]>
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 | ||
0989a788 | 116 | exp2 = rqst_exp_get_by_name(rqstp, mnt, mounts); |
1da177e4 LT |
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 | ||
1da177e4 LT |
171 | /* Lookup the name, but don't follow links */ |
172 | if (isdotent(name, len)) { | |
173 | if (len==1) | |
174 | dentry = dget(dparent); | |
175 | else if (dparent != exp->ex_dentry) { | |
176 | dentry = dget_parent(dparent); | |
177 | } else if (!EX_NOHIDE(exp)) | |
178 | dentry = dget(dparent); /* .. == . just like at / */ | |
179 | else { | |
180 | /* checking mountpoint crossing is very different when stepping up */ | |
181 | struct svc_export *exp2 = NULL; | |
182 | struct dentry *dp; | |
183 | struct vfsmount *mnt = mntget(exp->ex_mnt); | |
184 | dentry = dget(dparent); | |
185 | while(dentry == mnt->mnt_root && follow_up(&mnt, &dentry)) | |
186 | ; | |
187 | dp = dget_parent(dentry); | |
188 | dput(dentry); | |
189 | dentry = dp; | |
190 | ||
0989a788 | 191 | exp2 = rqst_exp_parent(rqstp, mnt, dentry); |
2d3bb252 BF |
192 | if (PTR_ERR(exp2) == -ENOENT) { |
193 | dput(dentry); | |
194 | dentry = dget(dparent); | |
195 | } else if (IS_ERR(exp2)) { | |
6264d69d | 196 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
197 | dput(dentry); |
198 | mntput(mnt); | |
199 | goto out_nfserr; | |
1da177e4 LT |
200 | } else { |
201 | exp_put(exp); | |
202 | exp = exp2; | |
203 | } | |
204 | mntput(mnt); | |
205 | } | |
206 | } else { | |
207 | fh_lock(fhp); | |
208 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 209 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
210 | if (IS_ERR(dentry)) |
211 | goto out_nfserr; | |
212 | /* | |
213 | * check if we have crossed a mount point ... | |
214 | */ | |
215 | if (d_mountpoint(dentry)) { | |
6264d69d | 216 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
217 | dput(dentry); |
218 | goto out_nfserr; | |
219 | } | |
220 | } | |
221 | } | |
222 | /* | |
223 | * Note: we compose the file handle now, but as the | |
224 | * dentry may be negative, it may need to be updated. | |
225 | */ | |
226 | err = fh_compose(resfh, exp, dentry, fhp); | |
227 | if (!err && !dentry->d_inode) | |
228 | err = nfserr_noent; | |
229 | dput(dentry); | |
230 | out: | |
231 | exp_put(exp); | |
232 | return err; | |
233 | ||
234 | out_nfserr: | |
6264d69d | 235 | err = nfserrno(host_err); |
1da177e4 LT |
236 | goto out; |
237 | } | |
238 | ||
239 | /* | |
240 | * Set various file attributes. | |
241 | * N.B. After this call fhp needs an fh_put | |
242 | */ | |
6264d69d | 243 | __be32 |
1da177e4 LT |
244 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
245 | int check_guard, time_t guardtime) | |
246 | { | |
247 | struct dentry *dentry; | |
248 | struct inode *inode; | |
249 | int accmode = MAY_SATTR; | |
250 | int ftype = 0; | |
251 | int imode; | |
6264d69d AV |
252 | __be32 err; |
253 | int host_err; | |
1da177e4 LT |
254 | int size_change = 0; |
255 | ||
256 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
257 | accmode |= MAY_WRITE|MAY_OWNER_OVERRIDE; | |
258 | if (iap->ia_valid & ATTR_SIZE) | |
259 | ftype = S_IFREG; | |
260 | ||
261 | /* Get inode */ | |
262 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 263 | if (err) |
1da177e4 LT |
264 | goto out; |
265 | ||
266 | dentry = fhp->fh_dentry; | |
267 | inode = dentry->d_inode; | |
268 | ||
15b7a1b8 N |
269 | /* Ignore any mode updates on symlinks */ |
270 | if (S_ISLNK(inode->i_mode)) | |
271 | iap->ia_valid &= ~ATTR_MODE; | |
272 | ||
273 | if (!iap->ia_valid) | |
274 | goto out; | |
275 | ||
1da177e4 LT |
276 | /* NFSv2 does not differentiate between "set-[ac]time-to-now" |
277 | * which only requires access, and "set-[ac]time-to-X" which | |
278 | * requires ownership. | |
279 | * So if it looks like it might be "set both to the same time which | |
280 | * is close to now", and if inode_change_ok fails, then we | |
281 | * convert to "set to now" instead of "set to explicit time" | |
282 | * | |
283 | * We only call inode_change_ok as the last test as technically | |
284 | * it is not an interface that we should be using. It is only | |
285 | * valid if the filesystem does not define it's own i_op->setattr. | |
286 | */ | |
287 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
288 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
289 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET | |
290 | && iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec | |
291 | ) { | |
292 | /* Looks probable. Now just make sure time is in the right ballpark. | |
293 | * Solaris, at least, doesn't seem to care what the time request is. | |
294 | * We require it be within 30 minutes of now. | |
295 | */ | |
296 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); | |
297 | if (delta<0) delta = -delta; | |
298 | if (delta < MAX_TOUCH_TIME_ERROR && | |
299 | inode_change_ok(inode, iap) != 0) { | |
300 | /* turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME | |
301 | * this will cause notify_change to set these times to "now" | |
302 | */ | |
303 | iap->ia_valid &= ~BOTH_TIME_SET; | |
304 | } | |
305 | } | |
306 | ||
307 | /* The size case is special. It changes the file as well as the attributes. */ | |
308 | if (iap->ia_valid & ATTR_SIZE) { | |
309 | if (iap->ia_size < inode->i_size) { | |
310 | err = nfsd_permission(fhp->fh_export, dentry, MAY_TRUNC|MAY_OWNER_OVERRIDE); | |
311 | if (err) | |
312 | goto out; | |
313 | } | |
314 | ||
315 | /* | |
316 | * If we are changing the size of the file, then | |
317 | * we need to break all leases. | |
318 | */ | |
6264d69d AV |
319 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
320 | if (host_err == -EWOULDBLOCK) | |
321 | host_err = -ETIMEDOUT; | |
322 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
323 | goto out_nfserr; |
324 | ||
6264d69d AV |
325 | host_err = get_write_access(inode); |
326 | if (host_err) | |
1da177e4 LT |
327 | goto out_nfserr; |
328 | ||
329 | size_change = 1; | |
6264d69d AV |
330 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
331 | if (host_err) { | |
1da177e4 LT |
332 | put_write_access(inode); |
333 | goto out_nfserr; | |
334 | } | |
335 | DQUOT_INIT(inode); | |
336 | } | |
337 | ||
338 | imode = inode->i_mode; | |
339 | if (iap->ia_valid & ATTR_MODE) { | |
340 | iap->ia_mode &= S_IALLUGO; | |
341 | imode = iap->ia_mode |= (imode & ~S_IALLUGO); | |
342 | } | |
343 | ||
344 | /* Revoke setuid/setgid bit on chown/chgrp */ | |
345 | if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) | |
346 | iap->ia_valid |= ATTR_KILL_SUID; | |
347 | if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid) | |
348 | iap->ia_valid |= ATTR_KILL_SGID; | |
349 | ||
350 | /* Change the attributes. */ | |
351 | ||
352 | iap->ia_valid |= ATTR_CTIME; | |
353 | ||
354 | err = nfserr_notsync; | |
355 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
356 | fh_lock(fhp); | |
6264d69d AV |
357 | host_err = notify_change(dentry, iap); |
358 | err = nfserrno(host_err); | |
1da177e4 LT |
359 | fh_unlock(fhp); |
360 | } | |
361 | if (size_change) | |
362 | put_write_access(inode); | |
363 | if (!err) | |
364 | if (EX_ISSYNC(fhp->fh_export)) | |
365 | write_inode_now(inode, 1); | |
366 | out: | |
367 | return err; | |
368 | ||
369 | out_nfserr: | |
6264d69d | 370 | err = nfserrno(host_err); |
1da177e4 LT |
371 | goto out; |
372 | } | |
373 | ||
5be196e5 CH |
374 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
375 | defined(CONFIG_NFSD_V3_ACL) || \ | |
376 | defined(CONFIG_NFSD_V4) | |
377 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
378 | { | |
379 | ssize_t buflen; | |
5be196e5 CH |
380 | |
381 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
382 | if (buflen <= 0) | |
383 | return buflen; | |
1da177e4 | 384 | |
5be196e5 CH |
385 | *buf = kmalloc(buflen, GFP_KERNEL); |
386 | if (!*buf) | |
387 | return -ENOMEM; | |
388 | ||
b5872b0d | 389 | return vfs_getxattr(dentry, key, *buf, buflen); |
5be196e5 CH |
390 | } |
391 | #endif | |
392 | ||
393 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
394 | static int |
395 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
396 | { | |
397 | int len; | |
398 | size_t buflen; | |
399 | char *buf = NULL; | |
400 | int error = 0; | |
1da177e4 LT |
401 | |
402 | buflen = posix_acl_xattr_size(pacl->a_count); | |
403 | buf = kmalloc(buflen, GFP_KERNEL); | |
404 | error = -ENOMEM; | |
405 | if (buf == NULL) | |
406 | goto out; | |
407 | ||
408 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
409 | if (len < 0) { | |
410 | error = len; | |
411 | goto out; | |
412 | } | |
413 | ||
5be196e5 | 414 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
415 | out: |
416 | kfree(buf); | |
417 | return error; | |
418 | } | |
419 | ||
6264d69d | 420 | __be32 |
1da177e4 LT |
421 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
422 | struct nfs4_acl *acl) | |
423 | { | |
6264d69d AV |
424 | __be32 error; |
425 | int host_error; | |
1da177e4 LT |
426 | struct dentry *dentry; |
427 | struct inode *inode; | |
428 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
429 | unsigned int flags = 0; | |
430 | ||
431 | /* Get inode */ | |
432 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, MAY_SATTR); | |
433 | if (error) | |
4b2ca38a | 434 | return error; |
1da177e4 LT |
435 | |
436 | dentry = fhp->fh_dentry; | |
437 | inode = dentry->d_inode; | |
438 | if (S_ISDIR(inode->i_mode)) | |
439 | flags = NFS4_ACL_DIR; | |
440 | ||
6264d69d AV |
441 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
442 | if (host_error == -EINVAL) { | |
4b2ca38a | 443 | return nfserr_attrnotsupp; |
6264d69d | 444 | } else if (host_error < 0) |
1da177e4 LT |
445 | goto out_nfserr; |
446 | ||
6264d69d AV |
447 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
448 | if (host_error < 0) | |
4b2ca38a | 449 | goto out_release; |
1da177e4 | 450 | |
4b2ca38a | 451 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 452 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 453 | |
4b2ca38a | 454 | out_release: |
1da177e4 LT |
455 | posix_acl_release(pacl); |
456 | posix_acl_release(dpacl); | |
1da177e4 | 457 | out_nfserr: |
f34f9242 | 458 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 459 | return nfserr_attrnotsupp; |
f34f9242 | 460 | else |
4b2ca38a | 461 | return nfserrno(host_error); |
1da177e4 LT |
462 | } |
463 | ||
464 | static struct posix_acl * | |
465 | _get_posix_acl(struct dentry *dentry, char *key) | |
466 | { | |
5be196e5 | 467 | void *buf = NULL; |
1da177e4 | 468 | struct posix_acl *pacl = NULL; |
5be196e5 | 469 | int buflen; |
1da177e4 | 470 | |
5be196e5 CH |
471 | buflen = nfsd_getxattr(dentry, key, &buf); |
472 | if (!buflen) | |
473 | buflen = -ENODATA; | |
474 | if (buflen <= 0) | |
475 | return ERR_PTR(buflen); | |
1da177e4 LT |
476 | |
477 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
478 | kfree(buf); |
479 | return pacl; | |
1da177e4 LT |
480 | } |
481 | ||
482 | int | |
483 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
484 | { | |
485 | struct inode *inode = dentry->d_inode; | |
486 | int error = 0; | |
487 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
488 | unsigned int flags = 0; | |
489 | ||
9a59f452 | 490 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
491 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
492 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
493 | if (IS_ERR(pacl)) { | |
494 | error = PTR_ERR(pacl); | |
495 | pacl = NULL; | |
496 | goto out; | |
497 | } | |
498 | ||
499 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 500 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
501 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
502 | dpacl = NULL; | |
503 | else if (IS_ERR(dpacl)) { | |
504 | error = PTR_ERR(dpacl); | |
505 | dpacl = NULL; | |
506 | goto out; | |
507 | } | |
508 | flags = NFS4_ACL_DIR; | |
509 | } | |
510 | ||
511 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
512 | if (IS_ERR(*acl)) { | |
513 | error = PTR_ERR(*acl); | |
514 | *acl = NULL; | |
515 | } | |
516 | out: | |
517 | posix_acl_release(pacl); | |
518 | posix_acl_release(dpacl); | |
519 | return error; | |
520 | } | |
521 | ||
522 | #endif /* defined(CONFIG_NFS_V4) */ | |
523 | ||
524 | #ifdef CONFIG_NFSD_V3 | |
525 | /* | |
526 | * Check server access rights to a file system object | |
527 | */ | |
528 | struct accessmap { | |
529 | u32 access; | |
530 | int how; | |
531 | }; | |
532 | static struct accessmap nfs3_regaccess[] = { | |
533 | { NFS3_ACCESS_READ, MAY_READ }, | |
534 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
535 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_TRUNC }, | |
536 | { NFS3_ACCESS_EXTEND, MAY_WRITE }, | |
537 | ||
538 | { 0, 0 } | |
539 | }; | |
540 | ||
541 | static struct accessmap nfs3_diraccess[] = { | |
542 | { NFS3_ACCESS_READ, MAY_READ }, | |
543 | { NFS3_ACCESS_LOOKUP, MAY_EXEC }, | |
544 | { NFS3_ACCESS_MODIFY, MAY_EXEC|MAY_WRITE|MAY_TRUNC }, | |
545 | { NFS3_ACCESS_EXTEND, MAY_EXEC|MAY_WRITE }, | |
546 | { NFS3_ACCESS_DELETE, MAY_REMOVE }, | |
547 | ||
548 | { 0, 0 } | |
549 | }; | |
550 | ||
551 | static struct accessmap nfs3_anyaccess[] = { | |
552 | /* Some clients - Solaris 2.6 at least, make an access call | |
553 | * to the server to check for access for things like /dev/null | |
554 | * (which really, the server doesn't care about). So | |
555 | * We provide simple access checking for them, looking | |
556 | * mainly at mode bits, and we make sure to ignore read-only | |
557 | * filesystem checks | |
558 | */ | |
559 | { NFS3_ACCESS_READ, MAY_READ }, | |
560 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
561 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
562 | { NFS3_ACCESS_EXTEND, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
563 | ||
564 | { 0, 0 } | |
565 | }; | |
566 | ||
6264d69d | 567 | __be32 |
1da177e4 LT |
568 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
569 | { | |
570 | struct accessmap *map; | |
571 | struct svc_export *export; | |
572 | struct dentry *dentry; | |
573 | u32 query, result = 0, sresult = 0; | |
6264d69d | 574 | __be32 error; |
1da177e4 LT |
575 | |
576 | error = fh_verify(rqstp, fhp, 0, MAY_NOP); | |
577 | if (error) | |
578 | goto out; | |
579 | ||
580 | export = fhp->fh_export; | |
581 | dentry = fhp->fh_dentry; | |
582 | ||
583 | if (S_ISREG(dentry->d_inode->i_mode)) | |
584 | map = nfs3_regaccess; | |
585 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
586 | map = nfs3_diraccess; | |
587 | else | |
588 | map = nfs3_anyaccess; | |
589 | ||
590 | ||
591 | query = *access; | |
592 | for (; map->access; map++) { | |
593 | if (map->access & query) { | |
6264d69d | 594 | __be32 err2; |
1da177e4 LT |
595 | |
596 | sresult |= map->access; | |
597 | ||
598 | err2 = nfsd_permission(export, dentry, map->how); | |
599 | switch (err2) { | |
600 | case nfs_ok: | |
601 | result |= map->access; | |
602 | break; | |
603 | ||
604 | /* the following error codes just mean the access was not allowed, | |
605 | * rather than an error occurred */ | |
606 | case nfserr_rofs: | |
607 | case nfserr_acces: | |
608 | case nfserr_perm: | |
609 | /* simply don't "or" in the access bit. */ | |
610 | break; | |
611 | default: | |
612 | error = err2; | |
613 | goto out; | |
614 | } | |
615 | } | |
616 | } | |
617 | *access = result; | |
618 | if (supported) | |
619 | *supported = sresult; | |
620 | ||
621 | out: | |
622 | return error; | |
623 | } | |
624 | #endif /* CONFIG_NFSD_V3 */ | |
625 | ||
626 | ||
627 | ||
628 | /* | |
629 | * Open an existing file or directory. | |
630 | * The access argument indicates the type of open (read/write/lock) | |
631 | * N.B. After this call fhp needs an fh_put | |
632 | */ | |
6264d69d | 633 | __be32 |
1da177e4 LT |
634 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
635 | int access, struct file **filp) | |
636 | { | |
637 | struct dentry *dentry; | |
638 | struct inode *inode; | |
6264d69d AV |
639 | int flags = O_RDONLY|O_LARGEFILE; |
640 | __be32 err; | |
641 | int host_err; | |
1da177e4 LT |
642 | |
643 | /* | |
644 | * If we get here, then the client has already done an "open", | |
645 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
646 | * in case a chmod has now revoked permission. | |
647 | */ | |
648 | err = fh_verify(rqstp, fhp, type, access | MAY_OWNER_OVERRIDE); | |
649 | if (err) | |
650 | goto out; | |
651 | ||
652 | dentry = fhp->fh_dentry; | |
653 | inode = dentry->d_inode; | |
654 | ||
655 | /* Disallow write access to files with the append-only bit set | |
656 | * or any access when mandatory locking enabled | |
657 | */ | |
658 | err = nfserr_perm; | |
659 | if (IS_APPEND(inode) && (access & MAY_WRITE)) | |
660 | goto out; | |
661 | if (IS_ISMNDLK(inode)) | |
662 | goto out; | |
663 | ||
664 | if (!inode->i_fop) | |
665 | goto out; | |
666 | ||
667 | /* | |
668 | * Check to see if there are any leases on this file. | |
669 | * This may block while leases are broken. | |
670 | */ | |
6264d69d AV |
671 | host_err = break_lease(inode, O_NONBLOCK | ((access & MAY_WRITE) ? FMODE_WRITE : 0)); |
672 | if (host_err == -EWOULDBLOCK) | |
673 | host_err = -ETIMEDOUT; | |
674 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
675 | goto out_nfserr; |
676 | ||
677 | if (access & MAY_WRITE) { | |
9ecb6a08 BF |
678 | if (access & MAY_READ) |
679 | flags = O_RDWR|O_LARGEFILE; | |
680 | else | |
681 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 LT |
682 | |
683 | DQUOT_INIT(inode); | |
684 | } | |
685 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_mnt), flags); | |
686 | if (IS_ERR(*filp)) | |
6264d69d | 687 | host_err = PTR_ERR(*filp); |
1da177e4 | 688 | out_nfserr: |
6264d69d | 689 | err = nfserrno(host_err); |
1da177e4 LT |
690 | out: |
691 | return err; | |
692 | } | |
693 | ||
694 | /* | |
695 | * Close a file. | |
696 | */ | |
697 | void | |
698 | nfsd_close(struct file *filp) | |
699 | { | |
700 | fput(filp); | |
701 | } | |
702 | ||
703 | /* | |
704 | * Sync a file | |
705 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
706 | * after it. | |
707 | */ | |
a334de28 | 708 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, |
99ac48f5 | 709 | const struct file_operations *fop) |
1da177e4 LT |
710 | { |
711 | struct inode *inode = dp->d_inode; | |
712 | int (*fsync) (struct file *, struct dentry *, int); | |
f193fbab | 713 | int err; |
1da177e4 | 714 | |
f193fbab YT |
715 | err = filemap_fdatawrite(inode->i_mapping); |
716 | if (err == 0 && fop && (fsync = fop->fsync)) | |
717 | err = fsync(filp, dp, 0); | |
718 | if (err == 0) | |
719 | err = filemap_fdatawait(inode->i_mapping); | |
a334de28 | 720 | |
f193fbab | 721 | return err; |
1da177e4 LT |
722 | } |
723 | ||
724 | ||
a334de28 | 725 | static int |
1da177e4 LT |
726 | nfsd_sync(struct file *filp) |
727 | { | |
a334de28 | 728 | int err; |
7eaa36e2 JJS |
729 | struct inode *inode = filp->f_path.dentry->d_inode; |
730 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
1b1dcc1b | 731 | mutex_lock(&inode->i_mutex); |
7eaa36e2 | 732 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); |
1b1dcc1b | 733 | mutex_unlock(&inode->i_mutex); |
a334de28 DS |
734 | |
735 | return err; | |
1da177e4 LT |
736 | } |
737 | ||
f193fbab | 738 | int |
1da177e4 LT |
739 | nfsd_sync_dir(struct dentry *dp) |
740 | { | |
f193fbab | 741 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
742 | } |
743 | ||
744 | /* | |
745 | * Obtain the readahead parameters for the file | |
746 | * specified by (dev, ino). | |
747 | */ | |
1da177e4 LT |
748 | |
749 | static inline struct raparms * | |
750 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
751 | { | |
752 | struct raparms *ra, **rap, **frap = NULL; | |
753 | int depth = 0; | |
fce1456a GB |
754 | unsigned int hash; |
755 | struct raparm_hbucket *rab; | |
756 | ||
757 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
758 | rab = &raparm_hash[hash]; | |
1da177e4 | 759 | |
fce1456a GB |
760 | spin_lock(&rab->pb_lock); |
761 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
762 | if (ra->p_ino == ino && ra->p_dev == dev) |
763 | goto found; | |
764 | depth++; | |
765 | if (ra->p_count == 0) | |
766 | frap = rap; | |
767 | } | |
768 | depth = nfsdstats.ra_size*11/10; | |
769 | if (!frap) { | |
fce1456a | 770 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
771 | return NULL; |
772 | } | |
773 | rap = frap; | |
774 | ra = *frap; | |
775 | ra->p_dev = dev; | |
776 | ra->p_ino = ino; | |
777 | ra->p_set = 0; | |
fce1456a | 778 | ra->p_hindex = hash; |
1da177e4 | 779 | found: |
fce1456a | 780 | if (rap != &rab->pb_head) { |
1da177e4 | 781 | *rap = ra->p_next; |
fce1456a GB |
782 | ra->p_next = rab->pb_head; |
783 | rab->pb_head = ra; | |
1da177e4 LT |
784 | } |
785 | ra->p_count++; | |
786 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 787 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
788 | return ra; |
789 | } | |
790 | ||
791 | /* | |
cf8208d0 JA |
792 | * Grab and keep cached pages associated with a file in the svc_rqst |
793 | * so that they can be passed to the network sendmsg/sendpage routines | |
794 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
795 | */ |
796 | static int | |
cf8208d0 JA |
797 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
798 | struct splice_desc *sd) | |
1da177e4 | 799 | { |
cf8208d0 | 800 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 801 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
802 | struct page *page = buf->page; |
803 | size_t size; | |
804 | int ret; | |
1da177e4 | 805 | |
cac36bb0 | 806 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
807 | if (unlikely(ret)) |
808 | return ret; | |
809 | ||
810 | size = sd->len; | |
1da177e4 LT |
811 | |
812 | if (rqstp->rq_res.page_len == 0) { | |
813 | get_page(page); | |
44524359 N |
814 | put_page(*pp); |
815 | *pp = page; | |
816 | rqstp->rq_resused++; | |
cf8208d0 | 817 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 818 | rqstp->rq_res.page_len = size; |
44524359 | 819 | } else if (page != pp[-1]) { |
1da177e4 | 820 | get_page(page); |
250f3915 N |
821 | if (*pp) |
822 | put_page(*pp); | |
44524359 N |
823 | *pp = page; |
824 | rqstp->rq_resused++; | |
1da177e4 | 825 | rqstp->rq_res.page_len += size; |
44524359 | 826 | } else |
1da177e4 | 827 | rqstp->rq_res.page_len += size; |
1da177e4 | 828 | |
1da177e4 LT |
829 | return size; |
830 | } | |
831 | ||
cf8208d0 JA |
832 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
833 | struct splice_desc *sd) | |
834 | { | |
835 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
836 | } | |
837 | ||
6264d69d | 838 | static __be32 |
1da177e4 LT |
839 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
840 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
841 | { | |
842 | struct inode *inode; | |
843 | struct raparms *ra; | |
844 | mm_segment_t oldfs; | |
6264d69d AV |
845 | __be32 err; |
846 | int host_err; | |
1da177e4 LT |
847 | |
848 | err = nfserr_perm; | |
7eaa36e2 | 849 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
850 | #ifdef MSNFS |
851 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
852 | (!lock_may_read(inode, offset, *count))) | |
853 | goto out; | |
854 | #endif | |
855 | ||
856 | /* Get readahead parameters */ | |
857 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
858 | ||
859 | if (ra && ra->p_set) | |
860 | file->f_ra = ra->p_ra; | |
861 | ||
cf8208d0 JA |
862 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
863 | struct splice_desc sd = { | |
864 | .len = 0, | |
865 | .total_len = *count, | |
866 | .pos = offset, | |
867 | .u.data = rqstp, | |
868 | }; | |
869 | ||
4fbef206 | 870 | rqstp->rq_resused = 1; |
cf8208d0 | 871 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
872 | } else { |
873 | oldfs = get_fs(); | |
874 | set_fs(KERNEL_DS); | |
6264d69d | 875 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
876 | set_fs(oldfs); |
877 | } | |
878 | ||
879 | /* Write back readahead params */ | |
880 | if (ra) { | |
fce1456a GB |
881 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
882 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
883 | ra->p_ra = file->f_ra; |
884 | ra->p_set = 1; | |
885 | ra->p_count--; | |
fce1456a | 886 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
887 | } |
888 | ||
6264d69d AV |
889 | if (host_err >= 0) { |
890 | nfsdstats.io_read += host_err; | |
891 | *count = host_err; | |
1da177e4 | 892 | err = 0; |
7eaa36e2 | 893 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 894 | } else |
6264d69d | 895 | err = nfserrno(host_err); |
1da177e4 LT |
896 | out: |
897 | return err; | |
898 | } | |
899 | ||
9f708e40 NB |
900 | static void kill_suid(struct dentry *dentry) |
901 | { | |
902 | struct iattr ia; | |
903 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID; | |
904 | ||
1b1dcc1b | 905 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 906 | notify_change(dentry, &ia); |
1b1dcc1b | 907 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
908 | } |
909 | ||
6264d69d | 910 | static __be32 |
1da177e4 LT |
911 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
912 | loff_t offset, struct kvec *vec, int vlen, | |
913 | unsigned long cnt, int *stablep) | |
914 | { | |
915 | struct svc_export *exp; | |
916 | struct dentry *dentry; | |
917 | struct inode *inode; | |
918 | mm_segment_t oldfs; | |
6264d69d AV |
919 | __be32 err = 0; |
920 | int host_err; | |
1da177e4 LT |
921 | int stable = *stablep; |
922 | ||
45bd3b3d | 923 | #ifdef MSNFS |
1da177e4 LT |
924 | err = nfserr_perm; |
925 | ||
1da177e4 | 926 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
7eaa36e2 | 927 | (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt))) |
1da177e4 LT |
928 | goto out; |
929 | #endif | |
930 | ||
7eaa36e2 | 931 | dentry = file->f_path.dentry; |
1da177e4 LT |
932 | inode = dentry->d_inode; |
933 | exp = fhp->fh_export; | |
934 | ||
935 | /* | |
936 | * Request sync writes if | |
937 | * - the sync export option has been set, or | |
938 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
939 | * - The file system doesn't support fsync(). | |
940 | * When gathered writes have been configured for this volume, | |
941 | * flushing the data to disk is handled separately below. | |
942 | */ | |
943 | ||
944 | if (file->f_op->fsync == 0) {/* COMMIT3 cannot work */ | |
945 | stable = 2; | |
946 | *stablep = 2; /* FILE_SYNC */ | |
947 | } | |
948 | ||
949 | if (!EX_ISSYNC(exp)) | |
950 | stable = 0; | |
951 | if (stable && !EX_WGATHER(exp)) | |
952 | file->f_flags |= O_SYNC; | |
953 | ||
954 | /* Write the data. */ | |
955 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 956 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 957 | set_fs(oldfs); |
6264d69d | 958 | if (host_err >= 0) { |
1da177e4 | 959 | nfsdstats.io_write += cnt; |
7eaa36e2 | 960 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
961 | } |
962 | ||
963 | /* clear setuid/setgid flag after write */ | |
6264d69d | 964 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 965 | kill_suid(dentry); |
1da177e4 | 966 | |
6264d69d | 967 | if (host_err >= 0 && stable) { |
1da177e4 LT |
968 | static ino_t last_ino; |
969 | static dev_t last_dev; | |
970 | ||
971 | /* | |
972 | * Gathered writes: If another process is currently | |
973 | * writing to the file, there's a high chance | |
974 | * this is another nfsd (triggered by a bulk write | |
975 | * from a client's biod). Rather than syncing the | |
976 | * file with each write request, we sleep for 10 msec. | |
977 | * | |
978 | * I don't know if this roughly approximates | |
979 | * C. Juszak's idea of gathered writes, but it's a | |
980 | * nice and simple solution (IMHO), and it seems to | |
981 | * work:-) | |
982 | */ | |
983 | if (EX_WGATHER(exp)) { | |
984 | if (atomic_read(&inode->i_writecount) > 1 | |
985 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
986 | dprintk("nfsd: write defer %d\n", current->pid); | |
987 | msleep(10); | |
988 | dprintk("nfsd: write resume %d\n", current->pid); | |
989 | } | |
990 | ||
991 | if (inode->i_state & I_DIRTY) { | |
992 | dprintk("nfsd: write sync %d\n", current->pid); | |
6264d69d | 993 | host_err=nfsd_sync(file); |
1da177e4 LT |
994 | } |
995 | #if 0 | |
996 | wake_up(&inode->i_wait); | |
997 | #endif | |
998 | } | |
999 | last_ino = inode->i_ino; | |
1000 | last_dev = inode->i_sb->s_dev; | |
1001 | } | |
1002 | ||
6264d69d AV |
1003 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
1004 | if (host_err >= 0) | |
1da177e4 LT |
1005 | err = 0; |
1006 | else | |
6264d69d | 1007 | err = nfserrno(host_err); |
1da177e4 LT |
1008 | out: |
1009 | return err; | |
1010 | } | |
1011 | ||
1012 | /* | |
1013 | * Read data from a file. count must contain the requested read count | |
1014 | * on entry. On return, *count contains the number of bytes actually read. | |
1015 | * N.B. After this call fhp needs an fh_put | |
1016 | */ | |
6264d69d | 1017 | __be32 |
1da177e4 LT |
1018 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1019 | loff_t offset, struct kvec *vec, int vlen, | |
1020 | unsigned long *count) | |
1021 | { | |
6264d69d | 1022 | __be32 err; |
1da177e4 LT |
1023 | |
1024 | if (file) { | |
1025 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1026 | MAY_READ|MAY_OWNER_OVERRIDE); | |
1027 | if (err) | |
1028 | goto out; | |
1029 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1030 | } else { | |
1031 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_READ, &file); | |
1032 | if (err) | |
1033 | goto out; | |
1034 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1035 | nfsd_close(file); | |
1036 | } | |
1037 | out: | |
1038 | return err; | |
1039 | } | |
1040 | ||
1041 | /* | |
1042 | * Write data to a file. | |
1043 | * The stable flag requests synchronous writes. | |
1044 | * N.B. After this call fhp needs an fh_put | |
1045 | */ | |
6264d69d | 1046 | __be32 |
1da177e4 LT |
1047 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1048 | loff_t offset, struct kvec *vec, int vlen, unsigned long cnt, | |
1049 | int *stablep) | |
1050 | { | |
6264d69d | 1051 | __be32 err = 0; |
1da177e4 LT |
1052 | |
1053 | if (file) { | |
1054 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1055 | MAY_WRITE|MAY_OWNER_OVERRIDE); | |
1056 | if (err) | |
1057 | goto out; | |
1058 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1059 | stablep); | |
1060 | } else { | |
1061 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file); | |
1062 | if (err) | |
1063 | goto out; | |
1064 | ||
1065 | if (cnt) | |
1066 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1067 | cnt, stablep); | |
1068 | nfsd_close(file); | |
1069 | } | |
1070 | out: | |
1071 | return err; | |
1072 | } | |
1073 | ||
1074 | #ifdef CONFIG_NFSD_V3 | |
1075 | /* | |
1076 | * Commit all pending writes to stable storage. | |
1077 | * Strictly speaking, we could sync just the indicated file region here, | |
1078 | * but there's currently no way we can ask the VFS to do so. | |
1079 | * | |
1080 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1081 | * data to the client, as locking happens lower down in the filesystem. | |
1082 | */ | |
6264d69d | 1083 | __be32 |
1da177e4 LT |
1084 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1085 | loff_t offset, unsigned long count) | |
1086 | { | |
1087 | struct file *file; | |
6264d69d | 1088 | __be32 err; |
1da177e4 LT |
1089 | |
1090 | if ((u64)count > ~(u64)offset) | |
1091 | return nfserr_inval; | |
1092 | ||
1093 | if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0) | |
1094 | return err; | |
1095 | if (EX_ISSYNC(fhp->fh_export)) { | |
1096 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1097 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1098 | } else { |
1099 | err = nfserr_notsupp; | |
1100 | } | |
1101 | } | |
1102 | ||
1103 | nfsd_close(file); | |
1104 | return err; | |
1105 | } | |
1106 | #endif /* CONFIG_NFSD_V3 */ | |
1107 | ||
1108 | /* | |
1109 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1110 | * not yet implemented. | |
1111 | * If the response fh has been verified, the parent directory should | |
1112 | * already be locked. Note that the parent directory is left locked. | |
1113 | * | |
1114 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1115 | */ | |
6264d69d | 1116 | __be32 |
1da177e4 LT |
1117 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1118 | char *fname, int flen, struct iattr *iap, | |
1119 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1120 | { | |
1121 | struct dentry *dentry, *dchild = NULL; | |
1122 | struct inode *dirp; | |
6264d69d AV |
1123 | __be32 err; |
1124 | int host_err; | |
1da177e4 LT |
1125 | |
1126 | err = nfserr_perm; | |
1127 | if (!flen) | |
1128 | goto out; | |
1129 | err = nfserr_exist; | |
1130 | if (isdotent(fname, flen)) | |
1131 | goto out; | |
1132 | ||
1133 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1134 | if (err) | |
1135 | goto out; | |
1136 | ||
1137 | dentry = fhp->fh_dentry; | |
1138 | dirp = dentry->d_inode; | |
1139 | ||
1140 | err = nfserr_notdir; | |
1141 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1142 | goto out; | |
1143 | /* | |
1144 | * Check whether the response file handle has been verified yet. | |
1145 | * If it has, the parent directory should already be locked. | |
1146 | */ | |
1147 | if (!resfhp->fh_dentry) { | |
1148 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1149 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1150 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1151 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1152 | if (IS_ERR(dchild)) |
1153 | goto out_nfserr; | |
1154 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1155 | if (err) | |
1156 | goto out; | |
1157 | } else { | |
1158 | /* called from nfsd_proc_create */ | |
1159 | dchild = dget(resfhp->fh_dentry); | |
1160 | if (!fhp->fh_locked) { | |
1161 | /* not actually possible */ | |
1162 | printk(KERN_ERR | |
1163 | "nfsd_create: parent %s/%s not locked!\n", | |
1164 | dentry->d_parent->d_name.name, | |
1165 | dentry->d_name.name); | |
d75f2b9f | 1166 | err = nfserr_io; |
1da177e4 LT |
1167 | goto out; |
1168 | } | |
1169 | } | |
1170 | /* | |
1171 | * Make sure the child dentry is still negative ... | |
1172 | */ | |
1173 | err = nfserr_exist; | |
1174 | if (dchild->d_inode) { | |
1175 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1176 | dentry->d_name.name, dchild->d_name.name); | |
1177 | goto out; | |
1178 | } | |
1179 | ||
1180 | if (!(iap->ia_valid & ATTR_MODE)) | |
1181 | iap->ia_mode = 0; | |
1182 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1183 | ||
1184 | /* | |
1185 | * Get the dir op function pointer. | |
1186 | */ | |
088406bc | 1187 | err = 0; |
1da177e4 LT |
1188 | switch (type) { |
1189 | case S_IFREG: | |
6264d69d | 1190 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1da177e4 LT |
1191 | break; |
1192 | case S_IFDIR: | |
6264d69d | 1193 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1194 | break; |
1195 | case S_IFCHR: | |
1196 | case S_IFBLK: | |
1197 | case S_IFIFO: | |
1198 | case S_IFSOCK: | |
6264d69d | 1199 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 LT |
1200 | break; |
1201 | default: | |
1202 | printk("nfsd: bad file type %o in nfsd_create\n", type); | |
6264d69d | 1203 | host_err = -EINVAL; |
1da177e4 | 1204 | } |
6264d69d | 1205 | if (host_err < 0) |
1da177e4 LT |
1206 | goto out_nfserr; |
1207 | ||
1208 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1209 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1210 | write_inode_now(dchild->d_inode, 1); |
1211 | } | |
1212 | ||
1213 | ||
1214 | /* Set file attributes. Mode has already been set and | |
1215 | * setting uid/gid works only for root. Irix appears to | |
1216 | * send along the gid when it tries to implement setgid | |
1217 | * directories via NFS. | |
1218 | */ | |
f193fbab | 1219 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1220 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab YT |
1221 | if (err2) |
1222 | err = err2; | |
1223 | } | |
1da177e4 LT |
1224 | /* |
1225 | * Update the file handle to get the new inode info. | |
1226 | */ | |
1227 | if (!err) | |
1228 | err = fh_update(resfhp); | |
1229 | out: | |
1230 | if (dchild && !IS_ERR(dchild)) | |
1231 | dput(dchild); | |
1232 | return err; | |
1233 | ||
1234 | out_nfserr: | |
6264d69d | 1235 | err = nfserrno(host_err); |
1da177e4 LT |
1236 | goto out; |
1237 | } | |
1238 | ||
1239 | #ifdef CONFIG_NFSD_V3 | |
1240 | /* | |
1241 | * NFSv3 version of nfsd_create | |
1242 | */ | |
6264d69d | 1243 | __be32 |
1da177e4 LT |
1244 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1245 | char *fname, int flen, struct iattr *iap, | |
1246 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1247 | int *truncp, int *created) |
1da177e4 LT |
1248 | { |
1249 | struct dentry *dentry, *dchild = NULL; | |
1250 | struct inode *dirp; | |
6264d69d AV |
1251 | __be32 err; |
1252 | int host_err; | |
1da177e4 | 1253 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1254 | |
1255 | err = nfserr_perm; | |
1256 | if (!flen) | |
1257 | goto out; | |
1258 | err = nfserr_exist; | |
1259 | if (isdotent(fname, flen)) | |
1260 | goto out; | |
1261 | if (!(iap->ia_valid & ATTR_MODE)) | |
1262 | iap->ia_mode = 0; | |
1263 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1264 | if (err) | |
1265 | goto out; | |
1266 | ||
1267 | dentry = fhp->fh_dentry; | |
1268 | dirp = dentry->d_inode; | |
1269 | ||
1270 | /* Get all the sanity checks out of the way before | |
1271 | * we lock the parent. */ | |
1272 | err = nfserr_notdir; | |
1273 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1274 | goto out; | |
12fd3520 | 1275 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1276 | |
1277 | /* | |
1278 | * Compose the response file handle. | |
1279 | */ | |
1280 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1281 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1282 | if (IS_ERR(dchild)) |
1283 | goto out_nfserr; | |
1284 | ||
1285 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1286 | if (err) | |
1287 | goto out; | |
1288 | ||
1289 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c PS |
1290 | /* solaris7 gets confused (bugid 4218508) if these have |
1291 | * the high bit set, so just clear the high bits. | |
1da177e4 LT |
1292 | */ |
1293 | v_mtime = verifier[0]&0x7fffffff; | |
1294 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1295 | } |
1296 | ||
1297 | if (dchild->d_inode) { | |
1298 | err = 0; | |
1299 | ||
1300 | switch (createmode) { | |
1301 | case NFS3_CREATE_UNCHECKED: | |
1302 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1303 | err = nfserr_exist; | |
1304 | else if (truncp) { | |
1305 | /* in nfsv4, we need to treat this case a little | |
1306 | * differently. we don't want to truncate the | |
1307 | * file now; this would be wrong if the OPEN | |
1308 | * fails for some other reason. furthermore, | |
1309 | * if the size is nonzero, we should ignore it | |
1310 | * according to spec! | |
1311 | */ | |
1312 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1313 | } | |
1314 | else { | |
1315 | iap->ia_valid &= ATTR_SIZE; | |
1316 | goto set_attr; | |
1317 | } | |
1318 | break; | |
1319 | case NFS3_CREATE_EXCLUSIVE: | |
1320 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1321 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1322 | && dchild->d_inode->i_size == 0 ) |
1323 | break; | |
1324 | /* fallthru */ | |
1325 | case NFS3_CREATE_GUARDED: | |
1326 | err = nfserr_exist; | |
1327 | } | |
1328 | goto out; | |
1329 | } | |
1330 | ||
6264d69d AV |
1331 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1332 | if (host_err < 0) | |
1da177e4 | 1333 | goto out_nfserr; |
81ac95c5 BF |
1334 | if (created) |
1335 | *created = 1; | |
1da177e4 LT |
1336 | |
1337 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1338 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1339 | /* setattr will sync the child (or not) */ |
1340 | } | |
1341 | ||
1da177e4 | 1342 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1343 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1344 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1345 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1346 | /* XXX someone who knows this better please fix it for nsec */ |
1347 | iap->ia_mtime.tv_sec = v_mtime; | |
1348 | iap->ia_atime.tv_sec = v_atime; | |
1349 | iap->ia_mtime.tv_nsec = 0; | |
1350 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1351 | } |
1352 | ||
1353 | /* Set file attributes. | |
1da177e4 LT |
1354 | * Irix appears to send along the gid when it tries to |
1355 | * implement setgid directories via NFS. Clear out all that cruft. | |
1356 | */ | |
1357 | set_attr: | |
c397852c | 1358 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1359 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab | 1360 | if (err2) |
45bd3b3d | 1361 | err = err2; |
f193fbab YT |
1362 | } |
1363 | ||
1364 | /* | |
1365 | * Update the filehandle to get the new inode info. | |
1366 | */ | |
1367 | if (!err) | |
1368 | err = fh_update(resfhp); | |
1da177e4 LT |
1369 | |
1370 | out: | |
1371 | fh_unlock(fhp); | |
1372 | if (dchild && !IS_ERR(dchild)) | |
1373 | dput(dchild); | |
1374 | return err; | |
1375 | ||
1376 | out_nfserr: | |
6264d69d | 1377 | err = nfserrno(host_err); |
1da177e4 LT |
1378 | goto out; |
1379 | } | |
1380 | #endif /* CONFIG_NFSD_V3 */ | |
1381 | ||
1382 | /* | |
1383 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1384 | * fits into the buffer. On return, it contains the true length. | |
1385 | * N.B. After this call fhp needs an fh_put | |
1386 | */ | |
6264d69d | 1387 | __be32 |
1da177e4 LT |
1388 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1389 | { | |
1390 | struct dentry *dentry; | |
1391 | struct inode *inode; | |
1392 | mm_segment_t oldfs; | |
6264d69d AV |
1393 | __be32 err; |
1394 | int host_err; | |
1da177e4 LT |
1395 | |
1396 | err = fh_verify(rqstp, fhp, S_IFLNK, MAY_NOP); | |
1397 | if (err) | |
1398 | goto out; | |
1399 | ||
1400 | dentry = fhp->fh_dentry; | |
1401 | inode = dentry->d_inode; | |
1402 | ||
1403 | err = nfserr_inval; | |
1404 | if (!inode->i_op || !inode->i_op->readlink) | |
1405 | goto out; | |
1406 | ||
1407 | touch_atime(fhp->fh_export->ex_mnt, dentry); | |
1408 | /* N.B. Why does this call need a get_fs()?? | |
1409 | * Remove the set_fs and watch the fireworks:-) --okir | |
1410 | */ | |
1411 | ||
1412 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1413 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1414 | set_fs(oldfs); |
1415 | ||
6264d69d | 1416 | if (host_err < 0) |
1da177e4 | 1417 | goto out_nfserr; |
6264d69d | 1418 | *lenp = host_err; |
1da177e4 LT |
1419 | err = 0; |
1420 | out: | |
1421 | return err; | |
1422 | ||
1423 | out_nfserr: | |
6264d69d | 1424 | err = nfserrno(host_err); |
1da177e4 LT |
1425 | goto out; |
1426 | } | |
1427 | ||
1428 | /* | |
1429 | * Create a symlink and look up its inode | |
1430 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1431 | */ | |
6264d69d | 1432 | __be32 |
1da177e4 LT |
1433 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1434 | char *fname, int flen, | |
1435 | char *path, int plen, | |
1436 | struct svc_fh *resfhp, | |
1437 | struct iattr *iap) | |
1438 | { | |
1439 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1440 | __be32 err, cerr; |
1441 | int host_err; | |
1da177e4 LT |
1442 | umode_t mode; |
1443 | ||
1444 | err = nfserr_noent; | |
1445 | if (!flen || !plen) | |
1446 | goto out; | |
1447 | err = nfserr_exist; | |
1448 | if (isdotent(fname, flen)) | |
1449 | goto out; | |
1450 | ||
1451 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1452 | if (err) | |
1453 | goto out; | |
1454 | fh_lock(fhp); | |
1455 | dentry = fhp->fh_dentry; | |
1456 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1457 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1458 | if (IS_ERR(dnew)) |
1459 | goto out_nfserr; | |
1460 | ||
1461 | mode = S_IALLUGO; | |
1462 | /* Only the MODE ATTRibute is even vaguely meaningful */ | |
1463 | if (iap && (iap->ia_valid & ATTR_MODE)) | |
1464 | mode = iap->ia_mode & S_IALLUGO; | |
1465 | ||
1466 | if (unlikely(path[plen] != 0)) { | |
1467 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1468 | if (path_alloced == NULL) | |
6264d69d | 1469 | host_err = -ENOMEM; |
1da177e4 LT |
1470 | else { |
1471 | strncpy(path_alloced, path, plen); | |
1472 | path_alloced[plen] = 0; | |
6264d69d | 1473 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode); |
1da177e4 LT |
1474 | kfree(path_alloced); |
1475 | } | |
1476 | } else | |
6264d69d | 1477 | host_err = vfs_symlink(dentry->d_inode, dnew, path, mode); |
1da177e4 | 1478 | |
6264d69d | 1479 | if (!host_err) { |
1da177e4 | 1480 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1481 | host_err = nfsd_sync_dir(dentry); |
1482 | } | |
1483 | err = nfserrno(host_err); | |
1da177e4 LT |
1484 | fh_unlock(fhp); |
1485 | ||
1486 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); | |
1487 | dput(dnew); | |
1488 | if (err==0) err = cerr; | |
1489 | out: | |
1490 | return err; | |
1491 | ||
1492 | out_nfserr: | |
6264d69d | 1493 | err = nfserrno(host_err); |
1da177e4 LT |
1494 | goto out; |
1495 | } | |
1496 | ||
1497 | /* | |
1498 | * Create a hardlink | |
1499 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1500 | */ | |
6264d69d | 1501 | __be32 |
1da177e4 LT |
1502 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1503 | char *name, int len, struct svc_fh *tfhp) | |
1504 | { | |
1505 | struct dentry *ddir, *dnew, *dold; | |
1506 | struct inode *dirp, *dest; | |
6264d69d AV |
1507 | __be32 err; |
1508 | int host_err; | |
1da177e4 LT |
1509 | |
1510 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_CREATE); | |
1511 | if (err) | |
1512 | goto out; | |
1513 | err = fh_verify(rqstp, tfhp, -S_IFDIR, MAY_NOP); | |
1514 | if (err) | |
1515 | goto out; | |
1516 | ||
1517 | err = nfserr_perm; | |
1518 | if (!len) | |
1519 | goto out; | |
1520 | err = nfserr_exist; | |
1521 | if (isdotent(name, len)) | |
1522 | goto out; | |
1523 | ||
12fd3520 | 1524 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1525 | ddir = ffhp->fh_dentry; |
1526 | dirp = ddir->d_inode; | |
1527 | ||
1528 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1529 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1530 | if (IS_ERR(dnew)) |
1531 | goto out_nfserr; | |
1532 | ||
1533 | dold = tfhp->fh_dentry; | |
1534 | dest = dold->d_inode; | |
1535 | ||
6264d69d AV |
1536 | host_err = vfs_link(dold, dirp, dnew); |
1537 | if (!host_err) { | |
1da177e4 | 1538 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1539 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1540 | write_inode_now(dest, 1); |
1541 | } | |
6264d69d | 1542 | err = 0; |
1da177e4 | 1543 | } else { |
6264d69d | 1544 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1545 | err = nfserr_acces; |
1546 | else | |
6264d69d | 1547 | err = nfserrno(host_err); |
1da177e4 LT |
1548 | } |
1549 | ||
1da177e4 | 1550 | dput(dnew); |
270d56e5 DR |
1551 | out_unlock: |
1552 | fh_unlock(ffhp); | |
1da177e4 LT |
1553 | out: |
1554 | return err; | |
1555 | ||
1556 | out_nfserr: | |
6264d69d | 1557 | err = nfserrno(host_err); |
270d56e5 | 1558 | goto out_unlock; |
1da177e4 LT |
1559 | } |
1560 | ||
1561 | /* | |
1562 | * Rename a file | |
1563 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1564 | */ | |
6264d69d | 1565 | __be32 |
1da177e4 LT |
1566 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1567 | struct svc_fh *tfhp, char *tname, int tlen) | |
1568 | { | |
1569 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1570 | struct inode *fdir, *tdir; | |
6264d69d AV |
1571 | __be32 err; |
1572 | int host_err; | |
1da177e4 LT |
1573 | |
1574 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_REMOVE); | |
1575 | if (err) | |
1576 | goto out; | |
1577 | err = fh_verify(rqstp, tfhp, S_IFDIR, MAY_CREATE); | |
1578 | if (err) | |
1579 | goto out; | |
1580 | ||
1581 | fdentry = ffhp->fh_dentry; | |
1582 | fdir = fdentry->d_inode; | |
1583 | ||
1584 | tdentry = tfhp->fh_dentry; | |
1585 | tdir = tdentry->d_inode; | |
1586 | ||
1587 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1588 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1589 | goto out; |
1590 | ||
1591 | err = nfserr_perm; | |
1592 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1593 | goto out; | |
1594 | ||
1595 | /* cannot use fh_lock as we need deadlock protective ordering | |
1596 | * so do it by hand */ | |
1597 | trap = lock_rename(tdentry, fdentry); | |
1598 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1599 | fill_pre_wcc(ffhp); | |
1600 | fill_pre_wcc(tfhp); | |
1601 | ||
1602 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1603 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1604 | if (IS_ERR(odentry)) |
1605 | goto out_nfserr; | |
1606 | ||
6264d69d | 1607 | host_err = -ENOENT; |
1da177e4 LT |
1608 | if (!odentry->d_inode) |
1609 | goto out_dput_old; | |
6264d69d | 1610 | host_err = -EINVAL; |
1da177e4 LT |
1611 | if (odentry == trap) |
1612 | goto out_dput_old; | |
1613 | ||
1614 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1615 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1616 | if (IS_ERR(ndentry)) |
1617 | goto out_dput_old; | |
6264d69d | 1618 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1619 | if (ndentry == trap) |
1620 | goto out_dput_new; | |
1621 | ||
1622 | #ifdef MSNFS | |
1623 | if ((ffhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1624 | ((atomic_read(&odentry->d_count) > 1) | |
1625 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1626 | host_err = -EPERM; |
1da177e4 LT |
1627 | } else |
1628 | #endif | |
6264d69d AV |
1629 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1630 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1631 | host_err = nfsd_sync_dir(tdentry); | |
1632 | if (!host_err) | |
1633 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1634 | } |
1635 | ||
1636 | out_dput_new: | |
1637 | dput(ndentry); | |
1638 | out_dput_old: | |
1639 | dput(odentry); | |
1640 | out_nfserr: | |
6264d69d | 1641 | err = nfserrno(host_err); |
1da177e4 LT |
1642 | |
1643 | /* we cannot reply on fh_unlock on the two filehandles, | |
1644 | * as that would do the wrong thing if the two directories | |
1645 | * were the same, so again we do it by hand | |
1646 | */ | |
1647 | fill_post_wcc(ffhp); | |
1648 | fill_post_wcc(tfhp); | |
1649 | unlock_rename(tdentry, fdentry); | |
1650 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1651 | ||
1652 | out: | |
1653 | return err; | |
1654 | } | |
1655 | ||
1656 | /* | |
1657 | * Unlink a file or directory | |
1658 | * N.B. After this call fhp needs an fh_put | |
1659 | */ | |
6264d69d | 1660 | __be32 |
1da177e4 LT |
1661 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1662 | char *fname, int flen) | |
1663 | { | |
1664 | struct dentry *dentry, *rdentry; | |
1665 | struct inode *dirp; | |
6264d69d AV |
1666 | __be32 err; |
1667 | int host_err; | |
1da177e4 LT |
1668 | |
1669 | err = nfserr_acces; | |
1670 | if (!flen || isdotent(fname, flen)) | |
1671 | goto out; | |
1672 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_REMOVE); | |
1673 | if (err) | |
1674 | goto out; | |
1675 | ||
12fd3520 | 1676 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1677 | dentry = fhp->fh_dentry; |
1678 | dirp = dentry->d_inode; | |
1679 | ||
1680 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1681 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1682 | if (IS_ERR(rdentry)) |
1683 | goto out_nfserr; | |
1684 | ||
1685 | if (!rdentry->d_inode) { | |
1686 | dput(rdentry); | |
1687 | err = nfserr_noent; | |
1688 | goto out; | |
1689 | } | |
1690 | ||
1691 | if (!type) | |
1692 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1693 | ||
1694 | if (type != S_IFDIR) { /* It's UNLINK */ | |
1695 | #ifdef MSNFS | |
1696 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1697 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1698 | host_err = -EPERM; |
1da177e4 LT |
1699 | } else |
1700 | #endif | |
6264d69d | 1701 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1702 | } else { /* It's RMDIR */ |
6264d69d | 1703 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1704 | } |
1705 | ||
1706 | dput(rdentry); | |
1707 | ||
6264d69d AV |
1708 | if (host_err) |
1709 | goto out_nfserr; | |
1710 | if (EX_ISSYNC(fhp->fh_export)) | |
1711 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 LT |
1712 | |
1713 | out_nfserr: | |
6264d69d | 1714 | err = nfserrno(host_err); |
f193fbab YT |
1715 | out: |
1716 | return err; | |
1da177e4 LT |
1717 | } |
1718 | ||
1719 | /* | |
1720 | * Read entries from a directory. | |
1721 | * The NFSv3/4 verifier we ignore for now. | |
1722 | */ | |
6264d69d | 1723 | __be32 |
1da177e4 | 1724 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1725 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1726 | { |
6264d69d AV |
1727 | __be32 err; |
1728 | int host_err; | |
1da177e4 LT |
1729 | struct file *file; |
1730 | loff_t offset = *offsetp; | |
1731 | ||
1732 | err = nfsd_open(rqstp, fhp, S_IFDIR, MAY_READ, &file); | |
1733 | if (err) | |
1734 | goto out; | |
1735 | ||
1736 | offset = vfs_llseek(file, offset, 0); | |
1737 | if (offset < 0) { | |
1738 | err = nfserrno((int)offset); | |
1739 | goto out_close; | |
1740 | } | |
1741 | ||
1742 | /* | |
1743 | * Read the directory entries. This silly loop is necessary because | |
1744 | * readdir() is not guaranteed to fill up the entire buffer, but | |
1745 | * may choose to do less. | |
1746 | */ | |
1747 | ||
1748 | do { | |
1749 | cdp->err = nfserr_eof; /* will be cleared on successful read */ | |
a0ad13ef | 1750 | host_err = vfs_readdir(file, func, cdp); |
6264d69d AV |
1751 | } while (host_err >=0 && cdp->err == nfs_ok); |
1752 | if (host_err) | |
1753 | err = nfserrno(host_err); | |
1da177e4 LT |
1754 | else |
1755 | err = cdp->err; | |
1756 | *offsetp = vfs_llseek(file, 0, 1); | |
1757 | ||
1758 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1759 | err = nfs_ok; /* can still be found in ->err */ | |
1760 | out_close: | |
1761 | nfsd_close(file); | |
1762 | out: | |
1763 | return err; | |
1764 | } | |
1765 | ||
1766 | /* | |
1767 | * Get file system stats | |
1768 | * N.B. After this call fhp needs an fh_put | |
1769 | */ | |
6264d69d | 1770 | __be32 |
1da177e4 LT |
1771 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat) |
1772 | { | |
6264d69d | 1773 | __be32 err = fh_verify(rqstp, fhp, 0, MAY_NOP); |
726c3342 | 1774 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
1775 | err = nfserr_io; |
1776 | return err; | |
1777 | } | |
1778 | ||
1779 | /* | |
1780 | * Check for a user's access permissions to this inode. | |
1781 | */ | |
6264d69d | 1782 | __be32 |
1da177e4 LT |
1783 | nfsd_permission(struct svc_export *exp, struct dentry *dentry, int acc) |
1784 | { | |
1785 | struct inode *inode = dentry->d_inode; | |
1786 | int err; | |
1787 | ||
1788 | if (acc == MAY_NOP) | |
1789 | return 0; | |
1790 | #if 0 | |
1791 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
1792 | acc, | |
1793 | (acc & MAY_READ)? " read" : "", | |
1794 | (acc & MAY_WRITE)? " write" : "", | |
1795 | (acc & MAY_EXEC)? " exec" : "", | |
1796 | (acc & MAY_SATTR)? " sattr" : "", | |
1797 | (acc & MAY_TRUNC)? " trunc" : "", | |
1798 | (acc & MAY_LOCK)? " lock" : "", | |
1799 | (acc & MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1800 | inode->i_mode, | |
1801 | IS_IMMUTABLE(inode)? " immut" : "", | |
1802 | IS_APPEND(inode)? " append" : "", | |
1803 | IS_RDONLY(inode)? " ro" : ""); | |
1804 | dprintk(" owner %d/%d user %d/%d\n", | |
1805 | inode->i_uid, inode->i_gid, current->fsuid, current->fsgid); | |
1806 | #endif | |
1807 | ||
1808 | /* Normally we reject any write/sattr etc access on a read-only file | |
1809 | * system. But if it is IRIX doing check on write-access for a | |
1810 | * device special file, we ignore rofs. | |
1811 | */ | |
1812 | if (!(acc & MAY_LOCAL_ACCESS)) | |
1813 | if (acc & (MAY_WRITE | MAY_SATTR | MAY_TRUNC)) { | |
1814 | if (EX_RDONLY(exp) || IS_RDONLY(inode)) | |
1815 | return nfserr_rofs; | |
1816 | if (/* (acc & MAY_WRITE) && */ IS_IMMUTABLE(inode)) | |
1817 | return nfserr_perm; | |
1818 | } | |
1819 | if ((acc & MAY_TRUNC) && IS_APPEND(inode)) | |
1820 | return nfserr_perm; | |
1821 | ||
1822 | if (acc & MAY_LOCK) { | |
1823 | /* If we cannot rely on authentication in NLM requests, | |
1824 | * just allow locks, otherwise require read permission, or | |
1825 | * ownership | |
1826 | */ | |
1827 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
1828 | return 0; | |
1829 | else | |
1830 | acc = MAY_READ | MAY_OWNER_OVERRIDE; | |
1831 | } | |
1832 | /* | |
1833 | * The file owner always gets access permission for accesses that | |
1834 | * would normally be checked at open time. This is to make | |
1835 | * file access work even when the client has done a fchmod(fd, 0). | |
1836 | * | |
1837 | * However, `cp foo bar' should fail nevertheless when bar is | |
1838 | * readonly. A sensible way to do this might be to reject all | |
1839 | * attempts to truncate a read-only file, because a creat() call | |
1840 | * always implies file truncation. | |
1841 | * ... but this isn't really fair. A process may reasonably call | |
1842 | * ftruncate on an open file descriptor on a file with perm 000. | |
1843 | * We must trust the client to do permission checking - using "ACCESS" | |
1844 | * with NFSv3. | |
1845 | */ | |
1846 | if ((acc & MAY_OWNER_OVERRIDE) && | |
1847 | inode->i_uid == current->fsuid) | |
1848 | return 0; | |
1849 | ||
1850 | err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL); | |
1851 | ||
1852 | /* Allow read access to binaries even when mode 111 */ | |
1853 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
1854 | acc == (MAY_READ | MAY_OWNER_OVERRIDE)) | |
1855 | err = permission(inode, MAY_EXEC, NULL); | |
1856 | ||
1857 | return err? nfserrno(err) : 0; | |
1858 | } | |
1859 | ||
1860 | void | |
1861 | nfsd_racache_shutdown(void) | |
1862 | { | |
fce1456a | 1863 | if (!raparml) |
1da177e4 LT |
1864 | return; |
1865 | dprintk("nfsd: freeing readahead buffers.\n"); | |
1866 | kfree(raparml); | |
fce1456a | 1867 | raparml = NULL; |
1da177e4 LT |
1868 | } |
1869 | /* | |
1870 | * Initialize readahead param cache | |
1871 | */ | |
1872 | int | |
1873 | nfsd_racache_init(int cache_size) | |
1874 | { | |
1875 | int i; | |
fce1456a GB |
1876 | int j = 0; |
1877 | int nperbucket; | |
1da177e4 | 1878 | |
fce1456a GB |
1879 | |
1880 | if (raparml) | |
1da177e4 | 1881 | return 0; |
fce1456a GB |
1882 | if (cache_size < 2*RAPARM_HASH_SIZE) |
1883 | cache_size = 2*RAPARM_HASH_SIZE; | |
4b3bb06b YB |
1884 | raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL); |
1885 | ||
1886 | if (!raparml) { | |
1da177e4 | 1887 | printk(KERN_WARNING |
4b3bb06b | 1888 | "nfsd: Could not allocate memory read-ahead cache.\n"); |
1da177e4 LT |
1889 | return -ENOMEM; |
1890 | } | |
4b3bb06b YB |
1891 | |
1892 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
1893 | for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) { | |
1894 | raparm_hash[i].pb_head = NULL; | |
1895 | spin_lock_init(&raparm_hash[i].pb_lock); | |
1896 | } | |
1897 | nperbucket = cache_size >> RAPARM_HASH_BITS; | |
1898 | for (i = 0; i < cache_size - 1; i++) { | |
1899 | if (i % nperbucket == 0) | |
1900 | raparm_hash[j++].pb_head = raparml + i; | |
1901 | if (i % nperbucket < nperbucket-1) | |
1902 | raparml[i].p_next = raparml + i + 1; | |
1903 | } | |
1904 | ||
1da177e4 LT |
1905 | nfsdstats.ra_size = cache_size; |
1906 | return 0; | |
1907 | } | |
a257cdd0 AG |
1908 | |
1909 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
1910 | struct posix_acl * | |
1911 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
1912 | { | |
1913 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1914 | char *name; | |
1915 | void *value = NULL; | |
1916 | ssize_t size; | |
1917 | struct posix_acl *acl; | |
1918 | ||
5be196e5 CH |
1919 | if (!IS_POSIXACL(inode)) |
1920 | return ERR_PTR(-EOPNOTSUPP); | |
1921 | ||
1922 | switch (type) { | |
1923 | case ACL_TYPE_ACCESS: | |
1924 | name = POSIX_ACL_XATTR_ACCESS; | |
1925 | break; | |
1926 | case ACL_TYPE_DEFAULT: | |
1927 | name = POSIX_ACL_XATTR_DEFAULT; | |
1928 | break; | |
1929 | default: | |
a257cdd0 | 1930 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
1931 | } |
1932 | ||
5be196e5 CH |
1933 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
1934 | if (size < 0) | |
1935 | return ERR_PTR(size); | |
a257cdd0 | 1936 | |
a257cdd0 | 1937 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
1938 | kfree(value); |
1939 | return acl; | |
1940 | } | |
1941 | ||
1942 | int | |
1943 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
1944 | { | |
1945 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1946 | char *name; | |
1947 | void *value = NULL; | |
1948 | size_t size; | |
1949 | int error; | |
1950 | ||
1951 | if (!IS_POSIXACL(inode) || !inode->i_op || | |
1952 | !inode->i_op->setxattr || !inode->i_op->removexattr) | |
1953 | return -EOPNOTSUPP; | |
1954 | switch(type) { | |
1955 | case ACL_TYPE_ACCESS: | |
334a13ec | 1956 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
1957 | break; |
1958 | case ACL_TYPE_DEFAULT: | |
334a13ec | 1959 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
1960 | break; |
1961 | default: | |
1962 | return -EOPNOTSUPP; | |
1963 | } | |
1964 | ||
1965 | if (acl && acl->a_count) { | |
334a13ec | 1966 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
1967 | value = kmalloc(size, GFP_KERNEL); |
1968 | if (!value) | |
1969 | return -ENOMEM; | |
9ccfc29c FM |
1970 | error = posix_acl_to_xattr(acl, value, size); |
1971 | if (error < 0) | |
a257cdd0 | 1972 | goto getout; |
9ccfc29c | 1973 | size = error; |
a257cdd0 AG |
1974 | } else |
1975 | size = 0; | |
1976 | ||
a257cdd0 | 1977 | if (size) |
5be196e5 | 1978 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
1979 | else { |
1980 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
1981 | error = 0; | |
1982 | else { | |
5be196e5 | 1983 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
1984 | if (error == -ENODATA) |
1985 | error = 0; | |
1986 | } | |
1987 | } | |
1988 | ||
1989 | getout: | |
1990 | kfree(value); | |
1991 | return error; | |
1992 | } | |
1993 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |