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