]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #define MSNFS /* HACK HACK */ |
2 | /* | |
3 | * linux/fs/nfsd/vfs.c | |
4 | * | |
5 | * File operations used by nfsd. Some of these have been ripped from | |
6 | * other parts of the kernel because they weren't exported, others | |
7 | * are partial duplicates with added or changed functionality. | |
8 | * | |
9 | * Note that several functions dget() the dentry upon which they want | |
10 | * to act, most notably those that create directory entries. Response | |
11 | * dentry's are dput()'d if necessary in the release callback. | |
12 | * So if you notice code paths that apparently fail to dput() the | |
13 | * dentry, don't worry--they have been taken care of. | |
14 | * | |
15 | * Copyright (C) 1995-1999 Olaf Kirch <okir@monad.swb.de> | |
16 | * Zerocpy NFS support (C) 2002 Hirokazu Takahashi <taka@valinux.co.jp> | |
17 | */ | |
18 | ||
1da177e4 LT |
19 | #include <linux/fs.h> |
20 | #include <linux/file.h> | |
d6b29d7c | 21 | #include <linux/splice.h> |
1da177e4 | 22 | #include <linux/fcntl.h> |
1da177e4 | 23 | #include <linux/namei.h> |
1da177e4 | 24 | #include <linux/delay.h> |
1da177e4 | 25 | #include <linux/quotaops.h> |
0eeca283 | 26 | #include <linux/fsnotify.h> |
1da177e4 | 27 | #include <linux/posix_acl_xattr.h> |
1da177e4 | 28 | #include <linux/xattr.h> |
9a74af21 BH |
29 | #include <linux/jhash.h> |
30 | #include <linux/ima.h> | |
31 | #include <asm/uaccess.h> | |
32 | ||
33 | #ifdef CONFIG_NFSD_V3 | |
34 | #include "xdr3.h" | |
35 | #endif /* CONFIG_NFSD_V3 */ | |
36 | ||
5be196e5 | 37 | #ifdef CONFIG_NFSD_V4 |
1da177e4 LT |
38 | #include <linux/nfs4_acl.h> |
39 | #include <linux/nfsd_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 | 74 | |
3c394dda SD |
75 | static inline int |
76 | nfsd_v4client(struct svc_rqst *rq) | |
77 | { | |
78 | return rq->rq_prog == NFS_PROGRAM && rq->rq_vers == 4; | |
79 | } | |
80 | ||
1da177e4 LT |
81 | /* |
82 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
83 | * a mount point. | |
e0bb89ef | 84 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
85 | * or nfs_ok having possibly changed *dpp and *expp |
86 | */ | |
87 | int | |
88 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
89 | struct svc_export **expp) | |
90 | { | |
91 | struct svc_export *exp = *expp, *exp2 = NULL; | |
92 | struct dentry *dentry = *dpp; | |
91c9fa8f AV |
93 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
94 | .dentry = dget(dentry)}; | |
6264d69d | 95 | int err = 0; |
1da177e4 | 96 | |
9393bd07 | 97 | while (d_mountpoint(path.dentry) && follow_down(&path)) |
91c9fa8f | 98 | ; |
1da177e4 | 99 | |
91c9fa8f | 100 | exp2 = rqst_exp_get_by_name(rqstp, &path); |
1da177e4 | 101 | if (IS_ERR(exp2)) { |
3b6cee7b BF |
102 | err = PTR_ERR(exp2); |
103 | /* | |
104 | * We normally allow NFS clients to continue | |
105 | * "underneath" a mountpoint that is not exported. | |
106 | * The exception is V4ROOT, where no traversal is ever | |
107 | * allowed without an explicit export of the new | |
108 | * directory. | |
109 | */ | |
110 | if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT)) | |
111 | err = 0; | |
91c9fa8f | 112 | path_put(&path); |
1da177e4 LT |
113 | goto out; |
114 | } | |
3c394dda SD |
115 | if (nfsd_v4client(rqstp) || |
116 | (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { | |
1da177e4 | 117 | /* successfully crossed mount point */ |
1644ccc8 | 118 | /* |
91c9fa8f AV |
119 | * This is subtle: path.dentry is *not* on path.mnt |
120 | * at this point. The only reason we are safe is that | |
121 | * original mnt is pinned down by exp, so we should | |
122 | * put path *before* putting exp | |
1644ccc8 | 123 | */ |
91c9fa8f AV |
124 | *dpp = path.dentry; |
125 | path.dentry = dentry; | |
1644ccc8 | 126 | *expp = exp2; |
91c9fa8f | 127 | exp2 = exp; |
1da177e4 | 128 | } |
91c9fa8f AV |
129 | path_put(&path); |
130 | exp_put(exp2); | |
1da177e4 LT |
131 | out: |
132 | return err; | |
133 | } | |
134 | ||
289ede45 BF |
135 | static void follow_to_parent(struct path *path) |
136 | { | |
137 | struct dentry *dp; | |
138 | ||
139 | while (path->dentry == path->mnt->mnt_root && follow_up(path)) | |
140 | ; | |
141 | dp = dget_parent(path->dentry); | |
142 | dput(path->dentry); | |
143 | path->dentry = dp; | |
144 | } | |
145 | ||
146 | static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp) | |
147 | { | |
148 | struct svc_export *exp2; | |
149 | struct path path = {.mnt = mntget((*exp)->ex_path.mnt), | |
150 | .dentry = dget(dparent)}; | |
151 | ||
152 | follow_to_parent(&path); | |
153 | ||
154 | exp2 = rqst_exp_parent(rqstp, &path); | |
155 | if (PTR_ERR(exp2) == -ENOENT) { | |
156 | *dentryp = dget(dparent); | |
157 | } else if (IS_ERR(exp2)) { | |
158 | path_put(&path); | |
159 | return PTR_ERR(exp2); | |
160 | } else { | |
161 | *dentryp = dget(path.dentry); | |
162 | exp_put(*exp); | |
163 | *exp = exp2; | |
164 | } | |
165 | path_put(&path); | |
166 | return 0; | |
167 | } | |
168 | ||
82ead7fe BF |
169 | /* |
170 | * For nfsd purposes, we treat V4ROOT exports as though there was an | |
171 | * export at *every* directory. | |
172 | */ | |
3227fa41 | 173 | int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp) |
82ead7fe BF |
174 | { |
175 | if (d_mountpoint(dentry)) | |
176 | return 1; | |
177 | if (!(exp->ex_flags & NFSEXP_V4ROOT)) | |
178 | return 0; | |
179 | return dentry->d_inode != NULL; | |
180 | } | |
181 | ||
6264d69d | 182 | __be32 |
6c0a654d | 183 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 184 | const char *name, unsigned int len, |
6c0a654d | 185 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
186 | { |
187 | struct svc_export *exp; | |
188 | struct dentry *dparent; | |
189 | struct dentry *dentry; | |
6264d69d AV |
190 | __be32 err; |
191 | int host_err; | |
1da177e4 LT |
192 | |
193 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
194 | ||
195 | /* Obtain dentry and export. */ | |
8837abca | 196 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
197 | if (err) |
198 | return err; | |
199 | ||
200 | dparent = fhp->fh_dentry; | |
201 | exp = fhp->fh_export; | |
202 | exp_get(exp); | |
203 | ||
1da177e4 LT |
204 | /* Lookup the name, but don't follow links */ |
205 | if (isdotent(name, len)) { | |
206 | if (len==1) | |
207 | dentry = dget(dparent); | |
54775491 | 208 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 209 | dentry = dget_parent(dparent); |
fed83811 | 210 | else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp)) |
1da177e4 LT |
211 | dentry = dget(dparent); /* .. == . just like at / */ |
212 | else { | |
213 | /* checking mountpoint crossing is very different when stepping up */ | |
289ede45 BF |
214 | host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry); |
215 | if (host_err) | |
1da177e4 | 216 | goto out_nfserr; |
1da177e4 LT |
217 | } |
218 | } else { | |
219 | fh_lock(fhp); | |
220 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 221 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
222 | if (IS_ERR(dentry)) |
223 | goto out_nfserr; | |
224 | /* | |
225 | * check if we have crossed a mount point ... | |
226 | */ | |
82ead7fe | 227 | if (nfsd_mountpoint(dentry, exp)) { |
6264d69d | 228 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
229 | dput(dentry); |
230 | goto out_nfserr; | |
231 | } | |
232 | } | |
233 | } | |
6c0a654d BF |
234 | *dentry_ret = dentry; |
235 | *exp_ret = exp; | |
236 | return 0; | |
237 | ||
238 | out_nfserr: | |
239 | exp_put(exp); | |
240 | return nfserrno(host_err); | |
241 | } | |
242 | ||
243 | /* | |
244 | * Look up one component of a pathname. | |
245 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
246 | * | |
247 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
248 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
249 | * accepted as it stands and the mounted directory is | |
250 | * returned. Otherwise the covered directory is returned. | |
251 | * NOTE: this mountpoint crossing is not supported properly by all | |
252 | * clients and is explicitly disallowed for NFSv3 | |
253 | * NeilBrown <neilb@cse.unsw.edu.au> | |
254 | */ | |
255 | __be32 | |
256 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 257 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
258 | { |
259 | struct svc_export *exp; | |
260 | struct dentry *dentry; | |
261 | __be32 err; | |
262 | ||
263 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
264 | if (err) | |
265 | return err; | |
32c1eb0c AA |
266 | err = check_nfsd_access(exp, rqstp); |
267 | if (err) | |
268 | goto out; | |
1da177e4 LT |
269 | /* |
270 | * Note: we compose the file handle now, but as the | |
271 | * dentry may be negative, it may need to be updated. | |
272 | */ | |
273 | err = fh_compose(resfh, exp, dentry, fhp); | |
274 | if (!err && !dentry->d_inode) | |
275 | err = nfserr_noent; | |
32c1eb0c | 276 | out: |
1da177e4 | 277 | dput(dentry); |
1da177e4 LT |
278 | exp_put(exp); |
279 | return err; | |
1da177e4 LT |
280 | } |
281 | ||
6c0a654d | 282 | |
1da177e4 LT |
283 | /* |
284 | * Set various file attributes. | |
285 | * N.B. After this call fhp needs an fh_put | |
286 | */ | |
6264d69d | 287 | __be32 |
1da177e4 LT |
288 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
289 | int check_guard, time_t guardtime) | |
290 | { | |
291 | struct dentry *dentry; | |
292 | struct inode *inode; | |
8837abca | 293 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 294 | int ftype = 0; |
6264d69d AV |
295 | __be32 err; |
296 | int host_err; | |
1da177e4 LT |
297 | int size_change = 0; |
298 | ||
299 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 300 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
301 | if (iap->ia_valid & ATTR_SIZE) |
302 | ftype = S_IFREG; | |
303 | ||
304 | /* Get inode */ | |
305 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 306 | if (err) |
1da177e4 LT |
307 | goto out; |
308 | ||
309 | dentry = fhp->fh_dentry; | |
310 | inode = dentry->d_inode; | |
311 | ||
15b7a1b8 N |
312 | /* Ignore any mode updates on symlinks */ |
313 | if (S_ISLNK(inode->i_mode)) | |
314 | iap->ia_valid &= ~ATTR_MODE; | |
315 | ||
316 | if (!iap->ia_valid) | |
317 | goto out; | |
318 | ||
9c85fca5 CH |
319 | /* |
320 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
321 | * which only requires access, and "set-[ac]time-to-X" which |
322 | * requires ownership. | |
323 | * So if it looks like it might be "set both to the same time which | |
324 | * is close to now", and if inode_change_ok fails, then we | |
325 | * convert to "set to now" instead of "set to explicit time" | |
326 | * | |
327 | * We only call inode_change_ok as the last test as technically | |
328 | * it is not an interface that we should be using. It is only | |
329 | * valid if the filesystem does not define it's own i_op->setattr. | |
330 | */ | |
331 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
332 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
333 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
334 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
335 | /* | |
336 | * Looks probable. | |
337 | * | |
338 | * Now just make sure time is in the right ballpark. | |
339 | * Solaris, at least, doesn't seem to care what the time | |
340 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 341 | */ |
9c85fca5 CH |
342 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
343 | if (delta < 0) | |
344 | delta = -delta; | |
345 | if (delta < MAX_TOUCH_TIME_ERROR && | |
346 | inode_change_ok(inode, iap) != 0) { | |
347 | /* | |
348 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
349 | * This will cause notify_change to set these times | |
350 | * to "now" | |
351 | */ | |
352 | iap->ia_valid &= ~BOTH_TIME_SET; | |
353 | } | |
1da177e4 LT |
354 | } |
355 | ||
9c85fca5 CH |
356 | /* |
357 | * The size case is special. | |
358 | * It changes the file as well as the attributes. | |
359 | */ | |
1da177e4 LT |
360 | if (iap->ia_valid & ATTR_SIZE) { |
361 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
362 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
363 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
364 | if (err) |
365 | goto out; | |
366 | } | |
367 | ||
368 | /* | |
369 | * If we are changing the size of the file, then | |
370 | * we need to break all leases. | |
371 | */ | |
6264d69d AV |
372 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
373 | if (host_err == -EWOULDBLOCK) | |
374 | host_err = -ETIMEDOUT; | |
375 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
376 | goto out_nfserr; |
377 | ||
6264d69d AV |
378 | host_err = get_write_access(inode); |
379 | if (host_err) | |
1da177e4 LT |
380 | goto out_nfserr; |
381 | ||
382 | size_change = 1; | |
6264d69d AV |
383 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
384 | if (host_err) { | |
1da177e4 LT |
385 | put_write_access(inode); |
386 | goto out_nfserr; | |
387 | } | |
90c0af05 | 388 | vfs_dq_init(inode); |
1da177e4 LT |
389 | } |
390 | ||
ca456252 | 391 | /* sanitize the mode change */ |
1da177e4 LT |
392 | if (iap->ia_valid & ATTR_MODE) { |
393 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 394 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
395 | } |
396 | ||
397 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
398 | if (!S_ISDIR(inode->i_mode) && |
399 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
400 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
401 | iap->ia_valid |= ATTR_KILL_PRIV; |
402 | if (iap->ia_valid & ATTR_MODE) { | |
403 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 404 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
405 | if (iap->ia_mode & S_IXGRP) |
406 | iap->ia_mode &= ~S_ISGID; | |
407 | } else { | |
408 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
409 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 410 | } |
1da177e4 LT |
411 | } |
412 | ||
1da177e4 LT |
413 | /* Change the attributes. */ |
414 | ||
415 | iap->ia_valid |= ATTR_CTIME; | |
416 | ||
417 | err = nfserr_notsync; | |
418 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
419 | fh_lock(fhp); | |
6264d69d AV |
420 | host_err = notify_change(dentry, iap); |
421 | err = nfserrno(host_err); | |
1da177e4 LT |
422 | fh_unlock(fhp); |
423 | } | |
424 | if (size_change) | |
425 | put_write_access(inode); | |
426 | if (!err) | |
427 | if (EX_ISSYNC(fhp->fh_export)) | |
428 | write_inode_now(inode, 1); | |
429 | out: | |
430 | return err; | |
431 | ||
432 | out_nfserr: | |
6264d69d | 433 | err = nfserrno(host_err); |
1da177e4 LT |
434 | goto out; |
435 | } | |
436 | ||
5be196e5 CH |
437 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
438 | defined(CONFIG_NFSD_V3_ACL) || \ | |
439 | defined(CONFIG_NFSD_V4) | |
440 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
441 | { | |
442 | ssize_t buflen; | |
6c6a426f | 443 | ssize_t ret; |
5be196e5 CH |
444 | |
445 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
446 | if (buflen <= 0) | |
447 | return buflen; | |
1da177e4 | 448 | |
5be196e5 CH |
449 | *buf = kmalloc(buflen, GFP_KERNEL); |
450 | if (!*buf) | |
451 | return -ENOMEM; | |
452 | ||
6c6a426f KK |
453 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
454 | if (ret < 0) | |
455 | kfree(*buf); | |
456 | return ret; | |
5be196e5 CH |
457 | } |
458 | #endif | |
459 | ||
460 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
461 | static int |
462 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
463 | { | |
464 | int len; | |
465 | size_t buflen; | |
466 | char *buf = NULL; | |
467 | int error = 0; | |
1da177e4 LT |
468 | |
469 | buflen = posix_acl_xattr_size(pacl->a_count); | |
470 | buf = kmalloc(buflen, GFP_KERNEL); | |
471 | error = -ENOMEM; | |
472 | if (buf == NULL) | |
473 | goto out; | |
474 | ||
475 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
476 | if (len < 0) { | |
477 | error = len; | |
478 | goto out; | |
479 | } | |
480 | ||
5be196e5 | 481 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
482 | out: |
483 | kfree(buf); | |
484 | return error; | |
485 | } | |
486 | ||
6264d69d | 487 | __be32 |
1da177e4 LT |
488 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
489 | struct nfs4_acl *acl) | |
490 | { | |
6264d69d AV |
491 | __be32 error; |
492 | int host_error; | |
1da177e4 LT |
493 | struct dentry *dentry; |
494 | struct inode *inode; | |
495 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
496 | unsigned int flags = 0; | |
497 | ||
498 | /* Get inode */ | |
8837abca | 499 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 500 | if (error) |
4b2ca38a | 501 | return error; |
1da177e4 LT |
502 | |
503 | dentry = fhp->fh_dentry; | |
504 | inode = dentry->d_inode; | |
505 | if (S_ISDIR(inode->i_mode)) | |
506 | flags = NFS4_ACL_DIR; | |
507 | ||
6264d69d AV |
508 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
509 | if (host_error == -EINVAL) { | |
4b2ca38a | 510 | return nfserr_attrnotsupp; |
6264d69d | 511 | } else if (host_error < 0) |
1da177e4 LT |
512 | goto out_nfserr; |
513 | ||
6264d69d AV |
514 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
515 | if (host_error < 0) | |
4b2ca38a | 516 | goto out_release; |
1da177e4 | 517 | |
4b2ca38a | 518 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 519 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 520 | |
4b2ca38a | 521 | out_release: |
1da177e4 LT |
522 | posix_acl_release(pacl); |
523 | posix_acl_release(dpacl); | |
1da177e4 | 524 | out_nfserr: |
f34f9242 | 525 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 526 | return nfserr_attrnotsupp; |
f34f9242 | 527 | else |
4b2ca38a | 528 | return nfserrno(host_error); |
1da177e4 LT |
529 | } |
530 | ||
531 | static struct posix_acl * | |
532 | _get_posix_acl(struct dentry *dentry, char *key) | |
533 | { | |
5be196e5 | 534 | void *buf = NULL; |
1da177e4 | 535 | struct posix_acl *pacl = NULL; |
5be196e5 | 536 | int buflen; |
1da177e4 | 537 | |
5be196e5 CH |
538 | buflen = nfsd_getxattr(dentry, key, &buf); |
539 | if (!buflen) | |
540 | buflen = -ENODATA; | |
541 | if (buflen <= 0) | |
542 | return ERR_PTR(buflen); | |
1da177e4 LT |
543 | |
544 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
545 | kfree(buf); |
546 | return pacl; | |
1da177e4 LT |
547 | } |
548 | ||
549 | int | |
550 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
551 | { | |
552 | struct inode *inode = dentry->d_inode; | |
553 | int error = 0; | |
554 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
555 | unsigned int flags = 0; | |
556 | ||
9a59f452 | 557 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
558 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
559 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
560 | if (IS_ERR(pacl)) { | |
561 | error = PTR_ERR(pacl); | |
562 | pacl = NULL; | |
563 | goto out; | |
564 | } | |
565 | ||
566 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 567 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
568 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
569 | dpacl = NULL; | |
570 | else if (IS_ERR(dpacl)) { | |
571 | error = PTR_ERR(dpacl); | |
572 | dpacl = NULL; | |
573 | goto out; | |
574 | } | |
575 | flags = NFS4_ACL_DIR; | |
576 | } | |
577 | ||
578 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
579 | if (IS_ERR(*acl)) { | |
580 | error = PTR_ERR(*acl); | |
581 | *acl = NULL; | |
582 | } | |
583 | out: | |
584 | posix_acl_release(pacl); | |
585 | posix_acl_release(dpacl); | |
586 | return error; | |
587 | } | |
588 | ||
589 | #endif /* defined(CONFIG_NFS_V4) */ | |
590 | ||
591 | #ifdef CONFIG_NFSD_V3 | |
592 | /* | |
593 | * Check server access rights to a file system object | |
594 | */ | |
595 | struct accessmap { | |
596 | u32 access; | |
597 | int how; | |
598 | }; | |
599 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
600 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
601 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
602 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
603 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
604 | |
605 | { 0, 0 } | |
606 | }; | |
607 | ||
608 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
609 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
610 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
611 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
612 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
613 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
614 | |
615 | { 0, 0 } | |
616 | }; | |
617 | ||
618 | static struct accessmap nfs3_anyaccess[] = { | |
619 | /* Some clients - Solaris 2.6 at least, make an access call | |
620 | * to the server to check for access for things like /dev/null | |
621 | * (which really, the server doesn't care about). So | |
622 | * We provide simple access checking for them, looking | |
623 | * mainly at mode bits, and we make sure to ignore read-only | |
624 | * filesystem checks | |
625 | */ | |
8837abca MS |
626 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
627 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
628 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
629 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
630 | |
631 | { 0, 0 } | |
632 | }; | |
633 | ||
6264d69d | 634 | __be32 |
1da177e4 LT |
635 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
636 | { | |
637 | struct accessmap *map; | |
638 | struct svc_export *export; | |
639 | struct dentry *dentry; | |
640 | u32 query, result = 0, sresult = 0; | |
6264d69d | 641 | __be32 error; |
1da177e4 | 642 | |
8837abca | 643 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
644 | if (error) |
645 | goto out; | |
646 | ||
647 | export = fhp->fh_export; | |
648 | dentry = fhp->fh_dentry; | |
649 | ||
650 | if (S_ISREG(dentry->d_inode->i_mode)) | |
651 | map = nfs3_regaccess; | |
652 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
653 | map = nfs3_diraccess; | |
654 | else | |
655 | map = nfs3_anyaccess; | |
656 | ||
657 | ||
658 | query = *access; | |
659 | for (; map->access; map++) { | |
660 | if (map->access & query) { | |
6264d69d | 661 | __be32 err2; |
1da177e4 LT |
662 | |
663 | sresult |= map->access; | |
664 | ||
0ec757df | 665 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
666 | switch (err2) { |
667 | case nfs_ok: | |
668 | result |= map->access; | |
669 | break; | |
670 | ||
671 | /* the following error codes just mean the access was not allowed, | |
672 | * rather than an error occurred */ | |
673 | case nfserr_rofs: | |
674 | case nfserr_acces: | |
675 | case nfserr_perm: | |
676 | /* simply don't "or" in the access bit. */ | |
677 | break; | |
678 | default: | |
679 | error = err2; | |
680 | goto out; | |
681 | } | |
682 | } | |
683 | } | |
684 | *access = result; | |
685 | if (supported) | |
686 | *supported = sresult; | |
687 | ||
688 | out: | |
689 | return error; | |
690 | } | |
691 | #endif /* CONFIG_NFSD_V3 */ | |
692 | ||
693 | ||
694 | ||
695 | /* | |
696 | * Open an existing file or directory. | |
697 | * The access argument indicates the type of open (read/write/lock) | |
698 | * N.B. After this call fhp needs an fh_put | |
699 | */ | |
6264d69d | 700 | __be32 |
1da177e4 LT |
701 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
702 | int access, struct file **filp) | |
703 | { | |
704 | struct dentry *dentry; | |
705 | struct inode *inode; | |
6264d69d AV |
706 | int flags = O_RDONLY|O_LARGEFILE; |
707 | __be32 err; | |
708 | int host_err; | |
1da177e4 | 709 | |
e0e81739 DH |
710 | validate_process_creds(); |
711 | ||
1da177e4 LT |
712 | /* |
713 | * If we get here, then the client has already done an "open", | |
714 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
715 | * in case a chmod has now revoked permission. | |
716 | */ | |
8837abca | 717 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
718 | if (err) |
719 | goto out; | |
720 | ||
721 | dentry = fhp->fh_dentry; | |
722 | inode = dentry->d_inode; | |
723 | ||
724 | /* Disallow write access to files with the append-only bit set | |
725 | * or any access when mandatory locking enabled | |
726 | */ | |
727 | err = nfserr_perm; | |
8837abca | 728 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 729 | goto out; |
5e7fc436 BF |
730 | /* |
731 | * We must ignore files (but only files) which might have mandatory | |
732 | * locks on them because there is no way to know if the accesser has | |
733 | * the lock. | |
734 | */ | |
735 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
736 | goto out; |
737 | ||
738 | if (!inode->i_fop) | |
739 | goto out; | |
740 | ||
741 | /* | |
742 | * Check to see if there are any leases on this file. | |
743 | * This may block while leases are broken. | |
744 | */ | |
8837abca | 745 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0)); |
6264d69d AV |
746 | if (host_err == -EWOULDBLOCK) |
747 | host_err = -ETIMEDOUT; | |
748 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
749 | goto out_nfserr; |
750 | ||
8837abca MS |
751 | if (access & NFSD_MAY_WRITE) { |
752 | if (access & NFSD_MAY_READ) | |
9ecb6a08 BF |
753 | flags = O_RDWR|O_LARGEFILE; |
754 | else | |
755 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 756 | |
90c0af05 | 757 | vfs_dq_init(inode); |
1da177e4 | 758 | } |
54775491 | 759 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
033a666c | 760 | flags, current_cred()); |
1da177e4 | 761 | if (IS_ERR(*filp)) |
6264d69d | 762 | host_err = PTR_ERR(*filp); |
14dba533 MZ |
763 | else |
764 | ima_counts_get(*filp); | |
1da177e4 | 765 | out_nfserr: |
6264d69d | 766 | err = nfserrno(host_err); |
1da177e4 | 767 | out: |
e0e81739 | 768 | validate_process_creds(); |
1da177e4 LT |
769 | return err; |
770 | } | |
771 | ||
772 | /* | |
773 | * Close a file. | |
774 | */ | |
775 | void | |
776 | nfsd_close(struct file *filp) | |
777 | { | |
778 | fput(filp); | |
779 | } | |
780 | ||
9a8d248e BF |
781 | /* |
782 | * Sync a file | |
783 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
784 | * after it. | |
785 | */ | |
786 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, | |
787 | const struct file_operations *fop) | |
788 | { | |
789 | struct inode *inode = dp->d_inode; | |
790 | int (*fsync) (struct file *, struct dentry *, int); | |
791 | int err; | |
792 | ||
793 | err = filemap_fdatawrite(inode->i_mapping); | |
794 | if (err == 0 && fop && (fsync = fop->fsync)) | |
795 | err = fsync(filp, dp, 0); | |
796 | if (err == 0) | |
797 | err = filemap_fdatawait(inode->i_mapping); | |
798 | ||
799 | return err; | |
800 | } | |
801 | ||
a334de28 | 802 | static int |
1da177e4 LT |
803 | nfsd_sync(struct file *filp) |
804 | { | |
9a8d248e BF |
805 | int err; |
806 | struct inode *inode = filp->f_path.dentry->d_inode; | |
807 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
808 | mutex_lock(&inode->i_mutex); | |
809 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); | |
810 | mutex_unlock(&inode->i_mutex); | |
811 | ||
812 | return err; | |
1da177e4 LT |
813 | } |
814 | ||
f193fbab | 815 | int |
9a8d248e | 816 | nfsd_sync_dir(struct dentry *dp) |
1da177e4 | 817 | { |
9a8d248e | 818 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
819 | } |
820 | ||
821 | /* | |
822 | * Obtain the readahead parameters for the file | |
823 | * specified by (dev, ino). | |
824 | */ | |
1da177e4 LT |
825 | |
826 | static inline struct raparms * | |
827 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
828 | { | |
829 | struct raparms *ra, **rap, **frap = NULL; | |
830 | int depth = 0; | |
fce1456a GB |
831 | unsigned int hash; |
832 | struct raparm_hbucket *rab; | |
833 | ||
834 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
835 | rab = &raparm_hash[hash]; | |
1da177e4 | 836 | |
fce1456a GB |
837 | spin_lock(&rab->pb_lock); |
838 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
839 | if (ra->p_ino == ino && ra->p_dev == dev) |
840 | goto found; | |
841 | depth++; | |
842 | if (ra->p_count == 0) | |
843 | frap = rap; | |
844 | } | |
845 | depth = nfsdstats.ra_size*11/10; | |
846 | if (!frap) { | |
fce1456a | 847 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
848 | return NULL; |
849 | } | |
850 | rap = frap; | |
851 | ra = *frap; | |
852 | ra->p_dev = dev; | |
853 | ra->p_ino = ino; | |
854 | ra->p_set = 0; | |
fce1456a | 855 | ra->p_hindex = hash; |
1da177e4 | 856 | found: |
fce1456a | 857 | if (rap != &rab->pb_head) { |
1da177e4 | 858 | *rap = ra->p_next; |
fce1456a GB |
859 | ra->p_next = rab->pb_head; |
860 | rab->pb_head = ra; | |
1da177e4 LT |
861 | } |
862 | ra->p_count++; | |
863 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 864 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
865 | return ra; |
866 | } | |
867 | ||
868 | /* | |
cf8208d0 JA |
869 | * Grab and keep cached pages associated with a file in the svc_rqst |
870 | * so that they can be passed to the network sendmsg/sendpage routines | |
871 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
872 | */ |
873 | static int | |
cf8208d0 JA |
874 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
875 | struct splice_desc *sd) | |
1da177e4 | 876 | { |
cf8208d0 | 877 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 878 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
879 | struct page *page = buf->page; |
880 | size_t size; | |
881 | int ret; | |
1da177e4 | 882 | |
cac36bb0 | 883 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
884 | if (unlikely(ret)) |
885 | return ret; | |
886 | ||
887 | size = sd->len; | |
1da177e4 LT |
888 | |
889 | if (rqstp->rq_res.page_len == 0) { | |
890 | get_page(page); | |
44524359 N |
891 | put_page(*pp); |
892 | *pp = page; | |
893 | rqstp->rq_resused++; | |
cf8208d0 | 894 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 895 | rqstp->rq_res.page_len = size; |
44524359 | 896 | } else if (page != pp[-1]) { |
1da177e4 | 897 | get_page(page); |
250f3915 N |
898 | if (*pp) |
899 | put_page(*pp); | |
44524359 N |
900 | *pp = page; |
901 | rqstp->rq_resused++; | |
1da177e4 | 902 | rqstp->rq_res.page_len += size; |
44524359 | 903 | } else |
1da177e4 | 904 | rqstp->rq_res.page_len += size; |
1da177e4 | 905 | |
1da177e4 LT |
906 | return size; |
907 | } | |
908 | ||
cf8208d0 JA |
909 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
910 | struct splice_desc *sd) | |
911 | { | |
912 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
913 | } | |
914 | ||
a8754bee DH |
915 | static inline int svc_msnfs(struct svc_fh *ffhp) |
916 | { | |
917 | #ifdef MSNFS | |
918 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
919 | #else | |
920 | return 0; | |
921 | #endif | |
922 | } | |
923 | ||
6264d69d | 924 | static __be32 |
1da177e4 LT |
925 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
926 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
927 | { | |
928 | struct inode *inode; | |
929 | struct raparms *ra; | |
930 | mm_segment_t oldfs; | |
6264d69d AV |
931 | __be32 err; |
932 | int host_err; | |
1da177e4 LT |
933 | |
934 | err = nfserr_perm; | |
7eaa36e2 | 935 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
936 | |
937 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 938 | goto out; |
1da177e4 LT |
939 | |
940 | /* Get readahead parameters */ | |
941 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
942 | ||
943 | if (ra && ra->p_set) | |
944 | file->f_ra = ra->p_ra; | |
945 | ||
cf8208d0 JA |
946 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
947 | struct splice_desc sd = { | |
948 | .len = 0, | |
949 | .total_len = *count, | |
950 | .pos = offset, | |
951 | .u.data = rqstp, | |
952 | }; | |
953 | ||
4fbef206 | 954 | rqstp->rq_resused = 1; |
cf8208d0 | 955 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
956 | } else { |
957 | oldfs = get_fs(); | |
958 | set_fs(KERNEL_DS); | |
6264d69d | 959 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
960 | set_fs(oldfs); |
961 | } | |
962 | ||
963 | /* Write back readahead params */ | |
964 | if (ra) { | |
fce1456a GB |
965 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
966 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
967 | ra->p_ra = file->f_ra; |
968 | ra->p_set = 1; | |
969 | ra->p_count--; | |
fce1456a | 970 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
971 | } |
972 | ||
6264d69d AV |
973 | if (host_err >= 0) { |
974 | nfsdstats.io_read += host_err; | |
975 | *count = host_err; | |
1da177e4 | 976 | err = 0; |
7eaa36e2 | 977 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 978 | } else |
6264d69d | 979 | err = nfserrno(host_err); |
1da177e4 LT |
980 | out: |
981 | return err; | |
982 | } | |
983 | ||
9f708e40 NB |
984 | static void kill_suid(struct dentry *dentry) |
985 | { | |
986 | struct iattr ia; | |
b5376771 | 987 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 988 | |
1b1dcc1b | 989 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 990 | notify_change(dentry, &ia); |
1b1dcc1b | 991 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
992 | } |
993 | ||
d911df7b BF |
994 | /* |
995 | * Gathered writes: If another process is currently writing to the file, | |
996 | * there's a high chance this is another nfsd (triggered by a bulk write | |
997 | * from a client's biod). Rather than syncing the file with each write | |
998 | * request, we sleep for 10 msec. | |
999 | * | |
1000 | * I don't know if this roughly approximates C. Juszak's idea of | |
1001 | * gathered writes, but it's a nice and simple solution (IMHO), and it | |
1002 | * seems to work:-) | |
1003 | * | |
1004 | * Note: we do this only in the NFSv2 case, since v3 and higher have a | |
1005 | * better tool (separate unstable writes and commits) for solving this | |
1006 | * problem. | |
1007 | */ | |
1008 | static int wait_for_concurrent_writes(struct file *file) | |
1009 | { | |
1010 | struct inode *inode = file->f_path.dentry->d_inode; | |
1011 | static ino_t last_ino; | |
1012 | static dev_t last_dev; | |
1013 | int err = 0; | |
1014 | ||
1015 | if (atomic_read(&inode->i_writecount) > 1 | |
1016 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
1017 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); | |
1018 | msleep(10); | |
1019 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); | |
1020 | } | |
1021 | ||
1022 | if (inode->i_state & I_DIRTY) { | |
1023 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); | |
1024 | err = nfsd_sync(file); | |
1025 | } | |
1026 | last_ino = inode->i_ino; | |
1027 | last_dev = inode->i_sb->s_dev; | |
1028 | return err; | |
1029 | } | |
1030 | ||
6264d69d | 1031 | static __be32 |
1da177e4 LT |
1032 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1033 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 1034 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
1035 | { |
1036 | struct svc_export *exp; | |
1037 | struct dentry *dentry; | |
1038 | struct inode *inode; | |
1039 | mm_segment_t oldfs; | |
6264d69d AV |
1040 | __be32 err = 0; |
1041 | int host_err; | |
1da177e4 | 1042 | int stable = *stablep; |
48e03bc5 | 1043 | int use_wgather; |
1da177e4 | 1044 | |
45bd3b3d | 1045 | #ifdef MSNFS |
1da177e4 LT |
1046 | err = nfserr_perm; |
1047 | ||
1da177e4 | 1048 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
31dec253 | 1049 | (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt))) |
1da177e4 LT |
1050 | goto out; |
1051 | #endif | |
1052 | ||
7eaa36e2 | 1053 | dentry = file->f_path.dentry; |
1da177e4 LT |
1054 | inode = dentry->d_inode; |
1055 | exp = fhp->fh_export; | |
1056 | ||
1057 | /* | |
1058 | * Request sync writes if | |
1059 | * - the sync export option has been set, or | |
1060 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
1061 | * - The file system doesn't support fsync(). | |
48e03bc5 | 1062 | * When NFSv2 gathered writes have been configured for this volume, |
1da177e4 LT |
1063 | * flushing the data to disk is handled separately below. |
1064 | */ | |
48e03bc5 | 1065 | use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp); |
1da177e4 | 1066 | |
3ba15148 | 1067 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
1068 | stable = 2; |
1069 | *stablep = 2; /* FILE_SYNC */ | |
1070 | } | |
1071 | ||
1072 | if (!EX_ISSYNC(exp)) | |
1073 | stable = 0; | |
48e03bc5 | 1074 | if (stable && !use_wgather) { |
db1dd4d3 | 1075 | spin_lock(&file->f_lock); |
1da177e4 | 1076 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
1077 | spin_unlock(&file->f_lock); |
1078 | } | |
1da177e4 LT |
1079 | |
1080 | /* Write the data. */ | |
1081 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1082 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1083 | set_fs(oldfs); |
e4636d53 BF |
1084 | if (host_err < 0) |
1085 | goto out_nfserr; | |
1086 | *cnt = host_err; | |
1087 | nfsdstats.io_write += host_err; | |
1088 | fsnotify_modify(file->f_path.dentry); | |
1da177e4 LT |
1089 | |
1090 | /* clear setuid/setgid flag after write */ | |
e4636d53 | 1091 | if (inode->i_mode & (S_ISUID | S_ISGID)) |
9f708e40 | 1092 | kill_suid(dentry); |
1da177e4 | 1093 | |
e4636d53 | 1094 | if (stable && use_wgather) |
d911df7b | 1095 | host_err = wait_for_concurrent_writes(file); |
1da177e4 | 1096 | |
e4636d53 | 1097 | out_nfserr: |
6264d69d | 1098 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1099 | if (host_err >= 0) |
1da177e4 | 1100 | err = 0; |
a0d24b29 | 1101 | else |
6264d69d | 1102 | err = nfserrno(host_err); |
1da177e4 LT |
1103 | out: |
1104 | return err; | |
1105 | } | |
1106 | ||
1107 | /* | |
1108 | * Read data from a file. count must contain the requested read count | |
1109 | * on entry. On return, *count contains the number of bytes actually read. | |
1110 | * N.B. After this call fhp needs an fh_put | |
1111 | */ | |
6264d69d | 1112 | __be32 |
1da177e4 LT |
1113 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1114 | loff_t offset, struct kvec *vec, int vlen, | |
1115 | unsigned long *count) | |
1116 | { | |
6264d69d | 1117 | __be32 err; |
1da177e4 LT |
1118 | |
1119 | if (file) { | |
0ec757df | 1120 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1121 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1122 | if (err) |
1123 | goto out; | |
1124 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1125 | } else { | |
8837abca | 1126 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); |
1da177e4 LT |
1127 | if (err) |
1128 | goto out; | |
1129 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1130 | nfsd_close(file); | |
1131 | } | |
1132 | out: | |
1133 | return err; | |
1134 | } | |
1135 | ||
1136 | /* | |
1137 | * Write data to a file. | |
1138 | * The stable flag requests synchronous writes. | |
1139 | * N.B. After this call fhp needs an fh_put | |
1140 | */ | |
6264d69d | 1141 | __be32 |
1da177e4 | 1142 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1143 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1144 | int *stablep) |
1145 | { | |
6264d69d | 1146 | __be32 err = 0; |
1da177e4 LT |
1147 | |
1148 | if (file) { | |
0ec757df | 1149 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1150 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1151 | if (err) |
1152 | goto out; | |
1153 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1154 | stablep); | |
1155 | } else { | |
8837abca | 1156 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1157 | if (err) |
1158 | goto out; | |
1159 | ||
1160 | if (cnt) | |
1161 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1162 | cnt, stablep); | |
1163 | nfsd_close(file); | |
1164 | } | |
1165 | out: | |
1166 | return err; | |
1167 | } | |
1168 | ||
1169 | #ifdef CONFIG_NFSD_V3 | |
1170 | /* | |
1171 | * Commit all pending writes to stable storage. | |
1172 | * Strictly speaking, we could sync just the indicated file region here, | |
1173 | * but there's currently no way we can ask the VFS to do so. | |
1174 | * | |
1175 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1176 | * data to the client, as locking happens lower down in the filesystem. | |
1177 | */ | |
6264d69d | 1178 | __be32 |
1da177e4 LT |
1179 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1180 | loff_t offset, unsigned long count) | |
1181 | { | |
1182 | struct file *file; | |
6264d69d | 1183 | __be32 err; |
1da177e4 LT |
1184 | |
1185 | if ((u64)count > ~(u64)offset) | |
1186 | return nfserr_inval; | |
1187 | ||
8837abca MS |
1188 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1189 | if (err) | |
1da177e4 LT |
1190 | return err; |
1191 | if (EX_ISSYNC(fhp->fh_export)) { | |
1192 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1193 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1194 | } else { |
1195 | err = nfserr_notsupp; | |
1196 | } | |
1197 | } | |
1198 | ||
1199 | nfsd_close(file); | |
1200 | return err; | |
1201 | } | |
1202 | #endif /* CONFIG_NFSD_V3 */ | |
1203 | ||
f2b0dee2 | 1204 | static __be32 |
5c002b3b BF |
1205 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1206 | struct iattr *iap) | |
1207 | { | |
1208 | /* | |
1209 | * Mode has already been set earlier in create: | |
1210 | */ | |
1211 | iap->ia_valid &= ~ATTR_MODE; | |
1212 | /* | |
1213 | * Setting uid/gid works only for root. Irix appears to | |
1214 | * send along the gid on create when it tries to implement | |
1215 | * setgid directories via NFS: | |
1216 | */ | |
5cc0a840 | 1217 | if (current_fsuid() != 0) |
5c002b3b BF |
1218 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1219 | if (iap->ia_valid) | |
1220 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1221 | return 0; | |
1222 | } | |
1223 | ||
4ac35c2f | 1224 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1225 | * setting size to 0 may fail for some specific file systems by the permission | |
1226 | * checking which requires WRITE permission but the mode is 000. | |
1227 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1228 | * 0 after file created. | |
1229 | * | |
1230 | * call this only after vfs_create() is called. | |
1231 | * */ | |
1232 | static void | |
1233 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1234 | { | |
1235 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1236 | iap->ia_valid &= ~ATTR_SIZE; | |
1237 | } | |
1238 | ||
1da177e4 LT |
1239 | /* |
1240 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1241 | * not yet implemented. | |
1242 | * If the response fh has been verified, the parent directory should | |
1243 | * already be locked. Note that the parent directory is left locked. | |
1244 | * | |
1245 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1246 | */ | |
6264d69d | 1247 | __be32 |
1da177e4 LT |
1248 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1249 | char *fname, int flen, struct iattr *iap, | |
1250 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1251 | { | |
1252 | struct dentry *dentry, *dchild = NULL; | |
1253 | struct inode *dirp; | |
6264d69d | 1254 | __be32 err; |
5c002b3b | 1255 | __be32 err2; |
6264d69d | 1256 | int host_err; |
1da177e4 LT |
1257 | |
1258 | err = nfserr_perm; | |
1259 | if (!flen) | |
1260 | goto out; | |
1261 | err = nfserr_exist; | |
1262 | if (isdotent(fname, flen)) | |
1263 | goto out; | |
1264 | ||
8837abca | 1265 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1266 | if (err) |
1267 | goto out; | |
1268 | ||
1269 | dentry = fhp->fh_dentry; | |
1270 | dirp = dentry->d_inode; | |
1271 | ||
1272 | err = nfserr_notdir; | |
acfa4380 | 1273 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1274 | goto out; |
1275 | /* | |
1276 | * Check whether the response file handle has been verified yet. | |
1277 | * If it has, the parent directory should already be locked. | |
1278 | */ | |
1279 | if (!resfhp->fh_dentry) { | |
1280 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1281 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1282 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1283 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1284 | if (IS_ERR(dchild)) |
1285 | goto out_nfserr; | |
1286 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1287 | if (err) | |
1288 | goto out; | |
1289 | } else { | |
1290 | /* called from nfsd_proc_create */ | |
1291 | dchild = dget(resfhp->fh_dentry); | |
1292 | if (!fhp->fh_locked) { | |
1293 | /* not actually possible */ | |
1294 | printk(KERN_ERR | |
1295 | "nfsd_create: parent %s/%s not locked!\n", | |
1296 | dentry->d_parent->d_name.name, | |
1297 | dentry->d_name.name); | |
d75f2b9f | 1298 | err = nfserr_io; |
1da177e4 LT |
1299 | goto out; |
1300 | } | |
1301 | } | |
1302 | /* | |
1303 | * Make sure the child dentry is still negative ... | |
1304 | */ | |
1305 | err = nfserr_exist; | |
1306 | if (dchild->d_inode) { | |
1307 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1308 | dentry->d_name.name, dchild->d_name.name); | |
1309 | goto out; | |
1310 | } | |
1311 | ||
1312 | if (!(iap->ia_valid & ATTR_MODE)) | |
1313 | iap->ia_mode = 0; | |
1314 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1315 | ||
07cad1d2 MS |
1316 | err = nfserr_inval; |
1317 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1318 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1319 | type); | |
1320 | goto out; | |
1321 | } | |
1322 | ||
1323 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1324 | if (host_err) | |
1325 | goto out_nfserr; | |
1326 | ||
1da177e4 LT |
1327 | /* |
1328 | * Get the dir op function pointer. | |
1329 | */ | |
088406bc | 1330 | err = 0; |
1da177e4 LT |
1331 | switch (type) { |
1332 | case S_IFREG: | |
6264d69d | 1333 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1334 | if (!host_err) |
1335 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1336 | break; |
1337 | case S_IFDIR: | |
6264d69d | 1338 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1339 | break; |
1340 | case S_IFCHR: | |
1341 | case S_IFBLK: | |
1342 | case S_IFIFO: | |
1343 | case S_IFSOCK: | |
6264d69d | 1344 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1345 | break; |
1da177e4 | 1346 | } |
463c3197 DH |
1347 | if (host_err < 0) { |
1348 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1349 | goto out_nfserr; |
463c3197 | 1350 | } |
1da177e4 LT |
1351 | |
1352 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1353 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1354 | write_inode_now(dchild->d_inode, 1); |
1355 | } | |
1356 | ||
5c002b3b BF |
1357 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1358 | if (err2) | |
1359 | err = err2; | |
463c3197 | 1360 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1361 | /* |
1362 | * Update the file handle to get the new inode info. | |
1363 | */ | |
1364 | if (!err) | |
1365 | err = fh_update(resfhp); | |
1366 | out: | |
1367 | if (dchild && !IS_ERR(dchild)) | |
1368 | dput(dchild); | |
1369 | return err; | |
1370 | ||
1371 | out_nfserr: | |
6264d69d | 1372 | err = nfserrno(host_err); |
1da177e4 LT |
1373 | goto out; |
1374 | } | |
1375 | ||
1376 | #ifdef CONFIG_NFSD_V3 | |
1377 | /* | |
1378 | * NFSv3 version of nfsd_create | |
1379 | */ | |
6264d69d | 1380 | __be32 |
1da177e4 LT |
1381 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1382 | char *fname, int flen, struct iattr *iap, | |
1383 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1384 | int *truncp, int *created) |
1da177e4 LT |
1385 | { |
1386 | struct dentry *dentry, *dchild = NULL; | |
1387 | struct inode *dirp; | |
6264d69d | 1388 | __be32 err; |
5c002b3b | 1389 | __be32 err2; |
6264d69d | 1390 | int host_err; |
1da177e4 | 1391 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1392 | |
1393 | err = nfserr_perm; | |
1394 | if (!flen) | |
1395 | goto out; | |
1396 | err = nfserr_exist; | |
1397 | if (isdotent(fname, flen)) | |
1398 | goto out; | |
1399 | if (!(iap->ia_valid & ATTR_MODE)) | |
1400 | iap->ia_mode = 0; | |
8837abca | 1401 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1402 | if (err) |
1403 | goto out; | |
1404 | ||
1405 | dentry = fhp->fh_dentry; | |
1406 | dirp = dentry->d_inode; | |
1407 | ||
1408 | /* Get all the sanity checks out of the way before | |
1409 | * we lock the parent. */ | |
1410 | err = nfserr_notdir; | |
acfa4380 | 1411 | if (!dirp->i_op->lookup) |
1da177e4 | 1412 | goto out; |
12fd3520 | 1413 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1414 | |
1415 | /* | |
1416 | * Compose the response file handle. | |
1417 | */ | |
1418 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1419 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1420 | if (IS_ERR(dchild)) |
1421 | goto out_nfserr; | |
1422 | ||
1423 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1424 | if (err) | |
1425 | goto out; | |
1426 | ||
1427 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1428 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1429 | * the high bit set, so just clear the high bits. If this is |
1430 | * ever changed to use different attrs for storing the | |
1431 | * verifier, then do_open_lookup() will also need to be fixed | |
1432 | * accordingly. | |
1da177e4 LT |
1433 | */ |
1434 | v_mtime = verifier[0]&0x7fffffff; | |
1435 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1436 | } |
1437 | ||
463c3197 DH |
1438 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1439 | if (host_err) | |
1440 | goto out_nfserr; | |
1da177e4 LT |
1441 | if (dchild->d_inode) { |
1442 | err = 0; | |
1443 | ||
1444 | switch (createmode) { | |
1445 | case NFS3_CREATE_UNCHECKED: | |
1446 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1447 | err = nfserr_exist; | |
1448 | else if (truncp) { | |
1449 | /* in nfsv4, we need to treat this case a little | |
1450 | * differently. we don't want to truncate the | |
1451 | * file now; this would be wrong if the OPEN | |
1452 | * fails for some other reason. furthermore, | |
1453 | * if the size is nonzero, we should ignore it | |
1454 | * according to spec! | |
1455 | */ | |
1456 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1457 | } | |
1458 | else { | |
1459 | iap->ia_valid &= ATTR_SIZE; | |
1460 | goto set_attr; | |
1461 | } | |
1462 | break; | |
1463 | case NFS3_CREATE_EXCLUSIVE: | |
1464 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1465 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1466 | && dchild->d_inode->i_size == 0 ) |
1467 | break; | |
1468 | /* fallthru */ | |
1469 | case NFS3_CREATE_GUARDED: | |
1470 | err = nfserr_exist; | |
1471 | } | |
463c3197 | 1472 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1473 | goto out; |
1474 | } | |
1475 | ||
6264d69d | 1476 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1477 | if (host_err < 0) { |
1478 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1479 | goto out_nfserr; |
463c3197 | 1480 | } |
81ac95c5 BF |
1481 | if (created) |
1482 | *created = 1; | |
1da177e4 LT |
1483 | |
1484 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1485 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1486 | /* setattr will sync the child (or not) */ |
1487 | } | |
1488 | ||
4ac35c2f | 1489 | nfsd_check_ignore_resizing(iap); |
1490 | ||
1da177e4 | 1491 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1492 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1493 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1494 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1495 | /* XXX someone who knows this better please fix it for nsec */ |
1496 | iap->ia_mtime.tv_sec = v_mtime; | |
1497 | iap->ia_atime.tv_sec = v_atime; | |
1498 | iap->ia_mtime.tv_nsec = 0; | |
1499 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1500 | } |
1501 | ||
1da177e4 | 1502 | set_attr: |
5c002b3b BF |
1503 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1504 | if (err2) | |
1505 | err = err2; | |
f193fbab | 1506 | |
463c3197 | 1507 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1508 | /* |
1509 | * Update the filehandle to get the new inode info. | |
1510 | */ | |
1511 | if (!err) | |
1512 | err = fh_update(resfhp); | |
1da177e4 LT |
1513 | |
1514 | out: | |
1515 | fh_unlock(fhp); | |
1516 | if (dchild && !IS_ERR(dchild)) | |
1517 | dput(dchild); | |
1518 | return err; | |
1519 | ||
1520 | out_nfserr: | |
6264d69d | 1521 | err = nfserrno(host_err); |
1da177e4 LT |
1522 | goto out; |
1523 | } | |
1524 | #endif /* CONFIG_NFSD_V3 */ | |
1525 | ||
1526 | /* | |
1527 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1528 | * fits into the buffer. On return, it contains the true length. | |
1529 | * N.B. After this call fhp needs an fh_put | |
1530 | */ | |
6264d69d | 1531 | __be32 |
1da177e4 LT |
1532 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1533 | { | |
1534 | struct dentry *dentry; | |
1535 | struct inode *inode; | |
1536 | mm_segment_t oldfs; | |
6264d69d AV |
1537 | __be32 err; |
1538 | int host_err; | |
1da177e4 | 1539 | |
8837abca | 1540 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1541 | if (err) |
1542 | goto out; | |
1543 | ||
1544 | dentry = fhp->fh_dentry; | |
1545 | inode = dentry->d_inode; | |
1546 | ||
1547 | err = nfserr_inval; | |
acfa4380 | 1548 | if (!inode->i_op->readlink) |
1da177e4 LT |
1549 | goto out; |
1550 | ||
54775491 | 1551 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1552 | /* N.B. Why does this call need a get_fs()?? |
1553 | * Remove the set_fs and watch the fireworks:-) --okir | |
1554 | */ | |
1555 | ||
1556 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1557 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1558 | set_fs(oldfs); |
1559 | ||
6264d69d | 1560 | if (host_err < 0) |
1da177e4 | 1561 | goto out_nfserr; |
6264d69d | 1562 | *lenp = host_err; |
1da177e4 LT |
1563 | err = 0; |
1564 | out: | |
1565 | return err; | |
1566 | ||
1567 | out_nfserr: | |
6264d69d | 1568 | err = nfserrno(host_err); |
1da177e4 LT |
1569 | goto out; |
1570 | } | |
1571 | ||
1572 | /* | |
1573 | * Create a symlink and look up its inode | |
1574 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1575 | */ | |
6264d69d | 1576 | __be32 |
1da177e4 LT |
1577 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1578 | char *fname, int flen, | |
1579 | char *path, int plen, | |
1580 | struct svc_fh *resfhp, | |
1581 | struct iattr *iap) | |
1582 | { | |
1583 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1584 | __be32 err, cerr; |
1585 | int host_err; | |
1da177e4 LT |
1586 | |
1587 | err = nfserr_noent; | |
1588 | if (!flen || !plen) | |
1589 | goto out; | |
1590 | err = nfserr_exist; | |
1591 | if (isdotent(fname, flen)) | |
1592 | goto out; | |
1593 | ||
8837abca | 1594 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1595 | if (err) |
1596 | goto out; | |
1597 | fh_lock(fhp); | |
1598 | dentry = fhp->fh_dentry; | |
1599 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1600 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1601 | if (IS_ERR(dnew)) |
1602 | goto out_nfserr; | |
1603 | ||
75c3f29d DH |
1604 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1605 | if (host_err) | |
1606 | goto out_nfserr; | |
1607 | ||
1da177e4 LT |
1608 | if (unlikely(path[plen] != 0)) { |
1609 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1610 | if (path_alloced == NULL) | |
6264d69d | 1611 | host_err = -ENOMEM; |
1da177e4 LT |
1612 | else { |
1613 | strncpy(path_alloced, path, plen); | |
1614 | path_alloced[plen] = 0; | |
db2e747b | 1615 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1616 | kfree(path_alloced); |
1617 | } | |
1618 | } else | |
db2e747b | 1619 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
1da177e4 | 1620 | |
6264d69d | 1621 | if (!host_err) { |
1da177e4 | 1622 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1623 | host_err = nfsd_sync_dir(dentry); |
1624 | } | |
1625 | err = nfserrno(host_err); | |
1da177e4 LT |
1626 | fh_unlock(fhp); |
1627 | ||
75c3f29d DH |
1628 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1629 | ||
1da177e4 LT |
1630 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1631 | dput(dnew); | |
1632 | if (err==0) err = cerr; | |
1633 | out: | |
1634 | return err; | |
1635 | ||
1636 | out_nfserr: | |
6264d69d | 1637 | err = nfserrno(host_err); |
1da177e4 LT |
1638 | goto out; |
1639 | } | |
1640 | ||
1641 | /* | |
1642 | * Create a hardlink | |
1643 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1644 | */ | |
6264d69d | 1645 | __be32 |
1da177e4 LT |
1646 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1647 | char *name, int len, struct svc_fh *tfhp) | |
1648 | { | |
1649 | struct dentry *ddir, *dnew, *dold; | |
1650 | struct inode *dirp, *dest; | |
6264d69d AV |
1651 | __be32 err; |
1652 | int host_err; | |
1da177e4 | 1653 | |
8837abca | 1654 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1655 | if (err) |
1656 | goto out; | |
8837abca | 1657 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1658 | if (err) |
1659 | goto out; | |
1660 | ||
1661 | err = nfserr_perm; | |
1662 | if (!len) | |
1663 | goto out; | |
1664 | err = nfserr_exist; | |
1665 | if (isdotent(name, len)) | |
1666 | goto out; | |
1667 | ||
12fd3520 | 1668 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1669 | ddir = ffhp->fh_dentry; |
1670 | dirp = ddir->d_inode; | |
1671 | ||
1672 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1673 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1674 | if (IS_ERR(dnew)) |
1675 | goto out_nfserr; | |
1676 | ||
1677 | dold = tfhp->fh_dentry; | |
1678 | dest = dold->d_inode; | |
1679 | ||
75c3f29d DH |
1680 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1681 | if (host_err) { | |
1682 | err = nfserrno(host_err); | |
1683 | goto out_dput; | |
1684 | } | |
6264d69d AV |
1685 | host_err = vfs_link(dold, dirp, dnew); |
1686 | if (!host_err) { | |
1da177e4 | 1687 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1688 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1689 | write_inode_now(dest, 1); |
1690 | } | |
6264d69d | 1691 | err = 0; |
1da177e4 | 1692 | } else { |
6264d69d | 1693 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1694 | err = nfserr_acces; |
1695 | else | |
6264d69d | 1696 | err = nfserrno(host_err); |
1da177e4 | 1697 | } |
75c3f29d DH |
1698 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1699 | out_dput: | |
1da177e4 | 1700 | dput(dnew); |
270d56e5 DR |
1701 | out_unlock: |
1702 | fh_unlock(ffhp); | |
1da177e4 LT |
1703 | out: |
1704 | return err; | |
1705 | ||
1706 | out_nfserr: | |
6264d69d | 1707 | err = nfserrno(host_err); |
270d56e5 | 1708 | goto out_unlock; |
1da177e4 LT |
1709 | } |
1710 | ||
1711 | /* | |
1712 | * Rename a file | |
1713 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1714 | */ | |
6264d69d | 1715 | __be32 |
1da177e4 LT |
1716 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1717 | struct svc_fh *tfhp, char *tname, int tlen) | |
1718 | { | |
1719 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1720 | struct inode *fdir, *tdir; | |
6264d69d AV |
1721 | __be32 err; |
1722 | int host_err; | |
1da177e4 | 1723 | |
8837abca | 1724 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1725 | if (err) |
1726 | goto out; | |
8837abca | 1727 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1728 | if (err) |
1729 | goto out; | |
1730 | ||
1731 | fdentry = ffhp->fh_dentry; | |
1732 | fdir = fdentry->d_inode; | |
1733 | ||
1734 | tdentry = tfhp->fh_dentry; | |
1735 | tdir = tdentry->d_inode; | |
1736 | ||
1737 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1738 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1739 | goto out; |
1740 | ||
1741 | err = nfserr_perm; | |
1742 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1743 | goto out; | |
1744 | ||
1745 | /* cannot use fh_lock as we need deadlock protective ordering | |
1746 | * so do it by hand */ | |
1747 | trap = lock_rename(tdentry, fdentry); | |
1748 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1749 | fill_pre_wcc(ffhp); | |
1750 | fill_pre_wcc(tfhp); | |
1751 | ||
1752 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1753 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1754 | if (IS_ERR(odentry)) |
1755 | goto out_nfserr; | |
1756 | ||
6264d69d | 1757 | host_err = -ENOENT; |
1da177e4 LT |
1758 | if (!odentry->d_inode) |
1759 | goto out_dput_old; | |
6264d69d | 1760 | host_err = -EINVAL; |
1da177e4 LT |
1761 | if (odentry == trap) |
1762 | goto out_dput_old; | |
1763 | ||
1764 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1765 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1766 | if (IS_ERR(ndentry)) |
1767 | goto out_dput_old; | |
6264d69d | 1768 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1769 | if (ndentry == trap) |
1770 | goto out_dput_new; | |
1771 | ||
9079b1eb | 1772 | if (svc_msnfs(ffhp) && |
1da177e4 LT |
1773 | ((atomic_read(&odentry->d_count) > 1) |
1774 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1775 | host_err = -EPERM; |
9079b1eb DH |
1776 | goto out_dput_new; |
1777 | } | |
1778 | ||
1779 | host_err = -EXDEV; | |
1780 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1781 | goto out_dput_new; | |
1782 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1783 | if (host_err) | |
1784 | goto out_dput_new; | |
1785 | ||
6264d69d AV |
1786 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1787 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1788 | host_err = nfsd_sync_dir(tdentry); | |
1789 | if (!host_err) | |
1790 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1791 | } |
1792 | ||
9079b1eb DH |
1793 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1794 | ||
1da177e4 LT |
1795 | out_dput_new: |
1796 | dput(ndentry); | |
1797 | out_dput_old: | |
1798 | dput(odentry); | |
1799 | out_nfserr: | |
6264d69d | 1800 | err = nfserrno(host_err); |
1da177e4 LT |
1801 | |
1802 | /* we cannot reply on fh_unlock on the two filehandles, | |
1803 | * as that would do the wrong thing if the two directories | |
1804 | * were the same, so again we do it by hand | |
1805 | */ | |
1806 | fill_post_wcc(ffhp); | |
1807 | fill_post_wcc(tfhp); | |
1808 | unlock_rename(tdentry, fdentry); | |
1809 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1810 | ||
1811 | out: | |
1812 | return err; | |
1813 | } | |
1814 | ||
1815 | /* | |
1816 | * Unlink a file or directory | |
1817 | * N.B. After this call fhp needs an fh_put | |
1818 | */ | |
6264d69d | 1819 | __be32 |
1da177e4 LT |
1820 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1821 | char *fname, int flen) | |
1822 | { | |
1823 | struct dentry *dentry, *rdentry; | |
1824 | struct inode *dirp; | |
6264d69d AV |
1825 | __be32 err; |
1826 | int host_err; | |
1da177e4 LT |
1827 | |
1828 | err = nfserr_acces; | |
1829 | if (!flen || isdotent(fname, flen)) | |
1830 | goto out; | |
8837abca | 1831 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1832 | if (err) |
1833 | goto out; | |
1834 | ||
12fd3520 | 1835 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1836 | dentry = fhp->fh_dentry; |
1837 | dirp = dentry->d_inode; | |
1838 | ||
1839 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1840 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1841 | if (IS_ERR(rdentry)) |
1842 | goto out_nfserr; | |
1843 | ||
1844 | if (!rdentry->d_inode) { | |
1845 | dput(rdentry); | |
1846 | err = nfserr_noent; | |
1847 | goto out; | |
1848 | } | |
1849 | ||
1850 | if (!type) | |
1851 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1852 | ||
0622753b DH |
1853 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1854 | if (host_err) | |
1855 | goto out_nfserr; | |
1856 | ||
1da177e4 LT |
1857 | if (type != S_IFDIR) { /* It's UNLINK */ |
1858 | #ifdef MSNFS | |
1859 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1860 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1861 | host_err = -EPERM; |
1da177e4 LT |
1862 | } else |
1863 | #endif | |
6264d69d | 1864 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1865 | } else { /* It's RMDIR */ |
6264d69d | 1866 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1867 | } |
1868 | ||
1869 | dput(rdentry); | |
1870 | ||
6264d69d | 1871 | if (host_err) |
0622753b | 1872 | goto out_drop; |
6264d69d AV |
1873 | if (EX_ISSYNC(fhp->fh_export)) |
1874 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 | 1875 | |
0622753b DH |
1876 | out_drop: |
1877 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1878 | out_nfserr: |
6264d69d | 1879 | err = nfserrno(host_err); |
f193fbab YT |
1880 | out: |
1881 | return err; | |
1da177e4 LT |
1882 | } |
1883 | ||
14f7dd63 DW |
1884 | /* |
1885 | * We do this buffering because we must not call back into the file | |
1886 | * system's ->lookup() method from the filldir callback. That may well | |
1887 | * deadlock a number of file systems. | |
1888 | * | |
1889 | * This is based heavily on the implementation of same in XFS. | |
1890 | */ | |
1891 | struct buffered_dirent { | |
1892 | u64 ino; | |
1893 | loff_t offset; | |
1894 | int namlen; | |
1895 | unsigned int d_type; | |
1896 | char name[]; | |
1897 | }; | |
1898 | ||
1899 | struct readdir_data { | |
1900 | char *dirent; | |
1901 | size_t used; | |
53c9c5c0 | 1902 | int full; |
14f7dd63 DW |
1903 | }; |
1904 | ||
1905 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1906 | loff_t offset, u64 ino, unsigned int d_type) | |
1907 | { | |
1908 | struct readdir_data *buf = __buf; | |
1909 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1910 | unsigned int reclen; | |
1911 | ||
1912 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1913 | if (buf->used + reclen > PAGE_SIZE) { |
1914 | buf->full = 1; | |
14f7dd63 | 1915 | return -EINVAL; |
53c9c5c0 | 1916 | } |
14f7dd63 DW |
1917 | |
1918 | de->namlen = namlen; | |
1919 | de->offset = offset; | |
1920 | de->ino = ino; | |
1921 | de->d_type = d_type; | |
1922 | memcpy(de->name, name, namlen); | |
1923 | buf->used += reclen; | |
1924 | ||
1925 | return 0; | |
1926 | } | |
1927 | ||
2f9092e1 DW |
1928 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
1929 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1930 | { |
14f7dd63 DW |
1931 | struct readdir_data buf; |
1932 | struct buffered_dirent *de; | |
2628b766 | 1933 | int host_err; |
14f7dd63 DW |
1934 | int size; |
1935 | loff_t offset; | |
2628b766 | 1936 | |
14f7dd63 DW |
1937 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1938 | if (!buf.dirent) | |
2f9092e1 | 1939 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
1940 | |
1941 | offset = *offsetp; | |
2628b766 | 1942 | |
14f7dd63 | 1943 | while (1) { |
2f9092e1 | 1944 | struct inode *dir_inode = file->f_path.dentry->d_inode; |
14f7dd63 DW |
1945 | unsigned int reclen; |
1946 | ||
b726e923 | 1947 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1948 | buf.used = 0; |
53c9c5c0 | 1949 | buf.full = 0; |
14f7dd63 DW |
1950 | |
1951 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1952 | if (buf.full) |
1953 | host_err = 0; | |
1954 | ||
1955 | if (host_err < 0) | |
14f7dd63 DW |
1956 | break; |
1957 | ||
1958 | size = buf.used; | |
1959 | ||
1960 | if (!size) | |
1961 | break; | |
1962 | ||
2f9092e1 DW |
1963 | /* |
1964 | * Various filldir functions may end up calling back into | |
1965 | * lookup_one_len() and the file system's ->lookup() method. | |
1966 | * These expect i_mutex to be held, as it would within readdir. | |
1967 | */ | |
1968 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
1969 | if (host_err) | |
1970 | break; | |
1971 | ||
14f7dd63 DW |
1972 | de = (struct buffered_dirent *)buf.dirent; |
1973 | while (size > 0) { | |
1974 | offset = de->offset; | |
1975 | ||
1976 | if (func(cdp, de->name, de->namlen, de->offset, | |
1977 | de->ino, de->d_type)) | |
2f9092e1 | 1978 | break; |
14f7dd63 DW |
1979 | |
1980 | if (cdp->err != nfs_ok) | |
2f9092e1 | 1981 | break; |
14f7dd63 DW |
1982 | |
1983 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1984 | sizeof(u64)); | |
1985 | size -= reclen; | |
1986 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1987 | } | |
2f9092e1 DW |
1988 | mutex_unlock(&dir_inode->i_mutex); |
1989 | if (size > 0) /* We bailed out early */ | |
1990 | break; | |
1991 | ||
c002a6c7 | 1992 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1993 | } |
1994 | ||
14f7dd63 | 1995 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
1996 | |
1997 | if (host_err) | |
1998 | return nfserrno(host_err); | |
14f7dd63 DW |
1999 | |
2000 | *offsetp = offset; | |
2001 | return cdp->err; | |
2628b766 DW |
2002 | } |
2003 | ||
1da177e4 LT |
2004 | /* |
2005 | * Read entries from a directory. | |
2006 | * The NFSv3/4 verifier we ignore for now. | |
2007 | */ | |
6264d69d | 2008 | __be32 |
1da177e4 | 2009 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 2010 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 2011 | { |
6264d69d | 2012 | __be32 err; |
1da177e4 LT |
2013 | struct file *file; |
2014 | loff_t offset = *offsetp; | |
2015 | ||
8837abca | 2016 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
2017 | if (err) |
2018 | goto out; | |
2019 | ||
2020 | offset = vfs_llseek(file, offset, 0); | |
2021 | if (offset < 0) { | |
2022 | err = nfserrno((int)offset); | |
2023 | goto out_close; | |
2024 | } | |
2025 | ||
14f7dd63 | 2026 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
2027 | |
2028 | if (err == nfserr_eof || err == nfserr_toosmall) | |
2029 | err = nfs_ok; /* can still be found in ->err */ | |
2030 | out_close: | |
2031 | nfsd_close(file); | |
2032 | out: | |
2033 | return err; | |
2034 | } | |
2035 | ||
2036 | /* | |
2037 | * Get file system stats | |
2038 | * N.B. After this call fhp needs an fh_put | |
2039 | */ | |
6264d69d | 2040 | __be32 |
04716e66 | 2041 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 2042 | { |
04716e66 | 2043 | __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); |
726c3342 | 2044 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
2045 | err = nfserr_io; |
2046 | return err; | |
2047 | } | |
2048 | ||
c7d51402 | 2049 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 2050 | { |
c7d51402 | 2051 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
2052 | } |
2053 | ||
1da177e4 LT |
2054 | /* |
2055 | * Check for a user's access permissions to this inode. | |
2056 | */ | |
6264d69d | 2057 | __be32 |
0ec757df BF |
2058 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2059 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2060 | { |
2061 | struct inode *inode = dentry->d_inode; | |
14dba533 | 2062 | struct path path; |
1da177e4 LT |
2063 | int err; |
2064 | ||
8837abca | 2065 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2066 | return 0; |
2067 | #if 0 | |
2068 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2069 | acc, | |
8837abca MS |
2070 | (acc & NFSD_MAY_READ)? " read" : "", |
2071 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2072 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2073 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2074 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2075 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2076 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2077 | inode->i_mode, |
2078 | IS_IMMUTABLE(inode)? " immut" : "", | |
2079 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2080 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2081 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2082 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2083 | #endif |
2084 | ||
2085 | /* Normally we reject any write/sattr etc access on a read-only file | |
2086 | * system. But if it is IRIX doing check on write-access for a | |
2087 | * device special file, we ignore rofs. | |
2088 | */ | |
8837abca MS |
2089 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2090 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2091 | if (exp_rdonly(rqstp, exp) || |
2092 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2093 | return nfserr_rofs; |
8837abca | 2094 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2095 | return nfserr_perm; |
2096 | } | |
8837abca | 2097 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2098 | return nfserr_perm; |
2099 | ||
8837abca | 2100 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2101 | /* If we cannot rely on authentication in NLM requests, |
2102 | * just allow locks, otherwise require read permission, or | |
2103 | * ownership | |
2104 | */ | |
2105 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2106 | return 0; | |
2107 | else | |
8837abca | 2108 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2109 | } |
2110 | /* | |
2111 | * The file owner always gets access permission for accesses that | |
2112 | * would normally be checked at open time. This is to make | |
2113 | * file access work even when the client has done a fchmod(fd, 0). | |
2114 | * | |
2115 | * However, `cp foo bar' should fail nevertheless when bar is | |
2116 | * readonly. A sensible way to do this might be to reject all | |
2117 | * attempts to truncate a read-only file, because a creat() call | |
2118 | * always implies file truncation. | |
2119 | * ... but this isn't really fair. A process may reasonably call | |
2120 | * ftruncate on an open file descriptor on a file with perm 000. | |
2121 | * We must trust the client to do permission checking - using "ACCESS" | |
2122 | * with NFSv3. | |
2123 | */ | |
8837abca | 2124 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2125 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2126 | return 0; |
2127 | ||
8837abca | 2128 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2129 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2130 | |
2131 | /* Allow read access to binaries even when mode 111 */ | |
2132 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2133 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2134 | err = inode_permission(inode, MAY_EXEC); |
14dba533 MZ |
2135 | if (err) |
2136 | goto nfsd_out; | |
1da177e4 | 2137 | |
14dba533 MZ |
2138 | /* Do integrity (permission) checking now, but defer incrementing |
2139 | * IMA counts to the actual file open. | |
2140 | */ | |
2141 | path.mnt = exp->ex_path.mnt; | |
2142 | path.dentry = dentry; | |
2143 | err = ima_path_check(&path, acc & (MAY_READ | MAY_WRITE | MAY_EXEC), | |
2144 | IMA_COUNT_LEAVE); | |
2145 | nfsd_out: | |
1da177e4 LT |
2146 | return err? nfserrno(err) : 0; |
2147 | } | |
2148 | ||
2149 | void | |
2150 | nfsd_racache_shutdown(void) | |
2151 | { | |
54a66e54 JL |
2152 | struct raparms *raparm, *last_raparm; |
2153 | unsigned int i; | |
2154 | ||
1da177e4 | 2155 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2156 | |
2157 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2158 | raparm = raparm_hash[i].pb_head; | |
2159 | while(raparm) { | |
2160 | last_raparm = raparm; | |
2161 | raparm = raparm->p_next; | |
2162 | kfree(last_raparm); | |
2163 | } | |
2164 | raparm_hash[i].pb_head = NULL; | |
2165 | } | |
1da177e4 LT |
2166 | } |
2167 | /* | |
2168 | * Initialize readahead param cache | |
2169 | */ | |
2170 | int | |
2171 | nfsd_racache_init(int cache_size) | |
2172 | { | |
2173 | int i; | |
fce1456a GB |
2174 | int j = 0; |
2175 | int nperbucket; | |
54a66e54 | 2176 | struct raparms **raparm = NULL; |
1da177e4 | 2177 | |
fce1456a | 2178 | |
54a66e54 | 2179 | if (raparm_hash[0].pb_head) |
1da177e4 | 2180 | return 0; |
54a66e54 JL |
2181 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2182 | if (nperbucket < 2) | |
2183 | nperbucket = 2; | |
2184 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2185 | |
2186 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2187 | |
2188 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2189 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2190 | |
2191 | raparm = &raparm_hash[i].pb_head; | |
2192 | for (j = 0; j < nperbucket; j++) { | |
2193 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2194 | if (!*raparm) | |
2195 | goto out_nomem; | |
2196 | raparm = &(*raparm)->p_next; | |
2197 | } | |
2198 | *raparm = NULL; | |
4b3bb06b YB |
2199 | } |
2200 | ||
1da177e4 LT |
2201 | nfsdstats.ra_size = cache_size; |
2202 | return 0; | |
54a66e54 JL |
2203 | |
2204 | out_nomem: | |
2205 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2206 | nfsd_racache_shutdown(); | |
2207 | return -ENOMEM; | |
1da177e4 | 2208 | } |
a257cdd0 AG |
2209 | |
2210 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2211 | struct posix_acl * | |
2212 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2213 | { | |
2214 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2215 | char *name; | |
2216 | void *value = NULL; | |
2217 | ssize_t size; | |
2218 | struct posix_acl *acl; | |
2219 | ||
5be196e5 CH |
2220 | if (!IS_POSIXACL(inode)) |
2221 | return ERR_PTR(-EOPNOTSUPP); | |
2222 | ||
2223 | switch (type) { | |
2224 | case ACL_TYPE_ACCESS: | |
2225 | name = POSIX_ACL_XATTR_ACCESS; | |
2226 | break; | |
2227 | case ACL_TYPE_DEFAULT: | |
2228 | name = POSIX_ACL_XATTR_DEFAULT; | |
2229 | break; | |
2230 | default: | |
a257cdd0 | 2231 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2232 | } |
2233 | ||
5be196e5 CH |
2234 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2235 | if (size < 0) | |
2236 | return ERR_PTR(size); | |
a257cdd0 | 2237 | |
a257cdd0 | 2238 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2239 | kfree(value); |
2240 | return acl; | |
2241 | } | |
2242 | ||
2243 | int | |
2244 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2245 | { | |
2246 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2247 | char *name; | |
2248 | void *value = NULL; | |
2249 | size_t size; | |
2250 | int error; | |
2251 | ||
acfa4380 | 2252 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2253 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2254 | return -EOPNOTSUPP; | |
2255 | switch(type) { | |
2256 | case ACL_TYPE_ACCESS: | |
334a13ec | 2257 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2258 | break; |
2259 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2260 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2261 | break; |
2262 | default: | |
2263 | return -EOPNOTSUPP; | |
2264 | } | |
2265 | ||
2266 | if (acl && acl->a_count) { | |
334a13ec | 2267 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2268 | value = kmalloc(size, GFP_KERNEL); |
2269 | if (!value) | |
2270 | return -ENOMEM; | |
9ccfc29c FM |
2271 | error = posix_acl_to_xattr(acl, value, size); |
2272 | if (error < 0) | |
a257cdd0 | 2273 | goto getout; |
9ccfc29c | 2274 | size = error; |
a257cdd0 AG |
2275 | } else |
2276 | size = 0; | |
2277 | ||
18f335af DH |
2278 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2279 | if (error) | |
2280 | goto getout; | |
a257cdd0 | 2281 | if (size) |
5be196e5 | 2282 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2283 | else { |
2284 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2285 | error = 0; | |
2286 | else { | |
5be196e5 | 2287 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2288 | if (error == -ENODATA) |
2289 | error = 0; | |
2290 | } | |
2291 | } | |
18f335af | 2292 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2293 | |
2294 | getout: | |
2295 | kfree(value); | |
2296 | return error; | |
2297 | } | |
2298 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |