]>
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> |
1da177e4 LT |
58 | |
59 | #include <asm/uaccess.h> | |
60 | ||
61 | #define NFSDDBG_FACILITY NFSDDBG_FILEOP | |
1da177e4 LT |
62 | |
63 | ||
64 | /* We must ignore files (but only files) which might have mandatory | |
65 | * locks on them because there is no way to know if the accesser has | |
66 | * the lock. | |
67 | */ | |
68 | #define IS_ISMNDLK(i) (S_ISREG((i)->i_mode) && MANDATORY_LOCK(i)) | |
69 | ||
70 | /* | |
71 | * This is a cache of readahead params that help us choose the proper | |
72 | * readahead strategy. Initially, we set all readahead parameters to 0 | |
73 | * and let the VFS handle things. | |
74 | * If you increase the number of cached files very much, you'll need to | |
75 | * add a hash table here. | |
76 | */ | |
77 | struct raparms { | |
78 | struct raparms *p_next; | |
79 | unsigned int p_count; | |
80 | ino_t p_ino; | |
81 | dev_t p_dev; | |
82 | int p_set; | |
83 | struct file_ra_state p_ra; | |
fce1456a | 84 | unsigned int p_hindex; |
1da177e4 LT |
85 | }; |
86 | ||
fce1456a GB |
87 | struct raparm_hbucket { |
88 | struct raparms *pb_head; | |
89 | spinlock_t pb_lock; | |
90 | } ____cacheline_aligned_in_smp; | |
91 | ||
1da177e4 | 92 | static struct raparms * raparml; |
fce1456a GB |
93 | #define RAPARM_HASH_BITS 4 |
94 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
95 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
96 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
97 | |
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; | |
110 | struct vfsmount *mnt = mntget(exp->ex_mnt); | |
111 | struct dentry *mounts = dget(dentry); | |
6264d69d | 112 | int err = 0; |
1da177e4 LT |
113 | |
114 | while (follow_down(&mnt,&mounts)&&d_mountpoint(mounts)); | |
115 | ||
116 | exp2 = exp_get_by_name(exp->ex_client, mnt, mounts, &rqstp->rq_chandle); | |
117 | if (IS_ERR(exp2)) { | |
118 | err = PTR_ERR(exp2); | |
119 | dput(mounts); | |
120 | mntput(mnt); | |
121 | goto out; | |
122 | } | |
123 | if (exp2 && ((exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2))) { | |
124 | /* successfully crossed mount point */ | |
125 | exp_put(exp); | |
126 | *expp = exp2; | |
127 | dput(dentry); | |
128 | *dpp = mounts; | |
129 | } else { | |
130 | if (exp2) exp_put(exp2); | |
131 | dput(mounts); | |
132 | } | |
133 | mntput(mnt); | |
134 | out: | |
135 | return err; | |
136 | } | |
137 | ||
138 | /* | |
139 | * Look up one component of a pathname. | |
140 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
141 | * | |
142 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
143 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
144 | * accepted as it stands and the mounted directory is | |
145 | * returned. Otherwise the covered directory is returned. | |
146 | * NOTE: this mountpoint crossing is not supported properly by all | |
147 | * clients and is explicitly disallowed for NFSv3 | |
148 | * NeilBrown <neilb@cse.unsw.edu.au> | |
149 | */ | |
6264d69d | 150 | __be32 |
1da177e4 LT |
151 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, |
152 | int len, struct svc_fh *resfh) | |
153 | { | |
154 | struct svc_export *exp; | |
155 | struct dentry *dparent; | |
156 | struct dentry *dentry; | |
6264d69d AV |
157 | __be32 err; |
158 | int host_err; | |
1da177e4 LT |
159 | |
160 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
161 | ||
162 | /* Obtain dentry and export. */ | |
163 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_EXEC); | |
164 | if (err) | |
165 | return err; | |
166 | ||
167 | dparent = fhp->fh_dentry; | |
168 | exp = fhp->fh_export; | |
169 | exp_get(exp); | |
170 | ||
171 | err = nfserr_acces; | |
172 | ||
173 | /* Lookup the name, but don't follow links */ | |
174 | if (isdotent(name, len)) { | |
175 | if (len==1) | |
176 | dentry = dget(dparent); | |
177 | else if (dparent != exp->ex_dentry) { | |
178 | dentry = dget_parent(dparent); | |
179 | } else if (!EX_NOHIDE(exp)) | |
180 | dentry = dget(dparent); /* .. == . just like at / */ | |
181 | else { | |
182 | /* checking mountpoint crossing is very different when stepping up */ | |
183 | struct svc_export *exp2 = NULL; | |
184 | struct dentry *dp; | |
185 | struct vfsmount *mnt = mntget(exp->ex_mnt); | |
186 | dentry = dget(dparent); | |
187 | while(dentry == mnt->mnt_root && follow_up(&mnt, &dentry)) | |
188 | ; | |
189 | dp = dget_parent(dentry); | |
190 | dput(dentry); | |
191 | dentry = dp; | |
192 | ||
193 | exp2 = exp_parent(exp->ex_client, mnt, dentry, | |
194 | &rqstp->rq_chandle); | |
195 | if (IS_ERR(exp2)) { | |
6264d69d | 196 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
197 | dput(dentry); |
198 | mntput(mnt); | |
199 | goto out_nfserr; | |
200 | } | |
201 | if (!exp2) { | |
202 | dput(dentry); | |
203 | dentry = dget(dparent); | |
204 | } else { | |
205 | exp_put(exp); | |
206 | exp = exp2; | |
207 | } | |
208 | mntput(mnt); | |
209 | } | |
210 | } else { | |
211 | fh_lock(fhp); | |
212 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 213 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
214 | if (IS_ERR(dentry)) |
215 | goto out_nfserr; | |
216 | /* | |
217 | * check if we have crossed a mount point ... | |
218 | */ | |
219 | if (d_mountpoint(dentry)) { | |
6264d69d | 220 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
221 | dput(dentry); |
222 | goto out_nfserr; | |
223 | } | |
224 | } | |
225 | } | |
226 | /* | |
227 | * Note: we compose the file handle now, but as the | |
228 | * dentry may be negative, it may need to be updated. | |
229 | */ | |
230 | err = fh_compose(resfh, exp, dentry, fhp); | |
231 | if (!err && !dentry->d_inode) | |
232 | err = nfserr_noent; | |
233 | dput(dentry); | |
234 | out: | |
235 | exp_put(exp); | |
236 | return err; | |
237 | ||
238 | out_nfserr: | |
6264d69d | 239 | err = nfserrno(host_err); |
1da177e4 LT |
240 | goto out; |
241 | } | |
242 | ||
243 | /* | |
244 | * Set various file attributes. | |
245 | * N.B. After this call fhp needs an fh_put | |
246 | */ | |
6264d69d | 247 | __be32 |
1da177e4 LT |
248 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
249 | int check_guard, time_t guardtime) | |
250 | { | |
251 | struct dentry *dentry; | |
252 | struct inode *inode; | |
253 | int accmode = MAY_SATTR; | |
254 | int ftype = 0; | |
255 | int imode; | |
6264d69d AV |
256 | __be32 err; |
257 | int host_err; | |
1da177e4 LT |
258 | int size_change = 0; |
259 | ||
260 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
261 | accmode |= MAY_WRITE|MAY_OWNER_OVERRIDE; | |
262 | if (iap->ia_valid & ATTR_SIZE) | |
263 | ftype = S_IFREG; | |
264 | ||
265 | /* Get inode */ | |
266 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 267 | if (err) |
1da177e4 LT |
268 | goto out; |
269 | ||
270 | dentry = fhp->fh_dentry; | |
271 | inode = dentry->d_inode; | |
272 | ||
15b7a1b8 N |
273 | /* Ignore any mode updates on symlinks */ |
274 | if (S_ISLNK(inode->i_mode)) | |
275 | iap->ia_valid &= ~ATTR_MODE; | |
276 | ||
277 | if (!iap->ia_valid) | |
278 | goto out; | |
279 | ||
1da177e4 LT |
280 | /* NFSv2 does not differentiate between "set-[ac]time-to-now" |
281 | * which only requires access, and "set-[ac]time-to-X" which | |
282 | * requires ownership. | |
283 | * So if it looks like it might be "set both to the same time which | |
284 | * is close to now", and if inode_change_ok fails, then we | |
285 | * convert to "set to now" instead of "set to explicit time" | |
286 | * | |
287 | * We only call inode_change_ok as the last test as technically | |
288 | * it is not an interface that we should be using. It is only | |
289 | * valid if the filesystem does not define it's own i_op->setattr. | |
290 | */ | |
291 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
292 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
293 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET | |
294 | && iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec | |
295 | ) { | |
296 | /* Looks probable. Now just make sure time is in the right ballpark. | |
297 | * Solaris, at least, doesn't seem to care what the time request is. | |
298 | * We require it be within 30 minutes of now. | |
299 | */ | |
300 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); | |
301 | if (delta<0) delta = -delta; | |
302 | if (delta < MAX_TOUCH_TIME_ERROR && | |
303 | inode_change_ok(inode, iap) != 0) { | |
304 | /* turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME | |
305 | * this will cause notify_change to set these times to "now" | |
306 | */ | |
307 | iap->ia_valid &= ~BOTH_TIME_SET; | |
308 | } | |
309 | } | |
310 | ||
311 | /* The size case is special. It changes the file as well as the attributes. */ | |
312 | if (iap->ia_valid & ATTR_SIZE) { | |
313 | if (iap->ia_size < inode->i_size) { | |
314 | err = nfsd_permission(fhp->fh_export, dentry, MAY_TRUNC|MAY_OWNER_OVERRIDE); | |
315 | if (err) | |
316 | goto out; | |
317 | } | |
318 | ||
319 | /* | |
320 | * If we are changing the size of the file, then | |
321 | * we need to break all leases. | |
322 | */ | |
6264d69d AV |
323 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
324 | if (host_err == -EWOULDBLOCK) | |
325 | host_err = -ETIMEDOUT; | |
326 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
327 | goto out_nfserr; |
328 | ||
6264d69d AV |
329 | host_err = get_write_access(inode); |
330 | if (host_err) | |
1da177e4 LT |
331 | goto out_nfserr; |
332 | ||
333 | size_change = 1; | |
6264d69d AV |
334 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
335 | if (host_err) { | |
1da177e4 LT |
336 | put_write_access(inode); |
337 | goto out_nfserr; | |
338 | } | |
339 | DQUOT_INIT(inode); | |
340 | } | |
341 | ||
342 | imode = inode->i_mode; | |
343 | if (iap->ia_valid & ATTR_MODE) { | |
344 | iap->ia_mode &= S_IALLUGO; | |
345 | imode = iap->ia_mode |= (imode & ~S_IALLUGO); | |
346 | } | |
347 | ||
348 | /* Revoke setuid/setgid bit on chown/chgrp */ | |
349 | if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) | |
350 | iap->ia_valid |= ATTR_KILL_SUID; | |
351 | if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid) | |
352 | iap->ia_valid |= ATTR_KILL_SGID; | |
353 | ||
354 | /* Change the attributes. */ | |
355 | ||
356 | iap->ia_valid |= ATTR_CTIME; | |
357 | ||
358 | err = nfserr_notsync; | |
359 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
360 | fh_lock(fhp); | |
6264d69d AV |
361 | host_err = notify_change(dentry, iap); |
362 | err = nfserrno(host_err); | |
1da177e4 LT |
363 | fh_unlock(fhp); |
364 | } | |
365 | if (size_change) | |
366 | put_write_access(inode); | |
367 | if (!err) | |
368 | if (EX_ISSYNC(fhp->fh_export)) | |
369 | write_inode_now(inode, 1); | |
370 | out: | |
371 | return err; | |
372 | ||
373 | out_nfserr: | |
6264d69d | 374 | err = nfserrno(host_err); |
1da177e4 LT |
375 | goto out; |
376 | } | |
377 | ||
5be196e5 CH |
378 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
379 | defined(CONFIG_NFSD_V3_ACL) || \ | |
380 | defined(CONFIG_NFSD_V4) | |
381 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
382 | { | |
383 | ssize_t buflen; | |
5be196e5 CH |
384 | |
385 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
386 | if (buflen <= 0) | |
387 | return buflen; | |
1da177e4 | 388 | |
5be196e5 CH |
389 | *buf = kmalloc(buflen, GFP_KERNEL); |
390 | if (!*buf) | |
391 | return -ENOMEM; | |
392 | ||
b5872b0d | 393 | return vfs_getxattr(dentry, key, *buf, buflen); |
5be196e5 CH |
394 | } |
395 | #endif | |
396 | ||
397 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
398 | static int |
399 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
400 | { | |
401 | int len; | |
402 | size_t buflen; | |
403 | char *buf = NULL; | |
404 | int error = 0; | |
1da177e4 LT |
405 | |
406 | buflen = posix_acl_xattr_size(pacl->a_count); | |
407 | buf = kmalloc(buflen, GFP_KERNEL); | |
408 | error = -ENOMEM; | |
409 | if (buf == NULL) | |
410 | goto out; | |
411 | ||
412 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
413 | if (len < 0) { | |
414 | error = len; | |
415 | goto out; | |
416 | } | |
417 | ||
5be196e5 | 418 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
419 | out: |
420 | kfree(buf); | |
421 | return error; | |
422 | } | |
423 | ||
6264d69d | 424 | __be32 |
1da177e4 LT |
425 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
426 | struct nfs4_acl *acl) | |
427 | { | |
6264d69d AV |
428 | __be32 error; |
429 | int host_error; | |
1da177e4 LT |
430 | struct dentry *dentry; |
431 | struct inode *inode; | |
432 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
433 | unsigned int flags = 0; | |
434 | ||
435 | /* Get inode */ | |
436 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, MAY_SATTR); | |
437 | if (error) | |
4b2ca38a | 438 | return error; |
1da177e4 LT |
439 | |
440 | dentry = fhp->fh_dentry; | |
441 | inode = dentry->d_inode; | |
442 | if (S_ISDIR(inode->i_mode)) | |
443 | flags = NFS4_ACL_DIR; | |
444 | ||
6264d69d AV |
445 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
446 | if (host_error == -EINVAL) { | |
4b2ca38a | 447 | return nfserr_attrnotsupp; |
6264d69d | 448 | } else if (host_error < 0) |
1da177e4 LT |
449 | goto out_nfserr; |
450 | ||
6264d69d AV |
451 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
452 | if (host_error < 0) | |
4b2ca38a | 453 | goto out_release; |
1da177e4 | 454 | |
4b2ca38a | 455 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 456 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 457 | |
4b2ca38a | 458 | out_release: |
1da177e4 LT |
459 | posix_acl_release(pacl); |
460 | posix_acl_release(dpacl); | |
1da177e4 | 461 | out_nfserr: |
f34f9242 | 462 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 463 | return nfserr_attrnotsupp; |
f34f9242 | 464 | else |
4b2ca38a | 465 | return nfserrno(host_error); |
1da177e4 LT |
466 | } |
467 | ||
468 | static struct posix_acl * | |
469 | _get_posix_acl(struct dentry *dentry, char *key) | |
470 | { | |
5be196e5 | 471 | void *buf = NULL; |
1da177e4 | 472 | struct posix_acl *pacl = NULL; |
5be196e5 | 473 | int buflen; |
1da177e4 | 474 | |
5be196e5 CH |
475 | buflen = nfsd_getxattr(dentry, key, &buf); |
476 | if (!buflen) | |
477 | buflen = -ENODATA; | |
478 | if (buflen <= 0) | |
479 | return ERR_PTR(buflen); | |
1da177e4 LT |
480 | |
481 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
482 | kfree(buf); |
483 | return pacl; | |
1da177e4 LT |
484 | } |
485 | ||
486 | int | |
487 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
488 | { | |
489 | struct inode *inode = dentry->d_inode; | |
490 | int error = 0; | |
491 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
492 | unsigned int flags = 0; | |
493 | ||
9a59f452 | 494 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
495 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
496 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
497 | if (IS_ERR(pacl)) { | |
498 | error = PTR_ERR(pacl); | |
499 | pacl = NULL; | |
500 | goto out; | |
501 | } | |
502 | ||
503 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 504 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
505 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
506 | dpacl = NULL; | |
507 | else if (IS_ERR(dpacl)) { | |
508 | error = PTR_ERR(dpacl); | |
509 | dpacl = NULL; | |
510 | goto out; | |
511 | } | |
512 | flags = NFS4_ACL_DIR; | |
513 | } | |
514 | ||
515 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
516 | if (IS_ERR(*acl)) { | |
517 | error = PTR_ERR(*acl); | |
518 | *acl = NULL; | |
519 | } | |
520 | out: | |
521 | posix_acl_release(pacl); | |
522 | posix_acl_release(dpacl); | |
523 | return error; | |
524 | } | |
525 | ||
526 | #endif /* defined(CONFIG_NFS_V4) */ | |
527 | ||
528 | #ifdef CONFIG_NFSD_V3 | |
529 | /* | |
530 | * Check server access rights to a file system object | |
531 | */ | |
532 | struct accessmap { | |
533 | u32 access; | |
534 | int how; | |
535 | }; | |
536 | static struct accessmap nfs3_regaccess[] = { | |
537 | { NFS3_ACCESS_READ, MAY_READ }, | |
538 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
539 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_TRUNC }, | |
540 | { NFS3_ACCESS_EXTEND, MAY_WRITE }, | |
541 | ||
542 | { 0, 0 } | |
543 | }; | |
544 | ||
545 | static struct accessmap nfs3_diraccess[] = { | |
546 | { NFS3_ACCESS_READ, MAY_READ }, | |
547 | { NFS3_ACCESS_LOOKUP, MAY_EXEC }, | |
548 | { NFS3_ACCESS_MODIFY, MAY_EXEC|MAY_WRITE|MAY_TRUNC }, | |
549 | { NFS3_ACCESS_EXTEND, MAY_EXEC|MAY_WRITE }, | |
550 | { NFS3_ACCESS_DELETE, MAY_REMOVE }, | |
551 | ||
552 | { 0, 0 } | |
553 | }; | |
554 | ||
555 | static struct accessmap nfs3_anyaccess[] = { | |
556 | /* Some clients - Solaris 2.6 at least, make an access call | |
557 | * to the server to check for access for things like /dev/null | |
558 | * (which really, the server doesn't care about). So | |
559 | * We provide simple access checking for them, looking | |
560 | * mainly at mode bits, and we make sure to ignore read-only | |
561 | * filesystem checks | |
562 | */ | |
563 | { NFS3_ACCESS_READ, MAY_READ }, | |
564 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
565 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
566 | { NFS3_ACCESS_EXTEND, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
567 | ||
568 | { 0, 0 } | |
569 | }; | |
570 | ||
6264d69d | 571 | __be32 |
1da177e4 LT |
572 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
573 | { | |
574 | struct accessmap *map; | |
575 | struct svc_export *export; | |
576 | struct dentry *dentry; | |
577 | u32 query, result = 0, sresult = 0; | |
6264d69d | 578 | __be32 error; |
1da177e4 LT |
579 | |
580 | error = fh_verify(rqstp, fhp, 0, MAY_NOP); | |
581 | if (error) | |
582 | goto out; | |
583 | ||
584 | export = fhp->fh_export; | |
585 | dentry = fhp->fh_dentry; | |
586 | ||
587 | if (S_ISREG(dentry->d_inode->i_mode)) | |
588 | map = nfs3_regaccess; | |
589 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
590 | map = nfs3_diraccess; | |
591 | else | |
592 | map = nfs3_anyaccess; | |
593 | ||
594 | ||
595 | query = *access; | |
596 | for (; map->access; map++) { | |
597 | if (map->access & query) { | |
6264d69d | 598 | __be32 err2; |
1da177e4 LT |
599 | |
600 | sresult |= map->access; | |
601 | ||
602 | err2 = nfsd_permission(export, dentry, map->how); | |
603 | switch (err2) { | |
604 | case nfs_ok: | |
605 | result |= map->access; | |
606 | break; | |
607 | ||
608 | /* the following error codes just mean the access was not allowed, | |
609 | * rather than an error occurred */ | |
610 | case nfserr_rofs: | |
611 | case nfserr_acces: | |
612 | case nfserr_perm: | |
613 | /* simply don't "or" in the access bit. */ | |
614 | break; | |
615 | default: | |
616 | error = err2; | |
617 | goto out; | |
618 | } | |
619 | } | |
620 | } | |
621 | *access = result; | |
622 | if (supported) | |
623 | *supported = sresult; | |
624 | ||
625 | out: | |
626 | return error; | |
627 | } | |
628 | #endif /* CONFIG_NFSD_V3 */ | |
629 | ||
630 | ||
631 | ||
632 | /* | |
633 | * Open an existing file or directory. | |
634 | * The access argument indicates the type of open (read/write/lock) | |
635 | * N.B. After this call fhp needs an fh_put | |
636 | */ | |
6264d69d | 637 | __be32 |
1da177e4 LT |
638 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
639 | int access, struct file **filp) | |
640 | { | |
641 | struct dentry *dentry; | |
642 | struct inode *inode; | |
6264d69d AV |
643 | int flags = O_RDONLY|O_LARGEFILE; |
644 | __be32 err; | |
645 | int host_err; | |
1da177e4 LT |
646 | |
647 | /* | |
648 | * If we get here, then the client has already done an "open", | |
649 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
650 | * in case a chmod has now revoked permission. | |
651 | */ | |
652 | err = fh_verify(rqstp, fhp, type, access | MAY_OWNER_OVERRIDE); | |
653 | if (err) | |
654 | goto out; | |
655 | ||
656 | dentry = fhp->fh_dentry; | |
657 | inode = dentry->d_inode; | |
658 | ||
659 | /* Disallow write access to files with the append-only bit set | |
660 | * or any access when mandatory locking enabled | |
661 | */ | |
662 | err = nfserr_perm; | |
663 | if (IS_APPEND(inode) && (access & MAY_WRITE)) | |
664 | goto out; | |
665 | if (IS_ISMNDLK(inode)) | |
666 | goto out; | |
667 | ||
668 | if (!inode->i_fop) | |
669 | goto out; | |
670 | ||
671 | /* | |
672 | * Check to see if there are any leases on this file. | |
673 | * This may block while leases are broken. | |
674 | */ | |
6264d69d AV |
675 | host_err = break_lease(inode, O_NONBLOCK | ((access & MAY_WRITE) ? FMODE_WRITE : 0)); |
676 | if (host_err == -EWOULDBLOCK) | |
677 | host_err = -ETIMEDOUT; | |
678 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
679 | goto out_nfserr; |
680 | ||
681 | if (access & MAY_WRITE) { | |
9ecb6a08 BF |
682 | if (access & MAY_READ) |
683 | flags = O_RDWR|O_LARGEFILE; | |
684 | else | |
685 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 LT |
686 | |
687 | DQUOT_INIT(inode); | |
688 | } | |
689 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_mnt), flags); | |
690 | if (IS_ERR(*filp)) | |
6264d69d | 691 | host_err = PTR_ERR(*filp); |
1da177e4 | 692 | out_nfserr: |
6264d69d | 693 | err = nfserrno(host_err); |
1da177e4 LT |
694 | out: |
695 | return err; | |
696 | } | |
697 | ||
698 | /* | |
699 | * Close a file. | |
700 | */ | |
701 | void | |
702 | nfsd_close(struct file *filp) | |
703 | { | |
704 | fput(filp); | |
705 | } | |
706 | ||
707 | /* | |
708 | * Sync a file | |
709 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
710 | * after it. | |
711 | */ | |
a334de28 | 712 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, |
99ac48f5 | 713 | const struct file_operations *fop) |
1da177e4 LT |
714 | { |
715 | struct inode *inode = dp->d_inode; | |
716 | int (*fsync) (struct file *, struct dentry *, int); | |
f193fbab | 717 | int err; |
1da177e4 | 718 | |
f193fbab YT |
719 | err = filemap_fdatawrite(inode->i_mapping); |
720 | if (err == 0 && fop && (fsync = fop->fsync)) | |
721 | err = fsync(filp, dp, 0); | |
722 | if (err == 0) | |
723 | err = filemap_fdatawait(inode->i_mapping); | |
a334de28 | 724 | |
f193fbab | 725 | return err; |
1da177e4 LT |
726 | } |
727 | ||
728 | ||
a334de28 | 729 | static int |
1da177e4 LT |
730 | nfsd_sync(struct file *filp) |
731 | { | |
a334de28 | 732 | int err; |
7eaa36e2 JJS |
733 | struct inode *inode = filp->f_path.dentry->d_inode; |
734 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
1b1dcc1b | 735 | mutex_lock(&inode->i_mutex); |
7eaa36e2 | 736 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); |
1b1dcc1b | 737 | mutex_unlock(&inode->i_mutex); |
a334de28 DS |
738 | |
739 | return err; | |
1da177e4 LT |
740 | } |
741 | ||
f193fbab | 742 | int |
1da177e4 LT |
743 | nfsd_sync_dir(struct dentry *dp) |
744 | { | |
f193fbab | 745 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
746 | } |
747 | ||
748 | /* | |
749 | * Obtain the readahead parameters for the file | |
750 | * specified by (dev, ino). | |
751 | */ | |
1da177e4 LT |
752 | |
753 | static inline struct raparms * | |
754 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
755 | { | |
756 | struct raparms *ra, **rap, **frap = NULL; | |
757 | int depth = 0; | |
fce1456a GB |
758 | unsigned int hash; |
759 | struct raparm_hbucket *rab; | |
760 | ||
761 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
762 | rab = &raparm_hash[hash]; | |
1da177e4 | 763 | |
fce1456a GB |
764 | spin_lock(&rab->pb_lock); |
765 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
766 | if (ra->p_ino == ino && ra->p_dev == dev) |
767 | goto found; | |
768 | depth++; | |
769 | if (ra->p_count == 0) | |
770 | frap = rap; | |
771 | } | |
772 | depth = nfsdstats.ra_size*11/10; | |
773 | if (!frap) { | |
fce1456a | 774 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
775 | return NULL; |
776 | } | |
777 | rap = frap; | |
778 | ra = *frap; | |
779 | ra->p_dev = dev; | |
780 | ra->p_ino = ino; | |
781 | ra->p_set = 0; | |
fce1456a | 782 | ra->p_hindex = hash; |
1da177e4 | 783 | found: |
fce1456a | 784 | if (rap != &rab->pb_head) { |
1da177e4 | 785 | *rap = ra->p_next; |
fce1456a GB |
786 | ra->p_next = rab->pb_head; |
787 | rab->pb_head = ra; | |
1da177e4 LT |
788 | } |
789 | ra->p_count++; | |
790 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 791 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
792 | return ra; |
793 | } | |
794 | ||
795 | /* | |
cf8208d0 JA |
796 | * Grab and keep cached pages associated with a file in the svc_rqst |
797 | * so that they can be passed to the network sendmsg/sendpage routines | |
798 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
799 | */ |
800 | static int | |
cf8208d0 JA |
801 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
802 | struct splice_desc *sd) | |
1da177e4 | 803 | { |
cf8208d0 | 804 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 805 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
806 | struct page *page = buf->page; |
807 | size_t size; | |
808 | int ret; | |
1da177e4 | 809 | |
cac36bb0 | 810 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
811 | if (unlikely(ret)) |
812 | return ret; | |
813 | ||
814 | size = sd->len; | |
1da177e4 LT |
815 | |
816 | if (rqstp->rq_res.page_len == 0) { | |
817 | get_page(page); | |
44524359 N |
818 | put_page(*pp); |
819 | *pp = page; | |
820 | rqstp->rq_resused++; | |
cf8208d0 | 821 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 822 | rqstp->rq_res.page_len = size; |
44524359 | 823 | } else if (page != pp[-1]) { |
1da177e4 | 824 | get_page(page); |
250f3915 N |
825 | if (*pp) |
826 | put_page(*pp); | |
44524359 N |
827 | *pp = page; |
828 | rqstp->rq_resused++; | |
1da177e4 | 829 | rqstp->rq_res.page_len += size; |
44524359 | 830 | } else |
1da177e4 | 831 | rqstp->rq_res.page_len += size; |
1da177e4 | 832 | |
1da177e4 LT |
833 | return size; |
834 | } | |
835 | ||
cf8208d0 JA |
836 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
837 | struct splice_desc *sd) | |
838 | { | |
839 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
840 | } | |
841 | ||
6264d69d | 842 | static __be32 |
1da177e4 LT |
843 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
844 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
845 | { | |
846 | struct inode *inode; | |
847 | struct raparms *ra; | |
848 | mm_segment_t oldfs; | |
6264d69d AV |
849 | __be32 err; |
850 | int host_err; | |
1da177e4 LT |
851 | |
852 | err = nfserr_perm; | |
7eaa36e2 | 853 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
854 | #ifdef MSNFS |
855 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
856 | (!lock_may_read(inode, offset, *count))) | |
857 | goto out; | |
858 | #endif | |
859 | ||
860 | /* Get readahead parameters */ | |
861 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
862 | ||
863 | if (ra && ra->p_set) | |
864 | file->f_ra = ra->p_ra; | |
865 | ||
cf8208d0 JA |
866 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
867 | struct splice_desc sd = { | |
868 | .len = 0, | |
869 | .total_len = *count, | |
870 | .pos = offset, | |
871 | .u.data = rqstp, | |
872 | }; | |
873 | ||
4fbef206 | 874 | rqstp->rq_resused = 1; |
cf8208d0 | 875 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
876 | } else { |
877 | oldfs = get_fs(); | |
878 | set_fs(KERNEL_DS); | |
6264d69d | 879 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
880 | set_fs(oldfs); |
881 | } | |
882 | ||
883 | /* Write back readahead params */ | |
884 | if (ra) { | |
fce1456a GB |
885 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
886 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
887 | ra->p_ra = file->f_ra; |
888 | ra->p_set = 1; | |
889 | ra->p_count--; | |
fce1456a | 890 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
891 | } |
892 | ||
6264d69d AV |
893 | if (host_err >= 0) { |
894 | nfsdstats.io_read += host_err; | |
895 | *count = host_err; | |
1da177e4 | 896 | err = 0; |
7eaa36e2 | 897 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 898 | } else |
6264d69d | 899 | err = nfserrno(host_err); |
1da177e4 LT |
900 | out: |
901 | return err; | |
902 | } | |
903 | ||
9f708e40 NB |
904 | static void kill_suid(struct dentry *dentry) |
905 | { | |
906 | struct iattr ia; | |
907 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID; | |
908 | ||
1b1dcc1b | 909 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 910 | notify_change(dentry, &ia); |
1b1dcc1b | 911 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
912 | } |
913 | ||
6264d69d | 914 | static __be32 |
1da177e4 LT |
915 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
916 | loff_t offset, struct kvec *vec, int vlen, | |
917 | unsigned long cnt, int *stablep) | |
918 | { | |
919 | struct svc_export *exp; | |
920 | struct dentry *dentry; | |
921 | struct inode *inode; | |
922 | mm_segment_t oldfs; | |
6264d69d AV |
923 | __be32 err = 0; |
924 | int host_err; | |
1da177e4 LT |
925 | int stable = *stablep; |
926 | ||
45bd3b3d | 927 | #ifdef MSNFS |
1da177e4 LT |
928 | err = nfserr_perm; |
929 | ||
1da177e4 | 930 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
7eaa36e2 | 931 | (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt))) |
1da177e4 LT |
932 | goto out; |
933 | #endif | |
934 | ||
7eaa36e2 | 935 | dentry = file->f_path.dentry; |
1da177e4 LT |
936 | inode = dentry->d_inode; |
937 | exp = fhp->fh_export; | |
938 | ||
939 | /* | |
940 | * Request sync writes if | |
941 | * - the sync export option has been set, or | |
942 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
943 | * - The file system doesn't support fsync(). | |
944 | * When gathered writes have been configured for this volume, | |
945 | * flushing the data to disk is handled separately below. | |
946 | */ | |
947 | ||
948 | if (file->f_op->fsync == 0) {/* COMMIT3 cannot work */ | |
949 | stable = 2; | |
950 | *stablep = 2; /* FILE_SYNC */ | |
951 | } | |
952 | ||
953 | if (!EX_ISSYNC(exp)) | |
954 | stable = 0; | |
955 | if (stable && !EX_WGATHER(exp)) | |
956 | file->f_flags |= O_SYNC; | |
957 | ||
958 | /* Write the data. */ | |
959 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 960 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 961 | set_fs(oldfs); |
6264d69d | 962 | if (host_err >= 0) { |
1da177e4 | 963 | nfsdstats.io_write += cnt; |
7eaa36e2 | 964 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
965 | } |
966 | ||
967 | /* clear setuid/setgid flag after write */ | |
6264d69d | 968 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 969 | kill_suid(dentry); |
1da177e4 | 970 | |
6264d69d | 971 | if (host_err >= 0 && stable) { |
1da177e4 LT |
972 | static ino_t last_ino; |
973 | static dev_t last_dev; | |
974 | ||
975 | /* | |
976 | * Gathered writes: If another process is currently | |
977 | * writing to the file, there's a high chance | |
978 | * this is another nfsd (triggered by a bulk write | |
979 | * from a client's biod). Rather than syncing the | |
980 | * file with each write request, we sleep for 10 msec. | |
981 | * | |
982 | * I don't know if this roughly approximates | |
983 | * C. Juszak's idea of gathered writes, but it's a | |
984 | * nice and simple solution (IMHO), and it seems to | |
985 | * work:-) | |
986 | */ | |
987 | if (EX_WGATHER(exp)) { | |
988 | if (atomic_read(&inode->i_writecount) > 1 | |
989 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
990 | dprintk("nfsd: write defer %d\n", current->pid); | |
991 | msleep(10); | |
992 | dprintk("nfsd: write resume %d\n", current->pid); | |
993 | } | |
994 | ||
995 | if (inode->i_state & I_DIRTY) { | |
996 | dprintk("nfsd: write sync %d\n", current->pid); | |
6264d69d | 997 | host_err=nfsd_sync(file); |
1da177e4 LT |
998 | } |
999 | #if 0 | |
1000 | wake_up(&inode->i_wait); | |
1001 | #endif | |
1002 | } | |
1003 | last_ino = inode->i_ino; | |
1004 | last_dev = inode->i_sb->s_dev; | |
1005 | } | |
1006 | ||
6264d69d AV |
1007 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
1008 | if (host_err >= 0) | |
1da177e4 LT |
1009 | err = 0; |
1010 | else | |
6264d69d | 1011 | err = nfserrno(host_err); |
1da177e4 LT |
1012 | out: |
1013 | return err; | |
1014 | } | |
1015 | ||
1016 | /* | |
1017 | * Read data from a file. count must contain the requested read count | |
1018 | * on entry. On return, *count contains the number of bytes actually read. | |
1019 | * N.B. After this call fhp needs an fh_put | |
1020 | */ | |
6264d69d | 1021 | __be32 |
1da177e4 LT |
1022 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1023 | loff_t offset, struct kvec *vec, int vlen, | |
1024 | unsigned long *count) | |
1025 | { | |
6264d69d | 1026 | __be32 err; |
1da177e4 LT |
1027 | |
1028 | if (file) { | |
1029 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1030 | MAY_READ|MAY_OWNER_OVERRIDE); | |
1031 | if (err) | |
1032 | goto out; | |
1033 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1034 | } else { | |
1035 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_READ, &file); | |
1036 | if (err) | |
1037 | goto out; | |
1038 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1039 | nfsd_close(file); | |
1040 | } | |
1041 | out: | |
1042 | return err; | |
1043 | } | |
1044 | ||
1045 | /* | |
1046 | * Write data to a file. | |
1047 | * The stable flag requests synchronous writes. | |
1048 | * N.B. After this call fhp needs an fh_put | |
1049 | */ | |
6264d69d | 1050 | __be32 |
1da177e4 LT |
1051 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1052 | loff_t offset, struct kvec *vec, int vlen, unsigned long cnt, | |
1053 | int *stablep) | |
1054 | { | |
6264d69d | 1055 | __be32 err = 0; |
1da177e4 LT |
1056 | |
1057 | if (file) { | |
1058 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1059 | MAY_WRITE|MAY_OWNER_OVERRIDE); | |
1060 | if (err) | |
1061 | goto out; | |
1062 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1063 | stablep); | |
1064 | } else { | |
1065 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file); | |
1066 | if (err) | |
1067 | goto out; | |
1068 | ||
1069 | if (cnt) | |
1070 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1071 | cnt, stablep); | |
1072 | nfsd_close(file); | |
1073 | } | |
1074 | out: | |
1075 | return err; | |
1076 | } | |
1077 | ||
1078 | #ifdef CONFIG_NFSD_V3 | |
1079 | /* | |
1080 | * Commit all pending writes to stable storage. | |
1081 | * Strictly speaking, we could sync just the indicated file region here, | |
1082 | * but there's currently no way we can ask the VFS to do so. | |
1083 | * | |
1084 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1085 | * data to the client, as locking happens lower down in the filesystem. | |
1086 | */ | |
6264d69d | 1087 | __be32 |
1da177e4 LT |
1088 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1089 | loff_t offset, unsigned long count) | |
1090 | { | |
1091 | struct file *file; | |
6264d69d | 1092 | __be32 err; |
1da177e4 LT |
1093 | |
1094 | if ((u64)count > ~(u64)offset) | |
1095 | return nfserr_inval; | |
1096 | ||
1097 | if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0) | |
1098 | return err; | |
1099 | if (EX_ISSYNC(fhp->fh_export)) { | |
1100 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1101 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1102 | } else { |
1103 | err = nfserr_notsupp; | |
1104 | } | |
1105 | } | |
1106 | ||
1107 | nfsd_close(file); | |
1108 | return err; | |
1109 | } | |
1110 | #endif /* CONFIG_NFSD_V3 */ | |
1111 | ||
1112 | /* | |
1113 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1114 | * not yet implemented. | |
1115 | * If the response fh has been verified, the parent directory should | |
1116 | * already be locked. Note that the parent directory is left locked. | |
1117 | * | |
1118 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1119 | */ | |
6264d69d | 1120 | __be32 |
1da177e4 LT |
1121 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1122 | char *fname, int flen, struct iattr *iap, | |
1123 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1124 | { | |
1125 | struct dentry *dentry, *dchild = NULL; | |
1126 | struct inode *dirp; | |
6264d69d AV |
1127 | __be32 err; |
1128 | int host_err; | |
1da177e4 LT |
1129 | |
1130 | err = nfserr_perm; | |
1131 | if (!flen) | |
1132 | goto out; | |
1133 | err = nfserr_exist; | |
1134 | if (isdotent(fname, flen)) | |
1135 | goto out; | |
1136 | ||
1137 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1138 | if (err) | |
1139 | goto out; | |
1140 | ||
1141 | dentry = fhp->fh_dentry; | |
1142 | dirp = dentry->d_inode; | |
1143 | ||
1144 | err = nfserr_notdir; | |
1145 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1146 | goto out; | |
1147 | /* | |
1148 | * Check whether the response file handle has been verified yet. | |
1149 | * If it has, the parent directory should already be locked. | |
1150 | */ | |
1151 | if (!resfhp->fh_dentry) { | |
1152 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1153 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1154 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1155 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1156 | if (IS_ERR(dchild)) |
1157 | goto out_nfserr; | |
1158 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1159 | if (err) | |
1160 | goto out; | |
1161 | } else { | |
1162 | /* called from nfsd_proc_create */ | |
1163 | dchild = dget(resfhp->fh_dentry); | |
1164 | if (!fhp->fh_locked) { | |
1165 | /* not actually possible */ | |
1166 | printk(KERN_ERR | |
1167 | "nfsd_create: parent %s/%s not locked!\n", | |
1168 | dentry->d_parent->d_name.name, | |
1169 | dentry->d_name.name); | |
d75f2b9f | 1170 | err = nfserr_io; |
1da177e4 LT |
1171 | goto out; |
1172 | } | |
1173 | } | |
1174 | /* | |
1175 | * Make sure the child dentry is still negative ... | |
1176 | */ | |
1177 | err = nfserr_exist; | |
1178 | if (dchild->d_inode) { | |
1179 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1180 | dentry->d_name.name, dchild->d_name.name); | |
1181 | goto out; | |
1182 | } | |
1183 | ||
1184 | if (!(iap->ia_valid & ATTR_MODE)) | |
1185 | iap->ia_mode = 0; | |
1186 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1187 | ||
1188 | /* | |
1189 | * Get the dir op function pointer. | |
1190 | */ | |
088406bc | 1191 | err = 0; |
1da177e4 LT |
1192 | switch (type) { |
1193 | case S_IFREG: | |
6264d69d | 1194 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1da177e4 LT |
1195 | break; |
1196 | case S_IFDIR: | |
6264d69d | 1197 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1198 | break; |
1199 | case S_IFCHR: | |
1200 | case S_IFBLK: | |
1201 | case S_IFIFO: | |
1202 | case S_IFSOCK: | |
6264d69d | 1203 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 LT |
1204 | break; |
1205 | default: | |
1206 | printk("nfsd: bad file type %o in nfsd_create\n", type); | |
6264d69d | 1207 | host_err = -EINVAL; |
1da177e4 | 1208 | } |
6264d69d | 1209 | if (host_err < 0) |
1da177e4 LT |
1210 | goto out_nfserr; |
1211 | ||
1212 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1213 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1214 | write_inode_now(dchild->d_inode, 1); |
1215 | } | |
1216 | ||
1217 | ||
1218 | /* Set file attributes. Mode has already been set and | |
1219 | * setting uid/gid works only for root. Irix appears to | |
1220 | * send along the gid when it tries to implement setgid | |
1221 | * directories via NFS. | |
1222 | */ | |
f193fbab | 1223 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1224 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab YT |
1225 | if (err2) |
1226 | err = err2; | |
1227 | } | |
1da177e4 LT |
1228 | /* |
1229 | * Update the file handle to get the new inode info. | |
1230 | */ | |
1231 | if (!err) | |
1232 | err = fh_update(resfhp); | |
1233 | out: | |
1234 | if (dchild && !IS_ERR(dchild)) | |
1235 | dput(dchild); | |
1236 | return err; | |
1237 | ||
1238 | out_nfserr: | |
6264d69d | 1239 | err = nfserrno(host_err); |
1da177e4 LT |
1240 | goto out; |
1241 | } | |
1242 | ||
1243 | #ifdef CONFIG_NFSD_V3 | |
1244 | /* | |
1245 | * NFSv3 version of nfsd_create | |
1246 | */ | |
6264d69d | 1247 | __be32 |
1da177e4 LT |
1248 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1249 | char *fname, int flen, struct iattr *iap, | |
1250 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1251 | int *truncp, int *created) |
1da177e4 LT |
1252 | { |
1253 | struct dentry *dentry, *dchild = NULL; | |
1254 | struct inode *dirp; | |
6264d69d AV |
1255 | __be32 err; |
1256 | int host_err; | |
1da177e4 | 1257 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1258 | |
1259 | err = nfserr_perm; | |
1260 | if (!flen) | |
1261 | goto out; | |
1262 | err = nfserr_exist; | |
1263 | if (isdotent(fname, flen)) | |
1264 | goto out; | |
1265 | if (!(iap->ia_valid & ATTR_MODE)) | |
1266 | iap->ia_mode = 0; | |
1267 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1268 | if (err) | |
1269 | goto out; | |
1270 | ||
1271 | dentry = fhp->fh_dentry; | |
1272 | dirp = dentry->d_inode; | |
1273 | ||
1274 | /* Get all the sanity checks out of the way before | |
1275 | * we lock the parent. */ | |
1276 | err = nfserr_notdir; | |
1277 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1278 | goto out; | |
12fd3520 | 1279 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1280 | |
1281 | /* | |
1282 | * Compose the response file handle. | |
1283 | */ | |
1284 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1285 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1286 | if (IS_ERR(dchild)) |
1287 | goto out_nfserr; | |
1288 | ||
1289 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1290 | if (err) | |
1291 | goto out; | |
1292 | ||
1293 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c PS |
1294 | /* solaris7 gets confused (bugid 4218508) if these have |
1295 | * the high bit set, so just clear the high bits. | |
1da177e4 LT |
1296 | */ |
1297 | v_mtime = verifier[0]&0x7fffffff; | |
1298 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1299 | } |
1300 | ||
1301 | if (dchild->d_inode) { | |
1302 | err = 0; | |
1303 | ||
1304 | switch (createmode) { | |
1305 | case NFS3_CREATE_UNCHECKED: | |
1306 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1307 | err = nfserr_exist; | |
1308 | else if (truncp) { | |
1309 | /* in nfsv4, we need to treat this case a little | |
1310 | * differently. we don't want to truncate the | |
1311 | * file now; this would be wrong if the OPEN | |
1312 | * fails for some other reason. furthermore, | |
1313 | * if the size is nonzero, we should ignore it | |
1314 | * according to spec! | |
1315 | */ | |
1316 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1317 | } | |
1318 | else { | |
1319 | iap->ia_valid &= ATTR_SIZE; | |
1320 | goto set_attr; | |
1321 | } | |
1322 | break; | |
1323 | case NFS3_CREATE_EXCLUSIVE: | |
1324 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1325 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1326 | && dchild->d_inode->i_size == 0 ) |
1327 | break; | |
1328 | /* fallthru */ | |
1329 | case NFS3_CREATE_GUARDED: | |
1330 | err = nfserr_exist; | |
1331 | } | |
1332 | goto out; | |
1333 | } | |
1334 | ||
6264d69d AV |
1335 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1336 | if (host_err < 0) | |
1da177e4 | 1337 | goto out_nfserr; |
81ac95c5 BF |
1338 | if (created) |
1339 | *created = 1; | |
1da177e4 LT |
1340 | |
1341 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1342 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1343 | /* setattr will sync the child (or not) */ |
1344 | } | |
1345 | ||
1da177e4 | 1346 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1347 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1348 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1349 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1350 | /* XXX someone who knows this better please fix it for nsec */ |
1351 | iap->ia_mtime.tv_sec = v_mtime; | |
1352 | iap->ia_atime.tv_sec = v_atime; | |
1353 | iap->ia_mtime.tv_nsec = 0; | |
1354 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1355 | } |
1356 | ||
1357 | /* Set file attributes. | |
1da177e4 LT |
1358 | * Irix appears to send along the gid when it tries to |
1359 | * implement setgid directories via NFS. Clear out all that cruft. | |
1360 | */ | |
1361 | set_attr: | |
c397852c | 1362 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1363 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab | 1364 | if (err2) |
45bd3b3d | 1365 | err = err2; |
f193fbab YT |
1366 | } |
1367 | ||
1368 | /* | |
1369 | * Update the filehandle to get the new inode info. | |
1370 | */ | |
1371 | if (!err) | |
1372 | err = fh_update(resfhp); | |
1da177e4 LT |
1373 | |
1374 | out: | |
1375 | fh_unlock(fhp); | |
1376 | if (dchild && !IS_ERR(dchild)) | |
1377 | dput(dchild); | |
1378 | return err; | |
1379 | ||
1380 | out_nfserr: | |
6264d69d | 1381 | err = nfserrno(host_err); |
1da177e4 LT |
1382 | goto out; |
1383 | } | |
1384 | #endif /* CONFIG_NFSD_V3 */ | |
1385 | ||
1386 | /* | |
1387 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1388 | * fits into the buffer. On return, it contains the true length. | |
1389 | * N.B. After this call fhp needs an fh_put | |
1390 | */ | |
6264d69d | 1391 | __be32 |
1da177e4 LT |
1392 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1393 | { | |
1394 | struct dentry *dentry; | |
1395 | struct inode *inode; | |
1396 | mm_segment_t oldfs; | |
6264d69d AV |
1397 | __be32 err; |
1398 | int host_err; | |
1da177e4 LT |
1399 | |
1400 | err = fh_verify(rqstp, fhp, S_IFLNK, MAY_NOP); | |
1401 | if (err) | |
1402 | goto out; | |
1403 | ||
1404 | dentry = fhp->fh_dentry; | |
1405 | inode = dentry->d_inode; | |
1406 | ||
1407 | err = nfserr_inval; | |
1408 | if (!inode->i_op || !inode->i_op->readlink) | |
1409 | goto out; | |
1410 | ||
1411 | touch_atime(fhp->fh_export->ex_mnt, dentry); | |
1412 | /* N.B. Why does this call need a get_fs()?? | |
1413 | * Remove the set_fs and watch the fireworks:-) --okir | |
1414 | */ | |
1415 | ||
1416 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1417 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1418 | set_fs(oldfs); |
1419 | ||
6264d69d | 1420 | if (host_err < 0) |
1da177e4 | 1421 | goto out_nfserr; |
6264d69d | 1422 | *lenp = host_err; |
1da177e4 LT |
1423 | err = 0; |
1424 | out: | |
1425 | return err; | |
1426 | ||
1427 | out_nfserr: | |
6264d69d | 1428 | err = nfserrno(host_err); |
1da177e4 LT |
1429 | goto out; |
1430 | } | |
1431 | ||
1432 | /* | |
1433 | * Create a symlink and look up its inode | |
1434 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1435 | */ | |
6264d69d | 1436 | __be32 |
1da177e4 LT |
1437 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1438 | char *fname, int flen, | |
1439 | char *path, int plen, | |
1440 | struct svc_fh *resfhp, | |
1441 | struct iattr *iap) | |
1442 | { | |
1443 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1444 | __be32 err, cerr; |
1445 | int host_err; | |
1da177e4 LT |
1446 | umode_t mode; |
1447 | ||
1448 | err = nfserr_noent; | |
1449 | if (!flen || !plen) | |
1450 | goto out; | |
1451 | err = nfserr_exist; | |
1452 | if (isdotent(fname, flen)) | |
1453 | goto out; | |
1454 | ||
1455 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1456 | if (err) | |
1457 | goto out; | |
1458 | fh_lock(fhp); | |
1459 | dentry = fhp->fh_dentry; | |
1460 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1461 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1462 | if (IS_ERR(dnew)) |
1463 | goto out_nfserr; | |
1464 | ||
1465 | mode = S_IALLUGO; | |
1466 | /* Only the MODE ATTRibute is even vaguely meaningful */ | |
1467 | if (iap && (iap->ia_valid & ATTR_MODE)) | |
1468 | mode = iap->ia_mode & S_IALLUGO; | |
1469 | ||
1470 | if (unlikely(path[plen] != 0)) { | |
1471 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1472 | if (path_alloced == NULL) | |
6264d69d | 1473 | host_err = -ENOMEM; |
1da177e4 LT |
1474 | else { |
1475 | strncpy(path_alloced, path, plen); | |
1476 | path_alloced[plen] = 0; | |
6264d69d | 1477 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode); |
1da177e4 LT |
1478 | kfree(path_alloced); |
1479 | } | |
1480 | } else | |
6264d69d | 1481 | host_err = vfs_symlink(dentry->d_inode, dnew, path, mode); |
1da177e4 | 1482 | |
6264d69d | 1483 | if (!host_err) { |
1da177e4 | 1484 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1485 | host_err = nfsd_sync_dir(dentry); |
1486 | } | |
1487 | err = nfserrno(host_err); | |
1da177e4 LT |
1488 | fh_unlock(fhp); |
1489 | ||
1490 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); | |
1491 | dput(dnew); | |
1492 | if (err==0) err = cerr; | |
1493 | out: | |
1494 | return err; | |
1495 | ||
1496 | out_nfserr: | |
6264d69d | 1497 | err = nfserrno(host_err); |
1da177e4 LT |
1498 | goto out; |
1499 | } | |
1500 | ||
1501 | /* | |
1502 | * Create a hardlink | |
1503 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1504 | */ | |
6264d69d | 1505 | __be32 |
1da177e4 LT |
1506 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1507 | char *name, int len, struct svc_fh *tfhp) | |
1508 | { | |
1509 | struct dentry *ddir, *dnew, *dold; | |
1510 | struct inode *dirp, *dest; | |
6264d69d AV |
1511 | __be32 err; |
1512 | int host_err; | |
1da177e4 LT |
1513 | |
1514 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_CREATE); | |
1515 | if (err) | |
1516 | goto out; | |
1517 | err = fh_verify(rqstp, tfhp, -S_IFDIR, MAY_NOP); | |
1518 | if (err) | |
1519 | goto out; | |
1520 | ||
1521 | err = nfserr_perm; | |
1522 | if (!len) | |
1523 | goto out; | |
1524 | err = nfserr_exist; | |
1525 | if (isdotent(name, len)) | |
1526 | goto out; | |
1527 | ||
12fd3520 | 1528 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1529 | ddir = ffhp->fh_dentry; |
1530 | dirp = ddir->d_inode; | |
1531 | ||
1532 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1533 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1534 | if (IS_ERR(dnew)) |
1535 | goto out_nfserr; | |
1536 | ||
1537 | dold = tfhp->fh_dentry; | |
1538 | dest = dold->d_inode; | |
1539 | ||
6264d69d AV |
1540 | host_err = vfs_link(dold, dirp, dnew); |
1541 | if (!host_err) { | |
1da177e4 | 1542 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1543 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1544 | write_inode_now(dest, 1); |
1545 | } | |
6264d69d | 1546 | err = 0; |
1da177e4 | 1547 | } else { |
6264d69d | 1548 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1549 | err = nfserr_acces; |
1550 | else | |
6264d69d | 1551 | err = nfserrno(host_err); |
1da177e4 LT |
1552 | } |
1553 | ||
1da177e4 | 1554 | dput(dnew); |
270d56e5 DR |
1555 | out_unlock: |
1556 | fh_unlock(ffhp); | |
1da177e4 LT |
1557 | out: |
1558 | return err; | |
1559 | ||
1560 | out_nfserr: | |
6264d69d | 1561 | err = nfserrno(host_err); |
270d56e5 | 1562 | goto out_unlock; |
1da177e4 LT |
1563 | } |
1564 | ||
1565 | /* | |
1566 | * Rename a file | |
1567 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1568 | */ | |
6264d69d | 1569 | __be32 |
1da177e4 LT |
1570 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1571 | struct svc_fh *tfhp, char *tname, int tlen) | |
1572 | { | |
1573 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1574 | struct inode *fdir, *tdir; | |
6264d69d AV |
1575 | __be32 err; |
1576 | int host_err; | |
1da177e4 LT |
1577 | |
1578 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_REMOVE); | |
1579 | if (err) | |
1580 | goto out; | |
1581 | err = fh_verify(rqstp, tfhp, S_IFDIR, MAY_CREATE); | |
1582 | if (err) | |
1583 | goto out; | |
1584 | ||
1585 | fdentry = ffhp->fh_dentry; | |
1586 | fdir = fdentry->d_inode; | |
1587 | ||
1588 | tdentry = tfhp->fh_dentry; | |
1589 | tdir = tdentry->d_inode; | |
1590 | ||
1591 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1592 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1593 | goto out; |
1594 | ||
1595 | err = nfserr_perm; | |
1596 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1597 | goto out; | |
1598 | ||
1599 | /* cannot use fh_lock as we need deadlock protective ordering | |
1600 | * so do it by hand */ | |
1601 | trap = lock_rename(tdentry, fdentry); | |
1602 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1603 | fill_pre_wcc(ffhp); | |
1604 | fill_pre_wcc(tfhp); | |
1605 | ||
1606 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1607 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1608 | if (IS_ERR(odentry)) |
1609 | goto out_nfserr; | |
1610 | ||
6264d69d | 1611 | host_err = -ENOENT; |
1da177e4 LT |
1612 | if (!odentry->d_inode) |
1613 | goto out_dput_old; | |
6264d69d | 1614 | host_err = -EINVAL; |
1da177e4 LT |
1615 | if (odentry == trap) |
1616 | goto out_dput_old; | |
1617 | ||
1618 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1619 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1620 | if (IS_ERR(ndentry)) |
1621 | goto out_dput_old; | |
6264d69d | 1622 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1623 | if (ndentry == trap) |
1624 | goto out_dput_new; | |
1625 | ||
1626 | #ifdef MSNFS | |
1627 | if ((ffhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1628 | ((atomic_read(&odentry->d_count) > 1) | |
1629 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1630 | host_err = -EPERM; |
1da177e4 LT |
1631 | } else |
1632 | #endif | |
6264d69d AV |
1633 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1634 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1635 | host_err = nfsd_sync_dir(tdentry); | |
1636 | if (!host_err) | |
1637 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1638 | } |
1639 | ||
1640 | out_dput_new: | |
1641 | dput(ndentry); | |
1642 | out_dput_old: | |
1643 | dput(odentry); | |
1644 | out_nfserr: | |
6264d69d | 1645 | err = nfserrno(host_err); |
1da177e4 LT |
1646 | |
1647 | /* we cannot reply on fh_unlock on the two filehandles, | |
1648 | * as that would do the wrong thing if the two directories | |
1649 | * were the same, so again we do it by hand | |
1650 | */ | |
1651 | fill_post_wcc(ffhp); | |
1652 | fill_post_wcc(tfhp); | |
1653 | unlock_rename(tdentry, fdentry); | |
1654 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1655 | ||
1656 | out: | |
1657 | return err; | |
1658 | } | |
1659 | ||
1660 | /* | |
1661 | * Unlink a file or directory | |
1662 | * N.B. After this call fhp needs an fh_put | |
1663 | */ | |
6264d69d | 1664 | __be32 |
1da177e4 LT |
1665 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1666 | char *fname, int flen) | |
1667 | { | |
1668 | struct dentry *dentry, *rdentry; | |
1669 | struct inode *dirp; | |
6264d69d AV |
1670 | __be32 err; |
1671 | int host_err; | |
1da177e4 LT |
1672 | |
1673 | err = nfserr_acces; | |
1674 | if (!flen || isdotent(fname, flen)) | |
1675 | goto out; | |
1676 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_REMOVE); | |
1677 | if (err) | |
1678 | goto out; | |
1679 | ||
12fd3520 | 1680 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1681 | dentry = fhp->fh_dentry; |
1682 | dirp = dentry->d_inode; | |
1683 | ||
1684 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1685 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1686 | if (IS_ERR(rdentry)) |
1687 | goto out_nfserr; | |
1688 | ||
1689 | if (!rdentry->d_inode) { | |
1690 | dput(rdentry); | |
1691 | err = nfserr_noent; | |
1692 | goto out; | |
1693 | } | |
1694 | ||
1695 | if (!type) | |
1696 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1697 | ||
1698 | if (type != S_IFDIR) { /* It's UNLINK */ | |
1699 | #ifdef MSNFS | |
1700 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1701 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1702 | host_err = -EPERM; |
1da177e4 LT |
1703 | } else |
1704 | #endif | |
6264d69d | 1705 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1706 | } else { /* It's RMDIR */ |
6264d69d | 1707 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1708 | } |
1709 | ||
1710 | dput(rdentry); | |
1711 | ||
6264d69d AV |
1712 | if (host_err) |
1713 | goto out_nfserr; | |
1714 | if (EX_ISSYNC(fhp->fh_export)) | |
1715 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 LT |
1716 | |
1717 | out_nfserr: | |
6264d69d | 1718 | err = nfserrno(host_err); |
f193fbab YT |
1719 | out: |
1720 | return err; | |
1da177e4 LT |
1721 | } |
1722 | ||
1723 | /* | |
1724 | * Read entries from a directory. | |
1725 | * The NFSv3/4 verifier we ignore for now. | |
1726 | */ | |
6264d69d | 1727 | __be32 |
1da177e4 | 1728 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1729 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1730 | { |
6264d69d AV |
1731 | __be32 err; |
1732 | int host_err; | |
1da177e4 LT |
1733 | struct file *file; |
1734 | loff_t offset = *offsetp; | |
1735 | ||
1736 | err = nfsd_open(rqstp, fhp, S_IFDIR, MAY_READ, &file); | |
1737 | if (err) | |
1738 | goto out; | |
1739 | ||
1740 | offset = vfs_llseek(file, offset, 0); | |
1741 | if (offset < 0) { | |
1742 | err = nfserrno((int)offset); | |
1743 | goto out_close; | |
1744 | } | |
1745 | ||
1746 | /* | |
1747 | * Read the directory entries. This silly loop is necessary because | |
1748 | * readdir() is not guaranteed to fill up the entire buffer, but | |
1749 | * may choose to do less. | |
1750 | */ | |
1751 | ||
1752 | do { | |
1753 | cdp->err = nfserr_eof; /* will be cleared on successful read */ | |
a0ad13ef | 1754 | host_err = vfs_readdir(file, func, cdp); |
6264d69d AV |
1755 | } while (host_err >=0 && cdp->err == nfs_ok); |
1756 | if (host_err) | |
1757 | err = nfserrno(host_err); | |
1da177e4 LT |
1758 | else |
1759 | err = cdp->err; | |
1760 | *offsetp = vfs_llseek(file, 0, 1); | |
1761 | ||
1762 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1763 | err = nfs_ok; /* can still be found in ->err */ | |
1764 | out_close: | |
1765 | nfsd_close(file); | |
1766 | out: | |
1767 | return err; | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * Get file system stats | |
1772 | * N.B. After this call fhp needs an fh_put | |
1773 | */ | |
6264d69d | 1774 | __be32 |
1da177e4 LT |
1775 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat) |
1776 | { | |
6264d69d | 1777 | __be32 err = fh_verify(rqstp, fhp, 0, MAY_NOP); |
726c3342 | 1778 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
1779 | err = nfserr_io; |
1780 | return err; | |
1781 | } | |
1782 | ||
1783 | /* | |
1784 | * Check for a user's access permissions to this inode. | |
1785 | */ | |
6264d69d | 1786 | __be32 |
1da177e4 LT |
1787 | nfsd_permission(struct svc_export *exp, struct dentry *dentry, int acc) |
1788 | { | |
1789 | struct inode *inode = dentry->d_inode; | |
1790 | int err; | |
1791 | ||
1792 | if (acc == MAY_NOP) | |
1793 | return 0; | |
1794 | #if 0 | |
1795 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
1796 | acc, | |
1797 | (acc & MAY_READ)? " read" : "", | |
1798 | (acc & MAY_WRITE)? " write" : "", | |
1799 | (acc & MAY_EXEC)? " exec" : "", | |
1800 | (acc & MAY_SATTR)? " sattr" : "", | |
1801 | (acc & MAY_TRUNC)? " trunc" : "", | |
1802 | (acc & MAY_LOCK)? " lock" : "", | |
1803 | (acc & MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1804 | inode->i_mode, | |
1805 | IS_IMMUTABLE(inode)? " immut" : "", | |
1806 | IS_APPEND(inode)? " append" : "", | |
1807 | IS_RDONLY(inode)? " ro" : ""); | |
1808 | dprintk(" owner %d/%d user %d/%d\n", | |
1809 | inode->i_uid, inode->i_gid, current->fsuid, current->fsgid); | |
1810 | #endif | |
1811 | ||
1812 | /* Normally we reject any write/sattr etc access on a read-only file | |
1813 | * system. But if it is IRIX doing check on write-access for a | |
1814 | * device special file, we ignore rofs. | |
1815 | */ | |
1816 | if (!(acc & MAY_LOCAL_ACCESS)) | |
1817 | if (acc & (MAY_WRITE | MAY_SATTR | MAY_TRUNC)) { | |
1818 | if (EX_RDONLY(exp) || IS_RDONLY(inode)) | |
1819 | return nfserr_rofs; | |
1820 | if (/* (acc & MAY_WRITE) && */ IS_IMMUTABLE(inode)) | |
1821 | return nfserr_perm; | |
1822 | } | |
1823 | if ((acc & MAY_TRUNC) && IS_APPEND(inode)) | |
1824 | return nfserr_perm; | |
1825 | ||
1826 | if (acc & MAY_LOCK) { | |
1827 | /* If we cannot rely on authentication in NLM requests, | |
1828 | * just allow locks, otherwise require read permission, or | |
1829 | * ownership | |
1830 | */ | |
1831 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
1832 | return 0; | |
1833 | else | |
1834 | acc = MAY_READ | MAY_OWNER_OVERRIDE; | |
1835 | } | |
1836 | /* | |
1837 | * The file owner always gets access permission for accesses that | |
1838 | * would normally be checked at open time. This is to make | |
1839 | * file access work even when the client has done a fchmod(fd, 0). | |
1840 | * | |
1841 | * However, `cp foo bar' should fail nevertheless when bar is | |
1842 | * readonly. A sensible way to do this might be to reject all | |
1843 | * attempts to truncate a read-only file, because a creat() call | |
1844 | * always implies file truncation. | |
1845 | * ... but this isn't really fair. A process may reasonably call | |
1846 | * ftruncate on an open file descriptor on a file with perm 000. | |
1847 | * We must trust the client to do permission checking - using "ACCESS" | |
1848 | * with NFSv3. | |
1849 | */ | |
1850 | if ((acc & MAY_OWNER_OVERRIDE) && | |
1851 | inode->i_uid == current->fsuid) | |
1852 | return 0; | |
1853 | ||
1854 | err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL); | |
1855 | ||
1856 | /* Allow read access to binaries even when mode 111 */ | |
1857 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
1858 | acc == (MAY_READ | MAY_OWNER_OVERRIDE)) | |
1859 | err = permission(inode, MAY_EXEC, NULL); | |
1860 | ||
1861 | return err? nfserrno(err) : 0; | |
1862 | } | |
1863 | ||
1864 | void | |
1865 | nfsd_racache_shutdown(void) | |
1866 | { | |
fce1456a | 1867 | if (!raparml) |
1da177e4 LT |
1868 | return; |
1869 | dprintk("nfsd: freeing readahead buffers.\n"); | |
1870 | kfree(raparml); | |
fce1456a | 1871 | raparml = NULL; |
1da177e4 LT |
1872 | } |
1873 | /* | |
1874 | * Initialize readahead param cache | |
1875 | */ | |
1876 | int | |
1877 | nfsd_racache_init(int cache_size) | |
1878 | { | |
1879 | int i; | |
fce1456a GB |
1880 | int j = 0; |
1881 | int nperbucket; | |
1da177e4 | 1882 | |
fce1456a GB |
1883 | |
1884 | if (raparml) | |
1da177e4 | 1885 | return 0; |
fce1456a GB |
1886 | if (cache_size < 2*RAPARM_HASH_SIZE) |
1887 | cache_size = 2*RAPARM_HASH_SIZE; | |
4b3bb06b YB |
1888 | raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL); |
1889 | ||
1890 | if (!raparml) { | |
1da177e4 | 1891 | printk(KERN_WARNING |
4b3bb06b | 1892 | "nfsd: Could not allocate memory read-ahead cache.\n"); |
1da177e4 LT |
1893 | return -ENOMEM; |
1894 | } | |
4b3bb06b YB |
1895 | |
1896 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
1897 | for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) { | |
1898 | raparm_hash[i].pb_head = NULL; | |
1899 | spin_lock_init(&raparm_hash[i].pb_lock); | |
1900 | } | |
1901 | nperbucket = cache_size >> RAPARM_HASH_BITS; | |
1902 | for (i = 0; i < cache_size - 1; i++) { | |
1903 | if (i % nperbucket == 0) | |
1904 | raparm_hash[j++].pb_head = raparml + i; | |
1905 | if (i % nperbucket < nperbucket-1) | |
1906 | raparml[i].p_next = raparml + i + 1; | |
1907 | } | |
1908 | ||
1da177e4 LT |
1909 | nfsdstats.ra_size = cache_size; |
1910 | return 0; | |
1911 | } | |
a257cdd0 AG |
1912 | |
1913 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
1914 | struct posix_acl * | |
1915 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
1916 | { | |
1917 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1918 | char *name; | |
1919 | void *value = NULL; | |
1920 | ssize_t size; | |
1921 | struct posix_acl *acl; | |
1922 | ||
5be196e5 CH |
1923 | if (!IS_POSIXACL(inode)) |
1924 | return ERR_PTR(-EOPNOTSUPP); | |
1925 | ||
1926 | switch (type) { | |
1927 | case ACL_TYPE_ACCESS: | |
1928 | name = POSIX_ACL_XATTR_ACCESS; | |
1929 | break; | |
1930 | case ACL_TYPE_DEFAULT: | |
1931 | name = POSIX_ACL_XATTR_DEFAULT; | |
1932 | break; | |
1933 | default: | |
a257cdd0 | 1934 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
1935 | } |
1936 | ||
5be196e5 CH |
1937 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
1938 | if (size < 0) | |
1939 | return ERR_PTR(size); | |
a257cdd0 | 1940 | |
a257cdd0 | 1941 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
1942 | kfree(value); |
1943 | return acl; | |
1944 | } | |
1945 | ||
1946 | int | |
1947 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
1948 | { | |
1949 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1950 | char *name; | |
1951 | void *value = NULL; | |
1952 | size_t size; | |
1953 | int error; | |
1954 | ||
1955 | if (!IS_POSIXACL(inode) || !inode->i_op || | |
1956 | !inode->i_op->setxattr || !inode->i_op->removexattr) | |
1957 | return -EOPNOTSUPP; | |
1958 | switch(type) { | |
1959 | case ACL_TYPE_ACCESS: | |
334a13ec | 1960 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
1961 | break; |
1962 | case ACL_TYPE_DEFAULT: | |
334a13ec | 1963 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
1964 | break; |
1965 | default: | |
1966 | return -EOPNOTSUPP; | |
1967 | } | |
1968 | ||
1969 | if (acl && acl->a_count) { | |
334a13ec | 1970 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
1971 | value = kmalloc(size, GFP_KERNEL); |
1972 | if (!value) | |
1973 | return -ENOMEM; | |
9ccfc29c FM |
1974 | error = posix_acl_to_xattr(acl, value, size); |
1975 | if (error < 0) | |
a257cdd0 | 1976 | goto getout; |
9ccfc29c | 1977 | size = error; |
a257cdd0 AG |
1978 | } else |
1979 | size = 0; | |
1980 | ||
a257cdd0 | 1981 | if (size) |
5be196e5 | 1982 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
1983 | else { |
1984 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
1985 | error = 0; | |
1986 | else { | |
5be196e5 | 1987 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
1988 | if (error == -ENODATA) |
1989 | error = 0; | |
1990 | } | |
1991 | } | |
1992 | ||
1993 | getout: | |
1994 | kfree(value); | |
1995 | return error; | |
1996 | } | |
1997 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |