afs: Provide a splice-read wrapper
[linux-block.git] / fs / nfsd / nfsfh.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4 2/*
1da177e4
LT
3 * NFS server file handle treatment.
4 *
5 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
6 * Portions Copyright (C) 1999 G. Allen Morris III <gam3@acm.org>
7 * Extensive rewrite by Neil Brown <neilb@cse.unsw.edu.au> Southern-Spring 1999
8 * ... and again Southern-Winter 2001 to support export_operations
9 */
10
a5694255 11#include <linux/exportfs.h>
1da177e4 12
32c1eb0c 13#include <linux/sunrpc/svcauth_gss.h>
9a74af21 14#include "nfsd.h"
0a3adade 15#include "vfs.h"
2e8138a2 16#include "auth.h"
f01274a9 17#include "trace.h"
1da177e4
LT
18
19#define NFSDDBG_FACILITY NFSDDBG_FH
1da177e4
LT
20
21
1da177e4
LT
22/*
23 * our acceptability function.
24 * if NOSUBTREECHECK, accept anything
25 * if not, require that we can walk up to exp->ex_dentry
26 * doing some checks on the 'x' bits
27 */
28static int nfsd_acceptable(void *expv, struct dentry *dentry)
29{
30 struct svc_export *exp = expv;
31 int rv;
32 struct dentry *tdentry;
33 struct dentry *parent;
34
35 if (exp->ex_flags & NFSEXP_NOSUBTREECHECK)
36 return 1;
37
38 tdentry = dget(dentry);
54775491 39 while (tdentry != exp->ex_path.dentry && !IS_ROOT(tdentry)) {
1da177e4
LT
40 /* make sure parents give x permission to user */
41 int err;
42 parent = dget_parent(tdentry);
4609e1f1 43 err = inode_permission(&nop_mnt_idmap,
47291baa 44 d_inode(parent), MAY_EXEC);
1da177e4
LT
45 if (err < 0) {
46 dput(parent);
47 break;
48 }
49 dput(tdentry);
50 tdentry = parent;
51 }
54775491 52 if (tdentry != exp->ex_path.dentry)
97e47fa1 53 dprintk("nfsd_acceptable failed at %p %pd\n", tdentry, tdentry);
54775491 54 rv = (tdentry == exp->ex_path.dentry);
1da177e4
LT
55 dput(tdentry);
56 return rv;
57}
58
59/* Type check. The correct error return for type mismatches does not seem to be
60 * generally agreed upon. SunOS seems to use EISDIR if file isn't S_IFREG; a
61 * comment in the NFSv3 spec says this is incorrect (implementation notes for
62 * the write call).
63 */
83b11340 64static inline __be32
e75b23f9
BF
65nfsd_mode_check(struct svc_rqst *rqstp, struct dentry *dentry,
66 umode_t requested)
1da177e4 67{
e75b23f9 68 umode_t mode = d_inode(dentry)->i_mode & S_IFMT;
e10f9e14
BF
69
70 if (requested == 0) /* the caller doesn't care */
71 return nfs_ok;
e75b23f9
BF
72 if (mode == requested) {
73 if (mode == S_IFDIR && !d_can_lookup(dentry)) {
74 WARN_ON_ONCE(1);
75 return nfserr_notdir;
76 }
e10f9e14 77 return nfs_ok;
e75b23f9 78 }
e10f9e14
BF
79 /*
80 * v4 has an error more specific than err_notdir which we should
81 * return in preference to err_notdir:
82 */
83 if (rqstp->rq_vers == 4 && mode == S_IFLNK)
84 return nfserr_symlink;
85 if (requested == S_IFDIR)
86 return nfserr_notdir;
87 if (mode == S_IFDIR)
88 return nfserr_isdir;
89 return nfserr_inval;
1da177e4
LT
90}
91
9d7ed135
BF
92static bool nfsd_originating_port_ok(struct svc_rqst *rqstp, int flags)
93{
94 if (flags & NFSEXP_INSECURE_PORT)
95 return true;
96 /* We don't require gss requests to use low ports: */
97 if (rqstp->rq_cred.cr_flavor >= RPC_AUTH_GSS)
98 return true;
99 return test_bit(RQ_SECURE, &rqstp->rq_flags);
100}
101
6fa02839
BF
102static __be32 nfsd_setuser_and_check_port(struct svc_rqst *rqstp,
103 struct svc_export *exp)
104{
12045a6e
BF
105 int flags = nfsexp_flags(rqstp, exp);
106
6fa02839 107 /* Check if the request originated from a secure port. */
9d7ed135 108 if (!nfsd_originating_port_ok(rqstp, flags)) {
5216a8e7 109 RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
a48fd0f9
KM
110 dprintk("nfsd: request from insecure port %s!\n",
111 svc_print_addr(rqstp, buf, sizeof(buf)));
6fa02839
BF
112 return nfserr_perm;
113 }
114
115 /* Set user creds for this exportpoint */
116 return nfserrno(nfsd_setuser(rqstp, exp));
117}
118
03a816b4
SD
119static inline __be32 check_pseudo_root(struct svc_rqst *rqstp,
120 struct dentry *dentry, struct svc_export *exp)
121{
122 if (!(exp->ex_flags & NFSEXP_V4ROOT))
123 return nfs_ok;
124 /*
125 * v2/v3 clients have no need for the V4ROOT export--they use
126 * the mount protocl instead; also, further V4ROOT checks may be
127 * in v4-specific code, in which case v2/v3 clients could bypass
128 * them.
129 */
130 if (!nfsd_v4client(rqstp))
131 return nfserr_stale;
132 /*
133 * We're exposing only the directories and symlinks that have to be
134 * traversed on the way to real exports:
135 */
e36cb0b8
DH
136 if (unlikely(!d_is_dir(dentry) &&
137 !d_is_symlink(dentry)))
03a816b4
SD
138 return nfserr_stale;
139 /*
140 * A pseudoroot export gives permission to access only one
141 * single directory; the kernel has to make another upcall
142 * before granting access to anything else under it:
143 */
144 if (unlikely(dentry != exp->ex_path.dentry))
145 return nfserr_stale;
146 return nfs_ok;
147}
148
03550fac
BF
149/*
150 * Use the given filehandle to look up the corresponding export and
151 * dentry. On success, the results are used to set fh_export and
152 * fh_dentry.
153 */
154static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)
155{
156 struct knfsd_fh *fh = &fhp->fh_handle;
c645a883 157 struct fid *fid = NULL;
03550fac
BF
158 struct svc_export *exp;
159 struct dentry *dentry;
160 int fileid_type;
161 int data_left = fh->fh_size/4;
c645a883 162 int len;
03550fac
BF
163 __be32 error;
164
165 error = nfserr_stale;
166 if (rqstp->rq_vers > 2)
167 error = nfserr_badhandle;
168 if (rqstp->rq_vers == 4 && fh->fh_size == 0)
169 return nfserr_nofilehandle;
170
c645a883
N
171 if (fh->fh_version != 1)
172 return error;
173
174 if (--data_left < 0)
175 return error;
176 if (fh->fh_auth_type != 0)
177 return error;
178 len = key_len(fh->fh_fsid_type) / 4;
179 if (len == 0)
180 return error;
181 if (fh->fh_fsid_type == FSID_MAJOR_MINOR) {
182 /* deprecated, convert to type 3 */
183 len = key_len(FSID_ENCODE_DEV)/4;
184 fh->fh_fsid_type = FSID_ENCODE_DEV;
185 /*
186 * struct knfsd_fh uses host-endian fields, which are
187 * sometimes used to hold net-endian values. This
188 * confuses sparse, so we must use __force here to
189 * keep it from complaining.
190 */
191 fh->fh_fsid[0] = new_encode_dev(MKDEV(ntohl((__force __be32)fh->fh_fsid[0]),
192 ntohl((__force __be32)fh->fh_fsid[1])));
193 fh->fh_fsid[1] = fh->fh_fsid[2];
03550fac 194 }
c645a883
N
195 data_left -= len;
196 if (data_left < 0)
197 return error;
198 exp = rqst_exp_find(rqstp, fh->fh_fsid_type, fh->fh_fsid);
199 fid = (struct fid *)(fh->fh_fsid + len);
03550fac
BF
200
201 error = nfserr_stale;
f01274a9
TM
202 if (IS_ERR(exp)) {
203 trace_nfsd_set_fh_dentry_badexport(rqstp, fhp, PTR_ERR(exp));
204
205 if (PTR_ERR(exp) == -ENOENT)
206 return error;
03550fac 207
03550fac 208 return nfserrno(PTR_ERR(exp));
f01274a9 209 }
03550fac 210
496d6c32
NB
211 if (exp->ex_flags & NFSEXP_NOSUBTREECHECK) {
212 /* Elevate privileges so that the lack of 'r' or 'x'
213 * permission on some parent directory will
214 * not stop exportfs_decode_fh from being able
215 * to reconnect a directory into the dentry cache.
216 * The same problem can affect "SUBTREECHECK" exports,
217 * but as nfsd_acceptable depends on correct
218 * access control settings being in effect, we cannot
219 * fix that case easily.
220 */
d84f4f99 221 struct cred *new = prepare_creds();
027bc41a
KM
222 if (!new) {
223 error = nfserrno(-ENOMEM);
224 goto out;
225 }
d84f4f99
DH
226 new->cap_effective =
227 cap_raise_nfsd_set(new->cap_effective,
228 new->cap_permitted);
229 put_cred(override_creds(new));
230 put_cred(new);
496d6c32
NB
231 } else {
232 error = nfsd_setuser_and_check_port(rqstp, exp);
233 if (error)
234 goto out;
235 }
03550fac
BF
236
237 /*
238 * Look up the dentry using the NFS file handle.
239 */
240 error = nfserr_stale;
241 if (rqstp->rq_vers > 2)
242 error = nfserr_badhandle;
243
c645a883 244 fileid_type = fh->fh_fileid_type;
03550fac
BF
245
246 if (fileid_type == FILEID_ROOT)
247 dentry = dget(exp->ex_path.dentry);
248 else {
2e19d10c
TM
249 dentry = exportfs_decode_fh_raw(exp->ex_path.mnt, fid,
250 data_left, fileid_type,
251 nfsd_acceptable, exp);
252 if (IS_ERR_OR_NULL(dentry)) {
f01274a9
TM
253 trace_nfsd_set_fh_dentry_badhandle(rqstp, fhp,
254 dentry ? PTR_ERR(dentry) : -ESTALE);
2e19d10c
TM
255 switch (PTR_ERR(dentry)) {
256 case -ENOMEM:
257 case -ETIMEDOUT:
258 break;
259 default:
260 dentry = ERR_PTR(-ESTALE);
261 }
262 }
03550fac
BF
263 }
264 if (dentry == NULL)
265 goto out;
266 if (IS_ERR(dentry)) {
267 if (PTR_ERR(dentry) != -EINVAL)
268 error = nfserrno(PTR_ERR(dentry));
269 goto out;
270 }
271
e36cb0b8 272 if (d_is_dir(dentry) &&
03550fac 273 (dentry->d_flags & DCACHE_DISCONNECTED)) {
97e47fa1
AV
274 printk("nfsd: find_fh_dentry returned a DISCONNECTED directory: %pd2\n",
275 dentry);
03550fac
BF
276 }
277
278 fhp->fh_dentry = dentry;
279 fhp->fh_export = exp;
daab110e
JL
280
281 switch (rqstp->rq_vers) {
716a8bc7
TM
282 case 4:
283 if (dentry->d_sb->s_export_op->flags & EXPORT_OP_NOATOMIC_ATTR)
284 fhp->fh_no_atomic_attr = true;
285 break;
daab110e
JL
286 case 3:
287 if (dentry->d_sb->s_export_op->flags & EXPORT_OP_NOWCC)
288 fhp->fh_no_wcc = true;
289 break;
290 case 2:
291 fhp->fh_no_wcc = true;
292 }
293
03550fac
BF
294 return 0;
295out:
296 exp_put(exp);
297 return error;
298}
299
b3d47676
BF
300/**
301 * fh_verify - filehandle lookup and access checking
302 * @rqstp: pointer to current rpc request
303 * @fhp: filehandle to be verified
304 * @type: expected type of object pointed to by filehandle
305 * @access: type of access needed to object
306 *
307 * Look up a dentry from the on-the-wire filehandle, check the client's
308 * access to the export, and set the current task's credentials.
309 *
310 * Regardless of success or failure of fh_verify(), fh_put() should be
311 * called on @fhp when the caller is finished with the filehandle.
312 *
313 * fh_verify() may be called multiple times on a given filehandle, for
314 * example, when processing an NFSv4 compound. The first call will look
315 * up a dentry using the on-the-wire filehandle. Subsequent calls will
316 * skip the lookup and just perform the other checks and possibly change
317 * the current task's credentials.
1da177e4 318 *
b3d47676
BF
319 * @type specifies the type of object expected using one of the S_IF*
320 * constants defined in include/linux/stat.h. The caller may use zero
321 * to indicate that it doesn't care, or a negative integer to indicate
322 * that it expects something not of the given type.
1da177e4 323 *
b3d47676 324 * @access is formed from the NFSD_MAY_* constants defined in
93f580a9 325 * fs/nfsd/vfs.h.
1da177e4 326 */
83b11340 327__be32
175a4eb7 328fh_verify(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, int access)
1da177e4 329{
20ad856e 330 struct svc_export *exp = NULL;
1da177e4 331 struct dentry *dentry;
03550fac 332 __be32 error;
1da177e4 333
1da177e4 334 if (!fhp->fh_dentry) {
03550fac 335 error = nfsd_set_fh_dentry(rqstp, fhp);
d1bbf14f
N
336 if (error)
337 goto out;
1da177e4 338 }
864f0f61
BF
339 dentry = fhp->fh_dentry;
340 exp = fhp->fh_export;
05138288
CL
341
342 trace_nfsd_fh_verify(rqstp, fhp, type, access);
343
864f0f61
BF
344 /*
345 * We still have to do all these permission checks, even when
346 * fh_dentry is already set:
347 * - fh_verify may be called multiple times with different
348 * "access" arguments (e.g. nfsd_proc_create calls
349 * fh_verify(...,NFSD_MAY_EXEC) first, then later (in
350 * nfsd_create) calls fh_verify(...,NFSD_MAY_CREATE).
351 * - in the NFSv4 case, the filehandle may have been filled
352 * in by fh_compose, and given a dentry, but further
353 * compound operations performed with that filehandle
354 * still need permissions checks. In the worst case, a
355 * mountpoint crossing may have changed the export
356 * options, and we may now need to use a different uid
357 * (for example, if different id-squashing options are in
358 * effect on the new filesystem).
359 */
03a816b4
SD
360 error = check_pseudo_root(rqstp, dentry, exp);
361 if (error)
362 goto out;
363
864f0f61
BF
364 error = nfsd_setuser_and_check_port(rqstp, exp);
365 if (error)
366 goto out;
7fc90ec9 367
e75b23f9 368 error = nfsd_mode_check(rqstp, dentry, type);
1da177e4
LT
369 if (error)
370 goto out;
371
04716e66
BF
372 /*
373 * pseudoflavor restrictions are not enforced on NLM,
374 * which clients virtually always use auth_sys for,
375 * even while using RPCSEC_GSS for NFS.
376 */
204f4ce7 377 if (access & NFSD_MAY_LOCK || access & NFSD_MAY_BYPASS_GSS)
04716e66
BF
378 goto skip_pseudoflavor_check;
379 /*
380 * Clients may expect to be able to use auth_sys during mount,
381 * even if they use gss for everything else; see section 2.3.2
382 * of rfc 2623.
383 */
384 if (access & NFSD_MAY_BYPASS_GSS_ON_ROOT
385 && exp->ex_path.dentry == dentry)
386 goto skip_pseudoflavor_check;
387
388 error = check_nfsd_access(exp, rqstp);
389 if (error)
390 goto out;
32c1eb0c 391
04716e66 392skip_pseudoflavor_check:
1da177e4 393 /* Finally, check access permissions. */
0ec757df 394 error = nfsd_permission(rqstp, exp, dentry, access);
1da177e4 395out:
93c128e7 396 trace_nfsd_fh_verify_err(rqstp, fhp, type, access, error);
1da177e4 397 if (error == nfserr_stale)
20ad856e 398 nfsd_stats_fh_stale_inc(exp);
1da177e4
LT
399 return error;
400}
401
402
403/*
404 * Compose a file handle for an NFS reply.
405 *
406 * Note that when first composed, the dentry may not yet have
407 * an inode. In this case a call to fh_update should be made
408 * before the fh goes out on the wire ...
409 */
6e91ea2b
CH
410static void _fh_update(struct svc_fh *fhp, struct svc_export *exp,
411 struct dentry *dentry)
1da177e4 412{
54775491 413 if (dentry != exp->ex_path.dentry) {
6e91ea2b 414 struct fid *fid = (struct fid *)
5409e46f 415 (fhp->fh_handle.fh_fsid + fhp->fh_handle.fh_size/4 - 1);
6e91ea2b
CH
416 int maxsize = (fhp->fh_maxsize - fhp->fh_handle.fh_size)/4;
417 int subtreecheck = !(exp->ex_flags & NFSEXP_NOSUBTREECHECK);
1da177e4 418
6e91ea2b
CH
419 fhp->fh_handle.fh_fileid_type =
420 exportfs_encode_fh(dentry, fid, &maxsize, subtreecheck);
421 fhp->fh_handle.fh_size += maxsize * 4;
422 } else {
423 fhp->fh_handle.fh_fileid_type = FILEID_ROOT;
424 }
1da177e4
LT
425}
426
8e498751
BF
427static bool is_root_export(struct svc_export *exp)
428{
429 return exp->ex_path.dentry == exp->ex_path.dentry->d_sb->s_root;
430}
431
432static struct super_block *exp_sb(struct svc_export *exp)
433{
fc64005c 434 return exp->ex_path.dentry->d_sb;
8e498751
BF
435}
436
437static bool fsid_type_ok_for_exp(u8 fsid_type, struct svc_export *exp)
438{
439 switch (fsid_type) {
440 case FSID_DEV:
441 if (!old_valid_dev(exp_sb(exp)->s_dev))
a677a783 442 return false;
df561f66 443 fallthrough;
8e498751
BF
444 case FSID_MAJOR_MINOR:
445 case FSID_ENCODE_DEV:
446 return exp_sb(exp)->s_type->fs_flags & FS_REQUIRES_DEV;
447 case FSID_NUM:
448 return exp->ex_flags & NFSEXP_FSID;
449 case FSID_UUID8:
450 case FSID_UUID16:
451 if (!is_root_export(exp))
a677a783 452 return false;
df561f66 453 fallthrough;
8e498751
BF
454 case FSID_UUID4_INUM:
455 case FSID_UUID16_INUM:
456 return exp->ex_uuid != NULL;
457 }
a677a783 458 return true;
8e498751
BF
459}
460
1da177e4 461
bc6c53d5
BF
462static void set_version_and_fsid_type(struct svc_fh *fhp, struct svc_export *exp, struct svc_fh *ref_fh)
463{
b41eeef1 464 u8 version;
bc6c53d5
BF
465 u8 fsid_type;
466retry:
b41eeef1 467 version = 1;
7e405364 468 if (ref_fh && ref_fh->fh_export == exp) {
982aedfd 469 version = ref_fh->fh_handle.fh_version;
b41eeef1
N
470 fsid_type = ref_fh->fh_handle.fh_fsid_type;
471
b41eeef1
N
472 ref_fh = NULL;
473
474 switch (version) {
475 case 0xca:
af6a4e28 476 fsid_type = FSID_DEV;
b41eeef1
N
477 break;
478 case 1:
479 break;
480 default:
481 goto retry;
482 }
483
8e498751
BF
484 /*
485 * As the fsid -> filesystem mapping was guided by
486 * user-space, there is no guarantee that the filesystem
487 * actually supports that fsid type. If it doesn't we
488 * loop around again without ref_fh set.
982aedfd 489 */
8e498751
BF
490 if (!fsid_type_ok_for_exp(fsid_type, exp))
491 goto retry;
30fa8c01
SD
492 } else if (exp->ex_flags & NFSEXP_FSID) {
493 fsid_type = FSID_NUM;
af6a4e28
N
494 } else if (exp->ex_uuid) {
495 if (fhp->fh_maxsize >= 64) {
8e498751 496 if (is_root_export(exp))
af6a4e28
N
497 fsid_type = FSID_UUID16;
498 else
499 fsid_type = FSID_UUID16_INUM;
500 } else {
8e498751 501 if (is_root_export(exp))
af6a4e28
N
502 fsid_type = FSID_UUID8;
503 else
504 fsid_type = FSID_UUID4_INUM;
505 }
bc6c53d5 506 } else if (!old_valid_dev(exp_sb(exp)->s_dev))
1da177e4 507 /* for newer device numbers, we must use a newer fsid format */
af6a4e28 508 fsid_type = FSID_ENCODE_DEV;
982aedfd 509 else
af6a4e28 510 fsid_type = FSID_DEV;
bc6c53d5
BF
511 fhp->fh_handle.fh_version = version;
512 if (version)
513 fhp->fh_handle.fh_fsid_type = fsid_type;
514}
515
516__be32
517fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry,
518 struct svc_fh *ref_fh)
519{
520 /* ref_fh is a reference file handle.
521 * if it is non-null and for the same filesystem, then we should compose
522 * a filehandle which is of the same version, where possible.
bc6c53d5
BF
523 */
524
2b0143b5 525 struct inode * inode = d_inode(dentry);
bc6c53d5
BF
526 dev_t ex_dev = exp_sb(exp)->s_dev;
527
97e47fa1 528 dprintk("nfsd: fh_compose(exp %02x:%02x/%ld %pd2, ino=%ld)\n",
bc6c53d5 529 MAJOR(ex_dev), MINOR(ex_dev),
2b0143b5 530 (long) d_inode(exp->ex_path.dentry)->i_ino,
97e47fa1 531 dentry,
bc6c53d5
BF
532 (inode ? inode->i_ino : 0));
533
534 /* Choose filehandle version and fsid type based on
535 * the reference filehandle (if it is in the same export)
536 * or the export options.
537 */
d28c442f 538 set_version_and_fsid_type(fhp, exp, ref_fh);
1da177e4 539
daab110e
JL
540 /* If we have a ref_fh, then copy the fh_no_wcc setting from it. */
541 fhp->fh_no_wcc = ref_fh ? ref_fh->fh_no_wcc : false;
542
1da177e4
LT
543 if (ref_fh == fhp)
544 fh_put(ref_fh);
545
dd8dd403 546 if (fhp->fh_dentry) {
97e47fa1
AV
547 printk(KERN_ERR "fh_compose: fh %pd2 not initialized!\n",
548 dentry);
1da177e4
LT
549 }
550 if (fhp->fh_maxsize < NFS_FHSIZE)
97e47fa1 551 printk(KERN_ERR "fh_compose: called with maxsize %d! %pd2\n",
982aedfd 552 fhp->fh_maxsize,
97e47fa1 553 dentry);
1da177e4
LT
554
555 fhp->fh_dentry = dget(dentry); /* our internal copy */
bf18f163 556 fhp->fh_export = exp_get(exp);
1da177e4 557
c645a883
N
558 fhp->fh_handle.fh_size =
559 key_len(fhp->fh_handle.fh_fsid_type) + 4;
560 fhp->fh_handle.fh_auth_type = 0;
561
562 mk_fsid(fhp->fh_handle.fh_fsid_type,
563 fhp->fh_handle.fh_fsid,
564 ex_dev,
565 d_inode(exp->ex_path.dentry)->i_ino,
566 exp->ex_fsid, exp->ex_uuid);
567
568 if (inode)
569 _fh_update(fhp, exp, dentry);
570 if (fhp->fh_handle.fh_fileid_type == FILEID_INVALID) {
571 fh_put(fhp);
572 return nfserr_opnotsupp;
1da177e4
LT
573 }
574
1da177e4
LT
575 return 0;
576}
577
578/*
579 * Update file handle information after changing a dentry.
580 * This is only called by nfsd_create, nfsd_create_v3 and nfsd_proc_create
581 */
83b11340 582__be32
1da177e4
LT
583fh_update(struct svc_fh *fhp)
584{
585 struct dentry *dentry;
982aedfd 586
1da177e4
LT
587 if (!fhp->fh_dentry)
588 goto out_bad;
589
590 dentry = fhp->fh_dentry;
2b0143b5 591 if (d_really_is_negative(dentry))
1da177e4 592 goto out_negative;
c645a883
N
593 if (fhp->fh_handle.fh_fileid_type != FILEID_ROOT)
594 return 0;
6e91ea2b 595
c645a883
N
596 _fh_update(fhp, fhp->fh_export, dentry);
597 if (fhp->fh_handle.fh_fileid_type == FILEID_INVALID)
598 return nfserr_opnotsupp;
1da177e4 599 return 0;
1da177e4
LT
600out_bad:
601 printk(KERN_ERR "fh_update: fh not verified!\n");
49e73720 602 return nfserr_serverfault;
1da177e4 603out_negative:
97e47fa1
AV
604 printk(KERN_ERR "fh_update: %pd2 still negative!\n",
605 dentry);
49e73720 606 return nfserr_serverfault;
1da177e4
LT
607}
608
fcb5e3fa
CL
609/**
610 * fh_fill_pre_attrs - Fill in pre-op attributes
611 * @fhp: file handle to be updated
612 *
613 */
614void fh_fill_pre_attrs(struct svc_fh *fhp)
615{
616 bool v4 = (fhp->fh_maxsize == NFS4_FHSIZE);
617 struct inode *inode;
618 struct kstat stat;
619 __be32 err;
620
621 if (fhp->fh_no_wcc || fhp->fh_pre_saved)
622 return;
623
624 inode = d_inode(fhp->fh_dentry);
625 err = fh_getattr(fhp, &stat);
626 if (err) {
627 /* Grab the times from inode anyway */
628 stat.mtime = inode->i_mtime;
629 stat.ctime = inode->i_ctime;
630 stat.size = inode->i_size;
638e3e7d
JL
631 if (v4 && IS_I_VERSION(inode)) {
632 stat.change_cookie = inode_query_iversion(inode);
633 stat.result_mask |= STATX_CHANGE_COOKIE;
634 }
fcb5e3fa
CL
635 }
636 if (v4)
637 fhp->fh_pre_change = nfsd4_change_attribute(&stat, inode);
638
639 fhp->fh_pre_mtime = stat.mtime;
640 fhp->fh_pre_ctime = stat.ctime;
641 fhp->fh_pre_size = stat.size;
642 fhp->fh_pre_saved = true;
643}
644
645/**
646 * fh_fill_post_attrs - Fill in post-op attributes
647 * @fhp: file handle to be updated
648 *
649 */
650void fh_fill_post_attrs(struct svc_fh *fhp)
651{
652 bool v4 = (fhp->fh_maxsize == NFS4_FHSIZE);
653 struct inode *inode = d_inode(fhp->fh_dentry);
654 __be32 err;
655
656 if (fhp->fh_no_wcc)
657 return;
658
659 if (fhp->fh_post_saved)
660 printk("nfsd: inode locked twice during operation.\n");
661
662 err = fh_getattr(fhp, &fhp->fh_post_attr);
663 if (err) {
664 fhp->fh_post_saved = false;
665 fhp->fh_post_attr.ctime = inode->i_ctime;
638e3e7d
JL
666 if (v4 && IS_I_VERSION(inode)) {
667 fhp->fh_post_attr.change_cookie = inode_query_iversion(inode);
668 fhp->fh_post_attr.result_mask |= STATX_CHANGE_COOKIE;
669 }
fcb5e3fa
CL
670 } else
671 fhp->fh_post_saved = true;
672 if (v4)
673 fhp->fh_post_change =
674 nfsd4_change_attribute(&fhp->fh_post_attr, inode);
675}
676
19d008b4
N
677/**
678 * fh_fill_both_attrs - Fill pre-op and post-op attributes
679 * @fhp: file handle to be updated
680 *
681 * This is used when the directory wasn't changed, but wcc attributes
682 * are needed anyway.
683 */
684void fh_fill_both_attrs(struct svc_fh *fhp)
685{
686 fh_fill_post_attrs(fhp);
687 if (!fhp->fh_post_saved)
688 return;
689 fhp->fh_pre_change = fhp->fh_post_change;
690 fhp->fh_pre_mtime = fhp->fh_post_attr.mtime;
691 fhp->fh_pre_ctime = fhp->fh_post_attr.ctime;
692 fhp->fh_pre_size = fhp->fh_post_attr.size;
693 fhp->fh_pre_saved = true;
694}
695
1da177e4
LT
696/*
697 * Release a file handle.
698 */
699void
700fh_put(struct svc_fh *fhp)
701{
702 struct dentry * dentry = fhp->fh_dentry;
703 struct svc_export * exp = fhp->fh_export;
704 if (dentry) {
1da177e4
LT
705 fhp->fh_dentry = NULL;
706 dput(dentry);
fcb5e3fa 707 fh_clear_pre_post_attrs(fhp);
1da177e4 708 }
4a55c101 709 fh_drop_write(fhp);
1da177e4 710 if (exp) {
a09581f2 711 exp_put(exp);
1da177e4
LT
712 fhp->fh_export = NULL;
713 }
daab110e 714 fhp->fh_no_wcc = false;
1da177e4
LT
715 return;
716}
717
718/*
719 * Shorthand for dprintk()'s
720 */
721char * SVCFH_fmt(struct svc_fh *fhp)
722{
723 struct knfsd_fh *fh = &fhp->fh_handle;
d8b26071 724 static char buf[2+1+1+64*3+1];
1da177e4 725
d8b26071
N
726 if (fh->fh_size < 0 || fh->fh_size> 64)
727 return "bad-fh";
728 sprintf(buf, "%d: %*ph", fh->fh_size, fh->fh_size, fh->fh_raw);
1da177e4
LT
729 return buf;
730}
af6a4e28 731
2c42f804 732enum fsid_source fsid_source(const struct svc_fh *fhp)
af6a4e28
N
733{
734 if (fhp->fh_handle.fh_version != 1)
735 return FSIDSOURCE_DEV;
736 switch(fhp->fh_handle.fh_fsid_type) {
737 case FSID_DEV:
738 case FSID_ENCODE_DEV:
739 case FSID_MAJOR_MINOR:
8e498751 740 if (exp_sb(fhp->fh_export)->s_type->fs_flags & FS_REQUIRES_DEV)
b8da0d1c
NB
741 return FSIDSOURCE_DEV;
742 break;
af6a4e28 743 case FSID_NUM:
af6a4e28
N
744 if (fhp->fh_export->ex_flags & NFSEXP_FSID)
745 return FSIDSOURCE_FSID;
b8da0d1c
NB
746 break;
747 default:
748 break;
af6a4e28 749 }
b8da0d1c
NB
750 /* either a UUID type filehandle, or the filehandle doesn't
751 * match the export.
752 */
753 if (fhp->fh_export->ex_flags & NFSEXP_FSID)
754 return FSIDSOURCE_FSID;
755 if (fhp->fh_export->ex_uuid)
756 return FSIDSOURCE_UUID;
757 return FSIDSOURCE_DEV;
af6a4e28 758}
3139b1d7
JL
759
760/*
638e3e7d
JL
761 * We could use i_version alone as the change attribute. However, i_version
762 * can go backwards on a regular file after an unclean shutdown. On its own
763 * that doesn't necessarily cause a problem, but if i_version goes backwards
764 * and then is incremented again it could reuse a value that was previously
765 * used before boot, and a client who queried the two values might incorrectly
766 * assume nothing changed.
767 *
768 * By using both ctime and the i_version counter we guarantee that as long as
769 * time doesn't go backwards we never reuse an old value. If the filesystem
770 * advertises STATX_ATTR_CHANGE_MONOTONIC, then this mitigation is not
771 * needed.
3139b1d7 772 *
638e3e7d
JL
773 * We only need to do this for regular files as well. For directories, we
774 * assume that the new change attr is always logged to stable storage in some
775 * fashion before the results can be seen.
3139b1d7
JL
776 */
777u64 nfsd4_change_attribute(struct kstat *stat, struct inode *inode)
778{
638e3e7d
JL
779 u64 chattr;
780
638e3e7d
JL
781 if (stat->result_mask & STATX_CHANGE_COOKIE) {
782 chattr = stat->change_cookie;
638e3e7d
JL
783 if (S_ISREG(inode->i_mode) &&
784 !(stat->attributes & STATX_ATTR_CHANGE_MONOTONIC)) {
785 chattr += (u64)stat->ctime.tv_sec << 30;
786 chattr += stat->ctime.tv_nsec;
787 }
788 } else {
789 chattr = time_to_chattr(&stat->ctime);
790 }
791 return chattr;
3139b1d7 792}