]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * File operations used by nfsd. Some of these have been ripped from |
3 | * other parts of the kernel because they weren't exported, others | |
4 | * are partial duplicates with added or changed functionality. | |
5 | * | |
6 | * Note that several functions dget() the dentry upon which they want | |
7 | * to act, most notably those that create directory entries. Response | |
8 | * dentry's are dput()'d if necessary in the release callback. | |
9 | * So if you notice code paths that apparently fail to dput() the | |
10 | * dentry, don't worry--they have been taken care of. | |
11 | * | |
12 | * Copyright (C) 1995-1999 Olaf Kirch <okir@monad.swb.de> | |
13 | * Zerocpy NFS support (C) 2002 Hirokazu Takahashi <taka@valinux.co.jp> | |
14 | */ | |
15 | ||
1da177e4 LT |
16 | #include <linux/fs.h> |
17 | #include <linux/file.h> | |
d6b29d7c | 18 | #include <linux/splice.h> |
1da177e4 | 19 | #include <linux/fcntl.h> |
1da177e4 | 20 | #include <linux/namei.h> |
1da177e4 | 21 | #include <linux/delay.h> |
0eeca283 | 22 | #include <linux/fsnotify.h> |
1da177e4 | 23 | #include <linux/posix_acl_xattr.h> |
1da177e4 | 24 | #include <linux/xattr.h> |
9a74af21 BH |
25 | #include <linux/jhash.h> |
26 | #include <linux/ima.h> | |
5a0e3ad6 | 27 | #include <linux/slab.h> |
9a74af21 | 28 | #include <asm/uaccess.h> |
f501912a BM |
29 | #include <linux/exportfs.h> |
30 | #include <linux/writeback.h> | |
18032ca0 | 31 | #include <linux/security.h> |
9a74af21 BH |
32 | |
33 | #ifdef CONFIG_NFSD_V3 | |
34 | #include "xdr3.h" | |
35 | #endif /* CONFIG_NFSD_V3 */ | |
36 | ||
5be196e5 | 37 | #ifdef CONFIG_NFSD_V4 |
2ca72e17 BF |
38 | #include "acl.h" |
39 | #include "idmap.h" | |
1da177e4 LT |
40 | #endif /* CONFIG_NFSD_V4 */ |
41 | ||
9a74af21 BH |
42 | #include "nfsd.h" |
43 | #include "vfs.h" | |
1da177e4 LT |
44 | |
45 | #define NFSDDBG_FACILITY NFSDDBG_FILEOP | |
1da177e4 LT |
46 | |
47 | ||
1da177e4 LT |
48 | /* |
49 | * This is a cache of readahead params that help us choose the proper | |
50 | * readahead strategy. Initially, we set all readahead parameters to 0 | |
51 | * and let the VFS handle things. | |
52 | * If you increase the number of cached files very much, you'll need to | |
53 | * add a hash table here. | |
54 | */ | |
55 | struct raparms { | |
56 | struct raparms *p_next; | |
57 | unsigned int p_count; | |
58 | ino_t p_ino; | |
59 | dev_t p_dev; | |
60 | int p_set; | |
61 | struct file_ra_state p_ra; | |
fce1456a | 62 | unsigned int p_hindex; |
1da177e4 LT |
63 | }; |
64 | ||
fce1456a GB |
65 | struct raparm_hbucket { |
66 | struct raparms *pb_head; | |
67 | spinlock_t pb_lock; | |
68 | } ____cacheline_aligned_in_smp; | |
69 | ||
fce1456a GB |
70 | #define RAPARM_HASH_BITS 4 |
71 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
72 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
73 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
74 | |
75 | /* | |
76 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
77 | * a mount point. | |
e0bb89ef | 78 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
79 | * or nfs_ok having possibly changed *dpp and *expp |
80 | */ | |
81 | int | |
82 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
83 | struct svc_export **expp) | |
84 | { | |
85 | struct svc_export *exp = *expp, *exp2 = NULL; | |
86 | struct dentry *dentry = *dpp; | |
91c9fa8f AV |
87 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
88 | .dentry = dget(dentry)}; | |
6264d69d | 89 | int err = 0; |
1da177e4 | 90 | |
7cc90cc3 | 91 | err = follow_down(&path); |
cc53ce53 DH |
92 | if (err < 0) |
93 | goto out; | |
1da177e4 | 94 | |
91c9fa8f | 95 | exp2 = rqst_exp_get_by_name(rqstp, &path); |
1da177e4 | 96 | if (IS_ERR(exp2)) { |
3b6cee7b BF |
97 | err = PTR_ERR(exp2); |
98 | /* | |
99 | * We normally allow NFS clients to continue | |
100 | * "underneath" a mountpoint that is not exported. | |
101 | * The exception is V4ROOT, where no traversal is ever | |
102 | * allowed without an explicit export of the new | |
103 | * directory. | |
104 | */ | |
105 | if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT)) | |
106 | err = 0; | |
91c9fa8f | 107 | path_put(&path); |
1da177e4 LT |
108 | goto out; |
109 | } | |
3c394dda SD |
110 | if (nfsd_v4client(rqstp) || |
111 | (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { | |
1da177e4 | 112 | /* successfully crossed mount point */ |
1644ccc8 | 113 | /* |
91c9fa8f AV |
114 | * This is subtle: path.dentry is *not* on path.mnt |
115 | * at this point. The only reason we are safe is that | |
116 | * original mnt is pinned down by exp, so we should | |
117 | * put path *before* putting exp | |
1644ccc8 | 118 | */ |
91c9fa8f AV |
119 | *dpp = path.dentry; |
120 | path.dentry = dentry; | |
1644ccc8 | 121 | *expp = exp2; |
91c9fa8f | 122 | exp2 = exp; |
1da177e4 | 123 | } |
91c9fa8f AV |
124 | path_put(&path); |
125 | exp_put(exp2); | |
1da177e4 LT |
126 | out: |
127 | return err; | |
128 | } | |
129 | ||
289ede45 BF |
130 | static void follow_to_parent(struct path *path) |
131 | { | |
132 | struct dentry *dp; | |
133 | ||
134 | while (path->dentry == path->mnt->mnt_root && follow_up(path)) | |
135 | ; | |
136 | dp = dget_parent(path->dentry); | |
137 | dput(path->dentry); | |
138 | path->dentry = dp; | |
139 | } | |
140 | ||
141 | static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp) | |
142 | { | |
143 | struct svc_export *exp2; | |
144 | struct path path = {.mnt = mntget((*exp)->ex_path.mnt), | |
145 | .dentry = dget(dparent)}; | |
146 | ||
147 | follow_to_parent(&path); | |
148 | ||
149 | exp2 = rqst_exp_parent(rqstp, &path); | |
150 | if (PTR_ERR(exp2) == -ENOENT) { | |
151 | *dentryp = dget(dparent); | |
152 | } else if (IS_ERR(exp2)) { | |
153 | path_put(&path); | |
154 | return PTR_ERR(exp2); | |
155 | } else { | |
156 | *dentryp = dget(path.dentry); | |
157 | exp_put(*exp); | |
158 | *exp = exp2; | |
159 | } | |
160 | path_put(&path); | |
161 | return 0; | |
162 | } | |
163 | ||
82ead7fe BF |
164 | /* |
165 | * For nfsd purposes, we treat V4ROOT exports as though there was an | |
166 | * export at *every* directory. | |
167 | */ | |
3227fa41 | 168 | int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp) |
82ead7fe BF |
169 | { |
170 | if (d_mountpoint(dentry)) | |
171 | return 1; | |
11fcee02 TM |
172 | if (nfsd4_is_junction(dentry)) |
173 | return 1; | |
82ead7fe BF |
174 | if (!(exp->ex_flags & NFSEXP_V4ROOT)) |
175 | return 0; | |
176 | return dentry->d_inode != NULL; | |
177 | } | |
178 | ||
6264d69d | 179 | __be32 |
6c0a654d | 180 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 181 | const char *name, unsigned int len, |
6c0a654d | 182 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
183 | { |
184 | struct svc_export *exp; | |
185 | struct dentry *dparent; | |
186 | struct dentry *dentry; | |
6264d69d | 187 | int host_err; |
1da177e4 LT |
188 | |
189 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
190 | ||
1da177e4 LT |
191 | dparent = fhp->fh_dentry; |
192 | exp = fhp->fh_export; | |
193 | exp_get(exp); | |
194 | ||
1da177e4 LT |
195 | /* Lookup the name, but don't follow links */ |
196 | if (isdotent(name, len)) { | |
197 | if (len==1) | |
198 | dentry = dget(dparent); | |
54775491 | 199 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 200 | dentry = dget_parent(dparent); |
fed83811 | 201 | else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp)) |
1da177e4 LT |
202 | dentry = dget(dparent); /* .. == . just like at / */ |
203 | else { | |
204 | /* checking mountpoint crossing is very different when stepping up */ | |
289ede45 BF |
205 | host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry); |
206 | if (host_err) | |
1da177e4 | 207 | goto out_nfserr; |
1da177e4 LT |
208 | } |
209 | } else { | |
210 | fh_lock(fhp); | |
211 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 212 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
213 | if (IS_ERR(dentry)) |
214 | goto out_nfserr; | |
215 | /* | |
216 | * check if we have crossed a mount point ... | |
217 | */ | |
82ead7fe | 218 | if (nfsd_mountpoint(dentry, exp)) { |
6264d69d | 219 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
220 | dput(dentry); |
221 | goto out_nfserr; | |
222 | } | |
223 | } | |
224 | } | |
6c0a654d BF |
225 | *dentry_ret = dentry; |
226 | *exp_ret = exp; | |
227 | return 0; | |
228 | ||
229 | out_nfserr: | |
230 | exp_put(exp); | |
231 | return nfserrno(host_err); | |
232 | } | |
233 | ||
234 | /* | |
235 | * Look up one component of a pathname. | |
236 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
237 | * | |
238 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
239 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
240 | * accepted as it stands and the mounted directory is | |
241 | * returned. Otherwise the covered directory is returned. | |
242 | * NOTE: this mountpoint crossing is not supported properly by all | |
243 | * clients and is explicitly disallowed for NFSv3 | |
244 | * NeilBrown <neilb@cse.unsw.edu.au> | |
245 | */ | |
246 | __be32 | |
247 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 248 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
249 | { |
250 | struct svc_export *exp; | |
251 | struct dentry *dentry; | |
252 | __be32 err; | |
253 | ||
29a78a3e BF |
254 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
255 | if (err) | |
256 | return err; | |
6c0a654d BF |
257 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); |
258 | if (err) | |
259 | return err; | |
32c1eb0c AA |
260 | err = check_nfsd_access(exp, rqstp); |
261 | if (err) | |
262 | goto out; | |
1da177e4 LT |
263 | /* |
264 | * Note: we compose the file handle now, but as the | |
265 | * dentry may be negative, it may need to be updated. | |
266 | */ | |
267 | err = fh_compose(resfh, exp, dentry, fhp); | |
268 | if (!err && !dentry->d_inode) | |
269 | err = nfserr_noent; | |
32c1eb0c | 270 | out: |
1da177e4 | 271 | dput(dentry); |
1da177e4 LT |
272 | exp_put(exp); |
273 | return err; | |
1da177e4 LT |
274 | } |
275 | ||
4795bb37 BF |
276 | static int nfsd_break_lease(struct inode *inode) |
277 | { | |
278 | if (!S_ISREG(inode->i_mode)) | |
279 | return 0; | |
280 | return break_lease(inode, O_WRONLY | O_NONBLOCK); | |
281 | } | |
282 | ||
f501912a BM |
283 | /* |
284 | * Commit metadata changes to stable storage. | |
285 | */ | |
286 | static int | |
287 | commit_metadata(struct svc_fh *fhp) | |
288 | { | |
289 | struct inode *inode = fhp->fh_dentry->d_inode; | |
290 | const struct export_operations *export_ops = inode->i_sb->s_export_op; | |
f501912a BM |
291 | |
292 | if (!EX_ISSYNC(fhp->fh_export)) | |
293 | return 0; | |
294 | ||
c3765016 CH |
295 | if (export_ops->commit_metadata) |
296 | return export_ops->commit_metadata(inode); | |
297 | return sync_inode_metadata(inode, 1); | |
f501912a | 298 | } |
6c0a654d | 299 | |
1da177e4 | 300 | /* |
818e5a22 CH |
301 | * Go over the attributes and take care of the small differences between |
302 | * NFS semantics and what Linux expects. | |
1da177e4 | 303 | */ |
818e5a22 CH |
304 | static void |
305 | nfsd_sanitize_attrs(struct inode *inode, struct iattr *iap) | |
1da177e4 | 306 | { |
9c85fca5 CH |
307 | /* |
308 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
309 | * which only requires access, and "set-[ac]time-to-X" which |
310 | * requires ownership. | |
311 | * So if it looks like it might be "set both to the same time which | |
312 | * is close to now", and if inode_change_ok fails, then we | |
313 | * convert to "set to now" instead of "set to explicit time" | |
314 | * | |
315 | * We only call inode_change_ok as the last test as technically | |
818e5a22 | 316 | * it is not an interface that we should be using. |
1da177e4 LT |
317 | */ |
318 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
319 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
320 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
321 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
322 | /* | |
323 | * Looks probable. | |
324 | * | |
325 | * Now just make sure time is in the right ballpark. | |
326 | * Solaris, at least, doesn't seem to care what the time | |
327 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 328 | */ |
9c85fca5 CH |
329 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
330 | if (delta < 0) | |
331 | delta = -delta; | |
332 | if (delta < MAX_TOUCH_TIME_ERROR && | |
333 | inode_change_ok(inode, iap) != 0) { | |
334 | /* | |
335 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
336 | * This will cause notify_change to set these times | |
337 | * to "now" | |
338 | */ | |
339 | iap->ia_valid &= ~BOTH_TIME_SET; | |
340 | } | |
1da177e4 | 341 | } |
1da177e4 | 342 | |
ca456252 | 343 | /* sanitize the mode change */ |
1da177e4 LT |
344 | if (iap->ia_valid & ATTR_MODE) { |
345 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 346 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
347 | } |
348 | ||
349 | /* Revoke setuid/setgid on chown */ | |
0953e620 | 350 | if (!S_ISDIR(inode->i_mode) && |
6fab8779 EB |
351 | (((iap->ia_valid & ATTR_UID) && !uid_eq(iap->ia_uid, inode->i_uid)) || |
352 | ((iap->ia_valid & ATTR_GID) && !gid_eq(iap->ia_gid, inode->i_gid)))) { | |
ca456252 JL |
353 | iap->ia_valid |= ATTR_KILL_PRIV; |
354 | if (iap->ia_valid & ATTR_MODE) { | |
355 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 356 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
357 | if (iap->ia_mode & S_IXGRP) |
358 | iap->ia_mode &= ~S_ISGID; | |
359 | } else { | |
360 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
361 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 362 | } |
1da177e4 | 363 | } |
818e5a22 CH |
364 | } |
365 | ||
366 | static __be32 | |
367 | nfsd_get_write_access(struct svc_rqst *rqstp, struct svc_fh *fhp, | |
368 | struct iattr *iap) | |
369 | { | |
370 | struct inode *inode = fhp->fh_dentry->d_inode; | |
371 | int host_err; | |
372 | ||
373 | if (iap->ia_size < inode->i_size) { | |
374 | __be32 err; | |
375 | ||
376 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, | |
377 | NFSD_MAY_TRUNC | NFSD_MAY_OWNER_OVERRIDE); | |
378 | if (err) | |
379 | return err; | |
380 | } | |
381 | ||
382 | host_err = get_write_access(inode); | |
383 | if (host_err) | |
384 | goto out_nfserrno; | |
385 | ||
386 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); | |
387 | if (host_err) | |
388 | goto out_put_write_access; | |
389 | return 0; | |
390 | ||
391 | out_put_write_access: | |
392 | put_write_access(inode); | |
393 | out_nfserrno: | |
394 | return nfserrno(host_err); | |
395 | } | |
396 | ||
397 | /* | |
398 | * Set various file attributes. After this call fhp needs an fh_put. | |
399 | */ | |
400 | __be32 | |
401 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, | |
402 | int check_guard, time_t guardtime) | |
403 | { | |
404 | struct dentry *dentry; | |
405 | struct inode *inode; | |
406 | int accmode = NFSD_MAY_SATTR; | |
407 | umode_t ftype = 0; | |
408 | __be32 err; | |
409 | int host_err; | |
410 | int size_change = 0; | |
411 | ||
412 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
413 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; | |
414 | if (iap->ia_valid & ATTR_SIZE) | |
415 | ftype = S_IFREG; | |
416 | ||
417 | /* Get inode */ | |
418 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
419 | if (err) | |
420 | goto out; | |
421 | ||
422 | dentry = fhp->fh_dentry; | |
423 | inode = dentry->d_inode; | |
424 | ||
425 | /* Ignore any mode updates on symlinks */ | |
426 | if (S_ISLNK(inode->i_mode)) | |
427 | iap->ia_valid &= ~ATTR_MODE; | |
428 | ||
429 | if (!iap->ia_valid) | |
430 | goto out; | |
1da177e4 | 431 | |
818e5a22 CH |
432 | nfsd_sanitize_attrs(inode, iap); |
433 | ||
434 | /* | |
435 | * The size case is special, it changes the file in addition to the | |
436 | * attributes. | |
437 | */ | |
438 | if (iap->ia_valid & ATTR_SIZE) { | |
439 | err = nfsd_get_write_access(rqstp, fhp, iap); | |
440 | if (err) | |
441 | goto out; | |
442 | size_change = 1; | |
443 | } | |
1da177e4 LT |
444 | |
445 | iap->ia_valid |= ATTR_CTIME; | |
446 | ||
987da479 CH |
447 | if (check_guard && guardtime != inode->i_ctime.tv_sec) { |
448 | err = nfserr_notsync; | |
449 | goto out_put_write_access; | |
1da177e4 | 450 | } |
987da479 CH |
451 | |
452 | host_err = nfsd_break_lease(inode); | |
453 | if (host_err) | |
454 | goto out_put_write_access_nfserror; | |
455 | ||
456 | fh_lock(fhp); | |
457 | host_err = notify_change(dentry, iap, NULL); | |
458 | fh_unlock(fhp); | |
459 | ||
460 | out_put_write_access_nfserror: | |
461 | err = nfserrno(host_err); | |
462 | out_put_write_access: | |
1da177e4 LT |
463 | if (size_change) |
464 | put_write_access(inode); | |
465 | if (!err) | |
b160fdab | 466 | commit_metadata(fhp); |
1da177e4 LT |
467 | out: |
468 | return err; | |
1da177e4 LT |
469 | } |
470 | ||
5be196e5 CH |
471 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
472 | defined(CONFIG_NFSD_V3_ACL) || \ | |
473 | defined(CONFIG_NFSD_V4) | |
474 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
475 | { | |
476 | ssize_t buflen; | |
6c6a426f | 477 | ssize_t ret; |
5be196e5 CH |
478 | |
479 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
480 | if (buflen <= 0) | |
481 | return buflen; | |
1da177e4 | 482 | |
5be196e5 CH |
483 | *buf = kmalloc(buflen, GFP_KERNEL); |
484 | if (!*buf) | |
485 | return -ENOMEM; | |
486 | ||
6c6a426f KK |
487 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
488 | if (ret < 0) | |
489 | kfree(*buf); | |
490 | return ret; | |
5be196e5 CH |
491 | } |
492 | #endif | |
493 | ||
494 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
495 | static int |
496 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
497 | { | |
498 | int len; | |
499 | size_t buflen; | |
500 | char *buf = NULL; | |
501 | int error = 0; | |
1da177e4 LT |
502 | |
503 | buflen = posix_acl_xattr_size(pacl->a_count); | |
504 | buf = kmalloc(buflen, GFP_KERNEL); | |
505 | error = -ENOMEM; | |
506 | if (buf == NULL) | |
507 | goto out; | |
508 | ||
5f3a4a28 | 509 | len = posix_acl_to_xattr(&init_user_ns, pacl, buf, buflen); |
1da177e4 LT |
510 | if (len < 0) { |
511 | error = len; | |
512 | goto out; | |
513 | } | |
514 | ||
5be196e5 | 515 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
516 | out: |
517 | kfree(buf); | |
518 | return error; | |
519 | } | |
520 | ||
6264d69d | 521 | __be32 |
1da177e4 LT |
522 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
523 | struct nfs4_acl *acl) | |
524 | { | |
6264d69d AV |
525 | __be32 error; |
526 | int host_error; | |
1da177e4 LT |
527 | struct dentry *dentry; |
528 | struct inode *inode; | |
529 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
530 | unsigned int flags = 0; | |
531 | ||
532 | /* Get inode */ | |
e281d810 | 533 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_SATTR); |
1da177e4 | 534 | if (error) |
4b2ca38a | 535 | return error; |
1da177e4 LT |
536 | |
537 | dentry = fhp->fh_dentry; | |
538 | inode = dentry->d_inode; | |
539 | if (S_ISDIR(inode->i_mode)) | |
540 | flags = NFS4_ACL_DIR; | |
541 | ||
6264d69d AV |
542 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
543 | if (host_error == -EINVAL) { | |
4b2ca38a | 544 | return nfserr_attrnotsupp; |
6264d69d | 545 | } else if (host_error < 0) |
1da177e4 LT |
546 | goto out_nfserr; |
547 | ||
6264d69d AV |
548 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
549 | if (host_error < 0) | |
4b2ca38a | 550 | goto out_release; |
1da177e4 | 551 | |
4b2ca38a | 552 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 553 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 554 | |
4b2ca38a | 555 | out_release: |
1da177e4 LT |
556 | posix_acl_release(pacl); |
557 | posix_acl_release(dpacl); | |
1da177e4 | 558 | out_nfserr: |
f34f9242 | 559 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 560 | return nfserr_attrnotsupp; |
f34f9242 | 561 | else |
4b2ca38a | 562 | return nfserrno(host_error); |
1da177e4 LT |
563 | } |
564 | ||
565 | static struct posix_acl * | |
566 | _get_posix_acl(struct dentry *dentry, char *key) | |
567 | { | |
5be196e5 | 568 | void *buf = NULL; |
1da177e4 | 569 | struct posix_acl *pacl = NULL; |
5be196e5 | 570 | int buflen; |
1da177e4 | 571 | |
5be196e5 CH |
572 | buflen = nfsd_getxattr(dentry, key, &buf); |
573 | if (!buflen) | |
574 | buflen = -ENODATA; | |
575 | if (buflen <= 0) | |
576 | return ERR_PTR(buflen); | |
1da177e4 | 577 | |
5f3a4a28 | 578 | pacl = posix_acl_from_xattr(&init_user_ns, buf, buflen); |
1da177e4 LT |
579 | kfree(buf); |
580 | return pacl; | |
1da177e4 LT |
581 | } |
582 | ||
583 | int | |
584 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
585 | { | |
586 | struct inode *inode = dentry->d_inode; | |
587 | int error = 0; | |
588 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
589 | unsigned int flags = 0; | |
590 | ||
9a59f452 | 591 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
592 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
593 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
594 | if (IS_ERR(pacl)) { | |
595 | error = PTR_ERR(pacl); | |
596 | pacl = NULL; | |
597 | goto out; | |
598 | } | |
599 | ||
600 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 601 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
602 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
603 | dpacl = NULL; | |
604 | else if (IS_ERR(dpacl)) { | |
605 | error = PTR_ERR(dpacl); | |
606 | dpacl = NULL; | |
607 | goto out; | |
608 | } | |
609 | flags = NFS4_ACL_DIR; | |
610 | } | |
611 | ||
612 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
613 | if (IS_ERR(*acl)) { | |
614 | error = PTR_ERR(*acl); | |
615 | *acl = NULL; | |
616 | } | |
617 | out: | |
618 | posix_acl_release(pacl); | |
619 | posix_acl_release(dpacl); | |
620 | return error; | |
621 | } | |
622 | ||
9b4146e8 CL |
623 | /* |
624 | * NFS junction information is stored in an extended attribute. | |
625 | */ | |
626 | #define NFSD_JUNCTION_XATTR_NAME XATTR_TRUSTED_PREFIX "junction.nfs" | |
627 | ||
628 | /** | |
629 | * nfsd4_is_junction - Test if an object could be an NFS junction | |
630 | * | |
631 | * @dentry: object to test | |
632 | * | |
633 | * Returns 1 if "dentry" appears to contain NFS junction information. | |
634 | * Otherwise 0 is returned. | |
635 | */ | |
11fcee02 TM |
636 | int nfsd4_is_junction(struct dentry *dentry) |
637 | { | |
638 | struct inode *inode = dentry->d_inode; | |
639 | ||
640 | if (inode == NULL) | |
641 | return 0; | |
642 | if (inode->i_mode & S_IXUGO) | |
643 | return 0; | |
644 | if (!(inode->i_mode & S_ISVTX)) | |
645 | return 0; | |
9b4146e8 | 646 | if (vfs_getxattr(dentry, NFSD_JUNCTION_XATTR_NAME, NULL, 0) <= 0) |
11fcee02 TM |
647 | return 0; |
648 | return 1; | |
649 | } | |
18032ca0 DQ |
650 | #ifdef CONFIG_NFSD_V4_SECURITY_LABEL |
651 | __be32 nfsd4_set_nfs4_label(struct svc_rqst *rqstp, struct svc_fh *fhp, | |
652 | struct xdr_netobj *label) | |
653 | { | |
654 | __be32 error; | |
655 | int host_error; | |
656 | struct dentry *dentry; | |
657 | ||
658 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); | |
659 | if (error) | |
660 | return error; | |
661 | ||
662 | dentry = fhp->fh_dentry; | |
663 | ||
664 | mutex_lock(&dentry->d_inode->i_mutex); | |
665 | host_error = security_inode_setsecctx(dentry, label->data, label->len); | |
666 | mutex_unlock(&dentry->d_inode->i_mutex); | |
667 | return nfserrno(host_error); | |
668 | } | |
669 | #else | |
670 | __be32 nfsd4_set_nfs4_label(struct svc_rqst *rqstp, struct svc_fh *fhp, | |
671 | struct xdr_netobj *label) | |
672 | { | |
673 | return nfserr_notsupp; | |
674 | } | |
675 | #endif | |
676 | ||
6a85d6c7 | 677 | #endif /* defined(CONFIG_NFSD_V4) */ |
1da177e4 LT |
678 | |
679 | #ifdef CONFIG_NFSD_V3 | |
680 | /* | |
681 | * Check server access rights to a file system object | |
682 | */ | |
683 | struct accessmap { | |
684 | u32 access; | |
685 | int how; | |
686 | }; | |
687 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
688 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
689 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
690 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
691 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
692 | |
693 | { 0, 0 } | |
694 | }; | |
695 | ||
696 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
697 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
698 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
699 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
700 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
701 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
702 | |
703 | { 0, 0 } | |
704 | }; | |
705 | ||
706 | static struct accessmap nfs3_anyaccess[] = { | |
707 | /* Some clients - Solaris 2.6 at least, make an access call | |
708 | * to the server to check for access for things like /dev/null | |
709 | * (which really, the server doesn't care about). So | |
710 | * We provide simple access checking for them, looking | |
711 | * mainly at mode bits, and we make sure to ignore read-only | |
712 | * filesystem checks | |
713 | */ | |
8837abca MS |
714 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
715 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
716 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
717 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
718 | |
719 | { 0, 0 } | |
720 | }; | |
721 | ||
6264d69d | 722 | __be32 |
1da177e4 LT |
723 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
724 | { | |
725 | struct accessmap *map; | |
726 | struct svc_export *export; | |
727 | struct dentry *dentry; | |
728 | u32 query, result = 0, sresult = 0; | |
6264d69d | 729 | __be32 error; |
1da177e4 | 730 | |
8837abca | 731 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
732 | if (error) |
733 | goto out; | |
734 | ||
735 | export = fhp->fh_export; | |
736 | dentry = fhp->fh_dentry; | |
737 | ||
738 | if (S_ISREG(dentry->d_inode->i_mode)) | |
739 | map = nfs3_regaccess; | |
740 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
741 | map = nfs3_diraccess; | |
742 | else | |
743 | map = nfs3_anyaccess; | |
744 | ||
745 | ||
746 | query = *access; | |
747 | for (; map->access; map++) { | |
748 | if (map->access & query) { | |
6264d69d | 749 | __be32 err2; |
1da177e4 LT |
750 | |
751 | sresult |= map->access; | |
752 | ||
0ec757df | 753 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
754 | switch (err2) { |
755 | case nfs_ok: | |
756 | result |= map->access; | |
757 | break; | |
758 | ||
759 | /* the following error codes just mean the access was not allowed, | |
760 | * rather than an error occurred */ | |
761 | case nfserr_rofs: | |
762 | case nfserr_acces: | |
763 | case nfserr_perm: | |
764 | /* simply don't "or" in the access bit. */ | |
765 | break; | |
766 | default: | |
767 | error = err2; | |
768 | goto out; | |
769 | } | |
770 | } | |
771 | } | |
772 | *access = result; | |
773 | if (supported) | |
774 | *supported = sresult; | |
775 | ||
776 | out: | |
777 | return error; | |
778 | } | |
779 | #endif /* CONFIG_NFSD_V3 */ | |
780 | ||
105f4622 BF |
781 | static int nfsd_open_break_lease(struct inode *inode, int access) |
782 | { | |
783 | unsigned int mode; | |
1da177e4 | 784 | |
105f4622 BF |
785 | if (access & NFSD_MAY_NOT_BREAK_LEASE) |
786 | return 0; | |
787 | mode = (access & NFSD_MAY_WRITE) ? O_WRONLY : O_RDONLY; | |
788 | return break_lease(inode, mode | O_NONBLOCK); | |
789 | } | |
1da177e4 LT |
790 | |
791 | /* | |
792 | * Open an existing file or directory. | |
999448a8 BS |
793 | * The may_flags argument indicates the type of open (read/write/lock) |
794 | * and additional flags. | |
1da177e4 LT |
795 | * N.B. After this call fhp needs an fh_put |
796 | */ | |
6264d69d | 797 | __be32 |
175a4eb7 | 798 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type, |
999448a8 | 799 | int may_flags, struct file **filp) |
1da177e4 | 800 | { |
765927b2 | 801 | struct path path; |
1da177e4 | 802 | struct inode *inode; |
6264d69d AV |
803 | int flags = O_RDONLY|O_LARGEFILE; |
804 | __be32 err; | |
91885258 | 805 | int host_err = 0; |
1da177e4 | 806 | |
e0e81739 DH |
807 | validate_process_creds(); |
808 | ||
1da177e4 LT |
809 | /* |
810 | * If we get here, then the client has already done an "open", | |
811 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
812 | * in case a chmod has now revoked permission. | |
d91d0b56 BF |
813 | * |
814 | * Arguably we should also allow the owner override for | |
815 | * directories, but we never have and it doesn't seem to have | |
816 | * caused anyone a problem. If we were to change this, note | |
817 | * also that our filldir callbacks would need a variant of | |
818 | * lookup_one_len that doesn't check permissions. | |
1da177e4 | 819 | */ |
d91d0b56 BF |
820 | if (type == S_IFREG) |
821 | may_flags |= NFSD_MAY_OWNER_OVERRIDE; | |
822 | err = fh_verify(rqstp, fhp, type, may_flags); | |
1da177e4 LT |
823 | if (err) |
824 | goto out; | |
825 | ||
765927b2 AV |
826 | path.mnt = fhp->fh_export->ex_path.mnt; |
827 | path.dentry = fhp->fh_dentry; | |
828 | inode = path.dentry->d_inode; | |
1da177e4 LT |
829 | |
830 | /* Disallow write access to files with the append-only bit set | |
831 | * or any access when mandatory locking enabled | |
832 | */ | |
833 | err = nfserr_perm; | |
999448a8 | 834 | if (IS_APPEND(inode) && (may_flags & NFSD_MAY_WRITE)) |
1da177e4 | 835 | goto out; |
5e7fc436 BF |
836 | /* |
837 | * We must ignore files (but only files) which might have mandatory | |
838 | * locks on them because there is no way to know if the accesser has | |
839 | * the lock. | |
840 | */ | |
841 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
842 | goto out; |
843 | ||
844 | if (!inode->i_fop) | |
845 | goto out; | |
846 | ||
999448a8 | 847 | host_err = nfsd_open_break_lease(inode, may_flags); |
6264d69d | 848 | if (host_err) /* NOMEM or WOULDBLOCK */ |
1da177e4 LT |
849 | goto out_nfserr; |
850 | ||
999448a8 BS |
851 | if (may_flags & NFSD_MAY_WRITE) { |
852 | if (may_flags & NFSD_MAY_READ) | |
9ecb6a08 BF |
853 | flags = O_RDWR|O_LARGEFILE; |
854 | else | |
855 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 856 | } |
765927b2 | 857 | *filp = dentry_open(&path, flags, current_cred()); |
e4daf1ff | 858 | if (IS_ERR(*filp)) { |
6264d69d | 859 | host_err = PTR_ERR(*filp); |
e4daf1ff HJ |
860 | *filp = NULL; |
861 | } else { | |
999448a8 BS |
862 | host_err = ima_file_check(*filp, may_flags); |
863 | ||
06effdbb BS |
864 | if (may_flags & NFSD_MAY_64BIT_COOKIE) |
865 | (*filp)->f_mode |= FMODE_64BITHASH; | |
866 | else | |
867 | (*filp)->f_mode |= FMODE_32BITHASH; | |
868 | } | |
869 | ||
1da177e4 | 870 | out_nfserr: |
6264d69d | 871 | err = nfserrno(host_err); |
1da177e4 | 872 | out: |
e0e81739 | 873 | validate_process_creds(); |
1da177e4 LT |
874 | return err; |
875 | } | |
876 | ||
877 | /* | |
878 | * Close a file. | |
879 | */ | |
880 | void | |
881 | nfsd_close(struct file *filp) | |
882 | { | |
883 | fput(filp); | |
884 | } | |
885 | ||
1da177e4 LT |
886 | /* |
887 | * Obtain the readahead parameters for the file | |
888 | * specified by (dev, ino). | |
889 | */ | |
1da177e4 LT |
890 | |
891 | static inline struct raparms * | |
892 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
893 | { | |
894 | struct raparms *ra, **rap, **frap = NULL; | |
895 | int depth = 0; | |
fce1456a GB |
896 | unsigned int hash; |
897 | struct raparm_hbucket *rab; | |
898 | ||
899 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
900 | rab = &raparm_hash[hash]; | |
1da177e4 | 901 | |
fce1456a GB |
902 | spin_lock(&rab->pb_lock); |
903 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
904 | if (ra->p_ino == ino && ra->p_dev == dev) |
905 | goto found; | |
906 | depth++; | |
907 | if (ra->p_count == 0) | |
908 | frap = rap; | |
909 | } | |
3aa6e0aa | 910 | depth = nfsdstats.ra_size; |
1da177e4 | 911 | if (!frap) { |
fce1456a | 912 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
913 | return NULL; |
914 | } | |
915 | rap = frap; | |
916 | ra = *frap; | |
917 | ra->p_dev = dev; | |
918 | ra->p_ino = ino; | |
919 | ra->p_set = 0; | |
fce1456a | 920 | ra->p_hindex = hash; |
1da177e4 | 921 | found: |
fce1456a | 922 | if (rap != &rab->pb_head) { |
1da177e4 | 923 | *rap = ra->p_next; |
fce1456a GB |
924 | ra->p_next = rab->pb_head; |
925 | rab->pb_head = ra; | |
1da177e4 LT |
926 | } |
927 | ra->p_count++; | |
928 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 929 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
930 | return ra; |
931 | } | |
932 | ||
933 | /* | |
cf8208d0 JA |
934 | * Grab and keep cached pages associated with a file in the svc_rqst |
935 | * so that they can be passed to the network sendmsg/sendpage routines | |
936 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
937 | */ |
938 | static int | |
cf8208d0 JA |
939 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
940 | struct splice_desc *sd) | |
1da177e4 | 941 | { |
cf8208d0 | 942 | struct svc_rqst *rqstp = sd->u.data; |
afc59400 | 943 | struct page **pp = rqstp->rq_next_page; |
cf8208d0 JA |
944 | struct page *page = buf->page; |
945 | size_t size; | |
cf8208d0 JA |
946 | |
947 | size = sd->len; | |
1da177e4 LT |
948 | |
949 | if (rqstp->rq_res.page_len == 0) { | |
950 | get_page(page); | |
afc59400 BF |
951 | put_page(*rqstp->rq_next_page); |
952 | *(rqstp->rq_next_page++) = page; | |
cf8208d0 | 953 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 954 | rqstp->rq_res.page_len = size; |
44524359 | 955 | } else if (page != pp[-1]) { |
1da177e4 | 956 | get_page(page); |
afc59400 BF |
957 | if (*rqstp->rq_next_page) |
958 | put_page(*rqstp->rq_next_page); | |
959 | *(rqstp->rq_next_page++) = page; | |
1da177e4 | 960 | rqstp->rq_res.page_len += size; |
44524359 | 961 | } else |
1da177e4 | 962 | rqstp->rq_res.page_len += size; |
1da177e4 | 963 | |
1da177e4 LT |
964 | return size; |
965 | } | |
966 | ||
cf8208d0 JA |
967 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
968 | struct splice_desc *sd) | |
969 | { | |
970 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
971 | } | |
972 | ||
6264d69d | 973 | static __be32 |
1da177e4 LT |
974 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
975 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
976 | { | |
1da177e4 | 977 | mm_segment_t oldfs; |
6264d69d AV |
978 | __be32 err; |
979 | int host_err; | |
1da177e4 LT |
980 | |
981 | err = nfserr_perm; | |
a8754bee | 982 | |
cf8208d0 JA |
983 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
984 | struct splice_desc sd = { | |
985 | .len = 0, | |
986 | .total_len = *count, | |
987 | .pos = offset, | |
988 | .u.data = rqstp, | |
989 | }; | |
990 | ||
afc59400 | 991 | rqstp->rq_next_page = rqstp->rq_respages + 1; |
cf8208d0 | 992 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
993 | } else { |
994 | oldfs = get_fs(); | |
995 | set_fs(KERNEL_DS); | |
6264d69d | 996 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
997 | set_fs(oldfs); |
998 | } | |
999 | ||
6264d69d AV |
1000 | if (host_err >= 0) { |
1001 | nfsdstats.io_read += host_err; | |
1002 | *count = host_err; | |
1da177e4 | 1003 | err = 0; |
2a12a9d7 | 1004 | fsnotify_access(file); |
1da177e4 | 1005 | } else |
6264d69d | 1006 | err = nfserrno(host_err); |
1da177e4 LT |
1007 | return err; |
1008 | } | |
1009 | ||
9f708e40 NB |
1010 | static void kill_suid(struct dentry *dentry) |
1011 | { | |
1012 | struct iattr ia; | |
b5376771 | 1013 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 1014 | |
1b1dcc1b | 1015 | mutex_lock(&dentry->d_inode->i_mutex); |
27ac0ffe BF |
1016 | /* |
1017 | * Note we call this on write, so notify_change will not | |
1018 | * encounter any conflicting delegations: | |
1019 | */ | |
1020 | notify_change(dentry, &ia, NULL); | |
1b1dcc1b | 1021 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
1022 | } |
1023 | ||
d911df7b BF |
1024 | /* |
1025 | * Gathered writes: If another process is currently writing to the file, | |
1026 | * there's a high chance this is another nfsd (triggered by a bulk write | |
1027 | * from a client's biod). Rather than syncing the file with each write | |
1028 | * request, we sleep for 10 msec. | |
1029 | * | |
1030 | * I don't know if this roughly approximates C. Juszak's idea of | |
1031 | * gathered writes, but it's a nice and simple solution (IMHO), and it | |
1032 | * seems to work:-) | |
1033 | * | |
1034 | * Note: we do this only in the NFSv2 case, since v3 and higher have a | |
1035 | * better tool (separate unstable writes and commits) for solving this | |
1036 | * problem. | |
1037 | */ | |
1038 | static int wait_for_concurrent_writes(struct file *file) | |
1039 | { | |
496ad9aa | 1040 | struct inode *inode = file_inode(file); |
d911df7b BF |
1041 | static ino_t last_ino; |
1042 | static dev_t last_dev; | |
1043 | int err = 0; | |
1044 | ||
1045 | if (atomic_read(&inode->i_writecount) > 1 | |
1046 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
1047 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); | |
1048 | msleep(10); | |
1049 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); | |
1050 | } | |
1051 | ||
1052 | if (inode->i_state & I_DIRTY) { | |
1053 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); | |
8018ab05 | 1054 | err = vfs_fsync(file, 0); |
d911df7b BF |
1055 | } |
1056 | last_ino = inode->i_ino; | |
1057 | last_dev = inode->i_sb->s_dev; | |
1058 | return err; | |
1059 | } | |
1060 | ||
6264d69d | 1061 | static __be32 |
1da177e4 LT |
1062 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1063 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 1064 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
1065 | { |
1066 | struct svc_export *exp; | |
1067 | struct dentry *dentry; | |
1068 | struct inode *inode; | |
1069 | mm_segment_t oldfs; | |
6264d69d AV |
1070 | __be32 err = 0; |
1071 | int host_err; | |
1da177e4 | 1072 | int stable = *stablep; |
48e03bc5 | 1073 | int use_wgather; |
e49dbbf3 | 1074 | loff_t pos = offset; |
1da177e4 | 1075 | |
7eaa36e2 | 1076 | dentry = file->f_path.dentry; |
1da177e4 LT |
1077 | inode = dentry->d_inode; |
1078 | exp = fhp->fh_export; | |
1079 | ||
48e03bc5 | 1080 | use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp); |
1da177e4 | 1081 | |
1da177e4 LT |
1082 | if (!EX_ISSYNC(exp)) |
1083 | stable = 0; | |
1da177e4 LT |
1084 | |
1085 | /* Write the data. */ | |
1086 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
e49dbbf3 | 1087 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &pos); |
1da177e4 | 1088 | set_fs(oldfs); |
e4636d53 BF |
1089 | if (host_err < 0) |
1090 | goto out_nfserr; | |
1091 | *cnt = host_err; | |
1092 | nfsdstats.io_write += host_err; | |
2a12a9d7 | 1093 | fsnotify_modify(file); |
1da177e4 LT |
1094 | |
1095 | /* clear setuid/setgid flag after write */ | |
e4636d53 | 1096 | if (inode->i_mode & (S_ISUID | S_ISGID)) |
9f708e40 | 1097 | kill_suid(dentry); |
1da177e4 | 1098 | |
face1502 BF |
1099 | if (stable) { |
1100 | if (use_wgather) | |
1101 | host_err = wait_for_concurrent_writes(file); | |
1102 | else | |
1103 | host_err = vfs_fsync_range(file, offset, offset+*cnt, 0); | |
1104 | } | |
1da177e4 | 1105 | |
e4636d53 | 1106 | out_nfserr: |
6264d69d | 1107 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1108 | if (host_err >= 0) |
1da177e4 | 1109 | err = 0; |
a0d24b29 | 1110 | else |
6264d69d | 1111 | err = nfserrno(host_err); |
1da177e4 LT |
1112 | return err; |
1113 | } | |
1114 | ||
1115 | /* | |
1116 | * Read data from a file. count must contain the requested read count | |
1117 | * on entry. On return, *count contains the number of bytes actually read. | |
1118 | * N.B. After this call fhp needs an fh_put | |
1119 | */ | |
039a87ca | 1120 | __be32 nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, |
fa0a2126 BF |
1121 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) |
1122 | { | |
1123 | struct file *file; | |
1124 | struct inode *inode; | |
1125 | struct raparms *ra; | |
1126 | __be32 err; | |
1127 | ||
1128 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); | |
1129 | if (err) | |
1130 | return err; | |
1131 | ||
496ad9aa | 1132 | inode = file_inode(file); |
fa0a2126 BF |
1133 | |
1134 | /* Get readahead parameters */ | |
1135 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
1136 | ||
1137 | if (ra && ra->p_set) | |
1138 | file->f_ra = ra->p_ra; | |
1139 | ||
1140 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1141 | ||
1142 | /* Write back readahead params */ | |
1143 | if (ra) { | |
1144 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; | |
1145 | spin_lock(&rab->pb_lock); | |
1146 | ra->p_ra = file->f_ra; | |
1147 | ra->p_set = 1; | |
1148 | ra->p_count--; | |
1149 | spin_unlock(&rab->pb_lock); | |
1150 | } | |
1151 | ||
1152 | nfsd_close(file); | |
1153 | return err; | |
1154 | } | |
1155 | ||
039a87ca | 1156 | /* As above, but use the provided file descriptor. */ |
6264d69d | 1157 | __be32 |
039a87ca | 1158 | nfsd_read_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1da177e4 LT |
1159 | loff_t offset, struct kvec *vec, int vlen, |
1160 | unsigned long *count) | |
1161 | { | |
6264d69d | 1162 | __be32 err; |
1da177e4 LT |
1163 | |
1164 | if (file) { | |
0ec757df | 1165 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1166 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1167 | if (err) |
1168 | goto out; | |
1169 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
039a87ca BF |
1170 | } else /* Note file may still be NULL in NFSv4 special stateid case: */ |
1171 | err = nfsd_read(rqstp, fhp, offset, vec, vlen, count); | |
1da177e4 LT |
1172 | out: |
1173 | return err; | |
1174 | } | |
1175 | ||
1176 | /* | |
1177 | * Write data to a file. | |
1178 | * The stable flag requests synchronous writes. | |
1179 | * N.B. After this call fhp needs an fh_put | |
1180 | */ | |
6264d69d | 1181 | __be32 |
1da177e4 | 1182 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1183 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1184 | int *stablep) |
1185 | { | |
6264d69d | 1186 | __be32 err = 0; |
1da177e4 LT |
1187 | |
1188 | if (file) { | |
0ec757df | 1189 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1190 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1191 | if (err) |
1192 | goto out; | |
1193 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1194 | stablep); | |
1195 | } else { | |
8837abca | 1196 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1197 | if (err) |
1198 | goto out; | |
1199 | ||
1200 | if (cnt) | |
1201 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1202 | cnt, stablep); | |
1203 | nfsd_close(file); | |
1204 | } | |
1205 | out: | |
1206 | return err; | |
1207 | } | |
1208 | ||
1209 | #ifdef CONFIG_NFSD_V3 | |
1210 | /* | |
1211 | * Commit all pending writes to stable storage. | |
aa696a6f TM |
1212 | * |
1213 | * Note: we only guarantee that data that lies within the range specified | |
1214 | * by the 'offset' and 'count' parameters will be synced. | |
1da177e4 LT |
1215 | * |
1216 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1217 | * data to the client, as locking happens lower down in the filesystem. | |
1218 | */ | |
6264d69d | 1219 | __be32 |
1da177e4 LT |
1220 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1221 | loff_t offset, unsigned long count) | |
1222 | { | |
1223 | struct file *file; | |
aa696a6f TM |
1224 | loff_t end = LLONG_MAX; |
1225 | __be32 err = nfserr_inval; | |
1da177e4 | 1226 | |
aa696a6f TM |
1227 | if (offset < 0) |
1228 | goto out; | |
1229 | if (count != 0) { | |
1230 | end = offset + (loff_t)count - 1; | |
1231 | if (end < offset) | |
1232 | goto out; | |
1233 | } | |
1da177e4 | 1234 | |
91885258 JL |
1235 | err = nfsd_open(rqstp, fhp, S_IFREG, |
1236 | NFSD_MAY_WRITE|NFSD_MAY_NOT_BREAK_LEASE, &file); | |
8837abca | 1237 | if (err) |
aa696a6f | 1238 | goto out; |
1da177e4 | 1239 | if (EX_ISSYNC(fhp->fh_export)) { |
8018ab05 | 1240 | int err2 = vfs_fsync_range(file, offset, end, 0); |
aa696a6f TM |
1241 | |
1242 | if (err2 != -EINVAL) | |
1243 | err = nfserrno(err2); | |
1244 | else | |
1da177e4 | 1245 | err = nfserr_notsupp; |
1da177e4 LT |
1246 | } |
1247 | ||
1248 | nfsd_close(file); | |
aa696a6f | 1249 | out: |
1da177e4 LT |
1250 | return err; |
1251 | } | |
1252 | #endif /* CONFIG_NFSD_V3 */ | |
1253 | ||
f2b0dee2 | 1254 | static __be32 |
5c002b3b BF |
1255 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1256 | struct iattr *iap) | |
1257 | { | |
1258 | /* | |
1259 | * Mode has already been set earlier in create: | |
1260 | */ | |
1261 | iap->ia_valid &= ~ATTR_MODE; | |
1262 | /* | |
1263 | * Setting uid/gid works only for root. Irix appears to | |
1264 | * send along the gid on create when it tries to implement | |
1265 | * setgid directories via NFS: | |
1266 | */ | |
6fab8779 | 1267 | if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) |
5c002b3b BF |
1268 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1269 | if (iap->ia_valid) | |
1270 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1271 | return 0; | |
1272 | } | |
1273 | ||
4ac35c2f | 1274 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1275 | * setting size to 0 may fail for some specific file systems by the permission | |
1276 | * checking which requires WRITE permission but the mode is 000. | |
1277 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1278 | * 0 after file created. | |
1279 | * | |
1280 | * call this only after vfs_create() is called. | |
1281 | * */ | |
1282 | static void | |
1283 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1284 | { | |
1285 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1286 | iap->ia_valid &= ~ATTR_SIZE; | |
1287 | } | |
1288 | ||
1da177e4 LT |
1289 | /* |
1290 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1291 | * not yet implemented. | |
1292 | * If the response fh has been verified, the parent directory should | |
1293 | * already be locked. Note that the parent directory is left locked. | |
1294 | * | |
1295 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1296 | */ | |
6264d69d | 1297 | __be32 |
1da177e4 LT |
1298 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1299 | char *fname, int flen, struct iattr *iap, | |
1300 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1301 | { | |
1302 | struct dentry *dentry, *dchild = NULL; | |
1303 | struct inode *dirp; | |
6264d69d | 1304 | __be32 err; |
5c002b3b | 1305 | __be32 err2; |
6264d69d | 1306 | int host_err; |
1da177e4 LT |
1307 | |
1308 | err = nfserr_perm; | |
1309 | if (!flen) | |
1310 | goto out; | |
1311 | err = nfserr_exist; | |
1312 | if (isdotent(fname, flen)) | |
1313 | goto out; | |
1314 | ||
8837abca | 1315 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1316 | if (err) |
1317 | goto out; | |
1318 | ||
1319 | dentry = fhp->fh_dentry; | |
1320 | dirp = dentry->d_inode; | |
1321 | ||
1322 | err = nfserr_notdir; | |
acfa4380 | 1323 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1324 | goto out; |
1325 | /* | |
1326 | * Check whether the response file handle has been verified yet. | |
1327 | * If it has, the parent directory should already be locked. | |
1328 | */ | |
1329 | if (!resfhp->fh_dentry) { | |
4a55c101 JK |
1330 | host_err = fh_want_write(fhp); |
1331 | if (host_err) | |
1332 | goto out_nfserr; | |
1333 | ||
1da177e4 | 1334 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ |
12fd3520 | 1335 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1336 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1337 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1338 | if (IS_ERR(dchild)) |
1339 | goto out_nfserr; | |
1340 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1341 | if (err) | |
1342 | goto out; | |
1343 | } else { | |
1344 | /* called from nfsd_proc_create */ | |
1345 | dchild = dget(resfhp->fh_dentry); | |
1346 | if (!fhp->fh_locked) { | |
1347 | /* not actually possible */ | |
1348 | printk(KERN_ERR | |
a6a9f18f AV |
1349 | "nfsd_create: parent %pd2 not locked!\n", |
1350 | dentry); | |
d75f2b9f | 1351 | err = nfserr_io; |
1da177e4 LT |
1352 | goto out; |
1353 | } | |
1354 | } | |
1355 | /* | |
1356 | * Make sure the child dentry is still negative ... | |
1357 | */ | |
1358 | err = nfserr_exist; | |
1359 | if (dchild->d_inode) { | |
a6a9f18f AV |
1360 | dprintk("nfsd_create: dentry %pd/%pd not negative!\n", |
1361 | dentry, dchild); | |
1da177e4 LT |
1362 | goto out; |
1363 | } | |
1364 | ||
1365 | if (!(iap->ia_valid & ATTR_MODE)) | |
1366 | iap->ia_mode = 0; | |
1367 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1368 | ||
07cad1d2 MS |
1369 | err = nfserr_inval; |
1370 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1371 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1372 | type); | |
1373 | goto out; | |
1374 | } | |
1375 | ||
1da177e4 LT |
1376 | /* |
1377 | * Get the dir op function pointer. | |
1378 | */ | |
088406bc | 1379 | err = 0; |
4a55c101 | 1380 | host_err = 0; |
1da177e4 LT |
1381 | switch (type) { |
1382 | case S_IFREG: | |
312b63fb | 1383 | host_err = vfs_create(dirp, dchild, iap->ia_mode, true); |
4ac35c2f | 1384 | if (!host_err) |
1385 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1386 | break; |
1387 | case S_IFDIR: | |
6264d69d | 1388 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1389 | break; |
1390 | case S_IFCHR: | |
1391 | case S_IFBLK: | |
1392 | case S_IFIFO: | |
1393 | case S_IFSOCK: | |
6264d69d | 1394 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1395 | break; |
1da177e4 | 1396 | } |
4a55c101 | 1397 | if (host_err < 0) |
1da177e4 LT |
1398 | goto out_nfserr; |
1399 | ||
f501912a | 1400 | err = nfsd_create_setattr(rqstp, resfhp, iap); |
1da177e4 | 1401 | |
f501912a BM |
1402 | /* |
1403 | * nfsd_setattr already committed the child. Transactional filesystems | |
1404 | * had a chance to commit changes for both parent and child | |
1405 | * simultaneously making the following commit_metadata a noop. | |
1406 | */ | |
1407 | err2 = nfserrno(commit_metadata(fhp)); | |
5c002b3b BF |
1408 | if (err2) |
1409 | err = err2; | |
1da177e4 LT |
1410 | /* |
1411 | * Update the file handle to get the new inode info. | |
1412 | */ | |
1413 | if (!err) | |
1414 | err = fh_update(resfhp); | |
1415 | out: | |
1416 | if (dchild && !IS_ERR(dchild)) | |
1417 | dput(dchild); | |
1418 | return err; | |
1419 | ||
1420 | out_nfserr: | |
6264d69d | 1421 | err = nfserrno(host_err); |
1da177e4 LT |
1422 | goto out; |
1423 | } | |
1424 | ||
1425 | #ifdef CONFIG_NFSD_V3 | |
ac6721a1 MJ |
1426 | |
1427 | static inline int nfsd_create_is_exclusive(int createmode) | |
1428 | { | |
1429 | return createmode == NFS3_CREATE_EXCLUSIVE | |
1430 | || createmode == NFS4_CREATE_EXCLUSIVE4_1; | |
1431 | } | |
1432 | ||
1da177e4 | 1433 | /* |
ac6721a1 | 1434 | * NFSv3 and NFSv4 version of nfsd_create |
1da177e4 | 1435 | */ |
6264d69d | 1436 | __be32 |
ac6721a1 | 1437 | do_nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1da177e4 LT |
1438 | char *fname, int flen, struct iattr *iap, |
1439 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
856121b2 | 1440 | bool *truncp, bool *created) |
1da177e4 LT |
1441 | { |
1442 | struct dentry *dentry, *dchild = NULL; | |
1443 | struct inode *dirp; | |
6264d69d AV |
1444 | __be32 err; |
1445 | int host_err; | |
1da177e4 | 1446 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1447 | |
1448 | err = nfserr_perm; | |
1449 | if (!flen) | |
1450 | goto out; | |
1451 | err = nfserr_exist; | |
1452 | if (isdotent(fname, flen)) | |
1453 | goto out; | |
1454 | if (!(iap->ia_valid & ATTR_MODE)) | |
1455 | iap->ia_mode = 0; | |
1574dff8 | 1456 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
1457 | if (err) |
1458 | goto out; | |
1459 | ||
1460 | dentry = fhp->fh_dentry; | |
1461 | dirp = dentry->d_inode; | |
1462 | ||
1463 | /* Get all the sanity checks out of the way before | |
1464 | * we lock the parent. */ | |
1465 | err = nfserr_notdir; | |
acfa4380 | 1466 | if (!dirp->i_op->lookup) |
1da177e4 | 1467 | goto out; |
4a55c101 JK |
1468 | |
1469 | host_err = fh_want_write(fhp); | |
1470 | if (host_err) | |
1471 | goto out_nfserr; | |
1472 | ||
12fd3520 | 1473 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1474 | |
1475 | /* | |
1476 | * Compose the response file handle. | |
1477 | */ | |
1478 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1479 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1480 | if (IS_ERR(dchild)) |
1481 | goto out_nfserr; | |
1482 | ||
1574dff8 SP |
1483 | /* If file doesn't exist, check for permissions to create one */ |
1484 | if (!dchild->d_inode) { | |
1485 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); | |
1486 | if (err) | |
1487 | goto out; | |
1488 | } | |
1489 | ||
1da177e4 LT |
1490 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); |
1491 | if (err) | |
1492 | goto out; | |
1493 | ||
ac6721a1 | 1494 | if (nfsd_create_is_exclusive(createmode)) { |
c397852c | 1495 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1496 | * the high bit set, so just clear the high bits. If this is |
1497 | * ever changed to use different attrs for storing the | |
1498 | * verifier, then do_open_lookup() will also need to be fixed | |
1499 | * accordingly. | |
1da177e4 LT |
1500 | */ |
1501 | v_mtime = verifier[0]&0x7fffffff; | |
1502 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1503 | } |
1504 | ||
1505 | if (dchild->d_inode) { | |
1506 | err = 0; | |
1507 | ||
1508 | switch (createmode) { | |
1509 | case NFS3_CREATE_UNCHECKED: | |
1510 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
9dc4e6c4 | 1511 | goto out; |
1da177e4 LT |
1512 | else if (truncp) { |
1513 | /* in nfsv4, we need to treat this case a little | |
1514 | * differently. we don't want to truncate the | |
1515 | * file now; this would be wrong if the OPEN | |
1516 | * fails for some other reason. furthermore, | |
1517 | * if the size is nonzero, we should ignore it | |
1518 | * according to spec! | |
1519 | */ | |
1520 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1521 | } | |
1522 | else { | |
1523 | iap->ia_valid &= ATTR_SIZE; | |
1524 | goto set_attr; | |
1525 | } | |
1526 | break; | |
1527 | case NFS3_CREATE_EXCLUSIVE: | |
1528 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1529 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
7007c90f NB |
1530 | && dchild->d_inode->i_size == 0 ) { |
1531 | if (created) | |
1532 | *created = 1; | |
1da177e4 | 1533 | break; |
7007c90f | 1534 | } |
ac6721a1 MJ |
1535 | case NFS4_CREATE_EXCLUSIVE4_1: |
1536 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1537 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
7007c90f NB |
1538 | && dchild->d_inode->i_size == 0 ) { |
1539 | if (created) | |
1540 | *created = 1; | |
ac6721a1 | 1541 | goto set_attr; |
7007c90f | 1542 | } |
1da177e4 LT |
1543 | /* fallthru */ |
1544 | case NFS3_CREATE_GUARDED: | |
1545 | err = nfserr_exist; | |
1546 | } | |
bad0dcff | 1547 | fh_drop_write(fhp); |
1da177e4 LT |
1548 | goto out; |
1549 | } | |
1550 | ||
312b63fb | 1551 | host_err = vfs_create(dirp, dchild, iap->ia_mode, true); |
463c3197 | 1552 | if (host_err < 0) { |
bad0dcff | 1553 | fh_drop_write(fhp); |
1da177e4 | 1554 | goto out_nfserr; |
463c3197 | 1555 | } |
81ac95c5 BF |
1556 | if (created) |
1557 | *created = 1; | |
1da177e4 | 1558 | |
4ac35c2f | 1559 | nfsd_check_ignore_resizing(iap); |
1560 | ||
ac6721a1 | 1561 | if (nfsd_create_is_exclusive(createmode)) { |
c397852c | 1562 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1563 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1564 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1565 | /* XXX someone who knows this better please fix it for nsec */ |
1566 | iap->ia_mtime.tv_sec = v_mtime; | |
1567 | iap->ia_atime.tv_sec = v_atime; | |
1568 | iap->ia_mtime.tv_nsec = 0; | |
1569 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1570 | } |
1571 | ||
1da177e4 | 1572 | set_attr: |
f501912a BM |
1573 | err = nfsd_create_setattr(rqstp, resfhp, iap); |
1574 | ||
1575 | /* | |
1576 | * nfsd_setattr already committed the child (and possibly also the parent). | |
1577 | */ | |
1578 | if (!err) | |
1579 | err = nfserrno(commit_metadata(fhp)); | |
f193fbab YT |
1580 | |
1581 | /* | |
1582 | * Update the filehandle to get the new inode info. | |
1583 | */ | |
1584 | if (!err) | |
1585 | err = fh_update(resfhp); | |
1da177e4 LT |
1586 | |
1587 | out: | |
1588 | fh_unlock(fhp); | |
1589 | if (dchild && !IS_ERR(dchild)) | |
1590 | dput(dchild); | |
4a55c101 | 1591 | fh_drop_write(fhp); |
1da177e4 LT |
1592 | return err; |
1593 | ||
1594 | out_nfserr: | |
6264d69d | 1595 | err = nfserrno(host_err); |
1da177e4 LT |
1596 | goto out; |
1597 | } | |
1598 | #endif /* CONFIG_NFSD_V3 */ | |
1599 | ||
1600 | /* | |
1601 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1602 | * fits into the buffer. On return, it contains the true length. | |
1603 | * N.B. After this call fhp needs an fh_put | |
1604 | */ | |
6264d69d | 1605 | __be32 |
1da177e4 LT |
1606 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1607 | { | |
1da177e4 LT |
1608 | struct inode *inode; |
1609 | mm_segment_t oldfs; | |
6264d69d AV |
1610 | __be32 err; |
1611 | int host_err; | |
68ac1234 | 1612 | struct path path; |
1da177e4 | 1613 | |
8837abca | 1614 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1615 | if (err) |
1616 | goto out; | |
1617 | ||
68ac1234 AV |
1618 | path.mnt = fhp->fh_export->ex_path.mnt; |
1619 | path.dentry = fhp->fh_dentry; | |
1620 | inode = path.dentry->d_inode; | |
1da177e4 LT |
1621 | |
1622 | err = nfserr_inval; | |
acfa4380 | 1623 | if (!inode->i_op->readlink) |
1da177e4 LT |
1624 | goto out; |
1625 | ||
68ac1234 | 1626 | touch_atime(&path); |
1da177e4 LT |
1627 | /* N.B. Why does this call need a get_fs()?? |
1628 | * Remove the set_fs and watch the fireworks:-) --okir | |
1629 | */ | |
1630 | ||
1631 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
fac7a17b | 1632 | host_err = inode->i_op->readlink(path.dentry, (char __user *)buf, *lenp); |
1da177e4 LT |
1633 | set_fs(oldfs); |
1634 | ||
6264d69d | 1635 | if (host_err < 0) |
1da177e4 | 1636 | goto out_nfserr; |
6264d69d | 1637 | *lenp = host_err; |
1da177e4 LT |
1638 | err = 0; |
1639 | out: | |
1640 | return err; | |
1641 | ||
1642 | out_nfserr: | |
6264d69d | 1643 | err = nfserrno(host_err); |
1da177e4 LT |
1644 | goto out; |
1645 | } | |
1646 | ||
1647 | /* | |
1648 | * Create a symlink and look up its inode | |
1649 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1650 | */ | |
6264d69d | 1651 | __be32 |
1da177e4 LT |
1652 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1653 | char *fname, int flen, | |
1654 | char *path, int plen, | |
1655 | struct svc_fh *resfhp, | |
1656 | struct iattr *iap) | |
1657 | { | |
1658 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1659 | __be32 err, cerr; |
1660 | int host_err; | |
1da177e4 LT |
1661 | |
1662 | err = nfserr_noent; | |
1663 | if (!flen || !plen) | |
1664 | goto out; | |
1665 | err = nfserr_exist; | |
1666 | if (isdotent(fname, flen)) | |
1667 | goto out; | |
1668 | ||
8837abca | 1669 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1670 | if (err) |
1671 | goto out; | |
4a55c101 JK |
1672 | |
1673 | host_err = fh_want_write(fhp); | |
1674 | if (host_err) | |
1675 | goto out_nfserr; | |
1676 | ||
1da177e4 LT |
1677 | fh_lock(fhp); |
1678 | dentry = fhp->fh_dentry; | |
1679 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1680 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1681 | if (IS_ERR(dnew)) |
1682 | goto out_nfserr; | |
1683 | ||
1da177e4 LT |
1684 | if (unlikely(path[plen] != 0)) { |
1685 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1686 | if (path_alloced == NULL) | |
6264d69d | 1687 | host_err = -ENOMEM; |
1da177e4 LT |
1688 | else { |
1689 | strncpy(path_alloced, path, plen); | |
1690 | path_alloced[plen] = 0; | |
db2e747b | 1691 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1692 | kfree(path_alloced); |
1693 | } | |
1694 | } else | |
db2e747b | 1695 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
6264d69d | 1696 | err = nfserrno(host_err); |
f501912a BM |
1697 | if (!err) |
1698 | err = nfserrno(commit_metadata(fhp)); | |
1da177e4 LT |
1699 | fh_unlock(fhp); |
1700 | ||
bad0dcff | 1701 | fh_drop_write(fhp); |
75c3f29d | 1702 | |
1da177e4 LT |
1703 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1704 | dput(dnew); | |
1705 | if (err==0) err = cerr; | |
1706 | out: | |
1707 | return err; | |
1708 | ||
1709 | out_nfserr: | |
6264d69d | 1710 | err = nfserrno(host_err); |
1da177e4 LT |
1711 | goto out; |
1712 | } | |
1713 | ||
1714 | /* | |
1715 | * Create a hardlink | |
1716 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1717 | */ | |
6264d69d | 1718 | __be32 |
1da177e4 LT |
1719 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1720 | char *name, int len, struct svc_fh *tfhp) | |
1721 | { | |
1722 | struct dentry *ddir, *dnew, *dold; | |
55b13354 | 1723 | struct inode *dirp; |
6264d69d AV |
1724 | __be32 err; |
1725 | int host_err; | |
1da177e4 | 1726 | |
8837abca | 1727 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1728 | if (err) |
1729 | goto out; | |
7d818a7b | 1730 | err = fh_verify(rqstp, tfhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
1731 | if (err) |
1732 | goto out; | |
7d818a7b BF |
1733 | err = nfserr_isdir; |
1734 | if (S_ISDIR(tfhp->fh_dentry->d_inode->i_mode)) | |
1735 | goto out; | |
1da177e4 LT |
1736 | err = nfserr_perm; |
1737 | if (!len) | |
1738 | goto out; | |
1739 | err = nfserr_exist; | |
1740 | if (isdotent(name, len)) | |
1741 | goto out; | |
1742 | ||
4a55c101 JK |
1743 | host_err = fh_want_write(tfhp); |
1744 | if (host_err) { | |
1745 | err = nfserrno(host_err); | |
1746 | goto out; | |
1747 | } | |
1748 | ||
12fd3520 | 1749 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1750 | ddir = ffhp->fh_dentry; |
1751 | dirp = ddir->d_inode; | |
1752 | ||
1753 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1754 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1755 | if (IS_ERR(dnew)) |
1756 | goto out_nfserr; | |
1757 | ||
1758 | dold = tfhp->fh_dentry; | |
1da177e4 | 1759 | |
4795bb37 BF |
1760 | err = nfserr_noent; |
1761 | if (!dold->d_inode) | |
4a55c101 | 1762 | goto out_dput; |
4795bb37 | 1763 | host_err = nfsd_break_lease(dold->d_inode); |
7d751f6f CB |
1764 | if (host_err) { |
1765 | err = nfserrno(host_err); | |
4a55c101 | 1766 | goto out_dput; |
7d751f6f | 1767 | } |
146a8595 | 1768 | host_err = vfs_link(dold, dirp, dnew, NULL); |
6264d69d | 1769 | if (!host_err) { |
f501912a BM |
1770 | err = nfserrno(commit_metadata(ffhp)); |
1771 | if (!err) | |
1772 | err = nfserrno(commit_metadata(tfhp)); | |
1da177e4 | 1773 | } else { |
6264d69d | 1774 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1775 | err = nfserr_acces; |
1776 | else | |
6264d69d | 1777 | err = nfserrno(host_err); |
1da177e4 | 1778 | } |
75c3f29d | 1779 | out_dput: |
1da177e4 | 1780 | dput(dnew); |
270d56e5 DR |
1781 | out_unlock: |
1782 | fh_unlock(ffhp); | |
4a55c101 | 1783 | fh_drop_write(tfhp); |
1da177e4 LT |
1784 | out: |
1785 | return err; | |
1786 | ||
1787 | out_nfserr: | |
6264d69d | 1788 | err = nfserrno(host_err); |
270d56e5 | 1789 | goto out_unlock; |
1da177e4 LT |
1790 | } |
1791 | ||
1792 | /* | |
1793 | * Rename a file | |
1794 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1795 | */ | |
6264d69d | 1796 | __be32 |
1da177e4 LT |
1797 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1798 | struct svc_fh *tfhp, char *tname, int tlen) | |
1799 | { | |
1800 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1801 | struct inode *fdir, *tdir; | |
6264d69d AV |
1802 | __be32 err; |
1803 | int host_err; | |
1da177e4 | 1804 | |
8837abca | 1805 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1806 | if (err) |
1807 | goto out; | |
8837abca | 1808 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1809 | if (err) |
1810 | goto out; | |
1811 | ||
1812 | fdentry = ffhp->fh_dentry; | |
1813 | fdir = fdentry->d_inode; | |
1814 | ||
1815 | tdentry = tfhp->fh_dentry; | |
1816 | tdir = tdentry->d_inode; | |
1817 | ||
1da177e4 LT |
1818 | err = nfserr_perm; |
1819 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1820 | goto out; | |
1821 | ||
4a55c101 JK |
1822 | host_err = fh_want_write(ffhp); |
1823 | if (host_err) { | |
1824 | err = nfserrno(host_err); | |
1825 | goto out; | |
1826 | } | |
1827 | ||
1da177e4 LT |
1828 | /* cannot use fh_lock as we need deadlock protective ordering |
1829 | * so do it by hand */ | |
1830 | trap = lock_rename(tdentry, fdentry); | |
1831 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1832 | fill_pre_wcc(ffhp); | |
1833 | fill_pre_wcc(tfhp); | |
1834 | ||
1835 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1836 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1837 | if (IS_ERR(odentry)) |
1838 | goto out_nfserr; | |
1839 | ||
6264d69d | 1840 | host_err = -ENOENT; |
1da177e4 LT |
1841 | if (!odentry->d_inode) |
1842 | goto out_dput_old; | |
6264d69d | 1843 | host_err = -EINVAL; |
1da177e4 LT |
1844 | if (odentry == trap) |
1845 | goto out_dput_old; | |
1846 | ||
1847 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1848 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1849 | if (IS_ERR(ndentry)) |
1850 | goto out_dput_old; | |
6264d69d | 1851 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1852 | if (ndentry == trap) |
1853 | goto out_dput_new; | |
1854 | ||
9079b1eb DH |
1855 | host_err = -EXDEV; |
1856 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1857 | goto out_dput_new; | |
aa387d6c BF |
1858 | if (ffhp->fh_export->ex_path.dentry != tfhp->fh_export->ex_path.dentry) |
1859 | goto out_dput_new; | |
9079b1eb | 1860 | |
4795bb37 | 1861 | host_err = nfsd_break_lease(odentry->d_inode); |
83f6b0c1 | 1862 | if (host_err) |
4a55c101 | 1863 | goto out_dput_new; |
83f6b0c1 BF |
1864 | if (ndentry->d_inode) { |
1865 | host_err = nfsd_break_lease(ndentry->d_inode); | |
1866 | if (host_err) | |
4a55c101 | 1867 | goto out_dput_new; |
83f6b0c1 | 1868 | } |
8e6d782c | 1869 | host_err = vfs_rename(fdir, odentry, tdir, ndentry, NULL); |
f501912a BM |
1870 | if (!host_err) { |
1871 | host_err = commit_metadata(tfhp); | |
6264d69d | 1872 | if (!host_err) |
f501912a | 1873 | host_err = commit_metadata(ffhp); |
1da177e4 | 1874 | } |
1da177e4 LT |
1875 | out_dput_new: |
1876 | dput(ndentry); | |
1877 | out_dput_old: | |
1878 | dput(odentry); | |
1879 | out_nfserr: | |
6264d69d | 1880 | err = nfserrno(host_err); |
1da177e4 LT |
1881 | |
1882 | /* we cannot reply on fh_unlock on the two filehandles, | |
1883 | * as that would do the wrong thing if the two directories | |
1884 | * were the same, so again we do it by hand | |
1885 | */ | |
1886 | fill_post_wcc(ffhp); | |
1887 | fill_post_wcc(tfhp); | |
1888 | unlock_rename(tdentry, fdentry); | |
1889 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
4a55c101 | 1890 | fh_drop_write(ffhp); |
1da177e4 LT |
1891 | |
1892 | out: | |
1893 | return err; | |
1894 | } | |
1895 | ||
1896 | /* | |
1897 | * Unlink a file or directory | |
1898 | * N.B. After this call fhp needs an fh_put | |
1899 | */ | |
6264d69d | 1900 | __be32 |
1da177e4 LT |
1901 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1902 | char *fname, int flen) | |
1903 | { | |
1904 | struct dentry *dentry, *rdentry; | |
1905 | struct inode *dirp; | |
6264d69d AV |
1906 | __be32 err; |
1907 | int host_err; | |
1da177e4 LT |
1908 | |
1909 | err = nfserr_acces; | |
1910 | if (!flen || isdotent(fname, flen)) | |
1911 | goto out; | |
8837abca | 1912 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1913 | if (err) |
1914 | goto out; | |
1915 | ||
4a55c101 JK |
1916 | host_err = fh_want_write(fhp); |
1917 | if (host_err) | |
1918 | goto out_nfserr; | |
1919 | ||
12fd3520 | 1920 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1921 | dentry = fhp->fh_dentry; |
1922 | dirp = dentry->d_inode; | |
1923 | ||
1924 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1925 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1926 | if (IS_ERR(rdentry)) |
1927 | goto out_nfserr; | |
1928 | ||
1929 | if (!rdentry->d_inode) { | |
1930 | dput(rdentry); | |
1931 | err = nfserr_noent; | |
1932 | goto out; | |
1933 | } | |
1934 | ||
1935 | if (!type) | |
1936 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1937 | ||
4795bb37 BF |
1938 | host_err = nfsd_break_lease(rdentry->d_inode); |
1939 | if (host_err) | |
4a55c101 | 1940 | goto out_put; |
9ce137ee | 1941 | if (type != S_IFDIR) |
b21996e3 | 1942 | host_err = vfs_unlink(dirp, rdentry, NULL); |
9ce137ee | 1943 | else |
6264d69d | 1944 | host_err = vfs_rmdir(dirp, rdentry); |
f501912a BM |
1945 | if (!host_err) |
1946 | host_err = commit_metadata(fhp); | |
541ce98c BF |
1947 | out_put: |
1948 | dput(rdentry); | |
1949 | ||
1da177e4 | 1950 | out_nfserr: |
6264d69d | 1951 | err = nfserrno(host_err); |
f193fbab YT |
1952 | out: |
1953 | return err; | |
1da177e4 LT |
1954 | } |
1955 | ||
14f7dd63 DW |
1956 | /* |
1957 | * We do this buffering because we must not call back into the file | |
1958 | * system's ->lookup() method from the filldir callback. That may well | |
1959 | * deadlock a number of file systems. | |
1960 | * | |
1961 | * This is based heavily on the implementation of same in XFS. | |
1962 | */ | |
1963 | struct buffered_dirent { | |
1964 | u64 ino; | |
1965 | loff_t offset; | |
1966 | int namlen; | |
1967 | unsigned int d_type; | |
1968 | char name[]; | |
1969 | }; | |
1970 | ||
1971 | struct readdir_data { | |
5c0ba4e0 | 1972 | struct dir_context ctx; |
14f7dd63 DW |
1973 | char *dirent; |
1974 | size_t used; | |
53c9c5c0 | 1975 | int full; |
14f7dd63 DW |
1976 | }; |
1977 | ||
1978 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1979 | loff_t offset, u64 ino, unsigned int d_type) | |
1980 | { | |
1981 | struct readdir_data *buf = __buf; | |
1982 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1983 | unsigned int reclen; | |
1984 | ||
1985 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1986 | if (buf->used + reclen > PAGE_SIZE) { |
1987 | buf->full = 1; | |
14f7dd63 | 1988 | return -EINVAL; |
53c9c5c0 | 1989 | } |
14f7dd63 DW |
1990 | |
1991 | de->namlen = namlen; | |
1992 | de->offset = offset; | |
1993 | de->ino = ino; | |
1994 | de->d_type = d_type; | |
1995 | memcpy(de->name, name, namlen); | |
1996 | buf->used += reclen; | |
1997 | ||
1998 | return 0; | |
1999 | } | |
2000 | ||
2f9092e1 DW |
2001 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
2002 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 2003 | { |
14f7dd63 | 2004 | struct buffered_dirent *de; |
2628b766 | 2005 | int host_err; |
14f7dd63 DW |
2006 | int size; |
2007 | loff_t offset; | |
ac6614b7 AV |
2008 | struct readdir_data buf = { |
2009 | .ctx.actor = nfsd_buffered_filldir, | |
2010 | .dirent = (void *)__get_free_page(GFP_KERNEL) | |
2011 | }; | |
2628b766 | 2012 | |
14f7dd63 | 2013 | if (!buf.dirent) |
2f9092e1 | 2014 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
2015 | |
2016 | offset = *offsetp; | |
2628b766 | 2017 | |
14f7dd63 | 2018 | while (1) { |
496ad9aa | 2019 | struct inode *dir_inode = file_inode(file); |
14f7dd63 DW |
2020 | unsigned int reclen; |
2021 | ||
b726e923 | 2022 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 2023 | buf.used = 0; |
53c9c5c0 | 2024 | buf.full = 0; |
14f7dd63 | 2025 | |
5c0ba4e0 | 2026 | host_err = iterate_dir(file, &buf.ctx); |
53c9c5c0 AV |
2027 | if (buf.full) |
2028 | host_err = 0; | |
2029 | ||
2030 | if (host_err < 0) | |
14f7dd63 DW |
2031 | break; |
2032 | ||
2033 | size = buf.used; | |
2034 | ||
2035 | if (!size) | |
2036 | break; | |
2037 | ||
2f9092e1 DW |
2038 | /* |
2039 | * Various filldir functions may end up calling back into | |
2040 | * lookup_one_len() and the file system's ->lookup() method. | |
2041 | * These expect i_mutex to be held, as it would within readdir. | |
2042 | */ | |
2043 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
2044 | if (host_err) | |
2045 | break; | |
2046 | ||
14f7dd63 DW |
2047 | de = (struct buffered_dirent *)buf.dirent; |
2048 | while (size > 0) { | |
2049 | offset = de->offset; | |
2050 | ||
2051 | if (func(cdp, de->name, de->namlen, de->offset, | |
2052 | de->ino, de->d_type)) | |
2f9092e1 | 2053 | break; |
14f7dd63 DW |
2054 | |
2055 | if (cdp->err != nfs_ok) | |
2f9092e1 | 2056 | break; |
14f7dd63 DW |
2057 | |
2058 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
2059 | sizeof(u64)); | |
2060 | size -= reclen; | |
2061 | de = (struct buffered_dirent *)((char *)de + reclen); | |
2062 | } | |
2f9092e1 DW |
2063 | mutex_unlock(&dir_inode->i_mutex); |
2064 | if (size > 0) /* We bailed out early */ | |
2065 | break; | |
2066 | ||
c002a6c7 | 2067 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
2068 | } |
2069 | ||
14f7dd63 | 2070 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
2071 | |
2072 | if (host_err) | |
2073 | return nfserrno(host_err); | |
14f7dd63 DW |
2074 | |
2075 | *offsetp = offset; | |
2076 | return cdp->err; | |
2628b766 DW |
2077 | } |
2078 | ||
1da177e4 LT |
2079 | /* |
2080 | * Read entries from a directory. | |
2081 | * The NFSv3/4 verifier we ignore for now. | |
2082 | */ | |
6264d69d | 2083 | __be32 |
1da177e4 | 2084 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 2085 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 2086 | { |
6264d69d | 2087 | __be32 err; |
1da177e4 LT |
2088 | struct file *file; |
2089 | loff_t offset = *offsetp; | |
06effdbb BS |
2090 | int may_flags = NFSD_MAY_READ; |
2091 | ||
2092 | /* NFSv2 only supports 32 bit cookies */ | |
2093 | if (rqstp->rq_vers > 2) | |
2094 | may_flags |= NFSD_MAY_64BIT_COOKIE; | |
1da177e4 | 2095 | |
06effdbb | 2096 | err = nfsd_open(rqstp, fhp, S_IFDIR, may_flags, &file); |
1da177e4 LT |
2097 | if (err) |
2098 | goto out; | |
2099 | ||
b108fe6b | 2100 | offset = vfs_llseek(file, offset, SEEK_SET); |
1da177e4 LT |
2101 | if (offset < 0) { |
2102 | err = nfserrno((int)offset); | |
2103 | goto out_close; | |
2104 | } | |
2105 | ||
14f7dd63 | 2106 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
2107 | |
2108 | if (err == nfserr_eof || err == nfserr_toosmall) | |
2109 | err = nfs_ok; /* can still be found in ->err */ | |
2110 | out_close: | |
2111 | nfsd_close(file); | |
2112 | out: | |
2113 | return err; | |
2114 | } | |
2115 | ||
2116 | /* | |
2117 | * Get file system stats | |
2118 | * N.B. After this call fhp needs an fh_put | |
2119 | */ | |
6264d69d | 2120 | __be32 |
04716e66 | 2121 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 2122 | { |
ebabe9a9 CH |
2123 | __be32 err; |
2124 | ||
2125 | err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); | |
f6360efb TI |
2126 | if (!err) { |
2127 | struct path path = { | |
2128 | .mnt = fhp->fh_export->ex_path.mnt, | |
2129 | .dentry = fhp->fh_dentry, | |
2130 | }; | |
2131 | if (vfs_statfs(&path, stat)) | |
2132 | err = nfserr_io; | |
2133 | } | |
1da177e4 LT |
2134 | return err; |
2135 | } | |
2136 | ||
c7d51402 | 2137 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 2138 | { |
c7d51402 | 2139 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
2140 | } |
2141 | ||
1da177e4 LT |
2142 | /* |
2143 | * Check for a user's access permissions to this inode. | |
2144 | */ | |
6264d69d | 2145 | __be32 |
0ec757df BF |
2146 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2147 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2148 | { |
2149 | struct inode *inode = dentry->d_inode; | |
2150 | int err; | |
2151 | ||
aea93397 | 2152 | if ((acc & NFSD_MAY_MASK) == NFSD_MAY_NOP) |
1da177e4 LT |
2153 | return 0; |
2154 | #if 0 | |
2155 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2156 | acc, | |
8837abca MS |
2157 | (acc & NFSD_MAY_READ)? " read" : "", |
2158 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2159 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2160 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2161 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2162 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2163 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2164 | inode->i_mode, |
2165 | IS_IMMUTABLE(inode)? " immut" : "", | |
2166 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2167 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2168 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2169 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2170 | #endif |
2171 | ||
2172 | /* Normally we reject any write/sattr etc access on a read-only file | |
2173 | * system. But if it is IRIX doing check on write-access for a | |
2174 | * device special file, we ignore rofs. | |
2175 | */ | |
8837abca MS |
2176 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2177 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2178 | if (exp_rdonly(rqstp, exp) || |
2179 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2180 | return nfserr_rofs; |
8837abca | 2181 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2182 | return nfserr_perm; |
2183 | } | |
8837abca | 2184 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2185 | return nfserr_perm; |
2186 | ||
8837abca | 2187 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2188 | /* If we cannot rely on authentication in NLM requests, |
2189 | * just allow locks, otherwise require read permission, or | |
2190 | * ownership | |
2191 | */ | |
2192 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2193 | return 0; | |
2194 | else | |
8837abca | 2195 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2196 | } |
2197 | /* | |
2198 | * The file owner always gets access permission for accesses that | |
2199 | * would normally be checked at open time. This is to make | |
2200 | * file access work even when the client has done a fchmod(fd, 0). | |
2201 | * | |
2202 | * However, `cp foo bar' should fail nevertheless when bar is | |
2203 | * readonly. A sensible way to do this might be to reject all | |
2204 | * attempts to truncate a read-only file, because a creat() call | |
2205 | * always implies file truncation. | |
2206 | * ... but this isn't really fair. A process may reasonably call | |
2207 | * ftruncate on an open file descriptor on a file with perm 000. | |
2208 | * We must trust the client to do permission checking - using "ACCESS" | |
2209 | * with NFSv3. | |
2210 | */ | |
8837abca | 2211 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
6fab8779 | 2212 | uid_eq(inode->i_uid, current_fsuid())) |
1da177e4 LT |
2213 | return 0; |
2214 | ||
8837abca | 2215 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2216 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2217 | |
2218 | /* Allow read access to binaries even when mode 111 */ | |
2219 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
a043226b BF |
2220 | (acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE) || |
2221 | acc == (NFSD_MAY_READ | NFSD_MAY_READ_IF_EXEC))) | |
f419a2e3 | 2222 | err = inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
2223 | |
2224 | return err? nfserrno(err) : 0; | |
2225 | } | |
2226 | ||
2227 | void | |
2228 | nfsd_racache_shutdown(void) | |
2229 | { | |
54a66e54 JL |
2230 | struct raparms *raparm, *last_raparm; |
2231 | unsigned int i; | |
2232 | ||
1da177e4 | 2233 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2234 | |
2235 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2236 | raparm = raparm_hash[i].pb_head; | |
2237 | while(raparm) { | |
2238 | last_raparm = raparm; | |
2239 | raparm = raparm->p_next; | |
2240 | kfree(last_raparm); | |
2241 | } | |
2242 | raparm_hash[i].pb_head = NULL; | |
2243 | } | |
1da177e4 LT |
2244 | } |
2245 | /* | |
2246 | * Initialize readahead param cache | |
2247 | */ | |
2248 | int | |
2249 | nfsd_racache_init(int cache_size) | |
2250 | { | |
2251 | int i; | |
fce1456a GB |
2252 | int j = 0; |
2253 | int nperbucket; | |
54a66e54 | 2254 | struct raparms **raparm = NULL; |
1da177e4 | 2255 | |
fce1456a | 2256 | |
54a66e54 | 2257 | if (raparm_hash[0].pb_head) |
1da177e4 | 2258 | return 0; |
54a66e54 JL |
2259 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2260 | if (nperbucket < 2) | |
2261 | nperbucket = 2; | |
2262 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2263 | |
2264 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2265 | |
2266 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2267 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2268 | |
2269 | raparm = &raparm_hash[i].pb_head; | |
2270 | for (j = 0; j < nperbucket; j++) { | |
2271 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2272 | if (!*raparm) | |
2273 | goto out_nomem; | |
2274 | raparm = &(*raparm)->p_next; | |
2275 | } | |
2276 | *raparm = NULL; | |
4b3bb06b YB |
2277 | } |
2278 | ||
1da177e4 LT |
2279 | nfsdstats.ra_size = cache_size; |
2280 | return 0; | |
54a66e54 JL |
2281 | |
2282 | out_nomem: | |
2283 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2284 | nfsd_racache_shutdown(); | |
2285 | return -ENOMEM; | |
1da177e4 | 2286 | } |
a257cdd0 AG |
2287 | |
2288 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2289 | struct posix_acl * | |
2290 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2291 | { | |
2292 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2293 | char *name; | |
2294 | void *value = NULL; | |
2295 | ssize_t size; | |
2296 | struct posix_acl *acl; | |
2297 | ||
5be196e5 CH |
2298 | if (!IS_POSIXACL(inode)) |
2299 | return ERR_PTR(-EOPNOTSUPP); | |
2300 | ||
2301 | switch (type) { | |
2302 | case ACL_TYPE_ACCESS: | |
2303 | name = POSIX_ACL_XATTR_ACCESS; | |
2304 | break; | |
2305 | case ACL_TYPE_DEFAULT: | |
2306 | name = POSIX_ACL_XATTR_DEFAULT; | |
2307 | break; | |
2308 | default: | |
a257cdd0 | 2309 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2310 | } |
2311 | ||
5be196e5 CH |
2312 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2313 | if (size < 0) | |
2314 | return ERR_PTR(size); | |
a257cdd0 | 2315 | |
5f3a4a28 | 2316 | acl = posix_acl_from_xattr(&init_user_ns, value, size); |
a257cdd0 AG |
2317 | kfree(value); |
2318 | return acl; | |
2319 | } | |
2320 | ||
2321 | int | |
2322 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2323 | { | |
2324 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2325 | char *name; | |
2326 | void *value = NULL; | |
2327 | size_t size; | |
2328 | int error; | |
2329 | ||
acfa4380 | 2330 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2331 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2332 | return -EOPNOTSUPP; | |
2333 | switch(type) { | |
2334 | case ACL_TYPE_ACCESS: | |
334a13ec | 2335 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2336 | break; |
2337 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2338 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2339 | break; |
2340 | default: | |
2341 | return -EOPNOTSUPP; | |
2342 | } | |
2343 | ||
2344 | if (acl && acl->a_count) { | |
334a13ec | 2345 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2346 | value = kmalloc(size, GFP_KERNEL); |
2347 | if (!value) | |
2348 | return -ENOMEM; | |
5f3a4a28 | 2349 | error = posix_acl_to_xattr(&init_user_ns, acl, value, size); |
9ccfc29c | 2350 | if (error < 0) |
a257cdd0 | 2351 | goto getout; |
9ccfc29c | 2352 | size = error; |
a257cdd0 AG |
2353 | } else |
2354 | size = 0; | |
2355 | ||
bad0dcff | 2356 | error = fh_want_write(fhp); |
18f335af DH |
2357 | if (error) |
2358 | goto getout; | |
a257cdd0 | 2359 | if (size) |
5be196e5 | 2360 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2361 | else { |
2362 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2363 | error = 0; | |
2364 | else { | |
5be196e5 | 2365 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2366 | if (error == -ENODATA) |
2367 | error = 0; | |
2368 | } | |
2369 | } | |
bad0dcff | 2370 | fh_drop_write(fhp); |
a257cdd0 AG |
2371 | |
2372 | getout: | |
2373 | kfree(value); | |
2374 | return error; | |
2375 | } | |
2376 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |