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 | 91 | |
3c394dda SD |
92 | static inline int |
93 | nfsd_v4client(struct svc_rqst *rq) | |
94 | { | |
95 | return rq->rq_prog == NFS_PROGRAM && rq->rq_vers == 4; | |
96 | } | |
97 | ||
1da177e4 LT |
98 | /* |
99 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
100 | * a mount point. | |
e0bb89ef | 101 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
102 | * or nfs_ok having possibly changed *dpp and *expp |
103 | */ | |
104 | int | |
105 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
106 | struct svc_export **expp) | |
107 | { | |
108 | struct svc_export *exp = *expp, *exp2 = NULL; | |
109 | struct dentry *dentry = *dpp; | |
91c9fa8f AV |
110 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
111 | .dentry = dget(dentry)}; | |
6264d69d | 112 | int err = 0; |
1da177e4 | 113 | |
9393bd07 | 114 | while (d_mountpoint(path.dentry) && follow_down(&path)) |
91c9fa8f | 115 | ; |
1da177e4 | 116 | |
91c9fa8f | 117 | exp2 = rqst_exp_get_by_name(rqstp, &path); |
1da177e4 | 118 | if (IS_ERR(exp2)) { |
a1033be7 NB |
119 | if (PTR_ERR(exp2) != -ENOENT) |
120 | err = PTR_ERR(exp2); | |
91c9fa8f | 121 | path_put(&path); |
1da177e4 LT |
122 | goto out; |
123 | } | |
3c394dda SD |
124 | if (nfsd_v4client(rqstp) || |
125 | (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { | |
1da177e4 | 126 | /* successfully crossed mount point */ |
1644ccc8 | 127 | /* |
91c9fa8f AV |
128 | * This is subtle: path.dentry is *not* on path.mnt |
129 | * at this point. The only reason we are safe is that | |
130 | * original mnt is pinned down by exp, so we should | |
131 | * put path *before* putting exp | |
1644ccc8 | 132 | */ |
91c9fa8f AV |
133 | *dpp = path.dentry; |
134 | path.dentry = dentry; | |
1644ccc8 | 135 | *expp = exp2; |
91c9fa8f | 136 | exp2 = exp; |
1da177e4 | 137 | } |
91c9fa8f AV |
138 | path_put(&path); |
139 | exp_put(exp2); | |
1da177e4 LT |
140 | out: |
141 | return err; | |
142 | } | |
143 | ||
6264d69d | 144 | __be32 |
6c0a654d | 145 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 146 | const char *name, unsigned int len, |
6c0a654d | 147 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
148 | { |
149 | struct svc_export *exp; | |
150 | struct dentry *dparent; | |
151 | struct dentry *dentry; | |
6264d69d AV |
152 | __be32 err; |
153 | int host_err; | |
1da177e4 LT |
154 | |
155 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
156 | ||
157 | /* Obtain dentry and export. */ | |
8837abca | 158 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
159 | if (err) |
160 | return err; | |
161 | ||
162 | dparent = fhp->fh_dentry; | |
163 | exp = fhp->fh_export; | |
164 | exp_get(exp); | |
165 | ||
1da177e4 LT |
166 | /* Lookup the name, but don't follow links */ |
167 | if (isdotent(name, len)) { | |
168 | if (len==1) | |
169 | dentry = dget(dparent); | |
54775491 | 170 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 171 | dentry = dget_parent(dparent); |
54775491 | 172 | else if (!EX_NOHIDE(exp)) |
1da177e4 LT |
173 | dentry = dget(dparent); /* .. == . just like at / */ |
174 | else { | |
175 | /* checking mountpoint crossing is very different when stepping up */ | |
176 | struct svc_export *exp2 = NULL; | |
177 | struct dentry *dp; | |
e64c390c AV |
178 | struct path path = {.mnt = mntget(exp->ex_path.mnt), |
179 | .dentry = dget(dparent)}; | |
180 | ||
181 | while (path.dentry == path.mnt->mnt_root && | |
bab77ebf | 182 | follow_up(&path)) |
1da177e4 | 183 | ; |
e64c390c AV |
184 | dp = dget_parent(path.dentry); |
185 | dput(path.dentry); | |
186 | path.dentry = dp; | |
1da177e4 | 187 | |
e64c390c | 188 | exp2 = rqst_exp_parent(rqstp, &path); |
2d3bb252 | 189 | if (PTR_ERR(exp2) == -ENOENT) { |
2d3bb252 BF |
190 | dentry = dget(dparent); |
191 | } else if (IS_ERR(exp2)) { | |
6264d69d | 192 | host_err = PTR_ERR(exp2); |
e64c390c | 193 | path_put(&path); |
1da177e4 | 194 | goto out_nfserr; |
1da177e4 | 195 | } else { |
e64c390c | 196 | dentry = dget(path.dentry); |
1da177e4 LT |
197 | exp_put(exp); |
198 | exp = exp2; | |
199 | } | |
e64c390c | 200 | path_put(&path); |
1da177e4 LT |
201 | } |
202 | } else { | |
203 | fh_lock(fhp); | |
204 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 205 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
206 | if (IS_ERR(dentry)) |
207 | goto out_nfserr; | |
208 | /* | |
209 | * check if we have crossed a mount point ... | |
210 | */ | |
211 | if (d_mountpoint(dentry)) { | |
6264d69d | 212 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
213 | dput(dentry); |
214 | goto out_nfserr; | |
215 | } | |
216 | } | |
217 | } | |
6c0a654d BF |
218 | *dentry_ret = dentry; |
219 | *exp_ret = exp; | |
220 | return 0; | |
221 | ||
222 | out_nfserr: | |
223 | exp_put(exp); | |
224 | return nfserrno(host_err); | |
225 | } | |
226 | ||
227 | /* | |
228 | * Look up one component of a pathname. | |
229 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
230 | * | |
231 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
232 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
233 | * accepted as it stands and the mounted directory is | |
234 | * returned. Otherwise the covered directory is returned. | |
235 | * NOTE: this mountpoint crossing is not supported properly by all | |
236 | * clients and is explicitly disallowed for NFSv3 | |
237 | * NeilBrown <neilb@cse.unsw.edu.au> | |
238 | */ | |
239 | __be32 | |
240 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 241 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
242 | { |
243 | struct svc_export *exp; | |
244 | struct dentry *dentry; | |
245 | __be32 err; | |
246 | ||
247 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
248 | if (err) | |
249 | return err; | |
32c1eb0c AA |
250 | err = check_nfsd_access(exp, rqstp); |
251 | if (err) | |
252 | goto out; | |
1da177e4 LT |
253 | /* |
254 | * Note: we compose the file handle now, but as the | |
255 | * dentry may be negative, it may need to be updated. | |
256 | */ | |
257 | err = fh_compose(resfh, exp, dentry, fhp); | |
258 | if (!err && !dentry->d_inode) | |
259 | err = nfserr_noent; | |
32c1eb0c | 260 | out: |
1da177e4 | 261 | dput(dentry); |
1da177e4 LT |
262 | exp_put(exp); |
263 | return err; | |
1da177e4 LT |
264 | } |
265 | ||
6c0a654d | 266 | |
1da177e4 LT |
267 | /* |
268 | * Set various file attributes. | |
269 | * N.B. After this call fhp needs an fh_put | |
270 | */ | |
6264d69d | 271 | __be32 |
1da177e4 LT |
272 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
273 | int check_guard, time_t guardtime) | |
274 | { | |
275 | struct dentry *dentry; | |
276 | struct inode *inode; | |
8837abca | 277 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 278 | int ftype = 0; |
6264d69d AV |
279 | __be32 err; |
280 | int host_err; | |
1da177e4 LT |
281 | int size_change = 0; |
282 | ||
283 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 284 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
285 | if (iap->ia_valid & ATTR_SIZE) |
286 | ftype = S_IFREG; | |
287 | ||
288 | /* Get inode */ | |
289 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 290 | if (err) |
1da177e4 LT |
291 | goto out; |
292 | ||
293 | dentry = fhp->fh_dentry; | |
294 | inode = dentry->d_inode; | |
295 | ||
15b7a1b8 N |
296 | /* Ignore any mode updates on symlinks */ |
297 | if (S_ISLNK(inode->i_mode)) | |
298 | iap->ia_valid &= ~ATTR_MODE; | |
299 | ||
300 | if (!iap->ia_valid) | |
301 | goto out; | |
302 | ||
9c85fca5 CH |
303 | /* |
304 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
305 | * which only requires access, and "set-[ac]time-to-X" which |
306 | * requires ownership. | |
307 | * So if it looks like it might be "set both to the same time which | |
308 | * is close to now", and if inode_change_ok fails, then we | |
309 | * convert to "set to now" instead of "set to explicit time" | |
310 | * | |
311 | * We only call inode_change_ok as the last test as technically | |
312 | * it is not an interface that we should be using. It is only | |
313 | * valid if the filesystem does not define it's own i_op->setattr. | |
314 | */ | |
315 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
316 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
317 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
318 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
319 | /* | |
320 | * Looks probable. | |
321 | * | |
322 | * Now just make sure time is in the right ballpark. | |
323 | * Solaris, at least, doesn't seem to care what the time | |
324 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 325 | */ |
9c85fca5 CH |
326 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
327 | if (delta < 0) | |
328 | delta = -delta; | |
329 | if (delta < MAX_TOUCH_TIME_ERROR && | |
330 | inode_change_ok(inode, iap) != 0) { | |
331 | /* | |
332 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
333 | * This will cause notify_change to set these times | |
334 | * to "now" | |
335 | */ | |
336 | iap->ia_valid &= ~BOTH_TIME_SET; | |
337 | } | |
1da177e4 LT |
338 | } |
339 | ||
9c85fca5 CH |
340 | /* |
341 | * The size case is special. | |
342 | * It changes the file as well as the attributes. | |
343 | */ | |
1da177e4 LT |
344 | if (iap->ia_valid & ATTR_SIZE) { |
345 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
346 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
347 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
348 | if (err) |
349 | goto out; | |
350 | } | |
351 | ||
352 | /* | |
353 | * If we are changing the size of the file, then | |
354 | * we need to break all leases. | |
355 | */ | |
6264d69d AV |
356 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
357 | if (host_err == -EWOULDBLOCK) | |
358 | host_err = -ETIMEDOUT; | |
359 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
360 | goto out_nfserr; |
361 | ||
6264d69d AV |
362 | host_err = get_write_access(inode); |
363 | if (host_err) | |
1da177e4 LT |
364 | goto out_nfserr; |
365 | ||
366 | size_change = 1; | |
6264d69d AV |
367 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
368 | if (host_err) { | |
1da177e4 LT |
369 | put_write_access(inode); |
370 | goto out_nfserr; | |
371 | } | |
90c0af05 | 372 | vfs_dq_init(inode); |
1da177e4 LT |
373 | } |
374 | ||
ca456252 | 375 | /* sanitize the mode change */ |
1da177e4 LT |
376 | if (iap->ia_valid & ATTR_MODE) { |
377 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 378 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
379 | } |
380 | ||
381 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
382 | if (!S_ISDIR(inode->i_mode) && |
383 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
384 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
385 | iap->ia_valid |= ATTR_KILL_PRIV; |
386 | if (iap->ia_valid & ATTR_MODE) { | |
387 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 388 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
389 | if (iap->ia_mode & S_IXGRP) |
390 | iap->ia_mode &= ~S_ISGID; | |
391 | } else { | |
392 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
393 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 394 | } |
1da177e4 LT |
395 | } |
396 | ||
1da177e4 LT |
397 | /* Change the attributes. */ |
398 | ||
399 | iap->ia_valid |= ATTR_CTIME; | |
400 | ||
401 | err = nfserr_notsync; | |
402 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
403 | fh_lock(fhp); | |
6264d69d AV |
404 | host_err = notify_change(dentry, iap); |
405 | err = nfserrno(host_err); | |
1da177e4 LT |
406 | fh_unlock(fhp); |
407 | } | |
408 | if (size_change) | |
409 | put_write_access(inode); | |
410 | if (!err) | |
411 | if (EX_ISSYNC(fhp->fh_export)) | |
412 | write_inode_now(inode, 1); | |
413 | out: | |
414 | return err; | |
415 | ||
416 | out_nfserr: | |
6264d69d | 417 | err = nfserrno(host_err); |
1da177e4 LT |
418 | goto out; |
419 | } | |
420 | ||
5be196e5 CH |
421 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
422 | defined(CONFIG_NFSD_V3_ACL) || \ | |
423 | defined(CONFIG_NFSD_V4) | |
424 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
425 | { | |
426 | ssize_t buflen; | |
6c6a426f | 427 | ssize_t ret; |
5be196e5 CH |
428 | |
429 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
430 | if (buflen <= 0) | |
431 | return buflen; | |
1da177e4 | 432 | |
5be196e5 CH |
433 | *buf = kmalloc(buflen, GFP_KERNEL); |
434 | if (!*buf) | |
435 | return -ENOMEM; | |
436 | ||
6c6a426f KK |
437 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
438 | if (ret < 0) | |
439 | kfree(*buf); | |
440 | return ret; | |
5be196e5 CH |
441 | } |
442 | #endif | |
443 | ||
444 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
445 | static int |
446 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
447 | { | |
448 | int len; | |
449 | size_t buflen; | |
450 | char *buf = NULL; | |
451 | int error = 0; | |
1da177e4 LT |
452 | |
453 | buflen = posix_acl_xattr_size(pacl->a_count); | |
454 | buf = kmalloc(buflen, GFP_KERNEL); | |
455 | error = -ENOMEM; | |
456 | if (buf == NULL) | |
457 | goto out; | |
458 | ||
459 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
460 | if (len < 0) { | |
461 | error = len; | |
462 | goto out; | |
463 | } | |
464 | ||
5be196e5 | 465 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
466 | out: |
467 | kfree(buf); | |
468 | return error; | |
469 | } | |
470 | ||
6264d69d | 471 | __be32 |
1da177e4 LT |
472 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
473 | struct nfs4_acl *acl) | |
474 | { | |
6264d69d AV |
475 | __be32 error; |
476 | int host_error; | |
1da177e4 LT |
477 | struct dentry *dentry; |
478 | struct inode *inode; | |
479 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
480 | unsigned int flags = 0; | |
481 | ||
482 | /* Get inode */ | |
8837abca | 483 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 484 | if (error) |
4b2ca38a | 485 | return error; |
1da177e4 LT |
486 | |
487 | dentry = fhp->fh_dentry; | |
488 | inode = dentry->d_inode; | |
489 | if (S_ISDIR(inode->i_mode)) | |
490 | flags = NFS4_ACL_DIR; | |
491 | ||
6264d69d AV |
492 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
493 | if (host_error == -EINVAL) { | |
4b2ca38a | 494 | return nfserr_attrnotsupp; |
6264d69d | 495 | } else if (host_error < 0) |
1da177e4 LT |
496 | goto out_nfserr; |
497 | ||
6264d69d AV |
498 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
499 | if (host_error < 0) | |
4b2ca38a | 500 | goto out_release; |
1da177e4 | 501 | |
4b2ca38a | 502 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 503 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 504 | |
4b2ca38a | 505 | out_release: |
1da177e4 LT |
506 | posix_acl_release(pacl); |
507 | posix_acl_release(dpacl); | |
1da177e4 | 508 | out_nfserr: |
f34f9242 | 509 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 510 | return nfserr_attrnotsupp; |
f34f9242 | 511 | else |
4b2ca38a | 512 | return nfserrno(host_error); |
1da177e4 LT |
513 | } |
514 | ||
515 | static struct posix_acl * | |
516 | _get_posix_acl(struct dentry *dentry, char *key) | |
517 | { | |
5be196e5 | 518 | void *buf = NULL; |
1da177e4 | 519 | struct posix_acl *pacl = NULL; |
5be196e5 | 520 | int buflen; |
1da177e4 | 521 | |
5be196e5 CH |
522 | buflen = nfsd_getxattr(dentry, key, &buf); |
523 | if (!buflen) | |
524 | buflen = -ENODATA; | |
525 | if (buflen <= 0) | |
526 | return ERR_PTR(buflen); | |
1da177e4 LT |
527 | |
528 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
529 | kfree(buf); |
530 | return pacl; | |
1da177e4 LT |
531 | } |
532 | ||
533 | int | |
534 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
535 | { | |
536 | struct inode *inode = dentry->d_inode; | |
537 | int error = 0; | |
538 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
539 | unsigned int flags = 0; | |
540 | ||
9a59f452 | 541 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
542 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
543 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
544 | if (IS_ERR(pacl)) { | |
545 | error = PTR_ERR(pacl); | |
546 | pacl = NULL; | |
547 | goto out; | |
548 | } | |
549 | ||
550 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 551 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
552 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
553 | dpacl = NULL; | |
554 | else if (IS_ERR(dpacl)) { | |
555 | error = PTR_ERR(dpacl); | |
556 | dpacl = NULL; | |
557 | goto out; | |
558 | } | |
559 | flags = NFS4_ACL_DIR; | |
560 | } | |
561 | ||
562 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
563 | if (IS_ERR(*acl)) { | |
564 | error = PTR_ERR(*acl); | |
565 | *acl = NULL; | |
566 | } | |
567 | out: | |
568 | posix_acl_release(pacl); | |
569 | posix_acl_release(dpacl); | |
570 | return error; | |
571 | } | |
572 | ||
573 | #endif /* defined(CONFIG_NFS_V4) */ | |
574 | ||
575 | #ifdef CONFIG_NFSD_V3 | |
576 | /* | |
577 | * Check server access rights to a file system object | |
578 | */ | |
579 | struct accessmap { | |
580 | u32 access; | |
581 | int how; | |
582 | }; | |
583 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
584 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
585 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
586 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
587 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
588 | |
589 | { 0, 0 } | |
590 | }; | |
591 | ||
592 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
593 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
594 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
595 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
596 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
597 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
598 | |
599 | { 0, 0 } | |
600 | }; | |
601 | ||
602 | static struct accessmap nfs3_anyaccess[] = { | |
603 | /* Some clients - Solaris 2.6 at least, make an access call | |
604 | * to the server to check for access for things like /dev/null | |
605 | * (which really, the server doesn't care about). So | |
606 | * We provide simple access checking for them, looking | |
607 | * mainly at mode bits, and we make sure to ignore read-only | |
608 | * filesystem checks | |
609 | */ | |
8837abca MS |
610 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
611 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
612 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
613 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
614 | |
615 | { 0, 0 } | |
616 | }; | |
617 | ||
6264d69d | 618 | __be32 |
1da177e4 LT |
619 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
620 | { | |
621 | struct accessmap *map; | |
622 | struct svc_export *export; | |
623 | struct dentry *dentry; | |
624 | u32 query, result = 0, sresult = 0; | |
6264d69d | 625 | __be32 error; |
1da177e4 | 626 | |
8837abca | 627 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
628 | if (error) |
629 | goto out; | |
630 | ||
631 | export = fhp->fh_export; | |
632 | dentry = fhp->fh_dentry; | |
633 | ||
634 | if (S_ISREG(dentry->d_inode->i_mode)) | |
635 | map = nfs3_regaccess; | |
636 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
637 | map = nfs3_diraccess; | |
638 | else | |
639 | map = nfs3_anyaccess; | |
640 | ||
641 | ||
642 | query = *access; | |
643 | for (; map->access; map++) { | |
644 | if (map->access & query) { | |
6264d69d | 645 | __be32 err2; |
1da177e4 LT |
646 | |
647 | sresult |= map->access; | |
648 | ||
0ec757df | 649 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
650 | switch (err2) { |
651 | case nfs_ok: | |
652 | result |= map->access; | |
653 | break; | |
654 | ||
655 | /* the following error codes just mean the access was not allowed, | |
656 | * rather than an error occurred */ | |
657 | case nfserr_rofs: | |
658 | case nfserr_acces: | |
659 | case nfserr_perm: | |
660 | /* simply don't "or" in the access bit. */ | |
661 | break; | |
662 | default: | |
663 | error = err2; | |
664 | goto out; | |
665 | } | |
666 | } | |
667 | } | |
668 | *access = result; | |
669 | if (supported) | |
670 | *supported = sresult; | |
671 | ||
672 | out: | |
673 | return error; | |
674 | } | |
675 | #endif /* CONFIG_NFSD_V3 */ | |
676 | ||
677 | ||
678 | ||
679 | /* | |
680 | * Open an existing file or directory. | |
681 | * The access argument indicates the type of open (read/write/lock) | |
682 | * N.B. After this call fhp needs an fh_put | |
683 | */ | |
6264d69d | 684 | __be32 |
1da177e4 LT |
685 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
686 | int access, struct file **filp) | |
687 | { | |
688 | struct dentry *dentry; | |
689 | struct inode *inode; | |
6264d69d AV |
690 | int flags = O_RDONLY|O_LARGEFILE; |
691 | __be32 err; | |
692 | int host_err; | |
1da177e4 | 693 | |
e0e81739 DH |
694 | validate_process_creds(); |
695 | ||
1da177e4 LT |
696 | /* |
697 | * If we get here, then the client has already done an "open", | |
698 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
699 | * in case a chmod has now revoked permission. | |
700 | */ | |
8837abca | 701 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
702 | if (err) |
703 | goto out; | |
704 | ||
705 | dentry = fhp->fh_dentry; | |
706 | inode = dentry->d_inode; | |
707 | ||
708 | /* Disallow write access to files with the append-only bit set | |
709 | * or any access when mandatory locking enabled | |
710 | */ | |
711 | err = nfserr_perm; | |
8837abca | 712 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 713 | goto out; |
5e7fc436 BF |
714 | /* |
715 | * We must ignore files (but only files) which might have mandatory | |
716 | * locks on them because there is no way to know if the accesser has | |
717 | * the lock. | |
718 | */ | |
719 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
720 | goto out; |
721 | ||
722 | if (!inode->i_fop) | |
723 | goto out; | |
724 | ||
725 | /* | |
726 | * Check to see if there are any leases on this file. | |
727 | * This may block while leases are broken. | |
728 | */ | |
8837abca | 729 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0)); |
6264d69d AV |
730 | if (host_err == -EWOULDBLOCK) |
731 | host_err = -ETIMEDOUT; | |
732 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
733 | goto out_nfserr; |
734 | ||
8837abca MS |
735 | if (access & NFSD_MAY_WRITE) { |
736 | if (access & NFSD_MAY_READ) | |
9ecb6a08 BF |
737 | flags = O_RDWR|O_LARGEFILE; |
738 | else | |
739 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 740 | |
90c0af05 | 741 | vfs_dq_init(inode); |
1da177e4 | 742 | } |
54775491 | 743 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
033a666c | 744 | flags, current_cred()); |
1da177e4 | 745 | if (IS_ERR(*filp)) |
6264d69d | 746 | host_err = PTR_ERR(*filp); |
1da177e4 | 747 | out_nfserr: |
6264d69d | 748 | err = nfserrno(host_err); |
1da177e4 | 749 | out: |
e0e81739 | 750 | validate_process_creds(); |
1da177e4 LT |
751 | return err; |
752 | } | |
753 | ||
754 | /* | |
755 | * Close a file. | |
756 | */ | |
757 | void | |
758 | nfsd_close(struct file *filp) | |
759 | { | |
760 | fput(filp); | |
761 | } | |
762 | ||
9a8d248e BF |
763 | /* |
764 | * Sync a file | |
765 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
766 | * after it. | |
767 | */ | |
768 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, | |
769 | const struct file_operations *fop) | |
770 | { | |
771 | struct inode *inode = dp->d_inode; | |
772 | int (*fsync) (struct file *, struct dentry *, int); | |
773 | int err; | |
774 | ||
775 | err = filemap_fdatawrite(inode->i_mapping); | |
776 | if (err == 0 && fop && (fsync = fop->fsync)) | |
777 | err = fsync(filp, dp, 0); | |
778 | if (err == 0) | |
779 | err = filemap_fdatawait(inode->i_mapping); | |
780 | ||
781 | return err; | |
782 | } | |
783 | ||
a334de28 | 784 | static int |
1da177e4 LT |
785 | nfsd_sync(struct file *filp) |
786 | { | |
9a8d248e BF |
787 | int err; |
788 | struct inode *inode = filp->f_path.dentry->d_inode; | |
789 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
790 | mutex_lock(&inode->i_mutex); | |
791 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); | |
792 | mutex_unlock(&inode->i_mutex); | |
793 | ||
794 | return err; | |
1da177e4 LT |
795 | } |
796 | ||
f193fbab | 797 | int |
9a8d248e | 798 | nfsd_sync_dir(struct dentry *dp) |
1da177e4 | 799 | { |
9a8d248e | 800 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
801 | } |
802 | ||
803 | /* | |
804 | * Obtain the readahead parameters for the file | |
805 | * specified by (dev, ino). | |
806 | */ | |
1da177e4 LT |
807 | |
808 | static inline struct raparms * | |
809 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
810 | { | |
811 | struct raparms *ra, **rap, **frap = NULL; | |
812 | int depth = 0; | |
fce1456a GB |
813 | unsigned int hash; |
814 | struct raparm_hbucket *rab; | |
815 | ||
816 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
817 | rab = &raparm_hash[hash]; | |
1da177e4 | 818 | |
fce1456a GB |
819 | spin_lock(&rab->pb_lock); |
820 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
821 | if (ra->p_ino == ino && ra->p_dev == dev) |
822 | goto found; | |
823 | depth++; | |
824 | if (ra->p_count == 0) | |
825 | frap = rap; | |
826 | } | |
827 | depth = nfsdstats.ra_size*11/10; | |
828 | if (!frap) { | |
fce1456a | 829 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
830 | return NULL; |
831 | } | |
832 | rap = frap; | |
833 | ra = *frap; | |
834 | ra->p_dev = dev; | |
835 | ra->p_ino = ino; | |
836 | ra->p_set = 0; | |
fce1456a | 837 | ra->p_hindex = hash; |
1da177e4 | 838 | found: |
fce1456a | 839 | if (rap != &rab->pb_head) { |
1da177e4 | 840 | *rap = ra->p_next; |
fce1456a GB |
841 | ra->p_next = rab->pb_head; |
842 | rab->pb_head = ra; | |
1da177e4 LT |
843 | } |
844 | ra->p_count++; | |
845 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 846 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
847 | return ra; |
848 | } | |
849 | ||
850 | /* | |
cf8208d0 JA |
851 | * Grab and keep cached pages associated with a file in the svc_rqst |
852 | * so that they can be passed to the network sendmsg/sendpage routines | |
853 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
854 | */ |
855 | static int | |
cf8208d0 JA |
856 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
857 | struct splice_desc *sd) | |
1da177e4 | 858 | { |
cf8208d0 | 859 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 860 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
861 | struct page *page = buf->page; |
862 | size_t size; | |
863 | int ret; | |
1da177e4 | 864 | |
cac36bb0 | 865 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
866 | if (unlikely(ret)) |
867 | return ret; | |
868 | ||
869 | size = sd->len; | |
1da177e4 LT |
870 | |
871 | if (rqstp->rq_res.page_len == 0) { | |
872 | get_page(page); | |
44524359 N |
873 | put_page(*pp); |
874 | *pp = page; | |
875 | rqstp->rq_resused++; | |
cf8208d0 | 876 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 877 | rqstp->rq_res.page_len = size; |
44524359 | 878 | } else if (page != pp[-1]) { |
1da177e4 | 879 | get_page(page); |
250f3915 N |
880 | if (*pp) |
881 | put_page(*pp); | |
44524359 N |
882 | *pp = page; |
883 | rqstp->rq_resused++; | |
1da177e4 | 884 | rqstp->rq_res.page_len += size; |
44524359 | 885 | } else |
1da177e4 | 886 | rqstp->rq_res.page_len += size; |
1da177e4 | 887 | |
1da177e4 LT |
888 | return size; |
889 | } | |
890 | ||
cf8208d0 JA |
891 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
892 | struct splice_desc *sd) | |
893 | { | |
894 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
895 | } | |
896 | ||
a8754bee DH |
897 | static inline int svc_msnfs(struct svc_fh *ffhp) |
898 | { | |
899 | #ifdef MSNFS | |
900 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
901 | #else | |
902 | return 0; | |
903 | #endif | |
904 | } | |
905 | ||
6264d69d | 906 | static __be32 |
1da177e4 LT |
907 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
908 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
909 | { | |
910 | struct inode *inode; | |
911 | struct raparms *ra; | |
912 | mm_segment_t oldfs; | |
6264d69d AV |
913 | __be32 err; |
914 | int host_err; | |
1da177e4 LT |
915 | |
916 | err = nfserr_perm; | |
7eaa36e2 | 917 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
918 | |
919 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 920 | goto out; |
1da177e4 LT |
921 | |
922 | /* Get readahead parameters */ | |
923 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
924 | ||
925 | if (ra && ra->p_set) | |
926 | file->f_ra = ra->p_ra; | |
927 | ||
cf8208d0 JA |
928 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
929 | struct splice_desc sd = { | |
930 | .len = 0, | |
931 | .total_len = *count, | |
932 | .pos = offset, | |
933 | .u.data = rqstp, | |
934 | }; | |
935 | ||
4fbef206 | 936 | rqstp->rq_resused = 1; |
cf8208d0 | 937 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
938 | } else { |
939 | oldfs = get_fs(); | |
940 | set_fs(KERNEL_DS); | |
6264d69d | 941 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
942 | set_fs(oldfs); |
943 | } | |
944 | ||
945 | /* Write back readahead params */ | |
946 | if (ra) { | |
fce1456a GB |
947 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
948 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
949 | ra->p_ra = file->f_ra; |
950 | ra->p_set = 1; | |
951 | ra->p_count--; | |
fce1456a | 952 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
953 | } |
954 | ||
6264d69d AV |
955 | if (host_err >= 0) { |
956 | nfsdstats.io_read += host_err; | |
957 | *count = host_err; | |
1da177e4 | 958 | err = 0; |
7eaa36e2 | 959 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 960 | } else |
6264d69d | 961 | err = nfserrno(host_err); |
1da177e4 LT |
962 | out: |
963 | return err; | |
964 | } | |
965 | ||
9f708e40 NB |
966 | static void kill_suid(struct dentry *dentry) |
967 | { | |
968 | struct iattr ia; | |
b5376771 | 969 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 970 | |
1b1dcc1b | 971 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 972 | notify_change(dentry, &ia); |
1b1dcc1b | 973 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
974 | } |
975 | ||
d911df7b BF |
976 | /* |
977 | * Gathered writes: If another process is currently writing to the file, | |
978 | * there's a high chance this is another nfsd (triggered by a bulk write | |
979 | * from a client's biod). Rather than syncing the file with each write | |
980 | * request, we sleep for 10 msec. | |
981 | * | |
982 | * I don't know if this roughly approximates C. Juszak's idea of | |
983 | * gathered writes, but it's a nice and simple solution (IMHO), and it | |
984 | * seems to work:-) | |
985 | * | |
986 | * Note: we do this only in the NFSv2 case, since v3 and higher have a | |
987 | * better tool (separate unstable writes and commits) for solving this | |
988 | * problem. | |
989 | */ | |
990 | static int wait_for_concurrent_writes(struct file *file) | |
991 | { | |
992 | struct inode *inode = file->f_path.dentry->d_inode; | |
993 | static ino_t last_ino; | |
994 | static dev_t last_dev; | |
995 | int err = 0; | |
996 | ||
997 | if (atomic_read(&inode->i_writecount) > 1 | |
998 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
999 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); | |
1000 | msleep(10); | |
1001 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); | |
1002 | } | |
1003 | ||
1004 | if (inode->i_state & I_DIRTY) { | |
1005 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); | |
1006 | err = nfsd_sync(file); | |
1007 | } | |
1008 | last_ino = inode->i_ino; | |
1009 | last_dev = inode->i_sb->s_dev; | |
1010 | return err; | |
1011 | } | |
1012 | ||
6264d69d | 1013 | static __be32 |
1da177e4 LT |
1014 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1015 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 1016 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
1017 | { |
1018 | struct svc_export *exp; | |
1019 | struct dentry *dentry; | |
1020 | struct inode *inode; | |
1021 | mm_segment_t oldfs; | |
6264d69d AV |
1022 | __be32 err = 0; |
1023 | int host_err; | |
1da177e4 | 1024 | int stable = *stablep; |
48e03bc5 | 1025 | int use_wgather; |
1da177e4 | 1026 | |
45bd3b3d | 1027 | #ifdef MSNFS |
1da177e4 LT |
1028 | err = nfserr_perm; |
1029 | ||
1da177e4 | 1030 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
31dec253 | 1031 | (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt))) |
1da177e4 LT |
1032 | goto out; |
1033 | #endif | |
1034 | ||
7eaa36e2 | 1035 | dentry = file->f_path.dentry; |
1da177e4 LT |
1036 | inode = dentry->d_inode; |
1037 | exp = fhp->fh_export; | |
1038 | ||
1039 | /* | |
1040 | * Request sync writes if | |
1041 | * - the sync export option has been set, or | |
1042 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
1043 | * - The file system doesn't support fsync(). | |
48e03bc5 | 1044 | * When NFSv2 gathered writes have been configured for this volume, |
1da177e4 LT |
1045 | * flushing the data to disk is handled separately below. |
1046 | */ | |
48e03bc5 | 1047 | use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp); |
1da177e4 | 1048 | |
3ba15148 | 1049 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
1050 | stable = 2; |
1051 | *stablep = 2; /* FILE_SYNC */ | |
1052 | } | |
1053 | ||
1054 | if (!EX_ISSYNC(exp)) | |
1055 | stable = 0; | |
48e03bc5 | 1056 | if (stable && !use_wgather) { |
db1dd4d3 | 1057 | spin_lock(&file->f_lock); |
1da177e4 | 1058 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
1059 | spin_unlock(&file->f_lock); |
1060 | } | |
1da177e4 LT |
1061 | |
1062 | /* Write the data. */ | |
1063 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1064 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1065 | set_fs(oldfs); |
e4636d53 BF |
1066 | if (host_err < 0) |
1067 | goto out_nfserr; | |
1068 | *cnt = host_err; | |
1069 | nfsdstats.io_write += host_err; | |
1070 | fsnotify_modify(file->f_path.dentry); | |
1da177e4 LT |
1071 | |
1072 | /* clear setuid/setgid flag after write */ | |
e4636d53 | 1073 | if (inode->i_mode & (S_ISUID | S_ISGID)) |
9f708e40 | 1074 | kill_suid(dentry); |
1da177e4 | 1075 | |
e4636d53 | 1076 | if (stable && use_wgather) |
d911df7b | 1077 | host_err = wait_for_concurrent_writes(file); |
1da177e4 | 1078 | |
e4636d53 | 1079 | out_nfserr: |
6264d69d | 1080 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1081 | if (host_err >= 0) |
1da177e4 | 1082 | err = 0; |
a0d24b29 | 1083 | else |
6264d69d | 1084 | err = nfserrno(host_err); |
1da177e4 LT |
1085 | out: |
1086 | return err; | |
1087 | } | |
1088 | ||
1089 | /* | |
1090 | * Read data from a file. count must contain the requested read count | |
1091 | * on entry. On return, *count contains the number of bytes actually read. | |
1092 | * N.B. After this call fhp needs an fh_put | |
1093 | */ | |
6264d69d | 1094 | __be32 |
1da177e4 LT |
1095 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1096 | loff_t offset, struct kvec *vec, int vlen, | |
1097 | unsigned long *count) | |
1098 | { | |
6264d69d | 1099 | __be32 err; |
1da177e4 LT |
1100 | |
1101 | if (file) { | |
0ec757df | 1102 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1103 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1104 | if (err) |
1105 | goto out; | |
1106 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1107 | } else { | |
8837abca | 1108 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); |
1da177e4 LT |
1109 | if (err) |
1110 | goto out; | |
1111 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1112 | nfsd_close(file); | |
1113 | } | |
1114 | out: | |
1115 | return err; | |
1116 | } | |
1117 | ||
1118 | /* | |
1119 | * Write data to a file. | |
1120 | * The stable flag requests synchronous writes. | |
1121 | * N.B. After this call fhp needs an fh_put | |
1122 | */ | |
6264d69d | 1123 | __be32 |
1da177e4 | 1124 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1125 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1126 | int *stablep) |
1127 | { | |
6264d69d | 1128 | __be32 err = 0; |
1da177e4 LT |
1129 | |
1130 | if (file) { | |
0ec757df | 1131 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1132 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1133 | if (err) |
1134 | goto out; | |
1135 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1136 | stablep); | |
1137 | } else { | |
8837abca | 1138 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1139 | if (err) |
1140 | goto out; | |
1141 | ||
1142 | if (cnt) | |
1143 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1144 | cnt, stablep); | |
1145 | nfsd_close(file); | |
1146 | } | |
1147 | out: | |
1148 | return err; | |
1149 | } | |
1150 | ||
1151 | #ifdef CONFIG_NFSD_V3 | |
1152 | /* | |
1153 | * Commit all pending writes to stable storage. | |
1154 | * Strictly speaking, we could sync just the indicated file region here, | |
1155 | * but there's currently no way we can ask the VFS to do so. | |
1156 | * | |
1157 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1158 | * data to the client, as locking happens lower down in the filesystem. | |
1159 | */ | |
6264d69d | 1160 | __be32 |
1da177e4 LT |
1161 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1162 | loff_t offset, unsigned long count) | |
1163 | { | |
1164 | struct file *file; | |
6264d69d | 1165 | __be32 err; |
1da177e4 LT |
1166 | |
1167 | if ((u64)count > ~(u64)offset) | |
1168 | return nfserr_inval; | |
1169 | ||
8837abca MS |
1170 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1171 | if (err) | |
1da177e4 LT |
1172 | return err; |
1173 | if (EX_ISSYNC(fhp->fh_export)) { | |
1174 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1175 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1176 | } else { |
1177 | err = nfserr_notsupp; | |
1178 | } | |
1179 | } | |
1180 | ||
1181 | nfsd_close(file); | |
1182 | return err; | |
1183 | } | |
1184 | #endif /* CONFIG_NFSD_V3 */ | |
1185 | ||
f2b0dee2 | 1186 | static __be32 |
5c002b3b BF |
1187 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1188 | struct iattr *iap) | |
1189 | { | |
1190 | /* | |
1191 | * Mode has already been set earlier in create: | |
1192 | */ | |
1193 | iap->ia_valid &= ~ATTR_MODE; | |
1194 | /* | |
1195 | * Setting uid/gid works only for root. Irix appears to | |
1196 | * send along the gid on create when it tries to implement | |
1197 | * setgid directories via NFS: | |
1198 | */ | |
5cc0a840 | 1199 | if (current_fsuid() != 0) |
5c002b3b BF |
1200 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1201 | if (iap->ia_valid) | |
1202 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1203 | return 0; | |
1204 | } | |
1205 | ||
4ac35c2f | 1206 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1207 | * setting size to 0 may fail for some specific file systems by the permission | |
1208 | * checking which requires WRITE permission but the mode is 000. | |
1209 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1210 | * 0 after file created. | |
1211 | * | |
1212 | * call this only after vfs_create() is called. | |
1213 | * */ | |
1214 | static void | |
1215 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1216 | { | |
1217 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1218 | iap->ia_valid &= ~ATTR_SIZE; | |
1219 | } | |
1220 | ||
1da177e4 LT |
1221 | /* |
1222 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1223 | * not yet implemented. | |
1224 | * If the response fh has been verified, the parent directory should | |
1225 | * already be locked. Note that the parent directory is left locked. | |
1226 | * | |
1227 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1228 | */ | |
6264d69d | 1229 | __be32 |
1da177e4 LT |
1230 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1231 | char *fname, int flen, struct iattr *iap, | |
1232 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1233 | { | |
1234 | struct dentry *dentry, *dchild = NULL; | |
1235 | struct inode *dirp; | |
6264d69d | 1236 | __be32 err; |
5c002b3b | 1237 | __be32 err2; |
6264d69d | 1238 | int host_err; |
1da177e4 LT |
1239 | |
1240 | err = nfserr_perm; | |
1241 | if (!flen) | |
1242 | goto out; | |
1243 | err = nfserr_exist; | |
1244 | if (isdotent(fname, flen)) | |
1245 | goto out; | |
1246 | ||
8837abca | 1247 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1248 | if (err) |
1249 | goto out; | |
1250 | ||
1251 | dentry = fhp->fh_dentry; | |
1252 | dirp = dentry->d_inode; | |
1253 | ||
1254 | err = nfserr_notdir; | |
acfa4380 | 1255 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1256 | goto out; |
1257 | /* | |
1258 | * Check whether the response file handle has been verified yet. | |
1259 | * If it has, the parent directory should already be locked. | |
1260 | */ | |
1261 | if (!resfhp->fh_dentry) { | |
1262 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1263 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1264 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1265 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1266 | if (IS_ERR(dchild)) |
1267 | goto out_nfserr; | |
1268 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1269 | if (err) | |
1270 | goto out; | |
1271 | } else { | |
1272 | /* called from nfsd_proc_create */ | |
1273 | dchild = dget(resfhp->fh_dentry); | |
1274 | if (!fhp->fh_locked) { | |
1275 | /* not actually possible */ | |
1276 | printk(KERN_ERR | |
1277 | "nfsd_create: parent %s/%s not locked!\n", | |
1278 | dentry->d_parent->d_name.name, | |
1279 | dentry->d_name.name); | |
d75f2b9f | 1280 | err = nfserr_io; |
1da177e4 LT |
1281 | goto out; |
1282 | } | |
1283 | } | |
1284 | /* | |
1285 | * Make sure the child dentry is still negative ... | |
1286 | */ | |
1287 | err = nfserr_exist; | |
1288 | if (dchild->d_inode) { | |
1289 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1290 | dentry->d_name.name, dchild->d_name.name); | |
1291 | goto out; | |
1292 | } | |
1293 | ||
1294 | if (!(iap->ia_valid & ATTR_MODE)) | |
1295 | iap->ia_mode = 0; | |
1296 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1297 | ||
07cad1d2 MS |
1298 | err = nfserr_inval; |
1299 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1300 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1301 | type); | |
1302 | goto out; | |
1303 | } | |
1304 | ||
1305 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1306 | if (host_err) | |
1307 | goto out_nfserr; | |
1308 | ||
1da177e4 LT |
1309 | /* |
1310 | * Get the dir op function pointer. | |
1311 | */ | |
088406bc | 1312 | err = 0; |
1da177e4 LT |
1313 | switch (type) { |
1314 | case S_IFREG: | |
6264d69d | 1315 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1316 | if (!host_err) |
1317 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1318 | break; |
1319 | case S_IFDIR: | |
6264d69d | 1320 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1321 | break; |
1322 | case S_IFCHR: | |
1323 | case S_IFBLK: | |
1324 | case S_IFIFO: | |
1325 | case S_IFSOCK: | |
6264d69d | 1326 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1327 | break; |
1da177e4 | 1328 | } |
463c3197 DH |
1329 | if (host_err < 0) { |
1330 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1331 | goto out_nfserr; |
463c3197 | 1332 | } |
1da177e4 LT |
1333 | |
1334 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1335 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1336 | write_inode_now(dchild->d_inode, 1); |
1337 | } | |
1338 | ||
5c002b3b BF |
1339 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1340 | if (err2) | |
1341 | err = err2; | |
463c3197 | 1342 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1343 | /* |
1344 | * Update the file handle to get the new inode info. | |
1345 | */ | |
1346 | if (!err) | |
1347 | err = fh_update(resfhp); | |
1348 | out: | |
1349 | if (dchild && !IS_ERR(dchild)) | |
1350 | dput(dchild); | |
1351 | return err; | |
1352 | ||
1353 | out_nfserr: | |
6264d69d | 1354 | err = nfserrno(host_err); |
1da177e4 LT |
1355 | goto out; |
1356 | } | |
1357 | ||
1358 | #ifdef CONFIG_NFSD_V3 | |
1359 | /* | |
1360 | * NFSv3 version of nfsd_create | |
1361 | */ | |
6264d69d | 1362 | __be32 |
1da177e4 LT |
1363 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1364 | char *fname, int flen, struct iattr *iap, | |
1365 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1366 | int *truncp, int *created) |
1da177e4 LT |
1367 | { |
1368 | struct dentry *dentry, *dchild = NULL; | |
1369 | struct inode *dirp; | |
6264d69d | 1370 | __be32 err; |
5c002b3b | 1371 | __be32 err2; |
6264d69d | 1372 | int host_err; |
1da177e4 | 1373 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1374 | |
1375 | err = nfserr_perm; | |
1376 | if (!flen) | |
1377 | goto out; | |
1378 | err = nfserr_exist; | |
1379 | if (isdotent(fname, flen)) | |
1380 | goto out; | |
1381 | if (!(iap->ia_valid & ATTR_MODE)) | |
1382 | iap->ia_mode = 0; | |
8837abca | 1383 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1384 | if (err) |
1385 | goto out; | |
1386 | ||
1387 | dentry = fhp->fh_dentry; | |
1388 | dirp = dentry->d_inode; | |
1389 | ||
1390 | /* Get all the sanity checks out of the way before | |
1391 | * we lock the parent. */ | |
1392 | err = nfserr_notdir; | |
acfa4380 | 1393 | if (!dirp->i_op->lookup) |
1da177e4 | 1394 | goto out; |
12fd3520 | 1395 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1396 | |
1397 | /* | |
1398 | * Compose the response file handle. | |
1399 | */ | |
1400 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1401 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1402 | if (IS_ERR(dchild)) |
1403 | goto out_nfserr; | |
1404 | ||
1405 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1406 | if (err) | |
1407 | goto out; | |
1408 | ||
1409 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1410 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1411 | * the high bit set, so just clear the high bits. If this is |
1412 | * ever changed to use different attrs for storing the | |
1413 | * verifier, then do_open_lookup() will also need to be fixed | |
1414 | * accordingly. | |
1da177e4 LT |
1415 | */ |
1416 | v_mtime = verifier[0]&0x7fffffff; | |
1417 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1418 | } |
1419 | ||
463c3197 DH |
1420 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1421 | if (host_err) | |
1422 | goto out_nfserr; | |
1da177e4 LT |
1423 | if (dchild->d_inode) { |
1424 | err = 0; | |
1425 | ||
1426 | switch (createmode) { | |
1427 | case NFS3_CREATE_UNCHECKED: | |
1428 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1429 | err = nfserr_exist; | |
1430 | else if (truncp) { | |
1431 | /* in nfsv4, we need to treat this case a little | |
1432 | * differently. we don't want to truncate the | |
1433 | * file now; this would be wrong if the OPEN | |
1434 | * fails for some other reason. furthermore, | |
1435 | * if the size is nonzero, we should ignore it | |
1436 | * according to spec! | |
1437 | */ | |
1438 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1439 | } | |
1440 | else { | |
1441 | iap->ia_valid &= ATTR_SIZE; | |
1442 | goto set_attr; | |
1443 | } | |
1444 | break; | |
1445 | case NFS3_CREATE_EXCLUSIVE: | |
1446 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1447 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1448 | && dchild->d_inode->i_size == 0 ) |
1449 | break; | |
1450 | /* fallthru */ | |
1451 | case NFS3_CREATE_GUARDED: | |
1452 | err = nfserr_exist; | |
1453 | } | |
463c3197 | 1454 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1455 | goto out; |
1456 | } | |
1457 | ||
6264d69d | 1458 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1459 | if (host_err < 0) { |
1460 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1461 | goto out_nfserr; |
463c3197 | 1462 | } |
81ac95c5 BF |
1463 | if (created) |
1464 | *created = 1; | |
1da177e4 LT |
1465 | |
1466 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1467 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1468 | /* setattr will sync the child (or not) */ |
1469 | } | |
1470 | ||
4ac35c2f | 1471 | nfsd_check_ignore_resizing(iap); |
1472 | ||
1da177e4 | 1473 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1474 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1475 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1476 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1477 | /* XXX someone who knows this better please fix it for nsec */ |
1478 | iap->ia_mtime.tv_sec = v_mtime; | |
1479 | iap->ia_atime.tv_sec = v_atime; | |
1480 | iap->ia_mtime.tv_nsec = 0; | |
1481 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1482 | } |
1483 | ||
1da177e4 | 1484 | set_attr: |
5c002b3b BF |
1485 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1486 | if (err2) | |
1487 | err = err2; | |
f193fbab | 1488 | |
463c3197 | 1489 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1490 | /* |
1491 | * Update the filehandle to get the new inode info. | |
1492 | */ | |
1493 | if (!err) | |
1494 | err = fh_update(resfhp); | |
1da177e4 LT |
1495 | |
1496 | out: | |
1497 | fh_unlock(fhp); | |
1498 | if (dchild && !IS_ERR(dchild)) | |
1499 | dput(dchild); | |
1500 | return err; | |
1501 | ||
1502 | out_nfserr: | |
6264d69d | 1503 | err = nfserrno(host_err); |
1da177e4 LT |
1504 | goto out; |
1505 | } | |
1506 | #endif /* CONFIG_NFSD_V3 */ | |
1507 | ||
1508 | /* | |
1509 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1510 | * fits into the buffer. On return, it contains the true length. | |
1511 | * N.B. After this call fhp needs an fh_put | |
1512 | */ | |
6264d69d | 1513 | __be32 |
1da177e4 LT |
1514 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1515 | { | |
1516 | struct dentry *dentry; | |
1517 | struct inode *inode; | |
1518 | mm_segment_t oldfs; | |
6264d69d AV |
1519 | __be32 err; |
1520 | int host_err; | |
1da177e4 | 1521 | |
8837abca | 1522 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1523 | if (err) |
1524 | goto out; | |
1525 | ||
1526 | dentry = fhp->fh_dentry; | |
1527 | inode = dentry->d_inode; | |
1528 | ||
1529 | err = nfserr_inval; | |
acfa4380 | 1530 | if (!inode->i_op->readlink) |
1da177e4 LT |
1531 | goto out; |
1532 | ||
54775491 | 1533 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1534 | /* N.B. Why does this call need a get_fs()?? |
1535 | * Remove the set_fs and watch the fireworks:-) --okir | |
1536 | */ | |
1537 | ||
1538 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1539 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1540 | set_fs(oldfs); |
1541 | ||
6264d69d | 1542 | if (host_err < 0) |
1da177e4 | 1543 | goto out_nfserr; |
6264d69d | 1544 | *lenp = host_err; |
1da177e4 LT |
1545 | err = 0; |
1546 | out: | |
1547 | return err; | |
1548 | ||
1549 | out_nfserr: | |
6264d69d | 1550 | err = nfserrno(host_err); |
1da177e4 LT |
1551 | goto out; |
1552 | } | |
1553 | ||
1554 | /* | |
1555 | * Create a symlink and look up its inode | |
1556 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1557 | */ | |
6264d69d | 1558 | __be32 |
1da177e4 LT |
1559 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1560 | char *fname, int flen, | |
1561 | char *path, int plen, | |
1562 | struct svc_fh *resfhp, | |
1563 | struct iattr *iap) | |
1564 | { | |
1565 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1566 | __be32 err, cerr; |
1567 | int host_err; | |
1da177e4 LT |
1568 | |
1569 | err = nfserr_noent; | |
1570 | if (!flen || !plen) | |
1571 | goto out; | |
1572 | err = nfserr_exist; | |
1573 | if (isdotent(fname, flen)) | |
1574 | goto out; | |
1575 | ||
8837abca | 1576 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1577 | if (err) |
1578 | goto out; | |
1579 | fh_lock(fhp); | |
1580 | dentry = fhp->fh_dentry; | |
1581 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1582 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1583 | if (IS_ERR(dnew)) |
1584 | goto out_nfserr; | |
1585 | ||
75c3f29d DH |
1586 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1587 | if (host_err) | |
1588 | goto out_nfserr; | |
1589 | ||
1da177e4 LT |
1590 | if (unlikely(path[plen] != 0)) { |
1591 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1592 | if (path_alloced == NULL) | |
6264d69d | 1593 | host_err = -ENOMEM; |
1da177e4 LT |
1594 | else { |
1595 | strncpy(path_alloced, path, plen); | |
1596 | path_alloced[plen] = 0; | |
db2e747b | 1597 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1598 | kfree(path_alloced); |
1599 | } | |
1600 | } else | |
db2e747b | 1601 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
1da177e4 | 1602 | |
6264d69d | 1603 | if (!host_err) { |
1da177e4 | 1604 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1605 | host_err = nfsd_sync_dir(dentry); |
1606 | } | |
1607 | err = nfserrno(host_err); | |
1da177e4 LT |
1608 | fh_unlock(fhp); |
1609 | ||
75c3f29d DH |
1610 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1611 | ||
1da177e4 LT |
1612 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1613 | dput(dnew); | |
1614 | if (err==0) err = cerr; | |
1615 | out: | |
1616 | return err; | |
1617 | ||
1618 | out_nfserr: | |
6264d69d | 1619 | err = nfserrno(host_err); |
1da177e4 LT |
1620 | goto out; |
1621 | } | |
1622 | ||
1623 | /* | |
1624 | * Create a hardlink | |
1625 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1626 | */ | |
6264d69d | 1627 | __be32 |
1da177e4 LT |
1628 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1629 | char *name, int len, struct svc_fh *tfhp) | |
1630 | { | |
1631 | struct dentry *ddir, *dnew, *dold; | |
1632 | struct inode *dirp, *dest; | |
6264d69d AV |
1633 | __be32 err; |
1634 | int host_err; | |
1da177e4 | 1635 | |
8837abca | 1636 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1637 | if (err) |
1638 | goto out; | |
8837abca | 1639 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1640 | if (err) |
1641 | goto out; | |
1642 | ||
1643 | err = nfserr_perm; | |
1644 | if (!len) | |
1645 | goto out; | |
1646 | err = nfserr_exist; | |
1647 | if (isdotent(name, len)) | |
1648 | goto out; | |
1649 | ||
12fd3520 | 1650 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1651 | ddir = ffhp->fh_dentry; |
1652 | dirp = ddir->d_inode; | |
1653 | ||
1654 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1655 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1656 | if (IS_ERR(dnew)) |
1657 | goto out_nfserr; | |
1658 | ||
1659 | dold = tfhp->fh_dentry; | |
1660 | dest = dold->d_inode; | |
1661 | ||
75c3f29d DH |
1662 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1663 | if (host_err) { | |
1664 | err = nfserrno(host_err); | |
1665 | goto out_dput; | |
1666 | } | |
6264d69d AV |
1667 | host_err = vfs_link(dold, dirp, dnew); |
1668 | if (!host_err) { | |
1da177e4 | 1669 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1670 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1671 | write_inode_now(dest, 1); |
1672 | } | |
6264d69d | 1673 | err = 0; |
1da177e4 | 1674 | } else { |
6264d69d | 1675 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1676 | err = nfserr_acces; |
1677 | else | |
6264d69d | 1678 | err = nfserrno(host_err); |
1da177e4 | 1679 | } |
75c3f29d DH |
1680 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1681 | out_dput: | |
1da177e4 | 1682 | dput(dnew); |
270d56e5 DR |
1683 | out_unlock: |
1684 | fh_unlock(ffhp); | |
1da177e4 LT |
1685 | out: |
1686 | return err; | |
1687 | ||
1688 | out_nfserr: | |
6264d69d | 1689 | err = nfserrno(host_err); |
270d56e5 | 1690 | goto out_unlock; |
1da177e4 LT |
1691 | } |
1692 | ||
1693 | /* | |
1694 | * Rename a file | |
1695 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1696 | */ | |
6264d69d | 1697 | __be32 |
1da177e4 LT |
1698 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1699 | struct svc_fh *tfhp, char *tname, int tlen) | |
1700 | { | |
1701 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1702 | struct inode *fdir, *tdir; | |
6264d69d AV |
1703 | __be32 err; |
1704 | int host_err; | |
1da177e4 | 1705 | |
8837abca | 1706 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1707 | if (err) |
1708 | goto out; | |
8837abca | 1709 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1710 | if (err) |
1711 | goto out; | |
1712 | ||
1713 | fdentry = ffhp->fh_dentry; | |
1714 | fdir = fdentry->d_inode; | |
1715 | ||
1716 | tdentry = tfhp->fh_dentry; | |
1717 | tdir = tdentry->d_inode; | |
1718 | ||
1719 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1720 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1721 | goto out; |
1722 | ||
1723 | err = nfserr_perm; | |
1724 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1725 | goto out; | |
1726 | ||
1727 | /* cannot use fh_lock as we need deadlock protective ordering | |
1728 | * so do it by hand */ | |
1729 | trap = lock_rename(tdentry, fdentry); | |
1730 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1731 | fill_pre_wcc(ffhp); | |
1732 | fill_pre_wcc(tfhp); | |
1733 | ||
1734 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1735 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1736 | if (IS_ERR(odentry)) |
1737 | goto out_nfserr; | |
1738 | ||
6264d69d | 1739 | host_err = -ENOENT; |
1da177e4 LT |
1740 | if (!odentry->d_inode) |
1741 | goto out_dput_old; | |
6264d69d | 1742 | host_err = -EINVAL; |
1da177e4 LT |
1743 | if (odentry == trap) |
1744 | goto out_dput_old; | |
1745 | ||
1746 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1747 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1748 | if (IS_ERR(ndentry)) |
1749 | goto out_dput_old; | |
6264d69d | 1750 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1751 | if (ndentry == trap) |
1752 | goto out_dput_new; | |
1753 | ||
9079b1eb | 1754 | if (svc_msnfs(ffhp) && |
1da177e4 LT |
1755 | ((atomic_read(&odentry->d_count) > 1) |
1756 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1757 | host_err = -EPERM; |
9079b1eb DH |
1758 | goto out_dput_new; |
1759 | } | |
1760 | ||
1761 | host_err = -EXDEV; | |
1762 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1763 | goto out_dput_new; | |
1764 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1765 | if (host_err) | |
1766 | goto out_dput_new; | |
1767 | ||
6264d69d AV |
1768 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1769 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1770 | host_err = nfsd_sync_dir(tdentry); | |
1771 | if (!host_err) | |
1772 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1773 | } |
1774 | ||
9079b1eb DH |
1775 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1776 | ||
1da177e4 LT |
1777 | out_dput_new: |
1778 | dput(ndentry); | |
1779 | out_dput_old: | |
1780 | dput(odentry); | |
1781 | out_nfserr: | |
6264d69d | 1782 | err = nfserrno(host_err); |
1da177e4 LT |
1783 | |
1784 | /* we cannot reply on fh_unlock on the two filehandles, | |
1785 | * as that would do the wrong thing if the two directories | |
1786 | * were the same, so again we do it by hand | |
1787 | */ | |
1788 | fill_post_wcc(ffhp); | |
1789 | fill_post_wcc(tfhp); | |
1790 | unlock_rename(tdentry, fdentry); | |
1791 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1792 | ||
1793 | out: | |
1794 | return err; | |
1795 | } | |
1796 | ||
1797 | /* | |
1798 | * Unlink a file or directory | |
1799 | * N.B. After this call fhp needs an fh_put | |
1800 | */ | |
6264d69d | 1801 | __be32 |
1da177e4 LT |
1802 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1803 | char *fname, int flen) | |
1804 | { | |
1805 | struct dentry *dentry, *rdentry; | |
1806 | struct inode *dirp; | |
6264d69d AV |
1807 | __be32 err; |
1808 | int host_err; | |
1da177e4 LT |
1809 | |
1810 | err = nfserr_acces; | |
1811 | if (!flen || isdotent(fname, flen)) | |
1812 | goto out; | |
8837abca | 1813 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1814 | if (err) |
1815 | goto out; | |
1816 | ||
12fd3520 | 1817 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1818 | dentry = fhp->fh_dentry; |
1819 | dirp = dentry->d_inode; | |
1820 | ||
1821 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1822 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1823 | if (IS_ERR(rdentry)) |
1824 | goto out_nfserr; | |
1825 | ||
1826 | if (!rdentry->d_inode) { | |
1827 | dput(rdentry); | |
1828 | err = nfserr_noent; | |
1829 | goto out; | |
1830 | } | |
1831 | ||
1832 | if (!type) | |
1833 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1834 | ||
0622753b DH |
1835 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1836 | if (host_err) | |
1837 | goto out_nfserr; | |
1838 | ||
1da177e4 LT |
1839 | if (type != S_IFDIR) { /* It's UNLINK */ |
1840 | #ifdef MSNFS | |
1841 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1842 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1843 | host_err = -EPERM; |
1da177e4 LT |
1844 | } else |
1845 | #endif | |
6264d69d | 1846 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1847 | } else { /* It's RMDIR */ |
6264d69d | 1848 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1849 | } |
1850 | ||
1851 | dput(rdentry); | |
1852 | ||
6264d69d | 1853 | if (host_err) |
0622753b | 1854 | goto out_drop; |
6264d69d AV |
1855 | if (EX_ISSYNC(fhp->fh_export)) |
1856 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 | 1857 | |
0622753b DH |
1858 | out_drop: |
1859 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1860 | out_nfserr: |
6264d69d | 1861 | err = nfserrno(host_err); |
f193fbab YT |
1862 | out: |
1863 | return err; | |
1da177e4 LT |
1864 | } |
1865 | ||
14f7dd63 DW |
1866 | /* |
1867 | * We do this buffering because we must not call back into the file | |
1868 | * system's ->lookup() method from the filldir callback. That may well | |
1869 | * deadlock a number of file systems. | |
1870 | * | |
1871 | * This is based heavily on the implementation of same in XFS. | |
1872 | */ | |
1873 | struct buffered_dirent { | |
1874 | u64 ino; | |
1875 | loff_t offset; | |
1876 | int namlen; | |
1877 | unsigned int d_type; | |
1878 | char name[]; | |
1879 | }; | |
1880 | ||
1881 | struct readdir_data { | |
1882 | char *dirent; | |
1883 | size_t used; | |
53c9c5c0 | 1884 | int full; |
14f7dd63 DW |
1885 | }; |
1886 | ||
1887 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1888 | loff_t offset, u64 ino, unsigned int d_type) | |
1889 | { | |
1890 | struct readdir_data *buf = __buf; | |
1891 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1892 | unsigned int reclen; | |
1893 | ||
1894 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1895 | if (buf->used + reclen > PAGE_SIZE) { |
1896 | buf->full = 1; | |
14f7dd63 | 1897 | return -EINVAL; |
53c9c5c0 | 1898 | } |
14f7dd63 DW |
1899 | |
1900 | de->namlen = namlen; | |
1901 | de->offset = offset; | |
1902 | de->ino = ino; | |
1903 | de->d_type = d_type; | |
1904 | memcpy(de->name, name, namlen); | |
1905 | buf->used += reclen; | |
1906 | ||
1907 | return 0; | |
1908 | } | |
1909 | ||
2f9092e1 DW |
1910 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
1911 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1912 | { |
14f7dd63 DW |
1913 | struct readdir_data buf; |
1914 | struct buffered_dirent *de; | |
2628b766 | 1915 | int host_err; |
14f7dd63 DW |
1916 | int size; |
1917 | loff_t offset; | |
2628b766 | 1918 | |
14f7dd63 DW |
1919 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1920 | if (!buf.dirent) | |
2f9092e1 | 1921 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
1922 | |
1923 | offset = *offsetp; | |
2628b766 | 1924 | |
14f7dd63 | 1925 | while (1) { |
2f9092e1 | 1926 | struct inode *dir_inode = file->f_path.dentry->d_inode; |
14f7dd63 DW |
1927 | unsigned int reclen; |
1928 | ||
b726e923 | 1929 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1930 | buf.used = 0; |
53c9c5c0 | 1931 | buf.full = 0; |
14f7dd63 DW |
1932 | |
1933 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1934 | if (buf.full) |
1935 | host_err = 0; | |
1936 | ||
1937 | if (host_err < 0) | |
14f7dd63 DW |
1938 | break; |
1939 | ||
1940 | size = buf.used; | |
1941 | ||
1942 | if (!size) | |
1943 | break; | |
1944 | ||
2f9092e1 DW |
1945 | /* |
1946 | * Various filldir functions may end up calling back into | |
1947 | * lookup_one_len() and the file system's ->lookup() method. | |
1948 | * These expect i_mutex to be held, as it would within readdir. | |
1949 | */ | |
1950 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
1951 | if (host_err) | |
1952 | break; | |
1953 | ||
14f7dd63 DW |
1954 | de = (struct buffered_dirent *)buf.dirent; |
1955 | while (size > 0) { | |
1956 | offset = de->offset; | |
1957 | ||
1958 | if (func(cdp, de->name, de->namlen, de->offset, | |
1959 | de->ino, de->d_type)) | |
2f9092e1 | 1960 | break; |
14f7dd63 DW |
1961 | |
1962 | if (cdp->err != nfs_ok) | |
2f9092e1 | 1963 | break; |
14f7dd63 DW |
1964 | |
1965 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1966 | sizeof(u64)); | |
1967 | size -= reclen; | |
1968 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1969 | } | |
2f9092e1 DW |
1970 | mutex_unlock(&dir_inode->i_mutex); |
1971 | if (size > 0) /* We bailed out early */ | |
1972 | break; | |
1973 | ||
c002a6c7 | 1974 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1975 | } |
1976 | ||
14f7dd63 | 1977 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
1978 | |
1979 | if (host_err) | |
1980 | return nfserrno(host_err); | |
14f7dd63 DW |
1981 | |
1982 | *offsetp = offset; | |
1983 | return cdp->err; | |
2628b766 DW |
1984 | } |
1985 | ||
1da177e4 LT |
1986 | /* |
1987 | * Read entries from a directory. | |
1988 | * The NFSv3/4 verifier we ignore for now. | |
1989 | */ | |
6264d69d | 1990 | __be32 |
1da177e4 | 1991 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1992 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1993 | { |
6264d69d | 1994 | __be32 err; |
1da177e4 LT |
1995 | struct file *file; |
1996 | loff_t offset = *offsetp; | |
1997 | ||
8837abca | 1998 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
1999 | if (err) |
2000 | goto out; | |
2001 | ||
2002 | offset = vfs_llseek(file, offset, 0); | |
2003 | if (offset < 0) { | |
2004 | err = nfserrno((int)offset); | |
2005 | goto out_close; | |
2006 | } | |
2007 | ||
14f7dd63 | 2008 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
2009 | |
2010 | if (err == nfserr_eof || err == nfserr_toosmall) | |
2011 | err = nfs_ok; /* can still be found in ->err */ | |
2012 | out_close: | |
2013 | nfsd_close(file); | |
2014 | out: | |
2015 | return err; | |
2016 | } | |
2017 | ||
2018 | /* | |
2019 | * Get file system stats | |
2020 | * N.B. After this call fhp needs an fh_put | |
2021 | */ | |
6264d69d | 2022 | __be32 |
04716e66 | 2023 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 2024 | { |
04716e66 | 2025 | __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); |
726c3342 | 2026 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
2027 | err = nfserr_io; |
2028 | return err; | |
2029 | } | |
2030 | ||
c7d51402 | 2031 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 2032 | { |
c7d51402 | 2033 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
2034 | } |
2035 | ||
1da177e4 LT |
2036 | /* |
2037 | * Check for a user's access permissions to this inode. | |
2038 | */ | |
6264d69d | 2039 | __be32 |
0ec757df BF |
2040 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2041 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2042 | { |
2043 | struct inode *inode = dentry->d_inode; | |
14dba533 | 2044 | struct path path; |
1da177e4 LT |
2045 | int err; |
2046 | ||
8837abca | 2047 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2048 | return 0; |
2049 | #if 0 | |
2050 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2051 | acc, | |
8837abca MS |
2052 | (acc & NFSD_MAY_READ)? " read" : "", |
2053 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2054 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2055 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2056 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2057 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2058 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2059 | inode->i_mode, |
2060 | IS_IMMUTABLE(inode)? " immut" : "", | |
2061 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2062 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2063 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2064 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2065 | #endif |
2066 | ||
2067 | /* Normally we reject any write/sattr etc access on a read-only file | |
2068 | * system. But if it is IRIX doing check on write-access for a | |
2069 | * device special file, we ignore rofs. | |
2070 | */ | |
8837abca MS |
2071 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2072 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2073 | if (exp_rdonly(rqstp, exp) || |
2074 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2075 | return nfserr_rofs; |
8837abca | 2076 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2077 | return nfserr_perm; |
2078 | } | |
8837abca | 2079 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2080 | return nfserr_perm; |
2081 | ||
8837abca | 2082 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2083 | /* If we cannot rely on authentication in NLM requests, |
2084 | * just allow locks, otherwise require read permission, or | |
2085 | * ownership | |
2086 | */ | |
2087 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2088 | return 0; | |
2089 | else | |
8837abca | 2090 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2091 | } |
2092 | /* | |
2093 | * The file owner always gets access permission for accesses that | |
2094 | * would normally be checked at open time. This is to make | |
2095 | * file access work even when the client has done a fchmod(fd, 0). | |
2096 | * | |
2097 | * However, `cp foo bar' should fail nevertheless when bar is | |
2098 | * readonly. A sensible way to do this might be to reject all | |
2099 | * attempts to truncate a read-only file, because a creat() call | |
2100 | * always implies file truncation. | |
2101 | * ... but this isn't really fair. A process may reasonably call | |
2102 | * ftruncate on an open file descriptor on a file with perm 000. | |
2103 | * We must trust the client to do permission checking - using "ACCESS" | |
2104 | * with NFSv3. | |
2105 | */ | |
8837abca | 2106 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2107 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2108 | return 0; |
2109 | ||
8837abca | 2110 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2111 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2112 | |
2113 | /* Allow read access to binaries even when mode 111 */ | |
2114 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2115 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2116 | err = inode_permission(inode, MAY_EXEC); |
14dba533 MZ |
2117 | if (err) |
2118 | goto nfsd_out; | |
1da177e4 | 2119 | |
14dba533 MZ |
2120 | /* Do integrity (permission) checking now, but defer incrementing |
2121 | * IMA counts to the actual file open. | |
2122 | */ | |
2123 | path.mnt = exp->ex_path.mnt; | |
2124 | path.dentry = dentry; | |
2125 | err = ima_path_check(&path, acc & (MAY_READ | MAY_WRITE | MAY_EXEC), | |
2126 | IMA_COUNT_LEAVE); | |
2127 | nfsd_out: | |
1da177e4 LT |
2128 | return err? nfserrno(err) : 0; |
2129 | } | |
2130 | ||
2131 | void | |
2132 | nfsd_racache_shutdown(void) | |
2133 | { | |
54a66e54 JL |
2134 | struct raparms *raparm, *last_raparm; |
2135 | unsigned int i; | |
2136 | ||
1da177e4 | 2137 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2138 | |
2139 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2140 | raparm = raparm_hash[i].pb_head; | |
2141 | while(raparm) { | |
2142 | last_raparm = raparm; | |
2143 | raparm = raparm->p_next; | |
2144 | kfree(last_raparm); | |
2145 | } | |
2146 | raparm_hash[i].pb_head = NULL; | |
2147 | } | |
1da177e4 LT |
2148 | } |
2149 | /* | |
2150 | * Initialize readahead param cache | |
2151 | */ | |
2152 | int | |
2153 | nfsd_racache_init(int cache_size) | |
2154 | { | |
2155 | int i; | |
fce1456a GB |
2156 | int j = 0; |
2157 | int nperbucket; | |
54a66e54 | 2158 | struct raparms **raparm = NULL; |
1da177e4 | 2159 | |
fce1456a | 2160 | |
54a66e54 | 2161 | if (raparm_hash[0].pb_head) |
1da177e4 | 2162 | return 0; |
54a66e54 JL |
2163 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2164 | if (nperbucket < 2) | |
2165 | nperbucket = 2; | |
2166 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2167 | |
2168 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2169 | |
2170 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2171 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2172 | |
2173 | raparm = &raparm_hash[i].pb_head; | |
2174 | for (j = 0; j < nperbucket; j++) { | |
2175 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2176 | if (!*raparm) | |
2177 | goto out_nomem; | |
2178 | raparm = &(*raparm)->p_next; | |
2179 | } | |
2180 | *raparm = NULL; | |
4b3bb06b YB |
2181 | } |
2182 | ||
1da177e4 LT |
2183 | nfsdstats.ra_size = cache_size; |
2184 | return 0; | |
54a66e54 JL |
2185 | |
2186 | out_nomem: | |
2187 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2188 | nfsd_racache_shutdown(); | |
2189 | return -ENOMEM; | |
1da177e4 | 2190 | } |
a257cdd0 AG |
2191 | |
2192 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2193 | struct posix_acl * | |
2194 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2195 | { | |
2196 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2197 | char *name; | |
2198 | void *value = NULL; | |
2199 | ssize_t size; | |
2200 | struct posix_acl *acl; | |
2201 | ||
5be196e5 CH |
2202 | if (!IS_POSIXACL(inode)) |
2203 | return ERR_PTR(-EOPNOTSUPP); | |
2204 | ||
2205 | switch (type) { | |
2206 | case ACL_TYPE_ACCESS: | |
2207 | name = POSIX_ACL_XATTR_ACCESS; | |
2208 | break; | |
2209 | case ACL_TYPE_DEFAULT: | |
2210 | name = POSIX_ACL_XATTR_DEFAULT; | |
2211 | break; | |
2212 | default: | |
a257cdd0 | 2213 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2214 | } |
2215 | ||
5be196e5 CH |
2216 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2217 | if (size < 0) | |
2218 | return ERR_PTR(size); | |
a257cdd0 | 2219 | |
a257cdd0 | 2220 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2221 | kfree(value); |
2222 | return acl; | |
2223 | } | |
2224 | ||
2225 | int | |
2226 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2227 | { | |
2228 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2229 | char *name; | |
2230 | void *value = NULL; | |
2231 | size_t size; | |
2232 | int error; | |
2233 | ||
acfa4380 | 2234 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2235 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2236 | return -EOPNOTSUPP; | |
2237 | switch(type) { | |
2238 | case ACL_TYPE_ACCESS: | |
334a13ec | 2239 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2240 | break; |
2241 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2242 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2243 | break; |
2244 | default: | |
2245 | return -EOPNOTSUPP; | |
2246 | } | |
2247 | ||
2248 | if (acl && acl->a_count) { | |
334a13ec | 2249 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2250 | value = kmalloc(size, GFP_KERNEL); |
2251 | if (!value) | |
2252 | return -ENOMEM; | |
9ccfc29c FM |
2253 | error = posix_acl_to_xattr(acl, value, size); |
2254 | if (error < 0) | |
a257cdd0 | 2255 | goto getout; |
9ccfc29c | 2256 | size = error; |
a257cdd0 AG |
2257 | } else |
2258 | size = 0; | |
2259 | ||
18f335af DH |
2260 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2261 | if (error) | |
2262 | goto getout; | |
a257cdd0 | 2263 | if (size) |
5be196e5 | 2264 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2265 | else { |
2266 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2267 | error = 0; | |
2268 | else { | |
5be196e5 | 2269 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2270 | if (error == -ENODATA) |
2271 | error = 0; | |
2272 | } | |
2273 | } | |
18f335af | 2274 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2275 | |
2276 | getout: | |
2277 | kfree(value); | |
2278 | return error; | |
2279 | } | |
2280 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |