]>
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 | ||
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 | ||
191 | exp2 = exp_parent(exp->ex_client, mnt, dentry, | |
192 | &rqstp->rq_chandle); | |
2d3bb252 BF |
193 | if (PTR_ERR(exp2) == -ENOENT) { |
194 | dput(dentry); | |
195 | dentry = dget(dparent); | |
196 | } else if (IS_ERR(exp2)) { | |
6264d69d | 197 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
198 | dput(dentry); |
199 | mntput(mnt); | |
200 | goto out_nfserr; | |
1da177e4 LT |
201 | } else { |
202 | exp_put(exp); | |
203 | exp = exp2; | |
204 | } | |
205 | mntput(mnt); | |
206 | } | |
207 | } else { | |
208 | fh_lock(fhp); | |
209 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 210 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
211 | if (IS_ERR(dentry)) |
212 | goto out_nfserr; | |
213 | /* | |
214 | * check if we have crossed a mount point ... | |
215 | */ | |
216 | if (d_mountpoint(dentry)) { | |
6264d69d | 217 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
218 | dput(dentry); |
219 | goto out_nfserr; | |
220 | } | |
221 | } | |
222 | } | |
223 | /* | |
224 | * Note: we compose the file handle now, but as the | |
225 | * dentry may be negative, it may need to be updated. | |
226 | */ | |
227 | err = fh_compose(resfh, exp, dentry, fhp); | |
228 | if (!err && !dentry->d_inode) | |
229 | err = nfserr_noent; | |
230 | dput(dentry); | |
231 | out: | |
232 | exp_put(exp); | |
233 | return err; | |
234 | ||
235 | out_nfserr: | |
6264d69d | 236 | err = nfserrno(host_err); |
1da177e4 LT |
237 | goto out; |
238 | } | |
239 | ||
240 | /* | |
241 | * Set various file attributes. | |
242 | * N.B. After this call fhp needs an fh_put | |
243 | */ | |
6264d69d | 244 | __be32 |
1da177e4 LT |
245 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
246 | int check_guard, time_t guardtime) | |
247 | { | |
248 | struct dentry *dentry; | |
249 | struct inode *inode; | |
250 | int accmode = MAY_SATTR; | |
251 | int ftype = 0; | |
252 | int imode; | |
6264d69d AV |
253 | __be32 err; |
254 | int host_err; | |
1da177e4 LT |
255 | int size_change = 0; |
256 | ||
257 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
258 | accmode |= MAY_WRITE|MAY_OWNER_OVERRIDE; | |
259 | if (iap->ia_valid & ATTR_SIZE) | |
260 | ftype = S_IFREG; | |
261 | ||
262 | /* Get inode */ | |
263 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 264 | if (err) |
1da177e4 LT |
265 | goto out; |
266 | ||
267 | dentry = fhp->fh_dentry; | |
268 | inode = dentry->d_inode; | |
269 | ||
15b7a1b8 N |
270 | /* Ignore any mode updates on symlinks */ |
271 | if (S_ISLNK(inode->i_mode)) | |
272 | iap->ia_valid &= ~ATTR_MODE; | |
273 | ||
274 | if (!iap->ia_valid) | |
275 | goto out; | |
276 | ||
1da177e4 LT |
277 | /* NFSv2 does not differentiate between "set-[ac]time-to-now" |
278 | * which only requires access, and "set-[ac]time-to-X" which | |
279 | * requires ownership. | |
280 | * So if it looks like it might be "set both to the same time which | |
281 | * is close to now", and if inode_change_ok fails, then we | |
282 | * convert to "set to now" instead of "set to explicit time" | |
283 | * | |
284 | * We only call inode_change_ok as the last test as technically | |
285 | * it is not an interface that we should be using. It is only | |
286 | * valid if the filesystem does not define it's own i_op->setattr. | |
287 | */ | |
288 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
289 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
290 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET | |
291 | && iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec | |
292 | ) { | |
293 | /* Looks probable. Now just make sure time is in the right ballpark. | |
294 | * Solaris, at least, doesn't seem to care what the time request is. | |
295 | * We require it be within 30 minutes of now. | |
296 | */ | |
297 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); | |
298 | if (delta<0) delta = -delta; | |
299 | if (delta < MAX_TOUCH_TIME_ERROR && | |
300 | inode_change_ok(inode, iap) != 0) { | |
301 | /* turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME | |
302 | * this will cause notify_change to set these times to "now" | |
303 | */ | |
304 | iap->ia_valid &= ~BOTH_TIME_SET; | |
305 | } | |
306 | } | |
307 | ||
308 | /* The size case is special. It changes the file as well as the attributes. */ | |
309 | if (iap->ia_valid & ATTR_SIZE) { | |
310 | if (iap->ia_size < inode->i_size) { | |
311 | err = nfsd_permission(fhp->fh_export, dentry, MAY_TRUNC|MAY_OWNER_OVERRIDE); | |
312 | if (err) | |
313 | goto out; | |
314 | } | |
315 | ||
316 | /* | |
317 | * If we are changing the size of the file, then | |
318 | * we need to break all leases. | |
319 | */ | |
6264d69d AV |
320 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
321 | if (host_err == -EWOULDBLOCK) | |
322 | host_err = -ETIMEDOUT; | |
323 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
324 | goto out_nfserr; |
325 | ||
6264d69d AV |
326 | host_err = get_write_access(inode); |
327 | if (host_err) | |
1da177e4 LT |
328 | goto out_nfserr; |
329 | ||
330 | size_change = 1; | |
6264d69d AV |
331 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
332 | if (host_err) { | |
1da177e4 LT |
333 | put_write_access(inode); |
334 | goto out_nfserr; | |
335 | } | |
336 | DQUOT_INIT(inode); | |
337 | } | |
338 | ||
339 | imode = inode->i_mode; | |
340 | if (iap->ia_valid & ATTR_MODE) { | |
341 | iap->ia_mode &= S_IALLUGO; | |
342 | imode = iap->ia_mode |= (imode & ~S_IALLUGO); | |
343 | } | |
344 | ||
345 | /* Revoke setuid/setgid bit on chown/chgrp */ | |
346 | if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) | |
347 | iap->ia_valid |= ATTR_KILL_SUID; | |
348 | if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid) | |
349 | iap->ia_valid |= ATTR_KILL_SGID; | |
350 | ||
351 | /* Change the attributes. */ | |
352 | ||
353 | iap->ia_valid |= ATTR_CTIME; | |
354 | ||
355 | err = nfserr_notsync; | |
356 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
357 | fh_lock(fhp); | |
6264d69d AV |
358 | host_err = notify_change(dentry, iap); |
359 | err = nfserrno(host_err); | |
1da177e4 LT |
360 | fh_unlock(fhp); |
361 | } | |
362 | if (size_change) | |
363 | put_write_access(inode); | |
364 | if (!err) | |
365 | if (EX_ISSYNC(fhp->fh_export)) | |
366 | write_inode_now(inode, 1); | |
367 | out: | |
368 | return err; | |
369 | ||
370 | out_nfserr: | |
6264d69d | 371 | err = nfserrno(host_err); |
1da177e4 LT |
372 | goto out; |
373 | } | |
374 | ||
5be196e5 CH |
375 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
376 | defined(CONFIG_NFSD_V3_ACL) || \ | |
377 | defined(CONFIG_NFSD_V4) | |
378 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
379 | { | |
380 | ssize_t buflen; | |
5be196e5 CH |
381 | |
382 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
383 | if (buflen <= 0) | |
384 | return buflen; | |
1da177e4 | 385 | |
5be196e5 CH |
386 | *buf = kmalloc(buflen, GFP_KERNEL); |
387 | if (!*buf) | |
388 | return -ENOMEM; | |
389 | ||
b5872b0d | 390 | return vfs_getxattr(dentry, key, *buf, buflen); |
5be196e5 CH |
391 | } |
392 | #endif | |
393 | ||
394 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
395 | static int |
396 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
397 | { | |
398 | int len; | |
399 | size_t buflen; | |
400 | char *buf = NULL; | |
401 | int error = 0; | |
1da177e4 LT |
402 | |
403 | buflen = posix_acl_xattr_size(pacl->a_count); | |
404 | buf = kmalloc(buflen, GFP_KERNEL); | |
405 | error = -ENOMEM; | |
406 | if (buf == NULL) | |
407 | goto out; | |
408 | ||
409 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
410 | if (len < 0) { | |
411 | error = len; | |
412 | goto out; | |
413 | } | |
414 | ||
5be196e5 | 415 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
416 | out: |
417 | kfree(buf); | |
418 | return error; | |
419 | } | |
420 | ||
6264d69d | 421 | __be32 |
1da177e4 LT |
422 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
423 | struct nfs4_acl *acl) | |
424 | { | |
6264d69d AV |
425 | __be32 error; |
426 | int host_error; | |
1da177e4 LT |
427 | struct dentry *dentry; |
428 | struct inode *inode; | |
429 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
430 | unsigned int flags = 0; | |
431 | ||
432 | /* Get inode */ | |
433 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, MAY_SATTR); | |
434 | if (error) | |
4b2ca38a | 435 | return error; |
1da177e4 LT |
436 | |
437 | dentry = fhp->fh_dentry; | |
438 | inode = dentry->d_inode; | |
439 | if (S_ISDIR(inode->i_mode)) | |
440 | flags = NFS4_ACL_DIR; | |
441 | ||
6264d69d AV |
442 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
443 | if (host_error == -EINVAL) { | |
4b2ca38a | 444 | return nfserr_attrnotsupp; |
6264d69d | 445 | } else if (host_error < 0) |
1da177e4 LT |
446 | goto out_nfserr; |
447 | ||
6264d69d AV |
448 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
449 | if (host_error < 0) | |
4b2ca38a | 450 | goto out_release; |
1da177e4 | 451 | |
4b2ca38a | 452 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 453 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 454 | |
4b2ca38a | 455 | out_release: |
1da177e4 LT |
456 | posix_acl_release(pacl); |
457 | posix_acl_release(dpacl); | |
1da177e4 | 458 | out_nfserr: |
f34f9242 | 459 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 460 | return nfserr_attrnotsupp; |
f34f9242 | 461 | else |
4b2ca38a | 462 | return nfserrno(host_error); |
1da177e4 LT |
463 | } |
464 | ||
465 | static struct posix_acl * | |
466 | _get_posix_acl(struct dentry *dentry, char *key) | |
467 | { | |
5be196e5 | 468 | void *buf = NULL; |
1da177e4 | 469 | struct posix_acl *pacl = NULL; |
5be196e5 | 470 | int buflen; |
1da177e4 | 471 | |
5be196e5 CH |
472 | buflen = nfsd_getxattr(dentry, key, &buf); |
473 | if (!buflen) | |
474 | buflen = -ENODATA; | |
475 | if (buflen <= 0) | |
476 | return ERR_PTR(buflen); | |
1da177e4 LT |
477 | |
478 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
479 | kfree(buf); |
480 | return pacl; | |
1da177e4 LT |
481 | } |
482 | ||
483 | int | |
484 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
485 | { | |
486 | struct inode *inode = dentry->d_inode; | |
487 | int error = 0; | |
488 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
489 | unsigned int flags = 0; | |
490 | ||
9a59f452 | 491 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
492 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
493 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
494 | if (IS_ERR(pacl)) { | |
495 | error = PTR_ERR(pacl); | |
496 | pacl = NULL; | |
497 | goto out; | |
498 | } | |
499 | ||
500 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 501 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
502 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
503 | dpacl = NULL; | |
504 | else if (IS_ERR(dpacl)) { | |
505 | error = PTR_ERR(dpacl); | |
506 | dpacl = NULL; | |
507 | goto out; | |
508 | } | |
509 | flags = NFS4_ACL_DIR; | |
510 | } | |
511 | ||
512 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
513 | if (IS_ERR(*acl)) { | |
514 | error = PTR_ERR(*acl); | |
515 | *acl = NULL; | |
516 | } | |
517 | out: | |
518 | posix_acl_release(pacl); | |
519 | posix_acl_release(dpacl); | |
520 | return error; | |
521 | } | |
522 | ||
523 | #endif /* defined(CONFIG_NFS_V4) */ | |
524 | ||
525 | #ifdef CONFIG_NFSD_V3 | |
526 | /* | |
527 | * Check server access rights to a file system object | |
528 | */ | |
529 | struct accessmap { | |
530 | u32 access; | |
531 | int how; | |
532 | }; | |
533 | static struct accessmap nfs3_regaccess[] = { | |
534 | { NFS3_ACCESS_READ, MAY_READ }, | |
535 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
536 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_TRUNC }, | |
537 | { NFS3_ACCESS_EXTEND, MAY_WRITE }, | |
538 | ||
539 | { 0, 0 } | |
540 | }; | |
541 | ||
542 | static struct accessmap nfs3_diraccess[] = { | |
543 | { NFS3_ACCESS_READ, MAY_READ }, | |
544 | { NFS3_ACCESS_LOOKUP, MAY_EXEC }, | |
545 | { NFS3_ACCESS_MODIFY, MAY_EXEC|MAY_WRITE|MAY_TRUNC }, | |
546 | { NFS3_ACCESS_EXTEND, MAY_EXEC|MAY_WRITE }, | |
547 | { NFS3_ACCESS_DELETE, MAY_REMOVE }, | |
548 | ||
549 | { 0, 0 } | |
550 | }; | |
551 | ||
552 | static struct accessmap nfs3_anyaccess[] = { | |
553 | /* Some clients - Solaris 2.6 at least, make an access call | |
554 | * to the server to check for access for things like /dev/null | |
555 | * (which really, the server doesn't care about). So | |
556 | * We provide simple access checking for them, looking | |
557 | * mainly at mode bits, and we make sure to ignore read-only | |
558 | * filesystem checks | |
559 | */ | |
560 | { NFS3_ACCESS_READ, MAY_READ }, | |
561 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
562 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
563 | { NFS3_ACCESS_EXTEND, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
564 | ||
565 | { 0, 0 } | |
566 | }; | |
567 | ||
6264d69d | 568 | __be32 |
1da177e4 LT |
569 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
570 | { | |
571 | struct accessmap *map; | |
572 | struct svc_export *export; | |
573 | struct dentry *dentry; | |
574 | u32 query, result = 0, sresult = 0; | |
6264d69d | 575 | __be32 error; |
1da177e4 LT |
576 | |
577 | error = fh_verify(rqstp, fhp, 0, MAY_NOP); | |
578 | if (error) | |
579 | goto out; | |
580 | ||
581 | export = fhp->fh_export; | |
582 | dentry = fhp->fh_dentry; | |
583 | ||
584 | if (S_ISREG(dentry->d_inode->i_mode)) | |
585 | map = nfs3_regaccess; | |
586 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
587 | map = nfs3_diraccess; | |
588 | else | |
589 | map = nfs3_anyaccess; | |
590 | ||
591 | ||
592 | query = *access; | |
593 | for (; map->access; map++) { | |
594 | if (map->access & query) { | |
6264d69d | 595 | __be32 err2; |
1da177e4 LT |
596 | |
597 | sresult |= map->access; | |
598 | ||
599 | err2 = nfsd_permission(export, dentry, map->how); | |
600 | switch (err2) { | |
601 | case nfs_ok: | |
602 | result |= map->access; | |
603 | break; | |
604 | ||
605 | /* the following error codes just mean the access was not allowed, | |
606 | * rather than an error occurred */ | |
607 | case nfserr_rofs: | |
608 | case nfserr_acces: | |
609 | case nfserr_perm: | |
610 | /* simply don't "or" in the access bit. */ | |
611 | break; | |
612 | default: | |
613 | error = err2; | |
614 | goto out; | |
615 | } | |
616 | } | |
617 | } | |
618 | *access = result; | |
619 | if (supported) | |
620 | *supported = sresult; | |
621 | ||
622 | out: | |
623 | return error; | |
624 | } | |
625 | #endif /* CONFIG_NFSD_V3 */ | |
626 | ||
627 | ||
628 | ||
629 | /* | |
630 | * Open an existing file or directory. | |
631 | * The access argument indicates the type of open (read/write/lock) | |
632 | * N.B. After this call fhp needs an fh_put | |
633 | */ | |
6264d69d | 634 | __be32 |
1da177e4 LT |
635 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
636 | int access, struct file **filp) | |
637 | { | |
638 | struct dentry *dentry; | |
639 | struct inode *inode; | |
6264d69d AV |
640 | int flags = O_RDONLY|O_LARGEFILE; |
641 | __be32 err; | |
642 | int host_err; | |
1da177e4 LT |
643 | |
644 | /* | |
645 | * If we get here, then the client has already done an "open", | |
646 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
647 | * in case a chmod has now revoked permission. | |
648 | */ | |
649 | err = fh_verify(rqstp, fhp, type, access | MAY_OWNER_OVERRIDE); | |
650 | if (err) | |
651 | goto out; | |
652 | ||
653 | dentry = fhp->fh_dentry; | |
654 | inode = dentry->d_inode; | |
655 | ||
656 | /* Disallow write access to files with the append-only bit set | |
657 | * or any access when mandatory locking enabled | |
658 | */ | |
659 | err = nfserr_perm; | |
660 | if (IS_APPEND(inode) && (access & MAY_WRITE)) | |
661 | goto out; | |
662 | if (IS_ISMNDLK(inode)) | |
663 | goto out; | |
664 | ||
665 | if (!inode->i_fop) | |
666 | goto out; | |
667 | ||
668 | /* | |
669 | * Check to see if there are any leases on this file. | |
670 | * This may block while leases are broken. | |
671 | */ | |
6264d69d AV |
672 | host_err = break_lease(inode, O_NONBLOCK | ((access & MAY_WRITE) ? FMODE_WRITE : 0)); |
673 | if (host_err == -EWOULDBLOCK) | |
674 | host_err = -ETIMEDOUT; | |
675 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
676 | goto out_nfserr; |
677 | ||
678 | if (access & MAY_WRITE) { | |
9ecb6a08 BF |
679 | if (access & MAY_READ) |
680 | flags = O_RDWR|O_LARGEFILE; | |
681 | else | |
682 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 LT |
683 | |
684 | DQUOT_INIT(inode); | |
685 | } | |
686 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_mnt), flags); | |
687 | if (IS_ERR(*filp)) | |
6264d69d | 688 | host_err = PTR_ERR(*filp); |
1da177e4 | 689 | out_nfserr: |
6264d69d | 690 | err = nfserrno(host_err); |
1da177e4 LT |
691 | out: |
692 | return err; | |
693 | } | |
694 | ||
695 | /* | |
696 | * Close a file. | |
697 | */ | |
698 | void | |
699 | nfsd_close(struct file *filp) | |
700 | { | |
701 | fput(filp); | |
702 | } | |
703 | ||
704 | /* | |
705 | * Sync a file | |
706 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
707 | * after it. | |
708 | */ | |
a334de28 | 709 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, |
99ac48f5 | 710 | const struct file_operations *fop) |
1da177e4 LT |
711 | { |
712 | struct inode *inode = dp->d_inode; | |
713 | int (*fsync) (struct file *, struct dentry *, int); | |
f193fbab | 714 | int err; |
1da177e4 | 715 | |
f193fbab YT |
716 | err = filemap_fdatawrite(inode->i_mapping); |
717 | if (err == 0 && fop && (fsync = fop->fsync)) | |
718 | err = fsync(filp, dp, 0); | |
719 | if (err == 0) | |
720 | err = filemap_fdatawait(inode->i_mapping); | |
a334de28 | 721 | |
f193fbab | 722 | return err; |
1da177e4 LT |
723 | } |
724 | ||
725 | ||
a334de28 | 726 | static int |
1da177e4 LT |
727 | nfsd_sync(struct file *filp) |
728 | { | |
a334de28 | 729 | int err; |
7eaa36e2 JJS |
730 | struct inode *inode = filp->f_path.dentry->d_inode; |
731 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
1b1dcc1b | 732 | mutex_lock(&inode->i_mutex); |
7eaa36e2 | 733 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); |
1b1dcc1b | 734 | mutex_unlock(&inode->i_mutex); |
a334de28 DS |
735 | |
736 | return err; | |
1da177e4 LT |
737 | } |
738 | ||
f193fbab | 739 | int |
1da177e4 LT |
740 | nfsd_sync_dir(struct dentry *dp) |
741 | { | |
f193fbab | 742 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
743 | } |
744 | ||
745 | /* | |
746 | * Obtain the readahead parameters for the file | |
747 | * specified by (dev, ino). | |
748 | */ | |
1da177e4 LT |
749 | |
750 | static inline struct raparms * | |
751 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
752 | { | |
753 | struct raparms *ra, **rap, **frap = NULL; | |
754 | int depth = 0; | |
fce1456a GB |
755 | unsigned int hash; |
756 | struct raparm_hbucket *rab; | |
757 | ||
758 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
759 | rab = &raparm_hash[hash]; | |
1da177e4 | 760 | |
fce1456a GB |
761 | spin_lock(&rab->pb_lock); |
762 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
763 | if (ra->p_ino == ino && ra->p_dev == dev) |
764 | goto found; | |
765 | depth++; | |
766 | if (ra->p_count == 0) | |
767 | frap = rap; | |
768 | } | |
769 | depth = nfsdstats.ra_size*11/10; | |
770 | if (!frap) { | |
fce1456a | 771 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
772 | return NULL; |
773 | } | |
774 | rap = frap; | |
775 | ra = *frap; | |
776 | ra->p_dev = dev; | |
777 | ra->p_ino = ino; | |
778 | ra->p_set = 0; | |
fce1456a | 779 | ra->p_hindex = hash; |
1da177e4 | 780 | found: |
fce1456a | 781 | if (rap != &rab->pb_head) { |
1da177e4 | 782 | *rap = ra->p_next; |
fce1456a GB |
783 | ra->p_next = rab->pb_head; |
784 | rab->pb_head = ra; | |
1da177e4 LT |
785 | } |
786 | ra->p_count++; | |
787 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 788 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
789 | return ra; |
790 | } | |
791 | ||
792 | /* | |
cf8208d0 JA |
793 | * Grab and keep cached pages associated with a file in the svc_rqst |
794 | * so that they can be passed to the network sendmsg/sendpage routines | |
795 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
796 | */ |
797 | static int | |
cf8208d0 JA |
798 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
799 | struct splice_desc *sd) | |
1da177e4 | 800 | { |
cf8208d0 | 801 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 802 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
803 | struct page *page = buf->page; |
804 | size_t size; | |
805 | int ret; | |
1da177e4 | 806 | |
cac36bb0 | 807 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
808 | if (unlikely(ret)) |
809 | return ret; | |
810 | ||
811 | size = sd->len; | |
1da177e4 LT |
812 | |
813 | if (rqstp->rq_res.page_len == 0) { | |
814 | get_page(page); | |
44524359 N |
815 | put_page(*pp); |
816 | *pp = page; | |
817 | rqstp->rq_resused++; | |
cf8208d0 | 818 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 819 | rqstp->rq_res.page_len = size; |
44524359 | 820 | } else if (page != pp[-1]) { |
1da177e4 | 821 | get_page(page); |
250f3915 N |
822 | if (*pp) |
823 | put_page(*pp); | |
44524359 N |
824 | *pp = page; |
825 | rqstp->rq_resused++; | |
1da177e4 | 826 | rqstp->rq_res.page_len += size; |
44524359 | 827 | } else |
1da177e4 | 828 | rqstp->rq_res.page_len += size; |
1da177e4 | 829 | |
1da177e4 LT |
830 | return size; |
831 | } | |
832 | ||
cf8208d0 JA |
833 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
834 | struct splice_desc *sd) | |
835 | { | |
836 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
837 | } | |
838 | ||
6264d69d | 839 | static __be32 |
1da177e4 LT |
840 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
841 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
842 | { | |
843 | struct inode *inode; | |
844 | struct raparms *ra; | |
845 | mm_segment_t oldfs; | |
6264d69d AV |
846 | __be32 err; |
847 | int host_err; | |
1da177e4 LT |
848 | |
849 | err = nfserr_perm; | |
7eaa36e2 | 850 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
851 | #ifdef MSNFS |
852 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
853 | (!lock_may_read(inode, offset, *count))) | |
854 | goto out; | |
855 | #endif | |
856 | ||
857 | /* Get readahead parameters */ | |
858 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
859 | ||
860 | if (ra && ra->p_set) | |
861 | file->f_ra = ra->p_ra; | |
862 | ||
cf8208d0 JA |
863 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
864 | struct splice_desc sd = { | |
865 | .len = 0, | |
866 | .total_len = *count, | |
867 | .pos = offset, | |
868 | .u.data = rqstp, | |
869 | }; | |
870 | ||
4fbef206 | 871 | rqstp->rq_resused = 1; |
cf8208d0 | 872 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
873 | } else { |
874 | oldfs = get_fs(); | |
875 | set_fs(KERNEL_DS); | |
6264d69d | 876 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
877 | set_fs(oldfs); |
878 | } | |
879 | ||
880 | /* Write back readahead params */ | |
881 | if (ra) { | |
fce1456a GB |
882 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
883 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
884 | ra->p_ra = file->f_ra; |
885 | ra->p_set = 1; | |
886 | ra->p_count--; | |
fce1456a | 887 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
888 | } |
889 | ||
6264d69d AV |
890 | if (host_err >= 0) { |
891 | nfsdstats.io_read += host_err; | |
892 | *count = host_err; | |
1da177e4 | 893 | err = 0; |
7eaa36e2 | 894 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 895 | } else |
6264d69d | 896 | err = nfserrno(host_err); |
1da177e4 LT |
897 | out: |
898 | return err; | |
899 | } | |
900 | ||
9f708e40 NB |
901 | static void kill_suid(struct dentry *dentry) |
902 | { | |
903 | struct iattr ia; | |
904 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID; | |
905 | ||
1b1dcc1b | 906 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 907 | notify_change(dentry, &ia); |
1b1dcc1b | 908 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
909 | } |
910 | ||
6264d69d | 911 | static __be32 |
1da177e4 LT |
912 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
913 | loff_t offset, struct kvec *vec, int vlen, | |
914 | unsigned long cnt, int *stablep) | |
915 | { | |
916 | struct svc_export *exp; | |
917 | struct dentry *dentry; | |
918 | struct inode *inode; | |
919 | mm_segment_t oldfs; | |
6264d69d AV |
920 | __be32 err = 0; |
921 | int host_err; | |
1da177e4 LT |
922 | int stable = *stablep; |
923 | ||
45bd3b3d | 924 | #ifdef MSNFS |
1da177e4 LT |
925 | err = nfserr_perm; |
926 | ||
1da177e4 | 927 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
7eaa36e2 | 928 | (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt))) |
1da177e4 LT |
929 | goto out; |
930 | #endif | |
931 | ||
7eaa36e2 | 932 | dentry = file->f_path.dentry; |
1da177e4 LT |
933 | inode = dentry->d_inode; |
934 | exp = fhp->fh_export; | |
935 | ||
936 | /* | |
937 | * Request sync writes if | |
938 | * - the sync export option has been set, or | |
939 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
940 | * - The file system doesn't support fsync(). | |
941 | * When gathered writes have been configured for this volume, | |
942 | * flushing the data to disk is handled separately below. | |
943 | */ | |
944 | ||
945 | if (file->f_op->fsync == 0) {/* COMMIT3 cannot work */ | |
946 | stable = 2; | |
947 | *stablep = 2; /* FILE_SYNC */ | |
948 | } | |
949 | ||
950 | if (!EX_ISSYNC(exp)) | |
951 | stable = 0; | |
952 | if (stable && !EX_WGATHER(exp)) | |
953 | file->f_flags |= O_SYNC; | |
954 | ||
955 | /* Write the data. */ | |
956 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 957 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 958 | set_fs(oldfs); |
6264d69d | 959 | if (host_err >= 0) { |
1da177e4 | 960 | nfsdstats.io_write += cnt; |
7eaa36e2 | 961 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
962 | } |
963 | ||
964 | /* clear setuid/setgid flag after write */ | |
6264d69d | 965 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 966 | kill_suid(dentry); |
1da177e4 | 967 | |
6264d69d | 968 | if (host_err >= 0 && stable) { |
1da177e4 LT |
969 | static ino_t last_ino; |
970 | static dev_t last_dev; | |
971 | ||
972 | /* | |
973 | * Gathered writes: If another process is currently | |
974 | * writing to the file, there's a high chance | |
975 | * this is another nfsd (triggered by a bulk write | |
976 | * from a client's biod). Rather than syncing the | |
977 | * file with each write request, we sleep for 10 msec. | |
978 | * | |
979 | * I don't know if this roughly approximates | |
980 | * C. Juszak's idea of gathered writes, but it's a | |
981 | * nice and simple solution (IMHO), and it seems to | |
982 | * work:-) | |
983 | */ | |
984 | if (EX_WGATHER(exp)) { | |
985 | if (atomic_read(&inode->i_writecount) > 1 | |
986 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
987 | dprintk("nfsd: write defer %d\n", current->pid); | |
988 | msleep(10); | |
989 | dprintk("nfsd: write resume %d\n", current->pid); | |
990 | } | |
991 | ||
992 | if (inode->i_state & I_DIRTY) { | |
993 | dprintk("nfsd: write sync %d\n", current->pid); | |
6264d69d | 994 | host_err=nfsd_sync(file); |
1da177e4 LT |
995 | } |
996 | #if 0 | |
997 | wake_up(&inode->i_wait); | |
998 | #endif | |
999 | } | |
1000 | last_ino = inode->i_ino; | |
1001 | last_dev = inode->i_sb->s_dev; | |
1002 | } | |
1003 | ||
6264d69d AV |
1004 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
1005 | if (host_err >= 0) | |
1da177e4 LT |
1006 | err = 0; |
1007 | else | |
6264d69d | 1008 | err = nfserrno(host_err); |
1da177e4 LT |
1009 | out: |
1010 | return err; | |
1011 | } | |
1012 | ||
1013 | /* | |
1014 | * Read data from a file. count must contain the requested read count | |
1015 | * on entry. On return, *count contains the number of bytes actually read. | |
1016 | * N.B. After this call fhp needs an fh_put | |
1017 | */ | |
6264d69d | 1018 | __be32 |
1da177e4 LT |
1019 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1020 | loff_t offset, struct kvec *vec, int vlen, | |
1021 | unsigned long *count) | |
1022 | { | |
6264d69d | 1023 | __be32 err; |
1da177e4 LT |
1024 | |
1025 | if (file) { | |
1026 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1027 | MAY_READ|MAY_OWNER_OVERRIDE); | |
1028 | if (err) | |
1029 | goto out; | |
1030 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1031 | } else { | |
1032 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_READ, &file); | |
1033 | if (err) | |
1034 | goto out; | |
1035 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1036 | nfsd_close(file); | |
1037 | } | |
1038 | out: | |
1039 | return err; | |
1040 | } | |
1041 | ||
1042 | /* | |
1043 | * Write data to a file. | |
1044 | * The stable flag requests synchronous writes. | |
1045 | * N.B. After this call fhp needs an fh_put | |
1046 | */ | |
6264d69d | 1047 | __be32 |
1da177e4 LT |
1048 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1049 | loff_t offset, struct kvec *vec, int vlen, unsigned long cnt, | |
1050 | int *stablep) | |
1051 | { | |
6264d69d | 1052 | __be32 err = 0; |
1da177e4 LT |
1053 | |
1054 | if (file) { | |
1055 | err = nfsd_permission(fhp->fh_export, fhp->fh_dentry, | |
1056 | MAY_WRITE|MAY_OWNER_OVERRIDE); | |
1057 | if (err) | |
1058 | goto out; | |
1059 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1060 | stablep); | |
1061 | } else { | |
1062 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file); | |
1063 | if (err) | |
1064 | goto out; | |
1065 | ||
1066 | if (cnt) | |
1067 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1068 | cnt, stablep); | |
1069 | nfsd_close(file); | |
1070 | } | |
1071 | out: | |
1072 | return err; | |
1073 | } | |
1074 | ||
1075 | #ifdef CONFIG_NFSD_V3 | |
1076 | /* | |
1077 | * Commit all pending writes to stable storage. | |
1078 | * Strictly speaking, we could sync just the indicated file region here, | |
1079 | * but there's currently no way we can ask the VFS to do so. | |
1080 | * | |
1081 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1082 | * data to the client, as locking happens lower down in the filesystem. | |
1083 | */ | |
6264d69d | 1084 | __be32 |
1da177e4 LT |
1085 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1086 | loff_t offset, unsigned long count) | |
1087 | { | |
1088 | struct file *file; | |
6264d69d | 1089 | __be32 err; |
1da177e4 LT |
1090 | |
1091 | if ((u64)count > ~(u64)offset) | |
1092 | return nfserr_inval; | |
1093 | ||
1094 | if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0) | |
1095 | return err; | |
1096 | if (EX_ISSYNC(fhp->fh_export)) { | |
1097 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1098 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1099 | } else { |
1100 | err = nfserr_notsupp; | |
1101 | } | |
1102 | } | |
1103 | ||
1104 | nfsd_close(file); | |
1105 | return err; | |
1106 | } | |
1107 | #endif /* CONFIG_NFSD_V3 */ | |
1108 | ||
1109 | /* | |
1110 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1111 | * not yet implemented. | |
1112 | * If the response fh has been verified, the parent directory should | |
1113 | * already be locked. Note that the parent directory is left locked. | |
1114 | * | |
1115 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1116 | */ | |
6264d69d | 1117 | __be32 |
1da177e4 LT |
1118 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1119 | char *fname, int flen, struct iattr *iap, | |
1120 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1121 | { | |
1122 | struct dentry *dentry, *dchild = NULL; | |
1123 | struct inode *dirp; | |
6264d69d AV |
1124 | __be32 err; |
1125 | int host_err; | |
1da177e4 LT |
1126 | |
1127 | err = nfserr_perm; | |
1128 | if (!flen) | |
1129 | goto out; | |
1130 | err = nfserr_exist; | |
1131 | if (isdotent(fname, flen)) | |
1132 | goto out; | |
1133 | ||
1134 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1135 | if (err) | |
1136 | goto out; | |
1137 | ||
1138 | dentry = fhp->fh_dentry; | |
1139 | dirp = dentry->d_inode; | |
1140 | ||
1141 | err = nfserr_notdir; | |
1142 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1143 | goto out; | |
1144 | /* | |
1145 | * Check whether the response file handle has been verified yet. | |
1146 | * If it has, the parent directory should already be locked. | |
1147 | */ | |
1148 | if (!resfhp->fh_dentry) { | |
1149 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1150 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1151 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1152 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1153 | if (IS_ERR(dchild)) |
1154 | goto out_nfserr; | |
1155 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1156 | if (err) | |
1157 | goto out; | |
1158 | } else { | |
1159 | /* called from nfsd_proc_create */ | |
1160 | dchild = dget(resfhp->fh_dentry); | |
1161 | if (!fhp->fh_locked) { | |
1162 | /* not actually possible */ | |
1163 | printk(KERN_ERR | |
1164 | "nfsd_create: parent %s/%s not locked!\n", | |
1165 | dentry->d_parent->d_name.name, | |
1166 | dentry->d_name.name); | |
d75f2b9f | 1167 | err = nfserr_io; |
1da177e4 LT |
1168 | goto out; |
1169 | } | |
1170 | } | |
1171 | /* | |
1172 | * Make sure the child dentry is still negative ... | |
1173 | */ | |
1174 | err = nfserr_exist; | |
1175 | if (dchild->d_inode) { | |
1176 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1177 | dentry->d_name.name, dchild->d_name.name); | |
1178 | goto out; | |
1179 | } | |
1180 | ||
1181 | if (!(iap->ia_valid & ATTR_MODE)) | |
1182 | iap->ia_mode = 0; | |
1183 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1184 | ||
1185 | /* | |
1186 | * Get the dir op function pointer. | |
1187 | */ | |
088406bc | 1188 | err = 0; |
1da177e4 LT |
1189 | switch (type) { |
1190 | case S_IFREG: | |
6264d69d | 1191 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1da177e4 LT |
1192 | break; |
1193 | case S_IFDIR: | |
6264d69d | 1194 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1195 | break; |
1196 | case S_IFCHR: | |
1197 | case S_IFBLK: | |
1198 | case S_IFIFO: | |
1199 | case S_IFSOCK: | |
6264d69d | 1200 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 LT |
1201 | break; |
1202 | default: | |
1203 | printk("nfsd: bad file type %o in nfsd_create\n", type); | |
6264d69d | 1204 | host_err = -EINVAL; |
1da177e4 | 1205 | } |
6264d69d | 1206 | if (host_err < 0) |
1da177e4 LT |
1207 | goto out_nfserr; |
1208 | ||
1209 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1210 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1211 | write_inode_now(dchild->d_inode, 1); |
1212 | } | |
1213 | ||
1214 | ||
1215 | /* Set file attributes. Mode has already been set and | |
1216 | * setting uid/gid works only for root. Irix appears to | |
1217 | * send along the gid when it tries to implement setgid | |
1218 | * directories via NFS. | |
1219 | */ | |
f193fbab | 1220 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1221 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab YT |
1222 | if (err2) |
1223 | err = err2; | |
1224 | } | |
1da177e4 LT |
1225 | /* |
1226 | * Update the file handle to get the new inode info. | |
1227 | */ | |
1228 | if (!err) | |
1229 | err = fh_update(resfhp); | |
1230 | out: | |
1231 | if (dchild && !IS_ERR(dchild)) | |
1232 | dput(dchild); | |
1233 | return err; | |
1234 | ||
1235 | out_nfserr: | |
6264d69d | 1236 | err = nfserrno(host_err); |
1da177e4 LT |
1237 | goto out; |
1238 | } | |
1239 | ||
1240 | #ifdef CONFIG_NFSD_V3 | |
1241 | /* | |
1242 | * NFSv3 version of nfsd_create | |
1243 | */ | |
6264d69d | 1244 | __be32 |
1da177e4 LT |
1245 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1246 | char *fname, int flen, struct iattr *iap, | |
1247 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1248 | int *truncp, int *created) |
1da177e4 LT |
1249 | { |
1250 | struct dentry *dentry, *dchild = NULL; | |
1251 | struct inode *dirp; | |
6264d69d AV |
1252 | __be32 err; |
1253 | int host_err; | |
1da177e4 | 1254 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1255 | |
1256 | err = nfserr_perm; | |
1257 | if (!flen) | |
1258 | goto out; | |
1259 | err = nfserr_exist; | |
1260 | if (isdotent(fname, flen)) | |
1261 | goto out; | |
1262 | if (!(iap->ia_valid & ATTR_MODE)) | |
1263 | iap->ia_mode = 0; | |
1264 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1265 | if (err) | |
1266 | goto out; | |
1267 | ||
1268 | dentry = fhp->fh_dentry; | |
1269 | dirp = dentry->d_inode; | |
1270 | ||
1271 | /* Get all the sanity checks out of the way before | |
1272 | * we lock the parent. */ | |
1273 | err = nfserr_notdir; | |
1274 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1275 | goto out; | |
12fd3520 | 1276 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1277 | |
1278 | /* | |
1279 | * Compose the response file handle. | |
1280 | */ | |
1281 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1282 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1283 | if (IS_ERR(dchild)) |
1284 | goto out_nfserr; | |
1285 | ||
1286 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1287 | if (err) | |
1288 | goto out; | |
1289 | ||
1290 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c PS |
1291 | /* solaris7 gets confused (bugid 4218508) if these have |
1292 | * the high bit set, so just clear the high bits. | |
1da177e4 LT |
1293 | */ |
1294 | v_mtime = verifier[0]&0x7fffffff; | |
1295 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1296 | } |
1297 | ||
1298 | if (dchild->d_inode) { | |
1299 | err = 0; | |
1300 | ||
1301 | switch (createmode) { | |
1302 | case NFS3_CREATE_UNCHECKED: | |
1303 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1304 | err = nfserr_exist; | |
1305 | else if (truncp) { | |
1306 | /* in nfsv4, we need to treat this case a little | |
1307 | * differently. we don't want to truncate the | |
1308 | * file now; this would be wrong if the OPEN | |
1309 | * fails for some other reason. furthermore, | |
1310 | * if the size is nonzero, we should ignore it | |
1311 | * according to spec! | |
1312 | */ | |
1313 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1314 | } | |
1315 | else { | |
1316 | iap->ia_valid &= ATTR_SIZE; | |
1317 | goto set_attr; | |
1318 | } | |
1319 | break; | |
1320 | case NFS3_CREATE_EXCLUSIVE: | |
1321 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1322 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1323 | && dchild->d_inode->i_size == 0 ) |
1324 | break; | |
1325 | /* fallthru */ | |
1326 | case NFS3_CREATE_GUARDED: | |
1327 | err = nfserr_exist; | |
1328 | } | |
1329 | goto out; | |
1330 | } | |
1331 | ||
6264d69d AV |
1332 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1333 | if (host_err < 0) | |
1da177e4 | 1334 | goto out_nfserr; |
81ac95c5 BF |
1335 | if (created) |
1336 | *created = 1; | |
1da177e4 LT |
1337 | |
1338 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1339 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1340 | /* setattr will sync the child (or not) */ |
1341 | } | |
1342 | ||
1da177e4 | 1343 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1344 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1345 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1346 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1347 | /* XXX someone who knows this better please fix it for nsec */ |
1348 | iap->ia_mtime.tv_sec = v_mtime; | |
1349 | iap->ia_atime.tv_sec = v_atime; | |
1350 | iap->ia_mtime.tv_nsec = 0; | |
1351 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1352 | } |
1353 | ||
1354 | /* Set file attributes. | |
1da177e4 LT |
1355 | * Irix appears to send along the gid when it tries to |
1356 | * implement setgid directories via NFS. Clear out all that cruft. | |
1357 | */ | |
1358 | set_attr: | |
c397852c | 1359 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1360 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab | 1361 | if (err2) |
45bd3b3d | 1362 | err = err2; |
f193fbab YT |
1363 | } |
1364 | ||
1365 | /* | |
1366 | * Update the filehandle to get the new inode info. | |
1367 | */ | |
1368 | if (!err) | |
1369 | err = fh_update(resfhp); | |
1da177e4 LT |
1370 | |
1371 | out: | |
1372 | fh_unlock(fhp); | |
1373 | if (dchild && !IS_ERR(dchild)) | |
1374 | dput(dchild); | |
1375 | return err; | |
1376 | ||
1377 | out_nfserr: | |
6264d69d | 1378 | err = nfserrno(host_err); |
1da177e4 LT |
1379 | goto out; |
1380 | } | |
1381 | #endif /* CONFIG_NFSD_V3 */ | |
1382 | ||
1383 | /* | |
1384 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1385 | * fits into the buffer. On return, it contains the true length. | |
1386 | * N.B. After this call fhp needs an fh_put | |
1387 | */ | |
6264d69d | 1388 | __be32 |
1da177e4 LT |
1389 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1390 | { | |
1391 | struct dentry *dentry; | |
1392 | struct inode *inode; | |
1393 | mm_segment_t oldfs; | |
6264d69d AV |
1394 | __be32 err; |
1395 | int host_err; | |
1da177e4 LT |
1396 | |
1397 | err = fh_verify(rqstp, fhp, S_IFLNK, MAY_NOP); | |
1398 | if (err) | |
1399 | goto out; | |
1400 | ||
1401 | dentry = fhp->fh_dentry; | |
1402 | inode = dentry->d_inode; | |
1403 | ||
1404 | err = nfserr_inval; | |
1405 | if (!inode->i_op || !inode->i_op->readlink) | |
1406 | goto out; | |
1407 | ||
1408 | touch_atime(fhp->fh_export->ex_mnt, dentry); | |
1409 | /* N.B. Why does this call need a get_fs()?? | |
1410 | * Remove the set_fs and watch the fireworks:-) --okir | |
1411 | */ | |
1412 | ||
1413 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1414 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1415 | set_fs(oldfs); |
1416 | ||
6264d69d | 1417 | if (host_err < 0) |
1da177e4 | 1418 | goto out_nfserr; |
6264d69d | 1419 | *lenp = host_err; |
1da177e4 LT |
1420 | err = 0; |
1421 | out: | |
1422 | return err; | |
1423 | ||
1424 | out_nfserr: | |
6264d69d | 1425 | err = nfserrno(host_err); |
1da177e4 LT |
1426 | goto out; |
1427 | } | |
1428 | ||
1429 | /* | |
1430 | * Create a symlink and look up its inode | |
1431 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1432 | */ | |
6264d69d | 1433 | __be32 |
1da177e4 LT |
1434 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1435 | char *fname, int flen, | |
1436 | char *path, int plen, | |
1437 | struct svc_fh *resfhp, | |
1438 | struct iattr *iap) | |
1439 | { | |
1440 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1441 | __be32 err, cerr; |
1442 | int host_err; | |
1da177e4 LT |
1443 | umode_t mode; |
1444 | ||
1445 | err = nfserr_noent; | |
1446 | if (!flen || !plen) | |
1447 | goto out; | |
1448 | err = nfserr_exist; | |
1449 | if (isdotent(fname, flen)) | |
1450 | goto out; | |
1451 | ||
1452 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1453 | if (err) | |
1454 | goto out; | |
1455 | fh_lock(fhp); | |
1456 | dentry = fhp->fh_dentry; | |
1457 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1458 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1459 | if (IS_ERR(dnew)) |
1460 | goto out_nfserr; | |
1461 | ||
1462 | mode = S_IALLUGO; | |
1463 | /* Only the MODE ATTRibute is even vaguely meaningful */ | |
1464 | if (iap && (iap->ia_valid & ATTR_MODE)) | |
1465 | mode = iap->ia_mode & S_IALLUGO; | |
1466 | ||
1467 | if (unlikely(path[plen] != 0)) { | |
1468 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1469 | if (path_alloced == NULL) | |
6264d69d | 1470 | host_err = -ENOMEM; |
1da177e4 LT |
1471 | else { |
1472 | strncpy(path_alloced, path, plen); | |
1473 | path_alloced[plen] = 0; | |
6264d69d | 1474 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode); |
1da177e4 LT |
1475 | kfree(path_alloced); |
1476 | } | |
1477 | } else | |
6264d69d | 1478 | host_err = vfs_symlink(dentry->d_inode, dnew, path, mode); |
1da177e4 | 1479 | |
6264d69d | 1480 | if (!host_err) { |
1da177e4 | 1481 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1482 | host_err = nfsd_sync_dir(dentry); |
1483 | } | |
1484 | err = nfserrno(host_err); | |
1da177e4 LT |
1485 | fh_unlock(fhp); |
1486 | ||
1487 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); | |
1488 | dput(dnew); | |
1489 | if (err==0) err = cerr; | |
1490 | out: | |
1491 | return err; | |
1492 | ||
1493 | out_nfserr: | |
6264d69d | 1494 | err = nfserrno(host_err); |
1da177e4 LT |
1495 | goto out; |
1496 | } | |
1497 | ||
1498 | /* | |
1499 | * Create a hardlink | |
1500 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1501 | */ | |
6264d69d | 1502 | __be32 |
1da177e4 LT |
1503 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1504 | char *name, int len, struct svc_fh *tfhp) | |
1505 | { | |
1506 | struct dentry *ddir, *dnew, *dold; | |
1507 | struct inode *dirp, *dest; | |
6264d69d AV |
1508 | __be32 err; |
1509 | int host_err; | |
1da177e4 LT |
1510 | |
1511 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_CREATE); | |
1512 | if (err) | |
1513 | goto out; | |
1514 | err = fh_verify(rqstp, tfhp, -S_IFDIR, MAY_NOP); | |
1515 | if (err) | |
1516 | goto out; | |
1517 | ||
1518 | err = nfserr_perm; | |
1519 | if (!len) | |
1520 | goto out; | |
1521 | err = nfserr_exist; | |
1522 | if (isdotent(name, len)) | |
1523 | goto out; | |
1524 | ||
12fd3520 | 1525 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1526 | ddir = ffhp->fh_dentry; |
1527 | dirp = ddir->d_inode; | |
1528 | ||
1529 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1530 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1531 | if (IS_ERR(dnew)) |
1532 | goto out_nfserr; | |
1533 | ||
1534 | dold = tfhp->fh_dentry; | |
1535 | dest = dold->d_inode; | |
1536 | ||
6264d69d AV |
1537 | host_err = vfs_link(dold, dirp, dnew); |
1538 | if (!host_err) { | |
1da177e4 | 1539 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1540 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1541 | write_inode_now(dest, 1); |
1542 | } | |
6264d69d | 1543 | err = 0; |
1da177e4 | 1544 | } else { |
6264d69d | 1545 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1546 | err = nfserr_acces; |
1547 | else | |
6264d69d | 1548 | err = nfserrno(host_err); |
1da177e4 LT |
1549 | } |
1550 | ||
1da177e4 | 1551 | dput(dnew); |
270d56e5 DR |
1552 | out_unlock: |
1553 | fh_unlock(ffhp); | |
1da177e4 LT |
1554 | out: |
1555 | return err; | |
1556 | ||
1557 | out_nfserr: | |
6264d69d | 1558 | err = nfserrno(host_err); |
270d56e5 | 1559 | goto out_unlock; |
1da177e4 LT |
1560 | } |
1561 | ||
1562 | /* | |
1563 | * Rename a file | |
1564 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1565 | */ | |
6264d69d | 1566 | __be32 |
1da177e4 LT |
1567 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1568 | struct svc_fh *tfhp, char *tname, int tlen) | |
1569 | { | |
1570 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1571 | struct inode *fdir, *tdir; | |
6264d69d AV |
1572 | __be32 err; |
1573 | int host_err; | |
1da177e4 LT |
1574 | |
1575 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_REMOVE); | |
1576 | if (err) | |
1577 | goto out; | |
1578 | err = fh_verify(rqstp, tfhp, S_IFDIR, MAY_CREATE); | |
1579 | if (err) | |
1580 | goto out; | |
1581 | ||
1582 | fdentry = ffhp->fh_dentry; | |
1583 | fdir = fdentry->d_inode; | |
1584 | ||
1585 | tdentry = tfhp->fh_dentry; | |
1586 | tdir = tdentry->d_inode; | |
1587 | ||
1588 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1589 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1590 | goto out; |
1591 | ||
1592 | err = nfserr_perm; | |
1593 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1594 | goto out; | |
1595 | ||
1596 | /* cannot use fh_lock as we need deadlock protective ordering | |
1597 | * so do it by hand */ | |
1598 | trap = lock_rename(tdentry, fdentry); | |
1599 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1600 | fill_pre_wcc(ffhp); | |
1601 | fill_pre_wcc(tfhp); | |
1602 | ||
1603 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1604 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1605 | if (IS_ERR(odentry)) |
1606 | goto out_nfserr; | |
1607 | ||
6264d69d | 1608 | host_err = -ENOENT; |
1da177e4 LT |
1609 | if (!odentry->d_inode) |
1610 | goto out_dput_old; | |
6264d69d | 1611 | host_err = -EINVAL; |
1da177e4 LT |
1612 | if (odentry == trap) |
1613 | goto out_dput_old; | |
1614 | ||
1615 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1616 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1617 | if (IS_ERR(ndentry)) |
1618 | goto out_dput_old; | |
6264d69d | 1619 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1620 | if (ndentry == trap) |
1621 | goto out_dput_new; | |
1622 | ||
1623 | #ifdef MSNFS | |
1624 | if ((ffhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1625 | ((atomic_read(&odentry->d_count) > 1) | |
1626 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1627 | host_err = -EPERM; |
1da177e4 LT |
1628 | } else |
1629 | #endif | |
6264d69d AV |
1630 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1631 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1632 | host_err = nfsd_sync_dir(tdentry); | |
1633 | if (!host_err) | |
1634 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1635 | } |
1636 | ||
1637 | out_dput_new: | |
1638 | dput(ndentry); | |
1639 | out_dput_old: | |
1640 | dput(odentry); | |
1641 | out_nfserr: | |
6264d69d | 1642 | err = nfserrno(host_err); |
1da177e4 LT |
1643 | |
1644 | /* we cannot reply on fh_unlock on the two filehandles, | |
1645 | * as that would do the wrong thing if the two directories | |
1646 | * were the same, so again we do it by hand | |
1647 | */ | |
1648 | fill_post_wcc(ffhp); | |
1649 | fill_post_wcc(tfhp); | |
1650 | unlock_rename(tdentry, fdentry); | |
1651 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1652 | ||
1653 | out: | |
1654 | return err; | |
1655 | } | |
1656 | ||
1657 | /* | |
1658 | * Unlink a file or directory | |
1659 | * N.B. After this call fhp needs an fh_put | |
1660 | */ | |
6264d69d | 1661 | __be32 |
1da177e4 LT |
1662 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1663 | char *fname, int flen) | |
1664 | { | |
1665 | struct dentry *dentry, *rdentry; | |
1666 | struct inode *dirp; | |
6264d69d AV |
1667 | __be32 err; |
1668 | int host_err; | |
1da177e4 LT |
1669 | |
1670 | err = nfserr_acces; | |
1671 | if (!flen || isdotent(fname, flen)) | |
1672 | goto out; | |
1673 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_REMOVE); | |
1674 | if (err) | |
1675 | goto out; | |
1676 | ||
12fd3520 | 1677 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1678 | dentry = fhp->fh_dentry; |
1679 | dirp = dentry->d_inode; | |
1680 | ||
1681 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1682 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1683 | if (IS_ERR(rdentry)) |
1684 | goto out_nfserr; | |
1685 | ||
1686 | if (!rdentry->d_inode) { | |
1687 | dput(rdentry); | |
1688 | err = nfserr_noent; | |
1689 | goto out; | |
1690 | } | |
1691 | ||
1692 | if (!type) | |
1693 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1694 | ||
1695 | if (type != S_IFDIR) { /* It's UNLINK */ | |
1696 | #ifdef MSNFS | |
1697 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1698 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1699 | host_err = -EPERM; |
1da177e4 LT |
1700 | } else |
1701 | #endif | |
6264d69d | 1702 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1703 | } else { /* It's RMDIR */ |
6264d69d | 1704 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1705 | } |
1706 | ||
1707 | dput(rdentry); | |
1708 | ||
6264d69d AV |
1709 | if (host_err) |
1710 | goto out_nfserr; | |
1711 | if (EX_ISSYNC(fhp->fh_export)) | |
1712 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 LT |
1713 | |
1714 | out_nfserr: | |
6264d69d | 1715 | err = nfserrno(host_err); |
f193fbab YT |
1716 | out: |
1717 | return err; | |
1da177e4 LT |
1718 | } |
1719 | ||
1720 | /* | |
1721 | * Read entries from a directory. | |
1722 | * The NFSv3/4 verifier we ignore for now. | |
1723 | */ | |
6264d69d | 1724 | __be32 |
1da177e4 | 1725 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1726 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1727 | { |
6264d69d AV |
1728 | __be32 err; |
1729 | int host_err; | |
1da177e4 LT |
1730 | struct file *file; |
1731 | loff_t offset = *offsetp; | |
1732 | ||
1733 | err = nfsd_open(rqstp, fhp, S_IFDIR, MAY_READ, &file); | |
1734 | if (err) | |
1735 | goto out; | |
1736 | ||
1737 | offset = vfs_llseek(file, offset, 0); | |
1738 | if (offset < 0) { | |
1739 | err = nfserrno((int)offset); | |
1740 | goto out_close; | |
1741 | } | |
1742 | ||
1743 | /* | |
1744 | * Read the directory entries. This silly loop is necessary because | |
1745 | * readdir() is not guaranteed to fill up the entire buffer, but | |
1746 | * may choose to do less. | |
1747 | */ | |
1748 | ||
1749 | do { | |
1750 | cdp->err = nfserr_eof; /* will be cleared on successful read */ | |
a0ad13ef | 1751 | host_err = vfs_readdir(file, func, cdp); |
6264d69d AV |
1752 | } while (host_err >=0 && cdp->err == nfs_ok); |
1753 | if (host_err) | |
1754 | err = nfserrno(host_err); | |
1da177e4 LT |
1755 | else |
1756 | err = cdp->err; | |
1757 | *offsetp = vfs_llseek(file, 0, 1); | |
1758 | ||
1759 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1760 | err = nfs_ok; /* can still be found in ->err */ | |
1761 | out_close: | |
1762 | nfsd_close(file); | |
1763 | out: | |
1764 | return err; | |
1765 | } | |
1766 | ||
1767 | /* | |
1768 | * Get file system stats | |
1769 | * N.B. After this call fhp needs an fh_put | |
1770 | */ | |
6264d69d | 1771 | __be32 |
1da177e4 LT |
1772 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat) |
1773 | { | |
6264d69d | 1774 | __be32 err = fh_verify(rqstp, fhp, 0, MAY_NOP); |
726c3342 | 1775 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
1776 | err = nfserr_io; |
1777 | return err; | |
1778 | } | |
1779 | ||
1780 | /* | |
1781 | * Check for a user's access permissions to this inode. | |
1782 | */ | |
6264d69d | 1783 | __be32 |
1da177e4 LT |
1784 | nfsd_permission(struct svc_export *exp, struct dentry *dentry, int acc) |
1785 | { | |
1786 | struct inode *inode = dentry->d_inode; | |
1787 | int err; | |
1788 | ||
1789 | if (acc == MAY_NOP) | |
1790 | return 0; | |
1791 | #if 0 | |
1792 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
1793 | acc, | |
1794 | (acc & MAY_READ)? " read" : "", | |
1795 | (acc & MAY_WRITE)? " write" : "", | |
1796 | (acc & MAY_EXEC)? " exec" : "", | |
1797 | (acc & MAY_SATTR)? " sattr" : "", | |
1798 | (acc & MAY_TRUNC)? " trunc" : "", | |
1799 | (acc & MAY_LOCK)? " lock" : "", | |
1800 | (acc & MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1801 | inode->i_mode, | |
1802 | IS_IMMUTABLE(inode)? " immut" : "", | |
1803 | IS_APPEND(inode)? " append" : "", | |
1804 | IS_RDONLY(inode)? " ro" : ""); | |
1805 | dprintk(" owner %d/%d user %d/%d\n", | |
1806 | inode->i_uid, inode->i_gid, current->fsuid, current->fsgid); | |
1807 | #endif | |
1808 | ||
1809 | /* Normally we reject any write/sattr etc access on a read-only file | |
1810 | * system. But if it is IRIX doing check on write-access for a | |
1811 | * device special file, we ignore rofs. | |
1812 | */ | |
1813 | if (!(acc & MAY_LOCAL_ACCESS)) | |
1814 | if (acc & (MAY_WRITE | MAY_SATTR | MAY_TRUNC)) { | |
1815 | if (EX_RDONLY(exp) || IS_RDONLY(inode)) | |
1816 | return nfserr_rofs; | |
1817 | if (/* (acc & MAY_WRITE) && */ IS_IMMUTABLE(inode)) | |
1818 | return nfserr_perm; | |
1819 | } | |
1820 | if ((acc & MAY_TRUNC) && IS_APPEND(inode)) | |
1821 | return nfserr_perm; | |
1822 | ||
1823 | if (acc & MAY_LOCK) { | |
1824 | /* If we cannot rely on authentication in NLM requests, | |
1825 | * just allow locks, otherwise require read permission, or | |
1826 | * ownership | |
1827 | */ | |
1828 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
1829 | return 0; | |
1830 | else | |
1831 | acc = MAY_READ | MAY_OWNER_OVERRIDE; | |
1832 | } | |
1833 | /* | |
1834 | * The file owner always gets access permission for accesses that | |
1835 | * would normally be checked at open time. This is to make | |
1836 | * file access work even when the client has done a fchmod(fd, 0). | |
1837 | * | |
1838 | * However, `cp foo bar' should fail nevertheless when bar is | |
1839 | * readonly. A sensible way to do this might be to reject all | |
1840 | * attempts to truncate a read-only file, because a creat() call | |
1841 | * always implies file truncation. | |
1842 | * ... but this isn't really fair. A process may reasonably call | |
1843 | * ftruncate on an open file descriptor on a file with perm 000. | |
1844 | * We must trust the client to do permission checking - using "ACCESS" | |
1845 | * with NFSv3. | |
1846 | */ | |
1847 | if ((acc & MAY_OWNER_OVERRIDE) && | |
1848 | inode->i_uid == current->fsuid) | |
1849 | return 0; | |
1850 | ||
1851 | err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL); | |
1852 | ||
1853 | /* Allow read access to binaries even when mode 111 */ | |
1854 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
1855 | acc == (MAY_READ | MAY_OWNER_OVERRIDE)) | |
1856 | err = permission(inode, MAY_EXEC, NULL); | |
1857 | ||
1858 | return err? nfserrno(err) : 0; | |
1859 | } | |
1860 | ||
1861 | void | |
1862 | nfsd_racache_shutdown(void) | |
1863 | { | |
fce1456a | 1864 | if (!raparml) |
1da177e4 LT |
1865 | return; |
1866 | dprintk("nfsd: freeing readahead buffers.\n"); | |
1867 | kfree(raparml); | |
fce1456a | 1868 | raparml = NULL; |
1da177e4 LT |
1869 | } |
1870 | /* | |
1871 | * Initialize readahead param cache | |
1872 | */ | |
1873 | int | |
1874 | nfsd_racache_init(int cache_size) | |
1875 | { | |
1876 | int i; | |
fce1456a GB |
1877 | int j = 0; |
1878 | int nperbucket; | |
1da177e4 | 1879 | |
fce1456a GB |
1880 | |
1881 | if (raparml) | |
1da177e4 | 1882 | return 0; |
fce1456a GB |
1883 | if (cache_size < 2*RAPARM_HASH_SIZE) |
1884 | cache_size = 2*RAPARM_HASH_SIZE; | |
4b3bb06b YB |
1885 | raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL); |
1886 | ||
1887 | if (!raparml) { | |
1da177e4 | 1888 | printk(KERN_WARNING |
4b3bb06b | 1889 | "nfsd: Could not allocate memory read-ahead cache.\n"); |
1da177e4 LT |
1890 | return -ENOMEM; |
1891 | } | |
4b3bb06b YB |
1892 | |
1893 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
1894 | for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) { | |
1895 | raparm_hash[i].pb_head = NULL; | |
1896 | spin_lock_init(&raparm_hash[i].pb_lock); | |
1897 | } | |
1898 | nperbucket = cache_size >> RAPARM_HASH_BITS; | |
1899 | for (i = 0; i < cache_size - 1; i++) { | |
1900 | if (i % nperbucket == 0) | |
1901 | raparm_hash[j++].pb_head = raparml + i; | |
1902 | if (i % nperbucket < nperbucket-1) | |
1903 | raparml[i].p_next = raparml + i + 1; | |
1904 | } | |
1905 | ||
1da177e4 LT |
1906 | nfsdstats.ra_size = cache_size; |
1907 | return 0; | |
1908 | } | |
a257cdd0 AG |
1909 | |
1910 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
1911 | struct posix_acl * | |
1912 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
1913 | { | |
1914 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1915 | char *name; | |
1916 | void *value = NULL; | |
1917 | ssize_t size; | |
1918 | struct posix_acl *acl; | |
1919 | ||
5be196e5 CH |
1920 | if (!IS_POSIXACL(inode)) |
1921 | return ERR_PTR(-EOPNOTSUPP); | |
1922 | ||
1923 | switch (type) { | |
1924 | case ACL_TYPE_ACCESS: | |
1925 | name = POSIX_ACL_XATTR_ACCESS; | |
1926 | break; | |
1927 | case ACL_TYPE_DEFAULT: | |
1928 | name = POSIX_ACL_XATTR_DEFAULT; | |
1929 | break; | |
1930 | default: | |
a257cdd0 | 1931 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
1932 | } |
1933 | ||
5be196e5 CH |
1934 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
1935 | if (size < 0) | |
1936 | return ERR_PTR(size); | |
a257cdd0 | 1937 | |
a257cdd0 | 1938 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
1939 | kfree(value); |
1940 | return acl; | |
1941 | } | |
1942 | ||
1943 | int | |
1944 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
1945 | { | |
1946 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1947 | char *name; | |
1948 | void *value = NULL; | |
1949 | size_t size; | |
1950 | int error; | |
1951 | ||
1952 | if (!IS_POSIXACL(inode) || !inode->i_op || | |
1953 | !inode->i_op->setxattr || !inode->i_op->removexattr) | |
1954 | return -EOPNOTSUPP; | |
1955 | switch(type) { | |
1956 | case ACL_TYPE_ACCESS: | |
334a13ec | 1957 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
1958 | break; |
1959 | case ACL_TYPE_DEFAULT: | |
334a13ec | 1960 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
1961 | break; |
1962 | default: | |
1963 | return -EOPNOTSUPP; | |
1964 | } | |
1965 | ||
1966 | if (acl && acl->a_count) { | |
334a13ec | 1967 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
1968 | value = kmalloc(size, GFP_KERNEL); |
1969 | if (!value) | |
1970 | return -ENOMEM; | |
9ccfc29c FM |
1971 | error = posix_acl_to_xattr(acl, value, size); |
1972 | if (error < 0) | |
a257cdd0 | 1973 | goto getout; |
9ccfc29c | 1974 | size = error; |
a257cdd0 AG |
1975 | } else |
1976 | size = 0; | |
1977 | ||
a257cdd0 | 1978 | if (size) |
5be196e5 | 1979 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
1980 | else { |
1981 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
1982 | error = 0; | |
1983 | else { | |
5be196e5 | 1984 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
1985 | if (error == -ENODATA) |
1986 | error = 0; | |
1987 | } | |
1988 | } | |
1989 | ||
1990 | getout: | |
1991 | kfree(value); | |
1992 | return error; | |
1993 | } | |
1994 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |