GNU Linux-libre 5.19-rc6-gnu
[releases.git] / fs / nfsd / nfs4proc.c
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 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
68 #include <linux/security.h>
69
70 static inline void
71 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
72 {
73         struct inode *inode = d_inode(resfh->fh_dentry);
74         int status;
75
76         inode_lock(inode);
77         status = security_inode_setsecctx(resfh->fh_dentry,
78                 label->data, label->len);
79         inode_unlock(inode);
80
81         if (status)
82                 /*
83                  * XXX: We should really fail the whole open, but we may
84                  * already have created a new file, so it may be too
85                  * late.  For now this seems the least of evils:
86                  */
87                 bmval[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
88
89         return;
90 }
91 #else
92 static inline void
93 nfsd4_security_inode_setsecctx(struct svc_fh *resfh, struct xdr_netobj *label, u32 *bmval)
94 { }
95 #endif
96
97 #define NFSDDBG_FACILITY                NFSDDBG_PROC
98
99 static u32 nfsd_attrmask[] = {
100         NFSD_WRITEABLE_ATTRS_WORD0,
101         NFSD_WRITEABLE_ATTRS_WORD1,
102         NFSD_WRITEABLE_ATTRS_WORD2
103 };
104
105 static u32 nfsd41_ex_attrmask[] = {
106         NFSD_SUPPATTR_EXCLCREAT_WORD0,
107         NFSD_SUPPATTR_EXCLCREAT_WORD1,
108         NFSD_SUPPATTR_EXCLCREAT_WORD2
109 };
110
111 static __be32
112 check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
113                    u32 *bmval, u32 *writable)
114 {
115         struct dentry *dentry = cstate->current_fh.fh_dentry;
116         struct svc_export *exp = cstate->current_fh.fh_export;
117
118         if (!nfsd_attrs_supported(cstate->minorversion, bmval))
119                 return nfserr_attrnotsupp;
120         if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
121                 return nfserr_attrnotsupp;
122         if ((bmval[2] & FATTR4_WORD2_SECURITY_LABEL) &&
123                         !(exp->ex_flags & NFSEXP_SECURITY_LABEL))
124                 return nfserr_attrnotsupp;
125         if (writable && !bmval_is_subset(bmval, writable))
126                 return nfserr_inval;
127         if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
128                         (bmval[1] & FATTR4_WORD1_MODE))
129                 return nfserr_inval;
130         return nfs_ok;
131 }
132
133 static __be32
134 nfsd4_check_open_attributes(struct svc_rqst *rqstp,
135         struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
136 {
137         __be32 status = nfs_ok;
138
139         if (open->op_create == NFS4_OPEN_CREATE) {
140                 if (open->op_createmode == NFS4_CREATE_UNCHECKED
141                     || open->op_createmode == NFS4_CREATE_GUARDED)
142                         status = check_attr_support(rqstp, cstate,
143                                         open->op_bmval, nfsd_attrmask);
144                 else if (open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1)
145                         status = check_attr_support(rqstp, cstate,
146                                         open->op_bmval, nfsd41_ex_attrmask);
147         }
148
149         return status;
150 }
151
152 static int
153 is_create_with_attrs(struct nfsd4_open *open)
154 {
155         return open->op_create == NFS4_OPEN_CREATE
156                 && (open->op_createmode == NFS4_CREATE_UNCHECKED
157                     || open->op_createmode == NFS4_CREATE_GUARDED
158                     || open->op_createmode == NFS4_CREATE_EXCLUSIVE4_1);
159 }
160
161 /*
162  * if error occurs when setting the acl, just clear the acl bit
163  * in the returned attr bitmap.
164  */
165 static void
166 do_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
167                 struct nfs4_acl *acl, u32 *bmval)
168 {
169         __be32 status;
170
171         status = nfsd4_set_nfs4_acl(rqstp, fhp, acl);
172         if (status)
173                 /*
174                  * We should probably fail the whole open at this point,
175                  * but we've already created the file, so it's too late;
176                  * So this seems the least of evils:
177                  */
178                 bmval[0] &= ~FATTR4_WORD0_ACL;
179 }
180
181 static inline void
182 fh_dup2(struct svc_fh *dst, struct svc_fh *src)
183 {
184         fh_put(dst);
185         dget(src->fh_dentry);
186         if (src->fh_export)
187                 exp_get(src->fh_export);
188         *dst = *src;
189 }
190
191 static __be32
192 do_open_permission(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open, int accmode)
193 {
194         __be32 status;
195
196         if (open->op_truncate &&
197                 !(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
198                 return nfserr_inval;
199
200         accmode |= NFSD_MAY_READ_IF_EXEC;
201
202         if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
203                 accmode |= NFSD_MAY_READ;
204         if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
205                 accmode |= (NFSD_MAY_WRITE | NFSD_MAY_TRUNC);
206         if (open->op_share_deny & NFS4_SHARE_DENY_READ)
207                 accmode |= NFSD_MAY_WRITE;
208
209         status = fh_verify(rqstp, current_fh, S_IFREG, accmode);
210
211         return status;
212 }
213
214 static __be32 nfsd_check_obj_isreg(struct svc_fh *fh)
215 {
216         umode_t mode = d_inode(fh->fh_dentry)->i_mode;
217
218         if (S_ISREG(mode))
219                 return nfs_ok;
220         if (S_ISDIR(mode))
221                 return nfserr_isdir;
222         /*
223          * Using err_symlink as our catch-all case may look odd; but
224          * there's no other obvious error for this case in 4.0, and we
225          * happen to know that it will cause the linux v4 client to do
226          * the right thing on attempts to open something other than a
227          * regular file.
228          */
229         return nfserr_symlink;
230 }
231
232 static void nfsd4_set_open_owner_reply_cache(struct nfsd4_compound_state *cstate, struct nfsd4_open *open, struct svc_fh *resfh)
233 {
234         if (nfsd4_has_session(cstate))
235                 return;
236         fh_copy_shallow(&open->op_openowner->oo_owner.so_replay.rp_openfh,
237                         &resfh->fh_handle);
238 }
239
240 static inline bool nfsd4_create_is_exclusive(int createmode)
241 {
242         return createmode == NFS4_CREATE_EXCLUSIVE ||
243                 createmode == NFS4_CREATE_EXCLUSIVE4_1;
244 }
245
246 static __be32
247 nfsd4_vfs_create(struct svc_fh *fhp, struct dentry *child,
248                  struct nfsd4_open *open)
249 {
250         struct file *filp;
251         struct path path;
252         int oflags;
253
254         oflags = O_CREAT | O_LARGEFILE;
255         switch (open->op_share_access & NFS4_SHARE_ACCESS_BOTH) {
256         case NFS4_SHARE_ACCESS_WRITE:
257                 oflags |= O_WRONLY;
258                 break;
259         case NFS4_SHARE_ACCESS_BOTH:
260                 oflags |= O_RDWR;
261                 break;
262         default:
263                 oflags |= O_RDONLY;
264         }
265
266         path.mnt = fhp->fh_export->ex_path.mnt;
267         path.dentry = child;
268         filp = dentry_create(&path, oflags, open->op_iattr.ia_mode,
269                              current_cred());
270         if (IS_ERR(filp))
271                 return nfserrno(PTR_ERR(filp));
272
273         open->op_filp = filp;
274         return nfs_ok;
275 }
276
277 /*
278  * Implement NFSv4's unchecked, guarded, and exclusive create
279  * semantics for regular files. Open state for this new file is
280  * subsequently fabricated in nfsd4_process_open2().
281  *
282  * Upon return, caller must release @fhp and @resfhp.
283  */
284 static __be32
285 nfsd4_create_file(struct svc_rqst *rqstp, struct svc_fh *fhp,
286                   struct svc_fh *resfhp, struct nfsd4_open *open)
287 {
288         struct iattr *iap = &open->op_iattr;
289         struct dentry *parent, *child;
290         __u32 v_mtime, v_atime;
291         struct inode *inode;
292         __be32 status;
293         int host_err;
294
295         if (isdotent(open->op_fname, open->op_fnamelen))
296                 return nfserr_exist;
297         if (!(iap->ia_valid & ATTR_MODE))
298                 iap->ia_mode = 0;
299
300         status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
301         if (status != nfs_ok)
302                 return status;
303         parent = fhp->fh_dentry;
304         inode = d_inode(parent);
305
306         host_err = fh_want_write(fhp);
307         if (host_err)
308                 return nfserrno(host_err);
309
310         fh_lock_nested(fhp, I_MUTEX_PARENT);
311
312         child = lookup_one_len(open->op_fname, parent, open->op_fnamelen);
313         if (IS_ERR(child)) {
314                 status = nfserrno(PTR_ERR(child));
315                 goto out;
316         }
317
318         if (d_really_is_negative(child)) {
319                 status = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
320                 if (status != nfs_ok)
321                         goto out;
322         }
323
324         status = fh_compose(resfhp, fhp->fh_export, child, fhp);
325         if (status != nfs_ok)
326                 goto out;
327
328         v_mtime = 0;
329         v_atime = 0;
330         if (nfsd4_create_is_exclusive(open->op_createmode)) {
331                 u32 *verifier = (u32 *)open->op_verf.data;
332
333                 /*
334                  * Solaris 7 gets confused (bugid 4218508) if these have
335                  * the high bit set, as do xfs filesystems without the
336                  * "bigtime" feature. So just clear the high bits. If this
337                  * is ever changed to use different attrs for storing the
338                  * verifier, then do_open_lookup() will also need to be
339                  * fixed accordingly.
340                  */
341                 v_mtime = verifier[0] & 0x7fffffff;
342                 v_atime = verifier[1] & 0x7fffffff;
343         }
344
345         if (d_really_is_positive(child)) {
346                 status = nfs_ok;
347
348                 switch (open->op_createmode) {
349                 case NFS4_CREATE_UNCHECKED:
350                         if (!d_is_reg(child))
351                                 break;
352
353                         /*
354                          * In NFSv4, we don't want to truncate the file
355                          * now. This would be wrong if the OPEN fails for
356                          * some other reason. Furthermore, if the size is
357                          * nonzero, we should ignore it according to spec!
358                          */
359                         open->op_truncate = (iap->ia_valid & ATTR_SIZE) &&
360                                                 !iap->ia_size;
361                         break;
362                 case NFS4_CREATE_GUARDED:
363                         status = nfserr_exist;
364                         break;
365                 case NFS4_CREATE_EXCLUSIVE:
366                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
367                             d_inode(child)->i_atime.tv_sec == v_atime &&
368                             d_inode(child)->i_size == 0) {
369                                 open->op_created = true;
370                                 break;          /* subtle */
371                         }
372                         status = nfserr_exist;
373                         break;
374                 case NFS4_CREATE_EXCLUSIVE4_1:
375                         if (d_inode(child)->i_mtime.tv_sec == v_mtime &&
376                             d_inode(child)->i_atime.tv_sec == v_atime &&
377                             d_inode(child)->i_size == 0) {
378                                 open->op_created = true;
379                                 goto set_attr;  /* subtle */
380                         }
381                         status = nfserr_exist;
382                 }
383                 goto out;
384         }
385
386         if (!IS_POSIXACL(inode))
387                 iap->ia_mode &= ~current_umask();
388
389         status = nfsd4_vfs_create(fhp, child, open);
390         if (status != nfs_ok)
391                 goto out;
392         open->op_created = true;
393
394         /* A newly created file already has a file size of zero. */
395         if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
396                 iap->ia_valid &= ~ATTR_SIZE;
397         if (nfsd4_create_is_exclusive(open->op_createmode)) {
398                 iap->ia_valid = ATTR_MTIME | ATTR_ATIME |
399                                 ATTR_MTIME_SET|ATTR_ATIME_SET;
400                 iap->ia_mtime.tv_sec = v_mtime;
401                 iap->ia_atime.tv_sec = v_atime;
402                 iap->ia_mtime.tv_nsec = 0;
403                 iap->ia_atime.tv_nsec = 0;
404         }
405
406 set_attr:
407         status = nfsd_create_setattr(rqstp, fhp, resfhp, iap);
408
409 out:
410         fh_unlock(fhp);
411         if (child && !IS_ERR(child))
412                 dput(child);
413         fh_drop_write(fhp);
414         return status;
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                 if (!status && open->op_label.len)
451                         nfsd4_security_inode_setsecctx(*resfh, &open->op_label, open->op_bmval);
452
453                 /*
454                  * Following rfc 3530 14.2.16, and rfc 5661 18.16.4
455                  * use the returned bitmask to indicate which attributes
456                  * we used to store the verifier:
457                  */
458                 if (nfsd4_create_is_exclusive(open->op_createmode) && status == 0)
459                         open->op_bmval[1] |= (FATTR4_WORD1_TIME_ACCESS |
460                                                 FATTR4_WORD1_TIME_MODIFY);
461         } else
462                 /*
463                  * Note this may exit with the parent still locked.
464                  * We will hold the lock until nfsd4_open's final
465                  * lookup, to prevent renames or unlinks until we've had
466                  * a chance to an acquire a delegation if appropriate.
467                  */
468                 status = nfsd_lookup(rqstp, current_fh,
469                                      open->op_fname, open->op_fnamelen, *resfh);
470         if (status)
471                 goto out;
472         status = nfsd_check_obj_isreg(*resfh);
473         if (status)
474                 goto out;
475
476         if (is_create_with_attrs(open) && open->op_acl != NULL)
477                 do_set_nfs4_acl(rqstp, *resfh, open->op_acl, open->op_bmval);
478
479         nfsd4_set_open_owner_reply_cache(cstate, open, *resfh);
480         accmode = NFSD_MAY_NOP;
481         if (open->op_created ||
482                         open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR)
483                 accmode |= NFSD_MAY_OWNER_OVERRIDE;
484         status = do_open_permission(rqstp, *resfh, open, accmode);
485         set_change_info(&open->op_cinfo, current_fh);
486 out:
487         return status;
488 }
489
490 static __be32
491 do_open_fhandle(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
492 {
493         struct svc_fh *current_fh = &cstate->current_fh;
494         __be32 status;
495         int accmode = 0;
496
497         /* We don't know the target directory, and therefore can not
498         * set the change info
499         */
500
501         memset(&open->op_cinfo, 0, sizeof(struct nfsd4_change_info));
502
503         nfsd4_set_open_owner_reply_cache(cstate, open, current_fh);
504
505         open->op_truncate = (open->op_iattr.ia_valid & ATTR_SIZE) &&
506                 (open->op_iattr.ia_size == 0);
507         /*
508          * In the delegation case, the client is telling us about an
509          * open that it *already* performed locally, some time ago.  We
510          * should let it succeed now if possible.
511          *
512          * In the case of a CLAIM_FH open, on the other hand, the client
513          * may be counting on us to enforce permissions (the Linux 4.1
514          * client uses this for normal opens, for example).
515          */
516         if (open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH)
517                 accmode = NFSD_MAY_OWNER_OVERRIDE;
518
519         status = do_open_permission(rqstp, current_fh, open, accmode);
520
521         return status;
522 }
523
524 static void
525 copy_clientid(clientid_t *clid, struct nfsd4_session *session)
526 {
527         struct nfsd4_sessionid *sid =
528                         (struct nfsd4_sessionid *)session->se_sessionid.data;
529
530         clid->cl_boot = sid->clientid.cl_boot;
531         clid->cl_id = sid->clientid.cl_id;
532 }
533
534 static __be32
535 nfsd4_open(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
536            union nfsd4_op_u *u)
537 {
538         struct nfsd4_open *open = &u->open;
539         __be32 status;
540         struct svc_fh *resfh = NULL;
541         struct net *net = SVC_NET(rqstp);
542         struct nfsd_net *nn = net_generic(net, nfsd_net_id);
543         bool reclaim = false;
544
545         dprintk("NFSD: nfsd4_open filename %.*s op_openowner %p\n",
546                 (int)open->op_fnamelen, open->op_fname,
547                 open->op_openowner);
548
549         open->op_filp = NULL;
550
551         /* This check required by spec. */
552         if (open->op_create && open->op_claim_type != NFS4_OPEN_CLAIM_NULL)
553                 return nfserr_inval;
554
555         open->op_created = false;
556         /*
557          * RFC5661 18.51.3
558          * Before RECLAIM_COMPLETE done, server should deny new lock
559          */
560         if (nfsd4_has_session(cstate) &&
561             !test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags) &&
562             open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
563                 return nfserr_grace;
564
565         if (nfsd4_has_session(cstate))
566                 copy_clientid(&open->op_clientid, cstate->session);
567
568         /* check seqid for replay. set nfs4_owner */
569         status = nfsd4_process_open1(cstate, open, nn);
570         if (status == nfserr_replay_me) {
571                 struct nfs4_replay *rp = &open->op_openowner->oo_owner.so_replay;
572                 fh_put(&cstate->current_fh);
573                 fh_copy_shallow(&cstate->current_fh.fh_handle,
574                                 &rp->rp_openfh);
575                 status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
576                 if (status)
577                         dprintk("nfsd4_open: replay failed"
578                                 " restoring previous filehandle\n");
579                 else
580                         status = nfserr_replay_me;
581         }
582         if (status)
583                 goto out;
584         if (open->op_xdr_error) {
585                 status = open->op_xdr_error;
586                 goto out;
587         }
588
589         status = nfsd4_check_open_attributes(rqstp, cstate, open);
590         if (status)
591                 goto out;
592
593         /* Openowner is now set, so sequence id will get bumped.  Now we need
594          * these checks before we do any creates: */
595         status = nfserr_grace;
596         if (opens_in_grace(net) && open->op_claim_type != NFS4_OPEN_CLAIM_PREVIOUS)
597                 goto out;
598         status = nfserr_no_grace;
599         if (!opens_in_grace(net) && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
600                 goto out;
601
602         switch (open->op_claim_type) {
603         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
604         case NFS4_OPEN_CLAIM_NULL:
605                 status = do_open_lookup(rqstp, cstate, open, &resfh);
606                 if (status)
607                         goto out;
608                 break;
609         case NFS4_OPEN_CLAIM_PREVIOUS:
610                 status = nfs4_check_open_reclaim(cstate->clp);
611                 if (status)
612                         goto out;
613                 open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
614                 reclaim = true;
615                 fallthrough;
616         case NFS4_OPEN_CLAIM_FH:
617         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
618                 status = do_open_fhandle(rqstp, cstate, open);
619                 if (status)
620                         goto out;
621                 resfh = &cstate->current_fh;
622                 break;
623         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
624         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
625                 status = nfserr_notsupp;
626                 goto out;
627         default:
628                 status = nfserr_inval;
629                 goto out;
630         }
631
632         status = nfsd4_process_open2(rqstp, resfh, open);
633         WARN(status && open->op_created,
634              "nfsd4_process_open2 failed to open newly-created file! status=%u\n",
635              be32_to_cpu(status));
636         if (reclaim && !status)
637                 nn->somebody_reclaimed = true;
638 out:
639         if (open->op_filp) {
640                 fput(open->op_filp);
641                 open->op_filp = NULL;
642         }
643         if (resfh && resfh != &cstate->current_fh) {
644                 fh_dup2(&cstate->current_fh, resfh);
645                 fh_put(resfh);
646                 kfree(resfh);
647         }
648         nfsd4_cleanup_open_state(cstate, open);
649         nfsd4_bump_seqid(cstate, status);
650         return status;
651 }
652
653 /*
654  * OPEN is the only seqid-mutating operation whose decoding can fail
655  * with a seqid-mutating error (specifically, decoding of user names in
656  * the attributes).  Therefore we have to do some processing to look up
657  * the stateowner so that we can bump the seqid.
658  */
659 static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_op *op)
660 {
661         struct nfsd4_open *open = &op->u.open;
662
663         if (!seqid_mutating_err(ntohl(op->status)))
664                 return op->status;
665         if (nfsd4_has_session(cstate))
666                 return op->status;
667         open->op_xdr_error = op->status;
668         return nfsd4_open(rqstp, cstate, &op->u);
669 }
670
671 /*
672  * filehandle-manipulating ops.
673  */
674 static __be32
675 nfsd4_getfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
676             union nfsd4_op_u *u)
677 {
678         u->getfh = &cstate->current_fh;
679         return nfs_ok;
680 }
681
682 static __be32
683 nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
684             union nfsd4_op_u *u)
685 {
686         struct nfsd4_putfh *putfh = &u->putfh;
687         __be32 ret;
688
689         fh_put(&cstate->current_fh);
690         cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
691         memcpy(&cstate->current_fh.fh_handle.fh_raw, putfh->pf_fhval,
692                putfh->pf_fhlen);
693         ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
694 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
695         if (ret == nfserr_stale && putfh->no_verify) {
696                 SET_FH_FLAG(&cstate->current_fh, NFSD4_FH_FOREIGN);
697                 ret = 0;
698         }
699 #endif
700         return ret;
701 }
702
703 static __be32
704 nfsd4_putrootfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
705                 union nfsd4_op_u *u)
706 {
707         __be32 status;
708
709         fh_put(&cstate->current_fh);
710         status = exp_pseudoroot(rqstp, &cstate->current_fh);
711         return status;
712 }
713
714 static __be32
715 nfsd4_restorefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
716                 union nfsd4_op_u *u)
717 {
718         if (!cstate->save_fh.fh_dentry)
719                 return nfserr_restorefh;
720
721         fh_dup2(&cstate->current_fh, &cstate->save_fh);
722         if (HAS_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG)) {
723                 memcpy(&cstate->current_stateid, &cstate->save_stateid, sizeof(stateid_t));
724                 SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
725         }
726         return nfs_ok;
727 }
728
729 static __be32
730 nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
731              union nfsd4_op_u *u)
732 {
733         fh_dup2(&cstate->save_fh, &cstate->current_fh);
734         if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG)) {
735                 memcpy(&cstate->save_stateid, &cstate->current_stateid, sizeof(stateid_t));
736                 SET_CSTATE_FLAG(cstate, SAVED_STATE_ID_FLAG);
737         }
738         return nfs_ok;
739 }
740
741 /*
742  * misc nfsv4 ops
743  */
744 static __be32
745 nfsd4_access(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
746              union nfsd4_op_u *u)
747 {
748         struct nfsd4_access *access = &u->access;
749         u32 access_full;
750
751         access_full = NFS3_ACCESS_FULL;
752         if (cstate->minorversion >= 2)
753                 access_full |= NFS4_ACCESS_XALIST | NFS4_ACCESS_XAREAD |
754                                NFS4_ACCESS_XAWRITE;
755
756         if (access->ac_req_access & ~access_full)
757                 return nfserr_inval;
758
759         access->ac_resp_access = access->ac_req_access;
760         return nfsd_access(rqstp, &cstate->current_fh, &access->ac_resp_access,
761                            &access->ac_supported);
762 }
763
764 static void gen_boot_verifier(nfs4_verifier *verifier, struct net *net)
765 {
766         __be32 *verf = (__be32 *)verifier->data;
767
768         BUILD_BUG_ON(2*sizeof(*verf) != sizeof(verifier->data));
769
770         nfsd_copy_write_verifier(verf, net_generic(net, nfsd_net_id));
771 }
772
773 static __be32
774 nfsd4_commit(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
775              union nfsd4_op_u *u)
776 {
777         struct nfsd4_commit *commit = &u->commit;
778
779         return nfsd_commit(rqstp, &cstate->current_fh, commit->co_offset,
780                              commit->co_count,
781                              (__be32 *)commit->co_verf.data);
782 }
783
784 static __be32
785 nfsd4_create(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
786              union nfsd4_op_u *u)
787 {
788         struct nfsd4_create *create = &u->create;
789         struct svc_fh resfh;
790         __be32 status;
791         dev_t rdev;
792
793         fh_init(&resfh, NFS4_FHSIZE);
794
795         status = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_NOP);
796         if (status)
797                 return status;
798
799         status = check_attr_support(rqstp, cstate, create->cr_bmval,
800                                     nfsd_attrmask);
801         if (status)
802                 return status;
803
804         current->fs->umask = create->cr_umask;
805         switch (create->cr_type) {
806         case NF4LNK:
807                 status = nfsd_symlink(rqstp, &cstate->current_fh,
808                                       create->cr_name, create->cr_namelen,
809                                       create->cr_data, &resfh);
810                 break;
811
812         case NF4BLK:
813                 status = nfserr_inval;
814                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
815                 if (MAJOR(rdev) != create->cr_specdata1 ||
816                     MINOR(rdev) != create->cr_specdata2)
817                         goto out_umask;
818                 status = nfsd_create(rqstp, &cstate->current_fh,
819                                      create->cr_name, create->cr_namelen,
820                                      &create->cr_iattr, S_IFBLK, rdev, &resfh);
821                 break;
822
823         case NF4CHR:
824                 status = nfserr_inval;
825                 rdev = MKDEV(create->cr_specdata1, create->cr_specdata2);
826                 if (MAJOR(rdev) != create->cr_specdata1 ||
827                     MINOR(rdev) != create->cr_specdata2)
828                         goto out_umask;
829                 status = nfsd_create(rqstp, &cstate->current_fh,
830                                      create->cr_name, create->cr_namelen,
831                                      &create->cr_iattr,S_IFCHR, rdev, &resfh);
832                 break;
833
834         case NF4SOCK:
835                 status = nfsd_create(rqstp, &cstate->current_fh,
836                                      create->cr_name, create->cr_namelen,
837                                      &create->cr_iattr, S_IFSOCK, 0, &resfh);
838                 break;
839
840         case NF4FIFO:
841                 status = nfsd_create(rqstp, &cstate->current_fh,
842                                      create->cr_name, create->cr_namelen,
843                                      &create->cr_iattr, S_IFIFO, 0, &resfh);
844                 break;
845
846         case NF4DIR:
847                 create->cr_iattr.ia_valid &= ~ATTR_SIZE;
848                 status = nfsd_create(rqstp, &cstate->current_fh,
849                                      create->cr_name, create->cr_namelen,
850                                      &create->cr_iattr, S_IFDIR, 0, &resfh);
851                 break;
852
853         default:
854                 status = nfserr_badtype;
855         }
856
857         if (status)
858                 goto out;
859
860         if (create->cr_label.len)
861                 nfsd4_security_inode_setsecctx(&resfh, &create->cr_label, create->cr_bmval);
862
863         if (create->cr_acl != NULL)
864                 do_set_nfs4_acl(rqstp, &resfh, create->cr_acl,
865                                 create->cr_bmval);
866
867         fh_unlock(&cstate->current_fh);
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         return status;
875 }
876
877 static __be32
878 nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
879               union nfsd4_op_u *u)
880 {
881         struct nfsd4_getattr *getattr = &u->getattr;
882         __be32 status;
883
884         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
885         if (status)
886                 return status;
887
888         if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
889                 return nfserr_inval;
890
891         getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
892         getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
893         getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
894
895         getattr->ga_fhp = &cstate->current_fh;
896         return nfs_ok;
897 }
898
899 static __be32
900 nfsd4_link(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
901            union nfsd4_op_u *u)
902 {
903         struct nfsd4_link *link = &u->link;
904         __be32 status;
905
906         status = nfsd_link(rqstp, &cstate->current_fh,
907                            link->li_name, link->li_namelen, &cstate->save_fh);
908         if (!status)
909                 set_change_info(&link->li_cinfo, &cstate->current_fh);
910         return status;
911 }
912
913 static __be32 nfsd4_do_lookupp(struct svc_rqst *rqstp, struct svc_fh *fh)
914 {
915         struct svc_fh tmp_fh;
916         __be32 ret;
917
918         fh_init(&tmp_fh, NFS4_FHSIZE);
919         ret = exp_pseudoroot(rqstp, &tmp_fh);
920         if (ret)
921                 return ret;
922         if (tmp_fh.fh_dentry == fh->fh_dentry) {
923                 fh_put(&tmp_fh);
924                 return nfserr_noent;
925         }
926         fh_put(&tmp_fh);
927         return nfsd_lookup(rqstp, fh, "..", 2, fh);
928 }
929
930 static __be32
931 nfsd4_lookupp(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
932               union nfsd4_op_u *u)
933 {
934         return nfsd4_do_lookupp(rqstp, &cstate->current_fh);
935 }
936
937 static __be32
938 nfsd4_lookup(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
939              union nfsd4_op_u *u)
940 {
941         return nfsd_lookup(rqstp, &cstate->current_fh,
942                            u->lookup.lo_name, u->lookup.lo_len,
943                            &cstate->current_fh);
944 }
945
946 static __be32
947 nfsd4_read(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
948            union nfsd4_op_u *u)
949 {
950         struct nfsd4_read *read = &u->read;
951         __be32 status;
952
953         read->rd_nf = NULL;
954
955         trace_nfsd_read_start(rqstp, &cstate->current_fh,
956                               read->rd_offset, read->rd_length);
957
958         read->rd_length = min_t(u32, read->rd_length, svc_max_payload(rqstp));
959         if (read->rd_offset > (u64)OFFSET_MAX)
960                 read->rd_offset = (u64)OFFSET_MAX;
961         if (read->rd_offset + read->rd_length > (u64)OFFSET_MAX)
962                 read->rd_length = (u64)OFFSET_MAX - read->rd_offset;
963
964         /*
965          * If we do a zero copy read, then a client will see read data
966          * that reflects the state of the file *after* performing the
967          * following compound.
968          *
969          * To ensure proper ordering, we therefore turn off zero copy if
970          * the client wants us to do more in this compound:
971          */
972         if (!nfsd4_last_compound_op(rqstp))
973                 __clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
974
975         /* check stateid */
976         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
977                                         &read->rd_stateid, RD_STATE,
978                                         &read->rd_nf, NULL);
979         if (status) {
980                 dprintk("NFSD: nfsd4_read: couldn't process stateid!\n");
981                 goto out;
982         }
983         status = nfs_ok;
984 out:
985         read->rd_rqstp = rqstp;
986         read->rd_fhp = &cstate->current_fh;
987         return status;
988 }
989
990
991 static void
992 nfsd4_read_release(union nfsd4_op_u *u)
993 {
994         if (u->read.rd_nf)
995                 nfsd_file_put(u->read.rd_nf);
996         trace_nfsd_read_done(u->read.rd_rqstp, u->read.rd_fhp,
997                              u->read.rd_offset, u->read.rd_length);
998 }
999
1000 static __be32
1001 nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1002               union nfsd4_op_u *u)
1003 {
1004         struct nfsd4_readdir *readdir = &u->readdir;
1005         u64 cookie = readdir->rd_cookie;
1006         static const nfs4_verifier zeroverf;
1007
1008         /* no need to check permission - this will be done in nfsd_readdir() */
1009
1010         if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
1011                 return nfserr_inval;
1012
1013         readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
1014         readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
1015         readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
1016
1017         if ((cookie == 1) || (cookie == 2) ||
1018             (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))
1019                 return nfserr_bad_cookie;
1020
1021         readdir->rd_rqstp = rqstp;
1022         readdir->rd_fhp = &cstate->current_fh;
1023         return nfs_ok;
1024 }
1025
1026 static __be32
1027 nfsd4_readlink(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1028                union nfsd4_op_u *u)
1029 {
1030         u->readlink.rl_rqstp = rqstp;
1031         u->readlink.rl_fhp = &cstate->current_fh;
1032         return nfs_ok;
1033 }
1034
1035 static __be32
1036 nfsd4_remove(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1037              union nfsd4_op_u *u)
1038 {
1039         struct nfsd4_remove *remove = &u->remove;
1040         __be32 status;
1041
1042         if (opens_in_grace(SVC_NET(rqstp)))
1043                 return nfserr_grace;
1044         status = nfsd_unlink(rqstp, &cstate->current_fh, 0,
1045                              remove->rm_name, remove->rm_namelen);
1046         if (!status) {
1047                 fh_unlock(&cstate->current_fh);
1048                 set_change_info(&remove->rm_cinfo, &cstate->current_fh);
1049         }
1050         return status;
1051 }
1052
1053 static __be32
1054 nfsd4_rename(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1055              union nfsd4_op_u *u)
1056 {
1057         struct nfsd4_rename *rename = &u->rename;
1058         __be32 status;
1059
1060         if (opens_in_grace(SVC_NET(rqstp)))
1061                 return nfserr_grace;
1062         status = nfsd_rename(rqstp, &cstate->save_fh, rename->rn_sname,
1063                              rename->rn_snamelen, &cstate->current_fh,
1064                              rename->rn_tname, rename->rn_tnamelen);
1065         if (status)
1066                 return status;
1067         set_change_info(&rename->rn_sinfo, &cstate->current_fh);
1068         set_change_info(&rename->rn_tinfo, &cstate->save_fh);
1069         return nfs_ok;
1070 }
1071
1072 static __be32
1073 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1074               union nfsd4_op_u *u)
1075 {
1076         struct nfsd4_secinfo *secinfo = &u->secinfo;
1077         struct svc_export *exp;
1078         struct dentry *dentry;
1079         __be32 err;
1080
1081         err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
1082         if (err)
1083                 return err;
1084         err = nfsd_lookup_dentry(rqstp, &cstate->current_fh,
1085                                     secinfo->si_name, secinfo->si_namelen,
1086                                     &exp, &dentry);
1087         if (err)
1088                 return err;
1089         fh_unlock(&cstate->current_fh);
1090         if (d_really_is_negative(dentry)) {
1091                 exp_put(exp);
1092                 err = nfserr_noent;
1093         } else
1094                 secinfo->si_exp = exp;
1095         dput(dentry);
1096         if (cstate->minorversion)
1097                 /* See rfc 5661 section 2.6.3.1.1.8 */
1098                 fh_put(&cstate->current_fh);
1099         return err;
1100 }
1101
1102 static __be32
1103 nfsd4_secinfo_no_name(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1104                 union nfsd4_op_u *u)
1105 {
1106         __be32 err;
1107
1108         switch (u->secinfo_no_name.sin_style) {
1109         case NFS4_SECINFO_STYLE4_CURRENT_FH:
1110                 break;
1111         case NFS4_SECINFO_STYLE4_PARENT:
1112                 err = nfsd4_do_lookupp(rqstp, &cstate->current_fh);
1113                 if (err)
1114                         return err;
1115                 break;
1116         default:
1117                 return nfserr_inval;
1118         }
1119
1120         u->secinfo_no_name.sin_exp = exp_get(cstate->current_fh.fh_export);
1121         fh_put(&cstate->current_fh);
1122         return nfs_ok;
1123 }
1124
1125 static void
1126 nfsd4_secinfo_release(union nfsd4_op_u *u)
1127 {
1128         if (u->secinfo.si_exp)
1129                 exp_put(u->secinfo.si_exp);
1130 }
1131
1132 static void
1133 nfsd4_secinfo_no_name_release(union nfsd4_op_u *u)
1134 {
1135         if (u->secinfo_no_name.sin_exp)
1136                 exp_put(u->secinfo_no_name.sin_exp);
1137 }
1138
1139 static __be32
1140 nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1141               union nfsd4_op_u *u)
1142 {
1143         struct nfsd4_setattr *setattr = &u->setattr;
1144         __be32 status = nfs_ok;
1145         int err;
1146
1147         if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
1148                 status = nfs4_preprocess_stateid_op(rqstp, cstate,
1149                                 &cstate->current_fh, &setattr->sa_stateid,
1150                                 WR_STATE, NULL, NULL);
1151                 if (status) {
1152                         dprintk("NFSD: nfsd4_setattr: couldn't process stateid!\n");
1153                         return status;
1154                 }
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         if (setattr->sa_acl != NULL)
1167                 status = nfsd4_set_nfs4_acl(rqstp, &cstate->current_fh,
1168                                             setattr->sa_acl);
1169         if (status)
1170                 goto out;
1171         if (setattr->sa_label.len)
1172                 status = nfsd4_set_nfs4_label(rqstp, &cstate->current_fh,
1173                                 &setattr->sa_label);
1174         if (status)
1175                 goto out;
1176         status = nfsd_setattr(rqstp, &cstate->current_fh, &setattr->sa_iattr,
1177                                 0, (time64_t)0);
1178 out:
1179         fh_drop_write(&cstate->current_fh);
1180         return status;
1181 }
1182
1183 static __be32
1184 nfsd4_write(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1185             union nfsd4_op_u *u)
1186 {
1187         struct nfsd4_write *write = &u->write;
1188         stateid_t *stateid = &write->wr_stateid;
1189         struct nfsd_file *nf = NULL;
1190         __be32 status = nfs_ok;
1191         unsigned long cnt;
1192         int nvecs;
1193
1194         if (write->wr_offset > (u64)OFFSET_MAX ||
1195             write->wr_offset + write->wr_buflen > (u64)OFFSET_MAX)
1196                 return nfserr_fbig;
1197
1198         cnt = write->wr_buflen;
1199         trace_nfsd_write_start(rqstp, &cstate->current_fh,
1200                                write->wr_offset, cnt);
1201         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1202                                                 stateid, WR_STATE, &nf, NULL);
1203         if (status) {
1204                 dprintk("NFSD: nfsd4_write: couldn't process stateid!\n");
1205                 return status;
1206         }
1207
1208         write->wr_how_written = write->wr_stable_how;
1209
1210         nvecs = svc_fill_write_vector(rqstp, &write->wr_payload);
1211         WARN_ON_ONCE(nvecs > ARRAY_SIZE(rqstp->rq_vec));
1212
1213         status = nfsd_vfs_write(rqstp, &cstate->current_fh, nf,
1214                                 write->wr_offset, rqstp->rq_vec, nvecs, &cnt,
1215                                 write->wr_how_written,
1216                                 (__be32 *)write->wr_verifier.data);
1217         nfsd_file_put(nf);
1218
1219         write->wr_bytes_written = cnt;
1220         trace_nfsd_write_done(rqstp, &cstate->current_fh,
1221                               write->wr_offset, cnt);
1222         return status;
1223 }
1224
1225 static __be32
1226 nfsd4_verify_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1227                   stateid_t *src_stateid, struct nfsd_file **src,
1228                   stateid_t *dst_stateid, struct nfsd_file **dst)
1229 {
1230         __be32 status;
1231
1232         if (!cstate->save_fh.fh_dentry)
1233                 return nfserr_nofilehandle;
1234
1235         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->save_fh,
1236                                             src_stateid, RD_STATE, src, NULL);
1237         if (status) {
1238                 dprintk("NFSD: %s: couldn't process src stateid!\n", __func__);
1239                 goto out;
1240         }
1241
1242         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1243                                             dst_stateid, WR_STATE, dst, NULL);
1244         if (status) {
1245                 dprintk("NFSD: %s: couldn't process dst stateid!\n", __func__);
1246                 goto out_put_src;
1247         }
1248
1249         /* fix up for NFS-specific error code */
1250         if (!S_ISREG(file_inode((*src)->nf_file)->i_mode) ||
1251             !S_ISREG(file_inode((*dst)->nf_file)->i_mode)) {
1252                 status = nfserr_wrong_type;
1253                 goto out_put_dst;
1254         }
1255
1256 out:
1257         return status;
1258 out_put_dst:
1259         nfsd_file_put(*dst);
1260 out_put_src:
1261         nfsd_file_put(*src);
1262         goto out;
1263 }
1264
1265 static __be32
1266 nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1267                 union nfsd4_op_u *u)
1268 {
1269         struct nfsd4_clone *clone = &u->clone;
1270         struct nfsd_file *src, *dst;
1271         __be32 status;
1272
1273         status = nfsd4_verify_copy(rqstp, cstate, &clone->cl_src_stateid, &src,
1274                                    &clone->cl_dst_stateid, &dst);
1275         if (status)
1276                 goto out;
1277
1278         status = nfsd4_clone_file_range(rqstp, src, clone->cl_src_pos,
1279                         dst, clone->cl_dst_pos, clone->cl_count,
1280                         EX_ISSYNC(cstate->current_fh.fh_export));
1281
1282         nfsd_file_put(dst);
1283         nfsd_file_put(src);
1284 out:
1285         return status;
1286 }
1287
1288 void nfs4_put_copy(struct nfsd4_copy *copy)
1289 {
1290         if (!refcount_dec_and_test(&copy->refcount))
1291                 return;
1292         kfree(copy);
1293 }
1294
1295 static bool
1296 check_and_set_stop_copy(struct nfsd4_copy *copy)
1297 {
1298         bool value;
1299
1300         spin_lock(&copy->cp_clp->async_lock);
1301         value = copy->stopped;
1302         if (!copy->stopped)
1303                 copy->stopped = true;
1304         spin_unlock(&copy->cp_clp->async_lock);
1305         return value;
1306 }
1307
1308 static void nfsd4_stop_copy(struct nfsd4_copy *copy)
1309 {
1310         /* only 1 thread should stop the copy */
1311         if (!check_and_set_stop_copy(copy))
1312                 kthread_stop(copy->copy_task);
1313         nfs4_put_copy(copy);
1314 }
1315
1316 static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp)
1317 {
1318         struct nfsd4_copy *copy = NULL;
1319
1320         spin_lock(&clp->async_lock);
1321         if (!list_empty(&clp->async_copies)) {
1322                 copy = list_first_entry(&clp->async_copies, struct nfsd4_copy,
1323                                         copies);
1324                 refcount_inc(&copy->refcount);
1325         }
1326         spin_unlock(&clp->async_lock);
1327         return copy;
1328 }
1329
1330 void nfsd4_shutdown_copy(struct nfs4_client *clp)
1331 {
1332         struct nfsd4_copy *copy;
1333
1334         while ((copy = nfsd4_get_copy(clp)) != NULL)
1335                 nfsd4_stop_copy(copy);
1336 }
1337 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
1338
1339 extern struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1340                                    struct nfs_fh *src_fh,
1341                                    nfs4_stateid *stateid);
1342 extern void nfs42_ssc_close(struct file *filep);
1343
1344 extern void nfs_sb_deactive(struct super_block *sb);
1345
1346 #define NFSD42_INTERSSC_MOUNTOPS "vers=4.2,addr=%s,sec=sys"
1347
1348 /*
1349  * setup a work entry in the ssc delayed unmount list.
1350  */
1351 static __be32 nfsd4_ssc_setup_dul(struct nfsd_net *nn, char *ipaddr,
1352                 struct nfsd4_ssc_umount_item **retwork, struct vfsmount **ss_mnt)
1353 {
1354         struct nfsd4_ssc_umount_item *ni = NULL;
1355         struct nfsd4_ssc_umount_item *work = NULL;
1356         struct nfsd4_ssc_umount_item *tmp;
1357         DEFINE_WAIT(wait);
1358
1359         *ss_mnt = NULL;
1360         *retwork = NULL;
1361         work = kzalloc(sizeof(*work), GFP_KERNEL);
1362 try_again:
1363         spin_lock(&nn->nfsd_ssc_lock);
1364         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1365                 if (strncmp(ni->nsui_ipaddr, ipaddr, sizeof(ni->nsui_ipaddr)))
1366                         continue;
1367                 /* found a match */
1368                 if (ni->nsui_busy) {
1369                         /*  wait - and try again */
1370                         prepare_to_wait(&nn->nfsd_ssc_waitq, &wait,
1371                                 TASK_INTERRUPTIBLE);
1372                         spin_unlock(&nn->nfsd_ssc_lock);
1373
1374                         /* allow 20secs for mount/unmount for now - revisit */
1375                         if (signal_pending(current) ||
1376                                         (schedule_timeout(20*HZ) == 0)) {
1377                                 kfree(work);
1378                                 return nfserr_eagain;
1379                         }
1380                         finish_wait(&nn->nfsd_ssc_waitq, &wait);
1381                         goto try_again;
1382                 }
1383                 *ss_mnt = ni->nsui_vfsmount;
1384                 refcount_inc(&ni->nsui_refcnt);
1385                 spin_unlock(&nn->nfsd_ssc_lock);
1386                 kfree(work);
1387
1388                 /* return vfsmount in ss_mnt */
1389                 return 0;
1390         }
1391         if (work) {
1392                 strncpy(work->nsui_ipaddr, ipaddr, sizeof(work->nsui_ipaddr));
1393                 refcount_set(&work->nsui_refcnt, 2);
1394                 work->nsui_busy = true;
1395                 list_add_tail(&work->nsui_list, &nn->nfsd_ssc_mount_list);
1396                 *retwork = work;
1397         }
1398         spin_unlock(&nn->nfsd_ssc_lock);
1399         return 0;
1400 }
1401
1402 static void nfsd4_ssc_update_dul_work(struct nfsd_net *nn,
1403                 struct nfsd4_ssc_umount_item *work, struct vfsmount *ss_mnt)
1404 {
1405         /* set nsui_vfsmount, clear busy flag and wakeup waiters */
1406         spin_lock(&nn->nfsd_ssc_lock);
1407         work->nsui_vfsmount = ss_mnt;
1408         work->nsui_busy = false;
1409         wake_up_all(&nn->nfsd_ssc_waitq);
1410         spin_unlock(&nn->nfsd_ssc_lock);
1411 }
1412
1413 static void nfsd4_ssc_cancel_dul_work(struct nfsd_net *nn,
1414                 struct nfsd4_ssc_umount_item *work)
1415 {
1416         spin_lock(&nn->nfsd_ssc_lock);
1417         list_del(&work->nsui_list);
1418         wake_up_all(&nn->nfsd_ssc_waitq);
1419         spin_unlock(&nn->nfsd_ssc_lock);
1420         kfree(work);
1421 }
1422
1423 /*
1424  * Support one copy source server for now.
1425  */
1426 static __be32
1427 nfsd4_interssc_connect(struct nl4_server *nss, struct svc_rqst *rqstp,
1428                        struct vfsmount **mount)
1429 {
1430         struct file_system_type *type;
1431         struct vfsmount *ss_mnt;
1432         struct nfs42_netaddr *naddr;
1433         struct sockaddr_storage tmp_addr;
1434         size_t tmp_addrlen, match_netid_len = 3;
1435         char *startsep = "", *endsep = "", *match_netid = "tcp";
1436         char *ipaddr, *dev_name, *raw_data;
1437         int len, raw_len;
1438         __be32 status = nfserr_inval;
1439         struct nfsd4_ssc_umount_item *work = NULL;
1440         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1441
1442         naddr = &nss->u.nl4_addr;
1443         tmp_addrlen = rpc_uaddr2sockaddr(SVC_NET(rqstp), naddr->addr,
1444                                          naddr->addr_len,
1445                                          (struct sockaddr *)&tmp_addr,
1446                                          sizeof(tmp_addr));
1447         if (tmp_addrlen == 0)
1448                 goto out_err;
1449
1450         if (tmp_addr.ss_family == AF_INET6) {
1451                 startsep = "[";
1452                 endsep = "]";
1453                 match_netid = "tcp6";
1454                 match_netid_len = 4;
1455         }
1456
1457         if (naddr->netid_len != match_netid_len ||
1458                 strncmp(naddr->netid, match_netid, naddr->netid_len))
1459                 goto out_err;
1460
1461         /* Construct the raw data for the vfs_kern_mount call */
1462         len = RPC_MAX_ADDRBUFLEN + 1;
1463         ipaddr = kzalloc(len, GFP_KERNEL);
1464         if (!ipaddr)
1465                 goto out_err;
1466
1467         rpc_ntop((struct sockaddr *)&tmp_addr, ipaddr, len);
1468
1469         /* 2 for ipv6 endsep and startsep. 3 for ":/" and trailing '/0'*/
1470
1471         raw_len = strlen(NFSD42_INTERSSC_MOUNTOPS) + strlen(ipaddr);
1472         raw_data = kzalloc(raw_len, GFP_KERNEL);
1473         if (!raw_data)
1474                 goto out_free_ipaddr;
1475
1476         snprintf(raw_data, raw_len, NFSD42_INTERSSC_MOUNTOPS, ipaddr);
1477
1478         status = nfserr_nodev;
1479         type = get_fs_type("nfs");
1480         if (!type)
1481                 goto out_free_rawdata;
1482
1483         /* Set the server:<export> for the vfs_kern_mount call */
1484         dev_name = kzalloc(len + 5, GFP_KERNEL);
1485         if (!dev_name)
1486                 goto out_free_rawdata;
1487         snprintf(dev_name, len + 5, "%s%s%s:/", startsep, ipaddr, endsep);
1488
1489         status = nfsd4_ssc_setup_dul(nn, ipaddr, &work, &ss_mnt);
1490         if (status)
1491                 goto out_free_devname;
1492         if (ss_mnt)
1493                 goto out_done;
1494
1495         /* Use an 'internal' mount: SB_KERNMOUNT -> MNT_INTERNAL */
1496         ss_mnt = vfs_kern_mount(type, SB_KERNMOUNT, dev_name, raw_data);
1497         module_put(type->owner);
1498         if (IS_ERR(ss_mnt)) {
1499                 status = nfserr_nodev;
1500                 if (work)
1501                         nfsd4_ssc_cancel_dul_work(nn, work);
1502                 goto out_free_devname;
1503         }
1504         if (work)
1505                 nfsd4_ssc_update_dul_work(nn, work, ss_mnt);
1506 out_done:
1507         status = 0;
1508         *mount = ss_mnt;
1509
1510 out_free_devname:
1511         kfree(dev_name);
1512 out_free_rawdata:
1513         kfree(raw_data);
1514 out_free_ipaddr:
1515         kfree(ipaddr);
1516 out_err:
1517         return status;
1518 }
1519
1520 static void
1521 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1522 {
1523         nfs_do_sb_deactive(ss_mnt->mnt_sb);
1524         mntput(ss_mnt);
1525 }
1526
1527 /*
1528  * Verify COPY destination stateid.
1529  *
1530  * Connect to the source server with NFSv4.1.
1531  * Create the source struct file for nfsd_copy_range.
1532  * Called with COPY cstate:
1533  *    SAVED_FH: source filehandle
1534  *    CURRENT_FH: destination filehandle
1535  */
1536 static __be32
1537 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1538                       struct nfsd4_compound_state *cstate,
1539                       struct nfsd4_copy *copy, struct vfsmount **mount)
1540 {
1541         struct svc_fh *s_fh = NULL;
1542         stateid_t *s_stid = &copy->cp_src_stateid;
1543         __be32 status = nfserr_inval;
1544
1545         /* Verify the destination stateid and set dst struct file*/
1546         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1547                                             &copy->cp_dst_stateid,
1548                                             WR_STATE, &copy->nf_dst, NULL);
1549         if (status)
1550                 goto out;
1551
1552         status = nfsd4_interssc_connect(&copy->cp_src, rqstp, mount);
1553         if (status)
1554                 goto out;
1555
1556         s_fh = &cstate->save_fh;
1557
1558         copy->c_fh.size = s_fh->fh_handle.fh_size;
1559         memcpy(copy->c_fh.data, &s_fh->fh_handle.fh_raw, copy->c_fh.size);
1560         copy->stateid.seqid = cpu_to_be32(s_stid->si_generation);
1561         memcpy(copy->stateid.other, (void *)&s_stid->si_opaque,
1562                sizeof(stateid_opaque_t));
1563
1564         status = 0;
1565 out:
1566         return status;
1567 }
1568
1569 static void
1570 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1571                         struct nfsd_file *dst)
1572 {
1573         bool found = false;
1574         long timeout;
1575         struct nfsd4_ssc_umount_item *tmp;
1576         struct nfsd4_ssc_umount_item *ni = NULL;
1577         struct nfsd_net *nn = net_generic(dst->nf_net, nfsd_net_id);
1578
1579         nfs42_ssc_close(src->nf_file);
1580         nfsd_file_put(dst);
1581         fput(src->nf_file);
1582
1583         if (!nn) {
1584                 mntput(ss_mnt);
1585                 return;
1586         }
1587         spin_lock(&nn->nfsd_ssc_lock);
1588         timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
1589         list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
1590                 if (ni->nsui_vfsmount->mnt_sb == ss_mnt->mnt_sb) {
1591                         list_del(&ni->nsui_list);
1592                         /*
1593                          * vfsmount can be shared by multiple exports,
1594                          * decrement refcnt. If the count drops to 1 it
1595                          * will be unmounted when nsui_expire expires.
1596                          */
1597                         refcount_dec(&ni->nsui_refcnt);
1598                         ni->nsui_expire = jiffies + timeout;
1599                         list_add_tail(&ni->nsui_list, &nn->nfsd_ssc_mount_list);
1600                         found = true;
1601                         break;
1602                 }
1603         }
1604         spin_unlock(&nn->nfsd_ssc_lock);
1605         if (!found) {
1606                 mntput(ss_mnt);
1607                 return;
1608         }
1609 }
1610
1611 #else /* CONFIG_NFSD_V4_2_INTER_SSC */
1612
1613 static __be32
1614 nfsd4_setup_inter_ssc(struct svc_rqst *rqstp,
1615                       struct nfsd4_compound_state *cstate,
1616                       struct nfsd4_copy *copy,
1617                       struct vfsmount **mount)
1618 {
1619         *mount = NULL;
1620         return nfserr_inval;
1621 }
1622
1623 static void
1624 nfsd4_cleanup_inter_ssc(struct vfsmount *ss_mnt, struct nfsd_file *src,
1625                         struct nfsd_file *dst)
1626 {
1627 }
1628
1629 static void
1630 nfsd4_interssc_disconnect(struct vfsmount *ss_mnt)
1631 {
1632 }
1633
1634 static struct file *nfs42_ssc_open(struct vfsmount *ss_mnt,
1635                                    struct nfs_fh *src_fh,
1636                                    nfs4_stateid *stateid)
1637 {
1638         return NULL;
1639 }
1640 #endif /* CONFIG_NFSD_V4_2_INTER_SSC */
1641
1642 static __be32
1643 nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
1644                       struct nfsd4_compound_state *cstate,
1645                       struct nfsd4_copy *copy)
1646 {
1647         return nfsd4_verify_copy(rqstp, cstate, &copy->cp_src_stateid,
1648                                  &copy->nf_src, &copy->cp_dst_stateid,
1649                                  &copy->nf_dst);
1650 }
1651
1652 static void
1653 nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
1654 {
1655         nfsd_file_put(src);
1656         nfsd_file_put(dst);
1657 }
1658
1659 static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
1660 {
1661         struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb);
1662
1663         nfs4_put_copy(copy);
1664 }
1665
1666 static int nfsd4_cb_offload_done(struct nfsd4_callback *cb,
1667                                  struct rpc_task *task)
1668 {
1669         return 1;
1670 }
1671
1672 static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = {
1673         .release = nfsd4_cb_offload_release,
1674         .done = nfsd4_cb_offload_done
1675 };
1676
1677 static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync)
1678 {
1679         copy->cp_res.wr_stable_how =
1680                 copy->committed ? NFS_FILE_SYNC : NFS_UNSTABLE;
1681         copy->cp_synchronous = sync;
1682         gen_boot_verifier(&copy->cp_res.wr_verifier, copy->cp_clp->net);
1683 }
1684
1685 static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy)
1686 {
1687         struct file *dst = copy->nf_dst->nf_file;
1688         struct file *src = copy->nf_src->nf_file;
1689         errseq_t since;
1690         ssize_t bytes_copied = 0;
1691         u64 bytes_total = copy->cp_count;
1692         u64 src_pos = copy->cp_src_pos;
1693         u64 dst_pos = copy->cp_dst_pos;
1694         int status;
1695
1696         /* See RFC 7862 p.67: */
1697         if (bytes_total == 0)
1698                 bytes_total = ULLONG_MAX;
1699         do {
1700                 if (kthread_should_stop())
1701                         break;
1702                 bytes_copied = nfsd_copy_file_range(src, src_pos, dst, dst_pos,
1703                                                     bytes_total);
1704                 if (bytes_copied <= 0)
1705                         break;
1706                 bytes_total -= bytes_copied;
1707                 copy->cp_res.wr_bytes_written += bytes_copied;
1708                 src_pos += bytes_copied;
1709                 dst_pos += bytes_copied;
1710         } while (bytes_total > 0 && !copy->cp_synchronous);
1711         /* for a non-zero asynchronous copy do a commit of data */
1712         if (!copy->cp_synchronous && copy->cp_res.wr_bytes_written > 0) {
1713                 since = READ_ONCE(dst->f_wb_err);
1714                 status = vfs_fsync_range(dst, copy->cp_dst_pos,
1715                                          copy->cp_res.wr_bytes_written, 0);
1716                 if (!status)
1717                         status = filemap_check_wb_err(dst->f_mapping, since);
1718                 if (!status)
1719                         copy->committed = true;
1720         }
1721         return bytes_copied;
1722 }
1723
1724 static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync)
1725 {
1726         __be32 status;
1727         ssize_t bytes;
1728
1729         bytes = _nfsd_copy_file_range(copy);
1730         /* for async copy, we ignore the error, client can always retry
1731          * to get the error
1732          */
1733         if (bytes < 0 && !copy->cp_res.wr_bytes_written)
1734                 status = nfserrno(bytes);
1735         else {
1736                 nfsd4_init_copy_res(copy, sync);
1737                 status = nfs_ok;
1738         }
1739
1740         if (!copy->cp_intra) /* Inter server SSC */
1741                 nfsd4_cleanup_inter_ssc(copy->ss_mnt, copy->nf_src,
1742                                         copy->nf_dst);
1743         else
1744                 nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
1745
1746         return status;
1747 }
1748
1749 static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
1750 {
1751         dst->cp_src_pos = src->cp_src_pos;
1752         dst->cp_dst_pos = src->cp_dst_pos;
1753         dst->cp_count = src->cp_count;
1754         dst->cp_synchronous = src->cp_synchronous;
1755         memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res));
1756         memcpy(&dst->fh, &src->fh, sizeof(src->fh));
1757         dst->cp_clp = src->cp_clp;
1758         dst->nf_dst = nfsd_file_get(src->nf_dst);
1759         dst->cp_intra = src->cp_intra;
1760         if (src->cp_intra) /* for inter, file_src doesn't exist yet */
1761                 dst->nf_src = nfsd_file_get(src->nf_src);
1762
1763         memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid));
1764         memcpy(&dst->cp_src, &src->cp_src, sizeof(struct nl4_server));
1765         memcpy(&dst->stateid, &src->stateid, sizeof(src->stateid));
1766         memcpy(&dst->c_fh, &src->c_fh, sizeof(src->c_fh));
1767         dst->ss_mnt = src->ss_mnt;
1768 }
1769
1770 static void cleanup_async_copy(struct nfsd4_copy *copy)
1771 {
1772         nfs4_free_copy_state(copy);
1773         nfsd_file_put(copy->nf_dst);
1774         if (copy->cp_intra)
1775                 nfsd_file_put(copy->nf_src);
1776         spin_lock(&copy->cp_clp->async_lock);
1777         list_del(&copy->copies);
1778         spin_unlock(&copy->cp_clp->async_lock);
1779         nfs4_put_copy(copy);
1780 }
1781
1782 static int nfsd4_do_async_copy(void *data)
1783 {
1784         struct nfsd4_copy *copy = (struct nfsd4_copy *)data;
1785         struct nfsd4_copy *cb_copy;
1786
1787         if (!copy->cp_intra) { /* Inter server SSC */
1788                 copy->nf_src = kzalloc(sizeof(struct nfsd_file), GFP_KERNEL);
1789                 if (!copy->nf_src) {
1790                         copy->nfserr = nfserr_serverfault;
1791                         nfsd4_interssc_disconnect(copy->ss_mnt);
1792                         goto do_callback;
1793                 }
1794                 copy->nf_src->nf_file = nfs42_ssc_open(copy->ss_mnt, &copy->c_fh,
1795                                               &copy->stateid);
1796                 if (IS_ERR(copy->nf_src->nf_file)) {
1797                         copy->nfserr = nfserr_offload_denied;
1798                         nfsd4_interssc_disconnect(copy->ss_mnt);
1799                         goto do_callback;
1800                 }
1801         }
1802
1803         copy->nfserr = nfsd4_do_copy(copy, 0);
1804 do_callback:
1805         cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1806         if (!cb_copy)
1807                 goto out;
1808         refcount_set(&cb_copy->refcount, 1);
1809         memcpy(&cb_copy->cp_res, &copy->cp_res, sizeof(copy->cp_res));
1810         cb_copy->cp_clp = copy->cp_clp;
1811         cb_copy->nfserr = copy->nfserr;
1812         memcpy(&cb_copy->fh, &copy->fh, sizeof(copy->fh));
1813         nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp,
1814                         &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD);
1815         trace_nfsd_cb_offload(copy->cp_clp, &copy->cp_res.cb_stateid,
1816                               &copy->fh, copy->cp_count, copy->nfserr);
1817         nfsd4_run_cb(&cb_copy->cp_cb);
1818 out:
1819         if (!copy->cp_intra)
1820                 kfree(copy->nf_src);
1821         cleanup_async_copy(copy);
1822         return 0;
1823 }
1824
1825 static __be32
1826 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1827                 union nfsd4_op_u *u)
1828 {
1829         struct nfsd4_copy *copy = &u->copy;
1830         __be32 status;
1831         struct nfsd4_copy *async_copy = NULL;
1832
1833         if (!copy->cp_intra) { /* Inter server SSC */
1834                 if (!inter_copy_offload_enable || copy->cp_synchronous) {
1835                         status = nfserr_notsupp;
1836                         goto out;
1837                 }
1838                 status = nfsd4_setup_inter_ssc(rqstp, cstate, copy,
1839                                 &copy->ss_mnt);
1840                 if (status)
1841                         return nfserr_offload_denied;
1842         } else {
1843                 status = nfsd4_setup_intra_ssc(rqstp, cstate, copy);
1844                 if (status)
1845                         return status;
1846         }
1847
1848         copy->cp_clp = cstate->clp;
1849         memcpy(&copy->fh, &cstate->current_fh.fh_handle,
1850                 sizeof(struct knfsd_fh));
1851         if (!copy->cp_synchronous) {
1852                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1853
1854                 status = nfserrno(-ENOMEM);
1855                 async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL);
1856                 if (!async_copy)
1857                         goto out_err;
1858                 if (!nfs4_init_copy_state(nn, copy))
1859                         goto out_err;
1860                 refcount_set(&async_copy->refcount, 1);
1861                 memcpy(&copy->cp_res.cb_stateid, &copy->cp_stateid.stid,
1862                         sizeof(copy->cp_res.cb_stateid));
1863                 dup_copy_fields(copy, async_copy);
1864                 async_copy->copy_task = kthread_create(nfsd4_do_async_copy,
1865                                 async_copy, "%s", "copy thread");
1866                 if (IS_ERR(async_copy->copy_task))
1867                         goto out_err;
1868                 spin_lock(&async_copy->cp_clp->async_lock);
1869                 list_add(&async_copy->copies,
1870                                 &async_copy->cp_clp->async_copies);
1871                 spin_unlock(&async_copy->cp_clp->async_lock);
1872                 wake_up_process(async_copy->copy_task);
1873                 status = nfs_ok;
1874         } else {
1875                 status = nfsd4_do_copy(copy, 1);
1876         }
1877 out:
1878         return status;
1879 out_err:
1880         if (async_copy)
1881                 cleanup_async_copy(async_copy);
1882         status = nfserrno(-ENOMEM);
1883         if (!copy->cp_intra)
1884                 nfsd4_interssc_disconnect(copy->ss_mnt);
1885         goto out;
1886 }
1887
1888 struct nfsd4_copy *
1889 find_async_copy(struct nfs4_client *clp, stateid_t *stateid)
1890 {
1891         struct nfsd4_copy *copy;
1892
1893         spin_lock(&clp->async_lock);
1894         list_for_each_entry(copy, &clp->async_copies, copies) {
1895                 if (memcmp(&copy->cp_stateid.stid, stateid, NFS4_STATEID_SIZE))
1896                         continue;
1897                 refcount_inc(&copy->refcount);
1898                 spin_unlock(&clp->async_lock);
1899                 return copy;
1900         }
1901         spin_unlock(&clp->async_lock);
1902         return NULL;
1903 }
1904
1905 static __be32
1906 nfsd4_offload_cancel(struct svc_rqst *rqstp,
1907                      struct nfsd4_compound_state *cstate,
1908                      union nfsd4_op_u *u)
1909 {
1910         struct nfsd4_offload_status *os = &u->offload_status;
1911         struct nfsd4_copy *copy;
1912         struct nfs4_client *clp = cstate->clp;
1913
1914         copy = find_async_copy(clp, &os->stateid);
1915         if (!copy) {
1916                 struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1917
1918                 return manage_cpntf_state(nn, &os->stateid, clp, NULL);
1919         } else
1920                 nfsd4_stop_copy(copy);
1921
1922         return nfs_ok;
1923 }
1924
1925 static __be32
1926 nfsd4_copy_notify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1927                   union nfsd4_op_u *u)
1928 {
1929         struct nfsd4_copy_notify *cn = &u->copy_notify;
1930         __be32 status;
1931         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1932         struct nfs4_stid *stid;
1933         struct nfs4_cpntf_state *cps;
1934         struct nfs4_client *clp = cstate->clp;
1935
1936         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1937                                         &cn->cpn_src_stateid, RD_STATE, NULL,
1938                                         &stid);
1939         if (status)
1940                 return status;
1941
1942         cn->cpn_sec = nn->nfsd4_lease;
1943         cn->cpn_nsec = 0;
1944
1945         status = nfserrno(-ENOMEM);
1946         cps = nfs4_alloc_init_cpntf_state(nn, stid);
1947         if (!cps)
1948                 goto out;
1949         memcpy(&cn->cpn_cnr_stateid, &cps->cp_stateid.stid, sizeof(stateid_t));
1950         memcpy(&cps->cp_p_stateid, &stid->sc_stateid, sizeof(stateid_t));
1951         memcpy(&cps->cp_p_clid, &clp->cl_clientid, sizeof(clientid_t));
1952
1953         /* For now, only return one server address in cpn_src, the
1954          * address used by the client to connect to this server.
1955          */
1956         cn->cpn_src.nl4_type = NL4_NETADDR;
1957         status = nfsd4_set_netaddr((struct sockaddr *)&rqstp->rq_daddr,
1958                                  &cn->cpn_src.u.nl4_addr);
1959         WARN_ON_ONCE(status);
1960         if (status) {
1961                 nfs4_put_cpntf_state(nn, cps);
1962                 goto out;
1963         }
1964 out:
1965         nfs4_put_stid(stid);
1966         return status;
1967 }
1968
1969 static __be32
1970 nfsd4_fallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
1971                 struct nfsd4_fallocate *fallocate, int flags)
1972 {
1973         __be32 status;
1974         struct nfsd_file *nf;
1975
1976         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
1977                                             &fallocate->falloc_stateid,
1978                                             WR_STATE, &nf, NULL);
1979         if (status != nfs_ok) {
1980                 dprintk("NFSD: nfsd4_fallocate: couldn't process stateid!\n");
1981                 return status;
1982         }
1983
1984         status = nfsd4_vfs_fallocate(rqstp, &cstate->current_fh, nf->nf_file,
1985                                      fallocate->falloc_offset,
1986                                      fallocate->falloc_length,
1987                                      flags);
1988         nfsd_file_put(nf);
1989         return status;
1990 }
1991 static __be32
1992 nfsd4_offload_status(struct svc_rqst *rqstp,
1993                      struct nfsd4_compound_state *cstate,
1994                      union nfsd4_op_u *u)
1995 {
1996         struct nfsd4_offload_status *os = &u->offload_status;
1997         __be32 status = 0;
1998         struct nfsd4_copy *copy;
1999         struct nfs4_client *clp = cstate->clp;
2000
2001         copy = find_async_copy(clp, &os->stateid);
2002         if (copy) {
2003                 os->count = copy->cp_res.wr_bytes_written;
2004                 nfs4_put_copy(copy);
2005         } else
2006                 status = nfserr_bad_stateid;
2007
2008         return status;
2009 }
2010
2011 static __be32
2012 nfsd4_allocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2013                union nfsd4_op_u *u)
2014 {
2015         return nfsd4_fallocate(rqstp, cstate, &u->allocate, 0);
2016 }
2017
2018 static __be32
2019 nfsd4_deallocate(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2020                  union nfsd4_op_u *u)
2021 {
2022         return nfsd4_fallocate(rqstp, cstate, &u->deallocate,
2023                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE);
2024 }
2025
2026 static __be32
2027 nfsd4_seek(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2028            union nfsd4_op_u *u)
2029 {
2030         struct nfsd4_seek *seek = &u->seek;
2031         int whence;
2032         __be32 status;
2033         struct nfsd_file *nf;
2034
2035         status = nfs4_preprocess_stateid_op(rqstp, cstate, &cstate->current_fh,
2036                                             &seek->seek_stateid,
2037                                             RD_STATE, &nf, NULL);
2038         if (status) {
2039                 dprintk("NFSD: nfsd4_seek: couldn't process stateid!\n");
2040                 return status;
2041         }
2042
2043         switch (seek->seek_whence) {
2044         case NFS4_CONTENT_DATA:
2045                 whence = SEEK_DATA;
2046                 break;
2047         case NFS4_CONTENT_HOLE:
2048                 whence = SEEK_HOLE;
2049                 break;
2050         default:
2051                 status = nfserr_union_notsupp;
2052                 goto out;
2053         }
2054
2055         /*
2056          * Note:  This call does change file->f_pos, but nothing in NFSD
2057          *        should ever file->f_pos.
2058          */
2059         seek->seek_pos = vfs_llseek(nf->nf_file, seek->seek_offset, whence);
2060         if (seek->seek_pos < 0)
2061                 status = nfserrno(seek->seek_pos);
2062         else if (seek->seek_pos >= i_size_read(file_inode(nf->nf_file)))
2063                 seek->seek_eof = true;
2064
2065 out:
2066         nfsd_file_put(nf);
2067         return status;
2068 }
2069
2070 /* This routine never returns NFS_OK!  If there are no other errors, it
2071  * will return NFSERR_SAME or NFSERR_NOT_SAME depending on whether the
2072  * attributes matched.  VERIFY is implemented by mapping NFSERR_SAME
2073  * to NFS_OK after the call; NVERIFY by mapping NFSERR_NOT_SAME to NFS_OK.
2074  */
2075 static __be32
2076 _nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2077              struct nfsd4_verify *verify)
2078 {
2079         __be32 *buf, *p;
2080         int count;
2081         __be32 status;
2082
2083         status = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_NOP);
2084         if (status)
2085                 return status;
2086
2087         status = check_attr_support(rqstp, cstate, verify->ve_bmval, NULL);
2088         if (status)
2089                 return status;
2090
2091         if ((verify->ve_bmval[0] & FATTR4_WORD0_RDATTR_ERROR)
2092             || (verify->ve_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1))
2093                 return nfserr_inval;
2094         if (verify->ve_attrlen & 3)
2095                 return nfserr_inval;
2096
2097         /* count in words:
2098          *   bitmap_len(1) + bitmap(2) + attr_len(1) = 4
2099          */
2100         count = 4 + (verify->ve_attrlen >> 2);
2101         buf = kmalloc(count << 2, GFP_KERNEL);
2102         if (!buf)
2103                 return nfserr_jukebox;
2104
2105         p = buf;
2106         status = nfsd4_encode_fattr_to_buf(&p, count, &cstate->current_fh,
2107                                     cstate->current_fh.fh_export,
2108                                     cstate->current_fh.fh_dentry,
2109                                     verify->ve_bmval,
2110                                     rqstp, 0);
2111         /*
2112          * If nfsd4_encode_fattr() ran out of space, assume that's because
2113          * the attributes are longer (hence different) than those given:
2114          */
2115         if (status == nfserr_resource)
2116                 status = nfserr_not_same;
2117         if (status)
2118                 goto out_kfree;
2119
2120         /* skip bitmap */
2121         p = buf + 1 + ntohl(buf[0]);
2122         status = nfserr_not_same;
2123         if (ntohl(*p++) != verify->ve_attrlen)
2124                 goto out_kfree;
2125         if (!memcmp(p, verify->ve_attrval, verify->ve_attrlen))
2126                 status = nfserr_same;
2127
2128 out_kfree:
2129         kfree(buf);
2130         return status;
2131 }
2132
2133 static __be32
2134 nfsd4_nverify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2135               union nfsd4_op_u *u)
2136 {
2137         __be32 status;
2138
2139         status = _nfsd4_verify(rqstp, cstate, &u->verify);
2140         return status == nfserr_not_same ? nfs_ok : status;
2141 }
2142
2143 static __be32
2144 nfsd4_verify(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2145              union nfsd4_op_u *u)
2146 {
2147         __be32 status;
2148
2149         status = _nfsd4_verify(rqstp, cstate, &u->nverify);
2150         return status == nfserr_same ? nfs_ok : status;
2151 }
2152
2153 #ifdef CONFIG_NFSD_PNFS
2154 static const struct nfsd4_layout_ops *
2155 nfsd4_layout_verify(struct svc_export *exp, unsigned int layout_type)
2156 {
2157         if (!exp->ex_layout_types) {
2158                 dprintk("%s: export does not support pNFS\n", __func__);
2159                 return NULL;
2160         }
2161
2162         if (layout_type >= LAYOUT_TYPE_MAX ||
2163             !(exp->ex_layout_types & (1 << layout_type))) {
2164                 dprintk("%s: layout type %d not supported\n",
2165                         __func__, layout_type);
2166                 return NULL;
2167         }
2168
2169         return nfsd4_layout_ops[layout_type];
2170 }
2171
2172 static __be32
2173 nfsd4_getdeviceinfo(struct svc_rqst *rqstp,
2174                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2175 {
2176         struct nfsd4_getdeviceinfo *gdp = &u->getdeviceinfo;
2177         const struct nfsd4_layout_ops *ops;
2178         struct nfsd4_deviceid_map *map;
2179         struct svc_export *exp;
2180         __be32 nfserr;
2181
2182         dprintk("%s: layout_type %u dev_id [0x%llx:0x%x] maxcnt %u\n",
2183                __func__,
2184                gdp->gd_layout_type,
2185                gdp->gd_devid.fsid_idx, gdp->gd_devid.generation,
2186                gdp->gd_maxcount);
2187
2188         map = nfsd4_find_devid_map(gdp->gd_devid.fsid_idx);
2189         if (!map) {
2190                 dprintk("%s: couldn't find device ID to export mapping!\n",
2191                         __func__);
2192                 return nfserr_noent;
2193         }
2194
2195         exp = rqst_exp_find(rqstp, map->fsid_type, map->fsid);
2196         if (IS_ERR(exp)) {
2197                 dprintk("%s: could not find device id\n", __func__);
2198                 return nfserr_noent;
2199         }
2200
2201         nfserr = nfserr_layoutunavailable;
2202         ops = nfsd4_layout_verify(exp, gdp->gd_layout_type);
2203         if (!ops)
2204                 goto out;
2205
2206         nfserr = nfs_ok;
2207         if (gdp->gd_maxcount != 0) {
2208                 nfserr = ops->proc_getdeviceinfo(exp->ex_path.mnt->mnt_sb,
2209                                 rqstp, cstate->clp, gdp);
2210         }
2211
2212         gdp->gd_notify_types &= ops->notify_types;
2213 out:
2214         exp_put(exp);
2215         return nfserr;
2216 }
2217
2218 static void
2219 nfsd4_getdeviceinfo_release(union nfsd4_op_u *u)
2220 {
2221         kfree(u->getdeviceinfo.gd_device);
2222 }
2223
2224 static __be32
2225 nfsd4_layoutget(struct svc_rqst *rqstp,
2226                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2227 {
2228         struct nfsd4_layoutget *lgp = &u->layoutget;
2229         struct svc_fh *current_fh = &cstate->current_fh;
2230         const struct nfsd4_layout_ops *ops;
2231         struct nfs4_layout_stateid *ls;
2232         __be32 nfserr;
2233         int accmode = NFSD_MAY_READ_IF_EXEC;
2234
2235         switch (lgp->lg_seg.iomode) {
2236         case IOMODE_READ:
2237                 accmode |= NFSD_MAY_READ;
2238                 break;
2239         case IOMODE_RW:
2240                 accmode |= NFSD_MAY_READ | NFSD_MAY_WRITE;
2241                 break;
2242         default:
2243                 dprintk("%s: invalid iomode %d\n",
2244                         __func__, lgp->lg_seg.iomode);
2245                 nfserr = nfserr_badiomode;
2246                 goto out;
2247         }
2248
2249         nfserr = fh_verify(rqstp, current_fh, 0, accmode);
2250         if (nfserr)
2251                 goto out;
2252
2253         nfserr = nfserr_layoutunavailable;
2254         ops = nfsd4_layout_verify(current_fh->fh_export, lgp->lg_layout_type);
2255         if (!ops)
2256                 goto out;
2257
2258         /*
2259          * Verify minlength and range as per RFC5661:
2260          *  o  If loga_length is less than loga_minlength,
2261          *     the metadata server MUST return NFS4ERR_INVAL.
2262          *  o  If the sum of loga_offset and loga_minlength exceeds
2263          *     NFS4_UINT64_MAX, and loga_minlength is not
2264          *     NFS4_UINT64_MAX, the error NFS4ERR_INVAL MUST result.
2265          *  o  If the sum of loga_offset and loga_length exceeds
2266          *     NFS4_UINT64_MAX, and loga_length is not NFS4_UINT64_MAX,
2267          *     the error NFS4ERR_INVAL MUST result.
2268          */
2269         nfserr = nfserr_inval;
2270         if (lgp->lg_seg.length < lgp->lg_minlength ||
2271             (lgp->lg_minlength != NFS4_MAX_UINT64 &&
2272              lgp->lg_minlength > NFS4_MAX_UINT64 - lgp->lg_seg.offset) ||
2273             (lgp->lg_seg.length != NFS4_MAX_UINT64 &&
2274              lgp->lg_seg.length > NFS4_MAX_UINT64 - lgp->lg_seg.offset))
2275                 goto out;
2276         if (lgp->lg_seg.length == 0)
2277                 goto out;
2278
2279         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lgp->lg_sid,
2280                                                 true, lgp->lg_layout_type, &ls);
2281         if (nfserr) {
2282                 trace_nfsd_layout_get_lookup_fail(&lgp->lg_sid);
2283                 goto out;
2284         }
2285
2286         nfserr = nfserr_recallconflict;
2287         if (atomic_read(&ls->ls_stid.sc_file->fi_lo_recalls))
2288                 goto out_put_stid;
2289
2290         nfserr = ops->proc_layoutget(d_inode(current_fh->fh_dentry),
2291                                      current_fh, lgp);
2292         if (nfserr)
2293                 goto out_put_stid;
2294
2295         nfserr = nfsd4_insert_layout(lgp, ls);
2296
2297 out_put_stid:
2298         mutex_unlock(&ls->ls_mutex);
2299         nfs4_put_stid(&ls->ls_stid);
2300 out:
2301         return nfserr;
2302 }
2303
2304 static void
2305 nfsd4_layoutget_release(union nfsd4_op_u *u)
2306 {
2307         kfree(u->layoutget.lg_content);
2308 }
2309
2310 static __be32
2311 nfsd4_layoutcommit(struct svc_rqst *rqstp,
2312                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2313 {
2314         struct nfsd4_layoutcommit *lcp = &u->layoutcommit;
2315         const struct nfsd4_layout_seg *seg = &lcp->lc_seg;
2316         struct svc_fh *current_fh = &cstate->current_fh;
2317         const struct nfsd4_layout_ops *ops;
2318         loff_t new_size = lcp->lc_last_wr + 1;
2319         struct inode *inode;
2320         struct nfs4_layout_stateid *ls;
2321         __be32 nfserr;
2322
2323         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_WRITE);
2324         if (nfserr)
2325                 goto out;
2326
2327         nfserr = nfserr_layoutunavailable;
2328         ops = nfsd4_layout_verify(current_fh->fh_export, lcp->lc_layout_type);
2329         if (!ops)
2330                 goto out;
2331         inode = d_inode(current_fh->fh_dentry);
2332
2333         nfserr = nfserr_inval;
2334         if (new_size <= seg->offset) {
2335                 dprintk("pnfsd: last write before layout segment\n");
2336                 goto out;
2337         }
2338         if (new_size > seg->offset + seg->length) {
2339                 dprintk("pnfsd: last write beyond layout segment\n");
2340                 goto out;
2341         }
2342         if (!lcp->lc_newoffset && new_size > i_size_read(inode)) {
2343                 dprintk("pnfsd: layoutcommit beyond EOF\n");
2344                 goto out;
2345         }
2346
2347         nfserr = nfsd4_preprocess_layout_stateid(rqstp, cstate, &lcp->lc_sid,
2348                                                 false, lcp->lc_layout_type,
2349                                                 &ls);
2350         if (nfserr) {
2351                 trace_nfsd_layout_commit_lookup_fail(&lcp->lc_sid);
2352                 /* fixup error code as per RFC5661 */
2353                 if (nfserr == nfserr_bad_stateid)
2354                         nfserr = nfserr_badlayout;
2355                 goto out;
2356         }
2357
2358         /* LAYOUTCOMMIT does not require any serialization */
2359         mutex_unlock(&ls->ls_mutex);
2360
2361         if (new_size > i_size_read(inode)) {
2362                 lcp->lc_size_chg = 1;
2363                 lcp->lc_newsize = new_size;
2364         } else {
2365                 lcp->lc_size_chg = 0;
2366         }
2367
2368         nfserr = ops->proc_layoutcommit(inode, lcp);
2369         nfs4_put_stid(&ls->ls_stid);
2370 out:
2371         return nfserr;
2372 }
2373
2374 static __be32
2375 nfsd4_layoutreturn(struct svc_rqst *rqstp,
2376                 struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
2377 {
2378         struct nfsd4_layoutreturn *lrp = &u->layoutreturn;
2379         struct svc_fh *current_fh = &cstate->current_fh;
2380         __be32 nfserr;
2381
2382         nfserr = fh_verify(rqstp, current_fh, 0, NFSD_MAY_NOP);
2383         if (nfserr)
2384                 goto out;
2385
2386         nfserr = nfserr_layoutunavailable;
2387         if (!nfsd4_layout_verify(current_fh->fh_export, lrp->lr_layout_type))
2388                 goto out;
2389
2390         switch (lrp->lr_seg.iomode) {
2391         case IOMODE_READ:
2392         case IOMODE_RW:
2393         case IOMODE_ANY:
2394                 break;
2395         default:
2396                 dprintk("%s: invalid iomode %d\n", __func__,
2397                         lrp->lr_seg.iomode);
2398                 nfserr = nfserr_inval;
2399                 goto out;
2400         }
2401
2402         switch (lrp->lr_return_type) {
2403         case RETURN_FILE:
2404                 nfserr = nfsd4_return_file_layouts(rqstp, cstate, lrp);
2405                 break;
2406         case RETURN_FSID:
2407         case RETURN_ALL:
2408                 nfserr = nfsd4_return_client_layouts(rqstp, cstate, lrp);
2409                 break;
2410         default:
2411                 dprintk("%s: invalid return_type %d\n", __func__,
2412                         lrp->lr_return_type);
2413                 nfserr = nfserr_inval;
2414                 break;
2415         }
2416 out:
2417         return nfserr;
2418 }
2419 #endif /* CONFIG_NFSD_PNFS */
2420
2421 static __be32
2422 nfsd4_getxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2423                union nfsd4_op_u *u)
2424 {
2425         struct nfsd4_getxattr *getxattr = &u->getxattr;
2426
2427         return nfsd_getxattr(rqstp, &cstate->current_fh,
2428                              getxattr->getxa_name, &getxattr->getxa_buf,
2429                              &getxattr->getxa_len);
2430 }
2431
2432 static __be32
2433 nfsd4_setxattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2434            union nfsd4_op_u *u)
2435 {
2436         struct nfsd4_setxattr *setxattr = &u->setxattr;
2437         __be32 ret;
2438
2439         if (opens_in_grace(SVC_NET(rqstp)))
2440                 return nfserr_grace;
2441
2442         ret = nfsd_setxattr(rqstp, &cstate->current_fh, setxattr->setxa_name,
2443                             setxattr->setxa_buf, setxattr->setxa_len,
2444                             setxattr->setxa_flags);
2445
2446         if (!ret)
2447                 set_change_info(&setxattr->setxa_cinfo, &cstate->current_fh);
2448
2449         return ret;
2450 }
2451
2452 static __be32
2453 nfsd4_listxattrs(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2454            union nfsd4_op_u *u)
2455 {
2456         /*
2457          * Get the entire list, then copy out only the user attributes
2458          * in the encode function.
2459          */
2460         return nfsd_listxattr(rqstp, &cstate->current_fh,
2461                              &u->listxattrs.lsxa_buf, &u->listxattrs.lsxa_len);
2462 }
2463
2464 static __be32
2465 nfsd4_removexattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
2466            union nfsd4_op_u *u)
2467 {
2468         struct nfsd4_removexattr *removexattr = &u->removexattr;
2469         __be32 ret;
2470
2471         if (opens_in_grace(SVC_NET(rqstp)))
2472                 return nfserr_grace;
2473
2474         ret = nfsd_removexattr(rqstp, &cstate->current_fh,
2475             removexattr->rmxa_name);
2476
2477         if (!ret)
2478                 set_change_info(&removexattr->rmxa_cinfo, &cstate->current_fh);
2479
2480         return ret;
2481 }
2482
2483 /*
2484  * NULL call.
2485  */
2486 static __be32
2487 nfsd4_proc_null(struct svc_rqst *rqstp)
2488 {
2489         return rpc_success;
2490 }
2491
2492 static inline void nfsd4_increment_op_stats(u32 opnum)
2493 {
2494         if (opnum >= FIRST_NFS4_OP && opnum <= LAST_NFS4_OP)
2495                 percpu_counter_inc(&nfsdstats.counter[NFSD_STATS_NFS4_OP(opnum)]);
2496 }
2497
2498 static const struct nfsd4_operation nfsd4_ops[];
2499
2500 static const char *nfsd4_op_name(unsigned opnum);
2501
2502 /*
2503  * Enforce NFSv4.1 COMPOUND ordering rules:
2504  *
2505  * Also note, enforced elsewhere:
2506  *      - SEQUENCE other than as first op results in
2507  *        NFS4ERR_SEQUENCE_POS. (Enforced in nfsd4_sequence().)
2508  *      - BIND_CONN_TO_SESSION must be the only op in its compound.
2509  *        (Enforced in nfsd4_bind_conn_to_session().)
2510  *      - DESTROY_SESSION must be the final operation in a compound, if
2511  *        sessionid's in SEQUENCE and DESTROY_SESSION are the same.
2512  *        (Enforced in nfsd4_destroy_session().)
2513  */
2514 static __be32 nfs41_check_op_ordering(struct nfsd4_compoundargs *args)
2515 {
2516         struct nfsd4_op *first_op = &args->ops[0];
2517
2518         /* These ordering requirements don't apply to NFSv4.0: */
2519         if (args->minorversion == 0)
2520                 return nfs_ok;
2521         /* This is weird, but OK, not our problem: */
2522         if (args->opcnt == 0)
2523                 return nfs_ok;
2524         if (first_op->status == nfserr_op_illegal)
2525                 return nfs_ok;
2526         if (!(nfsd4_ops[first_op->opnum].op_flags & ALLOWED_AS_FIRST_OP))
2527                 return nfserr_op_not_in_session;
2528         if (first_op->opnum == OP_SEQUENCE)
2529                 return nfs_ok;
2530         /*
2531          * So first_op is something allowed outside a session, like
2532          * EXCHANGE_ID; but then it has to be the only op in the
2533          * compound:
2534          */
2535         if (args->opcnt != 1)
2536                 return nfserr_not_only_op;
2537         return nfs_ok;
2538 }
2539
2540 const struct nfsd4_operation *OPDESC(struct nfsd4_op *op)
2541 {
2542         return &nfsd4_ops[op->opnum];
2543 }
2544
2545 bool nfsd4_cache_this_op(struct nfsd4_op *op)
2546 {
2547         if (op->opnum == OP_ILLEGAL)
2548                 return false;
2549         return OPDESC(op)->op_flags & OP_CACHEME;
2550 }
2551
2552 static bool need_wrongsec_check(struct svc_rqst *rqstp)
2553 {
2554         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2555         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
2556         struct nfsd4_op *this = &argp->ops[resp->opcnt - 1];
2557         struct nfsd4_op *next = &argp->ops[resp->opcnt];
2558         const struct nfsd4_operation *thisd = OPDESC(this);
2559         const struct nfsd4_operation *nextd;
2560
2561         /*
2562          * Most ops check wronsec on our own; only the putfh-like ops
2563          * have special rules.
2564          */
2565         if (!(thisd->op_flags & OP_IS_PUTFH_LIKE))
2566                 return false;
2567         /*
2568          * rfc 5661 2.6.3.1.1.6: don't bother erroring out a
2569          * put-filehandle operation if we're not going to use the
2570          * result:
2571          */
2572         if (argp->opcnt == resp->opcnt)
2573                 return false;
2574         if (next->opnum == OP_ILLEGAL)
2575                 return false;
2576         nextd = OPDESC(next);
2577         /*
2578          * Rest of 2.6.3.1.1: certain operations will return WRONGSEC
2579          * errors themselves as necessary; others should check for them
2580          * now:
2581          */
2582         return !(nextd->op_flags & OP_HANDLES_WRONGSEC);
2583 }
2584
2585 #ifdef CONFIG_NFSD_V4_2_INTER_SSC
2586 static void
2587 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2588 {
2589         struct nfsd4_op *op, *current_op = NULL, *saved_op = NULL;
2590         struct nfsd4_copy *copy;
2591         struct nfsd4_putfh *putfh;
2592         int i;
2593
2594         /* traverse all operation and if it's a COPY compound, mark the
2595          * source filehandle to skip verification
2596          */
2597         for (i = 0; i < args->opcnt; i++) {
2598                 op = &args->ops[i];
2599                 if (op->opnum == OP_PUTFH)
2600                         current_op = op;
2601                 else if (op->opnum == OP_SAVEFH)
2602                         saved_op = current_op;
2603                 else if (op->opnum == OP_RESTOREFH)
2604                         current_op = saved_op;
2605                 else if (op->opnum == OP_COPY) {
2606                         copy = (struct nfsd4_copy *)&op->u;
2607                         if (!saved_op) {
2608                                 op->status = nfserr_nofilehandle;
2609                                 return;
2610                         }
2611                         putfh = (struct nfsd4_putfh *)&saved_op->u;
2612                         if (!copy->cp_intra)
2613                                 putfh->no_verify = true;
2614                 }
2615         }
2616 }
2617 #else
2618 static void
2619 check_if_stalefh_allowed(struct nfsd4_compoundargs *args)
2620 {
2621 }
2622 #endif
2623
2624 /*
2625  * COMPOUND call.
2626  */
2627 static __be32
2628 nfsd4_proc_compound(struct svc_rqst *rqstp)
2629 {
2630         struct nfsd4_compoundargs *args = rqstp->rq_argp;
2631         struct nfsd4_compoundres *resp = rqstp->rq_resp;
2632         struct nfsd4_op *op;
2633         struct nfsd4_compound_state *cstate = &resp->cstate;
2634         struct svc_fh *current_fh = &cstate->current_fh;
2635         struct svc_fh *save_fh = &cstate->save_fh;
2636         struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2637         __be32          status;
2638
2639         resp->xdr = &rqstp->rq_res_stream;
2640         resp->statusp = resp->xdr->p;
2641
2642         /* reserve space for: NFS status code */
2643         xdr_reserve_space(resp->xdr, XDR_UNIT);
2644
2645         /* reserve space for: taglen, tag, and opcnt */
2646         xdr_reserve_space(resp->xdr, XDR_UNIT * 2 + args->taglen);
2647         resp->taglen = args->taglen;
2648         resp->tag = args->tag;
2649         resp->rqstp = rqstp;
2650         cstate->minorversion = args->minorversion;
2651         fh_init(current_fh, NFS4_FHSIZE);
2652         fh_init(save_fh, NFS4_FHSIZE);
2653
2654         /*
2655          * Don't use the deferral mechanism for NFSv4; compounds make it
2656          * too hard to avoid non-idempotency problems.
2657          */
2658         __clear_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2659
2660         /*
2661          * According to RFC3010, this takes precedence over all other errors.
2662          */
2663         status = nfserr_minor_vers_mismatch;
2664         if (nfsd_minorversion(nn, args->minorversion, NFSD_TEST) <= 0)
2665                 goto out;
2666         status = nfserr_resource;
2667         if (args->opcnt > NFSD_MAX_OPS_PER_COMPOUND)
2668                 goto out;
2669
2670         status = nfs41_check_op_ordering(args);
2671         if (status) {
2672                 op = &args->ops[0];
2673                 op->status = status;
2674                 resp->opcnt = 1;
2675                 goto encode_op;
2676         }
2677         check_if_stalefh_allowed(args);
2678
2679         rqstp->rq_lease_breaker = (void **)&cstate->clp;
2680
2681         trace_nfsd_compound(rqstp, args->opcnt);
2682         while (!status && resp->opcnt < args->opcnt) {
2683                 op = &args->ops[resp->opcnt++];
2684
2685                 /*
2686                  * The XDR decode routines may have pre-set op->status;
2687                  * for example, if there is a miscellaneous XDR error
2688                  * it will be set to nfserr_bad_xdr.
2689                  */
2690                 if (op->status) {
2691                         if (op->opnum == OP_OPEN)
2692                                 op->status = nfsd4_open_omfg(rqstp, cstate, op);
2693                         goto encode_op;
2694                 }
2695                 if (!current_fh->fh_dentry &&
2696                                 !HAS_FH_FLAG(current_fh, NFSD4_FH_FOREIGN)) {
2697                         if (!(op->opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
2698                                 op->status = nfserr_nofilehandle;
2699                                 goto encode_op;
2700                         }
2701                 } else if (current_fh->fh_export &&
2702                            current_fh->fh_export->ex_fslocs.migrated &&
2703                           !(op->opdesc->op_flags & ALLOWED_ON_ABSENT_FS)) {
2704                         op->status = nfserr_moved;
2705                         goto encode_op;
2706                 }
2707
2708                 fh_clear_pre_post_attrs(current_fh);
2709
2710                 /* If op is non-idempotent */
2711                 if (op->opdesc->op_flags & OP_MODIFIES_SOMETHING) {
2712                         /*
2713                          * Don't execute this op if we couldn't encode a
2714                          * succesful reply:
2715                          */
2716                         u32 plen = op->opdesc->op_rsize_bop(rqstp, op);
2717                         /*
2718                          * Plus if there's another operation, make sure
2719                          * we'll have space to at least encode an error:
2720                          */
2721                         if (resp->opcnt < args->opcnt)
2722                                 plen += COMPOUND_ERR_SLACK_SPACE;
2723                         op->status = nfsd4_check_resp_size(resp, plen);
2724                 }
2725
2726                 if (op->status)
2727                         goto encode_op;
2728
2729                 if (op->opdesc->op_get_currentstateid)
2730                         op->opdesc->op_get_currentstateid(cstate, &op->u);
2731                 op->status = op->opdesc->op_func(rqstp, cstate, &op->u);
2732
2733                 /* Only from SEQUENCE */
2734                 if (cstate->status == nfserr_replay_cache) {
2735                         dprintk("%s NFS4.1 replay from cache\n", __func__);
2736                         status = op->status;
2737                         goto out;
2738                 }
2739                 if (!op->status) {
2740                         if (op->opdesc->op_set_currentstateid)
2741                                 op->opdesc->op_set_currentstateid(cstate, &op->u);
2742
2743                         if (op->opdesc->op_flags & OP_CLEAR_STATEID)
2744                                 clear_current_stateid(cstate);
2745
2746                         if (current_fh->fh_export &&
2747                                         need_wrongsec_check(rqstp))
2748                                 op->status = check_nfsd_access(current_fh->fh_export, rqstp);
2749                 }
2750 encode_op:
2751                 if (op->status == nfserr_replay_me) {
2752                         op->replay = &cstate->replay_owner->so_replay;
2753                         nfsd4_encode_replay(resp->xdr, op);
2754                         status = op->status = op->replay->rp_status;
2755                 } else {
2756                         nfsd4_encode_operation(resp, op);
2757                         status = op->status;
2758                 }
2759
2760                 trace_nfsd_compound_status(args->opcnt, resp->opcnt, status,
2761                                            nfsd4_op_name(op->opnum));
2762
2763                 nfsd4_cstate_clear_replay(cstate);
2764                 nfsd4_increment_op_stats(op->opnum);
2765         }
2766
2767         fh_put(current_fh);
2768         fh_put(save_fh);
2769         BUG_ON(cstate->replay_owner);
2770 out:
2771         cstate->status = status;
2772         /* Reset deferral mechanism for RPC deferrals */
2773         __set_bit(RQ_USEDEFERRAL, &rqstp->rq_flags);
2774         return rpc_success;
2775 }
2776
2777 #define op_encode_hdr_size              (2)
2778 #define op_encode_stateid_maxsz         (XDR_QUADLEN(NFS4_STATEID_SIZE))
2779 #define op_encode_verifier_maxsz        (XDR_QUADLEN(NFS4_VERIFIER_SIZE))
2780 #define op_encode_change_info_maxsz     (5)
2781 #define nfs4_fattr_bitmap_maxsz         (4)
2782
2783 /* We'll fall back on returning no lockowner if run out of space: */
2784 #define op_encode_lockowner_maxsz       (0)
2785 #define op_encode_lock_denied_maxsz     (8 + op_encode_lockowner_maxsz)
2786
2787 #define nfs4_owner_maxsz                (1 + XDR_QUADLEN(IDMAP_NAMESZ))
2788
2789 #define op_encode_ace_maxsz             (3 + nfs4_owner_maxsz)
2790 #define op_encode_delegation_maxsz      (1 + op_encode_stateid_maxsz + 1 + \
2791                                          op_encode_ace_maxsz)
2792
2793 #define op_encode_channel_attrs_maxsz   (6 + 1 + 1)
2794
2795 static inline u32 nfsd4_only_status_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2796 {
2797         return (op_encode_hdr_size) * sizeof(__be32);
2798 }
2799
2800 static inline u32 nfsd4_status_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2801 {
2802         return (op_encode_hdr_size + op_encode_stateid_maxsz)* sizeof(__be32);
2803 }
2804
2805 static inline u32 nfsd4_access_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2806 {
2807         /* ac_supported, ac_resp_access */
2808         return (op_encode_hdr_size + 2)* sizeof(__be32);
2809 }
2810
2811 static inline u32 nfsd4_commit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2812 {
2813         return (op_encode_hdr_size + op_encode_verifier_maxsz) * sizeof(__be32);
2814 }
2815
2816 static inline u32 nfsd4_create_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2817 {
2818         return (op_encode_hdr_size + op_encode_change_info_maxsz
2819                 + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2820 }
2821
2822 /*
2823  * Note since this is an idempotent operation we won't insist on failing
2824  * the op prematurely if the estimate is too large.  We may turn off splice
2825  * reads unnecessarily.
2826  */
2827 static inline u32 nfsd4_getattr_rsize(struct svc_rqst *rqstp,
2828                                       struct nfsd4_op *op)
2829 {
2830         u32 *bmap = op->u.getattr.ga_bmval;
2831         u32 bmap0 = bmap[0], bmap1 = bmap[1], bmap2 = bmap[2];
2832         u32 ret = 0;
2833
2834         if (bmap0 & FATTR4_WORD0_ACL)
2835                 return svc_max_payload(rqstp);
2836         if (bmap0 & FATTR4_WORD0_FS_LOCATIONS)
2837                 return svc_max_payload(rqstp);
2838
2839         if (bmap1 & FATTR4_WORD1_OWNER) {
2840                 ret += IDMAP_NAMESZ + 4;
2841                 bmap1 &= ~FATTR4_WORD1_OWNER;
2842         }
2843         if (bmap1 & FATTR4_WORD1_OWNER_GROUP) {
2844                 ret += IDMAP_NAMESZ + 4;
2845                 bmap1 &= ~FATTR4_WORD1_OWNER_GROUP;
2846         }
2847         if (bmap0 & FATTR4_WORD0_FILEHANDLE) {
2848                 ret += NFS4_FHSIZE + 4;
2849                 bmap0 &= ~FATTR4_WORD0_FILEHANDLE;
2850         }
2851         if (bmap2 & FATTR4_WORD2_SECURITY_LABEL) {
2852                 ret += NFS4_MAXLABELLEN + 12;
2853                 bmap2 &= ~FATTR4_WORD2_SECURITY_LABEL;
2854         }
2855         /*
2856          * Largest of remaining attributes are 16 bytes (e.g.,
2857          * supported_attributes)
2858          */
2859         ret += 16 * (hweight32(bmap0) + hweight32(bmap1) + hweight32(bmap2));
2860         /* bitmask, length */
2861         ret += 20;
2862         return ret;
2863 }
2864
2865 static inline u32 nfsd4_getfh_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2866 {
2867         return (op_encode_hdr_size + 1) * sizeof(__be32) + NFS4_FHSIZE;
2868 }
2869
2870 static inline u32 nfsd4_link_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2871 {
2872         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2873                 * sizeof(__be32);
2874 }
2875
2876 static inline u32 nfsd4_lock_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2877 {
2878         return (op_encode_hdr_size + op_encode_lock_denied_maxsz)
2879                 * sizeof(__be32);
2880 }
2881
2882 static inline u32 nfsd4_open_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2883 {
2884         return (op_encode_hdr_size + op_encode_stateid_maxsz
2885                 + op_encode_change_info_maxsz + 1
2886                 + nfs4_fattr_bitmap_maxsz
2887                 + op_encode_delegation_maxsz) * sizeof(__be32);
2888 }
2889
2890 static inline u32 nfsd4_read_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2891 {
2892         u32 maxcount = 0, rlen = 0;
2893
2894         maxcount = svc_max_payload(rqstp);
2895         rlen = min(op->u.read.rd_length, maxcount);
2896
2897         return (op_encode_hdr_size + 2 + XDR_QUADLEN(rlen)) * sizeof(__be32);
2898 }
2899
2900 static inline u32 nfsd4_read_plus_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2901 {
2902         u32 maxcount = svc_max_payload(rqstp);
2903         u32 rlen = min(op->u.read.rd_length, maxcount);
2904         /*
2905          * If we detect that the file changed during hole encoding, then we
2906          * recover by encoding the remaining reply as data. This means we need
2907          * to set aside enough room to encode two data segments.
2908          */
2909         u32 seg_len = 2 * (1 + 2 + 1);
2910
2911         return (op_encode_hdr_size + 2 + seg_len + XDR_QUADLEN(rlen)) * sizeof(__be32);
2912 }
2913
2914 static inline u32 nfsd4_readdir_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2915 {
2916         u32 maxcount = 0, rlen = 0;
2917
2918         maxcount = svc_max_payload(rqstp);
2919         rlen = min(op->u.readdir.rd_maxcount, maxcount);
2920
2921         return (op_encode_hdr_size + op_encode_verifier_maxsz +
2922                 XDR_QUADLEN(rlen)) * sizeof(__be32);
2923 }
2924
2925 static inline u32 nfsd4_readlink_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2926 {
2927         return (op_encode_hdr_size + 1) * sizeof(__be32) + PAGE_SIZE;
2928 }
2929
2930 static inline u32 nfsd4_remove_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2931 {
2932         return (op_encode_hdr_size + op_encode_change_info_maxsz)
2933                 * sizeof(__be32);
2934 }
2935
2936 static inline u32 nfsd4_rename_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2937 {
2938         return (op_encode_hdr_size + op_encode_change_info_maxsz
2939                 + op_encode_change_info_maxsz) * sizeof(__be32);
2940 }
2941
2942 static inline u32 nfsd4_sequence_rsize(struct svc_rqst *rqstp,
2943                                        struct nfsd4_op *op)
2944 {
2945         return (op_encode_hdr_size
2946                 + XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5) * sizeof(__be32);
2947 }
2948
2949 static inline u32 nfsd4_test_stateid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2950 {
2951         return (op_encode_hdr_size + 1 + op->u.test_stateid.ts_num_ids)
2952                 * sizeof(__be32);
2953 }
2954
2955 static inline u32 nfsd4_setattr_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2956 {
2957         return (op_encode_hdr_size + nfs4_fattr_bitmap_maxsz) * sizeof(__be32);
2958 }
2959
2960 static inline u32 nfsd4_secinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2961 {
2962         return (op_encode_hdr_size + RPC_AUTH_MAXFLAVOR *
2963                 (4 + XDR_QUADLEN(GSS_OID_MAX_LEN))) * sizeof(__be32);
2964 }
2965
2966 static inline u32 nfsd4_setclientid_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2967 {
2968         return (op_encode_hdr_size + 2 + XDR_QUADLEN(NFS4_VERIFIER_SIZE)) *
2969                                                                 sizeof(__be32);
2970 }
2971
2972 static inline u32 nfsd4_write_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2973 {
2974         return (op_encode_hdr_size + 2 + op_encode_verifier_maxsz) * sizeof(__be32);
2975 }
2976
2977 static inline u32 nfsd4_exchange_id_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2978 {
2979         return (op_encode_hdr_size + 2 + 1 + /* eir_clientid, eir_sequenceid */\
2980                 1 + 1 + /* eir_flags, spr_how */\
2981                 4 + /* spo_must_enforce & _allow with bitmap */\
2982                 2 + /*eir_server_owner.so_minor_id */\
2983                 /* eir_server_owner.so_major_id<> */\
2984                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2985                 /* eir_server_scope<> */\
2986                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) + 1 +\
2987                 1 + /* eir_server_impl_id array length */\
2988                 0 /* ignored eir_server_impl_id contents */) * sizeof(__be32);
2989 }
2990
2991 static inline u32 nfsd4_bind_conn_to_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2992 {
2993         return (op_encode_hdr_size + \
2994                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* bctsr_sessid */\
2995                 2 /* bctsr_dir, use_conn_in_rdma_mode */) * sizeof(__be32);
2996 }
2997
2998 static inline u32 nfsd4_create_session_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
2999 {
3000         return (op_encode_hdr_size + \
3001                 XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + /* sessionid */\
3002                 2 + /* csr_sequence, csr_flags */\
3003                 op_encode_channel_attrs_maxsz + \
3004                 op_encode_channel_attrs_maxsz) * sizeof(__be32);
3005 }
3006
3007 static inline u32 nfsd4_copy_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3008 {
3009         return (op_encode_hdr_size +
3010                 1 /* wr_callback */ +
3011                 op_encode_stateid_maxsz /* wr_callback */ +
3012                 2 /* wr_count */ +
3013                 1 /* wr_committed */ +
3014                 op_encode_verifier_maxsz +
3015                 1 /* cr_consecutive */ +
3016                 1 /* cr_synchronous */) * sizeof(__be32);
3017 }
3018
3019 static inline u32 nfsd4_offload_status_rsize(struct svc_rqst *rqstp,
3020                                              struct nfsd4_op *op)
3021 {
3022         return (op_encode_hdr_size +
3023                 2 /* osr_count */ +
3024                 1 /* osr_complete<1> optional 0 for now */) * sizeof(__be32);
3025 }
3026
3027 static inline u32 nfsd4_copy_notify_rsize(struct svc_rqst *rqstp,
3028                                         struct nfsd4_op *op)
3029 {
3030         return (op_encode_hdr_size +
3031                 3 /* cnr_lease_time */ +
3032                 1 /* We support one cnr_source_server */ +
3033                 1 /* cnr_stateid seq */ +
3034                 op_encode_stateid_maxsz /* cnr_stateid */ +
3035                 1 /* num cnr_source_server*/ +
3036                 1 /* nl4_type */ +
3037                 1 /* nl4 size */ +
3038                 XDR_QUADLEN(NFS4_OPAQUE_LIMIT) /*nl4_loc + nl4_loc_sz */)
3039                 * sizeof(__be32);
3040 }
3041
3042 #ifdef CONFIG_NFSD_PNFS
3043 static inline u32 nfsd4_getdeviceinfo_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3044 {
3045         u32 maxcount = 0, rlen = 0;
3046
3047         maxcount = svc_max_payload(rqstp);
3048         rlen = min(op->u.getdeviceinfo.gd_maxcount, maxcount);
3049
3050         return (op_encode_hdr_size +
3051                 1 /* gd_layout_type*/ +
3052                 XDR_QUADLEN(rlen) +
3053                 2 /* gd_notify_types */) * sizeof(__be32);
3054 }
3055
3056 /*
3057  * At this stage we don't really know what layout driver will handle the request,
3058  * so we need to define an arbitrary upper bound here.
3059  */
3060 #define MAX_LAYOUT_SIZE         128
3061 static inline u32 nfsd4_layoutget_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3062 {
3063         return (op_encode_hdr_size +
3064                 1 /* logr_return_on_close */ +
3065                 op_encode_stateid_maxsz +
3066                 1 /* nr of layouts */ +
3067                 MAX_LAYOUT_SIZE) * sizeof(__be32);
3068 }
3069
3070 static inline u32 nfsd4_layoutcommit_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3071 {
3072         return (op_encode_hdr_size +
3073                 1 /* locr_newsize */ +
3074                 2 /* ns_size */) * sizeof(__be32);
3075 }
3076
3077 static inline u32 nfsd4_layoutreturn_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3078 {
3079         return (op_encode_hdr_size +
3080                 1 /* lrs_stateid */ +
3081                 op_encode_stateid_maxsz) * sizeof(__be32);
3082 }
3083 #endif /* CONFIG_NFSD_PNFS */
3084
3085
3086 static inline u32 nfsd4_seek_rsize(struct svc_rqst *rqstp, struct nfsd4_op *op)
3087 {
3088         return (op_encode_hdr_size + 3) * sizeof(__be32);
3089 }
3090
3091 static inline u32 nfsd4_getxattr_rsize(struct svc_rqst *rqstp,
3092                                        struct nfsd4_op *op)
3093 {
3094         u32 maxcount, rlen;
3095
3096         maxcount = svc_max_payload(rqstp);
3097         rlen = min_t(u32, XATTR_SIZE_MAX, maxcount);
3098
3099         return (op_encode_hdr_size + 1 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3100 }
3101
3102 static inline u32 nfsd4_setxattr_rsize(struct svc_rqst *rqstp,
3103                                        struct nfsd4_op *op)
3104 {
3105         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3106                 * sizeof(__be32);
3107 }
3108 static inline u32 nfsd4_listxattrs_rsize(struct svc_rqst *rqstp,
3109                                          struct nfsd4_op *op)
3110 {
3111         u32 maxcount, rlen;
3112
3113         maxcount = svc_max_payload(rqstp);
3114         rlen = min(op->u.listxattrs.lsxa_maxcount, maxcount);
3115
3116         return (op_encode_hdr_size + 4 + XDR_QUADLEN(rlen)) * sizeof(__be32);
3117 }
3118
3119 static inline u32 nfsd4_removexattr_rsize(struct svc_rqst *rqstp,
3120                                           struct nfsd4_op *op)
3121 {
3122         return (op_encode_hdr_size + op_encode_change_info_maxsz)
3123                 * sizeof(__be32);
3124 }
3125
3126
3127 static const struct nfsd4_operation nfsd4_ops[] = {
3128         [OP_ACCESS] = {
3129                 .op_func = nfsd4_access,
3130                 .op_name = "OP_ACCESS",
3131                 .op_rsize_bop = nfsd4_access_rsize,
3132         },
3133         [OP_CLOSE] = {
3134                 .op_func = nfsd4_close,
3135                 .op_flags = OP_MODIFIES_SOMETHING,
3136                 .op_name = "OP_CLOSE",
3137                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3138                 .op_get_currentstateid = nfsd4_get_closestateid,
3139                 .op_set_currentstateid = nfsd4_set_closestateid,
3140         },
3141         [OP_COMMIT] = {
3142                 .op_func = nfsd4_commit,
3143                 .op_flags = OP_MODIFIES_SOMETHING,
3144                 .op_name = "OP_COMMIT",
3145                 .op_rsize_bop = nfsd4_commit_rsize,
3146         },
3147         [OP_CREATE] = {
3148                 .op_func = nfsd4_create,
3149                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME | OP_CLEAR_STATEID,
3150                 .op_name = "OP_CREATE",
3151                 .op_rsize_bop = nfsd4_create_rsize,
3152         },
3153         [OP_DELEGRETURN] = {
3154                 .op_func = nfsd4_delegreturn,
3155                 .op_flags = OP_MODIFIES_SOMETHING,
3156                 .op_name = "OP_DELEGRETURN",
3157                 .op_rsize_bop = nfsd4_only_status_rsize,
3158                 .op_get_currentstateid = nfsd4_get_delegreturnstateid,
3159         },
3160         [OP_GETATTR] = {
3161                 .op_func = nfsd4_getattr,
3162                 .op_flags = ALLOWED_ON_ABSENT_FS,
3163                 .op_rsize_bop = nfsd4_getattr_rsize,
3164                 .op_name = "OP_GETATTR",
3165         },
3166         [OP_GETFH] = {
3167                 .op_func = nfsd4_getfh,
3168                 .op_name = "OP_GETFH",
3169                 .op_rsize_bop = nfsd4_getfh_rsize,
3170         },
3171         [OP_LINK] = {
3172                 .op_func = nfsd4_link,
3173                 .op_flags = ALLOWED_ON_ABSENT_FS | OP_MODIFIES_SOMETHING
3174                                 | OP_CACHEME,
3175                 .op_name = "OP_LINK",
3176                 .op_rsize_bop = nfsd4_link_rsize,
3177         },
3178         [OP_LOCK] = {
3179                 .op_func = nfsd4_lock,
3180                 .op_flags = OP_MODIFIES_SOMETHING |
3181                                 OP_NONTRIVIAL_ERROR_ENCODE,
3182                 .op_name = "OP_LOCK",
3183                 .op_rsize_bop = nfsd4_lock_rsize,
3184                 .op_set_currentstateid = nfsd4_set_lockstateid,
3185         },
3186         [OP_LOCKT] = {
3187                 .op_func = nfsd4_lockt,
3188                 .op_flags = OP_NONTRIVIAL_ERROR_ENCODE,
3189                 .op_name = "OP_LOCKT",
3190                 .op_rsize_bop = nfsd4_lock_rsize,
3191         },
3192         [OP_LOCKU] = {
3193                 .op_func = nfsd4_locku,
3194                 .op_flags = OP_MODIFIES_SOMETHING,
3195                 .op_name = "OP_LOCKU",
3196                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3197                 .op_get_currentstateid = nfsd4_get_lockustateid,
3198         },
3199         [OP_LOOKUP] = {
3200                 .op_func = nfsd4_lookup,
3201                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3202                 .op_name = "OP_LOOKUP",
3203                 .op_rsize_bop = nfsd4_only_status_rsize,
3204         },
3205         [OP_LOOKUPP] = {
3206                 .op_func = nfsd4_lookupp,
3207                 .op_flags = OP_HANDLES_WRONGSEC | OP_CLEAR_STATEID,
3208                 .op_name = "OP_LOOKUPP",
3209                 .op_rsize_bop = nfsd4_only_status_rsize,
3210         },
3211         [OP_NVERIFY] = {
3212                 .op_func = nfsd4_nverify,
3213                 .op_name = "OP_NVERIFY",
3214                 .op_rsize_bop = nfsd4_only_status_rsize,
3215         },
3216         [OP_OPEN] = {
3217                 .op_func = nfsd4_open,
3218                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3219                 .op_name = "OP_OPEN",
3220                 .op_rsize_bop = nfsd4_open_rsize,
3221                 .op_set_currentstateid = nfsd4_set_openstateid,
3222         },
3223         [OP_OPEN_CONFIRM] = {
3224                 .op_func = nfsd4_open_confirm,
3225                 .op_flags = OP_MODIFIES_SOMETHING,
3226                 .op_name = "OP_OPEN_CONFIRM",
3227                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3228         },
3229         [OP_OPEN_DOWNGRADE] = {
3230                 .op_func = nfsd4_open_downgrade,
3231                 .op_flags = OP_MODIFIES_SOMETHING,
3232                 .op_name = "OP_OPEN_DOWNGRADE",
3233                 .op_rsize_bop = nfsd4_status_stateid_rsize,
3234                 .op_get_currentstateid = nfsd4_get_opendowngradestateid,
3235                 .op_set_currentstateid = nfsd4_set_opendowngradestateid,
3236         },
3237         [OP_PUTFH] = {
3238                 .op_func = nfsd4_putfh,
3239                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3240                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3241                 .op_name = "OP_PUTFH",
3242                 .op_rsize_bop = nfsd4_only_status_rsize,
3243         },
3244         [OP_PUTPUBFH] = {
3245                 .op_func = nfsd4_putrootfh,
3246                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3247                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3248                 .op_name = "OP_PUTPUBFH",
3249                 .op_rsize_bop = nfsd4_only_status_rsize,
3250         },
3251         [OP_PUTROOTFH] = {
3252                 .op_func = nfsd4_putrootfh,
3253                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3254                                 | OP_IS_PUTFH_LIKE | OP_CLEAR_STATEID,
3255                 .op_name = "OP_PUTROOTFH",
3256                 .op_rsize_bop = nfsd4_only_status_rsize,
3257         },
3258         [OP_READ] = {
3259                 .op_func = nfsd4_read,
3260                 .op_release = nfsd4_read_release,
3261                 .op_name = "OP_READ",
3262                 .op_rsize_bop = nfsd4_read_rsize,
3263                 .op_get_currentstateid = nfsd4_get_readstateid,
3264         },
3265         [OP_READDIR] = {
3266                 .op_func = nfsd4_readdir,
3267                 .op_name = "OP_READDIR",
3268                 .op_rsize_bop = nfsd4_readdir_rsize,
3269         },
3270         [OP_READLINK] = {
3271                 .op_func = nfsd4_readlink,
3272                 .op_name = "OP_READLINK",
3273                 .op_rsize_bop = nfsd4_readlink_rsize,
3274         },
3275         [OP_REMOVE] = {
3276                 .op_func = nfsd4_remove,
3277                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3278                 .op_name = "OP_REMOVE",
3279                 .op_rsize_bop = nfsd4_remove_rsize,
3280         },
3281         [OP_RENAME] = {
3282                 .op_func = nfsd4_rename,
3283                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3284                 .op_name = "OP_RENAME",
3285                 .op_rsize_bop = nfsd4_rename_rsize,
3286         },
3287         [OP_RENEW] = {
3288                 .op_func = nfsd4_renew,
3289                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3290                                 | OP_MODIFIES_SOMETHING,
3291                 .op_name = "OP_RENEW",
3292                 .op_rsize_bop = nfsd4_only_status_rsize,
3293
3294         },
3295         [OP_RESTOREFH] = {
3296                 .op_func = nfsd4_restorefh,
3297                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3298                                 | OP_IS_PUTFH_LIKE | OP_MODIFIES_SOMETHING,
3299                 .op_name = "OP_RESTOREFH",
3300                 .op_rsize_bop = nfsd4_only_status_rsize,
3301         },
3302         [OP_SAVEFH] = {
3303                 .op_func = nfsd4_savefh,
3304                 .op_flags = OP_HANDLES_WRONGSEC | OP_MODIFIES_SOMETHING,
3305                 .op_name = "OP_SAVEFH",
3306                 .op_rsize_bop = nfsd4_only_status_rsize,
3307         },
3308         [OP_SECINFO] = {
3309                 .op_func = nfsd4_secinfo,
3310                 .op_release = nfsd4_secinfo_release,
3311                 .op_flags = OP_HANDLES_WRONGSEC,
3312                 .op_name = "OP_SECINFO",
3313                 .op_rsize_bop = nfsd4_secinfo_rsize,
3314         },
3315         [OP_SETATTR] = {
3316                 .op_func = nfsd4_setattr,
3317                 .op_name = "OP_SETATTR",
3318                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME
3319                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3320                 .op_rsize_bop = nfsd4_setattr_rsize,
3321                 .op_get_currentstateid = nfsd4_get_setattrstateid,
3322         },
3323         [OP_SETCLIENTID] = {
3324                 .op_func = nfsd4_setclientid,
3325                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3326                                 | OP_MODIFIES_SOMETHING | OP_CACHEME
3327                                 | OP_NONTRIVIAL_ERROR_ENCODE,
3328                 .op_name = "OP_SETCLIENTID",
3329                 .op_rsize_bop = nfsd4_setclientid_rsize,
3330         },
3331         [OP_SETCLIENTID_CONFIRM] = {
3332                 .op_func = nfsd4_setclientid_confirm,
3333                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3334                                 | OP_MODIFIES_SOMETHING | OP_CACHEME,
3335                 .op_name = "OP_SETCLIENTID_CONFIRM",
3336                 .op_rsize_bop = nfsd4_only_status_rsize,
3337         },
3338         [OP_VERIFY] = {
3339                 .op_func = nfsd4_verify,
3340                 .op_name = "OP_VERIFY",
3341                 .op_rsize_bop = nfsd4_only_status_rsize,
3342         },
3343         [OP_WRITE] = {
3344                 .op_func = nfsd4_write,
3345                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3346                 .op_name = "OP_WRITE",
3347                 .op_rsize_bop = nfsd4_write_rsize,
3348                 .op_get_currentstateid = nfsd4_get_writestateid,
3349         },
3350         [OP_RELEASE_LOCKOWNER] = {
3351                 .op_func = nfsd4_release_lockowner,
3352                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_ON_ABSENT_FS
3353                                 | OP_MODIFIES_SOMETHING,
3354                 .op_name = "OP_RELEASE_LOCKOWNER",
3355                 .op_rsize_bop = nfsd4_only_status_rsize,
3356         },
3357
3358         /* NFSv4.1 operations */
3359         [OP_EXCHANGE_ID] = {
3360                 .op_func = nfsd4_exchange_id,
3361                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3362                                 | OP_MODIFIES_SOMETHING,
3363                 .op_name = "OP_EXCHANGE_ID",
3364                 .op_rsize_bop = nfsd4_exchange_id_rsize,
3365         },
3366         [OP_BACKCHANNEL_CTL] = {
3367                 .op_func = nfsd4_backchannel_ctl,
3368                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3369                 .op_name = "OP_BACKCHANNEL_CTL",
3370                 .op_rsize_bop = nfsd4_only_status_rsize,
3371         },
3372         [OP_BIND_CONN_TO_SESSION] = {
3373                 .op_func = nfsd4_bind_conn_to_session,
3374                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3375                                 | OP_MODIFIES_SOMETHING,
3376                 .op_name = "OP_BIND_CONN_TO_SESSION",
3377                 .op_rsize_bop = nfsd4_bind_conn_to_session_rsize,
3378         },
3379         [OP_CREATE_SESSION] = {
3380                 .op_func = nfsd4_create_session,
3381                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3382                                 | OP_MODIFIES_SOMETHING,
3383                 .op_name = "OP_CREATE_SESSION",
3384                 .op_rsize_bop = nfsd4_create_session_rsize,
3385         },
3386         [OP_DESTROY_SESSION] = {
3387                 .op_func = nfsd4_destroy_session,
3388                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3389                                 | OP_MODIFIES_SOMETHING,
3390                 .op_name = "OP_DESTROY_SESSION",
3391                 .op_rsize_bop = nfsd4_only_status_rsize,
3392         },
3393         [OP_SEQUENCE] = {
3394                 .op_func = nfsd4_sequence,
3395                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP,
3396                 .op_name = "OP_SEQUENCE",
3397                 .op_rsize_bop = nfsd4_sequence_rsize,
3398         },
3399         [OP_DESTROY_CLIENTID] = {
3400                 .op_func = nfsd4_destroy_clientid,
3401                 .op_flags = ALLOWED_WITHOUT_FH | ALLOWED_AS_FIRST_OP
3402                                 | OP_MODIFIES_SOMETHING,
3403                 .op_name = "OP_DESTROY_CLIENTID",
3404                 .op_rsize_bop = nfsd4_only_status_rsize,
3405         },
3406         [OP_RECLAIM_COMPLETE] = {
3407                 .op_func = nfsd4_reclaim_complete,
3408                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3409                 .op_name = "OP_RECLAIM_COMPLETE",
3410                 .op_rsize_bop = nfsd4_only_status_rsize,
3411         },
3412         [OP_SECINFO_NO_NAME] = {
3413                 .op_func = nfsd4_secinfo_no_name,
3414                 .op_release = nfsd4_secinfo_no_name_release,
3415                 .op_flags = OP_HANDLES_WRONGSEC,
3416                 .op_name = "OP_SECINFO_NO_NAME",
3417                 .op_rsize_bop = nfsd4_secinfo_rsize,
3418         },
3419         [OP_TEST_STATEID] = {
3420                 .op_func = nfsd4_test_stateid,
3421                 .op_flags = ALLOWED_WITHOUT_FH,
3422                 .op_name = "OP_TEST_STATEID",
3423                 .op_rsize_bop = nfsd4_test_stateid_rsize,
3424         },
3425         [OP_FREE_STATEID] = {
3426                 .op_func = nfsd4_free_stateid,
3427                 .op_flags = ALLOWED_WITHOUT_FH | OP_MODIFIES_SOMETHING,
3428                 .op_name = "OP_FREE_STATEID",
3429                 .op_get_currentstateid = nfsd4_get_freestateid,
3430                 .op_rsize_bop = nfsd4_only_status_rsize,
3431         },
3432 #ifdef CONFIG_NFSD_PNFS
3433         [OP_GETDEVICEINFO] = {
3434                 .op_func = nfsd4_getdeviceinfo,
3435                 .op_release = nfsd4_getdeviceinfo_release,
3436                 .op_flags = ALLOWED_WITHOUT_FH,
3437                 .op_name = "OP_GETDEVICEINFO",
3438                 .op_rsize_bop = nfsd4_getdeviceinfo_rsize,
3439         },
3440         [OP_LAYOUTGET] = {
3441                 .op_func = nfsd4_layoutget,
3442                 .op_release = nfsd4_layoutget_release,
3443                 .op_flags = OP_MODIFIES_SOMETHING,
3444                 .op_name = "OP_LAYOUTGET",
3445                 .op_rsize_bop = nfsd4_layoutget_rsize,
3446         },
3447         [OP_LAYOUTCOMMIT] = {
3448                 .op_func = nfsd4_layoutcommit,
3449                 .op_flags = OP_MODIFIES_SOMETHING,
3450                 .op_name = "OP_LAYOUTCOMMIT",
3451                 .op_rsize_bop = nfsd4_layoutcommit_rsize,
3452         },
3453         [OP_LAYOUTRETURN] = {
3454                 .op_func = nfsd4_layoutreturn,
3455                 .op_flags = OP_MODIFIES_SOMETHING,
3456                 .op_name = "OP_LAYOUTRETURN",
3457                 .op_rsize_bop = nfsd4_layoutreturn_rsize,
3458         },
3459 #endif /* CONFIG_NFSD_PNFS */
3460
3461         /* NFSv4.2 operations */
3462         [OP_ALLOCATE] = {
3463                 .op_func = nfsd4_allocate,
3464                 .op_flags = OP_MODIFIES_SOMETHING,
3465                 .op_name = "OP_ALLOCATE",
3466                 .op_rsize_bop = nfsd4_only_status_rsize,
3467         },
3468         [OP_DEALLOCATE] = {
3469                 .op_func = nfsd4_deallocate,
3470                 .op_flags = OP_MODIFIES_SOMETHING,
3471                 .op_name = "OP_DEALLOCATE",
3472                 .op_rsize_bop = nfsd4_only_status_rsize,
3473         },
3474         [OP_CLONE] = {
3475                 .op_func = nfsd4_clone,
3476                 .op_flags = OP_MODIFIES_SOMETHING,
3477                 .op_name = "OP_CLONE",
3478                 .op_rsize_bop = nfsd4_only_status_rsize,
3479         },
3480         [OP_COPY] = {
3481                 .op_func = nfsd4_copy,
3482                 .op_flags = OP_MODIFIES_SOMETHING,
3483                 .op_name = "OP_COPY",
3484                 .op_rsize_bop = nfsd4_copy_rsize,
3485         },
3486         [OP_READ_PLUS] = {
3487                 .op_func = nfsd4_read,
3488                 .op_release = nfsd4_read_release,
3489                 .op_name = "OP_READ_PLUS",
3490                 .op_rsize_bop = nfsd4_read_plus_rsize,
3491                 .op_get_currentstateid = nfsd4_get_readstateid,
3492         },
3493         [OP_SEEK] = {
3494                 .op_func = nfsd4_seek,
3495                 .op_name = "OP_SEEK",
3496                 .op_rsize_bop = nfsd4_seek_rsize,
3497         },
3498         [OP_OFFLOAD_STATUS] = {
3499                 .op_func = nfsd4_offload_status,
3500                 .op_name = "OP_OFFLOAD_STATUS",
3501                 .op_rsize_bop = nfsd4_offload_status_rsize,
3502         },
3503         [OP_OFFLOAD_CANCEL] = {
3504                 .op_func = nfsd4_offload_cancel,
3505                 .op_flags = OP_MODIFIES_SOMETHING,
3506                 .op_name = "OP_OFFLOAD_CANCEL",
3507                 .op_rsize_bop = nfsd4_only_status_rsize,
3508         },
3509         [OP_COPY_NOTIFY] = {
3510                 .op_func = nfsd4_copy_notify,
3511                 .op_flags = OP_MODIFIES_SOMETHING,
3512                 .op_name = "OP_COPY_NOTIFY",
3513                 .op_rsize_bop = nfsd4_copy_notify_rsize,
3514         },
3515         [OP_GETXATTR] = {
3516                 .op_func = nfsd4_getxattr,
3517                 .op_name = "OP_GETXATTR",
3518                 .op_rsize_bop = nfsd4_getxattr_rsize,
3519         },
3520         [OP_SETXATTR] = {
3521                 .op_func = nfsd4_setxattr,
3522                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3523                 .op_name = "OP_SETXATTR",
3524                 .op_rsize_bop = nfsd4_setxattr_rsize,
3525         },
3526         [OP_LISTXATTRS] = {
3527                 .op_func = nfsd4_listxattrs,
3528                 .op_name = "OP_LISTXATTRS",
3529                 .op_rsize_bop = nfsd4_listxattrs_rsize,
3530         },
3531         [OP_REMOVEXATTR] = {
3532                 .op_func = nfsd4_removexattr,
3533                 .op_flags = OP_MODIFIES_SOMETHING | OP_CACHEME,
3534                 .op_name = "OP_REMOVEXATTR",
3535                 .op_rsize_bop = nfsd4_removexattr_rsize,
3536         },
3537 };
3538
3539 /**
3540  * nfsd4_spo_must_allow - Determine if the compound op contains an
3541  * operation that is allowed to be sent with machine credentials
3542  *
3543  * @rqstp: a pointer to the struct svc_rqst
3544  *
3545  * Checks to see if the compound contains a spo_must_allow op
3546  * and confirms that it was sent with the proper machine creds.
3547  */
3548
3549 bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
3550 {
3551         struct nfsd4_compoundres *resp = rqstp->rq_resp;
3552         struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3553         struct nfsd4_op *this;
3554         struct nfsd4_compound_state *cstate = &resp->cstate;
3555         struct nfs4_op_map *allow = &cstate->clp->cl_spo_must_allow;
3556         u32 opiter;
3557
3558         if (!cstate->minorversion)
3559                 return false;
3560
3561         if (cstate->spo_must_allowed)
3562                 return true;
3563
3564         opiter = resp->opcnt;
3565         while (opiter < argp->opcnt) {
3566                 this = &argp->ops[opiter++];
3567                 if (test_bit(this->opnum, allow->u.longs) &&
3568                         cstate->clp->cl_mach_cred &&
3569                         nfsd4_mach_creds_match(cstate->clp, rqstp)) {
3570                         cstate->spo_must_allowed = true;
3571                         return true;
3572                 }
3573         }
3574         cstate->spo_must_allowed = false;
3575         return false;
3576 }
3577
3578 int nfsd4_max_reply(struct svc_rqst *rqstp, struct nfsd4_op *op)
3579 {
3580         if (op->opnum == OP_ILLEGAL || op->status == nfserr_notsupp)
3581                 return op_encode_hdr_size * sizeof(__be32);
3582
3583         BUG_ON(OPDESC(op)->op_rsize_bop == NULL);
3584         return OPDESC(op)->op_rsize_bop(rqstp, op);
3585 }
3586
3587 void warn_on_nonidempotent_op(struct nfsd4_op *op)
3588 {
3589         if (OPDESC(op)->op_flags & OP_MODIFIES_SOMETHING) {
3590                 pr_err("unable to encode reply to nonidempotent op %u (%s)\n",
3591                         op->opnum, nfsd4_op_name(op->opnum));
3592                 WARN_ON_ONCE(1);
3593         }
3594 }
3595
3596 static const char *nfsd4_op_name(unsigned opnum)
3597 {
3598         if (opnum < ARRAY_SIZE(nfsd4_ops))
3599                 return nfsd4_ops[opnum].op_name;
3600         return "unknown_operation";
3601 }
3602
3603 static const struct svc_procedure nfsd_procedures4[2] = {
3604         [NFSPROC4_NULL] = {
3605                 .pc_func = nfsd4_proc_null,
3606                 .pc_decode = nfssvc_decode_voidarg,
3607                 .pc_encode = nfssvc_encode_voidres,
3608                 .pc_argsize = sizeof(struct nfsd_voidargs),
3609                 .pc_ressize = sizeof(struct nfsd_voidres),
3610                 .pc_cachetype = RC_NOCACHE,
3611                 .pc_xdrressize = 1,
3612                 .pc_name = "NULL",
3613         },
3614         [NFSPROC4_COMPOUND] = {
3615                 .pc_func = nfsd4_proc_compound,
3616                 .pc_decode = nfs4svc_decode_compoundargs,
3617                 .pc_encode = nfs4svc_encode_compoundres,
3618                 .pc_argsize = sizeof(struct nfsd4_compoundargs),
3619                 .pc_ressize = sizeof(struct nfsd4_compoundres),
3620                 .pc_release = nfsd4_release_compoundargs,
3621                 .pc_cachetype = RC_NOCACHE,
3622                 .pc_xdrressize = NFSD_BUFSIZE/4,
3623                 .pc_name = "COMPOUND",
3624         },
3625 };
3626
3627 static unsigned int nfsd_count3[ARRAY_SIZE(nfsd_procedures4)];
3628 const struct svc_version nfsd_version4 = {
3629         .vs_vers                = 4,
3630         .vs_nproc               = 2,
3631         .vs_proc                = nfsd_procedures4,
3632         .vs_count               = nfsd_count3,
3633         .vs_dispatch            = nfsd_dispatch,
3634         .vs_xdrsize             = NFS4_SVC_XDRSIZE,
3635         .vs_rpcb_optnl          = true,
3636         .vs_need_cong_ctrl      = true,
3637 };