]>
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/string.h> |
20 | #include <linux/time.h> | |
21 | #include <linux/errno.h> | |
22 | #include <linux/fs.h> | |
23 | #include <linux/file.h> | |
24 | #include <linux/mount.h> | |
25 | #include <linux/major.h> | |
d6b29d7c | 26 | #include <linux/splice.h> |
1da177e4 LT |
27 | #include <linux/proc_fs.h> |
28 | #include <linux/stat.h> | |
29 | #include <linux/fcntl.h> | |
30 | #include <linux/net.h> | |
31 | #include <linux/unistd.h> | |
32 | #include <linux/slab.h> | |
33 | #include <linux/pagemap.h> | |
34 | #include <linux/in.h> | |
35 | #include <linux/module.h> | |
36 | #include <linux/namei.h> | |
37 | #include <linux/vfs.h> | |
38 | #include <linux/delay.h> | |
39 | #include <linux/sunrpc/svc.h> | |
40 | #include <linux/nfsd/nfsd.h> | |
41 | #ifdef CONFIG_NFSD_V3 | |
42 | #include <linux/nfs3.h> | |
43 | #include <linux/nfsd/xdr3.h> | |
44 | #endif /* CONFIG_NFSD_V3 */ | |
45 | #include <linux/nfsd/nfsfh.h> | |
46 | #include <linux/quotaops.h> | |
0eeca283 | 47 | #include <linux/fsnotify.h> |
1da177e4 LT |
48 | #include <linux/posix_acl.h> |
49 | #include <linux/posix_acl_xattr.h> | |
1da177e4 | 50 | #include <linux/xattr.h> |
5be196e5 | 51 | #ifdef CONFIG_NFSD_V4 |
1da177e4 LT |
52 | #include <linux/nfs4.h> |
53 | #include <linux/nfs4_acl.h> | |
54 | #include <linux/nfsd_idmap.h> | |
55 | #include <linux/security.h> | |
56 | #endif /* CONFIG_NFSD_V4 */ | |
fce1456a | 57 | #include <linux/jhash.h> |
14dba533 | 58 | #include <linux/ima.h> |
1da177e4 LT |
59 | |
60 | #include <asm/uaccess.h> | |
61 | ||
62 | #define NFSDDBG_FACILITY NFSDDBG_FILEOP | |
1da177e4 LT |
63 | |
64 | ||
1da177e4 LT |
65 | /* |
66 | * This is a cache of readahead params that help us choose the proper | |
67 | * readahead strategy. Initially, we set all readahead parameters to 0 | |
68 | * and let the VFS handle things. | |
69 | * If you increase the number of cached files very much, you'll need to | |
70 | * add a hash table here. | |
71 | */ | |
72 | struct raparms { | |
73 | struct raparms *p_next; | |
74 | unsigned int p_count; | |
75 | ino_t p_ino; | |
76 | dev_t p_dev; | |
77 | int p_set; | |
78 | struct file_ra_state p_ra; | |
fce1456a | 79 | unsigned int p_hindex; |
1da177e4 LT |
80 | }; |
81 | ||
fce1456a GB |
82 | struct raparm_hbucket { |
83 | struct raparms *pb_head; | |
84 | spinlock_t pb_lock; | |
85 | } ____cacheline_aligned_in_smp; | |
86 | ||
fce1456a GB |
87 | #define RAPARM_HASH_BITS 4 |
88 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
89 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
90 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
91 | |
92 | /* | |
93 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
94 | * a mount point. | |
e0bb89ef | 95 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
96 | * or nfs_ok having possibly changed *dpp and *expp |
97 | */ | |
98 | int | |
99 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
100 | struct svc_export **expp) | |
101 | { | |
102 | struct svc_export *exp = *expp, *exp2 = NULL; | |
103 | struct dentry *dentry = *dpp; | |
91c9fa8f AV |
104 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
105 | .dentry = dget(dentry)}; | |
6264d69d | 106 | int err = 0; |
1da177e4 | 107 | |
91c9fa8f AV |
108 | while (follow_down(&path.mnt, &path.dentry) && |
109 | d_mountpoint(path.dentry)) | |
110 | ; | |
1da177e4 | 111 | |
91c9fa8f | 112 | exp2 = rqst_exp_get_by_name(rqstp, &path); |
1da177e4 | 113 | if (IS_ERR(exp2)) { |
a1033be7 NB |
114 | if (PTR_ERR(exp2) != -ENOENT) |
115 | err = PTR_ERR(exp2); | |
91c9fa8f | 116 | path_put(&path); |
1da177e4 LT |
117 | goto out; |
118 | } | |
5d3dbbea | 119 | if ((exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { |
1da177e4 | 120 | /* successfully crossed mount point */ |
1644ccc8 | 121 | /* |
91c9fa8f AV |
122 | * This is subtle: path.dentry is *not* on path.mnt |
123 | * at this point. The only reason we are safe is that | |
124 | * original mnt is pinned down by exp, so we should | |
125 | * put path *before* putting exp | |
1644ccc8 | 126 | */ |
91c9fa8f AV |
127 | *dpp = path.dentry; |
128 | path.dentry = dentry; | |
1644ccc8 | 129 | *expp = exp2; |
91c9fa8f | 130 | exp2 = exp; |
1da177e4 | 131 | } |
91c9fa8f AV |
132 | path_put(&path); |
133 | exp_put(exp2); | |
1da177e4 LT |
134 | out: |
135 | return err; | |
136 | } | |
137 | ||
6264d69d | 138 | __be32 |
6c0a654d | 139 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 140 | const char *name, unsigned int len, |
6c0a654d | 141 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
142 | { |
143 | struct svc_export *exp; | |
144 | struct dentry *dparent; | |
145 | struct dentry *dentry; | |
6264d69d AV |
146 | __be32 err; |
147 | int host_err; | |
1da177e4 LT |
148 | |
149 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
150 | ||
151 | /* Obtain dentry and export. */ | |
8837abca | 152 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
153 | if (err) |
154 | return err; | |
155 | ||
156 | dparent = fhp->fh_dentry; | |
157 | exp = fhp->fh_export; | |
158 | exp_get(exp); | |
159 | ||
1da177e4 LT |
160 | /* Lookup the name, but don't follow links */ |
161 | if (isdotent(name, len)) { | |
162 | if (len==1) | |
163 | dentry = dget(dparent); | |
54775491 | 164 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 165 | dentry = dget_parent(dparent); |
54775491 | 166 | else if (!EX_NOHIDE(exp)) |
1da177e4 LT |
167 | dentry = dget(dparent); /* .. == . just like at / */ |
168 | else { | |
169 | /* checking mountpoint crossing is very different when stepping up */ | |
170 | struct svc_export *exp2 = NULL; | |
171 | struct dentry *dp; | |
e64c390c AV |
172 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
173 | .dentry = dget(dparent)}; | |
174 | ||
175 | while (path.dentry == path.mnt->mnt_root && | |
bab77ebf | 176 | follow_up(&path)) |
1da177e4 | 177 | ; |
e64c390c AV |
178 | dp = dget_parent(path.dentry); |
179 | dput(path.dentry); | |
180 | path.dentry = dp; | |
1da177e4 | 181 | |
e64c390c | 182 | exp2 = rqst_exp_parent(rqstp, &path); |
2d3bb252 | 183 | if (PTR_ERR(exp2) == -ENOENT) { |
2d3bb252 BF |
184 | dentry = dget(dparent); |
185 | } else if (IS_ERR(exp2)) { | |
6264d69d | 186 | host_err = PTR_ERR(exp2); |
e64c390c | 187 | path_put(&path); |
1da177e4 | 188 | goto out_nfserr; |
1da177e4 | 189 | } else { |
e64c390c | 190 | dentry = dget(path.dentry); |
1da177e4 LT |
191 | exp_put(exp); |
192 | exp = exp2; | |
193 | } | |
e64c390c | 194 | path_put(&path); |
1da177e4 LT |
195 | } |
196 | } else { | |
197 | fh_lock(fhp); | |
198 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 199 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
200 | if (IS_ERR(dentry)) |
201 | goto out_nfserr; | |
202 | /* | |
203 | * check if we have crossed a mount point ... | |
204 | */ | |
205 | if (d_mountpoint(dentry)) { | |
6264d69d | 206 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
207 | dput(dentry); |
208 | goto out_nfserr; | |
209 | } | |
210 | } | |
211 | } | |
6c0a654d BF |
212 | *dentry_ret = dentry; |
213 | *exp_ret = exp; | |
214 | return 0; | |
215 | ||
216 | out_nfserr: | |
217 | exp_put(exp); | |
218 | return nfserrno(host_err); | |
219 | } | |
220 | ||
221 | /* | |
222 | * Look up one component of a pathname. | |
223 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
224 | * | |
225 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
226 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
227 | * accepted as it stands and the mounted directory is | |
228 | * returned. Otherwise the covered directory is returned. | |
229 | * NOTE: this mountpoint crossing is not supported properly by all | |
230 | * clients and is explicitly disallowed for NFSv3 | |
231 | * NeilBrown <neilb@cse.unsw.edu.au> | |
232 | */ | |
233 | __be32 | |
234 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 235 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
236 | { |
237 | struct svc_export *exp; | |
238 | struct dentry *dentry; | |
239 | __be32 err; | |
240 | ||
241 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
242 | if (err) | |
243 | return err; | |
32c1eb0c AA |
244 | err = check_nfsd_access(exp, rqstp); |
245 | if (err) | |
246 | goto out; | |
1da177e4 LT |
247 | /* |
248 | * Note: we compose the file handle now, but as the | |
249 | * dentry may be negative, it may need to be updated. | |
250 | */ | |
251 | err = fh_compose(resfh, exp, dentry, fhp); | |
252 | if (!err && !dentry->d_inode) | |
253 | err = nfserr_noent; | |
32c1eb0c | 254 | out: |
1da177e4 | 255 | dput(dentry); |
1da177e4 LT |
256 | exp_put(exp); |
257 | return err; | |
1da177e4 LT |
258 | } |
259 | ||
6c0a654d | 260 | |
1da177e4 LT |
261 | /* |
262 | * Set various file attributes. | |
263 | * N.B. After this call fhp needs an fh_put | |
264 | */ | |
6264d69d | 265 | __be32 |
1da177e4 LT |
266 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
267 | int check_guard, time_t guardtime) | |
268 | { | |
269 | struct dentry *dentry; | |
270 | struct inode *inode; | |
8837abca | 271 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 272 | int ftype = 0; |
6264d69d AV |
273 | __be32 err; |
274 | int host_err; | |
1da177e4 LT |
275 | int size_change = 0; |
276 | ||
277 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 278 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
279 | if (iap->ia_valid & ATTR_SIZE) |
280 | ftype = S_IFREG; | |
281 | ||
282 | /* Get inode */ | |
283 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 284 | if (err) |
1da177e4 LT |
285 | goto out; |
286 | ||
287 | dentry = fhp->fh_dentry; | |
288 | inode = dentry->d_inode; | |
289 | ||
15b7a1b8 N |
290 | /* Ignore any mode updates on symlinks */ |
291 | if (S_ISLNK(inode->i_mode)) | |
292 | iap->ia_valid &= ~ATTR_MODE; | |
293 | ||
294 | if (!iap->ia_valid) | |
295 | goto out; | |
296 | ||
9c85fca5 CH |
297 | /* |
298 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
299 | * which only requires access, and "set-[ac]time-to-X" which |
300 | * requires ownership. | |
301 | * So if it looks like it might be "set both to the same time which | |
302 | * is close to now", and if inode_change_ok fails, then we | |
303 | * convert to "set to now" instead of "set to explicit time" | |
304 | * | |
305 | * We only call inode_change_ok as the last test as technically | |
306 | * it is not an interface that we should be using. It is only | |
307 | * valid if the filesystem does not define it's own i_op->setattr. | |
308 | */ | |
309 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
310 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
311 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
312 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
313 | /* | |
314 | * Looks probable. | |
315 | * | |
316 | * Now just make sure time is in the right ballpark. | |
317 | * Solaris, at least, doesn't seem to care what the time | |
318 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 319 | */ |
9c85fca5 CH |
320 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
321 | if (delta < 0) | |
322 | delta = -delta; | |
323 | if (delta < MAX_TOUCH_TIME_ERROR && | |
324 | inode_change_ok(inode, iap) != 0) { | |
325 | /* | |
326 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
327 | * This will cause notify_change to set these times | |
328 | * to "now" | |
329 | */ | |
330 | iap->ia_valid &= ~BOTH_TIME_SET; | |
331 | } | |
1da177e4 LT |
332 | } |
333 | ||
9c85fca5 CH |
334 | /* |
335 | * The size case is special. | |
336 | * It changes the file as well as the attributes. | |
337 | */ | |
1da177e4 LT |
338 | if (iap->ia_valid & ATTR_SIZE) { |
339 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
340 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
341 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
342 | if (err) |
343 | goto out; | |
344 | } | |
345 | ||
346 | /* | |
347 | * If we are changing the size of the file, then | |
348 | * we need to break all leases. | |
349 | */ | |
6264d69d AV |
350 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
351 | if (host_err == -EWOULDBLOCK) | |
352 | host_err = -ETIMEDOUT; | |
353 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
354 | goto out_nfserr; |
355 | ||
6264d69d AV |
356 | host_err = get_write_access(inode); |
357 | if (host_err) | |
1da177e4 LT |
358 | goto out_nfserr; |
359 | ||
360 | size_change = 1; | |
6264d69d AV |
361 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
362 | if (host_err) { | |
1da177e4 LT |
363 | put_write_access(inode); |
364 | goto out_nfserr; | |
365 | } | |
90c0af05 | 366 | vfs_dq_init(inode); |
1da177e4 LT |
367 | } |
368 | ||
ca456252 | 369 | /* sanitize the mode change */ |
1da177e4 LT |
370 | if (iap->ia_valid & ATTR_MODE) { |
371 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 372 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
373 | } |
374 | ||
375 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
376 | if (!S_ISDIR(inode->i_mode) && |
377 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
378 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
379 | iap->ia_valid |= ATTR_KILL_PRIV; |
380 | if (iap->ia_valid & ATTR_MODE) { | |
381 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 382 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
383 | if (iap->ia_mode & S_IXGRP) |
384 | iap->ia_mode &= ~S_ISGID; | |
385 | } else { | |
386 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
387 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 388 | } |
1da177e4 LT |
389 | } |
390 | ||
1da177e4 LT |
391 | /* Change the attributes. */ |
392 | ||
393 | iap->ia_valid |= ATTR_CTIME; | |
394 | ||
395 | err = nfserr_notsync; | |
396 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
397 | fh_lock(fhp); | |
6264d69d AV |
398 | host_err = notify_change(dentry, iap); |
399 | err = nfserrno(host_err); | |
1da177e4 LT |
400 | fh_unlock(fhp); |
401 | } | |
402 | if (size_change) | |
403 | put_write_access(inode); | |
404 | if (!err) | |
405 | if (EX_ISSYNC(fhp->fh_export)) | |
406 | write_inode_now(inode, 1); | |
407 | out: | |
408 | return err; | |
409 | ||
410 | out_nfserr: | |
6264d69d | 411 | err = nfserrno(host_err); |
1da177e4 LT |
412 | goto out; |
413 | } | |
414 | ||
5be196e5 CH |
415 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
416 | defined(CONFIG_NFSD_V3_ACL) || \ | |
417 | defined(CONFIG_NFSD_V4) | |
418 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
419 | { | |
420 | ssize_t buflen; | |
6c6a426f | 421 | ssize_t ret; |
5be196e5 CH |
422 | |
423 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
424 | if (buflen <= 0) | |
425 | return buflen; | |
1da177e4 | 426 | |
5be196e5 CH |
427 | *buf = kmalloc(buflen, GFP_KERNEL); |
428 | if (!*buf) | |
429 | return -ENOMEM; | |
430 | ||
6c6a426f KK |
431 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
432 | if (ret < 0) | |
433 | kfree(*buf); | |
434 | return ret; | |
5be196e5 CH |
435 | } |
436 | #endif | |
437 | ||
438 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
439 | static int |
440 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
441 | { | |
442 | int len; | |
443 | size_t buflen; | |
444 | char *buf = NULL; | |
445 | int error = 0; | |
1da177e4 LT |
446 | |
447 | buflen = posix_acl_xattr_size(pacl->a_count); | |
448 | buf = kmalloc(buflen, GFP_KERNEL); | |
449 | error = -ENOMEM; | |
450 | if (buf == NULL) | |
451 | goto out; | |
452 | ||
453 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
454 | if (len < 0) { | |
455 | error = len; | |
456 | goto out; | |
457 | } | |
458 | ||
5be196e5 | 459 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
460 | out: |
461 | kfree(buf); | |
462 | return error; | |
463 | } | |
464 | ||
6264d69d | 465 | __be32 |
1da177e4 LT |
466 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
467 | struct nfs4_acl *acl) | |
468 | { | |
6264d69d AV |
469 | __be32 error; |
470 | int host_error; | |
1da177e4 LT |
471 | struct dentry *dentry; |
472 | struct inode *inode; | |
473 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
474 | unsigned int flags = 0; | |
475 | ||
476 | /* Get inode */ | |
8837abca | 477 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 478 | if (error) |
4b2ca38a | 479 | return error; |
1da177e4 LT |
480 | |
481 | dentry = fhp->fh_dentry; | |
482 | inode = dentry->d_inode; | |
483 | if (S_ISDIR(inode->i_mode)) | |
484 | flags = NFS4_ACL_DIR; | |
485 | ||
6264d69d AV |
486 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
487 | if (host_error == -EINVAL) { | |
4b2ca38a | 488 | return nfserr_attrnotsupp; |
6264d69d | 489 | } else if (host_error < 0) |
1da177e4 LT |
490 | goto out_nfserr; |
491 | ||
6264d69d AV |
492 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
493 | if (host_error < 0) | |
4b2ca38a | 494 | goto out_release; |
1da177e4 | 495 | |
4b2ca38a | 496 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 497 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 498 | |
4b2ca38a | 499 | out_release: |
1da177e4 LT |
500 | posix_acl_release(pacl); |
501 | posix_acl_release(dpacl); | |
1da177e4 | 502 | out_nfserr: |
f34f9242 | 503 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 504 | return nfserr_attrnotsupp; |
f34f9242 | 505 | else |
4b2ca38a | 506 | return nfserrno(host_error); |
1da177e4 LT |
507 | } |
508 | ||
509 | static struct posix_acl * | |
510 | _get_posix_acl(struct dentry *dentry, char *key) | |
511 | { | |
5be196e5 | 512 | void *buf = NULL; |
1da177e4 | 513 | struct posix_acl *pacl = NULL; |
5be196e5 | 514 | int buflen; |
1da177e4 | 515 | |
5be196e5 CH |
516 | buflen = nfsd_getxattr(dentry, key, &buf); |
517 | if (!buflen) | |
518 | buflen = -ENODATA; | |
519 | if (buflen <= 0) | |
520 | return ERR_PTR(buflen); | |
1da177e4 LT |
521 | |
522 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
523 | kfree(buf); |
524 | return pacl; | |
1da177e4 LT |
525 | } |
526 | ||
527 | int | |
528 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
529 | { | |
530 | struct inode *inode = dentry->d_inode; | |
531 | int error = 0; | |
532 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
533 | unsigned int flags = 0; | |
534 | ||
9a59f452 | 535 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
536 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
537 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
538 | if (IS_ERR(pacl)) { | |
539 | error = PTR_ERR(pacl); | |
540 | pacl = NULL; | |
541 | goto out; | |
542 | } | |
543 | ||
544 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 545 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
546 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
547 | dpacl = NULL; | |
548 | else if (IS_ERR(dpacl)) { | |
549 | error = PTR_ERR(dpacl); | |
550 | dpacl = NULL; | |
551 | goto out; | |
552 | } | |
553 | flags = NFS4_ACL_DIR; | |
554 | } | |
555 | ||
556 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
557 | if (IS_ERR(*acl)) { | |
558 | error = PTR_ERR(*acl); | |
559 | *acl = NULL; | |
560 | } | |
561 | out: | |
562 | posix_acl_release(pacl); | |
563 | posix_acl_release(dpacl); | |
564 | return error; | |
565 | } | |
566 | ||
567 | #endif /* defined(CONFIG_NFS_V4) */ | |
568 | ||
569 | #ifdef CONFIG_NFSD_V3 | |
570 | /* | |
571 | * Check server access rights to a file system object | |
572 | */ | |
573 | struct accessmap { | |
574 | u32 access; | |
575 | int how; | |
576 | }; | |
577 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
578 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
579 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
580 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
581 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
582 | |
583 | { 0, 0 } | |
584 | }; | |
585 | ||
586 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
587 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
588 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
589 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
590 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
591 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
592 | |
593 | { 0, 0 } | |
594 | }; | |
595 | ||
596 | static struct accessmap nfs3_anyaccess[] = { | |
597 | /* Some clients - Solaris 2.6 at least, make an access call | |
598 | * to the server to check for access for things like /dev/null | |
599 | * (which really, the server doesn't care about). So | |
600 | * We provide simple access checking for them, looking | |
601 | * mainly at mode bits, and we make sure to ignore read-only | |
602 | * filesystem checks | |
603 | */ | |
8837abca MS |
604 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
605 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
606 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
607 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
608 | |
609 | { 0, 0 } | |
610 | }; | |
611 | ||
6264d69d | 612 | __be32 |
1da177e4 LT |
613 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
614 | { | |
615 | struct accessmap *map; | |
616 | struct svc_export *export; | |
617 | struct dentry *dentry; | |
618 | u32 query, result = 0, sresult = 0; | |
6264d69d | 619 | __be32 error; |
1da177e4 | 620 | |
8837abca | 621 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
622 | if (error) |
623 | goto out; | |
624 | ||
625 | export = fhp->fh_export; | |
626 | dentry = fhp->fh_dentry; | |
627 | ||
628 | if (S_ISREG(dentry->d_inode->i_mode)) | |
629 | map = nfs3_regaccess; | |
630 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
631 | map = nfs3_diraccess; | |
632 | else | |
633 | map = nfs3_anyaccess; | |
634 | ||
635 | ||
636 | query = *access; | |
637 | for (; map->access; map++) { | |
638 | if (map->access & query) { | |
6264d69d | 639 | __be32 err2; |
1da177e4 LT |
640 | |
641 | sresult |= map->access; | |
642 | ||
0ec757df | 643 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
644 | switch (err2) { |
645 | case nfs_ok: | |
646 | result |= map->access; | |
647 | break; | |
648 | ||
649 | /* the following error codes just mean the access was not allowed, | |
650 | * rather than an error occurred */ | |
651 | case nfserr_rofs: | |
652 | case nfserr_acces: | |
653 | case nfserr_perm: | |
654 | /* simply don't "or" in the access bit. */ | |
655 | break; | |
656 | default: | |
657 | error = err2; | |
658 | goto out; | |
659 | } | |
660 | } | |
661 | } | |
662 | *access = result; | |
663 | if (supported) | |
664 | *supported = sresult; | |
665 | ||
666 | out: | |
667 | return error; | |
668 | } | |
669 | #endif /* CONFIG_NFSD_V3 */ | |
670 | ||
671 | ||
672 | ||
673 | /* | |
674 | * Open an existing file or directory. | |
675 | * The access argument indicates the type of open (read/write/lock) | |
676 | * N.B. After this call fhp needs an fh_put | |
677 | */ | |
6264d69d | 678 | __be32 |
1da177e4 LT |
679 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
680 | int access, struct file **filp) | |
681 | { | |
745ca247 | 682 | const struct cred *cred = current_cred(); |
1da177e4 LT |
683 | struct dentry *dentry; |
684 | struct inode *inode; | |
6264d69d AV |
685 | int flags = O_RDONLY|O_LARGEFILE; |
686 | __be32 err; | |
687 | int host_err; | |
1da177e4 LT |
688 | |
689 | /* | |
690 | * If we get here, then the client has already done an "open", | |
691 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
692 | * in case a chmod has now revoked permission. | |
693 | */ | |
8837abca | 694 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
695 | if (err) |
696 | goto out; | |
697 | ||
698 | dentry = fhp->fh_dentry; | |
699 | inode = dentry->d_inode; | |
700 | ||
701 | /* Disallow write access to files with the append-only bit set | |
702 | * or any access when mandatory locking enabled | |
703 | */ | |
704 | err = nfserr_perm; | |
8837abca | 705 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 706 | goto out; |
5e7fc436 BF |
707 | /* |
708 | * We must ignore files (but only files) which might have mandatory | |
709 | * locks on them because there is no way to know if the accesser has | |
710 | * the lock. | |
711 | */ | |
712 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
713 | goto out; |
714 | ||
715 | if (!inode->i_fop) | |
716 | goto out; | |
717 | ||
718 | /* | |
719 | * Check to see if there are any leases on this file. | |
720 | * This may block while leases are broken. | |
721 | */ | |
8837abca | 722 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0)); |
6264d69d AV |
723 | if (host_err == -EWOULDBLOCK) |
724 | host_err = -ETIMEDOUT; | |
725 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
726 | goto out_nfserr; |
727 | ||
8837abca MS |
728 | if (access & NFSD_MAY_WRITE) { |
729 | if (access & NFSD_MAY_READ) | |
9ecb6a08 BF |
730 | flags = O_RDWR|O_LARGEFILE; |
731 | else | |
732 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 733 | |
90c0af05 | 734 | vfs_dq_init(inode); |
1da177e4 | 735 | } |
54775491 | 736 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
745ca247 | 737 | flags, cred); |
1da177e4 | 738 | if (IS_ERR(*filp)) |
6264d69d | 739 | host_err = PTR_ERR(*filp); |
14dba533 MZ |
740 | else |
741 | ima_counts_get(*filp); | |
1da177e4 | 742 | out_nfserr: |
6264d69d | 743 | err = nfserrno(host_err); |
1da177e4 LT |
744 | out: |
745 | return err; | |
746 | } | |
747 | ||
748 | /* | |
749 | * Close a file. | |
750 | */ | |
751 | void | |
752 | nfsd_close(struct file *filp) | |
753 | { | |
754 | fput(filp); | |
755 | } | |
756 | ||
9a8d248e BF |
757 | /* |
758 | * Sync a file | |
759 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
760 | * after it. | |
761 | */ | |
762 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, | |
763 | const struct file_operations *fop) | |
764 | { | |
765 | struct inode *inode = dp->d_inode; | |
766 | int (*fsync) (struct file *, struct dentry *, int); | |
767 | int err; | |
768 | ||
769 | err = filemap_fdatawrite(inode->i_mapping); | |
770 | if (err == 0 && fop && (fsync = fop->fsync)) | |
771 | err = fsync(filp, dp, 0); | |
772 | if (err == 0) | |
773 | err = filemap_fdatawait(inode->i_mapping); | |
774 | ||
775 | return err; | |
776 | } | |
777 | ||
a334de28 | 778 | static int |
1da177e4 LT |
779 | nfsd_sync(struct file *filp) |
780 | { | |
9a8d248e BF |
781 | int err; |
782 | struct inode *inode = filp->f_path.dentry->d_inode; | |
783 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
784 | mutex_lock(&inode->i_mutex); | |
785 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); | |
786 | mutex_unlock(&inode->i_mutex); | |
787 | ||
788 | return err; | |
1da177e4 LT |
789 | } |
790 | ||
f193fbab | 791 | int |
9a8d248e | 792 | nfsd_sync_dir(struct dentry *dp) |
1da177e4 | 793 | { |
9a8d248e | 794 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
795 | } |
796 | ||
797 | /* | |
798 | * Obtain the readahead parameters for the file | |
799 | * specified by (dev, ino). | |
800 | */ | |
1da177e4 LT |
801 | |
802 | static inline struct raparms * | |
803 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
804 | { | |
805 | struct raparms *ra, **rap, **frap = NULL; | |
806 | int depth = 0; | |
fce1456a GB |
807 | unsigned int hash; |
808 | struct raparm_hbucket *rab; | |
809 | ||
810 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
811 | rab = &raparm_hash[hash]; | |
1da177e4 | 812 | |
fce1456a GB |
813 | spin_lock(&rab->pb_lock); |
814 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
815 | if (ra->p_ino == ino && ra->p_dev == dev) |
816 | goto found; | |
817 | depth++; | |
818 | if (ra->p_count == 0) | |
819 | frap = rap; | |
820 | } | |
821 | depth = nfsdstats.ra_size*11/10; | |
822 | if (!frap) { | |
fce1456a | 823 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
824 | return NULL; |
825 | } | |
826 | rap = frap; | |
827 | ra = *frap; | |
828 | ra->p_dev = dev; | |
829 | ra->p_ino = ino; | |
830 | ra->p_set = 0; | |
fce1456a | 831 | ra->p_hindex = hash; |
1da177e4 | 832 | found: |
fce1456a | 833 | if (rap != &rab->pb_head) { |
1da177e4 | 834 | *rap = ra->p_next; |
fce1456a GB |
835 | ra->p_next = rab->pb_head; |
836 | rab->pb_head = ra; | |
1da177e4 LT |
837 | } |
838 | ra->p_count++; | |
839 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 840 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
841 | return ra; |
842 | } | |
843 | ||
844 | /* | |
cf8208d0 JA |
845 | * Grab and keep cached pages associated with a file in the svc_rqst |
846 | * so that they can be passed to the network sendmsg/sendpage routines | |
847 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
848 | */ |
849 | static int | |
cf8208d0 JA |
850 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
851 | struct splice_desc *sd) | |
1da177e4 | 852 | { |
cf8208d0 | 853 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 854 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
855 | struct page *page = buf->page; |
856 | size_t size; | |
857 | int ret; | |
1da177e4 | 858 | |
cac36bb0 | 859 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
860 | if (unlikely(ret)) |
861 | return ret; | |
862 | ||
863 | size = sd->len; | |
1da177e4 LT |
864 | |
865 | if (rqstp->rq_res.page_len == 0) { | |
866 | get_page(page); | |
44524359 N |
867 | put_page(*pp); |
868 | *pp = page; | |
869 | rqstp->rq_resused++; | |
cf8208d0 | 870 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 871 | rqstp->rq_res.page_len = size; |
44524359 | 872 | } else if (page != pp[-1]) { |
1da177e4 | 873 | get_page(page); |
250f3915 N |
874 | if (*pp) |
875 | put_page(*pp); | |
44524359 N |
876 | *pp = page; |
877 | rqstp->rq_resused++; | |
1da177e4 | 878 | rqstp->rq_res.page_len += size; |
44524359 | 879 | } else |
1da177e4 | 880 | rqstp->rq_res.page_len += size; |
1da177e4 | 881 | |
1da177e4 LT |
882 | return size; |
883 | } | |
884 | ||
cf8208d0 JA |
885 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
886 | struct splice_desc *sd) | |
887 | { | |
888 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
889 | } | |
890 | ||
a8754bee DH |
891 | static inline int svc_msnfs(struct svc_fh *ffhp) |
892 | { | |
893 | #ifdef MSNFS | |
894 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
895 | #else | |
896 | return 0; | |
897 | #endif | |
898 | } | |
899 | ||
6264d69d | 900 | static __be32 |
1da177e4 LT |
901 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
902 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
903 | { | |
904 | struct inode *inode; | |
905 | struct raparms *ra; | |
906 | mm_segment_t oldfs; | |
6264d69d AV |
907 | __be32 err; |
908 | int host_err; | |
1da177e4 LT |
909 | |
910 | err = nfserr_perm; | |
7eaa36e2 | 911 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
912 | |
913 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 914 | goto out; |
1da177e4 LT |
915 | |
916 | /* Get readahead parameters */ | |
917 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
918 | ||
919 | if (ra && ra->p_set) | |
920 | file->f_ra = ra->p_ra; | |
921 | ||
cf8208d0 JA |
922 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
923 | struct splice_desc sd = { | |
924 | .len = 0, | |
925 | .total_len = *count, | |
926 | .pos = offset, | |
927 | .u.data = rqstp, | |
928 | }; | |
929 | ||
4fbef206 | 930 | rqstp->rq_resused = 1; |
cf8208d0 | 931 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
932 | } else { |
933 | oldfs = get_fs(); | |
934 | set_fs(KERNEL_DS); | |
6264d69d | 935 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
936 | set_fs(oldfs); |
937 | } | |
938 | ||
939 | /* Write back readahead params */ | |
940 | if (ra) { | |
fce1456a GB |
941 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
942 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
943 | ra->p_ra = file->f_ra; |
944 | ra->p_set = 1; | |
945 | ra->p_count--; | |
fce1456a | 946 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
947 | } |
948 | ||
6264d69d AV |
949 | if (host_err >= 0) { |
950 | nfsdstats.io_read += host_err; | |
951 | *count = host_err; | |
1da177e4 | 952 | err = 0; |
7eaa36e2 | 953 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 954 | } else |
6264d69d | 955 | err = nfserrno(host_err); |
1da177e4 LT |
956 | out: |
957 | return err; | |
958 | } | |
959 | ||
9f708e40 NB |
960 | static void kill_suid(struct dentry *dentry) |
961 | { | |
962 | struct iattr ia; | |
b5376771 | 963 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 964 | |
1b1dcc1b | 965 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 966 | notify_change(dentry, &ia); |
1b1dcc1b | 967 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
968 | } |
969 | ||
6264d69d | 970 | static __be32 |
1da177e4 LT |
971 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
972 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 973 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
974 | { |
975 | struct svc_export *exp; | |
976 | struct dentry *dentry; | |
977 | struct inode *inode; | |
978 | mm_segment_t oldfs; | |
6264d69d AV |
979 | __be32 err = 0; |
980 | int host_err; | |
1da177e4 LT |
981 | int stable = *stablep; |
982 | ||
45bd3b3d | 983 | #ifdef MSNFS |
1da177e4 LT |
984 | err = nfserr_perm; |
985 | ||
1da177e4 | 986 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
31dec253 | 987 | (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt))) |
1da177e4 LT |
988 | goto out; |
989 | #endif | |
990 | ||
7eaa36e2 | 991 | dentry = file->f_path.dentry; |
1da177e4 LT |
992 | inode = dentry->d_inode; |
993 | exp = fhp->fh_export; | |
994 | ||
995 | /* | |
996 | * Request sync writes if | |
997 | * - the sync export option has been set, or | |
998 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
999 | * - The file system doesn't support fsync(). | |
1000 | * When gathered writes have been configured for this volume, | |
1001 | * flushing the data to disk is handled separately below. | |
1002 | */ | |
1003 | ||
3ba15148 | 1004 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
1005 | stable = 2; |
1006 | *stablep = 2; /* FILE_SYNC */ | |
1007 | } | |
1008 | ||
1009 | if (!EX_ISSYNC(exp)) | |
1010 | stable = 0; | |
db1dd4d3 JC |
1011 | if (stable && !EX_WGATHER(exp)) { |
1012 | spin_lock(&file->f_lock); | |
1da177e4 | 1013 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
1014 | spin_unlock(&file->f_lock); |
1015 | } | |
1da177e4 LT |
1016 | |
1017 | /* Write the data. */ | |
1018 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1019 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1020 | set_fs(oldfs); |
6264d69d | 1021 | if (host_err >= 0) { |
a0d24b29 | 1022 | *cnt = host_err; |
31dec253 | 1023 | nfsdstats.io_write += host_err; |
7eaa36e2 | 1024 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
1025 | } |
1026 | ||
1027 | /* clear setuid/setgid flag after write */ | |
6264d69d | 1028 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 1029 | kill_suid(dentry); |
1da177e4 | 1030 | |
6264d69d | 1031 | if (host_err >= 0 && stable) { |
1da177e4 LT |
1032 | static ino_t last_ino; |
1033 | static dev_t last_dev; | |
1034 | ||
1035 | /* | |
1036 | * Gathered writes: If another process is currently | |
1037 | * writing to the file, there's a high chance | |
1038 | * this is another nfsd (triggered by a bulk write | |
1039 | * from a client's biod). Rather than syncing the | |
1040 | * file with each write request, we sleep for 10 msec. | |
1041 | * | |
1042 | * I don't know if this roughly approximates | |
1043 | * C. Juszak's idea of gathered writes, but it's a | |
1044 | * nice and simple solution (IMHO), and it seems to | |
1045 | * work:-) | |
1046 | */ | |
1047 | if (EX_WGATHER(exp)) { | |
1048 | if (atomic_read(&inode->i_writecount) > 1 | |
1049 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
ba25f9dc | 1050 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); |
1da177e4 | 1051 | msleep(10); |
ba25f9dc | 1052 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); |
1da177e4 LT |
1053 | } |
1054 | ||
1055 | if (inode->i_state & I_DIRTY) { | |
ba25f9dc | 1056 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); |
6264d69d | 1057 | host_err=nfsd_sync(file); |
1da177e4 LT |
1058 | } |
1059 | #if 0 | |
1060 | wake_up(&inode->i_wait); | |
1061 | #endif | |
1062 | } | |
1063 | last_ino = inode->i_ino; | |
1064 | last_dev = inode->i_sb->s_dev; | |
1065 | } | |
1066 | ||
6264d69d | 1067 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1068 | if (host_err >= 0) |
1da177e4 | 1069 | err = 0; |
a0d24b29 | 1070 | else |
6264d69d | 1071 | err = nfserrno(host_err); |
1da177e4 LT |
1072 | out: |
1073 | return err; | |
1074 | } | |
1075 | ||
1076 | /* | |
1077 | * Read data from a file. count must contain the requested read count | |
1078 | * on entry. On return, *count contains the number of bytes actually read. | |
1079 | * N.B. After this call fhp needs an fh_put | |
1080 | */ | |
6264d69d | 1081 | __be32 |
1da177e4 LT |
1082 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1083 | loff_t offset, struct kvec *vec, int vlen, | |
1084 | unsigned long *count) | |
1085 | { | |
6264d69d | 1086 | __be32 err; |
1da177e4 LT |
1087 | |
1088 | if (file) { | |
0ec757df | 1089 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1090 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1091 | if (err) |
1092 | goto out; | |
1093 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1094 | } else { | |
8837abca | 1095 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); |
1da177e4 LT |
1096 | if (err) |
1097 | goto out; | |
1098 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1099 | nfsd_close(file); | |
1100 | } | |
1101 | out: | |
1102 | return err; | |
1103 | } | |
1104 | ||
1105 | /* | |
1106 | * Write data to a file. | |
1107 | * The stable flag requests synchronous writes. | |
1108 | * N.B. After this call fhp needs an fh_put | |
1109 | */ | |
6264d69d | 1110 | __be32 |
1da177e4 | 1111 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1112 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1113 | int *stablep) |
1114 | { | |
6264d69d | 1115 | __be32 err = 0; |
1da177e4 LT |
1116 | |
1117 | if (file) { | |
0ec757df | 1118 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1119 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1120 | if (err) |
1121 | goto out; | |
1122 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1123 | stablep); | |
1124 | } else { | |
8837abca | 1125 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1126 | if (err) |
1127 | goto out; | |
1128 | ||
1129 | if (cnt) | |
1130 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1131 | cnt, stablep); | |
1132 | nfsd_close(file); | |
1133 | } | |
1134 | out: | |
1135 | return err; | |
1136 | } | |
1137 | ||
1138 | #ifdef CONFIG_NFSD_V3 | |
1139 | /* | |
1140 | * Commit all pending writes to stable storage. | |
1141 | * Strictly speaking, we could sync just the indicated file region here, | |
1142 | * but there's currently no way we can ask the VFS to do so. | |
1143 | * | |
1144 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1145 | * data to the client, as locking happens lower down in the filesystem. | |
1146 | */ | |
6264d69d | 1147 | __be32 |
1da177e4 LT |
1148 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1149 | loff_t offset, unsigned long count) | |
1150 | { | |
1151 | struct file *file; | |
6264d69d | 1152 | __be32 err; |
1da177e4 LT |
1153 | |
1154 | if ((u64)count > ~(u64)offset) | |
1155 | return nfserr_inval; | |
1156 | ||
8837abca MS |
1157 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1158 | if (err) | |
1da177e4 LT |
1159 | return err; |
1160 | if (EX_ISSYNC(fhp->fh_export)) { | |
1161 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1162 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1163 | } else { |
1164 | err = nfserr_notsupp; | |
1165 | } | |
1166 | } | |
1167 | ||
1168 | nfsd_close(file); | |
1169 | return err; | |
1170 | } | |
1171 | #endif /* CONFIG_NFSD_V3 */ | |
1172 | ||
f2b0dee2 | 1173 | static __be32 |
5c002b3b BF |
1174 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1175 | struct iattr *iap) | |
1176 | { | |
1177 | /* | |
1178 | * Mode has already been set earlier in create: | |
1179 | */ | |
1180 | iap->ia_valid &= ~ATTR_MODE; | |
1181 | /* | |
1182 | * Setting uid/gid works only for root. Irix appears to | |
1183 | * send along the gid on create when it tries to implement | |
1184 | * setgid directories via NFS: | |
1185 | */ | |
5cc0a840 | 1186 | if (current_fsuid() != 0) |
5c002b3b BF |
1187 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1188 | if (iap->ia_valid) | |
1189 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1190 | return 0; | |
1191 | } | |
1192 | ||
4ac35c2f | 1193 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1194 | * setting size to 0 may fail for some specific file systems by the permission | |
1195 | * checking which requires WRITE permission but the mode is 000. | |
1196 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1197 | * 0 after file created. | |
1198 | * | |
1199 | * call this only after vfs_create() is called. | |
1200 | * */ | |
1201 | static void | |
1202 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1203 | { | |
1204 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1205 | iap->ia_valid &= ~ATTR_SIZE; | |
1206 | } | |
1207 | ||
1da177e4 LT |
1208 | /* |
1209 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1210 | * not yet implemented. | |
1211 | * If the response fh has been verified, the parent directory should | |
1212 | * already be locked. Note that the parent directory is left locked. | |
1213 | * | |
1214 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1215 | */ | |
6264d69d | 1216 | __be32 |
1da177e4 LT |
1217 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1218 | char *fname, int flen, struct iattr *iap, | |
1219 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1220 | { | |
1221 | struct dentry *dentry, *dchild = NULL; | |
1222 | struct inode *dirp; | |
6264d69d | 1223 | __be32 err; |
5c002b3b | 1224 | __be32 err2; |
6264d69d | 1225 | int host_err; |
1da177e4 LT |
1226 | |
1227 | err = nfserr_perm; | |
1228 | if (!flen) | |
1229 | goto out; | |
1230 | err = nfserr_exist; | |
1231 | if (isdotent(fname, flen)) | |
1232 | goto out; | |
1233 | ||
8837abca | 1234 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1235 | if (err) |
1236 | goto out; | |
1237 | ||
1238 | dentry = fhp->fh_dentry; | |
1239 | dirp = dentry->d_inode; | |
1240 | ||
1241 | err = nfserr_notdir; | |
acfa4380 | 1242 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1243 | goto out; |
1244 | /* | |
1245 | * Check whether the response file handle has been verified yet. | |
1246 | * If it has, the parent directory should already be locked. | |
1247 | */ | |
1248 | if (!resfhp->fh_dentry) { | |
1249 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1250 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1251 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1252 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1253 | if (IS_ERR(dchild)) |
1254 | goto out_nfserr; | |
1255 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1256 | if (err) | |
1257 | goto out; | |
1258 | } else { | |
1259 | /* called from nfsd_proc_create */ | |
1260 | dchild = dget(resfhp->fh_dentry); | |
1261 | if (!fhp->fh_locked) { | |
1262 | /* not actually possible */ | |
1263 | printk(KERN_ERR | |
1264 | "nfsd_create: parent %s/%s not locked!\n", | |
1265 | dentry->d_parent->d_name.name, | |
1266 | dentry->d_name.name); | |
d75f2b9f | 1267 | err = nfserr_io; |
1da177e4 LT |
1268 | goto out; |
1269 | } | |
1270 | } | |
1271 | /* | |
1272 | * Make sure the child dentry is still negative ... | |
1273 | */ | |
1274 | err = nfserr_exist; | |
1275 | if (dchild->d_inode) { | |
1276 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1277 | dentry->d_name.name, dchild->d_name.name); | |
1278 | goto out; | |
1279 | } | |
1280 | ||
1281 | if (!(iap->ia_valid & ATTR_MODE)) | |
1282 | iap->ia_mode = 0; | |
1283 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1284 | ||
07cad1d2 MS |
1285 | err = nfserr_inval; |
1286 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1287 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1288 | type); | |
1289 | goto out; | |
1290 | } | |
1291 | ||
1292 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1293 | if (host_err) | |
1294 | goto out_nfserr; | |
1295 | ||
1da177e4 LT |
1296 | /* |
1297 | * Get the dir op function pointer. | |
1298 | */ | |
088406bc | 1299 | err = 0; |
1da177e4 LT |
1300 | switch (type) { |
1301 | case S_IFREG: | |
6264d69d | 1302 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1303 | if (!host_err) |
1304 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1305 | break; |
1306 | case S_IFDIR: | |
6264d69d | 1307 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1308 | break; |
1309 | case S_IFCHR: | |
1310 | case S_IFBLK: | |
1311 | case S_IFIFO: | |
1312 | case S_IFSOCK: | |
6264d69d | 1313 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1314 | break; |
1da177e4 | 1315 | } |
463c3197 DH |
1316 | if (host_err < 0) { |
1317 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1318 | goto out_nfserr; |
463c3197 | 1319 | } |
1da177e4 LT |
1320 | |
1321 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1322 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1323 | write_inode_now(dchild->d_inode, 1); |
1324 | } | |
1325 | ||
5c002b3b BF |
1326 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1327 | if (err2) | |
1328 | err = err2; | |
463c3197 | 1329 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1330 | /* |
1331 | * Update the file handle to get the new inode info. | |
1332 | */ | |
1333 | if (!err) | |
1334 | err = fh_update(resfhp); | |
1335 | out: | |
1336 | if (dchild && !IS_ERR(dchild)) | |
1337 | dput(dchild); | |
1338 | return err; | |
1339 | ||
1340 | out_nfserr: | |
6264d69d | 1341 | err = nfserrno(host_err); |
1da177e4 LT |
1342 | goto out; |
1343 | } | |
1344 | ||
1345 | #ifdef CONFIG_NFSD_V3 | |
1346 | /* | |
1347 | * NFSv3 version of nfsd_create | |
1348 | */ | |
6264d69d | 1349 | __be32 |
1da177e4 LT |
1350 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1351 | char *fname, int flen, struct iattr *iap, | |
1352 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1353 | int *truncp, int *created) |
1da177e4 LT |
1354 | { |
1355 | struct dentry *dentry, *dchild = NULL; | |
1356 | struct inode *dirp; | |
6264d69d | 1357 | __be32 err; |
5c002b3b | 1358 | __be32 err2; |
6264d69d | 1359 | int host_err; |
1da177e4 | 1360 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1361 | |
1362 | err = nfserr_perm; | |
1363 | if (!flen) | |
1364 | goto out; | |
1365 | err = nfserr_exist; | |
1366 | if (isdotent(fname, flen)) | |
1367 | goto out; | |
1368 | if (!(iap->ia_valid & ATTR_MODE)) | |
1369 | iap->ia_mode = 0; | |
8837abca | 1370 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1371 | if (err) |
1372 | goto out; | |
1373 | ||
1374 | dentry = fhp->fh_dentry; | |
1375 | dirp = dentry->d_inode; | |
1376 | ||
1377 | /* Get all the sanity checks out of the way before | |
1378 | * we lock the parent. */ | |
1379 | err = nfserr_notdir; | |
acfa4380 | 1380 | if (!dirp->i_op->lookup) |
1da177e4 | 1381 | goto out; |
12fd3520 | 1382 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1383 | |
1384 | /* | |
1385 | * Compose the response file handle. | |
1386 | */ | |
1387 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1388 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1389 | if (IS_ERR(dchild)) |
1390 | goto out_nfserr; | |
1391 | ||
1392 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1393 | if (err) | |
1394 | goto out; | |
1395 | ||
1396 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1397 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1398 | * the high bit set, so just clear the high bits. If this is |
1399 | * ever changed to use different attrs for storing the | |
1400 | * verifier, then do_open_lookup() will also need to be fixed | |
1401 | * accordingly. | |
1da177e4 LT |
1402 | */ |
1403 | v_mtime = verifier[0]&0x7fffffff; | |
1404 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1405 | } |
1406 | ||
463c3197 DH |
1407 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1408 | if (host_err) | |
1409 | goto out_nfserr; | |
1da177e4 LT |
1410 | if (dchild->d_inode) { |
1411 | err = 0; | |
1412 | ||
1413 | switch (createmode) { | |
1414 | case NFS3_CREATE_UNCHECKED: | |
1415 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1416 | err = nfserr_exist; | |
1417 | else if (truncp) { | |
1418 | /* in nfsv4, we need to treat this case a little | |
1419 | * differently. we don't want to truncate the | |
1420 | * file now; this would be wrong if the OPEN | |
1421 | * fails for some other reason. furthermore, | |
1422 | * if the size is nonzero, we should ignore it | |
1423 | * according to spec! | |
1424 | */ | |
1425 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1426 | } | |
1427 | else { | |
1428 | iap->ia_valid &= ATTR_SIZE; | |
1429 | goto set_attr; | |
1430 | } | |
1431 | break; | |
1432 | case NFS3_CREATE_EXCLUSIVE: | |
1433 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1434 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1435 | && dchild->d_inode->i_size == 0 ) |
1436 | break; | |
1437 | /* fallthru */ | |
1438 | case NFS3_CREATE_GUARDED: | |
1439 | err = nfserr_exist; | |
1440 | } | |
463c3197 | 1441 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1442 | goto out; |
1443 | } | |
1444 | ||
6264d69d | 1445 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1446 | if (host_err < 0) { |
1447 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1448 | goto out_nfserr; |
463c3197 | 1449 | } |
81ac95c5 BF |
1450 | if (created) |
1451 | *created = 1; | |
1da177e4 LT |
1452 | |
1453 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1454 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1455 | /* setattr will sync the child (or not) */ |
1456 | } | |
1457 | ||
4ac35c2f | 1458 | nfsd_check_ignore_resizing(iap); |
1459 | ||
1da177e4 | 1460 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1461 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1462 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1463 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1464 | /* XXX someone who knows this better please fix it for nsec */ |
1465 | iap->ia_mtime.tv_sec = v_mtime; | |
1466 | iap->ia_atime.tv_sec = v_atime; | |
1467 | iap->ia_mtime.tv_nsec = 0; | |
1468 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1469 | } |
1470 | ||
1da177e4 | 1471 | set_attr: |
5c002b3b BF |
1472 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1473 | if (err2) | |
1474 | err = err2; | |
f193fbab | 1475 | |
463c3197 | 1476 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1477 | /* |
1478 | * Update the filehandle to get the new inode info. | |
1479 | */ | |
1480 | if (!err) | |
1481 | err = fh_update(resfhp); | |
1da177e4 LT |
1482 | |
1483 | out: | |
1484 | fh_unlock(fhp); | |
1485 | if (dchild && !IS_ERR(dchild)) | |
1486 | dput(dchild); | |
1487 | return err; | |
1488 | ||
1489 | out_nfserr: | |
6264d69d | 1490 | err = nfserrno(host_err); |
1da177e4 LT |
1491 | goto out; |
1492 | } | |
1493 | #endif /* CONFIG_NFSD_V3 */ | |
1494 | ||
1495 | /* | |
1496 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1497 | * fits into the buffer. On return, it contains the true length. | |
1498 | * N.B. After this call fhp needs an fh_put | |
1499 | */ | |
6264d69d | 1500 | __be32 |
1da177e4 LT |
1501 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1502 | { | |
1503 | struct dentry *dentry; | |
1504 | struct inode *inode; | |
1505 | mm_segment_t oldfs; | |
6264d69d AV |
1506 | __be32 err; |
1507 | int host_err; | |
1da177e4 | 1508 | |
8837abca | 1509 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1510 | if (err) |
1511 | goto out; | |
1512 | ||
1513 | dentry = fhp->fh_dentry; | |
1514 | inode = dentry->d_inode; | |
1515 | ||
1516 | err = nfserr_inval; | |
acfa4380 | 1517 | if (!inode->i_op->readlink) |
1da177e4 LT |
1518 | goto out; |
1519 | ||
54775491 | 1520 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1521 | /* N.B. Why does this call need a get_fs()?? |
1522 | * Remove the set_fs and watch the fireworks:-) --okir | |
1523 | */ | |
1524 | ||
1525 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1526 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1527 | set_fs(oldfs); |
1528 | ||
6264d69d | 1529 | if (host_err < 0) |
1da177e4 | 1530 | goto out_nfserr; |
6264d69d | 1531 | *lenp = host_err; |
1da177e4 LT |
1532 | err = 0; |
1533 | out: | |
1534 | return err; | |
1535 | ||
1536 | out_nfserr: | |
6264d69d | 1537 | err = nfserrno(host_err); |
1da177e4 LT |
1538 | goto out; |
1539 | } | |
1540 | ||
1541 | /* | |
1542 | * Create a symlink and look up its inode | |
1543 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1544 | */ | |
6264d69d | 1545 | __be32 |
1da177e4 LT |
1546 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1547 | char *fname, int flen, | |
1548 | char *path, int plen, | |
1549 | struct svc_fh *resfhp, | |
1550 | struct iattr *iap) | |
1551 | { | |
1552 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1553 | __be32 err, cerr; |
1554 | int host_err; | |
1da177e4 LT |
1555 | |
1556 | err = nfserr_noent; | |
1557 | if (!flen || !plen) | |
1558 | goto out; | |
1559 | err = nfserr_exist; | |
1560 | if (isdotent(fname, flen)) | |
1561 | goto out; | |
1562 | ||
8837abca | 1563 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1564 | if (err) |
1565 | goto out; | |
1566 | fh_lock(fhp); | |
1567 | dentry = fhp->fh_dentry; | |
1568 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1569 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1570 | if (IS_ERR(dnew)) |
1571 | goto out_nfserr; | |
1572 | ||
75c3f29d DH |
1573 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1574 | if (host_err) | |
1575 | goto out_nfserr; | |
1576 | ||
1da177e4 LT |
1577 | if (unlikely(path[plen] != 0)) { |
1578 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1579 | if (path_alloced == NULL) | |
6264d69d | 1580 | host_err = -ENOMEM; |
1da177e4 LT |
1581 | else { |
1582 | strncpy(path_alloced, path, plen); | |
1583 | path_alloced[plen] = 0; | |
db2e747b | 1584 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1585 | kfree(path_alloced); |
1586 | } | |
1587 | } else | |
db2e747b | 1588 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
1da177e4 | 1589 | |
6264d69d | 1590 | if (!host_err) { |
1da177e4 | 1591 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1592 | host_err = nfsd_sync_dir(dentry); |
1593 | } | |
1594 | err = nfserrno(host_err); | |
1da177e4 LT |
1595 | fh_unlock(fhp); |
1596 | ||
75c3f29d DH |
1597 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1598 | ||
1da177e4 LT |
1599 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1600 | dput(dnew); | |
1601 | if (err==0) err = cerr; | |
1602 | out: | |
1603 | return err; | |
1604 | ||
1605 | out_nfserr: | |
6264d69d | 1606 | err = nfserrno(host_err); |
1da177e4 LT |
1607 | goto out; |
1608 | } | |
1609 | ||
1610 | /* | |
1611 | * Create a hardlink | |
1612 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1613 | */ | |
6264d69d | 1614 | __be32 |
1da177e4 LT |
1615 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1616 | char *name, int len, struct svc_fh *tfhp) | |
1617 | { | |
1618 | struct dentry *ddir, *dnew, *dold; | |
1619 | struct inode *dirp, *dest; | |
6264d69d AV |
1620 | __be32 err; |
1621 | int host_err; | |
1da177e4 | 1622 | |
8837abca | 1623 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1624 | if (err) |
1625 | goto out; | |
8837abca | 1626 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1627 | if (err) |
1628 | goto out; | |
1629 | ||
1630 | err = nfserr_perm; | |
1631 | if (!len) | |
1632 | goto out; | |
1633 | err = nfserr_exist; | |
1634 | if (isdotent(name, len)) | |
1635 | goto out; | |
1636 | ||
12fd3520 | 1637 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1638 | ddir = ffhp->fh_dentry; |
1639 | dirp = ddir->d_inode; | |
1640 | ||
1641 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1642 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1643 | if (IS_ERR(dnew)) |
1644 | goto out_nfserr; | |
1645 | ||
1646 | dold = tfhp->fh_dentry; | |
1647 | dest = dold->d_inode; | |
1648 | ||
75c3f29d DH |
1649 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1650 | if (host_err) { | |
1651 | err = nfserrno(host_err); | |
1652 | goto out_dput; | |
1653 | } | |
6264d69d AV |
1654 | host_err = vfs_link(dold, dirp, dnew); |
1655 | if (!host_err) { | |
1da177e4 | 1656 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1657 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1658 | write_inode_now(dest, 1); |
1659 | } | |
6264d69d | 1660 | err = 0; |
1da177e4 | 1661 | } else { |
6264d69d | 1662 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1663 | err = nfserr_acces; |
1664 | else | |
6264d69d | 1665 | err = nfserrno(host_err); |
1da177e4 | 1666 | } |
75c3f29d DH |
1667 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1668 | out_dput: | |
1da177e4 | 1669 | dput(dnew); |
270d56e5 DR |
1670 | out_unlock: |
1671 | fh_unlock(ffhp); | |
1da177e4 LT |
1672 | out: |
1673 | return err; | |
1674 | ||
1675 | out_nfserr: | |
6264d69d | 1676 | err = nfserrno(host_err); |
270d56e5 | 1677 | goto out_unlock; |
1da177e4 LT |
1678 | } |
1679 | ||
1680 | /* | |
1681 | * Rename a file | |
1682 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1683 | */ | |
6264d69d | 1684 | __be32 |
1da177e4 LT |
1685 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1686 | struct svc_fh *tfhp, char *tname, int tlen) | |
1687 | { | |
1688 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1689 | struct inode *fdir, *tdir; | |
6264d69d AV |
1690 | __be32 err; |
1691 | int host_err; | |
1da177e4 | 1692 | |
8837abca | 1693 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1694 | if (err) |
1695 | goto out; | |
8837abca | 1696 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1697 | if (err) |
1698 | goto out; | |
1699 | ||
1700 | fdentry = ffhp->fh_dentry; | |
1701 | fdir = fdentry->d_inode; | |
1702 | ||
1703 | tdentry = tfhp->fh_dentry; | |
1704 | tdir = tdentry->d_inode; | |
1705 | ||
1706 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1707 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1708 | goto out; |
1709 | ||
1710 | err = nfserr_perm; | |
1711 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1712 | goto out; | |
1713 | ||
1714 | /* cannot use fh_lock as we need deadlock protective ordering | |
1715 | * so do it by hand */ | |
1716 | trap = lock_rename(tdentry, fdentry); | |
1717 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1718 | fill_pre_wcc(ffhp); | |
1719 | fill_pre_wcc(tfhp); | |
1720 | ||
1721 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1722 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1723 | if (IS_ERR(odentry)) |
1724 | goto out_nfserr; | |
1725 | ||
6264d69d | 1726 | host_err = -ENOENT; |
1da177e4 LT |
1727 | if (!odentry->d_inode) |
1728 | goto out_dput_old; | |
6264d69d | 1729 | host_err = -EINVAL; |
1da177e4 LT |
1730 | if (odentry == trap) |
1731 | goto out_dput_old; | |
1732 | ||
1733 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1734 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1735 | if (IS_ERR(ndentry)) |
1736 | goto out_dput_old; | |
6264d69d | 1737 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1738 | if (ndentry == trap) |
1739 | goto out_dput_new; | |
1740 | ||
9079b1eb | 1741 | if (svc_msnfs(ffhp) && |
1da177e4 LT |
1742 | ((atomic_read(&odentry->d_count) > 1) |
1743 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1744 | host_err = -EPERM; |
9079b1eb DH |
1745 | goto out_dput_new; |
1746 | } | |
1747 | ||
1748 | host_err = -EXDEV; | |
1749 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1750 | goto out_dput_new; | |
1751 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1752 | if (host_err) | |
1753 | goto out_dput_new; | |
1754 | ||
6264d69d AV |
1755 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1756 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1757 | host_err = nfsd_sync_dir(tdentry); | |
1758 | if (!host_err) | |
1759 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1760 | } |
1761 | ||
9079b1eb DH |
1762 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1763 | ||
1da177e4 LT |
1764 | out_dput_new: |
1765 | dput(ndentry); | |
1766 | out_dput_old: | |
1767 | dput(odentry); | |
1768 | out_nfserr: | |
6264d69d | 1769 | err = nfserrno(host_err); |
1da177e4 LT |
1770 | |
1771 | /* we cannot reply on fh_unlock on the two filehandles, | |
1772 | * as that would do the wrong thing if the two directories | |
1773 | * were the same, so again we do it by hand | |
1774 | */ | |
1775 | fill_post_wcc(ffhp); | |
1776 | fill_post_wcc(tfhp); | |
1777 | unlock_rename(tdentry, fdentry); | |
1778 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1779 | ||
1780 | out: | |
1781 | return err; | |
1782 | } | |
1783 | ||
1784 | /* | |
1785 | * Unlink a file or directory | |
1786 | * N.B. After this call fhp needs an fh_put | |
1787 | */ | |
6264d69d | 1788 | __be32 |
1da177e4 LT |
1789 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1790 | char *fname, int flen) | |
1791 | { | |
1792 | struct dentry *dentry, *rdentry; | |
1793 | struct inode *dirp; | |
6264d69d AV |
1794 | __be32 err; |
1795 | int host_err; | |
1da177e4 LT |
1796 | |
1797 | err = nfserr_acces; | |
1798 | if (!flen || isdotent(fname, flen)) | |
1799 | goto out; | |
8837abca | 1800 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1801 | if (err) |
1802 | goto out; | |
1803 | ||
12fd3520 | 1804 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1805 | dentry = fhp->fh_dentry; |
1806 | dirp = dentry->d_inode; | |
1807 | ||
1808 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1809 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1810 | if (IS_ERR(rdentry)) |
1811 | goto out_nfserr; | |
1812 | ||
1813 | if (!rdentry->d_inode) { | |
1814 | dput(rdentry); | |
1815 | err = nfserr_noent; | |
1816 | goto out; | |
1817 | } | |
1818 | ||
1819 | if (!type) | |
1820 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1821 | ||
0622753b DH |
1822 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1823 | if (host_err) | |
1824 | goto out_nfserr; | |
1825 | ||
1da177e4 LT |
1826 | if (type != S_IFDIR) { /* It's UNLINK */ |
1827 | #ifdef MSNFS | |
1828 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1829 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1830 | host_err = -EPERM; |
1da177e4 LT |
1831 | } else |
1832 | #endif | |
6264d69d | 1833 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1834 | } else { /* It's RMDIR */ |
6264d69d | 1835 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1836 | } |
1837 | ||
1838 | dput(rdentry); | |
1839 | ||
6264d69d | 1840 | if (host_err) |
0622753b | 1841 | goto out_drop; |
6264d69d AV |
1842 | if (EX_ISSYNC(fhp->fh_export)) |
1843 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 | 1844 | |
0622753b DH |
1845 | out_drop: |
1846 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1847 | out_nfserr: |
6264d69d | 1848 | err = nfserrno(host_err); |
f193fbab YT |
1849 | out: |
1850 | return err; | |
1da177e4 LT |
1851 | } |
1852 | ||
14f7dd63 DW |
1853 | /* |
1854 | * We do this buffering because we must not call back into the file | |
1855 | * system's ->lookup() method from the filldir callback. That may well | |
1856 | * deadlock a number of file systems. | |
1857 | * | |
1858 | * This is based heavily on the implementation of same in XFS. | |
1859 | */ | |
1860 | struct buffered_dirent { | |
1861 | u64 ino; | |
1862 | loff_t offset; | |
1863 | int namlen; | |
1864 | unsigned int d_type; | |
1865 | char name[]; | |
1866 | }; | |
1867 | ||
1868 | struct readdir_data { | |
1869 | char *dirent; | |
1870 | size_t used; | |
53c9c5c0 | 1871 | int full; |
14f7dd63 DW |
1872 | }; |
1873 | ||
1874 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1875 | loff_t offset, u64 ino, unsigned int d_type) | |
1876 | { | |
1877 | struct readdir_data *buf = __buf; | |
1878 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1879 | unsigned int reclen; | |
1880 | ||
1881 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1882 | if (buf->used + reclen > PAGE_SIZE) { |
1883 | buf->full = 1; | |
14f7dd63 | 1884 | return -EINVAL; |
53c9c5c0 | 1885 | } |
14f7dd63 DW |
1886 | |
1887 | de->namlen = namlen; | |
1888 | de->offset = offset; | |
1889 | de->ino = ino; | |
1890 | de->d_type = d_type; | |
1891 | memcpy(de->name, name, namlen); | |
1892 | buf->used += reclen; | |
1893 | ||
1894 | return 0; | |
1895 | } | |
1896 | ||
2f9092e1 DW |
1897 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
1898 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1899 | { |
14f7dd63 DW |
1900 | struct readdir_data buf; |
1901 | struct buffered_dirent *de; | |
2628b766 | 1902 | int host_err; |
14f7dd63 DW |
1903 | int size; |
1904 | loff_t offset; | |
2628b766 | 1905 | |
14f7dd63 DW |
1906 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1907 | if (!buf.dirent) | |
2f9092e1 | 1908 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
1909 | |
1910 | offset = *offsetp; | |
2628b766 | 1911 | |
14f7dd63 | 1912 | while (1) { |
2f9092e1 | 1913 | struct inode *dir_inode = file->f_path.dentry->d_inode; |
14f7dd63 DW |
1914 | unsigned int reclen; |
1915 | ||
b726e923 | 1916 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1917 | buf.used = 0; |
53c9c5c0 | 1918 | buf.full = 0; |
14f7dd63 DW |
1919 | |
1920 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1921 | if (buf.full) |
1922 | host_err = 0; | |
1923 | ||
1924 | if (host_err < 0) | |
14f7dd63 DW |
1925 | break; |
1926 | ||
1927 | size = buf.used; | |
1928 | ||
1929 | if (!size) | |
1930 | break; | |
1931 | ||
2f9092e1 DW |
1932 | /* |
1933 | * Various filldir functions may end up calling back into | |
1934 | * lookup_one_len() and the file system's ->lookup() method. | |
1935 | * These expect i_mutex to be held, as it would within readdir. | |
1936 | */ | |
1937 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
1938 | if (host_err) | |
1939 | break; | |
1940 | ||
14f7dd63 DW |
1941 | de = (struct buffered_dirent *)buf.dirent; |
1942 | while (size > 0) { | |
1943 | offset = de->offset; | |
1944 | ||
1945 | if (func(cdp, de->name, de->namlen, de->offset, | |
1946 | de->ino, de->d_type)) | |
2f9092e1 | 1947 | break; |
14f7dd63 DW |
1948 | |
1949 | if (cdp->err != nfs_ok) | |
2f9092e1 | 1950 | break; |
14f7dd63 DW |
1951 | |
1952 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1953 | sizeof(u64)); | |
1954 | size -= reclen; | |
1955 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1956 | } | |
2f9092e1 DW |
1957 | mutex_unlock(&dir_inode->i_mutex); |
1958 | if (size > 0) /* We bailed out early */ | |
1959 | break; | |
1960 | ||
c002a6c7 | 1961 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1962 | } |
1963 | ||
14f7dd63 | 1964 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
1965 | |
1966 | if (host_err) | |
1967 | return nfserrno(host_err); | |
14f7dd63 DW |
1968 | |
1969 | *offsetp = offset; | |
1970 | return cdp->err; | |
2628b766 DW |
1971 | } |
1972 | ||
1da177e4 LT |
1973 | /* |
1974 | * Read entries from a directory. | |
1975 | * The NFSv3/4 verifier we ignore for now. | |
1976 | */ | |
6264d69d | 1977 | __be32 |
1da177e4 | 1978 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1979 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1980 | { |
6264d69d | 1981 | __be32 err; |
1da177e4 LT |
1982 | struct file *file; |
1983 | loff_t offset = *offsetp; | |
1984 | ||
8837abca | 1985 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
1986 | if (err) |
1987 | goto out; | |
1988 | ||
1989 | offset = vfs_llseek(file, offset, 0); | |
1990 | if (offset < 0) { | |
1991 | err = nfserrno((int)offset); | |
1992 | goto out_close; | |
1993 | } | |
1994 | ||
14f7dd63 | 1995 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
1996 | |
1997 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1998 | err = nfs_ok; /* can still be found in ->err */ | |
1999 | out_close: | |
2000 | nfsd_close(file); | |
2001 | out: | |
2002 | return err; | |
2003 | } | |
2004 | ||
2005 | /* | |
2006 | * Get file system stats | |
2007 | * N.B. After this call fhp needs an fh_put | |
2008 | */ | |
6264d69d | 2009 | __be32 |
04716e66 | 2010 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 2011 | { |
04716e66 | 2012 | __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); |
726c3342 | 2013 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
2014 | err = nfserr_io; |
2015 | return err; | |
2016 | } | |
2017 | ||
c7d51402 | 2018 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 2019 | { |
c7d51402 | 2020 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
2021 | } |
2022 | ||
1da177e4 LT |
2023 | /* |
2024 | * Check for a user's access permissions to this inode. | |
2025 | */ | |
6264d69d | 2026 | __be32 |
0ec757df BF |
2027 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2028 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2029 | { |
2030 | struct inode *inode = dentry->d_inode; | |
14dba533 | 2031 | struct path path; |
1da177e4 LT |
2032 | int err; |
2033 | ||
8837abca | 2034 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2035 | return 0; |
2036 | #if 0 | |
2037 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2038 | acc, | |
8837abca MS |
2039 | (acc & NFSD_MAY_READ)? " read" : "", |
2040 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2041 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2042 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2043 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2044 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2045 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2046 | inode->i_mode, |
2047 | IS_IMMUTABLE(inode)? " immut" : "", | |
2048 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2049 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2050 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2051 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2052 | #endif |
2053 | ||
2054 | /* Normally we reject any write/sattr etc access on a read-only file | |
2055 | * system. But if it is IRIX doing check on write-access for a | |
2056 | * device special file, we ignore rofs. | |
2057 | */ | |
8837abca MS |
2058 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2059 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2060 | if (exp_rdonly(rqstp, exp) || |
2061 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2062 | return nfserr_rofs; |
8837abca | 2063 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2064 | return nfserr_perm; |
2065 | } | |
8837abca | 2066 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2067 | return nfserr_perm; |
2068 | ||
8837abca | 2069 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2070 | /* If we cannot rely on authentication in NLM requests, |
2071 | * just allow locks, otherwise require read permission, or | |
2072 | * ownership | |
2073 | */ | |
2074 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2075 | return 0; | |
2076 | else | |
8837abca | 2077 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2078 | } |
2079 | /* | |
2080 | * The file owner always gets access permission for accesses that | |
2081 | * would normally be checked at open time. This is to make | |
2082 | * file access work even when the client has done a fchmod(fd, 0). | |
2083 | * | |
2084 | * However, `cp foo bar' should fail nevertheless when bar is | |
2085 | * readonly. A sensible way to do this might be to reject all | |
2086 | * attempts to truncate a read-only file, because a creat() call | |
2087 | * always implies file truncation. | |
2088 | * ... but this isn't really fair. A process may reasonably call | |
2089 | * ftruncate on an open file descriptor on a file with perm 000. | |
2090 | * We must trust the client to do permission checking - using "ACCESS" | |
2091 | * with NFSv3. | |
2092 | */ | |
8837abca | 2093 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2094 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2095 | return 0; |
2096 | ||
8837abca | 2097 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2098 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2099 | |
2100 | /* Allow read access to binaries even when mode 111 */ | |
2101 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2102 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2103 | err = inode_permission(inode, MAY_EXEC); |
14dba533 MZ |
2104 | if (err) |
2105 | goto nfsd_out; | |
1da177e4 | 2106 | |
14dba533 MZ |
2107 | /* Do integrity (permission) checking now, but defer incrementing |
2108 | * IMA counts to the actual file open. | |
2109 | */ | |
2110 | path.mnt = exp->ex_path.mnt; | |
2111 | path.dentry = dentry; | |
2112 | err = ima_path_check(&path, acc & (MAY_READ | MAY_WRITE | MAY_EXEC), | |
2113 | IMA_COUNT_LEAVE); | |
2114 | nfsd_out: | |
1da177e4 LT |
2115 | return err? nfserrno(err) : 0; |
2116 | } | |
2117 | ||
2118 | void | |
2119 | nfsd_racache_shutdown(void) | |
2120 | { | |
54a66e54 JL |
2121 | struct raparms *raparm, *last_raparm; |
2122 | unsigned int i; | |
2123 | ||
1da177e4 | 2124 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2125 | |
2126 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2127 | raparm = raparm_hash[i].pb_head; | |
2128 | while(raparm) { | |
2129 | last_raparm = raparm; | |
2130 | raparm = raparm->p_next; | |
2131 | kfree(last_raparm); | |
2132 | } | |
2133 | raparm_hash[i].pb_head = NULL; | |
2134 | } | |
1da177e4 LT |
2135 | } |
2136 | /* | |
2137 | * Initialize readahead param cache | |
2138 | */ | |
2139 | int | |
2140 | nfsd_racache_init(int cache_size) | |
2141 | { | |
2142 | int i; | |
fce1456a GB |
2143 | int j = 0; |
2144 | int nperbucket; | |
54a66e54 | 2145 | struct raparms **raparm = NULL; |
1da177e4 | 2146 | |
fce1456a | 2147 | |
54a66e54 | 2148 | if (raparm_hash[0].pb_head) |
1da177e4 | 2149 | return 0; |
54a66e54 JL |
2150 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2151 | if (nperbucket < 2) | |
2152 | nperbucket = 2; | |
2153 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2154 | |
2155 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2156 | |
2157 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2158 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2159 | |
2160 | raparm = &raparm_hash[i].pb_head; | |
2161 | for (j = 0; j < nperbucket; j++) { | |
2162 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2163 | if (!*raparm) | |
2164 | goto out_nomem; | |
2165 | raparm = &(*raparm)->p_next; | |
2166 | } | |
2167 | *raparm = NULL; | |
4b3bb06b YB |
2168 | } |
2169 | ||
1da177e4 LT |
2170 | nfsdstats.ra_size = cache_size; |
2171 | return 0; | |
54a66e54 JL |
2172 | |
2173 | out_nomem: | |
2174 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2175 | nfsd_racache_shutdown(); | |
2176 | return -ENOMEM; | |
1da177e4 | 2177 | } |
a257cdd0 AG |
2178 | |
2179 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2180 | struct posix_acl * | |
2181 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2182 | { | |
2183 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2184 | char *name; | |
2185 | void *value = NULL; | |
2186 | ssize_t size; | |
2187 | struct posix_acl *acl; | |
2188 | ||
5be196e5 CH |
2189 | if (!IS_POSIXACL(inode)) |
2190 | return ERR_PTR(-EOPNOTSUPP); | |
2191 | ||
2192 | switch (type) { | |
2193 | case ACL_TYPE_ACCESS: | |
2194 | name = POSIX_ACL_XATTR_ACCESS; | |
2195 | break; | |
2196 | case ACL_TYPE_DEFAULT: | |
2197 | name = POSIX_ACL_XATTR_DEFAULT; | |
2198 | break; | |
2199 | default: | |
a257cdd0 | 2200 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2201 | } |
2202 | ||
5be196e5 CH |
2203 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2204 | if (size < 0) | |
2205 | return ERR_PTR(size); | |
a257cdd0 | 2206 | |
a257cdd0 | 2207 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2208 | kfree(value); |
2209 | return acl; | |
2210 | } | |
2211 | ||
2212 | int | |
2213 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2214 | { | |
2215 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2216 | char *name; | |
2217 | void *value = NULL; | |
2218 | size_t size; | |
2219 | int error; | |
2220 | ||
acfa4380 | 2221 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2222 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2223 | return -EOPNOTSUPP; | |
2224 | switch(type) { | |
2225 | case ACL_TYPE_ACCESS: | |
334a13ec | 2226 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2227 | break; |
2228 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2229 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2230 | break; |
2231 | default: | |
2232 | return -EOPNOTSUPP; | |
2233 | } | |
2234 | ||
2235 | if (acl && acl->a_count) { | |
334a13ec | 2236 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2237 | value = kmalloc(size, GFP_KERNEL); |
2238 | if (!value) | |
2239 | return -ENOMEM; | |
9ccfc29c FM |
2240 | error = posix_acl_to_xattr(acl, value, size); |
2241 | if (error < 0) | |
a257cdd0 | 2242 | goto getout; |
9ccfc29c | 2243 | size = error; |
a257cdd0 AG |
2244 | } else |
2245 | size = 0; | |
2246 | ||
18f335af DH |
2247 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2248 | if (error) | |
2249 | goto getout; | |
a257cdd0 | 2250 | if (size) |
5be196e5 | 2251 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2252 | else { |
2253 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2254 | error = 0; | |
2255 | else { | |
5be196e5 | 2256 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2257 | if (error == -ENODATA) |
2258 | error = 0; | |
2259 | } | |
2260 | } | |
18f335af | 2261 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2262 | |
2263 | getout: | |
2264 | kfree(value); | |
2265 | return error; | |
2266 | } | |
2267 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |