		XDR_QUADLEN(rlen)) * sizeof(__be32);
2940 }
2941 
2942 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2943 				const struct nfsd4_op *op)
2944 {
2945 	return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2946 }
2947 
2948 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2949 			      const struct nfsd4_op *op)
2950 {
2951 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
2952 		* sizeof(__be32);
2953 }
2954 
2955 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
2956 			      const struct nfsd4_op *op)
2957 {
2958 	return (op_encode_hdr_size + op_encode_change_info_maxsz
2959 		+ op_encode_change_info_maxsz) * sizeof(__be32);
2960 }
2961 
2962 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
2963 				const struct nfsd4_op *op)
2964 {
2965 	return (op_encode_hdr_size
2966 		+ XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2967 }
2968 
2969 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
2970 				    const struct nfsd4_op *op)
2971 {
2972 	return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2973 		* sizeof(__be32);
2974 }
2975 
2976 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
2977 			       const struct nfsd4_op *op)
2978 {
2979 	return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2980 }
2981 
2982 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
2983 			       const struct nfsd4_op *op)
2984 {
2985 	return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2986 		(4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2987 }
2988 
2989 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
2990 				   const struct nfsd4_op *op)
2991 {
2992 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2993 								sizeof(__be32);
2994 }
2995 
2996 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
2997 			     const struct nfsd4_op *op)
2998 {
2999 	return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3000 }
3001 
3002 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3003 				   const struct nfsd4_op *op)
3004 {
3005 	return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3006 		1 + 1 + /* eir_flags, spr_how */\
3007 		4 + /* spo_must_enforce & _allow with bitmap */\
3008 		2 + /*eir_server_owner.so_minor_id */\
3009 		/* eir_server_owner.so_major_id<> */\
3010 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3011 		/* eir_server_scope<> */\
3012 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3013 		1 + /* eir_server_impl_id array length */\
3014 		0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3015 }
3016 
3017 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3018 					    const struct nfsd4_op *op)
3019 {
3020 	return (op_encode_hdr_size + \
3021 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3022 		2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3023 }
3024 
3025 static u32 nfsd4_create_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) + /* sessionid */\
3030 		2 + /* csr_sequence, csr_flags */\
3031 		op_encode_channel_attrs_maxsz + \
3032 		op_encode_channel_attrs_maxsz) * sizeof(__be32);
3033 }
3034 
3035 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3036 			    const struct nfsd4_op *op)
3037 {
3038 	return (op_encode_hdr_size +
3039 		1 /* wr_callback */ +
3040 		op_encode_stateid_maxsz /* wr_callback */ +
3041 		2 /* wr_count */ +
3042 		1 /* wr_committed */ +
3043 		op_encode_verifier_maxsz +
3044 		1 /* cr_consecutive */ +
3045 		1 /* cr_synchronous */) * sizeof(__be32);
3046 }
3047 
3048 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3049 				      const struct nfsd4_op *op)
3050 {
3051 	return (op_encode_hdr_size +
3052 		2 /* osr_count */ +
3053 		1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3054 }
3055 
3056 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3057 				   const struct nfsd4_op *op)
3058 {
3059 	return (op_encode_hdr_size +
3060 		3 /* cnr_lease_time */ +
3061 		1 /* We support one cnr_source_server */ +
3062 		1 /* cnr_stateid seq */ +
3063 		op_encode_stateid_maxsz /* cnr_stateid */ +
3064 		1 /* num cnr_source_server*/ +
3065 		1 /* nl4_type */ +
3066 		1 /* nl4 size */ +
3067 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3068 		* sizeof(__be32);
3069 }
3070 
3071 #ifdef CONFIG_NFSD_PNFS
3072 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3073 				     const struct nfsd4_op *op)
3074 {
3075 	u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3076 
3077 	return (op_encode_hdr_size +
3078 		1 /* gd_layout_type*/ +
3079 		XDR_QUADLEN(rlen) +
3080 		2 /* gd_notify_types */) * sizeof(__be32);
3081 }
3082 
3083 /*
3084  * At this stage we don't really know what layout driver will handle the request,
3085  * so we need to define an arbitrary upper bound here.
3086  */
3087 #define MAX_LAYOUT_SIZE		128
3088 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3089 				 const struct nfsd4_op *op)
3090 {
3091 	return (op_encode_hdr_size +
3092 		1 /* logr_return_on_close */ +
3093 		op_encode_stateid_maxsz +
3094 		1 /* nr of layouts */ +
3095 		MAX_LAYOUT_SIZE) * sizeof(__be32);
3096 }
3097 
3098 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3099 				    const struct nfsd4_op *op)
3100 {
3101 	return (op_encode_hdr_size +
3102 		1 /* locr_newsize */ +
3103 		2 /* ns_size */) * sizeof(__be32);
3104 }
3105 
3106 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3107 				    const struct nfsd4_op *op)
3108 {
3109 	return (op_encode_hdr_size +
3110 		1 /* lrs_stateid */ +
3111 		op_encode_stateid_maxsz) * sizeof(__be32);
3112 }
3113 #endif /* CONFIG_NFSD_PNFS */
3114 
3115 
3116 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3117 			    const struct nfsd4_op *op)
3118 {
3119 	return (op_encode_hdr_size + 3) * sizeof(__be32);
3120 }
3121 
3122 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3123 				const struct nfsd4_op *op)
3124 {
3125 	u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3126 
3127 	return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3128 }
3129 
3130 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3131 				const struct nfsd4_op *op)
3132 {
3133 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3134 		* sizeof(__be32);
3135 }
3136 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3137 				  const struct nfsd4_op *op)
3138 {
3139 	u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3140 
3141 	return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3142 }
3143 
3144 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3145 				   const struct nfsd4_op *op)
3146 {
3147 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3148 		* sizeof(__be32);
3149 }
3150 
3151 
3152 static const struct nfsd4_operation nfsd4_ops[] = {
3153 	[OP_ACCESS] = {
3154 		.op_func = nfsd4_access,
3155 		.op_name = "OP_ACCESS",
3156 		.op_rsize_bop = nfsd4_access_rsize,
3157 	},
3158 	[OP_CLOSE] = {
3159 		.op_func = nfsd4_close,
3160 		.op_flags = OP_MODIFIES_SOMETHING,
3161 		.op_name = "OP_CLOSE",
3162 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3163 		.op_get_currentstateid = nfsd4_get_closestateid,
3164 		.op_set_currentstateid = nfsd4_set_closestateid,
3165 	},
3166 	[OP_COMMIT] = {
3167 		.op_func = nfsd4_commit,
3168 		.op_flags = OP_MODIFIES_SOMETHING,
3169 		.op_name = "OP_COMMIT",
3170 		.op_rsize_bop = nfsd4_commit_rsize,
3171 	},
3172 	[OP_CREATE] = {
3173 		.op_func = nfsd4_create,
3174 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3175 		.op_name = "OP_CREATE",
3176 		.op_rsize_bop = nfsd4_create_rsize,
3177 	},
3178 	[OP_DELEGRETURN] = {
3179 		.op_func = nfsd4_delegreturn,
3180 		.op_flags = OP_MODIFIES_SOMETHING,
3181 		.op_name = "OP_DELEGRETURN",
3182 		.op_rsize_bop = nfsd4_only_status_rsize,
3183 		.op_get_currentstateid = nfsd4_get_delegreturnstateid,
3184 	},
3185 	[OP_GETATTR] = {
3186 		.op_func = nfsd4_getattr,
3187 		.op_flags = ALLOWED_ON_ABSENT_FS,
3188 		.op_rsize_bop = nfsd4_getattr_rsize,
3189 		.op_name = "OP_GETATTR",
3190 	},
3191 	[OP_GETFH] = {
3192 		.op_func = nfsd4_getfh,
3193 		.op_name = "OP_GETFH",
3194 		.op_rsize_bop = nfsd4_getfh_rsize,
3195 	},
3196 	[OP_LINK] = {
3197 		.op_func = nfsd4_link,
3198 		.op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3199 				| OP_CACHEME,
3200 		.op_name = "OP_LINK",
3201 		.op_rsize_bop = nfsd4_link_rsize,
3202 	},
3203 	[OP_LOCK] = {
3204 		.op_func = nfsd4_lock,
3205 		.op_flags = OP_MODIFIES_SOMETHING |
3206 				OP_NONTRIVIAL_ERROR_ENCODE,
3207 		.op_name = "OP_LOCK",
3208 		.op_rsize_bop = nfsd4_lock_rsize,
3209 		.op_set_currentstateid = nfsd4_set_lockstateid,
3210 	},
3211 	[OP_LOCKT] = {
3212 		.op_func = nfsd4_lockt,
3213 		.op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3214 		.op_name = "OP_LOCKT",
3215 		.op_rsize_bop = nfsd4_lock_rsize,
3216 	},
3217 	[OP_LOCKU] = {
3218 		.op_func = nfsd4_locku,
3219 		.op_flags = OP_MODIFIES_SOMETHING,
3220 		.op_name = "OP_LOCKU",
3221 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3222 		.op_get_currentstateid = nfsd4_get_lockustateid,
3223 	},
3224 	[OP_LOOKUP] = {
3225 		.op_func = nfsd4_lookup,
3226 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3227 		.op_name = "OP_LOOKUP",
3228 		.op_rsize_bop = nfsd4_only_status_rsize,
3229 	},
3230 	[OP_LOOKUPP] = {
3231 		.op_func = nfsd4_lookupp,
3232 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3233 		.op_name = "OP_LOOKUPP",
3234 		.op_rsize_bop = nfsd4_only_status_rsize,
3235 	},
3236 	[OP_NVERIFY] = {
3237 		.op_func = nfsd4_nverify,
3238 		.op_name = "OP_NVERIFY",
3239 		.op_rsize_bop = nfsd4_only_status_rsize,
3240 	},
3241 	[OP_OPEN] = {
3242 		.op_func = nfsd4_open,
3243 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3244 		.op_name = "OP_OPEN",
3245 		.op_rsize_bop = nfsd4_open_rsize,
3246 		.op_set_currentstateid = nfsd4_set_openstateid,
3247 	},
3248 	[OP_OPEN_CONFIRM] = {
3249 		.op_func = nfsd4_open_confirm,
3250 		.op_flags = OP_MODIFIES_SOMETHING,
3251 		.op_name = "OP_OPEN_CONFIRM",
3252 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3253 	},
3254 	[OP_OPEN_DOWNGRADE] = {
3255 		.op_func = nfsd4_open_downgrade,
3256 		.op_flags = OP_MODIFIES_SOMETHING,
3257 		.op_name = "OP_OPEN_DOWNGRADE",
3258 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3259 		.op_get_currentstateid = nfsd4_get_opendowngradestateid,
3260 		.op_set_currentstateid = nfsd4_set_opendowngradestateid,
3261 	},
3262 	[OP_PUTFH] = {
3263 		.op_func = nfsd4_putfh,
3264 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3265 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3266 		.op_name = "OP_PUTFH",
3267 		.op_rsize_bop = nfsd4_only_status_rsize,
3268 	},
3269 	[OP_PUTPUBFH] = {
3270 		.op_func = nfsd4_putrootfh,
3271 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3272 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3273 		.op_name = "OP_PUTPUBFH",
3274 		.op_rsize_bop = nfsd4_only_status_rsize,
3275 	},
3276 	[OP_PUTROOTFH] = {
3277 		.op_func = nfsd4_putrootfh,
3278 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3279 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3280 		.op_name = "OP_PUTROOTFH",
3281 		.op_rsize_bop = nfsd4_only_status_rsize,
3282 	},
3283 	[OP_READ] = {
3284 		.op_func = nfsd4_read,
3285 		.op_release = nfsd4_read_release,
3286 		.op_name = "OP_READ",
3287 		.op_rsize_bop = nfsd4_read_rsize,
3288 		.op_get_currentstateid = nfsd4_get_readstateid,
3289 	},
3290 	[OP_READDIR] = {
3291 		.op_func = nfsd4_readdir,
3292 		.op_name = "OP_READDIR",
3293 		.op_rsize_bop = nfsd4_readdir_rsize,
3294 	},
3295 	[OP_READLINK] = {
3296 		.op_func = nfsd4_readlink,
3297 		.op_name = "OP_READLINK",
3298 		.op_rsize_bop = nfsd4_readlink_rsize,
3299 	},
3300 	[OP_REMOVE] = {
3301 		.op_func = nfsd4_remove,
3302 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3303 		.op_name = "OP_REMOVE",
3304 		.op_rsize_bop = nfsd4_remove_rsize,
3305 	},
3306 	[OP_RENAME] = {
3307 		.op_func = nfsd4_rename,
3308 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3309 		.op_name = "OP_RENAME",
3310 		.op_rsize_bop = nfsd4_rename_rsize,
3311 	},
3312 	[OP_RENEW] = {
3313 		.op_func = nfsd4_renew,
3314 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3315 				| OP_MODIFIES_SOMETHING,
3316 		.op_name = "OP_RENEW",
3317 		.op_rsize_bop = nfsd4_only_status_rsize,
3318 
3319 	},
3320 	[OP_RESTOREFH] = {
3321 		.op_func = nfsd4_restorefh,
3322 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3323 				| OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3324 		.op_name = "OP_RESTOREFH",
3325 		.op_rsize_bop = nfsd4_only_status_rsize,
3326 	},
3327 	[OP_SAVEFH] = {
3328 		.op_func = nfsd4_savefh,
3329 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3330 		.op_name = "OP_SAVEFH",
3331 		.op_rsize_bop = nfsd4_only_status_rsize,
3332 	},
3333 	[OP_SECINFO] = {
3334 		.op_func = nfsd4_secinfo,
3335 		.op_release = nfsd4_secinfo_release,
3336 		.op_flags = OP_HANDLES_WRONGSEC,
3337 		.op_name = "OP_SECINFO",
3338 		.op_rsize_bop = nfsd4_secinfo_rsize,
3339 	},
3340 	[OP_SETATTR] = {
3341 		.op_func = nfsd4_setattr,
3342 		.op_name = "OP_SETATTR",
3343 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3344 				| OP_NONTRIVIAL_ERROR_ENCODE,
3345 		.op_rsize_bop = nfsd4_setattr_rsize,
3346 		.op_get_currentstateid = nfsd4_get_setattrstateid,
3347 	},
3348 	[OP_SETCLIENTID] = {
3349 		.op_func = nfsd4_setclientid,
3350 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3351 				| OP_MODIFIES_SOMETHING | OP_CACHEME
3352 				| OP_NONTRIVIAL_ERROR_ENCODE,
3353 		.op_name = "OP_SETCLIENTID",
3354 		.op_rsize_bop = nfsd4_setclientid_rsize,
3355 	},
3356 	[OP_SETCLIENTID_CONFIRM] = {
3357 		.op_func = nfsd4_setclientid_confirm,
3358 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3359 				| OP_MODIFIES_SOMETHING | OP_CACHEME,
3360 		.op_name = "OP_SETCLIENTID_CONFIRM",
3361 		.op_rsize_bop = nfsd4_only_status_rsize,
3362 	},
3363 	[OP_VERIFY] = {
3364 		.op_func = nfsd4_verify,
3365 		.op_name = "OP_VERIFY",
3366 		.op_rsize_bop = nfsd4_only_status_rsize,
3367 	},
3368 	[OP_WRITE] = {
3369 		.op_func = nfsd4_write,
3370 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3371 		.op_name = "OP_WRITE",
3372 		.op_rsize_bop = nfsd4_write_rsize,
3373 		.op_get_currentstateid = nfsd4_get_writestateid,
3374 	},
3375 	[OP_RELEASE_LOCKOWNER] = {
3376 		.op_func = nfsd4_release_lockowner,
3377 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3378 				| OP_MODIFIES_SOMETHING,
3379 		.op_name = "OP_RELEASE_LOCKOWNER",
3380 		.op_rsize_bop = nfsd4_only_status_rsize,
3381 	},
3382 
3383 	/* NFSv4.1 operations */
3384 	[OP_EXCHANGE_ID] = {
3385 		.op_func = nfsd4_exchange_id,
3386 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3387 				| OP_MODIFIES_SOMETHING,
3388 		.op_name = "OP_EXCHANGE_ID",
3389 		.op_rsize_bop = nfsd4_exchange_id_rsize,
3390 	},
3391 	[OP_BACKCHANNEL_CTL] = {
3392 		.op_func = nfsd4_backchannel_ctl,
3393 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3394 		.op_name = "OP_BACKCHANNEL_CTL",
3395 		.op_rsize_bop = nfsd4_only_status_rsize,
3396 	},
3397 	[OP_BIND_CONN_TO_SESSION] = {
3398 		.op_func = nfsd4_bind_conn_to_session,
3399 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3400 				| OP_MODIFIES_SOMETHING,
3401 		.op_name = "OP_BIND_CONN_TO_SESSION",
3402 		.op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3403 	},
3404 	[OP_CREATE_SESSION] = {
3405 		.op_func = nfsd4_create_session,
3406 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3407 				| OP_MODIFIES_SOMETHING,
3408 		.op_name = "OP_CREATE_SESSION",
3409 		.op_rsize_bop = nfsd4_create_session_rsize,
3410 	},
3411 	[OP_DESTROY_SESSION] = {
3412 		.op_func = nfsd4_destroy_session,
3413 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3414 				| OP_MODIFIES_SOMETHING,
3415 		.op_name = "OP_DESTROY_SESSION",
3416 		.op_rsize_bop = nfsd4_only_status_rsize,
3417 	},
3418 	[OP_SEQUENCE] = {
3419 		.op_func = nfsd4_sequence,
3420 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3421 		.op_name = "OP_SEQUENCE",
3422 		.op_rsize_bop = nfsd4_sequence_rsize,
3423 	},
3424 	[OP_DESTROY_CLIENTID] = {
3425 		.op_func = nfsd4_destroy_clientid,
3426 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3427 				| OP_MODIFIES_SOMETHING,
3428 		.op_name = "OP_DESTROY_CLIENTID",
3429 		.op_rsize_bop = nfsd4_only_status_rsize,
3430 	},
3431 	[OP_RECLAIM_COMPLETE] = {
3432 		.op_func = nfsd4_reclaim_complete,
3433 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3434 		.op_name = "OP_RECLAIM_COMPLETE",
3435 		.op_rsize_bop = nfsd4_only_status_rsize,
3436 	},
3437 	[OP_SECINFO_NO_NAME] = {
3438 		.op_func = nfsd4_secinfo_no_name,
3439 		.op_release = nfsd4_secinfo_no_name_release,
3440 		.op_flags = OP_HANDLES_WRONGSEC,
3441 		.op_name = "OP_SECINFO_NO_NAME",
3442 		.op_rsize_bop = nfsd4_secinfo_rsize,
3443 	},
3444 	[OP_TEST_STATEID] = {
3445 		.op_func = nfsd4_test_stateid,
3446 		.op_flags = ALLOWED_WITHOUT_FH,
3447 		.op_name = "OP_TEST_STATEID",
3448 		.op_rsize_bop = nfsd4_test_stateid_rsize,
3449 	},
3450 	[OP_FREE_STATEID] = {
3451 		.op_func = nfsd4_free_stateid,
3452 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3453 		.op_name = "OP_FREE_STATEID",
3454 		.op_get_currentstateid = nfsd4_get_freestateid,
3455 		.op_rsize_bop = nfsd4_only_status_rsize,
3456 	},
3457 #ifdef CONFIG_NFSD_PNFS
3458 	[OP_GETDEVICEINFO] = {
3459 		.op_func = nfsd4_getdeviceinfo,
3460 		.op_release = nfsd4_getdeviceinfo_release,
3461 		.op_flags = ALLOWED_WITHOUT_FH,
3462 		.op_name = "OP_GETDEVICEINFO",
3463 		.op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3464 	},
3465 	[OP_LAYOUTGET] = {
3466 		.op_func = nfsd4_layoutget,
3467 		.op_release = nfsd4_layoutget_release,
3468 		.op_flags = OP_MODIFIES_SOMETHING,
3469 		.op_name = "OP_LAYOUTGET",
3470 		.op_rsize_bop = nfsd4_layoutget_rsize,
3471 	},
3472 	[OP_LAYOUTCOMMIT] = {
3473 		.op_func = nfsd4_layoutcommit,
3474 		.op_flags = OP_MODIFIES_SOMETHING,
3475 		.op_name = "OP_LAYOUTCOMMIT",
3476 		.op_rsize_bop = nfsd4_layoutcommit_rsize,
3477 	},
3478 	[OP_LAYOUTRETURN] = {
3479 		.op_func = nfsd4_layoutreturn,
3480 		.op_flags = OP_MODIFIES_SOMETHING,
3481 		.op_name = "OP_LAYOUTRETURN",
3482 		.op_rsize_bop = nfsd4_layoutreturn_rsize,
3483 	},
3484 #endif /* CONFIG_NFSD_PNFS */
3485 
3486 	/* NFSv4.2 operations */
3487 	[OP_ALLOCATE] = {
3488 		.op_func = nfsd4_allocate,
3489 		.op_flags = OP_MODIFIES_SOMETHING,
3490 		.op_name = "OP_ALLOCATE",
3491 		.op_rsize_bop = nfsd4_only_status_rsize,
3492 	},
3493 	[OP_DEALLOCATE] = {
3494 		.op_func = nfsd4_deallocate,
3495 		.op_flags = OP_MODIFIES_SOMETHING,
3496 		.op_name = "OP_DEALLOCATE",
3497 		.op_rsize_bop = nfsd4_only_status_rsize,
3498 	},
3499 	[OP_CLONE] = {
3500 		.op_func = nfsd4_clone,
3501 		.op_flags = OP_MODIFIES_SOMETHING,
3502 		.op_name = "OP_CLONE",
3503 		.op_rsize_bop = nfsd4_only_status_rsize,
3504 	},
3505 	[OP_COPY] = {
3506 		.op_func = nfsd4_copy,
3507 		.op_flags = OP_MODIFIES_SOMETHING,
3508 		.op_name = "OP_COPY",
3509 		.op_rsize_bop = nfsd4_copy_rsize,
3510 	},
3511 	[OP_READ_PLUS] = {
3512 		.op_func = nfsd4_read,
3513 		.op_release = nfsd4_read_release,
3514 		.op_name = "OP_READ_PLUS",
3515 		.op_rsize_bop = nfsd4_read_plus_rsize,
3516 		.op_get_currentstateid = nfsd4_get_readstateid,
3517 	},
3518 	[OP_SEEK] = {
3519 		.op_func = nfsd4_seek,
3520 		.op_name = "OP_SEEK",
3521 		.op_rsize_bop = nfsd4_seek_rsize,
3522 	},
3523 	[OP_OFFLOAD_STATUS] = {
3524 		.op_func = nfsd4_offload_status,
3525 		.op_name = "OP_OFFLOAD_STATUS",
3526 		.op_rsize_bop = nfsd4_offload_status_rsize,
3527 	},
3528 	[OP_OFFLOAD_CANCEL] = {
3529 		.op_func = nfsd4_offload_cancel,
3530 		.op_flags = OP_MODIFIES_SOMETHING,
3531 		.op_name = "OP_OFFLOAD_CANCEL",
3532 		.op_rsize_bop = nfsd4_only_status_rsize,
3533 	},
3534 	[OP_COPY_NOTIFY] = {
3535 		.op_func = nfsd4_copy_notify,
3536 		.op_flags = OP_MODIFIES_SOMETHING,
3537 		.op_name = "OP_COPY_NOTIFY",
3538 		.op_rsize_bop = nfsd4_copy_notify_rsize,
3539 	},
3540 	[OP_GETXATTR] = {
3541 		.op_func = nfsd4_getxattr,
3542 		.op_name = "OP_GETXATTR",
3543 		.op_rsize_bop = nfsd4_getxattr_rsize,
3544 	},
3545 	[OP_SETXATTR] = {
3546 		.op_func = nfsd4_setxattr,
3547 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3548 		.op_name = "OP_SETXATTR",
3549 		.op_rsize_bop = nfsd4_setxattr_rsize,
3550 	},
3551 	[OP_LISTXATTRS] = {
3552 		.op_func = nfsd4_listxattrs,
3553 		.op_name = "OP_LISTXATTRS",
3554 		.op_rsize_bop = nfsd4_listxattrs_rsize,
3555 	},
3556 	[OP_REMOVEXATTR] = {
3557 		.op_func = nfsd4_removexattr,
3558 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3559 		.op_name = "OP_REMOVEXATTR",
3560 		.op_rsize_bop = nfsd4_removexattr_rsize,
3561 	},
3562 };
3563 
3564 /**
3565  * nfsd4_spo_must_allow - Determine if the compound op contains an
3566  * operation that is allowed to be sent with machine credentials
3567  *
3568  * @rqstp: a pointer to the struct svc_rqst
3569  *
3570  * Checks to see if the compound contains a spo_must_allow op
3571  * and confirms that it was sent with the proper machine creds.
3572  */
3573 
3574 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3575 {
3576 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
3577 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3578 	struct nfsd4_op *this;
3579 	struct nfsd4_compound_state *cstate = &resp->cstate;
3580 	struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3581 	u32 opiter;
3582 
3583 	if (!cstate->minorversion)
3584 		return false;
3585 
3586 	if (cstate->spo_must_allowed)
3587 		return true;
3588 
3589 	opiter = resp->opcnt;
3590 	while (opiter < argp->opcnt) {
3591 		this = &argp->ops[opiter++];
3592 		if (test_bit(this->opnum, allow->u.longs) &&
3593 			cstate->clp->cl_mach_cred &&
3594 			nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3595 			cstate->spo_must_allowed = true;
3596 			return true;
3597 		}
3598 	}
3599 	cstate->spo_must_allowed = false;
3600 	return false;
3601 }
3602 
3603 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3604 {
3605 	if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3606 		return op_encode_hdr_size * sizeof(__be32);
3607 
3608 	BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3609 	return OPDESC(op)->op_rsize_bop(rqstp, op);
3610 }
3611 
3612 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3613 {
3614 	if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3615 		pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3616 			op->opnum, nfsd4_op_name(op->opnum));
3617 		WARN_ON_ONCE(1);
3618 	}
3619 }
3620 
3621 static const char *nfsd4_op_name(unsigned opnum)
3622 {
3623 	if (opnum < ARRAY_SIZE(nfsd4_ops))
3624 		return nfsd4_ops[opnum].op_name;
3625 	return "unknown_operation";
3626 }
3627 
3628 static const struct svc_procedure nfsd_procedures4[2] = {
3629 	[NFSPROC4_NULL] = {
3630 		.pc_func = nfsd4_proc_null,
3631 		.pc_decode = nfssvc_decode_voidarg,
3632 		.pc_encode = nfssvc_encode_voidres,
3633 		.pc_argsize = sizeof(struct nfsd_voidargs),
3634 		.pc_argzero = sizeof(struct nfsd_voidargs),
3635 		.pc_ressize = sizeof(struct nfsd_voidres),
3636 		.pc_cachetype = RC_NOCACHE,
3637 		.pc_xdrressize = 1,
3638 		.pc_name = "NULL",
3639 	},
3640 	[NFSPROC4_COMPOUND] = {
3641 		.pc_func = nfsd4_proc_compound,
3642 		.pc_decode = nfs4svc_decode_compoundargs,
3643 		.pc_encode = nfs4svc_encode_compoundres,
3644 		.pc_argsize = sizeof(struct nfsd4_compoundargs),
3645 		.pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3646 		.pc_ressize = sizeof(struct nfsd4_compoundres),
3647 		.pc_release = nfsd4_release_compoundargs,
3648 		.pc_cachetype = RC_NOCACHE,
3649 		.pc_xdrressize = NFSD_BUFSIZE/4,
3650 		.pc_name = "COMPOUND",
3651 	},
3652 };
3653 
3654 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3655 			      nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3656 const struct svc_version nfsd_version4 = {
3657 	.vs_vers		= 4,
3658 	.vs_nproc		= ARRAY_SIZE(nfsd_procedures4),
3659 	.vs_proc		= nfsd_procedures4,
3660 	.vs_count		= nfsd_count4,
3661 	.vs_dispatch		= nfsd_dispatch,
3662 	.vs_xdrsize		= NFS4_SVC_XDRSIZE,
3663 	.vs_rpcb_optnl		= true,
3664 	.vs_need_cong_ctrl	= true,
3665 };
3666