xref: /linux/fs/nfsd/nfs4proc.c (revision 001821b0e79716c4e17c71d8e053a23599a7a508)
1 /*
2  *  Server-side procedures for NFSv4.
3  *
4  *  Copyright (c) 2002 The Regents of the University of Michigan.
5  *  All rights reserved.
6  *
7  *  Kendrick Smith <kmsmith@umich.edu>
8  *  Andy Adamson   <andros@umich.edu>
9  *
10  *  Redistribution and use in source and binary forms, with or without
11  *  modification, are permitted provided that the following conditions
12  *  are met:
13  *
14  *  1. Redistributions of source code must retain the above copyright
15  *     notice, this list of conditions and the following disclaimer.
16  *  2. Redistributions in binary form must reproduce the above copyright
17  *     notice, this list of conditions and the following disclaimer in the
18  *     documentation and/or other materials provided with the distribution.
19  *  3. Neither the name of the University nor the names of its
20  *     contributors may be used to endorse or promote products derived
21  *     from this software without specific prior written permission.
22  *
23  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34  */
35 #include <linux/fs_struct.h>
36 #include <linux/file.h>
37 #include <linux/falloc.h>
38 #include <linux/slab.h>
39 #include <linux/kthread.h>
40 #include <linux/namei.h>
41 
42 #include <linux/sunrpc/addr.h>
43 #include <linux/nfs_ssc.h>
44 
45 #include "idmap.h"
46 #include "cache.h"
47 #include "xdr4.h"
48 #include "vfs.h"
49 #include "current_stateid.h"
50 #include "netns.h"
51 #include "acl.h"
52 #include "pnfs.h"
53 #include "trace.h"
54 
55 static bool inter_copy_offload_enable;
56 module_param(inter_copy_offload_enable, bool, 0644);
57 MODULE_PARM_DESC(inter_copy_offload_enable,
58 		 "Enable inter server to server copy offload. Default: false");
59 
60 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
61 static int nfsd4_ssc_umount_timeout = 900000;		/* default to 15 mins */
62 module_param(nfsd4_ssc_umount_timeout, int, 0644);
63 MODULE_PARM_DESC(nfsd4_ssc_umount_timeout,
64 		"idle msecs before unmount export from source server");
65 #endif
66 
67 #define NFSDDBG_FACILITY		NFSDDBG_PROC
68 
69 static u32 nfsd_attrmask[] = {
70 	NFSD_WRITEABLE_ATTRS_WORD0,
71 	NFSD_WRITEABLE_ATTRS_WORD1,
72 	NFSD_WRITEABLE_ATTRS_WORD2
73 };
74 
75 static u32 nfsd41_ex_attrmask[] = {
76 	NFSD_SUPPATTR_EXCLCREAT_WORD0,
77 	NFSD_SUPPATTR_EXCLCREAT_WORD1,
78 	NFSD_SUPPATTR_EXCLCREAT_WORD2
79 };
80 
81 static __be32
82 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
83 		   u32 *bmval, u32 *writable)
84 {
85 	struct dentry *dentry = cstate->current_fh.fh_dentry;
86 	struct svc_export *exp = cstate->current_fh.fh_export;
87 
88 	if (!nfsd_attrs_supported(cstate->minorversion, bmval))
89 		return nfserr_attrnotsupp;
90 	if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
91 		return nfserr_attrnotsupp;
92 	if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
93 			!(exp->ex_flags & NFSEXP_SECURITY_LABEL))
94 		return nfserr_attrnotsupp;
95 	if (writable && !bmval_is_subset(bmval, writable))
96 		return nfserr_inval;
97 	if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
98 			(bmval[1] & FATTR4_WORD1_MODE))
99 		return nfserr_inval;
100 	return nfs_ok;
101 }
102 
103 static __be32
104 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
105 	struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
106 {
107 	__be32 status = nfs_ok;
108 
109 	if (open->op_create == NFS4_OPEN_CREATE) {
110 		if (open->op_createmode == NFS4_CREATE_UNCHECKED
111 		    || open->op_createmode == NFS4_CREATE_GUARDED)
112 			status = check_attr_support(rqstp, cstate,
113 					open->op_bmval, nfsd_attrmask);
114 		else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
115 			status = check_attr_support(rqstp, cstate,
116 					open->op_bmval, nfsd41_ex_attrmask);
117 	}
118 
119 	return status;
120 }
121 
122 static int
123 is_create_with_attrs(struct nfsd4_open *open)
124 {
125 	return open->op_create == NFS4_OPEN_CREATE
126 		&& (open->op_createmode == NFS4_CREATE_UNCHECKED
127 		    || open->op_createmode == NFS4_CREATE_GUARDED
128 		    || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
129 }
130 
131 static inline void
132 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
133 {
134 	fh_put(dst);
135 	dget(src->fh_dentry);
136 	if (src->fh_export)
137 		exp_get(src->fh_export);
138 	*dst = *src;
139 }
140 
141 static __be32
142 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
143 {
144 
145 	if (open->op_truncate &&
146 		!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
147 		return nfserr_inval;
148 
149 	accmode |= NFSD_MAY_READ_IF_EXEC;
150 
151 	if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
152 		accmode |= NFSD_MAY_READ;
153 	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
154 		accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
155 	if (open->op_share_deny & NFS4_SHARE_DENY_READ)
156 		accmode |= NFSD_MAY_WRITE;
157 
158 	return fh_verify(rqstp, current_fh, S_IFREG, accmode);
159 }
160 
161 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
162 {
163 	umode_t mode = d_inode(fh->fh_dentry)->i_mode;
164 
165 	if (S_ISREG(mode))
166 		return nfs_ok;
167 	if (S_ISDIR(mode))
168 		return nfserr_isdir;
169 	/*
170 	 * Using err_symlink as our catch-all case may look odd; but
171 	 * there's no other obvious error for this case in 4.0, and we
172 	 * happen to know that it will cause the linux v4 client to do
173 	 * the right thing on attempts to open something other than a
174 	 * regular file.
175 	 */
176 	return nfserr_symlink;
177 }
178 
179 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
180 {
181 	if (nfsd4_has_session(cstate))
182 		return;
183 	fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
184 			&resfh->fh_handle);
185 }
186 
187 static inline bool nfsd4_create_is_exclusive(int createmode)
188 {
189 	return createmode == NFS4_CREATE_EXCLUSIVE ||
190 		createmode == NFS4_CREATE_EXCLUSIVE4_1;
191 }
192 
193 static __be32
194 nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
195 		 struct nfsd4_open *open)
196 {
197 	struct file *filp;
198 	struct path path;
199 	int oflags;
200 
201 	oflags = O_CREAT | O_LARGEFILE;
202 	switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
203 	case NFS4_SHARE_ACCESS_WRITE:
204 		oflags |= O_WRONLY;
205 		break;
206 	case NFS4_SHARE_ACCESS_BOTH:
207 		oflags |= O_RDWR;
208 		break;
209 	default:
210 		oflags |= O_RDONLY;
211 	}
212 
213 	path.mnt = fhp->fh_export->ex_path.mnt;
214 	path.dentry = child;
215 	filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
216 			     current_cred());
217 	if (IS_ERR(filp))
218 		return nfserrno(PTR_ERR(filp));
219 
220 	open->op_filp = filp;
221 	return nfs_ok;
222 }
223 
224 /*
225  * Implement NFSv4's unchecked, guarded, and exclusive create
226  * semantics for regular files. Open state for this new file is
227  * subsequently fabricated in nfsd4_process_open2().
228  *
229  * Upon return, caller must release @fhp and @resfhp.
230  */
231 static __be32
232 nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
233 		  struct svc_fh *resfhp, struct nfsd4_open *open)
234 {
235 	struct iattr *iap = &open->op_iattr;
236 	struct nfsd_attrs attrs = {
237 		.na_iattr	= iap,
238 		.na_seclabel	= &open->op_label,
239 	};
240 	struct dentry *parent, *child;
241 	__u32 v_mtime, v_atime;
242 	struct inode *inode;
243 	__be32 status;
244 	int host_err;
245 
246 	if (isdotent(open->op_fname, open->op_fnamelen))
247 		return nfserr_exist;
248 	if (!(iap->ia_valid & ATTR_MODE))
249 		iap->ia_mode = 0;
250 
251 	status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
252 	if (status != nfs_ok)
253 		return status;
254 	parent = fhp->fh_dentry;
255 	inode = d_inode(parent);
256 
257 	host_err = fh_want_write(fhp);
258 	if (host_err)
259 		return nfserrno(host_err);
260 
261 	if (is_create_with_attrs(open))
262 		nfsd4_acl_to_attr(NF4REG, open->op_acl, &attrs);
263 
264 	inode_lock_nested(inode, I_MUTEX_PARENT);
265 
266 	child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
267 	if (IS_ERR(child)) {
268 		status = nfserrno(PTR_ERR(child));
269 		goto out;
270 	}
271 
272 	if (d_really_is_negative(child)) {
273 		status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
274 		if (status != nfs_ok)
275 			goto out;
276 	}
277 
278 	status = fh_compose(resfhp, fhp->fh_export, child, fhp);
279 	if (status != nfs_ok)
280 		goto out;
281 
282 	v_mtime = 0;
283 	v_atime = 0;
284 	if (nfsd4_create_is_exclusive(open->op_createmode)) {
285 		u32 *verifier = (u32 *)open->op_verf.data;
286 
287 		/*
288 		 * Solaris 7 gets confused (bugid 4218508) if these have
289 		 * the high bit set, as do xfs filesystems without the
290 		 * "bigtime" feature. So just clear the high bits. If this
291 		 * is ever changed to use different attrs for storing the
292 		 * verifier, then do_open_lookup() will also need to be
293 		 * fixed accordingly.
294 		 */
295 		v_mtime = verifier[0] & 0x7fffffff;
296 		v_atime = verifier[1] & 0x7fffffff;
297 	}
298 
299 	if (d_really_is_positive(child)) {
300 		/* NFSv4 protocol requires change attributes even though
301 		 * no change happened.
302 		 */
303 		status = fh_fill_both_attrs(fhp);
304 		if (status != nfs_ok)
305 			goto out;
306 
307 		switch (open->op_createmode) {
308 		case NFS4_CREATE_UNCHECKED:
309 			if (!d_is_reg(child))
310 				break;
311 
312 			/*
313 			 * In NFSv4, we don't want to truncate the file
314 			 * now. This would be wrong if the OPEN fails for
315 			 * some other reason. Furthermore, if the size is
316 			 * nonzero, we should ignore it according to spec!
317 			 */
318 			open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
319 						!iap->ia_size;
320 			break;
321 		case NFS4_CREATE_GUARDED:
322 			status = nfserr_exist;
323 			break;
324 		case NFS4_CREATE_EXCLUSIVE:
325 			if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
326 			    inode_get_atime_sec(d_inode(child)) == v_atime &&
327 			    d_inode(child)->i_size == 0) {
328 				open->op_created = true;
329 				break;		/* subtle */
330 			}
331 			status = nfserr_exist;
332 			break;
333 		case NFS4_CREATE_EXCLUSIVE4_1:
334 			if (inode_get_mtime_sec(d_inode(child)) == v_mtime &&
335 			    inode_get_atime_sec(d_inode(child)) == v_atime &&
336 			    d_inode(child)->i_size == 0) {
337 				open->op_created = true;
338 				goto set_attr;	/* subtle */
339 			}
340 			status = nfserr_exist;
341 		}
342 		goto out;
343 	}
344 
345 	if (!IS_POSIXACL(inode))
346 		iap->ia_mode &= ~current_umask();
347 
348 	status = fh_fill_pre_attrs(fhp);
349 	if (status != nfs_ok)
350 		goto out;
351 	status = nfsd4_vfs_create(fhp, child, open);
352 	if (status != nfs_ok)
353 		goto out;
354 	open->op_created = true;
355 	fh_fill_post_attrs(fhp);
356 
357 	/* A newly created file already has a file size of zero. */
358 	if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
359 		iap->ia_valid &= ~ATTR_SIZE;
360 	if (nfsd4_create_is_exclusive(open->op_createmode)) {
361 		iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
362 				ATTR_MTIME_SET|ATTR_ATIME_SET;
363 		iap->ia_mtime.tv_sec = v_mtime;
364 		iap->ia_atime.tv_sec = v_atime;
365 		iap->ia_mtime.tv_nsec = 0;
366 		iap->ia_atime.tv_nsec = 0;
367 	}
368 
369 set_attr:
370 	status = nfsd_create_setattr(rqstp, fhp, resfhp, &attrs);
371 
372 	if (attrs.na_labelerr)
373 		open->op_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
374 	if (attrs.na_aclerr)
375 		open->op_bmval[0] &= ~FATTR4_WORD0_ACL;
376 out:
377 	inode_unlock(inode);
378 	nfsd_attrs_free(&attrs);
379 	if (child && !IS_ERR(child))
380 		dput(child);
381 	fh_drop_write(fhp);
382 	return status;
383 }
384 
385 /**
386  * set_change_info - set up the change_info4 for a reply
387  * @cinfo: pointer to nfsd4_change_info to be populated
388  * @fhp: pointer to svc_fh to use as source
389  *
390  * Many operations in NFSv4 require change_info4 in the reply. This function
391  * populates that from the info that we (should!) have already collected. In
392  * the event that we didn't get any pre-attrs, just zero out both.
393  */
394 static void
395 set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp)
396 {
397 	cinfo->atomic = (u32)(fhp->fh_pre_saved && fhp->fh_post_saved && !fhp->fh_no_atomic_attr);
398 	cinfo->before_change = fhp->fh_pre_change;
399 	cinfo->after_change = fhp->fh_post_change;
400 
401 	/*
402 	 * If fetching the pre-change attributes failed, then we should
403 	 * have already failed the whole operation. We could have still
404 	 * failed to fetch post-change attributes however.
405 	 *
406 	 * If we didn't get post-op attrs, just zero-out the after
407 	 * field since we don't know what it should be. If the pre_saved
408 	 * field isn't set for some reason, throw warning and just copy
409 	 * whatever is in the after field.
410 	 */
411 	if (WARN_ON_ONCE(!fhp->fh_pre_saved))
412 		cinfo->before_change = 0;
413 	if (!fhp->fh_post_saved)
414 		cinfo->after_change = cinfo->before_change + 1;
415 }
416 
417 static __be32
418 do_open_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh **resfh)
419 {
420 	struct svc_fh *current_fh = &cstate->current_fh;
421 	int accmode;
422 	__be32 status;
423 
424 	*resfh = kmalloc(sizeof(struct svc_fh), GFP_KERNEL);
425 	if (!*resfh)
426 		return nfserr_jukebox;
427 	fh_init(*resfh, NFS4_FHSIZE);
428 	open->op_truncate = false;
429 
430 	if (open->op_create) {
431 		/* FIXME: check session persistence and pnfs flags.
432 		 * The nfsv4.1 spec requires the following semantics:
433 		 *
434 		 * Persistent   | pNFS   | Server REQUIRED | Client Allowed
435 		 * Reply Cache  | server |                 |
436 		 * -------------+--------+-----------------+--------------------
437 		 * no           | no     | EXCLUSIVE4_1    | EXCLUSIVE4_1
438 		 *              |        |                 | (SHOULD)
439 		 *              |        | and EXCLUSIVE4  | or EXCLUSIVE4
440 		 *              |        |                 | (SHOULD NOT)
441 		 * no           | yes    | EXCLUSIVE4_1    | EXCLUSIVE4_1
442 		 * yes          | no     | GUARDED4        | GUARDED4
443 		 * yes          | yes    | GUARDED4        | GUARDED4
444 		 */
445 
446 		current->fs->umask = open->op_umask;
447 		status = nfsd4_create_file(rqstp, current_fh, *resfh, open);
448 		current->fs->umask = 0;
449 
450 		/*
451 		 * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
452 		 * use the returned bitmask to indicate which attributes
453 		 * we used to store the verifier:
454 		 */
455 		if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
456 			open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
457 						FATTR4_WORD1_TIME_MODIFY);
458 	} else {
459 		status = nfsd_lookup(rqstp, current_fh,
460 				     open->op_fname, open->op_fnamelen, *resfh);
461 		if (status == nfs_ok)
462 			/* NFSv4 protocol requires change attributes even though
463 			 * no change happened.
464 			 */
465 			status = fh_fill_both_attrs(current_fh);
466 	}
467 	if (status)
468 		goto out;
469 	status = nfsd_check_obj_isreg(*resfh);
470 	if (status)
471 		goto out;
472 
473 	nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
474 	accmode = NFSD_MAY_NOP;
475 	if (open->op_created ||
476 			open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
477 		accmode |= NFSD_MAY_OWNER_OVERRIDE;
478 	status = do_open_permission(rqstp, *resfh, open, accmode);
479 	set_change_info(&open->op_cinfo, current_fh);
480 out:
481 	return status;
482 }
483 
484 static __be32
485 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
486 {
487 	struct svc_fh *current_fh = &cstate->current_fh;
488 	int accmode = 0;
489 
490 	/* We don't know the target directory, and therefore can not
491 	* set the change info
492 	*/
493 
494 	memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
495 
496 	nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
497 
498 	open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
499 		(open->op_iattr.ia_size == 0);
500 	/*
501 	 * In the delegation case, the client is telling us about an
502 	 * open that it *already* performed locally, some time ago.  We
503 	 * should let it succeed now if possible.
504 	 *
505 	 * In the case of a CLAIM_FH open, on the other hand, the client
506 	 * may be counting on us to enforce permissions (the Linux 4.1
507 	 * client uses this for normal opens, for example).
508 	 */
509 	if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
510 		accmode = NFSD_MAY_OWNER_OVERRIDE;
511 
512 	return do_open_permission(rqstp, current_fh, open, accmode);
513 }
514 
515 static void
516 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
517 {
518 	struct nfsd4_sessionid *sid =
519 			(struct nfsd4_sessionid *)session->se_sessionid.data;
520 
521 	clid->cl_boot = sid->clientid.cl_boot;
522 	clid->cl_id = sid->clientid.cl_id;
523 }
524 
525 static __be32
526 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
527 	   union nfsd4_op_u *u)
528 {
529 	struct nfsd4_open *open = &u->open;
530 	__be32 status;
531 	struct svc_fh *resfh = NULL;
532 	struct net *net = SVC_NET(rqstp);
533 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
534 	bool reclaim = false;
535 
536 	dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
537 		(int)open->op_fnamelen, open->op_fname,
538 		open->op_openowner);
539 
540 	open->op_filp = NULL;
541 	open->op_rqstp = rqstp;
542 
543 	/* This check required by spec. */
544 	if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
545 		return nfserr_inval;
546 
547 	open->op_created = false;
548 	/*
549 	 * RFC5661 18.51.3
550 	 * Before RECLAIM_COMPLETE done, server should deny new lock
551 	 */
552 	if (nfsd4_has_session(cstate) &&
553 	    !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
554 	    open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
555 		return nfserr_grace;
556 
557 	if (nfsd4_has_session(cstate))
558 		copy_clientid(&open->op_clientid, cstate->session);
559 
560 	/* check seqid for replay. set nfs4_owner */
561 	status = nfsd4_process_open1(cstate, open, nn);
562 	if (status == nfserr_replay_me) {
563 		struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
564 		fh_put(&cstate->current_fh);
565 		fh_copy_shallow(&cstate->current_fh.fh_handle,
566 				&rp->rp_openfh);
567 		status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
568 		if (status)
569 			dprintk("nfsd4_open: replay failed"
570 				" restoring previous filehandle\n");
571 		else
572 			status = nfserr_replay_me;
573 	}
574 	if (status)
575 		goto out;
576 	if (open->op_xdr_error) {
577 		status = open->op_xdr_error;
578 		goto out;
579 	}
580 
581 	status = nfsd4_check_open_attributes(rqstp, cstate, open);
582 	if (status)
583 		goto out;
584 
585 	/* Openowner is now set, so sequence id will get bumped.  Now we need
586 	 * these checks before we do any creates: */
587 	status = nfserr_grace;
588 	if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
589 		goto out;
590 	status = nfserr_no_grace;
591 	if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
592 		goto out;
593 
594 	switch (open->op_claim_type) {
595 	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
596 	case NFS4_OPEN_CLAIM_NULL:
597 		status = do_open_lookup(rqstp, cstate, open, &resfh);
598 		if (status)
599 			goto out;
600 		break;
601 	case NFS4_OPEN_CLAIM_PREVIOUS:
602 		status = nfs4_check_open_reclaim(cstate->clp);
603 		if (status)
604 			goto out;
605 		open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
606 		reclaim = true;
607 		fallthrough;
608 	case NFS4_OPEN_CLAIM_FH:
609 	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
610 		status = do_open_fhandle(rqstp, cstate, open);
611 		if (status)
612 			goto out;
613 		resfh = &cstate->current_fh;
614 		break;
615 	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
616 	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
617 		status = nfserr_notsupp;
618 		goto out;
619 	default:
620 		status = nfserr_inval;
621 		goto out;
622 	}
623 
624 	status = nfsd4_process_open2(rqstp, resfh, open);
625 	if (status && open->op_created)
626 		pr_warn("nfsd4_process_open2 failed to open newly-created file: status=%u\n",
627 			be32_to_cpu(status));
628 	if (reclaim && !status)
629 		nn->somebody_reclaimed = true;
630 out:
631 	if (open->op_filp) {
632 		fput(open->op_filp);
633 		open->op_filp = NULL;
634 	}
635 	if (resfh && resfh != &cstate->current_fh) {
636 		fh_dup2(&cstate->current_fh, resfh);
637 		fh_put(resfh);
638 		kfree(resfh);
639 	}
640 	nfsd4_cleanup_open_state(cstate, open);
641 	nfsd4_bump_seqid(cstate, status);
642 	return status;
643 }
644 
645 /*
646  * OPEN is the only seqid-mutating operation whose decoding can fail
647  * with a seqid-mutating error (specifically, decoding of user names in
648  * the attributes).  Therefore we have to do some processing to look up
649  * the stateowner so that we can bump the seqid.
650  */
651 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
652 {
653 	struct nfsd4_open *open = &op->u.open;
654 
655 	if (!seqid_mutating_err(ntohl(op->status)))
656 		return op->status;
657 	if (nfsd4_has_session(cstate))
658 		return op->status;
659 	open->op_xdr_error = op->status;
660 	return nfsd4_open(rqstp, cstate, &op->u);
661 }
662 
663 /*
664  * filehandle-manipulating ops.
665  */
666 static __be32
667 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
668 	    union nfsd4_op_u *u)
669 {
670 	u->getfh = &cstate->current_fh;
671 	return nfs_ok;
672 }
673 
674 static __be32
675 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
676 	    union nfsd4_op_u *u)
677 {
678 	struct nfsd4_putfh *putfh = &u->putfh;
679 	__be32 ret;
680 
681 	fh_put(&cstate->current_fh);
682 	cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
683 	memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
684 	       putfh->pf_fhlen);
685 	ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
686 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
687 	if (ret == nfserr_stale && putfh->no_verify) {
688 		SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
689 		ret = 0;
690 	}
691 #endif
692 	return ret;
693 }
694 
695 static __be32
696 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
697 		union nfsd4_op_u *u)
698 {
699 	fh_put(&cstate->current_fh);
700 
701 	return exp_pseudoroot(rqstp, &cstate->current_fh);
702 }
703 
704 static __be32
705 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
706 		union nfsd4_op_u *u)
707 {
708 	if (!cstate->save_fh.fh_dentry)
709 		return nfserr_restorefh;
710 
711 	fh_dup2(&cstate->current_fh, &cstate->save_fh);
712 	if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
713 		memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
714 		SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
715 	}
716 	return nfs_ok;
717 }
718 
719 static __be32
720 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
721 	     union nfsd4_op_u *u)
722 {
723 	fh_dup2(&cstate->save_fh, &cstate->current_fh);
724 	if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
725 		memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
726 		SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
727 	}
728 	return nfs_ok;
729 }
730 
731 /*
732  * misc nfsv4 ops
733  */
734 static __be32
735 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
736 	     union nfsd4_op_u *u)
737 {
738 	struct nfsd4_access *access = &u->access;
739 	u32 access_full;
740 
741 	access_full = NFS3_ACCESS_FULL;
742 	if (cstate->minorversion >= 2)
743 		access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
744 			       NFS4_ACCESS_XAWRITE;
745 
746 	if (access->ac_req_access & ~access_full)
747 		return nfserr_inval;
748 
749 	access->ac_resp_access = access->ac_req_access;
750 	return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
751 			   &access->ac_supported);
752 }
753 
754 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
755 {
756 	__be32 *verf = (__be32 *)verifier->data;
757 
758 	BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
759 
760 	nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
761 }
762 
763 static __be32
764 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
765 	     union nfsd4_op_u *u)
766 {
767 	struct nfsd4_commit *commit = &u->commit;
768 	struct nfsd_file *nf;
769 	__be32 status;
770 
771 	status = nfsd_file_acquire(rqstp, &cstate->current_fh, NFSD_MAY_WRITE |
772 				   NFSD_MAY_NOT_BREAK_LEASE, &nf);
773 	if (status != nfs_ok)
774 		return status;
775 
776 	status = nfsd_commit(rqstp, &cstate->current_fh, nf, commit->co_offset,
777 			     commit->co_count,
778 			     (__be32 *)commit->co_verf.data);
779 	nfsd_file_put(nf);
780 	return status;
781 }
782 
783 static __be32
784 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
785 	     union nfsd4_op_u *u)
786 {
787 	struct nfsd4_create *create = &u->create;
788 	struct nfsd_attrs attrs = {
789 		.na_iattr	= &create->cr_iattr,
790 		.na_seclabel	= &create->cr_label,
791 	};
792 	struct svc_fh resfh;
793 	__be32 status;
794 	dev_t rdev;
795 
796 	fh_init(&resfh, NFS4_FHSIZE);
797 
798 	status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
799 	if (status)
800 		return status;
801 
802 	status = check_attr_support(rqstp, cstate, create->cr_bmval,
803 				    nfsd_attrmask);
804 	if (status)
805 		return status;
806 
807 	status = nfsd4_acl_to_attr(create->cr_type, create->cr_acl, &attrs);
808 	current->fs->umask = create->cr_umask;
809 	switch (create->cr_type) {
810 	case NF4LNK:
811 		status = nfsd_symlink(rqstp, &cstate->current_fh,
812 				      create->cr_name, create->cr_namelen,
813 				      create->cr_data, &attrs, &resfh);
814 		break;
815 
816 	case NF4BLK:
817 		status = nfserr_inval;
818 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
819 		if (MAJOR(rdev) != create->cr_specdata1 ||
820 		    MINOR(rdev) != create->cr_specdata2)
821 			goto out_umask;
822 		status = nfsd_create(rqstp, &cstate->current_fh,
823 				     create->cr_name, create->cr_namelen,
824 				     &attrs, S_IFBLK, rdev, &resfh);
825 		break;
826 
827 	case NF4CHR:
828 		status = nfserr_inval;
829 		rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
830 		if (MAJOR(rdev) != create->cr_specdata1 ||
831 		    MINOR(rdev) != create->cr_specdata2)
832 			goto out_umask;
833 		status = nfsd_create(rqstp, &cstate->current_fh,
834 				     create->cr_name, create->cr_namelen,
835 				     &attrs, S_IFCHR, rdev, &resfh);
836 		break;
837 
838 	case NF4SOCK:
839 		status = nfsd_create(rqstp, &cstate->current_fh,
840 				     create->cr_name, create->cr_namelen,
841 				     &attrs, S_IFSOCK, 0, &resfh);
842 		break;
843 
844 	case NF4FIFO:
845 		status = nfsd_create(rqstp, &cstate->current_fh,
846 				     create->cr_name, create->cr_namelen,
847 				     &attrs, S_IFIFO, 0, &resfh);
848 		break;
849 
850 	case NF4DIR:
851 		create->cr_iattr.ia_valid &= ~ATTR_SIZE;
852 		status = nfsd_create(rqstp, &cstate->current_fh,
853 				     create->cr_name, create->cr_namelen,
854 				     &attrs, S_IFDIR, 0, &resfh);
855 		break;
856 
857 	default:
858 		status = nfserr_badtype;
859 	}
860 
861 	if (status)
862 		goto out;
863 
864 	if (attrs.na_labelerr)
865 		create->cr_bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
866 	if (attrs.na_aclerr)
867 		create->cr_bmval[0] &= ~FATTR4_WORD0_ACL;
868 	set_change_info(&create->cr_cinfo, &cstate->current_fh);
869 	fh_dup2(&cstate->current_fh, &resfh);
870 out:
871 	fh_put(&resfh);
872 out_umask:
873 	current->fs->umask = 0;
874 	nfsd_attrs_free(&attrs);
875 	return status;
876 }
877 
878 static __be32
879 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
880 	      union nfsd4_op_u *u)
881 {
882 	struct nfsd4_getattr *getattr = &u->getattr;
883 	__be32 status;
884 
885 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
886 	if (status)
887 		return status;
888 
889 	if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
890 		return nfserr_inval;
891 
892 	getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
893 	getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
894 	getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
895 
896 	getattr->ga_fhp = &cstate->current_fh;
897 	return nfs_ok;
898 }
899 
900 static __be32
901 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
902 	   union nfsd4_op_u *u)
903 {
904 	struct nfsd4_link *link = &u->link;
905 	__be32 status;
906 
907 	status = nfsd_link(rqstp, &cstate->current_fh,
908 			   link->li_name, link->li_namelen, &cstate->save_fh);
909 	if (!status)
910 		set_change_info(&link->li_cinfo, &cstate->current_fh);
911 	return status;
912 }
913 
914 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
915 {
916 	struct svc_fh tmp_fh;
917 	__be32 ret;
918 
919 	fh_init(&tmp_fh, NFS4_FHSIZE);
920 	ret = exp_pseudoroot(rqstp, &tmp_fh);
921 	if (ret)
922 		return ret;
923 	if (tmp_fh.fh_dentry == fh->fh_dentry) {
924 		fh_put(&tmp_fh);
925 		return nfserr_noent;
926 	}
927 	fh_put(&tmp_fh);
928 	return nfsd_lookup(rqstp, fh, "..", 2, fh);
929 }
930 
931 static __be32
932 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
933 	      union nfsd4_op_u *u)
934 {
935 	return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
936 }
937 
938 static __be32
939 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
940 	     union nfsd4_op_u *u)
941 {
942 	return nfsd_lookup(rqstp, &cstate->current_fh,
943 			   u->lookup.lo_name, u->lookup.lo_len,
944 			   &cstate->current_fh);
945 }
946 
947 static __be32
948 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
949 	   union nfsd4_op_u *u)
950 {
951 	struct nfsd4_read *read = &u->read;
952 	__be32 status;
953 
954 	read->rd_nf = NULL;
955 
956 	trace_nfsd_read_start(rqstp, &cstate->current_fh,
957 			      read->rd_offset, read->rd_length);
958 
959 	read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
960 	if (read->rd_offset > (u64)OFFSET_MAX)
961 		read->rd_offset = (u64)OFFSET_MAX;
962 	if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
963 		read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
964 
965 	/*
966 	 * If we do a zero copy read, then a client will see read data
967 	 * that reflects the state of the file *after* performing the
968 	 * following compound.
969 	 *
970 	 * To ensure proper ordering, we therefore turn off zero copy if
971 	 * the client wants us to do more in this compound:
972 	 */
973 	if (!nfsd4_last_compound_op(rqstp)) {
974 		struct nfsd4_compoundargs *argp = rqstp->rq_argp;
975 
976 		argp->splice_ok = false;
977 	}
978 
979 	/* check stateid */
980 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
981 					&read->rd_stateid, RD_STATE,
982 					&read->rd_nf, NULL);
983 
984 	read->rd_rqstp = rqstp;
985 	read->rd_fhp = &cstate->current_fh;
986 	return status;
987 }
988 
989 
990 static void
991 nfsd4_read_release(union nfsd4_op_u *u)
992 {
993 	if (u->read.rd_nf)
994 		nfsd_file_put(u->read.rd_nf);
995 	trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
996 			     u->read.rd_offset, u->read.rd_length);
997 }
998 
999 static __be32
1000 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1001 	      union nfsd4_op_u *u)
1002 {
1003 	struct nfsd4_readdir *readdir = &u->readdir;
1004 	u64 cookie = readdir->rd_cookie;
1005 	static const nfs4_verifier zeroverf;
1006 
1007 	/* no need to check permission - this will be done in nfsd_readdir() */
1008 
1009 	if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1010 		return nfserr_inval;
1011 
1012 	readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1013 	readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1014 	readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1015 
1016 	if ((cookie == 1) || (cookie == 2) ||
1017 	    (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1018 		return nfserr_bad_cookie;
1019 
1020 	readdir->rd_rqstp = rqstp;
1021 	readdir->rd_fhp = &cstate->current_fh;
1022 	return nfs_ok;
1023 }
1024 
1025 static __be32
1026 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1027 	       union nfsd4_op_u *u)
1028 {
1029 	u->readlink.rl_rqstp = rqstp;
1030 	u->readlink.rl_fhp = &cstate->current_fh;
1031 	return nfs_ok;
1032 }
1033 
1034 static __be32
1035 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1036 	     union nfsd4_op_u *u)
1037 {
1038 	struct nfsd4_remove *remove = &u->remove;
1039 	__be32 status;
1040 
1041 	if (opens_in_grace(SVC_NET(rqstp)))
1042 		return nfserr_grace;
1043 	status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1044 			     remove->rm_name, remove->rm_namelen);
1045 	if (!status)
1046 		set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1047 	return status;
1048 }
1049 
1050 static __be32
1051 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1052 	     union nfsd4_op_u *u)
1053 {
1054 	struct nfsd4_rename *rename = &u->rename;
1055 	__be32 status;
1056 
1057 	if (opens_in_grace(SVC_NET(rqstp)))
1058 		return nfserr_grace;
1059 	status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1060 			     rename->rn_snamelen, &cstate->current_fh,
1061 			     rename->rn_tname, rename->rn_tnamelen);
1062 	if (status)
1063 		return status;
1064 	set_change_info(&rename->rn_sinfo, &cstate->save_fh);
1065 	set_change_info(&rename->rn_tinfo, &cstate->current_fh);
1066 	return nfs_ok;
1067 }
1068 
1069 static __be32
1070 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1071 	      union nfsd4_op_u *u)
1072 {
1073 	struct nfsd4_secinfo *secinfo = &u->secinfo;
1074 	struct svc_export *exp;
1075 	struct dentry *dentry;
1076 	__be32 err;
1077 
1078 	err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1079 	if (err)
1080 		return err;
1081 	err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1082 				    secinfo->si_name, secinfo->si_namelen,
1083 				    &exp, &dentry);
1084 	if (err)
1085 		return err;
1086 	if (d_really_is_negative(dentry)) {
1087 		exp_put(exp);
1088 		err = nfserr_noent;
1089 	} else
1090 		secinfo->si_exp = exp;
1091 	dput(dentry);
1092 	if (cstate->minorversion)
1093 		/* See rfc 5661 section 2.6.3.1.1.8 */
1094 		fh_put(&cstate->current_fh);
1095 	return err;
1096 }
1097 
1098 static __be32
1099 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1100 		union nfsd4_op_u *u)
1101 {
1102 	__be32 err;
1103 
1104 	switch (u->secinfo_no_name.sin_style) {
1105 	case NFS4_SECINFO_STYLE4_CURRENT_FH:
1106 		break;
1107 	case NFS4_SECINFO_STYLE4_PARENT:
1108 		err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1109 		if (err)
1110 			return err;
1111 		break;
1112 	default:
1113 		return nfserr_inval;
1114 	}
1115 
1116 	u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1117 	fh_put(&cstate->current_fh);
1118 	return nfs_ok;
1119 }
1120 
1121 static void
1122 nfsd4_secinfo_release(union nfsd4_op_u *u)
1123 {
1124 	if (u->secinfo.si_exp)
1125 		exp_put(u->secinfo.si_exp);
1126 }
1127 
1128 static void
1129 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1130 {
1131 	if (u->secinfo_no_name.sin_exp)
1132 		exp_put(u->secinfo_no_name.sin_exp);
1133 }
1134 
1135 static __be32
1136 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1137 	      union nfsd4_op_u *u)
1138 {
1139 	struct nfsd4_setattr *setattr = &u->setattr;
1140 	struct nfsd_attrs attrs = {
1141 		.na_iattr	= &setattr->sa_iattr,
1142 		.na_seclabel	= &setattr->sa_label,
1143 	};
1144 	struct inode *inode;
1145 	__be32 status = nfs_ok;
1146 	bool save_no_wcc;
1147 	int err;
1148 
1149 	if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1150 		status = nfs4_preprocess_stateid_op(rqstp, cstate,
1151 				&cstate->current_fh, &setattr->sa_stateid,
1152 				WR_STATE, NULL, NULL);
1153 		if (status)
1154 			return status;
1155 	}
1156 	err = fh_want_write(&cstate->current_fh);
1157 	if (err)
1158 		return nfserrno(err);
1159 	status = nfs_ok;
1160 
1161 	status = check_attr_support(rqstp, cstate, setattr->sa_bmval,
1162 				    nfsd_attrmask);
1163 	if (status)
1164 		goto out;
1165 
1166 	inode = cstate->current_fh.fh_dentry->d_inode;
1167 	status = nfsd4_acl_to_attr(S_ISDIR(inode->i_mode) ? NF4DIR : NF4REG,
1168 				   setattr->sa_acl, &attrs);
1169 
1170 	if (status)
1171 		goto out;
1172 	save_no_wcc = cstate->current_fh.fh_no_wcc;
1173 	cstate->current_fh.fh_no_wcc = true;
1174 	status = nfsd_setattr(rqstp, &cstate->current_fh, &attrs, NULL);
1175 	cstate->current_fh.fh_no_wcc = save_no_wcc;
1176 	if (!status)
1177 		status = nfserrno(attrs.na_labelerr);
1178 	if (!status)
1179 		status = nfserrno(attrs.na_aclerr);
1180 out:
1181 	nfsd_attrs_free(&attrs);
1182 	fh_drop_write(&cstate->current_fh);
1183 	return status;
1184 }
1185 
1186 static __be32
1187 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1188 	    union nfsd4_op_u *u)
1189 {
1190 	struct nfsd4_write *write = &u->write;
1191 	stateid_t *stateid = &write->wr_stateid;
1192 	struct nfsd_file *nf = NULL;
1193 	__be32 status = nfs_ok;
1194 	unsigned long cnt;
1195 	int nvecs;
1196 
1197 	if (write->wr_offset > (u64)OFFSET_MAX ||
1198 	    write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1199 		return nfserr_fbig;
1200 
1201 	cnt = write->wr_buflen;
1202 	trace_nfsd_write_start(rqstp, &cstate->current_fh,
1203 			       write->wr_offset, cnt);
1204 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1205 						stateid, WR_STATE, &nf, NULL);
1206 	if (status)
1207 		return status;
1208 
1209 	write->wr_how_written = write->wr_stable_how;
1210 
1211 	nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1212 	WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1213 
1214 	status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1215 				write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1216 				write->wr_how_written,
1217 				(__be32 *)write->wr_verifier.data);
1218 	nfsd_file_put(nf);
1219 
1220 	write->wr_bytes_written = cnt;
1221 	trace_nfsd_write_done(rqstp, &cstate->current_fh,
1222 			      write->wr_offset, cnt);
1223 	return status;
1224 }
1225 
1226 static __be32
1227 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1228 		  stateid_t *src_stateid, struct nfsd_file **src,
1229 		  stateid_t *dst_stateid, struct nfsd_file **dst)
1230 {
1231 	__be32 status;
1232 
1233 	if (!cstate->save_fh.fh_dentry)
1234 		return nfserr_nofilehandle;
1235 
1236 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1237 					    src_stateid, RD_STATE, src, NULL);
1238 	if (status)
1239 		goto out;
1240 
1241 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1242 					    dst_stateid, WR_STATE, dst, NULL);
1243 	if (status)
1244 		goto out_put_src;
1245 
1246 	/* fix up for NFS-specific error code */
1247 	if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1248 	    !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1249 		status = nfserr_wrong_type;
1250 		goto out_put_dst;
1251 	}
1252 
1253 out:
1254 	return status;
1255 out_put_dst:
1256 	nfsd_file_put(*dst);
1257 	*dst = NULL;
1258 out_put_src:
1259 	nfsd_file_put(*src);
1260 	*src = NULL;
1261 	goto out;
1262 }
1263 
1264 static __be32
1265 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1266 		union nfsd4_op_u *u)
1267 {
1268 	struct nfsd4_clone *clone = &u->clone;
1269 	struct nfsd_file *src, *dst;
1270 	__be32 status;
1271 
1272 	status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1273 				   &clone->cl_dst_stateid, &dst);
1274 	if (status)
1275 		goto out;
1276 
1277 	status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1278 			dst, clone->cl_dst_pos, clone->cl_count,
1279 			EX_ISSYNC(cstate->current_fh.fh_export));
1280 
1281 	nfsd_file_put(dst);
1282 	nfsd_file_put(src);
1283 out:
1284 	return status;
1285 }
1286 
1287 static void nfs4_put_copy(struct nfsd4_copy *copy)
1288 {
1289 	if (!refcount_dec_and_test(&copy->refcount))
1290 		return;
1291 	kfree(copy->cp_src);
1292 	kfree(copy);
1293 }
1294 
1295 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1296 {
1297 	if (!test_and_set_bit(NFSD4_COPY_F_STOPPED, &copy->cp_flags))
1298 		kthread_stop(copy->copy_task);
1299 	nfs4_put_copy(copy);
1300 }
1301 
1302 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1303 {
1304 	struct nfsd4_copy *copy = NULL;
1305 
1306 	spin_lock(&clp->async_lock);
1307 	if (!list_empty(&clp->async_copies)) {
1308 		copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1309 					copies);
1310 		refcount_inc(&copy->refcount);
1311 	}
1312 	spin_unlock(&clp->async_lock);
1313 	return copy;
1314 }
1315 
1316 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1317 {
1318 	struct nfsd4_copy *copy;
1319 
1320 	while ((copy = nfsd4_get_copy(clp)) != NULL)
1321 		nfsd4_stop_copy(copy);
1322 }
1323 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1324 
1325 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1326 				   struct nfs_fh *src_fh,
1327 				   nfs4_stateid *stateid);
1328 extern void nfs42_ssc_close(struct file *filep);
1329 
1330 extern void nfs_sb_deactive(struct super_block *sb);
1331 
1332 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1333 
1334 /*
1335  * setup a work entry in the ssc delayed unmount list.
1336  */
1337 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1338 				  struct nfsd4_ssc_umount_item **nsui,
1339 				  struct svc_rqst *rqstp)
1340 {
1341 	struct nfsd4_ssc_umount_item *ni = NULL;
1342 	struct nfsd4_ssc_umount_item *work = NULL;
1343 	struct nfsd4_ssc_umount_item *tmp;
1344 	DEFINE_WAIT(wait);
1345 	__be32 status = 0;
1346 
1347 	*nsui = NULL;
1348 	work = kzalloc(sizeof(*work), GFP_KERNEL);
1349 try_again:
1350 	spin_lock(&nn->nfsd_ssc_lock);
1351 	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1352 		if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1353 			continue;
1354 		/* found a match */
1355 		if (ni->nsui_busy) {
1356 			/*  wait - and try again */
1357 			prepare_to_wait(&nn->nfsd_ssc_waitq, &wait, TASK_IDLE);
1358 			spin_unlock(&nn->nfsd_ssc_lock);
1359 
1360 			/* allow 20secs for mount/unmount for now - revisit */
1361 			if (svc_thread_should_stop(rqstp) ||
1362 					(schedule_timeout(20*HZ) == 0)) {
1363 				finish_wait(&nn->nfsd_ssc_waitq, &wait);
1364 				kfree(work);
1365 				return nfserr_eagain;
1366 			}
1367 			finish_wait(&nn->nfsd_ssc_waitq, &wait);
1368 			goto try_again;
1369 		}
1370 		*nsui = ni;
1371 		refcount_inc(&ni->nsui_refcnt);
1372 		spin_unlock(&nn->nfsd_ssc_lock);
1373 		kfree(work);
1374 
1375 		/* return vfsmount in (*nsui)->nsui_vfsmount */
1376 		return 0;
1377 	}
1378 	if (work) {
1379 		strscpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr) - 1);
1380 		refcount_set(&work->nsui_refcnt, 2);
1381 		work->nsui_busy = true;
1382 		list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1383 		*nsui = work;
1384 	} else
1385 		status = nfserr_resource;
1386 	spin_unlock(&nn->nfsd_ssc_lock);
1387 	return status;
1388 }
1389 
1390 static void nfsd4_ssc_update_dul(struct nfsd_net *nn,
1391 				 struct nfsd4_ssc_umount_item *nsui,
1392 				 struct vfsmount *ss_mnt)
1393 {
1394 	spin_lock(&nn->nfsd_ssc_lock);
1395 	nsui->nsui_vfsmount = ss_mnt;
1396 	nsui->nsui_busy = false;
1397 	wake_up_all(&nn->nfsd_ssc_waitq);
1398 	spin_unlock(&nn->nfsd_ssc_lock);
1399 }
1400 
1401 static void nfsd4_ssc_cancel_dul(struct nfsd_net *nn,
1402 				 struct nfsd4_ssc_umount_item *nsui)
1403 {
1404 	spin_lock(&nn->nfsd_ssc_lock);
1405 	list_del(&nsui->nsui_list);
1406 	wake_up_all(&nn->nfsd_ssc_waitq);
1407 	spin_unlock(&nn->nfsd_ssc_lock);
1408 	kfree(nsui);
1409 }
1410 
1411 /*
1412  * Support one copy source server for now.
1413  */
1414 static __be32
1415 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1416 		       struct nfsd4_ssc_umount_item **nsui)
1417 {
1418 	struct file_system_type *type;
1419 	struct vfsmount *ss_mnt;
1420 	struct nfs42_netaddr *naddr;
1421 	struct sockaddr_storage tmp_addr;
1422 	size_t tmp_addrlen, match_netid_len = 3;
1423 	char *startsep = "", *endsep = "", *match_netid = "tcp";
1424 	char *ipaddr, *dev_name, *raw_data;
1425 	int len, raw_len;
1426 	__be32 status = nfserr_inval;
1427 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1428 
1429 	naddr = &nss->u.nl4_addr;
1430 	tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1431 					 naddr->addr_len,
1432 					 (struct sockaddr *)&tmp_addr,
1433 					 sizeof(tmp_addr));
1434 	*nsui = NULL;
1435 	if (tmp_addrlen == 0)
1436 		goto out_err;
1437 
1438 	if (tmp_addr.ss_family == AF_INET6) {
1439 		startsep = "[";
1440 		endsep = "]";
1441 		match_netid = "tcp6";
1442 		match_netid_len = 4;
1443 	}
1444 
1445 	if (naddr->netid_len != match_netid_len ||
1446 		strncmp(naddr->netid, match_netid, naddr->netid_len))
1447 		goto out_err;
1448 
1449 	/* Construct the raw data for the vfs_kern_mount call */
1450 	len = RPC_MAX_ADDRBUFLEN + 1;
1451 	ipaddr = kzalloc(len, GFP_KERNEL);
1452 	if (!ipaddr)
1453 		goto out_err;
1454 
1455 	rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1456 
1457 	/* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1458 
1459 	raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1460 	raw_data = kzalloc(raw_len, GFP_KERNEL);
1461 	if (!raw_data)
1462 		goto out_free_ipaddr;
1463 
1464 	snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1465 
1466 	status = nfserr_nodev;
1467 	type = get_fs_type("nfs");
1468 	if (!type)
1469 		goto out_free_rawdata;
1470 
1471 	/* Set the server:<export> for the vfs_kern_mount call */
1472 	dev_name = kzalloc(len + 5, GFP_KERNEL);
1473 	if (!dev_name)
1474 		goto out_free_rawdata;
1475 	snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1476 
1477 	status = nfsd4_ssc_setup_dul(nn, ipaddr, nsui, rqstp);
1478 	if (status)
1479 		goto out_free_devname;
1480 	if ((*nsui)->nsui_vfsmount)
1481 		goto out_done;
1482 
1483 	/* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1484 	ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1485 	module_put(type->owner);
1486 	if (IS_ERR(ss_mnt)) {
1487 		status = nfserr_nodev;
1488 		nfsd4_ssc_cancel_dul(nn, *nsui);
1489 		goto out_free_devname;
1490 	}
1491 	nfsd4_ssc_update_dul(nn, *nsui, ss_mnt);
1492 out_done:
1493 	status = 0;
1494 
1495 out_free_devname:
1496 	kfree(dev_name);
1497 out_free_rawdata:
1498 	kfree(raw_data);
1499 out_free_ipaddr:
1500 	kfree(ipaddr);
1501 out_err:
1502 	return status;
1503 }
1504 
1505 /*
1506  * Verify COPY destination stateid.
1507  *
1508  * Connect to the source server with NFSv4.1.
1509  * Create the source struct file for nfsd_copy_range.
1510  * Called with COPY cstate:
1511  *    SAVED_FH: source filehandle
1512  *    CURRENT_FH: destination filehandle
1513  */
1514 static __be32
1515 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1516 		      struct nfsd4_compound_state *cstate,
1517 		      struct nfsd4_copy *copy)
1518 {
1519 	struct svc_fh *s_fh = NULL;
1520 	stateid_t *s_stid = &copy->cp_src_stateid;
1521 	__be32 status = nfserr_inval;
1522 
1523 	/* Verify the destination stateid and set dst struct file*/
1524 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1525 					    &copy->cp_dst_stateid,
1526 					    WR_STATE, &copy->nf_dst, NULL);
1527 	if (status)
1528 		goto out;
1529 
1530 	status = nfsd4_interssc_connect(copy->cp_src, rqstp, &copy->ss_nsui);
1531 	if (status)
1532 		goto out;
1533 
1534 	s_fh = &cstate->save_fh;
1535 
1536 	copy->c_fh.size = s_fh->fh_handle.fh_size;
1537 	memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1538 	copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1539 	memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1540 	       sizeof(stateid_opaque_t));
1541 
1542 	status = 0;
1543 out:
1544 	return status;
1545 }
1546 
1547 static void
1548 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1549 			struct nfsd_file *dst)
1550 {
1551 	struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1552 	long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1553 
1554 	nfs42_ssc_close(filp);
1555 	fput(filp);
1556 
1557 	spin_lock(&nn->nfsd_ssc_lock);
1558 	list_del(&nsui->nsui_list);
1559 	/*
1560 	 * vfsmount can be shared by multiple exports,
1561 	 * decrement refcnt. If the count drops to 1 it
1562 	 * will be unmounted when nsui_expire expires.
1563 	 */
1564 	refcount_dec(&nsui->nsui_refcnt);
1565 	nsui->nsui_expire = jiffies + timeout;
1566 	list_add_tail(&nsui->nsui_list, &nn->nfsd_ssc_mount_list);
1567 	spin_unlock(&nn->nfsd_ssc_lock);
1568 }
1569 
1570 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1571 
1572 static __be32
1573 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1574 		      struct nfsd4_compound_state *cstate,
1575 		      struct nfsd4_copy *copy)
1576 {
1577 	return nfserr_inval;
1578 }
1579 
1580 static void
1581 nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
1582 			struct nfsd_file *dst)
1583 {
1584 }
1585 
1586 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1587 				   struct nfs_fh *src_fh,
1588 				   nfs4_stateid *stateid)
1589 {
1590 	return NULL;
1591 }
1592 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1593 
1594 static __be32
1595 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1596 		      struct nfsd4_compound_state *cstate,
1597 		      struct nfsd4_copy *copy)
1598 {
1599 	return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1600 				 &copy->nf_src, &copy->cp_dst_stateid,
1601 				 &copy->nf_dst);
1602 }
1603 
1604 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1605 {
1606 	struct nfsd4_cb_offload *cbo =
1607 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1608 
1609 	kfree(cbo);
1610 }
1611 
1612 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1613 				 struct rpc_task *task)
1614 {
1615 	struct nfsd4_cb_offload *cbo =
1616 		container_of(cb, struct nfsd4_cb_offload, co_cb);
1617 
1618 	trace_nfsd_cb_offload_done(&cbo->co_res.cb_stateid, task);
1619 	return 1;
1620 }
1621 
1622 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1623 	.release = nfsd4_cb_offload_release,
1624 	.done = nfsd4_cb_offload_done
1625 };
1626 
1627 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1628 {
1629 	copy->cp_res.wr_stable_how =
1630 		test_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags) ?
1631 			NFS_FILE_SYNC : NFS_UNSTABLE;
1632 	nfsd4_copy_set_sync(copy, sync);
1633 	gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1634 }
1635 
1636 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy,
1637 				     struct file *dst,
1638 				     struct file *src)
1639 {
1640 	errseq_t since;
1641 	ssize_t bytes_copied = 0;
1642 	u64 bytes_total = copy->cp_count;
1643 	u64 src_pos = copy->cp_src_pos;
1644 	u64 dst_pos = copy->cp_dst_pos;
1645 	int status;
1646 	loff_t end;
1647 
1648 	/* See RFC 7862 p.67: */
1649 	if (bytes_total == 0)
1650 		bytes_total = ULLONG_MAX;
1651 	do {
1652 		/* Only async copies can be stopped here */
1653 		if (kthread_should_stop())
1654 			break;
1655 		bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1656 						    bytes_total);
1657 		if (bytes_copied <= 0)
1658 			break;
1659 		bytes_total -= bytes_copied;
1660 		copy->cp_res.wr_bytes_written += bytes_copied;
1661 		src_pos += bytes_copied;
1662 		dst_pos += bytes_copied;
1663 	} while (bytes_total > 0 && nfsd4_copy_is_async(copy));
1664 	/* for a non-zero asynchronous copy do a commit of data */
1665 	if (nfsd4_copy_is_async(copy) && copy->cp_res.wr_bytes_written > 0) {
1666 		since = READ_ONCE(dst->f_wb_err);
1667 		end = copy->cp_dst_pos + copy->cp_res.wr_bytes_written - 1;
1668 		status = vfs_fsync_range(dst, copy->cp_dst_pos, end, 0);
1669 		if (!status)
1670 			status = filemap_check_wb_err(dst->f_mapping, since);
1671 		if (!status)
1672 			set_bit(NFSD4_COPY_F_COMMITTED, &copy->cp_flags);
1673 	}
1674 	return bytes_copied;
1675 }
1676 
1677 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy,
1678 			    struct file *src, struct file *dst,
1679 			    bool sync)
1680 {
1681 	__be32 status;
1682 	ssize_t bytes;
1683 
1684 	bytes = _nfsd_copy_file_range(copy, dst, src);
1685 
1686 	/* for async copy, we ignore the error, client can always retry
1687 	 * to get the error
1688 	 */
1689 	if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1690 		status = nfserrno(bytes);
1691 	else {
1692 		nfsd4_init_copy_res(copy, sync);
1693 		status = nfs_ok;
1694 	}
1695 	return status;
1696 }
1697 
1698 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1699 {
1700 	dst->cp_src_pos = src->cp_src_pos;
1701 	dst->cp_dst_pos = src->cp_dst_pos;
1702 	dst->cp_count = src->cp_count;
1703 	dst->cp_flags = src->cp_flags;
1704 	memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1705 	memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1706 	dst->cp_clp = src->cp_clp;
1707 	dst->nf_dst = nfsd_file_get(src->nf_dst);
1708 	/* for inter, nf_src doesn't exist yet */
1709 	if (!nfsd4_ssc_is_inter(src))
1710 		dst->nf_src = nfsd_file_get(src->nf_src);
1711 
1712 	memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1713 	memcpy(dst->cp_src, src->cp_src, sizeof(struct nl4_server));
1714 	memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1715 	memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1716 	dst->ss_nsui = src->ss_nsui;
1717 }
1718 
1719 static void release_copy_files(struct nfsd4_copy *copy)
1720 {
1721 	if (copy->nf_src)
1722 		nfsd_file_put(copy->nf_src);
1723 	if (copy->nf_dst)
1724 		nfsd_file_put(copy->nf_dst);
1725 }
1726 
1727 static void cleanup_async_copy(struct nfsd4_copy *copy)
1728 {
1729 	nfs4_free_copy_state(copy);
1730 	release_copy_files(copy);
1731 	if (copy->cp_clp) {
1732 		spin_lock(&copy->cp_clp->async_lock);
1733 		if (!list_empty(&copy->copies))
1734 			list_del_init(&copy->copies);
1735 		spin_unlock(&copy->cp_clp->async_lock);
1736 	}
1737 	nfs4_put_copy(copy);
1738 }
1739 
1740 static void nfsd4_send_cb_offload(struct nfsd4_copy *copy)
1741 {
1742 	struct nfsd4_cb_offload *cbo;
1743 
1744 	cbo = kzalloc(sizeof(*cbo), GFP_KERNEL);
1745 	if (!cbo)
1746 		return;
1747 
1748 	memcpy(&cbo->co_res, &copy->cp_res, sizeof(copy->cp_res));
1749 	memcpy(&cbo->co_fh, &copy->fh, sizeof(copy->fh));
1750 	cbo->co_nfserr = copy->nfserr;
1751 
1752 	nfsd4_init_cb(&cbo->co_cb, copy->cp_clp, &nfsd4_cb_offload_ops,
1753 		      NFSPROC4_CLNT_CB_OFFLOAD);
1754 	trace_nfsd_cb_offload(copy->cp_clp, &cbo->co_res.cb_stateid,
1755 			      &cbo->co_fh, copy->cp_count, copy->nfserr);
1756 	nfsd4_run_cb(&cbo->co_cb);
1757 }
1758 
1759 /**
1760  * nfsd4_do_async_copy - kthread function for background server-side COPY
1761  * @data: arguments for COPY operation
1762  *
1763  * Return values:
1764  *   %0: Copy operation is done.
1765  */
1766 static int nfsd4_do_async_copy(void *data)
1767 {
1768 	struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1769 
1770 	trace_nfsd_copy_do_async(copy);
1771 	if (nfsd4_ssc_is_inter(copy)) {
1772 		struct file *filp;
1773 
1774 		filp = nfs42_ssc_open(copy->ss_nsui->nsui_vfsmount,
1775 				      &copy->c_fh, &copy->stateid);
1776 		if (IS_ERR(filp)) {
1777 			switch (PTR_ERR(filp)) {
1778 			case -EBADF:
1779 				copy->nfserr = nfserr_wrong_type;
1780 				break;
1781 			default:
1782 				copy->nfserr = nfserr_offload_denied;
1783 			}
1784 			/* ss_mnt will be unmounted by the laundromat */
1785 			goto do_callback;
1786 		}
1787 		copy->nfserr = nfsd4_do_copy(copy, filp, copy->nf_dst->nf_file,
1788 					     false);
1789 		nfsd4_cleanup_inter_ssc(copy->ss_nsui, filp, copy->nf_dst);
1790 	} else {
1791 		copy->nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1792 					     copy->nf_dst->nf_file, false);
1793 	}
1794 
1795 do_callback:
1796 	set_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags);
1797 	nfsd4_send_cb_offload(copy);
1798 	cleanup_async_copy(copy);
1799 	return 0;
1800 }
1801 
1802 static __be32
1803 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1804 		union nfsd4_op_u *u)
1805 {
1806 	struct nfsd4_copy *copy = &u->copy;
1807 	__be32 status;
1808 	struct nfsd4_copy *async_copy = NULL;
1809 
1810 	/*
1811 	 * Currently, async COPY is not reliable. Force all COPY
1812 	 * requests to be synchronous to avoid client application
1813 	 * hangs waiting for COPY completion.
1814 	 */
1815 	nfsd4_copy_set_sync(copy, true);
1816 
1817 	copy->cp_clp = cstate->clp;
1818 	if (nfsd4_ssc_is_inter(copy)) {
1819 		trace_nfsd_copy_inter(copy);
1820 		if (!inter_copy_offload_enable || nfsd4_copy_is_sync(copy)) {
1821 			status = nfserr_notsupp;
1822 			goto out;
1823 		}
1824 		status = nfsd4_setup_inter_ssc(rqstp, cstate, copy);
1825 		if (status) {
1826 			trace_nfsd_copy_done(copy, status);
1827 			return nfserr_offload_denied;
1828 		}
1829 	} else {
1830 		trace_nfsd_copy_intra(copy);
1831 		status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1832 		if (status) {
1833 			trace_nfsd_copy_done(copy, status);
1834 			return status;
1835 		}
1836 	}
1837 
1838 	memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1839 		sizeof(struct knfsd_fh));
1840 	if (nfsd4_copy_is_async(copy)) {
1841 		struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1842 
1843 		status = nfserrno(-ENOMEM);
1844 		async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1845 		if (!async_copy)
1846 			goto out_err;
1847 		INIT_LIST_HEAD(&async_copy->copies);
1848 		refcount_set(&async_copy->refcount, 1);
1849 		async_copy->cp_src = kmalloc(sizeof(*async_copy->cp_src), GFP_KERNEL);
1850 		if (!async_copy->cp_src)
1851 			goto out_err;
1852 		if (!nfs4_init_copy_state(nn, copy))
1853 			goto out_err;
1854 		memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.cs_stid,
1855 			sizeof(copy->cp_res.cb_stateid));
1856 		dup_copy_fields(copy, async_copy);
1857 		async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1858 				async_copy, "%s", "copy thread");
1859 		if (IS_ERR(async_copy->copy_task))
1860 			goto out_err;
1861 		spin_lock(&async_copy->cp_clp->async_lock);
1862 		list_add(&async_copy->copies,
1863 				&async_copy->cp_clp->async_copies);
1864 		spin_unlock(&async_copy->cp_clp->async_lock);
1865 		wake_up_process(async_copy->copy_task);
1866 		status = nfs_ok;
1867 	} else {
1868 		status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
1869 				       copy->nf_dst->nf_file, true);
1870 	}
1871 out:
1872 	trace_nfsd_copy_done(copy, status);
1873 	release_copy_files(copy);
1874 	return status;
1875 out_err:
1876 	if (nfsd4_ssc_is_inter(copy)) {
1877 		/*
1878 		 * Source's vfsmount of inter-copy will be unmounted
1879 		 * by the laundromat. Use copy instead of async_copy
1880 		 * since async_copy->ss_nsui might not be set yet.
1881 		 */
1882 		refcount_dec(&copy->ss_nsui->nsui_refcnt);
1883 	}
1884 	if (async_copy)
1885 		cleanup_async_copy(async_copy);
1886 	status = nfserrno(-ENOMEM);
1887 	goto out;
1888 }
1889 
1890 static struct nfsd4_copy *
1891 find_async_copy_locked(struct nfs4_client *clp, stateid_t *stateid)
1892 {
1893 	struct nfsd4_copy *copy;
1894 
1895 	lockdep_assert_held(&clp->async_lock);
1896 
1897 	list_for_each_entry(copy, &clp->async_copies, copies) {
1898 		if (memcmp(&copy->cp_stateid.cs_stid, stateid, NFS4_STATEID_SIZE))
1899 			continue;
1900 		return copy;
1901 	}
1902 	return NULL;
1903 }
1904 
1905 static struct nfsd4_copy *
1906 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1907 {
1908 	struct nfsd4_copy *copy;
1909 
1910 	spin_lock(&clp->async_lock);
1911 	copy = find_async_copy_locked(clp, stateid);
1912 	if (copy)
1913 		refcount_inc(&copy->refcount);
1914 	spin_unlock(&clp->async_lock);
1915 	return copy;
1916 }
1917 
1918 static __be32
1919 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1920 		     struct nfsd4_compound_state *cstate,
1921 		     union nfsd4_op_u *u)
1922 {
1923 	struct nfsd4_offload_status *os = &u->offload_status;
1924 	struct nfsd4_copy *copy;
1925 	struct nfs4_client *clp = cstate->clp;
1926 
1927 	copy = find_async_copy(clp, &os->stateid);
1928 	if (!copy) {
1929 		struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1930 
1931 		return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1932 	} else
1933 		nfsd4_stop_copy(copy);
1934 
1935 	return nfs_ok;
1936 }
1937 
1938 static __be32
1939 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1940 		  union nfsd4_op_u *u)
1941 {
1942 	struct nfsd4_copy_notify *cn = &u->copy_notify;
1943 	__be32 status;
1944 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1945 	struct nfs4_stid *stid;
1946 	struct nfs4_cpntf_state *cps;
1947 	struct nfs4_client *clp = cstate->clp;
1948 
1949 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1950 					&cn->cpn_src_stateid, RD_STATE, NULL,
1951 					&stid);
1952 	if (status)
1953 		return status;
1954 
1955 	cn->cpn_lease_time.tv_sec = nn->nfsd4_lease;
1956 	cn->cpn_lease_time.tv_nsec = 0;
1957 
1958 	status = nfserrno(-ENOMEM);
1959 	cps = nfs4_alloc_init_cpntf_state(nn, stid);
1960 	if (!cps)
1961 		goto out;
1962 	memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.cs_stid, sizeof(stateid_t));
1963 	memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1964 	memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1965 
1966 	/* For now, only return one server address in cpn_src, the
1967 	 * address used by the client to connect to this server.
1968 	 */
1969 	cn->cpn_src->nl4_type = NL4_NETADDR;
1970 	status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1971 				 &cn->cpn_src->u.nl4_addr);
1972 	WARN_ON_ONCE(status);
1973 	if (status) {
1974 		nfs4_put_cpntf_state(nn, cps);
1975 		goto out;
1976 	}
1977 out:
1978 	nfs4_put_stid(stid);
1979 	return status;
1980 }
1981 
1982 static __be32
1983 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1984 		struct nfsd4_fallocate *fallocate, int flags)
1985 {
1986 	__be32 status;
1987 	struct nfsd_file *nf;
1988 
1989 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1990 					    &fallocate->falloc_stateid,
1991 					    WR_STATE, &nf, NULL);
1992 	if (status != nfs_ok)
1993 		return status;
1994 
1995 	status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1996 				     fallocate->falloc_offset,
1997 				     fallocate->falloc_length,
1998 				     flags);
1999 	nfsd_file_put(nf);
2000 	return status;
2001 }
2002 
2003 static __be32
2004 nfsd4_offload_status(struct svc_rqst *rqstp,
2005 		     struct nfsd4_compound_state *cstate,
2006 		     union nfsd4_op_u *u)
2007 {
2008 	struct nfsd4_offload_status *os = &u->offload_status;
2009 	__be32 status = nfs_ok;
2010 	struct nfsd4_copy *copy;
2011 	struct nfs4_client *clp = cstate->clp;
2012 
2013 	os->completed = false;
2014 	spin_lock(&clp->async_lock);
2015 	copy = find_async_copy_locked(clp, &os->stateid);
2016 	if (copy) {
2017 		os->count = copy->cp_res.wr_bytes_written;
2018 		if (test_bit(NFSD4_COPY_F_COMPLETED, &copy->cp_flags)) {
2019 			os->completed = true;
2020 			os->status = copy->nfserr;
2021 		}
2022 	} else
2023 		status = nfserr_bad_stateid;
2024 	spin_unlock(&clp->async_lock);
2025 
2026 	return status;
2027 }
2028 
2029 static __be32
2030 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2031 	       union nfsd4_op_u *u)
2032 {
2033 	return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2034 }
2035 
2036 static __be32
2037 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2038 		 union nfsd4_op_u *u)
2039 {
2040 	return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2041 			       FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2042 }
2043 
2044 static __be32
2045 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2046 	   union nfsd4_op_u *u)
2047 {
2048 	struct nfsd4_seek *seek = &u->seek;
2049 	int whence;
2050 	__be32 status;
2051 	struct nfsd_file *nf;
2052 
2053 	status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2054 					    &seek->seek_stateid,
2055 					    RD_STATE, &nf, NULL);
2056 	if (status)
2057 		return status;
2058 
2059 	switch (seek->seek_whence) {
2060 	case NFS4_CONTENT_DATA:
2061 		whence = SEEK_DATA;
2062 		break;
2063 	case NFS4_CONTENT_HOLE:
2064 		whence = SEEK_HOLE;
2065 		break;
2066 	default:
2067 		status = nfserr_union_notsupp;
2068 		goto out;
2069 	}
2070 
2071 	/*
2072 	 * Note:  This call does change file->f_pos, but nothing in NFSD
2073 	 *        should ever file->f_pos.
2074 	 */
2075 	seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2076 	if (seek->seek_pos < 0)
2077 		status = nfserrno(seek->seek_pos);
2078 	else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2079 		seek->seek_eof = true;
2080 
2081 out:
2082 	nfsd_file_put(nf);
2083 	return status;
2084 }
2085 
2086 /* This routine never returns NFS_OK!  If there are no other errors, it
2087  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2088  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2089  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2090  */
2091 static __be32
2092 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2093 	     struct nfsd4_verify *verify)
2094 {
2095 	__be32 *buf, *p;
2096 	int count;
2097 	__be32 status;
2098 
2099 	status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2100 	if (status)
2101 		return status;
2102 
2103 	status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2104 	if (status)
2105 		return status;
2106 
2107 	if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2108 	    || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2109 		return nfserr_inval;
2110 	if (verify->ve_attrlen & 3)
2111 		return nfserr_inval;
2112 
2113 	/* count in words:
2114 	 *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2115 	 */
2116 	count = 4 + (verify->ve_attrlen >> 2);
2117 	buf = kmalloc(count << 2, GFP_KERNEL);
2118 	if (!buf)
2119 		return nfserr_jukebox;
2120 
2121 	p = buf;
2122 	status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2123 				    cstate->current_fh.fh_export,
2124 				    cstate->current_fh.fh_dentry,
2125 				    verify->ve_bmval,
2126 				    rqstp, 0);
2127 	/*
2128 	 * If nfsd4_encode_fattr() ran out of space, assume that's because
2129 	 * the attributes are longer (hence different) than those given:
2130 	 */
2131 	if (status == nfserr_resource)
2132 		status = nfserr_not_same;
2133 	if (status)
2134 		goto out_kfree;
2135 
2136 	/* skip bitmap */
2137 	p = buf + 1 + ntohl(buf[0]);
2138 	status = nfserr_not_same;
2139 	if (ntohl(*p++) != verify->ve_attrlen)
2140 		goto out_kfree;
2141 	if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2142 		status = nfserr_same;
2143 
2144 out_kfree:
2145 	kfree(buf);
2146 	return status;
2147 }
2148 
2149 static __be32
2150 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2151 	      union nfsd4_op_u *u)
2152 {
2153 	__be32 status;
2154 
2155 	status = _nfsd4_verify(rqstp, cstate, &u->verify);
2156 	return status == nfserr_not_same ? nfs_ok : status;
2157 }
2158 
2159 static __be32
2160 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2161 	     union nfsd4_op_u *u)
2162 {
2163 	__be32 status;
2164 
2165 	status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2166 	return status == nfserr_same ? nfs_ok : status;
2167 }
2168 
2169 static __be32
2170 nfsd4_get_dir_delegation(struct svc_rqst *rqstp,
2171 			 struct nfsd4_compound_state *cstate,
2172 			 union nfsd4_op_u *u)
2173 {
2174 	struct nfsd4_get_dir_delegation *gdd = &u->get_dir_delegation;
2175 
2176 	/*
2177 	 * RFC 8881, section 18.39.3 says:
2178 	 *
2179 	 * "The server may refuse to grant the delegation. In that case, the
2180 	 *  server will return NFS4ERR_DIRDELEG_UNAVAIL."
2181 	 *
2182 	 * This is sub-optimal, since it means that the server would need to
2183 	 * abort compound processing just because the delegation wasn't
2184 	 * available. RFC8881bis should change this to allow the server to
2185 	 * return NFS4_OK with a non-fatal status of GDD4_UNAVAIL in this
2186 	 * situation.
2187 	 */
2188 	gdd->gddrnf_status = GDD4_UNAVAIL;
2189 	return nfs_ok;
2190 }
2191 
2192 #ifdef CONFIG_NFSD_PNFS
2193 static const struct nfsd4_layout_ops *
2194 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2195 {
2196 	if (!exp->ex_layout_types) {
2197 		dprintk("%s: export does not support pNFS\n", __func__);
2198 		return NULL;
2199 	}
2200 
2201 	if (layout_type >= LAYOUT_TYPE_MAX ||
2202 	    !(exp->ex_layout_types & (1 << layout_type))) {
2203 		dprintk("%s: layout type %d not supported\n",
2204 			__func__, layout_type);
2205 		return NULL;
2206 	}
2207 
2208 	return nfsd4_layout_ops[layout_type];
2209 }
2210 
2211 static __be32
2212 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2213 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2214 {
2215 	struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2216 	const struct nfsd4_layout_ops *ops;
2217 	struct nfsd4_deviceid_map *map;
2218 	struct svc_export *exp;
2219 	__be32 nfserr;
2220 
2221 	dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2222 	       __func__,
2223 	       gdp->gd_layout_type,
2224 	       gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2225 	       gdp->gd_maxcount);
2226 
2227 	map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2228 	if (!map) {
2229 		dprintk("%s: couldn't find device ID to export mapping!\n",
2230 			__func__);
2231 		return nfserr_noent;
2232 	}
2233 
2234 	exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2235 	if (IS_ERR(exp)) {
2236 		dprintk("%s: could not find device id\n", __func__);
2237 		return nfserr_noent;
2238 	}
2239 
2240 	nfserr = nfserr_layoutunavailable;
2241 	ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2242 	if (!ops)
2243 		goto out;
2244 
2245 	nfserr = nfs_ok;
2246 	if (gdp->gd_maxcount != 0) {
2247 		nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2248 				rqstp, cstate->clp, gdp);
2249 	}
2250 
2251 	gdp->gd_notify_types &= ops->notify_types;
2252 out:
2253 	exp_put(exp);
2254 	return nfserr;
2255 }
2256 
2257 static void
2258 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2259 {
2260 	kfree(u->getdeviceinfo.gd_device);
2261 }
2262 
2263 static __be32
2264 nfsd4_layoutget(struct svc_rqst *rqstp,
2265 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2266 {
2267 	struct nfsd4_layoutget *lgp = &u->layoutget;
2268 	struct svc_fh *current_fh = &cstate->current_fh;
2269 	const struct nfsd4_layout_ops *ops;
2270 	struct nfs4_layout_stateid *ls;
2271 	__be32 nfserr;
2272 	int accmode = NFSD_MAY_READ_IF_EXEC;
2273 
2274 	switch (lgp->lg_seg.iomode) {
2275 	case IOMODE_READ:
2276 		accmode |= NFSD_MAY_READ;
2277 		break;
2278 	case IOMODE_RW:
2279 		accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2280 		break;
2281 	default:
2282 		dprintk("%s: invalid iomode %d\n",
2283 			__func__, lgp->lg_seg.iomode);
2284 		nfserr = nfserr_badiomode;
2285 		goto out;
2286 	}
2287 
2288 	nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2289 	if (nfserr)
2290 		goto out;
2291 
2292 	nfserr = nfserr_layoutunavailable;
2293 	ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2294 	if (!ops)
2295 		goto out;
2296 
2297 	/*
2298 	 * Verify minlength and range as per RFC5661:
2299 	 *  o  If loga_length is less than loga_minlength,
2300 	 *     the metadata server MUST return NFS4ERR_INVAL.
2301 	 *  o  If the sum of loga_offset and loga_minlength exceeds
2302 	 *     NFS4_UINT64_MAX, and loga_minlength is not
2303 	 *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2304 	 *  o  If the sum of loga_offset and loga_length exceeds
2305 	 *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2306 	 *     the error NFS4ERR_INVAL MUST result.
2307 	 */
2308 	nfserr = nfserr_inval;
2309 	if (lgp->lg_seg.length < lgp->lg_minlength ||
2310 	    (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2311 	     lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2312 	    (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2313 	     lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2314 		goto out;
2315 	if (lgp->lg_seg.length == 0)
2316 		goto out;
2317 
2318 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2319 						true, lgp->lg_layout_type, &ls);
2320 	if (nfserr) {
2321 		trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2322 		goto out;
2323 	}
2324 
2325 	nfserr = nfserr_recallconflict;
2326 	if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2327 		goto out_put_stid;
2328 
2329 	nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2330 				     current_fh, lgp);
2331 	if (nfserr)
2332 		goto out_put_stid;
2333 
2334 	nfserr = nfsd4_insert_layout(lgp, ls);
2335 
2336 out_put_stid:
2337 	mutex_unlock(&ls->ls_mutex);
2338 	nfs4_put_stid(&ls->ls_stid);
2339 out:
2340 	return nfserr;
2341 }
2342 
2343 static void
2344 nfsd4_layoutget_release(union nfsd4_op_u *u)
2345 {
2346 	kfree(u->layoutget.lg_content);
2347 }
2348 
2349 static __be32
2350 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2351 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2352 {
2353 	struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2354 	const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2355 	struct svc_fh *current_fh = &cstate->current_fh;
2356 	const struct nfsd4_layout_ops *ops;
2357 	loff_t new_size = lcp->lc_last_wr + 1;
2358 	struct inode *inode;
2359 	struct nfs4_layout_stateid *ls;
2360 	__be32 nfserr;
2361 
2362 	nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2363 	if (nfserr)
2364 		goto out;
2365 
2366 	nfserr = nfserr_layoutunavailable;
2367 	ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2368 	if (!ops)
2369 		goto out;
2370 	inode = d_inode(current_fh->fh_dentry);
2371 
2372 	nfserr = nfserr_inval;
2373 	if (new_size <= seg->offset) {
2374 		dprintk("pnfsd: last write before layout segment\n");
2375 		goto out;
2376 	}
2377 	if (new_size > seg->offset + seg->length) {
2378 		dprintk("pnfsd: last write beyond layout segment\n");
2379 		goto out;
2380 	}
2381 	if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2382 		dprintk("pnfsd: layoutcommit beyond EOF\n");
2383 		goto out;
2384 	}
2385 
2386 	nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2387 						false, lcp->lc_layout_type,
2388 						&ls);
2389 	if (nfserr) {
2390 		trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2391 		/* fixup error code as per RFC5661 */
2392 		if (nfserr == nfserr_bad_stateid)
2393 			nfserr = nfserr_badlayout;
2394 		goto out;
2395 	}
2396 
2397 	/* LAYOUTCOMMIT does not require any serialization */
2398 	mutex_unlock(&ls->ls_mutex);
2399 
2400 	if (new_size > i_size_read(inode)) {
2401 		lcp->lc_size_chg = true;
2402 		lcp->lc_newsize = new_size;
2403 	} else {
2404 		lcp->lc_size_chg = false;
2405 	}
2406 
2407 	nfserr = ops->proc_layoutcommit(inode, lcp);
2408 	nfs4_put_stid(&ls->ls_stid);
2409 out:
2410 	return nfserr;
2411 }
2412 
2413 static __be32
2414 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2415 		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2416 {
2417 	struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2418 	struct svc_fh *current_fh = &cstate->current_fh;
2419 	__be32 nfserr;
2420 
2421 	nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2422 	if (nfserr)
2423 		goto out;
2424 
2425 	nfserr = nfserr_layoutunavailable;
2426 	if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2427 		goto out;
2428 
2429 	switch (lrp->lr_seg.iomode) {
2430 	case IOMODE_READ:
2431 	case IOMODE_RW:
2432 	case IOMODE_ANY:
2433 		break;
2434 	default:
2435 		dprintk("%s: invalid iomode %d\n", __func__,
2436 			lrp->lr_seg.iomode);
2437 		nfserr = nfserr_inval;
2438 		goto out;
2439 	}
2440 
2441 	switch (lrp->lr_return_type) {
2442 	case RETURN_FILE:
2443 		nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2444 		break;
2445 	case RETURN_FSID:
2446 	case RETURN_ALL:
2447 		nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2448 		break;
2449 	default:
2450 		dprintk("%s: invalid return_type %d\n", __func__,
2451 			lrp->lr_return_type);
2452 		nfserr = nfserr_inval;
2453 		break;
2454 	}
2455 out:
2456 	return nfserr;
2457 }
2458 #endif /* CONFIG_NFSD_PNFS */
2459 
2460 static __be32
2461 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2462 	       union nfsd4_op_u *u)
2463 {
2464 	struct nfsd4_getxattr *getxattr = &u->getxattr;
2465 
2466 	return nfsd_getxattr(rqstp, &cstate->current_fh,
2467 			     getxattr->getxa_name, &getxattr->getxa_buf,
2468 			     &getxattr->getxa_len);
2469 }
2470 
2471 static __be32
2472 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2473 	   union nfsd4_op_u *u)
2474 {
2475 	struct nfsd4_setxattr *setxattr = &u->setxattr;
2476 	__be32 ret;
2477 
2478 	if (opens_in_grace(SVC_NET(rqstp)))
2479 		return nfserr_grace;
2480 
2481 	ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2482 			    setxattr->setxa_buf, setxattr->setxa_len,
2483 			    setxattr->setxa_flags);
2484 
2485 	if (!ret)
2486 		set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2487 
2488 	return ret;
2489 }
2490 
2491 static __be32
2492 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2493 	   union nfsd4_op_u *u)
2494 {
2495 	/*
2496 	 * Get the entire list, then copy out only the user attributes
2497 	 * in the encode function.
2498 	 */
2499 	return nfsd_listxattr(rqstp, &cstate->current_fh,
2500 			     &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2501 }
2502 
2503 static __be32
2504 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2505 	   union nfsd4_op_u *u)
2506 {
2507 	struct nfsd4_removexattr *removexattr = &u->removexattr;
2508 	__be32 ret;
2509 
2510 	if (opens_in_grace(SVC_NET(rqstp)))
2511 		return nfserr_grace;
2512 
2513 	ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2514 	    removexattr->rmxa_name);
2515 
2516 	if (!ret)
2517 		set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2518 
2519 	return ret;
2520 }
2521 
2522 /*
2523  * NULL call.
2524  */
2525 static __be32
2526 nfsd4_proc_null(struct svc_rqst *rqstp)
2527 {
2528 	return rpc_success;
2529 }
2530 
2531 static inline void nfsd4_increment_op_stats(struct nfsd_net *nn, u32 opnum)
2532 {
2533 	if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2534 		percpu_counter_inc(&nn->counter[NFSD_STATS_NFS4_OP(opnum)]);
2535 }
2536 
2537 static const struct nfsd4_operation nfsd4_ops[];
2538 
2539 static const char *nfsd4_op_name(unsigned opnum);
2540 
2541 /*
2542  * Enforce NFSv4.1 COMPOUND ordering rules:
2543  *
2544  * Also note, enforced elsewhere:
2545  *	- SEQUENCE other than as first op results in
2546  *	  NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2547  *	- BIND_CONN_TO_SESSION must be the only op in its compound.
2548  *	  (Enforced in nfsd4_bind_conn_to_session().)
2549  *	- DESTROY_SESSION must be the final operation in a compound, if
2550  *	  sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2551  *	  (Enforced in nfsd4_destroy_session().)
2552  */
2553 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2554 {
2555 	struct nfsd4_op *first_op = &args->ops[0];
2556 
2557 	/* These ordering requirements don't apply to NFSv4.0: */
2558 	if (args->minorversion == 0)
2559 		return nfs_ok;
2560 	/* This is weird, but OK, not our problem: */
2561 	if (args->opcnt == 0)
2562 		return nfs_ok;
2563 	if (first_op->status == nfserr_op_illegal)
2564 		return nfs_ok;
2565 	if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2566 		return nfserr_op_not_in_session;
2567 	if (first_op->opnum == OP_SEQUENCE)
2568 		return nfs_ok;
2569 	/*
2570 	 * So first_op is something allowed outside a session, like
2571 	 * EXCHANGE_ID; but then it has to be the only op in the
2572 	 * compound:
2573 	 */
2574 	if (args->opcnt != 1)
2575 		return nfserr_not_only_op;
2576 	return nfs_ok;
2577 }
2578 
2579 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2580 {
2581 	return &nfsd4_ops[op->opnum];
2582 }
2583 
2584 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2585 {
2586 	if (op->opnum == OP_ILLEGAL)
2587 		return false;
2588 	return OPDESC(op)->op_flags & OP_CACHEME;
2589 }
2590 
2591 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2592 {
2593 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2594 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2595 	struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2596 	struct nfsd4_op *next = &argp->ops[resp->opcnt];
2597 	const struct nfsd4_operation *thisd = OPDESC(this);
2598 	const struct nfsd4_operation *nextd;
2599 
2600 	/*
2601 	 * Most ops check wronsec on our own; only the putfh-like ops
2602 	 * have special rules.
2603 	 */
2604 	if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2605 		return false;
2606 	/*
2607 	 * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2608 	 * put-filehandle operation if we're not going to use the
2609 	 * result:
2610 	 */
2611 	if (argp->opcnt == resp->opcnt)
2612 		return false;
2613 	if (next->opnum == OP_ILLEGAL)
2614 		return false;
2615 	nextd = OPDESC(next);
2616 	/*
2617 	 * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2618 	 * errors themselves as necessary; others should check for them
2619 	 * now:
2620 	 */
2621 	return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2622 }
2623 
2624 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2625 static void
2626 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2627 {
2628 	struct nfsd4_op	*op, *current_op = NULL, *saved_op = NULL;
2629 	struct nfsd4_copy *copy;
2630 	struct nfsd4_putfh *putfh;
2631 	int i;
2632 
2633 	/* traverse all operation and if it's a COPY compound, mark the
2634 	 * source filehandle to skip verification
2635 	 */
2636 	for (i = 0; i < args->opcnt; i++) {
2637 		op = &args->ops[i];
2638 		if (op->opnum == OP_PUTFH)
2639 			current_op = op;
2640 		else if (op->opnum == OP_SAVEFH)
2641 			saved_op = current_op;
2642 		else if (op->opnum == OP_RESTOREFH)
2643 			current_op = saved_op;
2644 		else if (op->opnum == OP_COPY) {
2645 			copy = (struct nfsd4_copy *)&op->u;
2646 			if (!saved_op) {
2647 				op->status = nfserr_nofilehandle;
2648 				return;
2649 			}
2650 			putfh = (struct nfsd4_putfh *)&saved_op->u;
2651 			if (nfsd4_ssc_is_inter(copy))
2652 				putfh->no_verify = true;
2653 		}
2654 	}
2655 }
2656 #else
2657 static void
2658 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2659 {
2660 }
2661 #endif
2662 
2663 /*
2664  * COMPOUND call.
2665  */
2666 static __be32
2667 nfsd4_proc_compound(struct svc_rqst *rqstp)
2668 {
2669 	struct nfsd4_compoundargs *args = rqstp->rq_argp;
2670 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
2671 	struct nfsd4_op	*op;
2672 	struct nfsd4_compound_state *cstate = &resp->cstate;
2673 	struct svc_fh *current_fh = &cstate->current_fh;
2674 	struct svc_fh *save_fh = &cstate->save_fh;
2675 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2676 	__be32		status;
2677 
2678 	resp->xdr = &rqstp->rq_res_stream;
2679 	resp->statusp = resp->xdr->p;
2680 
2681 	/* reserve space for: NFS status code */
2682 	xdr_reserve_space(resp->xdr, XDR_UNIT);
2683 
2684 	/* reserve space for: taglen, tag, and opcnt */
2685 	xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2686 	resp->taglen = args->taglen;
2687 	resp->tag = args->tag;
2688 	resp->rqstp = rqstp;
2689 	cstate->minorversion = args->minorversion;
2690 	fh_init(current_fh, NFS4_FHSIZE);
2691 	fh_init(save_fh, NFS4_FHSIZE);
2692 	/*
2693 	 * Don't use the deferral mechanism for NFSv4; compounds make it
2694 	 * too hard to avoid non-idempotency problems.
2695 	 */
2696 	clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2697 
2698 	/*
2699 	 * According to RFC3010, this takes precedence over all other errors.
2700 	 */
2701 	status = nfserr_minor_vers_mismatch;
2702 	if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2703 		goto out;
2704 
2705 	status = nfs41_check_op_ordering(args);
2706 	if (status) {
2707 		op = &args->ops[0];
2708 		op->status = status;
2709 		resp->opcnt = 1;
2710 		goto encode_op;
2711 	}
2712 	check_if_stalefh_allowed(args);
2713 
2714 	rqstp->rq_lease_breaker = (void **)&cstate->clp;
2715 
2716 	trace_nfsd_compound(rqstp, args->tag, args->taglen, args->client_opcnt);
2717 	while (!status && resp->opcnt < args->opcnt) {
2718 		op = &args->ops[resp->opcnt++];
2719 
2720 		if (unlikely(resp->opcnt == NFSD_MAX_OPS_PER_COMPOUND)) {
2721 			/* If there are still more operations to process,
2722 			 * stop here and report NFS4ERR_RESOURCE. */
2723 			if (cstate->minorversion == 0 &&
2724 			    args->client_opcnt > resp->opcnt) {
2725 				op->status = nfserr_resource;
2726 				goto encode_op;
2727 			}
2728 		}
2729 
2730 		/*
2731 		 * The XDR decode routines may have pre-set op->status;
2732 		 * for example, if there is a miscellaneous XDR error
2733 		 * it will be set to nfserr_bad_xdr.
2734 		 */
2735 		if (op->status) {
2736 			if (op->opnum == OP_OPEN)
2737 				op->status = nfsd4_open_omfg(rqstp, cstate, op);
2738 			goto encode_op;
2739 		}
2740 		if (!current_fh->fh_dentry &&
2741 				!HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2742 			if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2743 				op->status = nfserr_nofilehandle;
2744 				goto encode_op;
2745 			}
2746 		} else if (current_fh->fh_export &&
2747 			   current_fh->fh_export->ex_fslocs.migrated &&
2748 			  !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2749 			op->status = nfserr_moved;
2750 			goto encode_op;
2751 		}
2752 
2753 		fh_clear_pre_post_attrs(current_fh);
2754 
2755 		/* If op is non-idempotent */
2756 		if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2757 			/*
2758 			 * Don't execute this op if we couldn't encode a
2759 			 * successful reply:
2760 			 */
2761 			u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2762 			/*
2763 			 * Plus if there's another operation, make sure
2764 			 * we'll have space to at least encode an error:
2765 			 */
2766 			if (resp->opcnt < args->opcnt)
2767 				plen += COMPOUND_ERR_SLACK_SPACE;
2768 			op->status = nfsd4_check_resp_size(resp, plen);
2769 		}
2770 
2771 		if (op->status)
2772 			goto encode_op;
2773 
2774 		if (op->opdesc->op_get_currentstateid)
2775 			op->opdesc->op_get_currentstateid(cstate, &op->u);
2776 		op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2777 
2778 		/* Only from SEQUENCE */
2779 		if (cstate->status == nfserr_replay_cache) {
2780 			dprintk("%s NFS4.1 replay from cache\n", __func__);
2781 			status = op->status;
2782 			goto out;
2783 		}
2784 		if (!op->status) {
2785 			if (op->opdesc->op_set_currentstateid)
2786 				op->opdesc->op_set_currentstateid(cstate, &op->u);
2787 
2788 			if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2789 				clear_current_stateid(cstate);
2790 
2791 			if (current_fh->fh_export &&
2792 					need_wrongsec_check(rqstp))
2793 				op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2794 		}
2795 encode_op:
2796 		if (op->status == nfserr_replay_me) {
2797 			op->replay = &cstate->replay_owner->so_replay;
2798 			nfsd4_encode_replay(resp->xdr, op);
2799 			status = op->status = op->replay->rp_status;
2800 		} else {
2801 			nfsd4_encode_operation(resp, op);
2802 			status = op->status;
2803 		}
2804 
2805 		trace_nfsd_compound_status(args->client_opcnt, resp->opcnt,
2806 					   status, nfsd4_op_name(op->opnum));
2807 
2808 		nfsd4_cstate_clear_replay(cstate);
2809 		nfsd4_increment_op_stats(nn, op->opnum);
2810 	}
2811 
2812 	fh_put(current_fh);
2813 	fh_put(save_fh);
2814 	BUG_ON(cstate->replay_owner);
2815 out:
2816 	cstate->status = status;
2817 	/* Reset deferral mechanism for RPC deferrals */
2818 	set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2819 	return rpc_success;
2820 }
2821 
2822 #define op_encode_hdr_size		(2)
2823 #define op_encode_stateid_maxsz		(XDR_QUADLEN(NFS4_STATEID_SIZE))
2824 #define op_encode_verifier_maxsz	(XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2825 #define op_encode_change_info_maxsz	(5)
2826 #define nfs4_fattr_bitmap_maxsz		(4)
2827 
2828 /* We'll fall back on returning no lockowner if run out of space: */
2829 #define op_encode_lockowner_maxsz	(0)
2830 #define op_encode_lock_denied_maxsz	(8 + op_encode_lockowner_maxsz)
2831 
2832 #define nfs4_owner_maxsz		(1 + XDR_QUADLEN(IDMAP_NAMESZ))
2833 
2834 #define op_encode_ace_maxsz		(3 + nfs4_owner_maxsz)
2835 #define op_encode_delegation_maxsz	(1 + op_encode_stateid_maxsz + 1 + \
2836 					 op_encode_ace_maxsz)
2837 
2838 #define op_encode_channel_attrs_maxsz	(6 + 1 + 1)
2839 
2840 /*
2841  * The _rsize() helpers are invoked by the NFSv4 COMPOUND decoder, which
2842  * is called before sunrpc sets rq_res.buflen. Thus we have to compute
2843  * the maximum payload size here, based on transport limits and the size
2844  * of the remaining space in the rq_pages array.
2845  */
2846 static u32 nfsd4_max_payload(const struct svc_rqst *rqstp)
2847 {
2848 	u32 buflen;
2849 
2850 	buflen = (rqstp->rq_page_end - rqstp->rq_next_page) * PAGE_SIZE;
2851 	buflen -= rqstp->rq_auth_slack;
2852 	buflen -= rqstp->rq_res.head[0].iov_len;
2853 	return min_t(u32, buflen, svc_max_payload(rqstp));
2854 }
2855 
2856 static u32 nfsd4_only_status_rsize(const struct svc_rqst *rqstp,
2857 				   const struct nfsd4_op *op)
2858 {
2859 	return (op_encode_hdr_size) * sizeof(__be32);
2860 }
2861 
2862 static u32 nfsd4_status_stateid_rsize(const struct svc_rqst *rqstp,
2863 				      const struct nfsd4_op *op)
2864 {
2865 	return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2866 }
2867 
2868 static u32 nfsd4_access_rsize(const struct svc_rqst *rqstp,
2869 			      const struct nfsd4_op *op)
2870 {
2871 	/* ac_supported, ac_resp_access */
2872 	return (op_encode_hdr_size + 2)* sizeof(__be32);
2873 }
2874 
2875 static u32 nfsd4_commit_rsize(const struct svc_rqst *rqstp,
2876 			      const struct nfsd4_op *op)
2877 {
2878 	return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2879 }
2880 
2881 static u32 nfsd4_create_rsize(const struct svc_rqst *rqstp,
2882 			      const struct nfsd4_op *op)
2883 {
2884 	return (op_encode_hdr_size + op_encode_change_info_maxsz
2885 		+ nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2886 }
2887 
2888 /*
2889  * Note since this is an idempotent operation we won't insist on failing
2890  * the op prematurely if the estimate is too large.  We may turn off splice
2891  * reads unnecessarily.
2892  */
2893 static u32 nfsd4_getattr_rsize(const struct svc_rqst *rqstp,
2894 			       const struct nfsd4_op *op)
2895 {
2896 	const u32 *bmap = op->u.getattr.ga_bmval;
2897 	u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2898 	u32 ret = 0;
2899 
2900 	if (bmap0 & FATTR4_WORD0_ACL)
2901 		return nfsd4_max_payload(rqstp);
2902 	if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2903 		return nfsd4_max_payload(rqstp);
2904 
2905 	if (bmap1 & FATTR4_WORD1_OWNER) {
2906 		ret += IDMAP_NAMESZ + 4;
2907 		bmap1 &= ~FATTR4_WORD1_OWNER;
2908 	}
2909 	if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2910 		ret += IDMAP_NAMESZ + 4;
2911 		bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2912 	}
2913 	if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2914 		ret += NFS4_FHSIZE + 4;
2915 		bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2916 	}
2917 	if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2918 		ret += NFS4_MAXLABELLEN + 12;
2919 		bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2920 	}
2921 	/*
2922 	 * Largest of remaining attributes are 16 bytes (e.g.,
2923 	 * supported_attributes)
2924 	 */
2925 	ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2926 	/* bitmask, length */
2927 	ret += 20;
2928 	return ret;
2929 }
2930 
2931 static u32 nfsd4_getfh_rsize(const struct svc_rqst *rqstp,
2932 			     const struct nfsd4_op *op)
2933 {
2934 	return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2935 }
2936 
2937 static u32 nfsd4_link_rsize(const struct svc_rqst *rqstp,
2938 			    const struct nfsd4_op *op)
2939 {
2940 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
2941 		* sizeof(__be32);
2942 }
2943 
2944 static u32 nfsd4_lock_rsize(const struct svc_rqst *rqstp,
2945 			    const struct nfsd4_op *op)
2946 {
2947 	return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2948 		* sizeof(__be32);
2949 }
2950 
2951 static u32 nfsd4_open_rsize(const struct svc_rqst *rqstp,
2952 			    const struct nfsd4_op *op)
2953 {
2954 	return (op_encode_hdr_size + op_encode_stateid_maxsz
2955 		+ op_encode_change_info_maxsz + 1
2956 		+ nfs4_fattr_bitmap_maxsz
2957 		+ op_encode_delegation_maxsz) * sizeof(__be32);
2958 }
2959 
2960 static u32 nfsd4_read_rsize(const struct svc_rqst *rqstp,
2961 			    const struct nfsd4_op *op)
2962 {
2963 	u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2964 
2965 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2966 }
2967 
2968 static u32 nfsd4_read_plus_rsize(const struct svc_rqst *rqstp,
2969 				 const struct nfsd4_op *op)
2970 {
2971 	u32 rlen = min(op->u.read.rd_length, nfsd4_max_payload(rqstp));
2972 	/*
2973 	 * If we detect that the file changed during hole encoding, then we
2974 	 * recover by encoding the remaining reply as data. This means we need
2975 	 * to set aside enough room to encode two data segments.
2976 	 */
2977 	u32 seg_len = 2 * (1 + 2 + 1);
2978 
2979 	return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2980 }
2981 
2982 static u32 nfsd4_readdir_rsize(const struct svc_rqst *rqstp,
2983 			       const struct nfsd4_op *op)
2984 {
2985 	u32 rlen = min(op->u.readdir.rd_maxcount, nfsd4_max_payload(rqstp));
2986 
2987 	return (op_encode_hdr_size + op_encode_verifier_maxsz +
2988 		XDR_QUADLEN(rlen)) * sizeof(__be32);
2989 }
2990 
2991 static u32 nfsd4_readlink_rsize(const struct svc_rqst *rqstp,
2992 				const struct nfsd4_op *op)
2993 {
2994 	return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2995 }
2996 
2997 static u32 nfsd4_remove_rsize(const struct svc_rqst *rqstp,
2998 			      const struct nfsd4_op *op)
2999 {
3000 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3001 		* sizeof(__be32);
3002 }
3003 
3004 static u32 nfsd4_rename_rsize(const struct svc_rqst *rqstp,
3005 			      const struct nfsd4_op *op)
3006 {
3007 	return (op_encode_hdr_size + op_encode_change_info_maxsz
3008 		+ op_encode_change_info_maxsz) * sizeof(__be32);
3009 }
3010 
3011 static u32 nfsd4_sequence_rsize(const struct svc_rqst *rqstp,
3012 				const struct nfsd4_op *op)
3013 {
3014 	return (op_encode_hdr_size
3015 		+ XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
3016 }
3017 
3018 static u32 nfsd4_test_stateid_rsize(const struct svc_rqst *rqstp,
3019 				    const struct nfsd4_op *op)
3020 {
3021 	return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
3022 		* sizeof(__be32);
3023 }
3024 
3025 static u32 nfsd4_setattr_rsize(const struct svc_rqst *rqstp,
3026 			       const struct nfsd4_op *op)
3027 {
3028 	return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
3029 }
3030 
3031 static u32 nfsd4_secinfo_rsize(const struct svc_rqst *rqstp,
3032 			       const struct nfsd4_op *op)
3033 {
3034 	return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
3035 		(4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
3036 }
3037 
3038 static u32 nfsd4_setclientid_rsize(const struct svc_rqst *rqstp,
3039 				   const struct nfsd4_op *op)
3040 {
3041 	return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
3042 								sizeof(__be32);
3043 }
3044 
3045 static u32 nfsd4_write_rsize(const struct svc_rqst *rqstp,
3046 			     const struct nfsd4_op *op)
3047 {
3048 	return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
3049 }
3050 
3051 static u32 nfsd4_exchange_id_rsize(const struct svc_rqst *rqstp,
3052 				   const struct nfsd4_op *op)
3053 {
3054 	return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
3055 		1 + 1 + /* eir_flags, spr_how */\
3056 		4 + /* spo_must_enforce & _allow with bitmap */\
3057 		2 + /*eir_server_owner.so_minor_id */\
3058 		/* eir_server_owner.so_major_id<> */\
3059 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3060 		/* eir_server_scope<> */\
3061 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
3062 		1 + /* eir_server_impl_id array length */\
3063 		0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
3064 }
3065 
3066 static u32 nfsd4_bind_conn_to_session_rsize(const struct svc_rqst *rqstp,
3067 					    const struct nfsd4_op *op)
3068 {
3069 	return (op_encode_hdr_size + \
3070 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
3071 		2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
3072 }
3073 
3074 static u32 nfsd4_create_session_rsize(const struct svc_rqst *rqstp,
3075 				      const struct nfsd4_op *op)
3076 {
3077 	return (op_encode_hdr_size + \
3078 		XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3079 		2 + /* csr_sequence, csr_flags */\
3080 		op_encode_channel_attrs_maxsz + \
3081 		op_encode_channel_attrs_maxsz) * sizeof(__be32);
3082 }
3083 
3084 static u32 nfsd4_copy_rsize(const struct svc_rqst *rqstp,
3085 			    const struct nfsd4_op *op)
3086 {
3087 	return (op_encode_hdr_size +
3088 		1 /* wr_callback */ +
3089 		op_encode_stateid_maxsz /* wr_callback */ +
3090 		2 /* wr_count */ +
3091 		1 /* wr_committed */ +
3092 		op_encode_verifier_maxsz +
3093 		1 /* cr_consecutive */ +
3094 		1 /* cr_synchronous */) * sizeof(__be32);
3095 }
3096 
3097 static u32 nfsd4_offload_status_rsize(const struct svc_rqst *rqstp,
3098 				      const struct nfsd4_op *op)
3099 {
3100 	return (op_encode_hdr_size +
3101 		2 /* osr_count */ +
3102 		1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3103 }
3104 
3105 static u32 nfsd4_copy_notify_rsize(const struct svc_rqst *rqstp,
3106 				   const struct nfsd4_op *op)
3107 {
3108 	return (op_encode_hdr_size +
3109 		3 /* cnr_lease_time */ +
3110 		1 /* We support one cnr_source_server */ +
3111 		1 /* cnr_stateid seq */ +
3112 		op_encode_stateid_maxsz /* cnr_stateid */ +
3113 		1 /* num cnr_source_server*/ +
3114 		1 /* nl4_type */ +
3115 		1 /* nl4 size */ +
3116 		XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3117 		* sizeof(__be32);
3118 }
3119 
3120 static u32 nfsd4_get_dir_delegation_rsize(const struct svc_rqst *rqstp,
3121 					  const struct nfsd4_op *op)
3122 {
3123 	return (op_encode_hdr_size +
3124 		1 /* gddr_status */ +
3125 		op_encode_verifier_maxsz +
3126 		op_encode_stateid_maxsz +
3127 		2 /* gddr_notification */ +
3128 		2 /* gddr_child_attributes */ +
3129 		2 /* gddr_dir_attributes */);
3130 }
3131 
3132 #ifdef CONFIG_NFSD_PNFS
3133 static u32 nfsd4_getdeviceinfo_rsize(const struct svc_rqst *rqstp,
3134 				     const struct nfsd4_op *op)
3135 {
3136 	u32 rlen = min(op->u.getdeviceinfo.gd_maxcount, nfsd4_max_payload(rqstp));
3137 
3138 	return (op_encode_hdr_size +
3139 		1 /* gd_layout_type*/ +
3140 		XDR_QUADLEN(rlen) +
3141 		2 /* gd_notify_types */) * sizeof(__be32);
3142 }
3143 
3144 /*
3145  * At this stage we don't really know what layout driver will handle the request,
3146  * so we need to define an arbitrary upper bound here.
3147  */
3148 #define MAX_LAYOUT_SIZE		128
3149 static u32 nfsd4_layoutget_rsize(const struct svc_rqst *rqstp,
3150 				 const struct nfsd4_op *op)
3151 {
3152 	return (op_encode_hdr_size +
3153 		1 /* logr_return_on_close */ +
3154 		op_encode_stateid_maxsz +
3155 		1 /* nr of layouts */ +
3156 		MAX_LAYOUT_SIZE) * sizeof(__be32);
3157 }
3158 
3159 static u32 nfsd4_layoutcommit_rsize(const struct svc_rqst *rqstp,
3160 				    const struct nfsd4_op *op)
3161 {
3162 	return (op_encode_hdr_size +
3163 		1 /* locr_newsize */ +
3164 		2 /* ns_size */) * sizeof(__be32);
3165 }
3166 
3167 static u32 nfsd4_layoutreturn_rsize(const struct svc_rqst *rqstp,
3168 				    const struct nfsd4_op *op)
3169 {
3170 	return (op_encode_hdr_size +
3171 		1 /* lrs_stateid */ +
3172 		op_encode_stateid_maxsz) * sizeof(__be32);
3173 }
3174 #endif /* CONFIG_NFSD_PNFS */
3175 
3176 
3177 static u32 nfsd4_seek_rsize(const struct svc_rqst *rqstp,
3178 			    const struct nfsd4_op *op)
3179 {
3180 	return (op_encode_hdr_size + 3) * sizeof(__be32);
3181 }
3182 
3183 static u32 nfsd4_getxattr_rsize(const struct svc_rqst *rqstp,
3184 				const struct nfsd4_op *op)
3185 {
3186 	u32 rlen = min_t(u32, XATTR_SIZE_MAX, nfsd4_max_payload(rqstp));
3187 
3188 	return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3189 }
3190 
3191 static u32 nfsd4_setxattr_rsize(const struct svc_rqst *rqstp,
3192 				const struct nfsd4_op *op)
3193 {
3194 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3195 		* sizeof(__be32);
3196 }
3197 static u32 nfsd4_listxattrs_rsize(const struct svc_rqst *rqstp,
3198 				  const struct nfsd4_op *op)
3199 {
3200 	u32 rlen = min(op->u.listxattrs.lsxa_maxcount, nfsd4_max_payload(rqstp));
3201 
3202 	return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3203 }
3204 
3205 static u32 nfsd4_removexattr_rsize(const struct svc_rqst *rqstp,
3206 				   const struct nfsd4_op *op)
3207 {
3208 	return (op_encode_hdr_size + op_encode_change_info_maxsz)
3209 		* sizeof(__be32);
3210 }
3211 
3212 
3213 static const struct nfsd4_operation nfsd4_ops[] = {
3214 	[OP_ACCESS] = {
3215 		.op_func = nfsd4_access,
3216 		.op_name = "OP_ACCESS",
3217 		.op_rsize_bop = nfsd4_access_rsize,
3218 	},
3219 	[OP_CLOSE] = {
3220 		.op_func = nfsd4_close,
3221 		.op_flags = OP_MODIFIES_SOMETHING,
3222 		.op_name = "OP_CLOSE",
3223 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3224 		.op_get_currentstateid = nfsd4_get_closestateid,
3225 		.op_set_currentstateid = nfsd4_set_closestateid,
3226 	},
3227 	[OP_COMMIT] = {
3228 		.op_func = nfsd4_commit,
3229 		.op_flags = OP_MODIFIES_SOMETHING,
3230 		.op_name = "OP_COMMIT",
3231 		.op_rsize_bop = nfsd4_commit_rsize,
3232 	},
3233 	[OP_CREATE] = {
3234 		.op_func = nfsd4_create,
3235 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3236 		.op_name = "OP_CREATE",
3237 		.op_rsize_bop = nfsd4_create_rsize,
3238 	},
3239 	[OP_DELEGRETURN] = {
3240 		.op_func = nfsd4_delegreturn,
3241 		.op_flags = OP_MODIFIES_SOMETHING,
3242 		.op_name = "OP_DELEGRETURN",
3243 		.op_rsize_bop = nfsd4_only_status_rsize,
3244 		.op_get_currentstateid = nfsd4_get_delegreturnstateid,
3245 	},
3246 	[OP_GETATTR] = {
3247 		.op_func = nfsd4_getattr,
3248 		.op_flags = ALLOWED_ON_ABSENT_FS,
3249 		.op_rsize_bop = nfsd4_getattr_rsize,
3250 		.op_name = "OP_GETATTR",
3251 	},
3252 	[OP_GETFH] = {
3253 		.op_func = nfsd4_getfh,
3254 		.op_name = "OP_GETFH",
3255 		.op_rsize_bop = nfsd4_getfh_rsize,
3256 	},
3257 	[OP_LINK] = {
3258 		.op_func = nfsd4_link,
3259 		.op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3260 				| OP_CACHEME,
3261 		.op_name = "OP_LINK",
3262 		.op_rsize_bop = nfsd4_link_rsize,
3263 	},
3264 	[OP_LOCK] = {
3265 		.op_func = nfsd4_lock,
3266 		.op_release = nfsd4_lock_release,
3267 		.op_flags = OP_MODIFIES_SOMETHING |
3268 				OP_NONTRIVIAL_ERROR_ENCODE,
3269 		.op_name = "OP_LOCK",
3270 		.op_rsize_bop = nfsd4_lock_rsize,
3271 		.op_set_currentstateid = nfsd4_set_lockstateid,
3272 	},
3273 	[OP_LOCKT] = {
3274 		.op_func = nfsd4_lockt,
3275 		.op_release = nfsd4_lockt_release,
3276 		.op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3277 		.op_name = "OP_LOCKT",
3278 		.op_rsize_bop = nfsd4_lock_rsize,
3279 	},
3280 	[OP_LOCKU] = {
3281 		.op_func = nfsd4_locku,
3282 		.op_flags = OP_MODIFIES_SOMETHING,
3283 		.op_name = "OP_LOCKU",
3284 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3285 		.op_get_currentstateid = nfsd4_get_lockustateid,
3286 	},
3287 	[OP_LOOKUP] = {
3288 		.op_func = nfsd4_lookup,
3289 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3290 		.op_name = "OP_LOOKUP",
3291 		.op_rsize_bop = nfsd4_only_status_rsize,
3292 	},
3293 	[OP_LOOKUPP] = {
3294 		.op_func = nfsd4_lookupp,
3295 		.op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3296 		.op_name = "OP_LOOKUPP",
3297 		.op_rsize_bop = nfsd4_only_status_rsize,
3298 	},
3299 	[OP_NVERIFY] = {
3300 		.op_func = nfsd4_nverify,
3301 		.op_name = "OP_NVERIFY",
3302 		.op_rsize_bop = nfsd4_only_status_rsize,
3303 	},
3304 	[OP_OPEN] = {
3305 		.op_func = nfsd4_open,
3306 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3307 		.op_name = "OP_OPEN",
3308 		.op_rsize_bop = nfsd4_open_rsize,
3309 		.op_set_currentstateid = nfsd4_set_openstateid,
3310 	},
3311 	[OP_OPEN_CONFIRM] = {
3312 		.op_func = nfsd4_open_confirm,
3313 		.op_flags = OP_MODIFIES_SOMETHING,
3314 		.op_name = "OP_OPEN_CONFIRM",
3315 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3316 	},
3317 	[OP_OPEN_DOWNGRADE] = {
3318 		.op_func = nfsd4_open_downgrade,
3319 		.op_flags = OP_MODIFIES_SOMETHING,
3320 		.op_name = "OP_OPEN_DOWNGRADE",
3321 		.op_rsize_bop = nfsd4_status_stateid_rsize,
3322 		.op_get_currentstateid = nfsd4_get_opendowngradestateid,
3323 		.op_set_currentstateid = nfsd4_set_opendowngradestateid,
3324 	},
3325 	[OP_PUTFH] = {
3326 		.op_func = nfsd4_putfh,
3327 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3328 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3329 		.op_name = "OP_PUTFH",
3330 		.op_rsize_bop = nfsd4_only_status_rsize,
3331 	},
3332 	[OP_PUTPUBFH] = {
3333 		.op_func = nfsd4_putrootfh,
3334 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3335 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3336 		.op_name = "OP_PUTPUBFH",
3337 		.op_rsize_bop = nfsd4_only_status_rsize,
3338 	},
3339 	[OP_PUTROOTFH] = {
3340 		.op_func = nfsd4_putrootfh,
3341 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3342 				| OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3343 		.op_name = "OP_PUTROOTFH",
3344 		.op_rsize_bop = nfsd4_only_status_rsize,
3345 	},
3346 	[OP_READ] = {
3347 		.op_func = nfsd4_read,
3348 		.op_release = nfsd4_read_release,
3349 		.op_name = "OP_READ",
3350 		.op_rsize_bop = nfsd4_read_rsize,
3351 		.op_get_currentstateid = nfsd4_get_readstateid,
3352 	},
3353 	[OP_READDIR] = {
3354 		.op_func = nfsd4_readdir,
3355 		.op_name = "OP_READDIR",
3356 		.op_rsize_bop = nfsd4_readdir_rsize,
3357 	},
3358 	[OP_READLINK] = {
3359 		.op_func = nfsd4_readlink,
3360 		.op_name = "OP_READLINK",
3361 		.op_rsize_bop = nfsd4_readlink_rsize,
3362 	},
3363 	[OP_REMOVE] = {
3364 		.op_func = nfsd4_remove,
3365 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3366 		.op_name = "OP_REMOVE",
3367 		.op_rsize_bop = nfsd4_remove_rsize,
3368 	},
3369 	[OP_RENAME] = {
3370 		.op_func = nfsd4_rename,
3371 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3372 		.op_name = "OP_RENAME",
3373 		.op_rsize_bop = nfsd4_rename_rsize,
3374 	},
3375 	[OP_RENEW] = {
3376 		.op_func = nfsd4_renew,
3377 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3378 				| OP_MODIFIES_SOMETHING,
3379 		.op_name = "OP_RENEW",
3380 		.op_rsize_bop = nfsd4_only_status_rsize,
3381 
3382 	},
3383 	[OP_RESTOREFH] = {
3384 		.op_func = nfsd4_restorefh,
3385 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3386 				| OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3387 		.op_name = "OP_RESTOREFH",
3388 		.op_rsize_bop = nfsd4_only_status_rsize,
3389 	},
3390 	[OP_SAVEFH] = {
3391 		.op_func = nfsd4_savefh,
3392 		.op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3393 		.op_name = "OP_SAVEFH",
3394 		.op_rsize_bop = nfsd4_only_status_rsize,
3395 	},
3396 	[OP_SECINFO] = {
3397 		.op_func = nfsd4_secinfo,
3398 		.op_release = nfsd4_secinfo_release,
3399 		.op_flags = OP_HANDLES_WRONGSEC,
3400 		.op_name = "OP_SECINFO",
3401 		.op_rsize_bop = nfsd4_secinfo_rsize,
3402 	},
3403 	[OP_SETATTR] = {
3404 		.op_func = nfsd4_setattr,
3405 		.op_name = "OP_SETATTR",
3406 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3407 				| OP_NONTRIVIAL_ERROR_ENCODE,
3408 		.op_rsize_bop = nfsd4_setattr_rsize,
3409 		.op_get_currentstateid = nfsd4_get_setattrstateid,
3410 	},
3411 	[OP_SETCLIENTID] = {
3412 		.op_func = nfsd4_setclientid,
3413 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3414 				| OP_MODIFIES_SOMETHING | OP_CACHEME
3415 				| OP_NONTRIVIAL_ERROR_ENCODE,
3416 		.op_name = "OP_SETCLIENTID",
3417 		.op_rsize_bop = nfsd4_setclientid_rsize,
3418 	},
3419 	[OP_SETCLIENTID_CONFIRM] = {
3420 		.op_func = nfsd4_setclientid_confirm,
3421 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3422 				| OP_MODIFIES_SOMETHING | OP_CACHEME,
3423 		.op_name = "OP_SETCLIENTID_CONFIRM",
3424 		.op_rsize_bop = nfsd4_only_status_rsize,
3425 	},
3426 	[OP_VERIFY] = {
3427 		.op_func = nfsd4_verify,
3428 		.op_name = "OP_VERIFY",
3429 		.op_rsize_bop = nfsd4_only_status_rsize,
3430 	},
3431 	[OP_WRITE] = {
3432 		.op_func = nfsd4_write,
3433 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3434 		.op_name = "OP_WRITE",
3435 		.op_rsize_bop = nfsd4_write_rsize,
3436 		.op_get_currentstateid = nfsd4_get_writestateid,
3437 	},
3438 	[OP_RELEASE_LOCKOWNER] = {
3439 		.op_func = nfsd4_release_lockowner,
3440 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3441 				| OP_MODIFIES_SOMETHING,
3442 		.op_name = "OP_RELEASE_LOCKOWNER",
3443 		.op_rsize_bop = nfsd4_only_status_rsize,
3444 	},
3445 
3446 	/* NFSv4.1 operations */
3447 	[OP_EXCHANGE_ID] = {
3448 		.op_func = nfsd4_exchange_id,
3449 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3450 				| OP_MODIFIES_SOMETHING,
3451 		.op_name = "OP_EXCHANGE_ID",
3452 		.op_rsize_bop = nfsd4_exchange_id_rsize,
3453 	},
3454 	[OP_BACKCHANNEL_CTL] = {
3455 		.op_func = nfsd4_backchannel_ctl,
3456 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3457 		.op_name = "OP_BACKCHANNEL_CTL",
3458 		.op_rsize_bop = nfsd4_only_status_rsize,
3459 	},
3460 	[OP_BIND_CONN_TO_SESSION] = {
3461 		.op_func = nfsd4_bind_conn_to_session,
3462 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3463 				| OP_MODIFIES_SOMETHING,
3464 		.op_name = "OP_BIND_CONN_TO_SESSION",
3465 		.op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3466 	},
3467 	[OP_CREATE_SESSION] = {
3468 		.op_func = nfsd4_create_session,
3469 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3470 				| OP_MODIFIES_SOMETHING,
3471 		.op_name = "OP_CREATE_SESSION",
3472 		.op_rsize_bop = nfsd4_create_session_rsize,
3473 	},
3474 	[OP_DESTROY_SESSION] = {
3475 		.op_func = nfsd4_destroy_session,
3476 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3477 				| OP_MODIFIES_SOMETHING,
3478 		.op_name = "OP_DESTROY_SESSION",
3479 		.op_rsize_bop = nfsd4_only_status_rsize,
3480 	},
3481 	[OP_SEQUENCE] = {
3482 		.op_func = nfsd4_sequence,
3483 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3484 		.op_name = "OP_SEQUENCE",
3485 		.op_rsize_bop = nfsd4_sequence_rsize,
3486 	},
3487 	[OP_DESTROY_CLIENTID] = {
3488 		.op_func = nfsd4_destroy_clientid,
3489 		.op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3490 				| OP_MODIFIES_SOMETHING,
3491 		.op_name = "OP_DESTROY_CLIENTID",
3492 		.op_rsize_bop = nfsd4_only_status_rsize,
3493 	},
3494 	[OP_RECLAIM_COMPLETE] = {
3495 		.op_func = nfsd4_reclaim_complete,
3496 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3497 		.op_name = "OP_RECLAIM_COMPLETE",
3498 		.op_rsize_bop = nfsd4_only_status_rsize,
3499 	},
3500 	[OP_SECINFO_NO_NAME] = {
3501 		.op_func = nfsd4_secinfo_no_name,
3502 		.op_release = nfsd4_secinfo_no_name_release,
3503 		.op_flags = OP_HANDLES_WRONGSEC,
3504 		.op_name = "OP_SECINFO_NO_NAME",
3505 		.op_rsize_bop = nfsd4_secinfo_rsize,
3506 	},
3507 	[OP_TEST_STATEID] = {
3508 		.op_func = nfsd4_test_stateid,
3509 		.op_flags = ALLOWED_WITHOUT_FH,
3510 		.op_name = "OP_TEST_STATEID",
3511 		.op_rsize_bop = nfsd4_test_stateid_rsize,
3512 	},
3513 	[OP_FREE_STATEID] = {
3514 		.op_func = nfsd4_free_stateid,
3515 		.op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3516 		.op_name = "OP_FREE_STATEID",
3517 		.op_get_currentstateid = nfsd4_get_freestateid,
3518 		.op_rsize_bop = nfsd4_only_status_rsize,
3519 	},
3520 	[OP_GET_DIR_DELEGATION] = {
3521 		.op_func = nfsd4_get_dir_delegation,
3522 		.op_flags = OP_MODIFIES_SOMETHING,
3523 		.op_name = "OP_GET_DIR_DELEGATION",
3524 		.op_rsize_bop = nfsd4_get_dir_delegation_rsize,
3525 	},
3526 #ifdef CONFIG_NFSD_PNFS
3527 	[OP_GETDEVICEINFO] = {
3528 		.op_func = nfsd4_getdeviceinfo,
3529 		.op_release = nfsd4_getdeviceinfo_release,
3530 		.op_flags = ALLOWED_WITHOUT_FH,
3531 		.op_name = "OP_GETDEVICEINFO",
3532 		.op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3533 	},
3534 	[OP_LAYOUTGET] = {
3535 		.op_func = nfsd4_layoutget,
3536 		.op_release = nfsd4_layoutget_release,
3537 		.op_flags = OP_MODIFIES_SOMETHING,
3538 		.op_name = "OP_LAYOUTGET",
3539 		.op_rsize_bop = nfsd4_layoutget_rsize,
3540 	},
3541 	[OP_LAYOUTCOMMIT] = {
3542 		.op_func = nfsd4_layoutcommit,
3543 		.op_flags = OP_MODIFIES_SOMETHING,
3544 		.op_name = "OP_LAYOUTCOMMIT",
3545 		.op_rsize_bop = nfsd4_layoutcommit_rsize,
3546 	},
3547 	[OP_LAYOUTRETURN] = {
3548 		.op_func = nfsd4_layoutreturn,
3549 		.op_flags = OP_MODIFIES_SOMETHING,
3550 		.op_name = "OP_LAYOUTRETURN",
3551 		.op_rsize_bop = nfsd4_layoutreturn_rsize,
3552 	},
3553 #endif /* CONFIG_NFSD_PNFS */
3554 
3555 	/* NFSv4.2 operations */
3556 	[OP_ALLOCATE] = {
3557 		.op_func = nfsd4_allocate,
3558 		.op_flags = OP_MODIFIES_SOMETHING,
3559 		.op_name = "OP_ALLOCATE",
3560 		.op_rsize_bop = nfsd4_only_status_rsize,
3561 	},
3562 	[OP_DEALLOCATE] = {
3563 		.op_func = nfsd4_deallocate,
3564 		.op_flags = OP_MODIFIES_SOMETHING,
3565 		.op_name = "OP_DEALLOCATE",
3566 		.op_rsize_bop = nfsd4_only_status_rsize,
3567 	},
3568 	[OP_CLONE] = {
3569 		.op_func = nfsd4_clone,
3570 		.op_flags = OP_MODIFIES_SOMETHING,
3571 		.op_name = "OP_CLONE",
3572 		.op_rsize_bop = nfsd4_only_status_rsize,
3573 	},
3574 	[OP_COPY] = {
3575 		.op_func = nfsd4_copy,
3576 		.op_flags = OP_MODIFIES_SOMETHING,
3577 		.op_name = "OP_COPY",
3578 		.op_rsize_bop = nfsd4_copy_rsize,
3579 	},
3580 	[OP_READ_PLUS] = {
3581 		.op_func = nfsd4_read,
3582 		.op_release = nfsd4_read_release,
3583 		.op_name = "OP_READ_PLUS",
3584 		.op_rsize_bop = nfsd4_read_plus_rsize,
3585 		.op_get_currentstateid = nfsd4_get_readstateid,
3586 	},
3587 	[OP_SEEK] = {
3588 		.op_func = nfsd4_seek,
3589 		.op_name = "OP_SEEK",
3590 		.op_rsize_bop = nfsd4_seek_rsize,
3591 	},
3592 	[OP_OFFLOAD_STATUS] = {
3593 		.op_func = nfsd4_offload_status,
3594 		.op_name = "OP_OFFLOAD_STATUS",
3595 		.op_rsize_bop = nfsd4_offload_status_rsize,
3596 	},
3597 	[OP_OFFLOAD_CANCEL] = {
3598 		.op_func = nfsd4_offload_cancel,
3599 		.op_flags = OP_MODIFIES_SOMETHING,
3600 		.op_name = "OP_OFFLOAD_CANCEL",
3601 		.op_rsize_bop = nfsd4_only_status_rsize,
3602 	},
3603 	[OP_COPY_NOTIFY] = {
3604 		.op_func = nfsd4_copy_notify,
3605 		.op_flags = OP_MODIFIES_SOMETHING,
3606 		.op_name = "OP_COPY_NOTIFY",
3607 		.op_rsize_bop = nfsd4_copy_notify_rsize,
3608 	},
3609 	[OP_GETXATTR] = {
3610 		.op_func = nfsd4_getxattr,
3611 		.op_name = "OP_GETXATTR",
3612 		.op_rsize_bop = nfsd4_getxattr_rsize,
3613 	},
3614 	[OP_SETXATTR] = {
3615 		.op_func = nfsd4_setxattr,
3616 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3617 		.op_name = "OP_SETXATTR",
3618 		.op_rsize_bop = nfsd4_setxattr_rsize,
3619 	},
3620 	[OP_LISTXATTRS] = {
3621 		.op_func = nfsd4_listxattrs,
3622 		.op_name = "OP_LISTXATTRS",
3623 		.op_rsize_bop = nfsd4_listxattrs_rsize,
3624 	},
3625 	[OP_REMOVEXATTR] = {
3626 		.op_func = nfsd4_removexattr,
3627 		.op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3628 		.op_name = "OP_REMOVEXATTR",
3629 		.op_rsize_bop = nfsd4_removexattr_rsize,
3630 	},
3631 };
3632 
3633 /**
3634  * nfsd4_spo_must_allow - Determine if the compound op contains an
3635  * operation that is allowed to be sent with machine credentials
3636  *
3637  * @rqstp: a pointer to the struct svc_rqst
3638  *
3639  * Checks to see if the compound contains a spo_must_allow op
3640  * and confirms that it was sent with the proper machine creds.
3641  */
3642 
3643 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3644 {
3645 	struct nfsd4_compoundres *resp = rqstp->rq_resp;
3646 	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3647 	struct nfsd4_op *this;
3648 	struct nfsd4_compound_state *cstate = &resp->cstate;
3649 	struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3650 	u32 opiter;
3651 
3652 	if (!cstate->minorversion)
3653 		return false;
3654 
3655 	if (cstate->spo_must_allowed)
3656 		return true;
3657 
3658 	opiter = resp->opcnt;
3659 	while (opiter < argp->opcnt) {
3660 		this = &argp->ops[opiter++];
3661 		if (test_bit(this->opnum, allow->u.longs) &&
3662 			cstate->clp->cl_mach_cred &&
3663 			nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3664 			cstate->spo_must_allowed = true;
3665 			return true;
3666 		}
3667 	}
3668 	cstate->spo_must_allowed = false;
3669 	return false;
3670 }
3671 
3672 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3673 {
3674 	if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3675 		return op_encode_hdr_size * sizeof(__be32);
3676 
3677 	BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3678 	return OPDESC(op)->op_rsize_bop(rqstp, op);
3679 }
3680 
3681 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3682 {
3683 	if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3684 		pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3685 			op->opnum, nfsd4_op_name(op->opnum));
3686 		WARN_ON_ONCE(1);
3687 	}
3688 }
3689 
3690 static const char *nfsd4_op_name(unsigned opnum)
3691 {
3692 	if (opnum < ARRAY_SIZE(nfsd4_ops))
3693 		return nfsd4_ops[opnum].op_name;
3694 	return "unknown_operation";
3695 }
3696 
3697 static const struct svc_procedure nfsd_procedures4[2] = {
3698 	[NFSPROC4_NULL] = {
3699 		.pc_func = nfsd4_proc_null,
3700 		.pc_decode = nfssvc_decode_voidarg,
3701 		.pc_encode = nfssvc_encode_voidres,
3702 		.pc_argsize = sizeof(struct nfsd_voidargs),
3703 		.pc_argzero = sizeof(struct nfsd_voidargs),
3704 		.pc_ressize = sizeof(struct nfsd_voidres),
3705 		.pc_cachetype = RC_NOCACHE,
3706 		.pc_xdrressize = 1,
3707 		.pc_name = "NULL",
3708 	},
3709 	[NFSPROC4_COMPOUND] = {
3710 		.pc_func = nfsd4_proc_compound,
3711 		.pc_decode = nfs4svc_decode_compoundargs,
3712 		.pc_encode = nfs4svc_encode_compoundres,
3713 		.pc_argsize = sizeof(struct nfsd4_compoundargs),
3714 		.pc_argzero = offsetof(struct nfsd4_compoundargs, iops),
3715 		.pc_ressize = sizeof(struct nfsd4_compoundres),
3716 		.pc_release = nfsd4_release_compoundargs,
3717 		.pc_cachetype = RC_NOCACHE,
3718 		.pc_xdrressize = NFSD_BUFSIZE/4,
3719 		.pc_name = "COMPOUND",
3720 	},
3721 };
3722 
3723 static DEFINE_PER_CPU_ALIGNED(unsigned long,
3724 			      nfsd_count4[ARRAY_SIZE(nfsd_procedures4)]);
3725 const struct svc_version nfsd_version4 = {
3726 	.vs_vers		= 4,
3727 	.vs_nproc		= ARRAY_SIZE(nfsd_procedures4),
3728 	.vs_proc		= nfsd_procedures4,
3729 	.vs_count		= nfsd_count4,
3730 	.vs_dispatch		= nfsd_dispatch,
3731 	.vs_xdrsize		= NFS4_SVC_XDRSIZE,
3732 	.vs_rpcb_optnl		= true,
3733 	.vs_need_cong_ctrl	= true,
3734 };
3735