]>
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 | ||
1da177e4 LT |
64 | /* |
65 | * This is a cache of readahead params that help us choose the proper | |
66 | * readahead strategy. Initially, we set all readahead parameters to 0 | |
67 | * and let the VFS handle things. | |
68 | * If you increase the number of cached files very much, you'll need to | |
69 | * add a hash table here. | |
70 | */ | |
71 | struct raparms { | |
72 | struct raparms *p_next; | |
73 | unsigned int p_count; | |
74 | ino_t p_ino; | |
75 | dev_t p_dev; | |
76 | int p_set; | |
77 | struct file_ra_state p_ra; | |
fce1456a | 78 | unsigned int p_hindex; |
1da177e4 LT |
79 | }; |
80 | ||
fce1456a GB |
81 | struct raparm_hbucket { |
82 | struct raparms *pb_head; | |
83 | spinlock_t pb_lock; | |
84 | } ____cacheline_aligned_in_smp; | |
85 | ||
fce1456a GB |
86 | #define RAPARM_HASH_BITS 4 |
87 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
88 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
89 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
90 | |
91 | /* | |
92 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
93 | * a mount point. | |
e0bb89ef | 94 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
95 | * or nfs_ok having possibly changed *dpp and *expp |
96 | */ | |
97 | int | |
98 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
99 | struct svc_export **expp) | |
100 | { | |
101 | struct svc_export *exp = *expp, *exp2 = NULL; | |
102 | struct dentry *dentry = *dpp; | |
54775491 | 103 | struct vfsmount *mnt = mntget(exp->ex_path.mnt); |
1da177e4 | 104 | struct dentry *mounts = dget(dentry); |
6264d69d | 105 | int err = 0; |
1da177e4 LT |
106 | |
107 | while (follow_down(&mnt,&mounts)&&d_mountpoint(mounts)); | |
108 | ||
0989a788 | 109 | exp2 = rqst_exp_get_by_name(rqstp, mnt, mounts); |
1da177e4 | 110 | if (IS_ERR(exp2)) { |
a1033be7 NB |
111 | if (PTR_ERR(exp2) != -ENOENT) |
112 | err = PTR_ERR(exp2); | |
1da177e4 LT |
113 | dput(mounts); |
114 | mntput(mnt); | |
115 | goto out; | |
116 | } | |
5d3dbbea | 117 | if ((exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { |
1da177e4 LT |
118 | /* successfully crossed mount point */ |
119 | exp_put(exp); | |
120 | *expp = exp2; | |
121 | dput(dentry); | |
122 | *dpp = mounts; | |
123 | } else { | |
5d3dbbea | 124 | exp_put(exp2); |
1da177e4 LT |
125 | dput(mounts); |
126 | } | |
127 | mntput(mnt); | |
128 | out: | |
129 | return err; | |
130 | } | |
131 | ||
6264d69d | 132 | __be32 |
6c0a654d | 133 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 134 | const char *name, unsigned int len, |
6c0a654d | 135 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
136 | { |
137 | struct svc_export *exp; | |
138 | struct dentry *dparent; | |
139 | struct dentry *dentry; | |
6264d69d AV |
140 | __be32 err; |
141 | int host_err; | |
1da177e4 LT |
142 | |
143 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
144 | ||
145 | /* Obtain dentry and export. */ | |
8837abca | 146 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
147 | if (err) |
148 | return err; | |
149 | ||
150 | dparent = fhp->fh_dentry; | |
151 | exp = fhp->fh_export; | |
152 | exp_get(exp); | |
153 | ||
1da177e4 LT |
154 | /* Lookup the name, but don't follow links */ |
155 | if (isdotent(name, len)) { | |
156 | if (len==1) | |
157 | dentry = dget(dparent); | |
54775491 | 158 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 159 | dentry = dget_parent(dparent); |
54775491 | 160 | else if (!EX_NOHIDE(exp)) |
1da177e4 LT |
161 | dentry = dget(dparent); /* .. == . just like at / */ |
162 | else { | |
163 | /* checking mountpoint crossing is very different when stepping up */ | |
164 | struct svc_export *exp2 = NULL; | |
165 | struct dentry *dp; | |
54775491 | 166 | struct vfsmount *mnt = mntget(exp->ex_path.mnt); |
1da177e4 LT |
167 | dentry = dget(dparent); |
168 | while(dentry == mnt->mnt_root && follow_up(&mnt, &dentry)) | |
169 | ; | |
170 | dp = dget_parent(dentry); | |
171 | dput(dentry); | |
172 | dentry = dp; | |
173 | ||
0989a788 | 174 | exp2 = rqst_exp_parent(rqstp, mnt, dentry); |
2d3bb252 BF |
175 | if (PTR_ERR(exp2) == -ENOENT) { |
176 | dput(dentry); | |
177 | dentry = dget(dparent); | |
178 | } else if (IS_ERR(exp2)) { | |
6264d69d | 179 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
180 | dput(dentry); |
181 | mntput(mnt); | |
182 | goto out_nfserr; | |
1da177e4 LT |
183 | } else { |
184 | exp_put(exp); | |
185 | exp = exp2; | |
186 | } | |
187 | mntput(mnt); | |
188 | } | |
189 | } else { | |
190 | fh_lock(fhp); | |
191 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 192 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
193 | if (IS_ERR(dentry)) |
194 | goto out_nfserr; | |
195 | /* | |
196 | * check if we have crossed a mount point ... | |
197 | */ | |
198 | if (d_mountpoint(dentry)) { | |
6264d69d | 199 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
200 | dput(dentry); |
201 | goto out_nfserr; | |
202 | } | |
203 | } | |
204 | } | |
6c0a654d BF |
205 | *dentry_ret = dentry; |
206 | *exp_ret = exp; | |
207 | return 0; | |
208 | ||
209 | out_nfserr: | |
210 | exp_put(exp); | |
211 | return nfserrno(host_err); | |
212 | } | |
213 | ||
214 | /* | |
215 | * Look up one component of a pathname. | |
216 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
217 | * | |
218 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
219 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
220 | * accepted as it stands and the mounted directory is | |
221 | * returned. Otherwise the covered directory is returned. | |
222 | * NOTE: this mountpoint crossing is not supported properly by all | |
223 | * clients and is explicitly disallowed for NFSv3 | |
224 | * NeilBrown <neilb@cse.unsw.edu.au> | |
225 | */ | |
226 | __be32 | |
227 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 228 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
229 | { |
230 | struct svc_export *exp; | |
231 | struct dentry *dentry; | |
232 | __be32 err; | |
233 | ||
234 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
235 | if (err) | |
236 | return err; | |
32c1eb0c AA |
237 | err = check_nfsd_access(exp, rqstp); |
238 | if (err) | |
239 | goto out; | |
1da177e4 LT |
240 | /* |
241 | * Note: we compose the file handle now, but as the | |
242 | * dentry may be negative, it may need to be updated. | |
243 | */ | |
244 | err = fh_compose(resfh, exp, dentry, fhp); | |
245 | if (!err && !dentry->d_inode) | |
246 | err = nfserr_noent; | |
32c1eb0c | 247 | out: |
1da177e4 | 248 | dput(dentry); |
1da177e4 LT |
249 | exp_put(exp); |
250 | return err; | |
1da177e4 LT |
251 | } |
252 | ||
6c0a654d | 253 | |
1da177e4 LT |
254 | /* |
255 | * Set various file attributes. | |
256 | * N.B. After this call fhp needs an fh_put | |
257 | */ | |
6264d69d | 258 | __be32 |
1da177e4 LT |
259 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
260 | int check_guard, time_t guardtime) | |
261 | { | |
262 | struct dentry *dentry; | |
263 | struct inode *inode; | |
8837abca | 264 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 265 | int ftype = 0; |
6264d69d AV |
266 | __be32 err; |
267 | int host_err; | |
1da177e4 LT |
268 | int size_change = 0; |
269 | ||
270 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 271 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
272 | if (iap->ia_valid & ATTR_SIZE) |
273 | ftype = S_IFREG; | |
274 | ||
275 | /* Get inode */ | |
276 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 277 | if (err) |
1da177e4 LT |
278 | goto out; |
279 | ||
280 | dentry = fhp->fh_dentry; | |
281 | inode = dentry->d_inode; | |
282 | ||
15b7a1b8 N |
283 | /* Ignore any mode updates on symlinks */ |
284 | if (S_ISLNK(inode->i_mode)) | |
285 | iap->ia_valid &= ~ATTR_MODE; | |
286 | ||
287 | if (!iap->ia_valid) | |
288 | goto out; | |
289 | ||
9c85fca5 CH |
290 | /* |
291 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
292 | * which only requires access, and "set-[ac]time-to-X" which |
293 | * requires ownership. | |
294 | * So if it looks like it might be "set both to the same time which | |
295 | * is close to now", and if inode_change_ok fails, then we | |
296 | * convert to "set to now" instead of "set to explicit time" | |
297 | * | |
298 | * We only call inode_change_ok as the last test as technically | |
299 | * it is not an interface that we should be using. It is only | |
300 | * valid if the filesystem does not define it's own i_op->setattr. | |
301 | */ | |
302 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
303 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
304 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
305 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
306 | /* | |
307 | * Looks probable. | |
308 | * | |
309 | * Now just make sure time is in the right ballpark. | |
310 | * Solaris, at least, doesn't seem to care what the time | |
311 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 312 | */ |
9c85fca5 CH |
313 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
314 | if (delta < 0) | |
315 | delta = -delta; | |
316 | if (delta < MAX_TOUCH_TIME_ERROR && | |
317 | inode_change_ok(inode, iap) != 0) { | |
318 | /* | |
319 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
320 | * This will cause notify_change to set these times | |
321 | * to "now" | |
322 | */ | |
323 | iap->ia_valid &= ~BOTH_TIME_SET; | |
324 | } | |
1da177e4 LT |
325 | } |
326 | ||
9c85fca5 CH |
327 | /* |
328 | * The size case is special. | |
329 | * It changes the file as well as the attributes. | |
330 | */ | |
1da177e4 LT |
331 | if (iap->ia_valid & ATTR_SIZE) { |
332 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
333 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
334 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
335 | if (err) |
336 | goto out; | |
337 | } | |
338 | ||
339 | /* | |
340 | * If we are changing the size of the file, then | |
341 | * we need to break all leases. | |
342 | */ | |
6264d69d AV |
343 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
344 | if (host_err == -EWOULDBLOCK) | |
345 | host_err = -ETIMEDOUT; | |
346 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
347 | goto out_nfserr; |
348 | ||
6264d69d AV |
349 | host_err = get_write_access(inode); |
350 | if (host_err) | |
1da177e4 LT |
351 | goto out_nfserr; |
352 | ||
353 | size_change = 1; | |
6264d69d AV |
354 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
355 | if (host_err) { | |
1da177e4 LT |
356 | put_write_access(inode); |
357 | goto out_nfserr; | |
358 | } | |
90c0af05 | 359 | vfs_dq_init(inode); |
1da177e4 LT |
360 | } |
361 | ||
ca456252 | 362 | /* sanitize the mode change */ |
1da177e4 LT |
363 | if (iap->ia_valid & ATTR_MODE) { |
364 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 365 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
366 | } |
367 | ||
368 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
369 | if (!S_ISDIR(inode->i_mode) && |
370 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
371 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
372 | iap->ia_valid |= ATTR_KILL_PRIV; |
373 | if (iap->ia_valid & ATTR_MODE) { | |
374 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 375 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
376 | if (iap->ia_mode & S_IXGRP) |
377 | iap->ia_mode &= ~S_ISGID; | |
378 | } else { | |
379 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
380 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 381 | } |
1da177e4 LT |
382 | } |
383 | ||
1da177e4 LT |
384 | /* Change the attributes. */ |
385 | ||
386 | iap->ia_valid |= ATTR_CTIME; | |
387 | ||
388 | err = nfserr_notsync; | |
389 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
390 | fh_lock(fhp); | |
6264d69d AV |
391 | host_err = notify_change(dentry, iap); |
392 | err = nfserrno(host_err); | |
1da177e4 LT |
393 | fh_unlock(fhp); |
394 | } | |
395 | if (size_change) | |
396 | put_write_access(inode); | |
397 | if (!err) | |
398 | if (EX_ISSYNC(fhp->fh_export)) | |
399 | write_inode_now(inode, 1); | |
400 | out: | |
401 | return err; | |
402 | ||
403 | out_nfserr: | |
6264d69d | 404 | err = nfserrno(host_err); |
1da177e4 LT |
405 | goto out; |
406 | } | |
407 | ||
5be196e5 CH |
408 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
409 | defined(CONFIG_NFSD_V3_ACL) || \ | |
410 | defined(CONFIG_NFSD_V4) | |
411 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
412 | { | |
413 | ssize_t buflen; | |
6c6a426f | 414 | ssize_t ret; |
5be196e5 CH |
415 | |
416 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
417 | if (buflen <= 0) | |
418 | return buflen; | |
1da177e4 | 419 | |
5be196e5 CH |
420 | *buf = kmalloc(buflen, GFP_KERNEL); |
421 | if (!*buf) | |
422 | return -ENOMEM; | |
423 | ||
6c6a426f KK |
424 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
425 | if (ret < 0) | |
426 | kfree(*buf); | |
427 | return ret; | |
5be196e5 CH |
428 | } |
429 | #endif | |
430 | ||
431 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
432 | static int |
433 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
434 | { | |
435 | int len; | |
436 | size_t buflen; | |
437 | char *buf = NULL; | |
438 | int error = 0; | |
1da177e4 LT |
439 | |
440 | buflen = posix_acl_xattr_size(pacl->a_count); | |
441 | buf = kmalloc(buflen, GFP_KERNEL); | |
442 | error = -ENOMEM; | |
443 | if (buf == NULL) | |
444 | goto out; | |
445 | ||
446 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
447 | if (len < 0) { | |
448 | error = len; | |
449 | goto out; | |
450 | } | |
451 | ||
5be196e5 | 452 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
453 | out: |
454 | kfree(buf); | |
455 | return error; | |
456 | } | |
457 | ||
6264d69d | 458 | __be32 |
1da177e4 LT |
459 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
460 | struct nfs4_acl *acl) | |
461 | { | |
6264d69d AV |
462 | __be32 error; |
463 | int host_error; | |
1da177e4 LT |
464 | struct dentry *dentry; |
465 | struct inode *inode; | |
466 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
467 | unsigned int flags = 0; | |
468 | ||
469 | /* Get inode */ | |
8837abca | 470 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 471 | if (error) |
4b2ca38a | 472 | return error; |
1da177e4 LT |
473 | |
474 | dentry = fhp->fh_dentry; | |
475 | inode = dentry->d_inode; | |
476 | if (S_ISDIR(inode->i_mode)) | |
477 | flags = NFS4_ACL_DIR; | |
478 | ||
6264d69d AV |
479 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
480 | if (host_error == -EINVAL) { | |
4b2ca38a | 481 | return nfserr_attrnotsupp; |
6264d69d | 482 | } else if (host_error < 0) |
1da177e4 LT |
483 | goto out_nfserr; |
484 | ||
6264d69d AV |
485 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
486 | if (host_error < 0) | |
4b2ca38a | 487 | goto out_release; |
1da177e4 | 488 | |
4b2ca38a | 489 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 490 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 491 | |
4b2ca38a | 492 | out_release: |
1da177e4 LT |
493 | posix_acl_release(pacl); |
494 | posix_acl_release(dpacl); | |
1da177e4 | 495 | out_nfserr: |
f34f9242 | 496 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 497 | return nfserr_attrnotsupp; |
f34f9242 | 498 | else |
4b2ca38a | 499 | return nfserrno(host_error); |
1da177e4 LT |
500 | } |
501 | ||
502 | static struct posix_acl * | |
503 | _get_posix_acl(struct dentry *dentry, char *key) | |
504 | { | |
5be196e5 | 505 | void *buf = NULL; |
1da177e4 | 506 | struct posix_acl *pacl = NULL; |
5be196e5 | 507 | int buflen; |
1da177e4 | 508 | |
5be196e5 CH |
509 | buflen = nfsd_getxattr(dentry, key, &buf); |
510 | if (!buflen) | |
511 | buflen = -ENODATA; | |
512 | if (buflen <= 0) | |
513 | return ERR_PTR(buflen); | |
1da177e4 LT |
514 | |
515 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
516 | kfree(buf); |
517 | return pacl; | |
1da177e4 LT |
518 | } |
519 | ||
520 | int | |
521 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
522 | { | |
523 | struct inode *inode = dentry->d_inode; | |
524 | int error = 0; | |
525 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
526 | unsigned int flags = 0; | |
527 | ||
9a59f452 | 528 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
529 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
530 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
531 | if (IS_ERR(pacl)) { | |
532 | error = PTR_ERR(pacl); | |
533 | pacl = NULL; | |
534 | goto out; | |
535 | } | |
536 | ||
537 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 538 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
539 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
540 | dpacl = NULL; | |
541 | else if (IS_ERR(dpacl)) { | |
542 | error = PTR_ERR(dpacl); | |
543 | dpacl = NULL; | |
544 | goto out; | |
545 | } | |
546 | flags = NFS4_ACL_DIR; | |
547 | } | |
548 | ||
549 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
550 | if (IS_ERR(*acl)) { | |
551 | error = PTR_ERR(*acl); | |
552 | *acl = NULL; | |
553 | } | |
554 | out: | |
555 | posix_acl_release(pacl); | |
556 | posix_acl_release(dpacl); | |
557 | return error; | |
558 | } | |
559 | ||
560 | #endif /* defined(CONFIG_NFS_V4) */ | |
561 | ||
562 | #ifdef CONFIG_NFSD_V3 | |
563 | /* | |
564 | * Check server access rights to a file system object | |
565 | */ | |
566 | struct accessmap { | |
567 | u32 access; | |
568 | int how; | |
569 | }; | |
570 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
571 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
572 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
573 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
574 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
575 | |
576 | { 0, 0 } | |
577 | }; | |
578 | ||
579 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
580 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
581 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
582 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
583 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
584 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
585 | |
586 | { 0, 0 } | |
587 | }; | |
588 | ||
589 | static struct accessmap nfs3_anyaccess[] = { | |
590 | /* Some clients - Solaris 2.6 at least, make an access call | |
591 | * to the server to check for access for things like /dev/null | |
592 | * (which really, the server doesn't care about). So | |
593 | * We provide simple access checking for them, looking | |
594 | * mainly at mode bits, and we make sure to ignore read-only | |
595 | * filesystem checks | |
596 | */ | |
8837abca MS |
597 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
598 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
599 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
600 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
601 | |
602 | { 0, 0 } | |
603 | }; | |
604 | ||
6264d69d | 605 | __be32 |
1da177e4 LT |
606 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
607 | { | |
608 | struct accessmap *map; | |
609 | struct svc_export *export; | |
610 | struct dentry *dentry; | |
611 | u32 query, result = 0, sresult = 0; | |
6264d69d | 612 | __be32 error; |
1da177e4 | 613 | |
8837abca | 614 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
615 | if (error) |
616 | goto out; | |
617 | ||
618 | export = fhp->fh_export; | |
619 | dentry = fhp->fh_dentry; | |
620 | ||
621 | if (S_ISREG(dentry->d_inode->i_mode)) | |
622 | map = nfs3_regaccess; | |
623 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
624 | map = nfs3_diraccess; | |
625 | else | |
626 | map = nfs3_anyaccess; | |
627 | ||
628 | ||
629 | query = *access; | |
630 | for (; map->access; map++) { | |
631 | if (map->access & query) { | |
6264d69d | 632 | __be32 err2; |
1da177e4 LT |
633 | |
634 | sresult |= map->access; | |
635 | ||
0ec757df | 636 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
637 | switch (err2) { |
638 | case nfs_ok: | |
639 | result |= map->access; | |
640 | break; | |
641 | ||
642 | /* the following error codes just mean the access was not allowed, | |
643 | * rather than an error occurred */ | |
644 | case nfserr_rofs: | |
645 | case nfserr_acces: | |
646 | case nfserr_perm: | |
647 | /* simply don't "or" in the access bit. */ | |
648 | break; | |
649 | default: | |
650 | error = err2; | |
651 | goto out; | |
652 | } | |
653 | } | |
654 | } | |
655 | *access = result; | |
656 | if (supported) | |
657 | *supported = sresult; | |
658 | ||
659 | out: | |
660 | return error; | |
661 | } | |
662 | #endif /* CONFIG_NFSD_V3 */ | |
663 | ||
664 | ||
665 | ||
666 | /* | |
667 | * Open an existing file or directory. | |
668 | * The access argument indicates the type of open (read/write/lock) | |
669 | * N.B. After this call fhp needs an fh_put | |
670 | */ | |
6264d69d | 671 | __be32 |
1da177e4 LT |
672 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
673 | int access, struct file **filp) | |
674 | { | |
745ca247 | 675 | const struct cred *cred = current_cred(); |
1da177e4 LT |
676 | struct dentry *dentry; |
677 | struct inode *inode; | |
6264d69d AV |
678 | int flags = O_RDONLY|O_LARGEFILE; |
679 | __be32 err; | |
680 | int host_err; | |
1da177e4 LT |
681 | |
682 | /* | |
683 | * If we get here, then the client has already done an "open", | |
684 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
685 | * in case a chmod has now revoked permission. | |
686 | */ | |
8837abca | 687 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
688 | if (err) |
689 | goto out; | |
690 | ||
691 | dentry = fhp->fh_dentry; | |
692 | inode = dentry->d_inode; | |
693 | ||
694 | /* Disallow write access to files with the append-only bit set | |
695 | * or any access when mandatory locking enabled | |
696 | */ | |
697 | err = nfserr_perm; | |
8837abca | 698 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 699 | goto out; |
5e7fc436 BF |
700 | /* |
701 | * We must ignore files (but only files) which might have mandatory | |
702 | * locks on them because there is no way to know if the accesser has | |
703 | * the lock. | |
704 | */ | |
705 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
706 | goto out; |
707 | ||
708 | if (!inode->i_fop) | |
709 | goto out; | |
710 | ||
711 | /* | |
712 | * Check to see if there are any leases on this file. | |
713 | * This may block while leases are broken. | |
714 | */ | |
8837abca | 715 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0)); |
6264d69d AV |
716 | if (host_err == -EWOULDBLOCK) |
717 | host_err = -ETIMEDOUT; | |
718 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
719 | goto out_nfserr; |
720 | ||
8837abca MS |
721 | if (access & NFSD_MAY_WRITE) { |
722 | if (access & NFSD_MAY_READ) | |
9ecb6a08 BF |
723 | flags = O_RDWR|O_LARGEFILE; |
724 | else | |
725 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 726 | |
90c0af05 | 727 | vfs_dq_init(inode); |
1da177e4 | 728 | } |
54775491 | 729 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
745ca247 | 730 | flags, cred); |
1da177e4 | 731 | if (IS_ERR(*filp)) |
6264d69d | 732 | host_err = PTR_ERR(*filp); |
1da177e4 | 733 | out_nfserr: |
6264d69d | 734 | err = nfserrno(host_err); |
1da177e4 LT |
735 | out: |
736 | return err; | |
737 | } | |
738 | ||
739 | /* | |
740 | * Close a file. | |
741 | */ | |
742 | void | |
743 | nfsd_close(struct file *filp) | |
744 | { | |
745 | fput(filp); | |
746 | } | |
747 | ||
9a8d248e BF |
748 | /* |
749 | * Sync a file | |
750 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
751 | * after it. | |
752 | */ | |
753 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, | |
754 | const struct file_operations *fop) | |
755 | { | |
756 | struct inode *inode = dp->d_inode; | |
757 | int (*fsync) (struct file *, struct dentry *, int); | |
758 | int err; | |
759 | ||
760 | err = filemap_fdatawrite(inode->i_mapping); | |
761 | if (err == 0 && fop && (fsync = fop->fsync)) | |
762 | err = fsync(filp, dp, 0); | |
763 | if (err == 0) | |
764 | err = filemap_fdatawait(inode->i_mapping); | |
765 | ||
766 | return err; | |
767 | } | |
768 | ||
a334de28 | 769 | static int |
1da177e4 LT |
770 | nfsd_sync(struct file *filp) |
771 | { | |
9a8d248e BF |
772 | int err; |
773 | struct inode *inode = filp->f_path.dentry->d_inode; | |
774 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
775 | mutex_lock(&inode->i_mutex); | |
776 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); | |
777 | mutex_unlock(&inode->i_mutex); | |
778 | ||
779 | return err; | |
1da177e4 LT |
780 | } |
781 | ||
f193fbab | 782 | int |
9a8d248e | 783 | nfsd_sync_dir(struct dentry *dp) |
1da177e4 | 784 | { |
9a8d248e | 785 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
786 | } |
787 | ||
788 | /* | |
789 | * Obtain the readahead parameters for the file | |
790 | * specified by (dev, ino). | |
791 | */ | |
1da177e4 LT |
792 | |
793 | static inline struct raparms * | |
794 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
795 | { | |
796 | struct raparms *ra, **rap, **frap = NULL; | |
797 | int depth = 0; | |
fce1456a GB |
798 | unsigned int hash; |
799 | struct raparm_hbucket *rab; | |
800 | ||
801 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
802 | rab = &raparm_hash[hash]; | |
1da177e4 | 803 | |
fce1456a GB |
804 | spin_lock(&rab->pb_lock); |
805 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
806 | if (ra->p_ino == ino && ra->p_dev == dev) |
807 | goto found; | |
808 | depth++; | |
809 | if (ra->p_count == 0) | |
810 | frap = rap; | |
811 | } | |
812 | depth = nfsdstats.ra_size*11/10; | |
813 | if (!frap) { | |
fce1456a | 814 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
815 | return NULL; |
816 | } | |
817 | rap = frap; | |
818 | ra = *frap; | |
819 | ra->p_dev = dev; | |
820 | ra->p_ino = ino; | |
821 | ra->p_set = 0; | |
fce1456a | 822 | ra->p_hindex = hash; |
1da177e4 | 823 | found: |
fce1456a | 824 | if (rap != &rab->pb_head) { |
1da177e4 | 825 | *rap = ra->p_next; |
fce1456a GB |
826 | ra->p_next = rab->pb_head; |
827 | rab->pb_head = ra; | |
1da177e4 LT |
828 | } |
829 | ra->p_count++; | |
830 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 831 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
832 | return ra; |
833 | } | |
834 | ||
835 | /* | |
cf8208d0 JA |
836 | * Grab and keep cached pages associated with a file in the svc_rqst |
837 | * so that they can be passed to the network sendmsg/sendpage routines | |
838 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
839 | */ |
840 | static int | |
cf8208d0 JA |
841 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
842 | struct splice_desc *sd) | |
1da177e4 | 843 | { |
cf8208d0 | 844 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 845 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
846 | struct page *page = buf->page; |
847 | size_t size; | |
848 | int ret; | |
1da177e4 | 849 | |
cac36bb0 | 850 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
851 | if (unlikely(ret)) |
852 | return ret; | |
853 | ||
854 | size = sd->len; | |
1da177e4 LT |
855 | |
856 | if (rqstp->rq_res.page_len == 0) { | |
857 | get_page(page); | |
44524359 N |
858 | put_page(*pp); |
859 | *pp = page; | |
860 | rqstp->rq_resused++; | |
cf8208d0 | 861 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 862 | rqstp->rq_res.page_len = size; |
44524359 | 863 | } else if (page != pp[-1]) { |
1da177e4 | 864 | get_page(page); |
250f3915 N |
865 | if (*pp) |
866 | put_page(*pp); | |
44524359 N |
867 | *pp = page; |
868 | rqstp->rq_resused++; | |
1da177e4 | 869 | rqstp->rq_res.page_len += size; |
44524359 | 870 | } else |
1da177e4 | 871 | rqstp->rq_res.page_len += size; |
1da177e4 | 872 | |
1da177e4 LT |
873 | return size; |
874 | } | |
875 | ||
cf8208d0 JA |
876 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
877 | struct splice_desc *sd) | |
878 | { | |
879 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
880 | } | |
881 | ||
a8754bee DH |
882 | static inline int svc_msnfs(struct svc_fh *ffhp) |
883 | { | |
884 | #ifdef MSNFS | |
885 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
886 | #else | |
887 | return 0; | |
888 | #endif | |
889 | } | |
890 | ||
6264d69d | 891 | static __be32 |
1da177e4 LT |
892 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
893 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
894 | { | |
895 | struct inode *inode; | |
896 | struct raparms *ra; | |
897 | mm_segment_t oldfs; | |
6264d69d AV |
898 | __be32 err; |
899 | int host_err; | |
1da177e4 LT |
900 | |
901 | err = nfserr_perm; | |
7eaa36e2 | 902 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
903 | |
904 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 905 | goto out; |
1da177e4 LT |
906 | |
907 | /* Get readahead parameters */ | |
908 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
909 | ||
910 | if (ra && ra->p_set) | |
911 | file->f_ra = ra->p_ra; | |
912 | ||
cf8208d0 JA |
913 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
914 | struct splice_desc sd = { | |
915 | .len = 0, | |
916 | .total_len = *count, | |
917 | .pos = offset, | |
918 | .u.data = rqstp, | |
919 | }; | |
920 | ||
4fbef206 | 921 | rqstp->rq_resused = 1; |
cf8208d0 | 922 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
923 | } else { |
924 | oldfs = get_fs(); | |
925 | set_fs(KERNEL_DS); | |
6264d69d | 926 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
927 | set_fs(oldfs); |
928 | } | |
929 | ||
930 | /* Write back readahead params */ | |
931 | if (ra) { | |
fce1456a GB |
932 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
933 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
934 | ra->p_ra = file->f_ra; |
935 | ra->p_set = 1; | |
936 | ra->p_count--; | |
fce1456a | 937 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
938 | } |
939 | ||
6264d69d AV |
940 | if (host_err >= 0) { |
941 | nfsdstats.io_read += host_err; | |
942 | *count = host_err; | |
1da177e4 | 943 | err = 0; |
7eaa36e2 | 944 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 945 | } else |
6264d69d | 946 | err = nfserrno(host_err); |
1da177e4 LT |
947 | out: |
948 | return err; | |
949 | } | |
950 | ||
9f708e40 NB |
951 | static void kill_suid(struct dentry *dentry) |
952 | { | |
953 | struct iattr ia; | |
b5376771 | 954 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 955 | |
1b1dcc1b | 956 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 957 | notify_change(dentry, &ia); |
1b1dcc1b | 958 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
959 | } |
960 | ||
6264d69d | 961 | static __be32 |
1da177e4 LT |
962 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
963 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 964 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
965 | { |
966 | struct svc_export *exp; | |
967 | struct dentry *dentry; | |
968 | struct inode *inode; | |
969 | mm_segment_t oldfs; | |
6264d69d AV |
970 | __be32 err = 0; |
971 | int host_err; | |
1da177e4 LT |
972 | int stable = *stablep; |
973 | ||
45bd3b3d | 974 | #ifdef MSNFS |
1da177e4 LT |
975 | err = nfserr_perm; |
976 | ||
1da177e4 | 977 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
31dec253 | 978 | (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt))) |
1da177e4 LT |
979 | goto out; |
980 | #endif | |
981 | ||
7eaa36e2 | 982 | dentry = file->f_path.dentry; |
1da177e4 LT |
983 | inode = dentry->d_inode; |
984 | exp = fhp->fh_export; | |
985 | ||
986 | /* | |
987 | * Request sync writes if | |
988 | * - the sync export option has been set, or | |
989 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
990 | * - The file system doesn't support fsync(). | |
991 | * When gathered writes have been configured for this volume, | |
992 | * flushing the data to disk is handled separately below. | |
993 | */ | |
994 | ||
3ba15148 | 995 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
996 | stable = 2; |
997 | *stablep = 2; /* FILE_SYNC */ | |
998 | } | |
999 | ||
1000 | if (!EX_ISSYNC(exp)) | |
1001 | stable = 0; | |
db1dd4d3 JC |
1002 | if (stable && !EX_WGATHER(exp)) { |
1003 | spin_lock(&file->f_lock); | |
1da177e4 | 1004 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
1005 | spin_unlock(&file->f_lock); |
1006 | } | |
1da177e4 LT |
1007 | |
1008 | /* Write the data. */ | |
1009 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1010 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1011 | set_fs(oldfs); |
6264d69d | 1012 | if (host_err >= 0) { |
31dec253 | 1013 | nfsdstats.io_write += host_err; |
7eaa36e2 | 1014 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
1015 | } |
1016 | ||
1017 | /* clear setuid/setgid flag after write */ | |
6264d69d | 1018 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 1019 | kill_suid(dentry); |
1da177e4 | 1020 | |
6264d69d | 1021 | if (host_err >= 0 && stable) { |
1da177e4 LT |
1022 | static ino_t last_ino; |
1023 | static dev_t last_dev; | |
1024 | ||
1025 | /* | |
1026 | * Gathered writes: If another process is currently | |
1027 | * writing to the file, there's a high chance | |
1028 | * this is another nfsd (triggered by a bulk write | |
1029 | * from a client's biod). Rather than syncing the | |
1030 | * file with each write request, we sleep for 10 msec. | |
1031 | * | |
1032 | * I don't know if this roughly approximates | |
1033 | * C. Juszak's idea of gathered writes, but it's a | |
1034 | * nice and simple solution (IMHO), and it seems to | |
1035 | * work:-) | |
1036 | */ | |
1037 | if (EX_WGATHER(exp)) { | |
1038 | if (atomic_read(&inode->i_writecount) > 1 | |
1039 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
ba25f9dc | 1040 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); |
1da177e4 | 1041 | msleep(10); |
ba25f9dc | 1042 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); |
1da177e4 LT |
1043 | } |
1044 | ||
1045 | if (inode->i_state & I_DIRTY) { | |
ba25f9dc | 1046 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); |
6264d69d | 1047 | host_err=nfsd_sync(file); |
1da177e4 LT |
1048 | } |
1049 | #if 0 | |
1050 | wake_up(&inode->i_wait); | |
1051 | #endif | |
1052 | } | |
1053 | last_ino = inode->i_ino; | |
1054 | last_dev = inode->i_sb->s_dev; | |
1055 | } | |
1056 | ||
6264d69d | 1057 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
31dec253 | 1058 | if (host_err >= 0) { |
1da177e4 | 1059 | err = 0; |
31dec253 DS |
1060 | *cnt = host_err; |
1061 | } else | |
6264d69d | 1062 | err = nfserrno(host_err); |
1da177e4 LT |
1063 | out: |
1064 | return err; | |
1065 | } | |
1066 | ||
1067 | /* | |
1068 | * Read data from a file. count must contain the requested read count | |
1069 | * on entry. On return, *count contains the number of bytes actually read. | |
1070 | * N.B. After this call fhp needs an fh_put | |
1071 | */ | |
6264d69d | 1072 | __be32 |
1da177e4 LT |
1073 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1074 | loff_t offset, struct kvec *vec, int vlen, | |
1075 | unsigned long *count) | |
1076 | { | |
6264d69d | 1077 | __be32 err; |
1da177e4 LT |
1078 | |
1079 | if (file) { | |
0ec757df | 1080 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1081 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1082 | if (err) |
1083 | goto out; | |
1084 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1085 | } else { | |
8837abca | 1086 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); |
1da177e4 LT |
1087 | if (err) |
1088 | goto out; | |
1089 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1090 | nfsd_close(file); | |
1091 | } | |
1092 | out: | |
1093 | return err; | |
1094 | } | |
1095 | ||
1096 | /* | |
1097 | * Write data to a file. | |
1098 | * The stable flag requests synchronous writes. | |
1099 | * N.B. After this call fhp needs an fh_put | |
1100 | */ | |
6264d69d | 1101 | __be32 |
1da177e4 | 1102 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1103 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1104 | int *stablep) |
1105 | { | |
6264d69d | 1106 | __be32 err = 0; |
1da177e4 LT |
1107 | |
1108 | if (file) { | |
0ec757df | 1109 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1110 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1111 | if (err) |
1112 | goto out; | |
1113 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1114 | stablep); | |
1115 | } else { | |
8837abca | 1116 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1117 | if (err) |
1118 | goto out; | |
1119 | ||
1120 | if (cnt) | |
1121 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1122 | cnt, stablep); | |
1123 | nfsd_close(file); | |
1124 | } | |
1125 | out: | |
1126 | return err; | |
1127 | } | |
1128 | ||
1129 | #ifdef CONFIG_NFSD_V3 | |
1130 | /* | |
1131 | * Commit all pending writes to stable storage. | |
1132 | * Strictly speaking, we could sync just the indicated file region here, | |
1133 | * but there's currently no way we can ask the VFS to do so. | |
1134 | * | |
1135 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1136 | * data to the client, as locking happens lower down in the filesystem. | |
1137 | */ | |
6264d69d | 1138 | __be32 |
1da177e4 LT |
1139 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1140 | loff_t offset, unsigned long count) | |
1141 | { | |
1142 | struct file *file; | |
6264d69d | 1143 | __be32 err; |
1da177e4 LT |
1144 | |
1145 | if ((u64)count > ~(u64)offset) | |
1146 | return nfserr_inval; | |
1147 | ||
8837abca MS |
1148 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1149 | if (err) | |
1da177e4 LT |
1150 | return err; |
1151 | if (EX_ISSYNC(fhp->fh_export)) { | |
1152 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1153 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1154 | } else { |
1155 | err = nfserr_notsupp; | |
1156 | } | |
1157 | } | |
1158 | ||
1159 | nfsd_close(file); | |
1160 | return err; | |
1161 | } | |
1162 | #endif /* CONFIG_NFSD_V3 */ | |
1163 | ||
f2b0dee2 | 1164 | static __be32 |
5c002b3b BF |
1165 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1166 | struct iattr *iap) | |
1167 | { | |
1168 | /* | |
1169 | * Mode has already been set earlier in create: | |
1170 | */ | |
1171 | iap->ia_valid &= ~ATTR_MODE; | |
1172 | /* | |
1173 | * Setting uid/gid works only for root. Irix appears to | |
1174 | * send along the gid on create when it tries to implement | |
1175 | * setgid directories via NFS: | |
1176 | */ | |
5cc0a840 | 1177 | if (current_fsuid() != 0) |
5c002b3b BF |
1178 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1179 | if (iap->ia_valid) | |
1180 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1181 | return 0; | |
1182 | } | |
1183 | ||
4ac35c2f | 1184 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1185 | * setting size to 0 may fail for some specific file systems by the permission | |
1186 | * checking which requires WRITE permission but the mode is 000. | |
1187 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1188 | * 0 after file created. | |
1189 | * | |
1190 | * call this only after vfs_create() is called. | |
1191 | * */ | |
1192 | static void | |
1193 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1194 | { | |
1195 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1196 | iap->ia_valid &= ~ATTR_SIZE; | |
1197 | } | |
1198 | ||
1da177e4 LT |
1199 | /* |
1200 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1201 | * not yet implemented. | |
1202 | * If the response fh has been verified, the parent directory should | |
1203 | * already be locked. Note that the parent directory is left locked. | |
1204 | * | |
1205 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1206 | */ | |
6264d69d | 1207 | __be32 |
1da177e4 LT |
1208 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1209 | char *fname, int flen, struct iattr *iap, | |
1210 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1211 | { | |
1212 | struct dentry *dentry, *dchild = NULL; | |
1213 | struct inode *dirp; | |
6264d69d | 1214 | __be32 err; |
5c002b3b | 1215 | __be32 err2; |
6264d69d | 1216 | int host_err; |
1da177e4 LT |
1217 | |
1218 | err = nfserr_perm; | |
1219 | if (!flen) | |
1220 | goto out; | |
1221 | err = nfserr_exist; | |
1222 | if (isdotent(fname, flen)) | |
1223 | goto out; | |
1224 | ||
8837abca | 1225 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1226 | if (err) |
1227 | goto out; | |
1228 | ||
1229 | dentry = fhp->fh_dentry; | |
1230 | dirp = dentry->d_inode; | |
1231 | ||
1232 | err = nfserr_notdir; | |
acfa4380 | 1233 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1234 | goto out; |
1235 | /* | |
1236 | * Check whether the response file handle has been verified yet. | |
1237 | * If it has, the parent directory should already be locked. | |
1238 | */ | |
1239 | if (!resfhp->fh_dentry) { | |
1240 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1241 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1242 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1243 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1244 | if (IS_ERR(dchild)) |
1245 | goto out_nfserr; | |
1246 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1247 | if (err) | |
1248 | goto out; | |
1249 | } else { | |
1250 | /* called from nfsd_proc_create */ | |
1251 | dchild = dget(resfhp->fh_dentry); | |
1252 | if (!fhp->fh_locked) { | |
1253 | /* not actually possible */ | |
1254 | printk(KERN_ERR | |
1255 | "nfsd_create: parent %s/%s not locked!\n", | |
1256 | dentry->d_parent->d_name.name, | |
1257 | dentry->d_name.name); | |
d75f2b9f | 1258 | err = nfserr_io; |
1da177e4 LT |
1259 | goto out; |
1260 | } | |
1261 | } | |
1262 | /* | |
1263 | * Make sure the child dentry is still negative ... | |
1264 | */ | |
1265 | err = nfserr_exist; | |
1266 | if (dchild->d_inode) { | |
1267 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1268 | dentry->d_name.name, dchild->d_name.name); | |
1269 | goto out; | |
1270 | } | |
1271 | ||
1272 | if (!(iap->ia_valid & ATTR_MODE)) | |
1273 | iap->ia_mode = 0; | |
1274 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1275 | ||
07cad1d2 MS |
1276 | err = nfserr_inval; |
1277 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1278 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1279 | type); | |
1280 | goto out; | |
1281 | } | |
1282 | ||
1283 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1284 | if (host_err) | |
1285 | goto out_nfserr; | |
1286 | ||
1da177e4 LT |
1287 | /* |
1288 | * Get the dir op function pointer. | |
1289 | */ | |
088406bc | 1290 | err = 0; |
1da177e4 LT |
1291 | switch (type) { |
1292 | case S_IFREG: | |
6264d69d | 1293 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1294 | if (!host_err) |
1295 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1296 | break; |
1297 | case S_IFDIR: | |
6264d69d | 1298 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1299 | break; |
1300 | case S_IFCHR: | |
1301 | case S_IFBLK: | |
1302 | case S_IFIFO: | |
1303 | case S_IFSOCK: | |
6264d69d | 1304 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1305 | break; |
1da177e4 | 1306 | } |
463c3197 DH |
1307 | if (host_err < 0) { |
1308 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1309 | goto out_nfserr; |
463c3197 | 1310 | } |
1da177e4 LT |
1311 | |
1312 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1313 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1314 | write_inode_now(dchild->d_inode, 1); |
1315 | } | |
1316 | ||
5c002b3b BF |
1317 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1318 | if (err2) | |
1319 | err = err2; | |
463c3197 | 1320 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1321 | /* |
1322 | * Update the file handle to get the new inode info. | |
1323 | */ | |
1324 | if (!err) | |
1325 | err = fh_update(resfhp); | |
1326 | out: | |
1327 | if (dchild && !IS_ERR(dchild)) | |
1328 | dput(dchild); | |
1329 | return err; | |
1330 | ||
1331 | out_nfserr: | |
6264d69d | 1332 | err = nfserrno(host_err); |
1da177e4 LT |
1333 | goto out; |
1334 | } | |
1335 | ||
1336 | #ifdef CONFIG_NFSD_V3 | |
1337 | /* | |
1338 | * NFSv3 version of nfsd_create | |
1339 | */ | |
6264d69d | 1340 | __be32 |
1da177e4 LT |
1341 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1342 | char *fname, int flen, struct iattr *iap, | |
1343 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1344 | int *truncp, int *created) |
1da177e4 LT |
1345 | { |
1346 | struct dentry *dentry, *dchild = NULL; | |
1347 | struct inode *dirp; | |
6264d69d | 1348 | __be32 err; |
5c002b3b | 1349 | __be32 err2; |
6264d69d | 1350 | int host_err; |
1da177e4 | 1351 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1352 | |
1353 | err = nfserr_perm; | |
1354 | if (!flen) | |
1355 | goto out; | |
1356 | err = nfserr_exist; | |
1357 | if (isdotent(fname, flen)) | |
1358 | goto out; | |
1359 | if (!(iap->ia_valid & ATTR_MODE)) | |
1360 | iap->ia_mode = 0; | |
8837abca | 1361 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1362 | if (err) |
1363 | goto out; | |
1364 | ||
1365 | dentry = fhp->fh_dentry; | |
1366 | dirp = dentry->d_inode; | |
1367 | ||
1368 | /* Get all the sanity checks out of the way before | |
1369 | * we lock the parent. */ | |
1370 | err = nfserr_notdir; | |
acfa4380 | 1371 | if (!dirp->i_op->lookup) |
1da177e4 | 1372 | goto out; |
12fd3520 | 1373 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1374 | |
1375 | /* | |
1376 | * Compose the response file handle. | |
1377 | */ | |
1378 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1379 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1380 | if (IS_ERR(dchild)) |
1381 | goto out_nfserr; | |
1382 | ||
1383 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1384 | if (err) | |
1385 | goto out; | |
1386 | ||
1387 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1388 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1389 | * the high bit set, so just clear the high bits. If this is |
1390 | * ever changed to use different attrs for storing the | |
1391 | * verifier, then do_open_lookup() will also need to be fixed | |
1392 | * accordingly. | |
1da177e4 LT |
1393 | */ |
1394 | v_mtime = verifier[0]&0x7fffffff; | |
1395 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1396 | } |
1397 | ||
463c3197 DH |
1398 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1399 | if (host_err) | |
1400 | goto out_nfserr; | |
1da177e4 LT |
1401 | if (dchild->d_inode) { |
1402 | err = 0; | |
1403 | ||
1404 | switch (createmode) { | |
1405 | case NFS3_CREATE_UNCHECKED: | |
1406 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1407 | err = nfserr_exist; | |
1408 | else if (truncp) { | |
1409 | /* in nfsv4, we need to treat this case a little | |
1410 | * differently. we don't want to truncate the | |
1411 | * file now; this would be wrong if the OPEN | |
1412 | * fails for some other reason. furthermore, | |
1413 | * if the size is nonzero, we should ignore it | |
1414 | * according to spec! | |
1415 | */ | |
1416 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1417 | } | |
1418 | else { | |
1419 | iap->ia_valid &= ATTR_SIZE; | |
1420 | goto set_attr; | |
1421 | } | |
1422 | break; | |
1423 | case NFS3_CREATE_EXCLUSIVE: | |
1424 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1425 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1426 | && dchild->d_inode->i_size == 0 ) |
1427 | break; | |
1428 | /* fallthru */ | |
1429 | case NFS3_CREATE_GUARDED: | |
1430 | err = nfserr_exist; | |
1431 | } | |
463c3197 | 1432 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1433 | goto out; |
1434 | } | |
1435 | ||
6264d69d | 1436 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1437 | if (host_err < 0) { |
1438 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1439 | goto out_nfserr; |
463c3197 | 1440 | } |
81ac95c5 BF |
1441 | if (created) |
1442 | *created = 1; | |
1da177e4 LT |
1443 | |
1444 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1445 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1446 | /* setattr will sync the child (or not) */ |
1447 | } | |
1448 | ||
4ac35c2f | 1449 | nfsd_check_ignore_resizing(iap); |
1450 | ||
1da177e4 | 1451 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1452 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1453 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1454 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1455 | /* XXX someone who knows this better please fix it for nsec */ |
1456 | iap->ia_mtime.tv_sec = v_mtime; | |
1457 | iap->ia_atime.tv_sec = v_atime; | |
1458 | iap->ia_mtime.tv_nsec = 0; | |
1459 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1460 | } |
1461 | ||
1da177e4 | 1462 | set_attr: |
5c002b3b BF |
1463 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1464 | if (err2) | |
1465 | err = err2; | |
f193fbab | 1466 | |
463c3197 | 1467 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1468 | /* |
1469 | * Update the filehandle to get the new inode info. | |
1470 | */ | |
1471 | if (!err) | |
1472 | err = fh_update(resfhp); | |
1da177e4 LT |
1473 | |
1474 | out: | |
1475 | fh_unlock(fhp); | |
1476 | if (dchild && !IS_ERR(dchild)) | |
1477 | dput(dchild); | |
1478 | return err; | |
1479 | ||
1480 | out_nfserr: | |
6264d69d | 1481 | err = nfserrno(host_err); |
1da177e4 LT |
1482 | goto out; |
1483 | } | |
1484 | #endif /* CONFIG_NFSD_V3 */ | |
1485 | ||
1486 | /* | |
1487 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1488 | * fits into the buffer. On return, it contains the true length. | |
1489 | * N.B. After this call fhp needs an fh_put | |
1490 | */ | |
6264d69d | 1491 | __be32 |
1da177e4 LT |
1492 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1493 | { | |
1494 | struct dentry *dentry; | |
1495 | struct inode *inode; | |
1496 | mm_segment_t oldfs; | |
6264d69d AV |
1497 | __be32 err; |
1498 | int host_err; | |
1da177e4 | 1499 | |
8837abca | 1500 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1501 | if (err) |
1502 | goto out; | |
1503 | ||
1504 | dentry = fhp->fh_dentry; | |
1505 | inode = dentry->d_inode; | |
1506 | ||
1507 | err = nfserr_inval; | |
acfa4380 | 1508 | if (!inode->i_op->readlink) |
1da177e4 LT |
1509 | goto out; |
1510 | ||
54775491 | 1511 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1512 | /* N.B. Why does this call need a get_fs()?? |
1513 | * Remove the set_fs and watch the fireworks:-) --okir | |
1514 | */ | |
1515 | ||
1516 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1517 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1518 | set_fs(oldfs); |
1519 | ||
6264d69d | 1520 | if (host_err < 0) |
1da177e4 | 1521 | goto out_nfserr; |
6264d69d | 1522 | *lenp = host_err; |
1da177e4 LT |
1523 | err = 0; |
1524 | out: | |
1525 | return err; | |
1526 | ||
1527 | out_nfserr: | |
6264d69d | 1528 | err = nfserrno(host_err); |
1da177e4 LT |
1529 | goto out; |
1530 | } | |
1531 | ||
1532 | /* | |
1533 | * Create a symlink and look up its inode | |
1534 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1535 | */ | |
6264d69d | 1536 | __be32 |
1da177e4 LT |
1537 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1538 | char *fname, int flen, | |
1539 | char *path, int plen, | |
1540 | struct svc_fh *resfhp, | |
1541 | struct iattr *iap) | |
1542 | { | |
1543 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1544 | __be32 err, cerr; |
1545 | int host_err; | |
1da177e4 LT |
1546 | |
1547 | err = nfserr_noent; | |
1548 | if (!flen || !plen) | |
1549 | goto out; | |
1550 | err = nfserr_exist; | |
1551 | if (isdotent(fname, flen)) | |
1552 | goto out; | |
1553 | ||
8837abca | 1554 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1555 | if (err) |
1556 | goto out; | |
1557 | fh_lock(fhp); | |
1558 | dentry = fhp->fh_dentry; | |
1559 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1560 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1561 | if (IS_ERR(dnew)) |
1562 | goto out_nfserr; | |
1563 | ||
75c3f29d DH |
1564 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1565 | if (host_err) | |
1566 | goto out_nfserr; | |
1567 | ||
1da177e4 LT |
1568 | if (unlikely(path[plen] != 0)) { |
1569 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1570 | if (path_alloced == NULL) | |
6264d69d | 1571 | host_err = -ENOMEM; |
1da177e4 LT |
1572 | else { |
1573 | strncpy(path_alloced, path, plen); | |
1574 | path_alloced[plen] = 0; | |
db2e747b | 1575 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1576 | kfree(path_alloced); |
1577 | } | |
1578 | } else | |
db2e747b | 1579 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
1da177e4 | 1580 | |
6264d69d | 1581 | if (!host_err) { |
1da177e4 | 1582 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1583 | host_err = nfsd_sync_dir(dentry); |
1584 | } | |
1585 | err = nfserrno(host_err); | |
1da177e4 LT |
1586 | fh_unlock(fhp); |
1587 | ||
75c3f29d DH |
1588 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1589 | ||
1da177e4 LT |
1590 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1591 | dput(dnew); | |
1592 | if (err==0) err = cerr; | |
1593 | out: | |
1594 | return err; | |
1595 | ||
1596 | out_nfserr: | |
6264d69d | 1597 | err = nfserrno(host_err); |
1da177e4 LT |
1598 | goto out; |
1599 | } | |
1600 | ||
1601 | /* | |
1602 | * Create a hardlink | |
1603 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1604 | */ | |
6264d69d | 1605 | __be32 |
1da177e4 LT |
1606 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1607 | char *name, int len, struct svc_fh *tfhp) | |
1608 | { | |
1609 | struct dentry *ddir, *dnew, *dold; | |
1610 | struct inode *dirp, *dest; | |
6264d69d AV |
1611 | __be32 err; |
1612 | int host_err; | |
1da177e4 | 1613 | |
8837abca | 1614 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1615 | if (err) |
1616 | goto out; | |
8837abca | 1617 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1618 | if (err) |
1619 | goto out; | |
1620 | ||
1621 | err = nfserr_perm; | |
1622 | if (!len) | |
1623 | goto out; | |
1624 | err = nfserr_exist; | |
1625 | if (isdotent(name, len)) | |
1626 | goto out; | |
1627 | ||
12fd3520 | 1628 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1629 | ddir = ffhp->fh_dentry; |
1630 | dirp = ddir->d_inode; | |
1631 | ||
1632 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1633 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1634 | if (IS_ERR(dnew)) |
1635 | goto out_nfserr; | |
1636 | ||
1637 | dold = tfhp->fh_dentry; | |
1638 | dest = dold->d_inode; | |
1639 | ||
75c3f29d DH |
1640 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1641 | if (host_err) { | |
1642 | err = nfserrno(host_err); | |
1643 | goto out_dput; | |
1644 | } | |
6264d69d AV |
1645 | host_err = vfs_link(dold, dirp, dnew); |
1646 | if (!host_err) { | |
1da177e4 | 1647 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1648 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1649 | write_inode_now(dest, 1); |
1650 | } | |
6264d69d | 1651 | err = 0; |
1da177e4 | 1652 | } else { |
6264d69d | 1653 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1654 | err = nfserr_acces; |
1655 | else | |
6264d69d | 1656 | err = nfserrno(host_err); |
1da177e4 | 1657 | } |
75c3f29d DH |
1658 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1659 | out_dput: | |
1da177e4 | 1660 | dput(dnew); |
270d56e5 DR |
1661 | out_unlock: |
1662 | fh_unlock(ffhp); | |
1da177e4 LT |
1663 | out: |
1664 | return err; | |
1665 | ||
1666 | out_nfserr: | |
6264d69d | 1667 | err = nfserrno(host_err); |
270d56e5 | 1668 | goto out_unlock; |
1da177e4 LT |
1669 | } |
1670 | ||
1671 | /* | |
1672 | * Rename a file | |
1673 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1674 | */ | |
6264d69d | 1675 | __be32 |
1da177e4 LT |
1676 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1677 | struct svc_fh *tfhp, char *tname, int tlen) | |
1678 | { | |
1679 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1680 | struct inode *fdir, *tdir; | |
6264d69d AV |
1681 | __be32 err; |
1682 | int host_err; | |
1da177e4 | 1683 | |
8837abca | 1684 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1685 | if (err) |
1686 | goto out; | |
8837abca | 1687 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1688 | if (err) |
1689 | goto out; | |
1690 | ||
1691 | fdentry = ffhp->fh_dentry; | |
1692 | fdir = fdentry->d_inode; | |
1693 | ||
1694 | tdentry = tfhp->fh_dentry; | |
1695 | tdir = tdentry->d_inode; | |
1696 | ||
1697 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1698 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1699 | goto out; |
1700 | ||
1701 | err = nfserr_perm; | |
1702 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1703 | goto out; | |
1704 | ||
1705 | /* cannot use fh_lock as we need deadlock protective ordering | |
1706 | * so do it by hand */ | |
1707 | trap = lock_rename(tdentry, fdentry); | |
1708 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1709 | fill_pre_wcc(ffhp); | |
1710 | fill_pre_wcc(tfhp); | |
1711 | ||
1712 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1713 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1714 | if (IS_ERR(odentry)) |
1715 | goto out_nfserr; | |
1716 | ||
6264d69d | 1717 | host_err = -ENOENT; |
1da177e4 LT |
1718 | if (!odentry->d_inode) |
1719 | goto out_dput_old; | |
6264d69d | 1720 | host_err = -EINVAL; |
1da177e4 LT |
1721 | if (odentry == trap) |
1722 | goto out_dput_old; | |
1723 | ||
1724 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1725 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1726 | if (IS_ERR(ndentry)) |
1727 | goto out_dput_old; | |
6264d69d | 1728 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1729 | if (ndentry == trap) |
1730 | goto out_dput_new; | |
1731 | ||
9079b1eb | 1732 | if (svc_msnfs(ffhp) && |
1da177e4 LT |
1733 | ((atomic_read(&odentry->d_count) > 1) |
1734 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1735 | host_err = -EPERM; |
9079b1eb DH |
1736 | goto out_dput_new; |
1737 | } | |
1738 | ||
1739 | host_err = -EXDEV; | |
1740 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1741 | goto out_dput_new; | |
1742 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1743 | if (host_err) | |
1744 | goto out_dput_new; | |
1745 | ||
6264d69d AV |
1746 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1747 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1748 | host_err = nfsd_sync_dir(tdentry); | |
1749 | if (!host_err) | |
1750 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1751 | } |
1752 | ||
9079b1eb DH |
1753 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1754 | ||
1da177e4 LT |
1755 | out_dput_new: |
1756 | dput(ndentry); | |
1757 | out_dput_old: | |
1758 | dput(odentry); | |
1759 | out_nfserr: | |
6264d69d | 1760 | err = nfserrno(host_err); |
1da177e4 LT |
1761 | |
1762 | /* we cannot reply on fh_unlock on the two filehandles, | |
1763 | * as that would do the wrong thing if the two directories | |
1764 | * were the same, so again we do it by hand | |
1765 | */ | |
1766 | fill_post_wcc(ffhp); | |
1767 | fill_post_wcc(tfhp); | |
1768 | unlock_rename(tdentry, fdentry); | |
1769 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1770 | ||
1771 | out: | |
1772 | return err; | |
1773 | } | |
1774 | ||
1775 | /* | |
1776 | * Unlink a file or directory | |
1777 | * N.B. After this call fhp needs an fh_put | |
1778 | */ | |
6264d69d | 1779 | __be32 |
1da177e4 LT |
1780 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1781 | char *fname, int flen) | |
1782 | { | |
1783 | struct dentry *dentry, *rdentry; | |
1784 | struct inode *dirp; | |
6264d69d AV |
1785 | __be32 err; |
1786 | int host_err; | |
1da177e4 LT |
1787 | |
1788 | err = nfserr_acces; | |
1789 | if (!flen || isdotent(fname, flen)) | |
1790 | goto out; | |
8837abca | 1791 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1792 | if (err) |
1793 | goto out; | |
1794 | ||
12fd3520 | 1795 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1796 | dentry = fhp->fh_dentry; |
1797 | dirp = dentry->d_inode; | |
1798 | ||
1799 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1800 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1801 | if (IS_ERR(rdentry)) |
1802 | goto out_nfserr; | |
1803 | ||
1804 | if (!rdentry->d_inode) { | |
1805 | dput(rdentry); | |
1806 | err = nfserr_noent; | |
1807 | goto out; | |
1808 | } | |
1809 | ||
1810 | if (!type) | |
1811 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1812 | ||
0622753b DH |
1813 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1814 | if (host_err) | |
1815 | goto out_nfserr; | |
1816 | ||
1da177e4 LT |
1817 | if (type != S_IFDIR) { /* It's UNLINK */ |
1818 | #ifdef MSNFS | |
1819 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1820 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1821 | host_err = -EPERM; |
1da177e4 LT |
1822 | } else |
1823 | #endif | |
6264d69d | 1824 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1825 | } else { /* It's RMDIR */ |
6264d69d | 1826 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1827 | } |
1828 | ||
1829 | dput(rdentry); | |
1830 | ||
6264d69d | 1831 | if (host_err) |
0622753b | 1832 | goto out_drop; |
6264d69d AV |
1833 | if (EX_ISSYNC(fhp->fh_export)) |
1834 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 | 1835 | |
0622753b DH |
1836 | out_drop: |
1837 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1838 | out_nfserr: |
6264d69d | 1839 | err = nfserrno(host_err); |
f193fbab YT |
1840 | out: |
1841 | return err; | |
1da177e4 LT |
1842 | } |
1843 | ||
14f7dd63 DW |
1844 | /* |
1845 | * We do this buffering because we must not call back into the file | |
1846 | * system's ->lookup() method from the filldir callback. That may well | |
1847 | * deadlock a number of file systems. | |
1848 | * | |
1849 | * This is based heavily on the implementation of same in XFS. | |
1850 | */ | |
1851 | struct buffered_dirent { | |
1852 | u64 ino; | |
1853 | loff_t offset; | |
1854 | int namlen; | |
1855 | unsigned int d_type; | |
1856 | char name[]; | |
1857 | }; | |
1858 | ||
1859 | struct readdir_data { | |
1860 | char *dirent; | |
1861 | size_t used; | |
53c9c5c0 | 1862 | int full; |
14f7dd63 DW |
1863 | }; |
1864 | ||
1865 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1866 | loff_t offset, u64 ino, unsigned int d_type) | |
1867 | { | |
1868 | struct readdir_data *buf = __buf; | |
1869 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1870 | unsigned int reclen; | |
1871 | ||
1872 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1873 | if (buf->used + reclen > PAGE_SIZE) { |
1874 | buf->full = 1; | |
14f7dd63 | 1875 | return -EINVAL; |
53c9c5c0 | 1876 | } |
14f7dd63 DW |
1877 | |
1878 | de->namlen = namlen; | |
1879 | de->offset = offset; | |
1880 | de->ino = ino; | |
1881 | de->d_type = d_type; | |
1882 | memcpy(de->name, name, namlen); | |
1883 | buf->used += reclen; | |
1884 | ||
1885 | return 0; | |
1886 | } | |
1887 | ||
1888 | static int nfsd_buffered_readdir(struct file *file, filldir_t func, | |
1889 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1890 | { |
14f7dd63 DW |
1891 | struct readdir_data buf; |
1892 | struct buffered_dirent *de; | |
2628b766 | 1893 | int host_err; |
14f7dd63 DW |
1894 | int size; |
1895 | loff_t offset; | |
2628b766 | 1896 | |
14f7dd63 DW |
1897 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1898 | if (!buf.dirent) | |
1899 | return -ENOMEM; | |
1900 | ||
1901 | offset = *offsetp; | |
2628b766 | 1902 | |
14f7dd63 DW |
1903 | while (1) { |
1904 | unsigned int reclen; | |
1905 | ||
b726e923 | 1906 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1907 | buf.used = 0; |
53c9c5c0 | 1908 | buf.full = 0; |
14f7dd63 DW |
1909 | |
1910 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1911 | if (buf.full) |
1912 | host_err = 0; | |
1913 | ||
1914 | if (host_err < 0) | |
14f7dd63 DW |
1915 | break; |
1916 | ||
1917 | size = buf.used; | |
1918 | ||
1919 | if (!size) | |
1920 | break; | |
1921 | ||
14f7dd63 DW |
1922 | de = (struct buffered_dirent *)buf.dirent; |
1923 | while (size > 0) { | |
1924 | offset = de->offset; | |
1925 | ||
1926 | if (func(cdp, de->name, de->namlen, de->offset, | |
1927 | de->ino, de->d_type)) | |
1928 | goto done; | |
1929 | ||
1930 | if (cdp->err != nfs_ok) | |
1931 | goto done; | |
1932 | ||
1933 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1934 | sizeof(u64)); | |
1935 | size -= reclen; | |
1936 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1937 | } | |
c002a6c7 | 1938 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1939 | } |
1940 | ||
1941 | done: | |
1942 | free_page((unsigned long)(buf.dirent)); | |
2628b766 DW |
1943 | |
1944 | if (host_err) | |
1945 | return nfserrno(host_err); | |
14f7dd63 DW |
1946 | |
1947 | *offsetp = offset; | |
1948 | return cdp->err; | |
2628b766 DW |
1949 | } |
1950 | ||
1da177e4 LT |
1951 | /* |
1952 | * Read entries from a directory. | |
1953 | * The NFSv3/4 verifier we ignore for now. | |
1954 | */ | |
6264d69d | 1955 | __be32 |
1da177e4 | 1956 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1957 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1958 | { |
6264d69d | 1959 | __be32 err; |
1da177e4 LT |
1960 | struct file *file; |
1961 | loff_t offset = *offsetp; | |
1962 | ||
8837abca | 1963 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
1964 | if (err) |
1965 | goto out; | |
1966 | ||
1967 | offset = vfs_llseek(file, offset, 0); | |
1968 | if (offset < 0) { | |
1969 | err = nfserrno((int)offset); | |
1970 | goto out_close; | |
1971 | } | |
1972 | ||
14f7dd63 | 1973 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
1974 | |
1975 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1976 | err = nfs_ok; /* can still be found in ->err */ | |
1977 | out_close: | |
1978 | nfsd_close(file); | |
1979 | out: | |
1980 | return err; | |
1981 | } | |
1982 | ||
1983 | /* | |
1984 | * Get file system stats | |
1985 | * N.B. After this call fhp needs an fh_put | |
1986 | */ | |
6264d69d | 1987 | __be32 |
04716e66 | 1988 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 1989 | { |
04716e66 | 1990 | __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); |
726c3342 | 1991 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
1992 | err = nfserr_io; |
1993 | return err; | |
1994 | } | |
1995 | ||
c7d51402 | 1996 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 1997 | { |
c7d51402 | 1998 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
1999 | } |
2000 | ||
1da177e4 LT |
2001 | /* |
2002 | * Check for a user's access permissions to this inode. | |
2003 | */ | |
6264d69d | 2004 | __be32 |
0ec757df BF |
2005 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2006 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2007 | { |
2008 | struct inode *inode = dentry->d_inode; | |
2009 | int err; | |
2010 | ||
8837abca | 2011 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2012 | return 0; |
2013 | #if 0 | |
2014 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2015 | acc, | |
8837abca MS |
2016 | (acc & NFSD_MAY_READ)? " read" : "", |
2017 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2018 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2019 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2020 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2021 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2022 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2023 | inode->i_mode, |
2024 | IS_IMMUTABLE(inode)? " immut" : "", | |
2025 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2026 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2027 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2028 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2029 | #endif |
2030 | ||
2031 | /* Normally we reject any write/sattr etc access on a read-only file | |
2032 | * system. But if it is IRIX doing check on write-access for a | |
2033 | * device special file, we ignore rofs. | |
2034 | */ | |
8837abca MS |
2035 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2036 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2037 | if (exp_rdonly(rqstp, exp) || |
2038 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2039 | return nfserr_rofs; |
8837abca | 2040 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2041 | return nfserr_perm; |
2042 | } | |
8837abca | 2043 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2044 | return nfserr_perm; |
2045 | ||
8837abca | 2046 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2047 | /* If we cannot rely on authentication in NLM requests, |
2048 | * just allow locks, otherwise require read permission, or | |
2049 | * ownership | |
2050 | */ | |
2051 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2052 | return 0; | |
2053 | else | |
8837abca | 2054 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2055 | } |
2056 | /* | |
2057 | * The file owner always gets access permission for accesses that | |
2058 | * would normally be checked at open time. This is to make | |
2059 | * file access work even when the client has done a fchmod(fd, 0). | |
2060 | * | |
2061 | * However, `cp foo bar' should fail nevertheless when bar is | |
2062 | * readonly. A sensible way to do this might be to reject all | |
2063 | * attempts to truncate a read-only file, because a creat() call | |
2064 | * always implies file truncation. | |
2065 | * ... but this isn't really fair. A process may reasonably call | |
2066 | * ftruncate on an open file descriptor on a file with perm 000. | |
2067 | * We must trust the client to do permission checking - using "ACCESS" | |
2068 | * with NFSv3. | |
2069 | */ | |
8837abca | 2070 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2071 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2072 | return 0; |
2073 | ||
8837abca | 2074 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2075 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2076 | |
2077 | /* Allow read access to binaries even when mode 111 */ | |
2078 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2079 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2080 | err = inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
2081 | |
2082 | return err? nfserrno(err) : 0; | |
2083 | } | |
2084 | ||
2085 | void | |
2086 | nfsd_racache_shutdown(void) | |
2087 | { | |
54a66e54 JL |
2088 | struct raparms *raparm, *last_raparm; |
2089 | unsigned int i; | |
2090 | ||
1da177e4 | 2091 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2092 | |
2093 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2094 | raparm = raparm_hash[i].pb_head; | |
2095 | while(raparm) { | |
2096 | last_raparm = raparm; | |
2097 | raparm = raparm->p_next; | |
2098 | kfree(last_raparm); | |
2099 | } | |
2100 | raparm_hash[i].pb_head = NULL; | |
2101 | } | |
1da177e4 LT |
2102 | } |
2103 | /* | |
2104 | * Initialize readahead param cache | |
2105 | */ | |
2106 | int | |
2107 | nfsd_racache_init(int cache_size) | |
2108 | { | |
2109 | int i; | |
fce1456a GB |
2110 | int j = 0; |
2111 | int nperbucket; | |
54a66e54 | 2112 | struct raparms **raparm = NULL; |
1da177e4 | 2113 | |
fce1456a | 2114 | |
54a66e54 | 2115 | if (raparm_hash[0].pb_head) |
1da177e4 | 2116 | return 0; |
54a66e54 JL |
2117 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2118 | if (nperbucket < 2) | |
2119 | nperbucket = 2; | |
2120 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2121 | |
2122 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2123 | |
2124 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2125 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2126 | |
2127 | raparm = &raparm_hash[i].pb_head; | |
2128 | for (j = 0; j < nperbucket; j++) { | |
2129 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2130 | if (!*raparm) | |
2131 | goto out_nomem; | |
2132 | raparm = &(*raparm)->p_next; | |
2133 | } | |
2134 | *raparm = NULL; | |
4b3bb06b YB |
2135 | } |
2136 | ||
1da177e4 LT |
2137 | nfsdstats.ra_size = cache_size; |
2138 | return 0; | |
54a66e54 JL |
2139 | |
2140 | out_nomem: | |
2141 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2142 | nfsd_racache_shutdown(); | |
2143 | return -ENOMEM; | |
1da177e4 | 2144 | } |
a257cdd0 AG |
2145 | |
2146 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2147 | struct posix_acl * | |
2148 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2149 | { | |
2150 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2151 | char *name; | |
2152 | void *value = NULL; | |
2153 | ssize_t size; | |
2154 | struct posix_acl *acl; | |
2155 | ||
5be196e5 CH |
2156 | if (!IS_POSIXACL(inode)) |
2157 | return ERR_PTR(-EOPNOTSUPP); | |
2158 | ||
2159 | switch (type) { | |
2160 | case ACL_TYPE_ACCESS: | |
2161 | name = POSIX_ACL_XATTR_ACCESS; | |
2162 | break; | |
2163 | case ACL_TYPE_DEFAULT: | |
2164 | name = POSIX_ACL_XATTR_DEFAULT; | |
2165 | break; | |
2166 | default: | |
a257cdd0 | 2167 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2168 | } |
2169 | ||
5be196e5 CH |
2170 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2171 | if (size < 0) | |
2172 | return ERR_PTR(size); | |
a257cdd0 | 2173 | |
a257cdd0 | 2174 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2175 | kfree(value); |
2176 | return acl; | |
2177 | } | |
2178 | ||
2179 | int | |
2180 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2181 | { | |
2182 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2183 | char *name; | |
2184 | void *value = NULL; | |
2185 | size_t size; | |
2186 | int error; | |
2187 | ||
acfa4380 | 2188 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2189 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2190 | return -EOPNOTSUPP; | |
2191 | switch(type) { | |
2192 | case ACL_TYPE_ACCESS: | |
334a13ec | 2193 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2194 | break; |
2195 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2196 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2197 | break; |
2198 | default: | |
2199 | return -EOPNOTSUPP; | |
2200 | } | |
2201 | ||
2202 | if (acl && acl->a_count) { | |
334a13ec | 2203 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2204 | value = kmalloc(size, GFP_KERNEL); |
2205 | if (!value) | |
2206 | return -ENOMEM; | |
9ccfc29c FM |
2207 | error = posix_acl_to_xattr(acl, value, size); |
2208 | if (error < 0) | |
a257cdd0 | 2209 | goto getout; |
9ccfc29c | 2210 | size = error; |
a257cdd0 AG |
2211 | } else |
2212 | size = 0; | |
2213 | ||
18f335af DH |
2214 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2215 | if (error) | |
2216 | goto getout; | |
a257cdd0 | 2217 | if (size) |
5be196e5 | 2218 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2219 | else { |
2220 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2221 | error = 0; | |
2222 | else { | |
5be196e5 | 2223 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2224 | if (error == -ENODATA) |
2225 | error = 0; | |
2226 | } | |
2227 | } | |
18f335af | 2228 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2229 | |
2230 | getout: | |
2231 | kfree(value); | |
2232 | return error; | |
2233 | } | |
2234 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |