]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * File operations used by nfsd. Some of these have been ripped from |
3 | * other parts of the kernel because they weren't exported, others | |
4 | * are partial duplicates with added or changed functionality. | |
5 | * | |
6 | * Note that several functions dget() the dentry upon which they want | |
7 | * to act, most notably those that create directory entries. Response | |
8 | * dentry's are dput()'d if necessary in the release callback. | |
9 | * So if you notice code paths that apparently fail to dput() the | |
10 | * dentry, don't worry--they have been taken care of. | |
11 | * | |
12 | * Copyright (C) 1995-1999 Olaf Kirch <okir@monad.swb.de> | |
13 | * Zerocpy NFS support (C) 2002 Hirokazu Takahashi <taka@valinux.co.jp> | |
14 | */ | |
15 | ||
1da177e4 LT |
16 | #include <linux/fs.h> |
17 | #include <linux/file.h> | |
d6b29d7c | 18 | #include <linux/splice.h> |
1da177e4 | 19 | #include <linux/fcntl.h> |
1da177e4 | 20 | #include <linux/namei.h> |
1da177e4 | 21 | #include <linux/delay.h> |
0eeca283 | 22 | #include <linux/fsnotify.h> |
1da177e4 | 23 | #include <linux/posix_acl_xattr.h> |
1da177e4 | 24 | #include <linux/xattr.h> |
9a74af21 BH |
25 | #include <linux/jhash.h> |
26 | #include <linux/ima.h> | |
5a0e3ad6 | 27 | #include <linux/slab.h> |
9a74af21 | 28 | #include <asm/uaccess.h> |
f501912a BM |
29 | #include <linux/exportfs.h> |
30 | #include <linux/writeback.h> | |
9a74af21 BH |
31 | |
32 | #ifdef CONFIG_NFSD_V3 | |
33 | #include "xdr3.h" | |
34 | #endif /* CONFIG_NFSD_V3 */ | |
35 | ||
5be196e5 | 36 | #ifdef CONFIG_NFSD_V4 |
2ca72e17 BF |
37 | #include "acl.h" |
38 | #include "idmap.h" | |
1da177e4 LT |
39 | #endif /* CONFIG_NFSD_V4 */ |
40 | ||
9a74af21 BH |
41 | #include "nfsd.h" |
42 | #include "vfs.h" | |
1da177e4 LT |
43 | |
44 | #define NFSDDBG_FACILITY NFSDDBG_FILEOP | |
1da177e4 LT |
45 | |
46 | ||
1da177e4 LT |
47 | /* |
48 | * This is a cache of readahead params that help us choose the proper | |
49 | * readahead strategy. Initially, we set all readahead parameters to 0 | |
50 | * and let the VFS handle things. | |
51 | * If you increase the number of cached files very much, you'll need to | |
52 | * add a hash table here. | |
53 | */ | |
54 | struct raparms { | |
55 | struct raparms *p_next; | |
56 | unsigned int p_count; | |
57 | ino_t p_ino; | |
58 | dev_t p_dev; | |
59 | int p_set; | |
60 | struct file_ra_state p_ra; | |
fce1456a | 61 | unsigned int p_hindex; |
1da177e4 LT |
62 | }; |
63 | ||
fce1456a GB |
64 | struct raparm_hbucket { |
65 | struct raparms *pb_head; | |
66 | spinlock_t pb_lock; | |
67 | } ____cacheline_aligned_in_smp; | |
68 | ||
fce1456a GB |
69 | #define RAPARM_HASH_BITS 4 |
70 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
71 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
72 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
73 | |
74 | /* | |
75 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
76 | * a mount point. | |
e0bb89ef | 77 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
78 | * or nfs_ok having possibly changed *dpp and *expp |
79 | */ | |
80 | int | |
81 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
82 | struct svc_export **expp) | |
83 | { | |
84 | struct svc_export *exp = *expp, *exp2 = NULL; | |
85 | struct dentry *dentry = *dpp; | |
91c9fa8f AV |
86 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
87 | .dentry = dget(dentry)}; | |
6264d69d | 88 | int err = 0; |
1da177e4 | 89 | |
9393bd07 | 90 | while (d_mountpoint(path.dentry) && follow_down(&path)) |
91c9fa8f | 91 | ; |
1da177e4 | 92 | |
91c9fa8f | 93 | exp2 = rqst_exp_get_by_name(rqstp, &path); |
1da177e4 | 94 | if (IS_ERR(exp2)) { |
3b6cee7b BF |
95 | err = PTR_ERR(exp2); |
96 | /* | |
97 | * We normally allow NFS clients to continue | |
98 | * "underneath" a mountpoint that is not exported. | |
99 | * The exception is V4ROOT, where no traversal is ever | |
100 | * allowed without an explicit export of the new | |
101 | * directory. | |
102 | */ | |
103 | if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT)) | |
104 | err = 0; | |
91c9fa8f | 105 | path_put(&path); |
1da177e4 LT |
106 | goto out; |
107 | } | |
3c394dda SD |
108 | if (nfsd_v4client(rqstp) || |
109 | (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { | |
1da177e4 | 110 | /* successfully crossed mount point */ |
1644ccc8 | 111 | /* |
91c9fa8f AV |
112 | * This is subtle: path.dentry is *not* on path.mnt |
113 | * at this point. The only reason we are safe is that | |
114 | * original mnt is pinned down by exp, so we should | |
115 | * put path *before* putting exp | |
1644ccc8 | 116 | */ |
91c9fa8f AV |
117 | *dpp = path.dentry; |
118 | path.dentry = dentry; | |
1644ccc8 | 119 | *expp = exp2; |
91c9fa8f | 120 | exp2 = exp; |
1da177e4 | 121 | } |
91c9fa8f AV |
122 | path_put(&path); |
123 | exp_put(exp2); | |
1da177e4 LT |
124 | out: |
125 | return err; | |
126 | } | |
127 | ||
289ede45 BF |
128 | static void follow_to_parent(struct path *path) |
129 | { | |
130 | struct dentry *dp; | |
131 | ||
132 | while (path->dentry == path->mnt->mnt_root && follow_up(path)) | |
133 | ; | |
134 | dp = dget_parent(path->dentry); | |
135 | dput(path->dentry); | |
136 | path->dentry = dp; | |
137 | } | |
138 | ||
139 | static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp) | |
140 | { | |
141 | struct svc_export *exp2; | |
142 | struct path path = {.mnt = mntget((*exp)->ex_path.mnt), | |
143 | .dentry = dget(dparent)}; | |
144 | ||
145 | follow_to_parent(&path); | |
146 | ||
147 | exp2 = rqst_exp_parent(rqstp, &path); | |
148 | if (PTR_ERR(exp2) == -ENOENT) { | |
149 | *dentryp = dget(dparent); | |
150 | } else if (IS_ERR(exp2)) { | |
151 | path_put(&path); | |
152 | return PTR_ERR(exp2); | |
153 | } else { | |
154 | *dentryp = dget(path.dentry); | |
155 | exp_put(*exp); | |
156 | *exp = exp2; | |
157 | } | |
158 | path_put(&path); | |
159 | return 0; | |
160 | } | |
161 | ||
82ead7fe BF |
162 | /* |
163 | * For nfsd purposes, we treat V4ROOT exports as though there was an | |
164 | * export at *every* directory. | |
165 | */ | |
3227fa41 | 166 | int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp) |
82ead7fe BF |
167 | { |
168 | if (d_mountpoint(dentry)) | |
169 | return 1; | |
170 | if (!(exp->ex_flags & NFSEXP_V4ROOT)) | |
171 | return 0; | |
172 | return dentry->d_inode != NULL; | |
173 | } | |
174 | ||
6264d69d | 175 | __be32 |
6c0a654d | 176 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 177 | const char *name, unsigned int len, |
6c0a654d | 178 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
179 | { |
180 | struct svc_export *exp; | |
181 | struct dentry *dparent; | |
182 | struct dentry *dentry; | |
6264d69d AV |
183 | __be32 err; |
184 | int host_err; | |
1da177e4 LT |
185 | |
186 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
187 | ||
188 | /* Obtain dentry and export. */ | |
8837abca | 189 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
190 | if (err) |
191 | return err; | |
192 | ||
193 | dparent = fhp->fh_dentry; | |
194 | exp = fhp->fh_export; | |
195 | exp_get(exp); | |
196 | ||
1da177e4 LT |
197 | /* Lookup the name, but don't follow links */ |
198 | if (isdotent(name, len)) { | |
199 | if (len==1) | |
200 | dentry = dget(dparent); | |
54775491 | 201 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 202 | dentry = dget_parent(dparent); |
fed83811 | 203 | else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp)) |
1da177e4 LT |
204 | dentry = dget(dparent); /* .. == . just like at / */ |
205 | else { | |
206 | /* checking mountpoint crossing is very different when stepping up */ | |
289ede45 BF |
207 | host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry); |
208 | if (host_err) | |
1da177e4 | 209 | goto out_nfserr; |
1da177e4 LT |
210 | } |
211 | } else { | |
212 | fh_lock(fhp); | |
213 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 214 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
215 | if (IS_ERR(dentry)) |
216 | goto out_nfserr; | |
217 | /* | |
218 | * check if we have crossed a mount point ... | |
219 | */ | |
82ead7fe | 220 | if (nfsd_mountpoint(dentry, exp)) { |
6264d69d | 221 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
222 | dput(dentry); |
223 | goto out_nfserr; | |
224 | } | |
225 | } | |
226 | } | |
6c0a654d BF |
227 | *dentry_ret = dentry; |
228 | *exp_ret = exp; | |
229 | return 0; | |
230 | ||
231 | out_nfserr: | |
232 | exp_put(exp); | |
233 | return nfserrno(host_err); | |
234 | } | |
235 | ||
236 | /* | |
237 | * Look up one component of a pathname. | |
238 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
239 | * | |
240 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
241 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
242 | * accepted as it stands and the mounted directory is | |
243 | * returned. Otherwise the covered directory is returned. | |
244 | * NOTE: this mountpoint crossing is not supported properly by all | |
245 | * clients and is explicitly disallowed for NFSv3 | |
246 | * NeilBrown <neilb@cse.unsw.edu.au> | |
247 | */ | |
248 | __be32 | |
249 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 250 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
251 | { |
252 | struct svc_export *exp; | |
253 | struct dentry *dentry; | |
254 | __be32 err; | |
255 | ||
256 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
257 | if (err) | |
258 | return err; | |
32c1eb0c AA |
259 | err = check_nfsd_access(exp, rqstp); |
260 | if (err) | |
261 | goto out; | |
1da177e4 LT |
262 | /* |
263 | * Note: we compose the file handle now, but as the | |
264 | * dentry may be negative, it may need to be updated. | |
265 | */ | |
266 | err = fh_compose(resfh, exp, dentry, fhp); | |
267 | if (!err && !dentry->d_inode) | |
268 | err = nfserr_noent; | |
32c1eb0c | 269 | out: |
1da177e4 | 270 | dput(dentry); |
1da177e4 LT |
271 | exp_put(exp); |
272 | return err; | |
1da177e4 LT |
273 | } |
274 | ||
f501912a BM |
275 | /* |
276 | * Commit metadata changes to stable storage. | |
277 | */ | |
278 | static int | |
279 | commit_metadata(struct svc_fh *fhp) | |
280 | { | |
281 | struct inode *inode = fhp->fh_dentry->d_inode; | |
282 | const struct export_operations *export_ops = inode->i_sb->s_export_op; | |
f501912a BM |
283 | |
284 | if (!EX_ISSYNC(fhp->fh_export)) | |
285 | return 0; | |
286 | ||
c3765016 CH |
287 | if (export_ops->commit_metadata) |
288 | return export_ops->commit_metadata(inode); | |
289 | return sync_inode_metadata(inode, 1); | |
f501912a | 290 | } |
6c0a654d | 291 | |
1da177e4 LT |
292 | /* |
293 | * Set various file attributes. | |
294 | * N.B. After this call fhp needs an fh_put | |
295 | */ | |
6264d69d | 296 | __be32 |
1da177e4 LT |
297 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
298 | int check_guard, time_t guardtime) | |
299 | { | |
300 | struct dentry *dentry; | |
301 | struct inode *inode; | |
8837abca | 302 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 303 | int ftype = 0; |
6264d69d AV |
304 | __be32 err; |
305 | int host_err; | |
1da177e4 LT |
306 | int size_change = 0; |
307 | ||
308 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 309 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
310 | if (iap->ia_valid & ATTR_SIZE) |
311 | ftype = S_IFREG; | |
312 | ||
313 | /* Get inode */ | |
314 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 315 | if (err) |
1da177e4 LT |
316 | goto out; |
317 | ||
318 | dentry = fhp->fh_dentry; | |
319 | inode = dentry->d_inode; | |
320 | ||
15b7a1b8 N |
321 | /* Ignore any mode updates on symlinks */ |
322 | if (S_ISLNK(inode->i_mode)) | |
323 | iap->ia_valid &= ~ATTR_MODE; | |
324 | ||
325 | if (!iap->ia_valid) | |
326 | goto out; | |
327 | ||
9c85fca5 CH |
328 | /* |
329 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
330 | * which only requires access, and "set-[ac]time-to-X" which |
331 | * requires ownership. | |
332 | * So if it looks like it might be "set both to the same time which | |
333 | * is close to now", and if inode_change_ok fails, then we | |
334 | * convert to "set to now" instead of "set to explicit time" | |
335 | * | |
336 | * We only call inode_change_ok as the last test as technically | |
337 | * it is not an interface that we should be using. It is only | |
338 | * valid if the filesystem does not define it's own i_op->setattr. | |
339 | */ | |
340 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
341 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
342 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
343 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
344 | /* | |
345 | * Looks probable. | |
346 | * | |
347 | * Now just make sure time is in the right ballpark. | |
348 | * Solaris, at least, doesn't seem to care what the time | |
349 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 350 | */ |
9c85fca5 CH |
351 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
352 | if (delta < 0) | |
353 | delta = -delta; | |
354 | if (delta < MAX_TOUCH_TIME_ERROR && | |
355 | inode_change_ok(inode, iap) != 0) { | |
356 | /* | |
357 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
358 | * This will cause notify_change to set these times | |
359 | * to "now" | |
360 | */ | |
361 | iap->ia_valid &= ~BOTH_TIME_SET; | |
362 | } | |
1da177e4 LT |
363 | } |
364 | ||
9c85fca5 CH |
365 | /* |
366 | * The size case is special. | |
367 | * It changes the file as well as the attributes. | |
368 | */ | |
1da177e4 LT |
369 | if (iap->ia_valid & ATTR_SIZE) { |
370 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
371 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
372 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
373 | if (err) |
374 | goto out; | |
375 | } | |
376 | ||
6264d69d AV |
377 | host_err = get_write_access(inode); |
378 | if (host_err) | |
1da177e4 LT |
379 | goto out_nfserr; |
380 | ||
381 | size_change = 1; | |
6264d69d AV |
382 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
383 | if (host_err) { | |
1da177e4 LT |
384 | put_write_access(inode); |
385 | goto out_nfserr; | |
386 | } | |
1da177e4 LT |
387 | } |
388 | ||
ca456252 | 389 | /* sanitize the mode change */ |
1da177e4 LT |
390 | if (iap->ia_valid & ATTR_MODE) { |
391 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 392 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
393 | } |
394 | ||
395 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
396 | if (!S_ISDIR(inode->i_mode) && |
397 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
398 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
399 | iap->ia_valid |= ATTR_KILL_PRIV; |
400 | if (iap->ia_valid & ATTR_MODE) { | |
401 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 402 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
403 | if (iap->ia_mode & S_IXGRP) |
404 | iap->ia_mode &= ~S_ISGID; | |
405 | } else { | |
406 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
407 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 408 | } |
1da177e4 LT |
409 | } |
410 | ||
1da177e4 LT |
411 | /* Change the attributes. */ |
412 | ||
413 | iap->ia_valid |= ATTR_CTIME; | |
414 | ||
415 | err = nfserr_notsync; | |
416 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
6a76bebe BF |
417 | host_err = break_lease(inode, O_WRONLY | O_NONBLOCK); |
418 | if (host_err) | |
419 | goto out_nfserr; | |
1da177e4 | 420 | fh_lock(fhp); |
6a76bebe | 421 | |
6264d69d AV |
422 | host_err = notify_change(dentry, iap); |
423 | err = nfserrno(host_err); | |
1da177e4 LT |
424 | fh_unlock(fhp); |
425 | } | |
426 | if (size_change) | |
427 | put_write_access(inode); | |
428 | if (!err) | |
b160fdab | 429 | commit_metadata(fhp); |
1da177e4 LT |
430 | out: |
431 | return err; | |
432 | ||
433 | out_nfserr: | |
6264d69d | 434 | err = nfserrno(host_err); |
1da177e4 LT |
435 | goto out; |
436 | } | |
437 | ||
5be196e5 CH |
438 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
439 | defined(CONFIG_NFSD_V3_ACL) || \ | |
440 | defined(CONFIG_NFSD_V4) | |
441 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
442 | { | |
443 | ssize_t buflen; | |
6c6a426f | 444 | ssize_t ret; |
5be196e5 CH |
445 | |
446 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
447 | if (buflen <= 0) | |
448 | return buflen; | |
1da177e4 | 449 | |
5be196e5 CH |
450 | *buf = kmalloc(buflen, GFP_KERNEL); |
451 | if (!*buf) | |
452 | return -ENOMEM; | |
453 | ||
6c6a426f KK |
454 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
455 | if (ret < 0) | |
456 | kfree(*buf); | |
457 | return ret; | |
5be196e5 CH |
458 | } |
459 | #endif | |
460 | ||
461 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
462 | static int |
463 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
464 | { | |
465 | int len; | |
466 | size_t buflen; | |
467 | char *buf = NULL; | |
468 | int error = 0; | |
1da177e4 LT |
469 | |
470 | buflen = posix_acl_xattr_size(pacl->a_count); | |
471 | buf = kmalloc(buflen, GFP_KERNEL); | |
472 | error = -ENOMEM; | |
473 | if (buf == NULL) | |
474 | goto out; | |
475 | ||
476 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
477 | if (len < 0) { | |
478 | error = len; | |
479 | goto out; | |
480 | } | |
481 | ||
5be196e5 | 482 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
483 | out: |
484 | kfree(buf); | |
485 | return error; | |
486 | } | |
487 | ||
6264d69d | 488 | __be32 |
1da177e4 LT |
489 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
490 | struct nfs4_acl *acl) | |
491 | { | |
6264d69d AV |
492 | __be32 error; |
493 | int host_error; | |
1da177e4 LT |
494 | struct dentry *dentry; |
495 | struct inode *inode; | |
496 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
497 | unsigned int flags = 0; | |
498 | ||
499 | /* Get inode */ | |
8837abca | 500 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 501 | if (error) |
4b2ca38a | 502 | return error; |
1da177e4 LT |
503 | |
504 | dentry = fhp->fh_dentry; | |
505 | inode = dentry->d_inode; | |
506 | if (S_ISDIR(inode->i_mode)) | |
507 | flags = NFS4_ACL_DIR; | |
508 | ||
6264d69d AV |
509 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
510 | if (host_error == -EINVAL) { | |
4b2ca38a | 511 | return nfserr_attrnotsupp; |
6264d69d | 512 | } else if (host_error < 0) |
1da177e4 LT |
513 | goto out_nfserr; |
514 | ||
6264d69d AV |
515 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
516 | if (host_error < 0) | |
4b2ca38a | 517 | goto out_release; |
1da177e4 | 518 | |
4b2ca38a | 519 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 520 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 521 | |
4b2ca38a | 522 | out_release: |
1da177e4 LT |
523 | posix_acl_release(pacl); |
524 | posix_acl_release(dpacl); | |
1da177e4 | 525 | out_nfserr: |
f34f9242 | 526 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 527 | return nfserr_attrnotsupp; |
f34f9242 | 528 | else |
4b2ca38a | 529 | return nfserrno(host_error); |
1da177e4 LT |
530 | } |
531 | ||
532 | static struct posix_acl * | |
533 | _get_posix_acl(struct dentry *dentry, char *key) | |
534 | { | |
5be196e5 | 535 | void *buf = NULL; |
1da177e4 | 536 | struct posix_acl *pacl = NULL; |
5be196e5 | 537 | int buflen; |
1da177e4 | 538 | |
5be196e5 CH |
539 | buflen = nfsd_getxattr(dentry, key, &buf); |
540 | if (!buflen) | |
541 | buflen = -ENODATA; | |
542 | if (buflen <= 0) | |
543 | return ERR_PTR(buflen); | |
1da177e4 LT |
544 | |
545 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
546 | kfree(buf); |
547 | return pacl; | |
1da177e4 LT |
548 | } |
549 | ||
550 | int | |
551 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
552 | { | |
553 | struct inode *inode = dentry->d_inode; | |
554 | int error = 0; | |
555 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
556 | unsigned int flags = 0; | |
557 | ||
9a59f452 | 558 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
559 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
560 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
561 | if (IS_ERR(pacl)) { | |
562 | error = PTR_ERR(pacl); | |
563 | pacl = NULL; | |
564 | goto out; | |
565 | } | |
566 | ||
567 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 568 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
569 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
570 | dpacl = NULL; | |
571 | else if (IS_ERR(dpacl)) { | |
572 | error = PTR_ERR(dpacl); | |
573 | dpacl = NULL; | |
574 | goto out; | |
575 | } | |
576 | flags = NFS4_ACL_DIR; | |
577 | } | |
578 | ||
579 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
580 | if (IS_ERR(*acl)) { | |
581 | error = PTR_ERR(*acl); | |
582 | *acl = NULL; | |
583 | } | |
584 | out: | |
585 | posix_acl_release(pacl); | |
586 | posix_acl_release(dpacl); | |
587 | return error; | |
588 | } | |
589 | ||
6a85d6c7 | 590 | #endif /* defined(CONFIG_NFSD_V4) */ |
1da177e4 LT |
591 | |
592 | #ifdef CONFIG_NFSD_V3 | |
593 | /* | |
594 | * Check server access rights to a file system object | |
595 | */ | |
596 | struct accessmap { | |
597 | u32 access; | |
598 | int how; | |
599 | }; | |
600 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
601 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
602 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
603 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
604 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
605 | |
606 | { 0, 0 } | |
607 | }; | |
608 | ||
609 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
610 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
611 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
612 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
613 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
614 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
615 | |
616 | { 0, 0 } | |
617 | }; | |
618 | ||
619 | static struct accessmap nfs3_anyaccess[] = { | |
620 | /* Some clients - Solaris 2.6 at least, make an access call | |
621 | * to the server to check for access for things like /dev/null | |
622 | * (which really, the server doesn't care about). So | |
623 | * We provide simple access checking for them, looking | |
624 | * mainly at mode bits, and we make sure to ignore read-only | |
625 | * filesystem checks | |
626 | */ | |
8837abca MS |
627 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
628 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
629 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
630 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
631 | |
632 | { 0, 0 } | |
633 | }; | |
634 | ||
6264d69d | 635 | __be32 |
1da177e4 LT |
636 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
637 | { | |
638 | struct accessmap *map; | |
639 | struct svc_export *export; | |
640 | struct dentry *dentry; | |
641 | u32 query, result = 0, sresult = 0; | |
6264d69d | 642 | __be32 error; |
1da177e4 | 643 | |
8837abca | 644 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
645 | if (error) |
646 | goto out; | |
647 | ||
648 | export = fhp->fh_export; | |
649 | dentry = fhp->fh_dentry; | |
650 | ||
651 | if (S_ISREG(dentry->d_inode->i_mode)) | |
652 | map = nfs3_regaccess; | |
653 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
654 | map = nfs3_diraccess; | |
655 | else | |
656 | map = nfs3_anyaccess; | |
657 | ||
658 | ||
659 | query = *access; | |
660 | for (; map->access; map++) { | |
661 | if (map->access & query) { | |
6264d69d | 662 | __be32 err2; |
1da177e4 LT |
663 | |
664 | sresult |= map->access; | |
665 | ||
0ec757df | 666 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
667 | switch (err2) { |
668 | case nfs_ok: | |
669 | result |= map->access; | |
670 | break; | |
671 | ||
672 | /* the following error codes just mean the access was not allowed, | |
673 | * rather than an error occurred */ | |
674 | case nfserr_rofs: | |
675 | case nfserr_acces: | |
676 | case nfserr_perm: | |
677 | /* simply don't "or" in the access bit. */ | |
678 | break; | |
679 | default: | |
680 | error = err2; | |
681 | goto out; | |
682 | } | |
683 | } | |
684 | } | |
685 | *access = result; | |
686 | if (supported) | |
687 | *supported = sresult; | |
688 | ||
689 | out: | |
690 | return error; | |
691 | } | |
692 | #endif /* CONFIG_NFSD_V3 */ | |
693 | ||
694 | ||
695 | ||
696 | /* | |
697 | * Open an existing file or directory. | |
698 | * The access argument indicates the type of open (read/write/lock) | |
699 | * N.B. After this call fhp needs an fh_put | |
700 | */ | |
6264d69d | 701 | __be32 |
1da177e4 LT |
702 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
703 | int access, struct file **filp) | |
704 | { | |
705 | struct dentry *dentry; | |
706 | struct inode *inode; | |
6264d69d AV |
707 | int flags = O_RDONLY|O_LARGEFILE; |
708 | __be32 err; | |
91885258 | 709 | int host_err = 0; |
1da177e4 | 710 | |
e0e81739 DH |
711 | validate_process_creds(); |
712 | ||
1da177e4 LT |
713 | /* |
714 | * If we get here, then the client has already done an "open", | |
715 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
716 | * in case a chmod has now revoked permission. | |
717 | */ | |
8837abca | 718 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
719 | if (err) |
720 | goto out; | |
721 | ||
722 | dentry = fhp->fh_dentry; | |
723 | inode = dentry->d_inode; | |
724 | ||
725 | /* Disallow write access to files with the append-only bit set | |
726 | * or any access when mandatory locking enabled | |
727 | */ | |
728 | err = nfserr_perm; | |
8837abca | 729 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 730 | goto out; |
5e7fc436 BF |
731 | /* |
732 | * We must ignore files (but only files) which might have mandatory | |
733 | * locks on them because there is no way to know if the accesser has | |
734 | * the lock. | |
735 | */ | |
736 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
737 | goto out; |
738 | ||
739 | if (!inode->i_fop) | |
740 | goto out; | |
741 | ||
742 | /* | |
743 | * Check to see if there are any leases on this file. | |
744 | * This may block while leases are broken. | |
745 | */ | |
91885258 JL |
746 | if (!(access & NFSD_MAY_NOT_BREAK_LEASE)) |
747 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? O_WRONLY : 0)); | |
6264d69d | 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 | } |
54775491 | 757 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
033a666c | 758 | flags, current_cred()); |
1da177e4 | 759 | if (IS_ERR(*filp)) |
6264d69d | 760 | host_err = PTR_ERR(*filp); |
aeaa5ccd CE |
761 | else |
762 | host_err = ima_file_check(*filp, access); | |
1da177e4 | 763 | out_nfserr: |
6264d69d | 764 | err = nfserrno(host_err); |
1da177e4 | 765 | out: |
e0e81739 | 766 | validate_process_creds(); |
1da177e4 LT |
767 | return err; |
768 | } | |
769 | ||
770 | /* | |
771 | * Close a file. | |
772 | */ | |
773 | void | |
774 | nfsd_close(struct file *filp) | |
775 | { | |
776 | fput(filp); | |
777 | } | |
778 | ||
1da177e4 LT |
779 | /* |
780 | * Obtain the readahead parameters for the file | |
781 | * specified by (dev, ino). | |
782 | */ | |
1da177e4 LT |
783 | |
784 | static inline struct raparms * | |
785 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
786 | { | |
787 | struct raparms *ra, **rap, **frap = NULL; | |
788 | int depth = 0; | |
fce1456a GB |
789 | unsigned int hash; |
790 | struct raparm_hbucket *rab; | |
791 | ||
792 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
793 | rab = &raparm_hash[hash]; | |
1da177e4 | 794 | |
fce1456a GB |
795 | spin_lock(&rab->pb_lock); |
796 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
797 | if (ra->p_ino == ino && ra->p_dev == dev) |
798 | goto found; | |
799 | depth++; | |
800 | if (ra->p_count == 0) | |
801 | frap = rap; | |
802 | } | |
803 | depth = nfsdstats.ra_size*11/10; | |
804 | if (!frap) { | |
fce1456a | 805 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
806 | return NULL; |
807 | } | |
808 | rap = frap; | |
809 | ra = *frap; | |
810 | ra->p_dev = dev; | |
811 | ra->p_ino = ino; | |
812 | ra->p_set = 0; | |
fce1456a | 813 | ra->p_hindex = hash; |
1da177e4 | 814 | found: |
fce1456a | 815 | if (rap != &rab->pb_head) { |
1da177e4 | 816 | *rap = ra->p_next; |
fce1456a GB |
817 | ra->p_next = rab->pb_head; |
818 | rab->pb_head = ra; | |
1da177e4 LT |
819 | } |
820 | ra->p_count++; | |
821 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 822 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
823 | return ra; |
824 | } | |
825 | ||
826 | /* | |
cf8208d0 JA |
827 | * Grab and keep cached pages associated with a file in the svc_rqst |
828 | * so that they can be passed to the network sendmsg/sendpage routines | |
829 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
830 | */ |
831 | static int | |
cf8208d0 JA |
832 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
833 | struct splice_desc *sd) | |
1da177e4 | 834 | { |
cf8208d0 | 835 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 836 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
837 | struct page *page = buf->page; |
838 | size_t size; | |
839 | int ret; | |
1da177e4 | 840 | |
cac36bb0 | 841 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
842 | if (unlikely(ret)) |
843 | return ret; | |
844 | ||
845 | size = sd->len; | |
1da177e4 LT |
846 | |
847 | if (rqstp->rq_res.page_len == 0) { | |
848 | get_page(page); | |
44524359 N |
849 | put_page(*pp); |
850 | *pp = page; | |
851 | rqstp->rq_resused++; | |
cf8208d0 | 852 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 853 | rqstp->rq_res.page_len = size; |
44524359 | 854 | } else if (page != pp[-1]) { |
1da177e4 | 855 | get_page(page); |
250f3915 N |
856 | if (*pp) |
857 | put_page(*pp); | |
44524359 N |
858 | *pp = page; |
859 | rqstp->rq_resused++; | |
1da177e4 | 860 | rqstp->rq_res.page_len += size; |
44524359 | 861 | } else |
1da177e4 | 862 | rqstp->rq_res.page_len += size; |
1da177e4 | 863 | |
1da177e4 LT |
864 | return size; |
865 | } | |
866 | ||
cf8208d0 JA |
867 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
868 | struct splice_desc *sd) | |
869 | { | |
870 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
871 | } | |
872 | ||
6264d69d | 873 | static __be32 |
1da177e4 LT |
874 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
875 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
876 | { | |
877 | struct inode *inode; | |
1da177e4 | 878 | mm_segment_t oldfs; |
6264d69d AV |
879 | __be32 err; |
880 | int host_err; | |
1da177e4 LT |
881 | |
882 | err = nfserr_perm; | |
7eaa36e2 | 883 | inode = file->f_path.dentry->d_inode; |
a8754bee | 884 | |
cf8208d0 JA |
885 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
886 | struct splice_desc sd = { | |
887 | .len = 0, | |
888 | .total_len = *count, | |
889 | .pos = offset, | |
890 | .u.data = rqstp, | |
891 | }; | |
892 | ||
4fbef206 | 893 | rqstp->rq_resused = 1; |
cf8208d0 | 894 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
895 | } else { |
896 | oldfs = get_fs(); | |
897 | set_fs(KERNEL_DS); | |
6264d69d | 898 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
899 | set_fs(oldfs); |
900 | } | |
901 | ||
6264d69d AV |
902 | if (host_err >= 0) { |
903 | nfsdstats.io_read += host_err; | |
904 | *count = host_err; | |
1da177e4 | 905 | err = 0; |
2a12a9d7 | 906 | fsnotify_access(file); |
1da177e4 | 907 | } else |
6264d69d | 908 | err = nfserrno(host_err); |
1da177e4 LT |
909 | return err; |
910 | } | |
911 | ||
9f708e40 NB |
912 | static void kill_suid(struct dentry *dentry) |
913 | { | |
914 | struct iattr ia; | |
b5376771 | 915 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 916 | |
1b1dcc1b | 917 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 918 | notify_change(dentry, &ia); |
1b1dcc1b | 919 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
920 | } |
921 | ||
d911df7b BF |
922 | /* |
923 | * Gathered writes: If another process is currently writing to the file, | |
924 | * there's a high chance this is another nfsd (triggered by a bulk write | |
925 | * from a client's biod). Rather than syncing the file with each write | |
926 | * request, we sleep for 10 msec. | |
927 | * | |
928 | * I don't know if this roughly approximates C. Juszak's idea of | |
929 | * gathered writes, but it's a nice and simple solution (IMHO), and it | |
930 | * seems to work:-) | |
931 | * | |
932 | * Note: we do this only in the NFSv2 case, since v3 and higher have a | |
933 | * better tool (separate unstable writes and commits) for solving this | |
934 | * problem. | |
935 | */ | |
936 | static int wait_for_concurrent_writes(struct file *file) | |
937 | { | |
938 | struct inode *inode = file->f_path.dentry->d_inode; | |
939 | static ino_t last_ino; | |
940 | static dev_t last_dev; | |
941 | int err = 0; | |
942 | ||
943 | if (atomic_read(&inode->i_writecount) > 1 | |
944 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
945 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); | |
946 | msleep(10); | |
947 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); | |
948 | } | |
949 | ||
950 | if (inode->i_state & I_DIRTY) { | |
951 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); | |
8018ab05 | 952 | err = vfs_fsync(file, 0); |
d911df7b BF |
953 | } |
954 | last_ino = inode->i_ino; | |
955 | last_dev = inode->i_sb->s_dev; | |
956 | return err; | |
957 | } | |
958 | ||
6264d69d | 959 | static __be32 |
1da177e4 LT |
960 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
961 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 962 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
963 | { |
964 | struct svc_export *exp; | |
965 | struct dentry *dentry; | |
966 | struct inode *inode; | |
967 | mm_segment_t oldfs; | |
6264d69d AV |
968 | __be32 err = 0; |
969 | int host_err; | |
1da177e4 | 970 | int stable = *stablep; |
48e03bc5 | 971 | int use_wgather; |
1da177e4 | 972 | |
7eaa36e2 | 973 | dentry = file->f_path.dentry; |
1da177e4 LT |
974 | inode = dentry->d_inode; |
975 | exp = fhp->fh_export; | |
976 | ||
977 | /* | |
978 | * Request sync writes if | |
979 | * - the sync export option has been set, or | |
980 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
981 | * - The file system doesn't support fsync(). | |
48e03bc5 | 982 | * When NFSv2 gathered writes have been configured for this volume, |
1da177e4 LT |
983 | * flushing the data to disk is handled separately below. |
984 | */ | |
48e03bc5 | 985 | use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp); |
1da177e4 | 986 | |
3ba15148 | 987 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
988 | stable = 2; |
989 | *stablep = 2; /* FILE_SYNC */ | |
990 | } | |
991 | ||
992 | if (!EX_ISSYNC(exp)) | |
993 | stable = 0; | |
48e03bc5 | 994 | if (stable && !use_wgather) { |
db1dd4d3 | 995 | spin_lock(&file->f_lock); |
1da177e4 | 996 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
997 | spin_unlock(&file->f_lock); |
998 | } | |
1da177e4 LT |
999 | |
1000 | /* Write the data. */ | |
1001 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1002 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1003 | set_fs(oldfs); |
e4636d53 BF |
1004 | if (host_err < 0) |
1005 | goto out_nfserr; | |
1006 | *cnt = host_err; | |
1007 | nfsdstats.io_write += host_err; | |
2a12a9d7 | 1008 | fsnotify_modify(file); |
1da177e4 LT |
1009 | |
1010 | /* clear setuid/setgid flag after write */ | |
e4636d53 | 1011 | if (inode->i_mode & (S_ISUID | S_ISGID)) |
9f708e40 | 1012 | kill_suid(dentry); |
1da177e4 | 1013 | |
e4636d53 | 1014 | if (stable && use_wgather) |
d911df7b | 1015 | host_err = wait_for_concurrent_writes(file); |
1da177e4 | 1016 | |
e4636d53 | 1017 | out_nfserr: |
6264d69d | 1018 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1019 | if (host_err >= 0) |
1da177e4 | 1020 | err = 0; |
a0d24b29 | 1021 | else |
6264d69d | 1022 | err = nfserrno(host_err); |
1da177e4 LT |
1023 | return err; |
1024 | } | |
1025 | ||
1026 | /* | |
1027 | * Read data from a file. count must contain the requested read count | |
1028 | * on entry. On return, *count contains the number of bytes actually read. | |
1029 | * N.B. After this call fhp needs an fh_put | |
1030 | */ | |
039a87ca | 1031 | __be32 nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, |
fa0a2126 BF |
1032 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) |
1033 | { | |
1034 | struct file *file; | |
1035 | struct inode *inode; | |
1036 | struct raparms *ra; | |
1037 | __be32 err; | |
1038 | ||
1039 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); | |
1040 | if (err) | |
1041 | return err; | |
1042 | ||
1043 | inode = file->f_path.dentry->d_inode; | |
1044 | ||
1045 | /* Get readahead parameters */ | |
1046 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
1047 | ||
1048 | if (ra && ra->p_set) | |
1049 | file->f_ra = ra->p_ra; | |
1050 | ||
1051 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1052 | ||
1053 | /* Write back readahead params */ | |
1054 | if (ra) { | |
1055 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; | |
1056 | spin_lock(&rab->pb_lock); | |
1057 | ra->p_ra = file->f_ra; | |
1058 | ra->p_set = 1; | |
1059 | ra->p_count--; | |
1060 | spin_unlock(&rab->pb_lock); | |
1061 | } | |
1062 | ||
1063 | nfsd_close(file); | |
1064 | return err; | |
1065 | } | |
1066 | ||
039a87ca | 1067 | /* As above, but use the provided file descriptor. */ |
6264d69d | 1068 | __be32 |
039a87ca | 1069 | nfsd_read_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1da177e4 LT |
1070 | loff_t offset, struct kvec *vec, int vlen, |
1071 | unsigned long *count) | |
1072 | { | |
6264d69d | 1073 | __be32 err; |
1da177e4 LT |
1074 | |
1075 | if (file) { | |
0ec757df | 1076 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1077 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1078 | if (err) |
1079 | goto out; | |
1080 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
039a87ca BF |
1081 | } else /* Note file may still be NULL in NFSv4 special stateid case: */ |
1082 | err = nfsd_read(rqstp, fhp, offset, vec, vlen, count); | |
1da177e4 LT |
1083 | out: |
1084 | return err; | |
1085 | } | |
1086 | ||
1087 | /* | |
1088 | * Write data to a file. | |
1089 | * The stable flag requests synchronous writes. | |
1090 | * N.B. After this call fhp needs an fh_put | |
1091 | */ | |
6264d69d | 1092 | __be32 |
1da177e4 | 1093 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1094 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1095 | int *stablep) |
1096 | { | |
6264d69d | 1097 | __be32 err = 0; |
1da177e4 LT |
1098 | |
1099 | if (file) { | |
0ec757df | 1100 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1101 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1102 | if (err) |
1103 | goto out; | |
1104 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1105 | stablep); | |
1106 | } else { | |
8837abca | 1107 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1108 | if (err) |
1109 | goto out; | |
1110 | ||
1111 | if (cnt) | |
1112 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1113 | cnt, stablep); | |
1114 | nfsd_close(file); | |
1115 | } | |
1116 | out: | |
1117 | return err; | |
1118 | } | |
1119 | ||
1120 | #ifdef CONFIG_NFSD_V3 | |
1121 | /* | |
1122 | * Commit all pending writes to stable storage. | |
aa696a6f TM |
1123 | * |
1124 | * Note: we only guarantee that data that lies within the range specified | |
1125 | * by the 'offset' and 'count' parameters will be synced. | |
1da177e4 LT |
1126 | * |
1127 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1128 | * data to the client, as locking happens lower down in the filesystem. | |
1129 | */ | |
6264d69d | 1130 | __be32 |
1da177e4 LT |
1131 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1132 | loff_t offset, unsigned long count) | |
1133 | { | |
1134 | struct file *file; | |
aa696a6f TM |
1135 | loff_t end = LLONG_MAX; |
1136 | __be32 err = nfserr_inval; | |
1da177e4 | 1137 | |
aa696a6f TM |
1138 | if (offset < 0) |
1139 | goto out; | |
1140 | if (count != 0) { | |
1141 | end = offset + (loff_t)count - 1; | |
1142 | if (end < offset) | |
1143 | goto out; | |
1144 | } | |
1da177e4 | 1145 | |
91885258 JL |
1146 | err = nfsd_open(rqstp, fhp, S_IFREG, |
1147 | NFSD_MAY_WRITE|NFSD_MAY_NOT_BREAK_LEASE, &file); | |
8837abca | 1148 | if (err) |
aa696a6f | 1149 | goto out; |
1da177e4 | 1150 | if (EX_ISSYNC(fhp->fh_export)) { |
8018ab05 | 1151 | int err2 = vfs_fsync_range(file, offset, end, 0); |
aa696a6f TM |
1152 | |
1153 | if (err2 != -EINVAL) | |
1154 | err = nfserrno(err2); | |
1155 | else | |
1da177e4 | 1156 | err = nfserr_notsupp; |
1da177e4 LT |
1157 | } |
1158 | ||
1159 | nfsd_close(file); | |
aa696a6f | 1160 | out: |
1da177e4 LT |
1161 | return err; |
1162 | } | |
1163 | #endif /* CONFIG_NFSD_V3 */ | |
1164 | ||
f2b0dee2 | 1165 | static __be32 |
5c002b3b BF |
1166 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1167 | struct iattr *iap) | |
1168 | { | |
1169 | /* | |
1170 | * Mode has already been set earlier in create: | |
1171 | */ | |
1172 | iap->ia_valid &= ~ATTR_MODE; | |
1173 | /* | |
1174 | * Setting uid/gid works only for root. Irix appears to | |
1175 | * send along the gid on create when it tries to implement | |
1176 | * setgid directories via NFS: | |
1177 | */ | |
5cc0a840 | 1178 | if (current_fsuid() != 0) |
5c002b3b BF |
1179 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1180 | if (iap->ia_valid) | |
1181 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1182 | return 0; | |
1183 | } | |
1184 | ||
4ac35c2f | 1185 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1186 | * setting size to 0 may fail for some specific file systems by the permission | |
1187 | * checking which requires WRITE permission but the mode is 000. | |
1188 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1189 | * 0 after file created. | |
1190 | * | |
1191 | * call this only after vfs_create() is called. | |
1192 | * */ | |
1193 | static void | |
1194 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1195 | { | |
1196 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1197 | iap->ia_valid &= ~ATTR_SIZE; | |
1198 | } | |
1199 | ||
1da177e4 LT |
1200 | /* |
1201 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1202 | * not yet implemented. | |
1203 | * If the response fh has been verified, the parent directory should | |
1204 | * already be locked. Note that the parent directory is left locked. | |
1205 | * | |
1206 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1207 | */ | |
6264d69d | 1208 | __be32 |
1da177e4 LT |
1209 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1210 | char *fname, int flen, struct iattr *iap, | |
1211 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1212 | { | |
1213 | struct dentry *dentry, *dchild = NULL; | |
1214 | struct inode *dirp; | |
6264d69d | 1215 | __be32 err; |
5c002b3b | 1216 | __be32 err2; |
6264d69d | 1217 | int host_err; |
1da177e4 LT |
1218 | |
1219 | err = nfserr_perm; | |
1220 | if (!flen) | |
1221 | goto out; | |
1222 | err = nfserr_exist; | |
1223 | if (isdotent(fname, flen)) | |
1224 | goto out; | |
1225 | ||
8837abca | 1226 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1227 | if (err) |
1228 | goto out; | |
1229 | ||
1230 | dentry = fhp->fh_dentry; | |
1231 | dirp = dentry->d_inode; | |
1232 | ||
1233 | err = nfserr_notdir; | |
acfa4380 | 1234 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1235 | goto out; |
1236 | /* | |
1237 | * Check whether the response file handle has been verified yet. | |
1238 | * If it has, the parent directory should already be locked. | |
1239 | */ | |
1240 | if (!resfhp->fh_dentry) { | |
1241 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1242 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1243 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1244 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1245 | if (IS_ERR(dchild)) |
1246 | goto out_nfserr; | |
1247 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1248 | if (err) | |
1249 | goto out; | |
1250 | } else { | |
1251 | /* called from nfsd_proc_create */ | |
1252 | dchild = dget(resfhp->fh_dentry); | |
1253 | if (!fhp->fh_locked) { | |
1254 | /* not actually possible */ | |
1255 | printk(KERN_ERR | |
1256 | "nfsd_create: parent %s/%s not locked!\n", | |
1257 | dentry->d_parent->d_name.name, | |
1258 | dentry->d_name.name); | |
d75f2b9f | 1259 | err = nfserr_io; |
1da177e4 LT |
1260 | goto out; |
1261 | } | |
1262 | } | |
1263 | /* | |
1264 | * Make sure the child dentry is still negative ... | |
1265 | */ | |
1266 | err = nfserr_exist; | |
1267 | if (dchild->d_inode) { | |
1268 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1269 | dentry->d_name.name, dchild->d_name.name); | |
1270 | goto out; | |
1271 | } | |
1272 | ||
1273 | if (!(iap->ia_valid & ATTR_MODE)) | |
1274 | iap->ia_mode = 0; | |
1275 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1276 | ||
07cad1d2 MS |
1277 | err = nfserr_inval; |
1278 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1279 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1280 | type); | |
1281 | goto out; | |
1282 | } | |
1283 | ||
1284 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1285 | if (host_err) | |
1286 | goto out_nfserr; | |
1287 | ||
1da177e4 LT |
1288 | /* |
1289 | * Get the dir op function pointer. | |
1290 | */ | |
088406bc | 1291 | err = 0; |
1da177e4 LT |
1292 | switch (type) { |
1293 | case S_IFREG: | |
6264d69d | 1294 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1295 | if (!host_err) |
1296 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1297 | break; |
1298 | case S_IFDIR: | |
6264d69d | 1299 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1300 | break; |
1301 | case S_IFCHR: | |
1302 | case S_IFBLK: | |
1303 | case S_IFIFO: | |
1304 | case S_IFSOCK: | |
6264d69d | 1305 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1306 | break; |
1da177e4 | 1307 | } |
463c3197 DH |
1308 | if (host_err < 0) { |
1309 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1310 | goto out_nfserr; |
463c3197 | 1311 | } |
1da177e4 | 1312 | |
f501912a | 1313 | err = nfsd_create_setattr(rqstp, resfhp, iap); |
1da177e4 | 1314 | |
f501912a BM |
1315 | /* |
1316 | * nfsd_setattr already committed the child. Transactional filesystems | |
1317 | * had a chance to commit changes for both parent and child | |
1318 | * simultaneously making the following commit_metadata a noop. | |
1319 | */ | |
1320 | err2 = nfserrno(commit_metadata(fhp)); | |
5c002b3b BF |
1321 | if (err2) |
1322 | err = err2; | |
463c3197 | 1323 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1324 | /* |
1325 | * Update the file handle to get the new inode info. | |
1326 | */ | |
1327 | if (!err) | |
1328 | err = fh_update(resfhp); | |
1329 | out: | |
1330 | if (dchild && !IS_ERR(dchild)) | |
1331 | dput(dchild); | |
1332 | return err; | |
1333 | ||
1334 | out_nfserr: | |
6264d69d | 1335 | err = nfserrno(host_err); |
1da177e4 LT |
1336 | goto out; |
1337 | } | |
1338 | ||
1339 | #ifdef CONFIG_NFSD_V3 | |
1340 | /* | |
1341 | * NFSv3 version of nfsd_create | |
1342 | */ | |
6264d69d | 1343 | __be32 |
1da177e4 LT |
1344 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1345 | char *fname, int flen, struct iattr *iap, | |
1346 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1347 | int *truncp, int *created) |
1da177e4 LT |
1348 | { |
1349 | struct dentry *dentry, *dchild = NULL; | |
1350 | struct inode *dirp; | |
6264d69d AV |
1351 | __be32 err; |
1352 | int host_err; | |
1da177e4 | 1353 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1354 | |
1355 | err = nfserr_perm; | |
1356 | if (!flen) | |
1357 | goto out; | |
1358 | err = nfserr_exist; | |
1359 | if (isdotent(fname, flen)) | |
1360 | goto out; | |
1361 | if (!(iap->ia_valid & ATTR_MODE)) | |
1362 | iap->ia_mode = 0; | |
8837abca | 1363 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1364 | if (err) |
1365 | goto out; | |
1366 | ||
1367 | dentry = fhp->fh_dentry; | |
1368 | dirp = dentry->d_inode; | |
1369 | ||
1370 | /* Get all the sanity checks out of the way before | |
1371 | * we lock the parent. */ | |
1372 | err = nfserr_notdir; | |
acfa4380 | 1373 | if (!dirp->i_op->lookup) |
1da177e4 | 1374 | goto out; |
12fd3520 | 1375 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1376 | |
1377 | /* | |
1378 | * Compose the response file handle. | |
1379 | */ | |
1380 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1381 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1382 | if (IS_ERR(dchild)) |
1383 | goto out_nfserr; | |
1384 | ||
1385 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1386 | if (err) | |
1387 | goto out; | |
1388 | ||
1389 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1390 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1391 | * the high bit set, so just clear the high bits. If this is |
1392 | * ever changed to use different attrs for storing the | |
1393 | * verifier, then do_open_lookup() will also need to be fixed | |
1394 | * accordingly. | |
1da177e4 LT |
1395 | */ |
1396 | v_mtime = verifier[0]&0x7fffffff; | |
1397 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1398 | } |
1399 | ||
463c3197 DH |
1400 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1401 | if (host_err) | |
1402 | goto out_nfserr; | |
1da177e4 LT |
1403 | if (dchild->d_inode) { |
1404 | err = 0; | |
1405 | ||
1406 | switch (createmode) { | |
1407 | case NFS3_CREATE_UNCHECKED: | |
1408 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1409 | err = nfserr_exist; | |
1410 | else if (truncp) { | |
1411 | /* in nfsv4, we need to treat this case a little | |
1412 | * differently. we don't want to truncate the | |
1413 | * file now; this would be wrong if the OPEN | |
1414 | * fails for some other reason. furthermore, | |
1415 | * if the size is nonzero, we should ignore it | |
1416 | * according to spec! | |
1417 | */ | |
1418 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1419 | } | |
1420 | else { | |
1421 | iap->ia_valid &= ATTR_SIZE; | |
1422 | goto set_attr; | |
1423 | } | |
1424 | break; | |
1425 | case NFS3_CREATE_EXCLUSIVE: | |
1426 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1427 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1428 | && dchild->d_inode->i_size == 0 ) |
1429 | break; | |
1430 | /* fallthru */ | |
1431 | case NFS3_CREATE_GUARDED: | |
1432 | err = nfserr_exist; | |
1433 | } | |
463c3197 | 1434 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1435 | goto out; |
1436 | } | |
1437 | ||
6264d69d | 1438 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1439 | if (host_err < 0) { |
1440 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1441 | goto out_nfserr; |
463c3197 | 1442 | } |
81ac95c5 BF |
1443 | if (created) |
1444 | *created = 1; | |
1da177e4 | 1445 | |
4ac35c2f | 1446 | nfsd_check_ignore_resizing(iap); |
1447 | ||
1da177e4 | 1448 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1449 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1450 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1451 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1452 | /* XXX someone who knows this better please fix it for nsec */ |
1453 | iap->ia_mtime.tv_sec = v_mtime; | |
1454 | iap->ia_atime.tv_sec = v_atime; | |
1455 | iap->ia_mtime.tv_nsec = 0; | |
1456 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1457 | } |
1458 | ||
1da177e4 | 1459 | set_attr: |
f501912a BM |
1460 | err = nfsd_create_setattr(rqstp, resfhp, iap); |
1461 | ||
1462 | /* | |
1463 | * nfsd_setattr already committed the child (and possibly also the parent). | |
1464 | */ | |
1465 | if (!err) | |
1466 | err = nfserrno(commit_metadata(fhp)); | |
f193fbab | 1467 | |
463c3197 | 1468 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1469 | /* |
1470 | * Update the filehandle to get the new inode info. | |
1471 | */ | |
1472 | if (!err) | |
1473 | err = fh_update(resfhp); | |
1da177e4 LT |
1474 | |
1475 | out: | |
1476 | fh_unlock(fhp); | |
1477 | if (dchild && !IS_ERR(dchild)) | |
1478 | dput(dchild); | |
1479 | return err; | |
1480 | ||
1481 | out_nfserr: | |
6264d69d | 1482 | err = nfserrno(host_err); |
1da177e4 LT |
1483 | goto out; |
1484 | } | |
1485 | #endif /* CONFIG_NFSD_V3 */ | |
1486 | ||
1487 | /* | |
1488 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1489 | * fits into the buffer. On return, it contains the true length. | |
1490 | * N.B. After this call fhp needs an fh_put | |
1491 | */ | |
6264d69d | 1492 | __be32 |
1da177e4 LT |
1493 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1494 | { | |
1495 | struct dentry *dentry; | |
1496 | struct inode *inode; | |
1497 | mm_segment_t oldfs; | |
6264d69d AV |
1498 | __be32 err; |
1499 | int host_err; | |
1da177e4 | 1500 | |
8837abca | 1501 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1502 | if (err) |
1503 | goto out; | |
1504 | ||
1505 | dentry = fhp->fh_dentry; | |
1506 | inode = dentry->d_inode; | |
1507 | ||
1508 | err = nfserr_inval; | |
acfa4380 | 1509 | if (!inode->i_op->readlink) |
1da177e4 LT |
1510 | goto out; |
1511 | ||
54775491 | 1512 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1513 | /* N.B. Why does this call need a get_fs()?? |
1514 | * Remove the set_fs and watch the fireworks:-) --okir | |
1515 | */ | |
1516 | ||
1517 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1518 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1519 | set_fs(oldfs); |
1520 | ||
6264d69d | 1521 | if (host_err < 0) |
1da177e4 | 1522 | goto out_nfserr; |
6264d69d | 1523 | *lenp = host_err; |
1da177e4 LT |
1524 | err = 0; |
1525 | out: | |
1526 | return err; | |
1527 | ||
1528 | out_nfserr: | |
6264d69d | 1529 | err = nfserrno(host_err); |
1da177e4 LT |
1530 | goto out; |
1531 | } | |
1532 | ||
1533 | /* | |
1534 | * Create a symlink and look up its inode | |
1535 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1536 | */ | |
6264d69d | 1537 | __be32 |
1da177e4 LT |
1538 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1539 | char *fname, int flen, | |
1540 | char *path, int plen, | |
1541 | struct svc_fh *resfhp, | |
1542 | struct iattr *iap) | |
1543 | { | |
1544 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1545 | __be32 err, cerr; |
1546 | int host_err; | |
1da177e4 LT |
1547 | |
1548 | err = nfserr_noent; | |
1549 | if (!flen || !plen) | |
1550 | goto out; | |
1551 | err = nfserr_exist; | |
1552 | if (isdotent(fname, flen)) | |
1553 | goto out; | |
1554 | ||
8837abca | 1555 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1556 | if (err) |
1557 | goto out; | |
1558 | fh_lock(fhp); | |
1559 | dentry = fhp->fh_dentry; | |
1560 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1561 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1562 | if (IS_ERR(dnew)) |
1563 | goto out_nfserr; | |
1564 | ||
75c3f29d DH |
1565 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1566 | if (host_err) | |
1567 | goto out_nfserr; | |
1568 | ||
1da177e4 LT |
1569 | if (unlikely(path[plen] != 0)) { |
1570 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1571 | if (path_alloced == NULL) | |
6264d69d | 1572 | host_err = -ENOMEM; |
1da177e4 LT |
1573 | else { |
1574 | strncpy(path_alloced, path, plen); | |
1575 | path_alloced[plen] = 0; | |
db2e747b | 1576 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1577 | kfree(path_alloced); |
1578 | } | |
1579 | } else | |
db2e747b | 1580 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
6264d69d | 1581 | err = nfserrno(host_err); |
f501912a BM |
1582 | if (!err) |
1583 | err = nfserrno(commit_metadata(fhp)); | |
1da177e4 LT |
1584 | fh_unlock(fhp); |
1585 | ||
75c3f29d DH |
1586 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1587 | ||
1da177e4 LT |
1588 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1589 | dput(dnew); | |
1590 | if (err==0) err = cerr; | |
1591 | out: | |
1592 | return err; | |
1593 | ||
1594 | out_nfserr: | |
6264d69d | 1595 | err = nfserrno(host_err); |
1da177e4 LT |
1596 | goto out; |
1597 | } | |
1598 | ||
1599 | /* | |
1600 | * Create a hardlink | |
1601 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1602 | */ | |
6264d69d | 1603 | __be32 |
1da177e4 LT |
1604 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1605 | char *name, int len, struct svc_fh *tfhp) | |
1606 | { | |
1607 | struct dentry *ddir, *dnew, *dold; | |
55b13354 | 1608 | struct inode *dirp; |
6264d69d AV |
1609 | __be32 err; |
1610 | int host_err; | |
1da177e4 | 1611 | |
8837abca | 1612 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1613 | if (err) |
1614 | goto out; | |
8837abca | 1615 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1616 | if (err) |
1617 | goto out; | |
1618 | ||
1619 | err = nfserr_perm; | |
1620 | if (!len) | |
1621 | goto out; | |
1622 | err = nfserr_exist; | |
1623 | if (isdotent(name, len)) | |
1624 | goto out; | |
1625 | ||
12fd3520 | 1626 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1627 | ddir = ffhp->fh_dentry; |
1628 | dirp = ddir->d_inode; | |
1629 | ||
1630 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1631 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1632 | if (IS_ERR(dnew)) |
1633 | goto out_nfserr; | |
1634 | ||
1635 | dold = tfhp->fh_dentry; | |
1da177e4 | 1636 | |
75c3f29d DH |
1637 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1638 | if (host_err) { | |
1639 | err = nfserrno(host_err); | |
1640 | goto out_dput; | |
1641 | } | |
6264d69d AV |
1642 | host_err = vfs_link(dold, dirp, dnew); |
1643 | if (!host_err) { | |
f501912a BM |
1644 | err = nfserrno(commit_metadata(ffhp)); |
1645 | if (!err) | |
1646 | err = nfserrno(commit_metadata(tfhp)); | |
1da177e4 | 1647 | } else { |
6264d69d | 1648 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1649 | err = nfserr_acces; |
1650 | else | |
6264d69d | 1651 | err = nfserrno(host_err); |
1da177e4 | 1652 | } |
75c3f29d DH |
1653 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1654 | out_dput: | |
1da177e4 | 1655 | dput(dnew); |
270d56e5 DR |
1656 | out_unlock: |
1657 | fh_unlock(ffhp); | |
1da177e4 LT |
1658 | out: |
1659 | return err; | |
1660 | ||
1661 | out_nfserr: | |
6264d69d | 1662 | err = nfserrno(host_err); |
270d56e5 | 1663 | goto out_unlock; |
1da177e4 LT |
1664 | } |
1665 | ||
1666 | /* | |
1667 | * Rename a file | |
1668 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1669 | */ | |
6264d69d | 1670 | __be32 |
1da177e4 LT |
1671 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1672 | struct svc_fh *tfhp, char *tname, int tlen) | |
1673 | { | |
1674 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1675 | struct inode *fdir, *tdir; | |
6264d69d AV |
1676 | __be32 err; |
1677 | int host_err; | |
1da177e4 | 1678 | |
8837abca | 1679 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1680 | if (err) |
1681 | goto out; | |
8837abca | 1682 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1683 | if (err) |
1684 | goto out; | |
1685 | ||
1686 | fdentry = ffhp->fh_dentry; | |
1687 | fdir = fdentry->d_inode; | |
1688 | ||
1689 | tdentry = tfhp->fh_dentry; | |
1690 | tdir = tdentry->d_inode; | |
1691 | ||
1692 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1693 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1694 | goto out; |
1695 | ||
1696 | err = nfserr_perm; | |
1697 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1698 | goto out; | |
1699 | ||
1700 | /* cannot use fh_lock as we need deadlock protective ordering | |
1701 | * so do it by hand */ | |
1702 | trap = lock_rename(tdentry, fdentry); | |
1703 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1704 | fill_pre_wcc(ffhp); | |
1705 | fill_pre_wcc(tfhp); | |
1706 | ||
1707 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1708 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1709 | if (IS_ERR(odentry)) |
1710 | goto out_nfserr; | |
1711 | ||
6264d69d | 1712 | host_err = -ENOENT; |
1da177e4 LT |
1713 | if (!odentry->d_inode) |
1714 | goto out_dput_old; | |
6264d69d | 1715 | host_err = -EINVAL; |
1da177e4 LT |
1716 | if (odentry == trap) |
1717 | goto out_dput_old; | |
1718 | ||
1719 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1720 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1721 | if (IS_ERR(ndentry)) |
1722 | goto out_dput_old; | |
6264d69d | 1723 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1724 | if (ndentry == trap) |
1725 | goto out_dput_new; | |
1726 | ||
9079b1eb DH |
1727 | host_err = -EXDEV; |
1728 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1729 | goto out_dput_new; | |
1730 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1731 | if (host_err) | |
1732 | goto out_dput_new; | |
1733 | ||
6264d69d | 1734 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
f501912a BM |
1735 | if (!host_err) { |
1736 | host_err = commit_metadata(tfhp); | |
6264d69d | 1737 | if (!host_err) |
f501912a | 1738 | host_err = commit_metadata(ffhp); |
1da177e4 LT |
1739 | } |
1740 | ||
9079b1eb DH |
1741 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1742 | ||
1da177e4 LT |
1743 | out_dput_new: |
1744 | dput(ndentry); | |
1745 | out_dput_old: | |
1746 | dput(odentry); | |
1747 | out_nfserr: | |
6264d69d | 1748 | err = nfserrno(host_err); |
1da177e4 LT |
1749 | |
1750 | /* we cannot reply on fh_unlock on the two filehandles, | |
1751 | * as that would do the wrong thing if the two directories | |
1752 | * were the same, so again we do it by hand | |
1753 | */ | |
1754 | fill_post_wcc(ffhp); | |
1755 | fill_post_wcc(tfhp); | |
1756 | unlock_rename(tdentry, fdentry); | |
1757 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1758 | ||
1759 | out: | |
1760 | return err; | |
1761 | } | |
1762 | ||
1763 | /* | |
1764 | * Unlink a file or directory | |
1765 | * N.B. After this call fhp needs an fh_put | |
1766 | */ | |
6264d69d | 1767 | __be32 |
1da177e4 LT |
1768 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1769 | char *fname, int flen) | |
1770 | { | |
1771 | struct dentry *dentry, *rdentry; | |
1772 | struct inode *dirp; | |
6264d69d AV |
1773 | __be32 err; |
1774 | int host_err; | |
1da177e4 LT |
1775 | |
1776 | err = nfserr_acces; | |
1777 | if (!flen || isdotent(fname, flen)) | |
1778 | goto out; | |
8837abca | 1779 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1780 | if (err) |
1781 | goto out; | |
1782 | ||
12fd3520 | 1783 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1784 | dentry = fhp->fh_dentry; |
1785 | dirp = dentry->d_inode; | |
1786 | ||
1787 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1788 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1789 | if (IS_ERR(rdentry)) |
1790 | goto out_nfserr; | |
1791 | ||
1792 | if (!rdentry->d_inode) { | |
1793 | dput(rdentry); | |
1794 | err = nfserr_noent; | |
1795 | goto out; | |
1796 | } | |
1797 | ||
1798 | if (!type) | |
1799 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1800 | ||
0622753b DH |
1801 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1802 | if (host_err) | |
1803 | goto out_nfserr; | |
1804 | ||
9ce137ee | 1805 | if (type != S_IFDIR) |
6264d69d | 1806 | host_err = vfs_unlink(dirp, rdentry); |
9ce137ee | 1807 | else |
6264d69d | 1808 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1809 | |
1810 | dput(rdentry); | |
1811 | ||
f501912a BM |
1812 | if (!host_err) |
1813 | host_err = commit_metadata(fhp); | |
1da177e4 | 1814 | |
0622753b | 1815 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 | 1816 | out_nfserr: |
6264d69d | 1817 | err = nfserrno(host_err); |
f193fbab YT |
1818 | out: |
1819 | return err; | |
1da177e4 LT |
1820 | } |
1821 | ||
14f7dd63 DW |
1822 | /* |
1823 | * We do this buffering because we must not call back into the file | |
1824 | * system's ->lookup() method from the filldir callback. That may well | |
1825 | * deadlock a number of file systems. | |
1826 | * | |
1827 | * This is based heavily on the implementation of same in XFS. | |
1828 | */ | |
1829 | struct buffered_dirent { | |
1830 | u64 ino; | |
1831 | loff_t offset; | |
1832 | int namlen; | |
1833 | unsigned int d_type; | |
1834 | char name[]; | |
1835 | }; | |
1836 | ||
1837 | struct readdir_data { | |
1838 | char *dirent; | |
1839 | size_t used; | |
53c9c5c0 | 1840 | int full; |
14f7dd63 DW |
1841 | }; |
1842 | ||
1843 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1844 | loff_t offset, u64 ino, unsigned int d_type) | |
1845 | { | |
1846 | struct readdir_data *buf = __buf; | |
1847 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1848 | unsigned int reclen; | |
1849 | ||
1850 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1851 | if (buf->used + reclen > PAGE_SIZE) { |
1852 | buf->full = 1; | |
14f7dd63 | 1853 | return -EINVAL; |
53c9c5c0 | 1854 | } |
14f7dd63 DW |
1855 | |
1856 | de->namlen = namlen; | |
1857 | de->offset = offset; | |
1858 | de->ino = ino; | |
1859 | de->d_type = d_type; | |
1860 | memcpy(de->name, name, namlen); | |
1861 | buf->used += reclen; | |
1862 | ||
1863 | return 0; | |
1864 | } | |
1865 | ||
2f9092e1 DW |
1866 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
1867 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1868 | { |
14f7dd63 DW |
1869 | struct readdir_data buf; |
1870 | struct buffered_dirent *de; | |
2628b766 | 1871 | int host_err; |
14f7dd63 DW |
1872 | int size; |
1873 | loff_t offset; | |
2628b766 | 1874 | |
14f7dd63 DW |
1875 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1876 | if (!buf.dirent) | |
2f9092e1 | 1877 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
1878 | |
1879 | offset = *offsetp; | |
2628b766 | 1880 | |
14f7dd63 | 1881 | while (1) { |
2f9092e1 | 1882 | struct inode *dir_inode = file->f_path.dentry->d_inode; |
14f7dd63 DW |
1883 | unsigned int reclen; |
1884 | ||
b726e923 | 1885 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1886 | buf.used = 0; |
53c9c5c0 | 1887 | buf.full = 0; |
14f7dd63 DW |
1888 | |
1889 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1890 | if (buf.full) |
1891 | host_err = 0; | |
1892 | ||
1893 | if (host_err < 0) | |
14f7dd63 DW |
1894 | break; |
1895 | ||
1896 | size = buf.used; | |
1897 | ||
1898 | if (!size) | |
1899 | break; | |
1900 | ||
2f9092e1 DW |
1901 | /* |
1902 | * Various filldir functions may end up calling back into | |
1903 | * lookup_one_len() and the file system's ->lookup() method. | |
1904 | * These expect i_mutex to be held, as it would within readdir. | |
1905 | */ | |
1906 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
1907 | if (host_err) | |
1908 | break; | |
1909 | ||
14f7dd63 DW |
1910 | de = (struct buffered_dirent *)buf.dirent; |
1911 | while (size > 0) { | |
1912 | offset = de->offset; | |
1913 | ||
1914 | if (func(cdp, de->name, de->namlen, de->offset, | |
1915 | de->ino, de->d_type)) | |
2f9092e1 | 1916 | break; |
14f7dd63 DW |
1917 | |
1918 | if (cdp->err != nfs_ok) | |
2f9092e1 | 1919 | break; |
14f7dd63 DW |
1920 | |
1921 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1922 | sizeof(u64)); | |
1923 | size -= reclen; | |
1924 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1925 | } | |
2f9092e1 DW |
1926 | mutex_unlock(&dir_inode->i_mutex); |
1927 | if (size > 0) /* We bailed out early */ | |
1928 | break; | |
1929 | ||
c002a6c7 | 1930 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1931 | } |
1932 | ||
14f7dd63 | 1933 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
1934 | |
1935 | if (host_err) | |
1936 | return nfserrno(host_err); | |
14f7dd63 DW |
1937 | |
1938 | *offsetp = offset; | |
1939 | return cdp->err; | |
2628b766 DW |
1940 | } |
1941 | ||
1da177e4 LT |
1942 | /* |
1943 | * Read entries from a directory. | |
1944 | * The NFSv3/4 verifier we ignore for now. | |
1945 | */ | |
6264d69d | 1946 | __be32 |
1da177e4 | 1947 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1948 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1949 | { |
6264d69d | 1950 | __be32 err; |
1da177e4 LT |
1951 | struct file *file; |
1952 | loff_t offset = *offsetp; | |
1953 | ||
8837abca | 1954 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
1955 | if (err) |
1956 | goto out; | |
1957 | ||
1958 | offset = vfs_llseek(file, offset, 0); | |
1959 | if (offset < 0) { | |
1960 | err = nfserrno((int)offset); | |
1961 | goto out_close; | |
1962 | } | |
1963 | ||
14f7dd63 | 1964 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
1965 | |
1966 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1967 | err = nfs_ok; /* can still be found in ->err */ | |
1968 | out_close: | |
1969 | nfsd_close(file); | |
1970 | out: | |
1971 | return err; | |
1972 | } | |
1973 | ||
1974 | /* | |
1975 | * Get file system stats | |
1976 | * N.B. After this call fhp needs an fh_put | |
1977 | */ | |
6264d69d | 1978 | __be32 |
04716e66 | 1979 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 1980 | { |
ebabe9a9 CH |
1981 | __be32 err; |
1982 | ||
1983 | err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); | |
f6360efb TI |
1984 | if (!err) { |
1985 | struct path path = { | |
1986 | .mnt = fhp->fh_export->ex_path.mnt, | |
1987 | .dentry = fhp->fh_dentry, | |
1988 | }; | |
1989 | if (vfs_statfs(&path, stat)) | |
1990 | err = nfserr_io; | |
1991 | } | |
1da177e4 LT |
1992 | return err; |
1993 | } | |
1994 | ||
c7d51402 | 1995 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 1996 | { |
c7d51402 | 1997 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
1998 | } |
1999 | ||
1da177e4 LT |
2000 | /* |
2001 | * Check for a user's access permissions to this inode. | |
2002 | */ | |
6264d69d | 2003 | __be32 |
0ec757df BF |
2004 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2005 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2006 | { |
2007 | struct inode *inode = dentry->d_inode; | |
2008 | int err; | |
2009 | ||
8837abca | 2010 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2011 | return 0; |
2012 | #if 0 | |
2013 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2014 | acc, | |
8837abca MS |
2015 | (acc & NFSD_MAY_READ)? " read" : "", |
2016 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2017 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2018 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2019 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2020 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2021 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2022 | inode->i_mode, |
2023 | IS_IMMUTABLE(inode)? " immut" : "", | |
2024 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2025 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2026 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2027 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2028 | #endif |
2029 | ||
2030 | /* Normally we reject any write/sattr etc access on a read-only file | |
2031 | * system. But if it is IRIX doing check on write-access for a | |
2032 | * device special file, we ignore rofs. | |
2033 | */ | |
8837abca MS |
2034 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2035 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2036 | if (exp_rdonly(rqstp, exp) || |
2037 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2038 | return nfserr_rofs; |
8837abca | 2039 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2040 | return nfserr_perm; |
2041 | } | |
8837abca | 2042 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2043 | return nfserr_perm; |
2044 | ||
8837abca | 2045 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2046 | /* If we cannot rely on authentication in NLM requests, |
2047 | * just allow locks, otherwise require read permission, or | |
2048 | * ownership | |
2049 | */ | |
2050 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2051 | return 0; | |
2052 | else | |
8837abca | 2053 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2054 | } |
2055 | /* | |
2056 | * The file owner always gets access permission for accesses that | |
2057 | * would normally be checked at open time. This is to make | |
2058 | * file access work even when the client has done a fchmod(fd, 0). | |
2059 | * | |
2060 | * However, `cp foo bar' should fail nevertheless when bar is | |
2061 | * readonly. A sensible way to do this might be to reject all | |
2062 | * attempts to truncate a read-only file, because a creat() call | |
2063 | * always implies file truncation. | |
2064 | * ... but this isn't really fair. A process may reasonably call | |
2065 | * ftruncate on an open file descriptor on a file with perm 000. | |
2066 | * We must trust the client to do permission checking - using "ACCESS" | |
2067 | * with NFSv3. | |
2068 | */ | |
8837abca | 2069 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2070 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2071 | return 0; |
2072 | ||
8837abca | 2073 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2074 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2075 | |
2076 | /* Allow read access to binaries even when mode 111 */ | |
2077 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2078 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2079 | err = inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
2080 | |
2081 | return err? nfserrno(err) : 0; | |
2082 | } | |
2083 | ||
2084 | void | |
2085 | nfsd_racache_shutdown(void) | |
2086 | { | |
54a66e54 JL |
2087 | struct raparms *raparm, *last_raparm; |
2088 | unsigned int i; | |
2089 | ||
1da177e4 | 2090 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2091 | |
2092 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2093 | raparm = raparm_hash[i].pb_head; | |
2094 | while(raparm) { | |
2095 | last_raparm = raparm; | |
2096 | raparm = raparm->p_next; | |
2097 | kfree(last_raparm); | |
2098 | } | |
2099 | raparm_hash[i].pb_head = NULL; | |
2100 | } | |
1da177e4 LT |
2101 | } |
2102 | /* | |
2103 | * Initialize readahead param cache | |
2104 | */ | |
2105 | int | |
2106 | nfsd_racache_init(int cache_size) | |
2107 | { | |
2108 | int i; | |
fce1456a GB |
2109 | int j = 0; |
2110 | int nperbucket; | |
54a66e54 | 2111 | struct raparms **raparm = NULL; |
1da177e4 | 2112 | |
fce1456a | 2113 | |
54a66e54 | 2114 | if (raparm_hash[0].pb_head) |
1da177e4 | 2115 | return 0; |
54a66e54 JL |
2116 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2117 | if (nperbucket < 2) | |
2118 | nperbucket = 2; | |
2119 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2120 | |
2121 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2122 | |
2123 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2124 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2125 | |
2126 | raparm = &raparm_hash[i].pb_head; | |
2127 | for (j = 0; j < nperbucket; j++) { | |
2128 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2129 | if (!*raparm) | |
2130 | goto out_nomem; | |
2131 | raparm = &(*raparm)->p_next; | |
2132 | } | |
2133 | *raparm = NULL; | |
4b3bb06b YB |
2134 | } |
2135 | ||
1da177e4 LT |
2136 | nfsdstats.ra_size = cache_size; |
2137 | return 0; | |
54a66e54 JL |
2138 | |
2139 | out_nomem: | |
2140 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2141 | nfsd_racache_shutdown(); | |
2142 | return -ENOMEM; | |
1da177e4 | 2143 | } |
a257cdd0 AG |
2144 | |
2145 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2146 | struct posix_acl * | |
2147 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2148 | { | |
2149 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2150 | char *name; | |
2151 | void *value = NULL; | |
2152 | ssize_t size; | |
2153 | struct posix_acl *acl; | |
2154 | ||
5be196e5 CH |
2155 | if (!IS_POSIXACL(inode)) |
2156 | return ERR_PTR(-EOPNOTSUPP); | |
2157 | ||
2158 | switch (type) { | |
2159 | case ACL_TYPE_ACCESS: | |
2160 | name = POSIX_ACL_XATTR_ACCESS; | |
2161 | break; | |
2162 | case ACL_TYPE_DEFAULT: | |
2163 | name = POSIX_ACL_XATTR_DEFAULT; | |
2164 | break; | |
2165 | default: | |
a257cdd0 | 2166 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2167 | } |
2168 | ||
5be196e5 CH |
2169 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2170 | if (size < 0) | |
2171 | return ERR_PTR(size); | |
a257cdd0 | 2172 | |
a257cdd0 | 2173 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2174 | kfree(value); |
2175 | return acl; | |
2176 | } | |
2177 | ||
2178 | int | |
2179 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2180 | { | |
2181 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2182 | char *name; | |
2183 | void *value = NULL; | |
2184 | size_t size; | |
2185 | int error; | |
2186 | ||
acfa4380 | 2187 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2188 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2189 | return -EOPNOTSUPP; | |
2190 | switch(type) { | |
2191 | case ACL_TYPE_ACCESS: | |
334a13ec | 2192 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2193 | break; |
2194 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2195 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2196 | break; |
2197 | default: | |
2198 | return -EOPNOTSUPP; | |
2199 | } | |
2200 | ||
2201 | if (acl && acl->a_count) { | |
334a13ec | 2202 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2203 | value = kmalloc(size, GFP_KERNEL); |
2204 | if (!value) | |
2205 | return -ENOMEM; | |
9ccfc29c FM |
2206 | error = posix_acl_to_xattr(acl, value, size); |
2207 | if (error < 0) | |
a257cdd0 | 2208 | goto getout; |
9ccfc29c | 2209 | size = error; |
a257cdd0 AG |
2210 | } else |
2211 | size = 0; | |
2212 | ||
18f335af DH |
2213 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2214 | if (error) | |
2215 | goto getout; | |
a257cdd0 | 2216 | if (size) |
5be196e5 | 2217 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2218 | else { |
2219 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2220 | error = 0; | |
2221 | else { | |
5be196e5 | 2222 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2223 | if (error == -ENODATA) |
2224 | error = 0; | |
2225 | } | |
2226 | } | |
18f335af | 2227 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2228 | |
2229 | getout: | |
2230 | kfree(value); | |
2231 | return error; | |
2232 | } | |
2233 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |