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 | 118 | /* successfully crossed mount point */ |
1644ccc8 AV |
119 | /* |
120 | * This is subtle: dentry is *not* under mnt at this point. | |
121 | * The only reason we are safe is that original mnt is pinned | |
122 | * down by exp, so we should dput before putting exp. | |
123 | */ | |
1da177e4 LT |
124 | dput(dentry); |
125 | *dpp = mounts; | |
1644ccc8 AV |
126 | exp_put(exp); |
127 | *expp = exp2; | |
1da177e4 | 128 | } else { |
5d3dbbea | 129 | exp_put(exp2); |
1da177e4 LT |
130 | dput(mounts); |
131 | } | |
132 | mntput(mnt); | |
133 | out: | |
134 | return err; | |
135 | } | |
136 | ||
6264d69d | 137 | __be32 |
6c0a654d | 138 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
5a022fc8 | 139 | const char *name, unsigned int len, |
6c0a654d | 140 | struct svc_export **exp_ret, struct dentry **dentry_ret) |
1da177e4 LT |
141 | { |
142 | struct svc_export *exp; | |
143 | struct dentry *dparent; | |
144 | struct dentry *dentry; | |
6264d69d AV |
145 | __be32 err; |
146 | int host_err; | |
1da177e4 LT |
147 | |
148 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
149 | ||
150 | /* Obtain dentry and export. */ | |
8837abca | 151 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC); |
1da177e4 LT |
152 | if (err) |
153 | return err; | |
154 | ||
155 | dparent = fhp->fh_dentry; | |
156 | exp = fhp->fh_export; | |
157 | exp_get(exp); | |
158 | ||
1da177e4 LT |
159 | /* Lookup the name, but don't follow links */ |
160 | if (isdotent(name, len)) { | |
161 | if (len==1) | |
162 | dentry = dget(dparent); | |
54775491 | 163 | else if (dparent != exp->ex_path.dentry) |
1da177e4 | 164 | dentry = dget_parent(dparent); |
54775491 | 165 | else if (!EX_NOHIDE(exp)) |
1da177e4 LT |
166 | dentry = dget(dparent); /* .. == . just like at / */ |
167 | else { | |
168 | /* checking mountpoint crossing is very different when stepping up */ | |
169 | struct svc_export *exp2 = NULL; | |
170 | struct dentry *dp; | |
54775491 | 171 | struct vfsmount *mnt = mntget(exp->ex_path.mnt); |
1da177e4 LT |
172 | dentry = dget(dparent); |
173 | while(dentry == mnt->mnt_root && follow_up(&mnt, &dentry)) | |
174 | ; | |
175 | dp = dget_parent(dentry); | |
176 | dput(dentry); | |
177 | dentry = dp; | |
178 | ||
0989a788 | 179 | exp2 = rqst_exp_parent(rqstp, mnt, dentry); |
2d3bb252 BF |
180 | if (PTR_ERR(exp2) == -ENOENT) { |
181 | dput(dentry); | |
182 | dentry = dget(dparent); | |
183 | } else if (IS_ERR(exp2)) { | |
6264d69d | 184 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
185 | dput(dentry); |
186 | mntput(mnt); | |
187 | goto out_nfserr; | |
1da177e4 LT |
188 | } else { |
189 | exp_put(exp); | |
190 | exp = exp2; | |
191 | } | |
192 | mntput(mnt); | |
193 | } | |
194 | } else { | |
195 | fh_lock(fhp); | |
196 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 197 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
198 | if (IS_ERR(dentry)) |
199 | goto out_nfserr; | |
200 | /* | |
201 | * check if we have crossed a mount point ... | |
202 | */ | |
203 | if (d_mountpoint(dentry)) { | |
6264d69d | 204 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
205 | dput(dentry); |
206 | goto out_nfserr; | |
207 | } | |
208 | } | |
209 | } | |
6c0a654d BF |
210 | *dentry_ret = dentry; |
211 | *exp_ret = exp; | |
212 | return 0; | |
213 | ||
214 | out_nfserr: | |
215 | exp_put(exp); | |
216 | return nfserrno(host_err); | |
217 | } | |
218 | ||
219 | /* | |
220 | * Look up one component of a pathname. | |
221 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
222 | * | |
223 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
224 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
225 | * accepted as it stands and the mounted directory is | |
226 | * returned. Otherwise the covered directory is returned. | |
227 | * NOTE: this mountpoint crossing is not supported properly by all | |
228 | * clients and is explicitly disallowed for NFSv3 | |
229 | * NeilBrown <neilb@cse.unsw.edu.au> | |
230 | */ | |
231 | __be32 | |
232 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
5a022fc8 | 233 | unsigned int len, struct svc_fh *resfh) |
6c0a654d BF |
234 | { |
235 | struct svc_export *exp; | |
236 | struct dentry *dentry; | |
237 | __be32 err; | |
238 | ||
239 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
240 | if (err) | |
241 | return err; | |
32c1eb0c AA |
242 | err = check_nfsd_access(exp, rqstp); |
243 | if (err) | |
244 | goto out; | |
1da177e4 LT |
245 | /* |
246 | * Note: we compose the file handle now, but as the | |
247 | * dentry may be negative, it may need to be updated. | |
248 | */ | |
249 | err = fh_compose(resfh, exp, dentry, fhp); | |
250 | if (!err && !dentry->d_inode) | |
251 | err = nfserr_noent; | |
32c1eb0c | 252 | out: |
1da177e4 | 253 | dput(dentry); |
1da177e4 LT |
254 | exp_put(exp); |
255 | return err; | |
1da177e4 LT |
256 | } |
257 | ||
6c0a654d | 258 | |
1da177e4 LT |
259 | /* |
260 | * Set various file attributes. | |
261 | * N.B. After this call fhp needs an fh_put | |
262 | */ | |
6264d69d | 263 | __be32 |
1da177e4 LT |
264 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
265 | int check_guard, time_t guardtime) | |
266 | { | |
267 | struct dentry *dentry; | |
268 | struct inode *inode; | |
8837abca | 269 | int accmode = NFSD_MAY_SATTR; |
1da177e4 | 270 | int ftype = 0; |
6264d69d AV |
271 | __be32 err; |
272 | int host_err; | |
1da177e4 LT |
273 | int size_change = 0; |
274 | ||
275 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
8837abca | 276 | accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
277 | if (iap->ia_valid & ATTR_SIZE) |
278 | ftype = S_IFREG; | |
279 | ||
280 | /* Get inode */ | |
281 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 282 | if (err) |
1da177e4 LT |
283 | goto out; |
284 | ||
285 | dentry = fhp->fh_dentry; | |
286 | inode = dentry->d_inode; | |
287 | ||
15b7a1b8 N |
288 | /* Ignore any mode updates on symlinks */ |
289 | if (S_ISLNK(inode->i_mode)) | |
290 | iap->ia_valid &= ~ATTR_MODE; | |
291 | ||
292 | if (!iap->ia_valid) | |
293 | goto out; | |
294 | ||
9c85fca5 CH |
295 | /* |
296 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
297 | * which only requires access, and "set-[ac]time-to-X" which |
298 | * requires ownership. | |
299 | * So if it looks like it might be "set both to the same time which | |
300 | * is close to now", and if inode_change_ok fails, then we | |
301 | * convert to "set to now" instead of "set to explicit time" | |
302 | * | |
303 | * We only call inode_change_ok as the last test as technically | |
304 | * it is not an interface that we should be using. It is only | |
305 | * valid if the filesystem does not define it's own i_op->setattr. | |
306 | */ | |
307 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
308 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
309 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
310 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
311 | /* | |
312 | * Looks probable. | |
313 | * | |
314 | * Now just make sure time is in the right ballpark. | |
315 | * Solaris, at least, doesn't seem to care what the time | |
316 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 317 | */ |
9c85fca5 CH |
318 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
319 | if (delta < 0) | |
320 | delta = -delta; | |
321 | if (delta < MAX_TOUCH_TIME_ERROR && | |
322 | inode_change_ok(inode, iap) != 0) { | |
323 | /* | |
324 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
325 | * This will cause notify_change to set these times | |
326 | * to "now" | |
327 | */ | |
328 | iap->ia_valid &= ~BOTH_TIME_SET; | |
329 | } | |
1da177e4 LT |
330 | } |
331 | ||
9c85fca5 CH |
332 | /* |
333 | * The size case is special. | |
334 | * It changes the file as well as the attributes. | |
335 | */ | |
1da177e4 LT |
336 | if (iap->ia_valid & ATTR_SIZE) { |
337 | if (iap->ia_size < inode->i_size) { | |
8837abca MS |
338 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, |
339 | NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE); | |
1da177e4 LT |
340 | if (err) |
341 | goto out; | |
342 | } | |
343 | ||
344 | /* | |
345 | * If we are changing the size of the file, then | |
346 | * we need to break all leases. | |
347 | */ | |
6264d69d AV |
348 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
349 | if (host_err == -EWOULDBLOCK) | |
350 | host_err = -ETIMEDOUT; | |
351 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
352 | goto out_nfserr; |
353 | ||
6264d69d AV |
354 | host_err = get_write_access(inode); |
355 | if (host_err) | |
1da177e4 LT |
356 | goto out_nfserr; |
357 | ||
358 | size_change = 1; | |
6264d69d AV |
359 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
360 | if (host_err) { | |
1da177e4 LT |
361 | put_write_access(inode); |
362 | goto out_nfserr; | |
363 | } | |
90c0af05 | 364 | vfs_dq_init(inode); |
1da177e4 LT |
365 | } |
366 | ||
ca456252 | 367 | /* sanitize the mode change */ |
1da177e4 LT |
368 | if (iap->ia_valid & ATTR_MODE) { |
369 | iap->ia_mode &= S_IALLUGO; | |
dee3209d | 370 | iap->ia_mode |= (inode->i_mode & ~S_IALLUGO); |
ca456252 JL |
371 | } |
372 | ||
373 | /* Revoke setuid/setgid on chown */ | |
0953e620 SP |
374 | if (!S_ISDIR(inode->i_mode) && |
375 | (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) || | |
376 | ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) { | |
ca456252 JL |
377 | iap->ia_valid |= ATTR_KILL_PRIV; |
378 | if (iap->ia_valid & ATTR_MODE) { | |
379 | /* we're setting mode too, just clear the s*id bits */ | |
8a0ce7d9 | 380 | iap->ia_mode &= ~S_ISUID; |
ca456252 JL |
381 | if (iap->ia_mode & S_IXGRP) |
382 | iap->ia_mode &= ~S_ISGID; | |
383 | } else { | |
384 | /* set ATTR_KILL_* bits and let VFS handle it */ | |
385 | iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID); | |
8a0ce7d9 | 386 | } |
1da177e4 LT |
387 | } |
388 | ||
1da177e4 LT |
389 | /* Change the attributes. */ |
390 | ||
391 | iap->ia_valid |= ATTR_CTIME; | |
392 | ||
393 | err = nfserr_notsync; | |
394 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
395 | fh_lock(fhp); | |
6264d69d AV |
396 | host_err = notify_change(dentry, iap); |
397 | err = nfserrno(host_err); | |
1da177e4 LT |
398 | fh_unlock(fhp); |
399 | } | |
400 | if (size_change) | |
401 | put_write_access(inode); | |
402 | if (!err) | |
403 | if (EX_ISSYNC(fhp->fh_export)) | |
404 | write_inode_now(inode, 1); | |
405 | out: | |
406 | return err; | |
407 | ||
408 | out_nfserr: | |
6264d69d | 409 | err = nfserrno(host_err); |
1da177e4 LT |
410 | goto out; |
411 | } | |
412 | ||
5be196e5 CH |
413 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
414 | defined(CONFIG_NFSD_V3_ACL) || \ | |
415 | defined(CONFIG_NFSD_V4) | |
416 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
417 | { | |
418 | ssize_t buflen; | |
6c6a426f | 419 | ssize_t ret; |
5be196e5 CH |
420 | |
421 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
422 | if (buflen <= 0) | |
423 | return buflen; | |
1da177e4 | 424 | |
5be196e5 CH |
425 | *buf = kmalloc(buflen, GFP_KERNEL); |
426 | if (!*buf) | |
427 | return -ENOMEM; | |
428 | ||
6c6a426f KK |
429 | ret = vfs_getxattr(dentry, key, *buf, buflen); |
430 | if (ret < 0) | |
431 | kfree(*buf); | |
432 | return ret; | |
5be196e5 CH |
433 | } |
434 | #endif | |
435 | ||
436 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
437 | static int |
438 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
439 | { | |
440 | int len; | |
441 | size_t buflen; | |
442 | char *buf = NULL; | |
443 | int error = 0; | |
1da177e4 LT |
444 | |
445 | buflen = posix_acl_xattr_size(pacl->a_count); | |
446 | buf = kmalloc(buflen, GFP_KERNEL); | |
447 | error = -ENOMEM; | |
448 | if (buf == NULL) | |
449 | goto out; | |
450 | ||
451 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
452 | if (len < 0) { | |
453 | error = len; | |
454 | goto out; | |
455 | } | |
456 | ||
5be196e5 | 457 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
458 | out: |
459 | kfree(buf); | |
460 | return error; | |
461 | } | |
462 | ||
6264d69d | 463 | __be32 |
1da177e4 LT |
464 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
465 | struct nfs4_acl *acl) | |
466 | { | |
6264d69d AV |
467 | __be32 error; |
468 | int host_error; | |
1da177e4 LT |
469 | struct dentry *dentry; |
470 | struct inode *inode; | |
471 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
472 | unsigned int flags = 0; | |
473 | ||
474 | /* Get inode */ | |
8837abca | 475 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR); |
1da177e4 | 476 | if (error) |
4b2ca38a | 477 | return error; |
1da177e4 LT |
478 | |
479 | dentry = fhp->fh_dentry; | |
480 | inode = dentry->d_inode; | |
481 | if (S_ISDIR(inode->i_mode)) | |
482 | flags = NFS4_ACL_DIR; | |
483 | ||
6264d69d AV |
484 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
485 | if (host_error == -EINVAL) { | |
4b2ca38a | 486 | return nfserr_attrnotsupp; |
6264d69d | 487 | } else if (host_error < 0) |
1da177e4 LT |
488 | goto out_nfserr; |
489 | ||
6264d69d AV |
490 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
491 | if (host_error < 0) | |
4b2ca38a | 492 | goto out_release; |
1da177e4 | 493 | |
4b2ca38a | 494 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 495 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 496 | |
4b2ca38a | 497 | out_release: |
1da177e4 LT |
498 | posix_acl_release(pacl); |
499 | posix_acl_release(dpacl); | |
1da177e4 | 500 | out_nfserr: |
f34f9242 | 501 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 502 | return nfserr_attrnotsupp; |
f34f9242 | 503 | else |
4b2ca38a | 504 | return nfserrno(host_error); |
1da177e4 LT |
505 | } |
506 | ||
507 | static struct posix_acl * | |
508 | _get_posix_acl(struct dentry *dentry, char *key) | |
509 | { | |
5be196e5 | 510 | void *buf = NULL; |
1da177e4 | 511 | struct posix_acl *pacl = NULL; |
5be196e5 | 512 | int buflen; |
1da177e4 | 513 | |
5be196e5 CH |
514 | buflen = nfsd_getxattr(dentry, key, &buf); |
515 | if (!buflen) | |
516 | buflen = -ENODATA; | |
517 | if (buflen <= 0) | |
518 | return ERR_PTR(buflen); | |
1da177e4 LT |
519 | |
520 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
521 | kfree(buf); |
522 | return pacl; | |
1da177e4 LT |
523 | } |
524 | ||
525 | int | |
526 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
527 | { | |
528 | struct inode *inode = dentry->d_inode; | |
529 | int error = 0; | |
530 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
531 | unsigned int flags = 0; | |
532 | ||
9a59f452 | 533 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
534 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
535 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
536 | if (IS_ERR(pacl)) { | |
537 | error = PTR_ERR(pacl); | |
538 | pacl = NULL; | |
539 | goto out; | |
540 | } | |
541 | ||
542 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 543 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
544 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
545 | dpacl = NULL; | |
546 | else if (IS_ERR(dpacl)) { | |
547 | error = PTR_ERR(dpacl); | |
548 | dpacl = NULL; | |
549 | goto out; | |
550 | } | |
551 | flags = NFS4_ACL_DIR; | |
552 | } | |
553 | ||
554 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
555 | if (IS_ERR(*acl)) { | |
556 | error = PTR_ERR(*acl); | |
557 | *acl = NULL; | |
558 | } | |
559 | out: | |
560 | posix_acl_release(pacl); | |
561 | posix_acl_release(dpacl); | |
562 | return error; | |
563 | } | |
564 | ||
565 | #endif /* defined(CONFIG_NFS_V4) */ | |
566 | ||
567 | #ifdef CONFIG_NFSD_V3 | |
568 | /* | |
569 | * Check server access rights to a file system object | |
570 | */ | |
571 | struct accessmap { | |
572 | u32 access; | |
573 | int how; | |
574 | }; | |
575 | static struct accessmap nfs3_regaccess[] = { | |
8837abca MS |
576 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
577 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
578 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC }, | |
579 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE }, | |
1da177e4 LT |
580 | |
581 | { 0, 0 } | |
582 | }; | |
583 | ||
584 | static struct accessmap nfs3_diraccess[] = { | |
8837abca MS |
585 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
586 | { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC }, | |
587 | { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC}, | |
588 | { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE }, | |
589 | { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE }, | |
1da177e4 LT |
590 | |
591 | { 0, 0 } | |
592 | }; | |
593 | ||
594 | static struct accessmap nfs3_anyaccess[] = { | |
595 | /* Some clients - Solaris 2.6 at least, make an access call | |
596 | * to the server to check for access for things like /dev/null | |
597 | * (which really, the server doesn't care about). So | |
598 | * We provide simple access checking for them, looking | |
599 | * mainly at mode bits, and we make sure to ignore read-only | |
600 | * filesystem checks | |
601 | */ | |
8837abca MS |
602 | { NFS3_ACCESS_READ, NFSD_MAY_READ }, |
603 | { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC }, | |
604 | { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
605 | { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS }, | |
1da177e4 LT |
606 | |
607 | { 0, 0 } | |
608 | }; | |
609 | ||
6264d69d | 610 | __be32 |
1da177e4 LT |
611 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
612 | { | |
613 | struct accessmap *map; | |
614 | struct svc_export *export; | |
615 | struct dentry *dentry; | |
616 | u32 query, result = 0, sresult = 0; | |
6264d69d | 617 | __be32 error; |
1da177e4 | 618 | |
8837abca | 619 | error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP); |
1da177e4 LT |
620 | if (error) |
621 | goto out; | |
622 | ||
623 | export = fhp->fh_export; | |
624 | dentry = fhp->fh_dentry; | |
625 | ||
626 | if (S_ISREG(dentry->d_inode->i_mode)) | |
627 | map = nfs3_regaccess; | |
628 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
629 | map = nfs3_diraccess; | |
630 | else | |
631 | map = nfs3_anyaccess; | |
632 | ||
633 | ||
634 | query = *access; | |
635 | for (; map->access; map++) { | |
636 | if (map->access & query) { | |
6264d69d | 637 | __be32 err2; |
1da177e4 LT |
638 | |
639 | sresult |= map->access; | |
640 | ||
0ec757df | 641 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
642 | switch (err2) { |
643 | case nfs_ok: | |
644 | result |= map->access; | |
645 | break; | |
646 | ||
647 | /* the following error codes just mean the access was not allowed, | |
648 | * rather than an error occurred */ | |
649 | case nfserr_rofs: | |
650 | case nfserr_acces: | |
651 | case nfserr_perm: | |
652 | /* simply don't "or" in the access bit. */ | |
653 | break; | |
654 | default: | |
655 | error = err2; | |
656 | goto out; | |
657 | } | |
658 | } | |
659 | } | |
660 | *access = result; | |
661 | if (supported) | |
662 | *supported = sresult; | |
663 | ||
664 | out: | |
665 | return error; | |
666 | } | |
667 | #endif /* CONFIG_NFSD_V3 */ | |
668 | ||
669 | ||
670 | ||
671 | /* | |
672 | * Open an existing file or directory. | |
673 | * The access argument indicates the type of open (read/write/lock) | |
674 | * N.B. After this call fhp needs an fh_put | |
675 | */ | |
6264d69d | 676 | __be32 |
1da177e4 LT |
677 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
678 | int access, struct file **filp) | |
679 | { | |
745ca247 | 680 | const struct cred *cred = current_cred(); |
1da177e4 LT |
681 | struct dentry *dentry; |
682 | struct inode *inode; | |
6264d69d AV |
683 | int flags = O_RDONLY|O_LARGEFILE; |
684 | __be32 err; | |
685 | int host_err; | |
1da177e4 LT |
686 | |
687 | /* | |
688 | * If we get here, then the client has already done an "open", | |
689 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
690 | * in case a chmod has now revoked permission. | |
691 | */ | |
8837abca | 692 | err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
693 | if (err) |
694 | goto out; | |
695 | ||
696 | dentry = fhp->fh_dentry; | |
697 | inode = dentry->d_inode; | |
698 | ||
699 | /* Disallow write access to files with the append-only bit set | |
700 | * or any access when mandatory locking enabled | |
701 | */ | |
702 | err = nfserr_perm; | |
8837abca | 703 | if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE)) |
1da177e4 | 704 | goto out; |
5e7fc436 BF |
705 | /* |
706 | * We must ignore files (but only files) which might have mandatory | |
707 | * locks on them because there is no way to know if the accesser has | |
708 | * the lock. | |
709 | */ | |
710 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
711 | goto out; |
712 | ||
713 | if (!inode->i_fop) | |
714 | goto out; | |
715 | ||
716 | /* | |
717 | * Check to see if there are any leases on this file. | |
718 | * This may block while leases are broken. | |
719 | */ | |
8837abca | 720 | host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0)); |
6264d69d AV |
721 | if (host_err == -EWOULDBLOCK) |
722 | host_err = -ETIMEDOUT; | |
723 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
724 | goto out_nfserr; |
725 | ||
8837abca MS |
726 | if (access & NFSD_MAY_WRITE) { |
727 | if (access & NFSD_MAY_READ) | |
9ecb6a08 BF |
728 | flags = O_RDWR|O_LARGEFILE; |
729 | else | |
730 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 | 731 | |
90c0af05 | 732 | vfs_dq_init(inode); |
1da177e4 | 733 | } |
54775491 | 734 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt), |
745ca247 | 735 | flags, cred); |
1da177e4 | 736 | if (IS_ERR(*filp)) |
6264d69d | 737 | host_err = PTR_ERR(*filp); |
1da177e4 | 738 | out_nfserr: |
6264d69d | 739 | err = nfserrno(host_err); |
1da177e4 LT |
740 | out: |
741 | return err; | |
742 | } | |
743 | ||
744 | /* | |
745 | * Close a file. | |
746 | */ | |
747 | void | |
748 | nfsd_close(struct file *filp) | |
749 | { | |
750 | fput(filp); | |
751 | } | |
752 | ||
9a8d248e BF |
753 | /* |
754 | * Sync a file | |
755 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
756 | * after it. | |
757 | */ | |
758 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, | |
759 | const struct file_operations *fop) | |
760 | { | |
761 | struct inode *inode = dp->d_inode; | |
762 | int (*fsync) (struct file *, struct dentry *, int); | |
763 | int err; | |
764 | ||
765 | err = filemap_fdatawrite(inode->i_mapping); | |
766 | if (err == 0 && fop && (fsync = fop->fsync)) | |
767 | err = fsync(filp, dp, 0); | |
768 | if (err == 0) | |
769 | err = filemap_fdatawait(inode->i_mapping); | |
770 | ||
771 | return err; | |
772 | } | |
773 | ||
a334de28 | 774 | static int |
1da177e4 LT |
775 | nfsd_sync(struct file *filp) |
776 | { | |
9a8d248e BF |
777 | int err; |
778 | struct inode *inode = filp->f_path.dentry->d_inode; | |
779 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
780 | mutex_lock(&inode->i_mutex); | |
781 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); | |
782 | mutex_unlock(&inode->i_mutex); | |
783 | ||
784 | return err; | |
1da177e4 LT |
785 | } |
786 | ||
f193fbab | 787 | int |
9a8d248e | 788 | nfsd_sync_dir(struct dentry *dp) |
1da177e4 | 789 | { |
9a8d248e | 790 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
791 | } |
792 | ||
793 | /* | |
794 | * Obtain the readahead parameters for the file | |
795 | * specified by (dev, ino). | |
796 | */ | |
1da177e4 LT |
797 | |
798 | static inline struct raparms * | |
799 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
800 | { | |
801 | struct raparms *ra, **rap, **frap = NULL; | |
802 | int depth = 0; | |
fce1456a GB |
803 | unsigned int hash; |
804 | struct raparm_hbucket *rab; | |
805 | ||
806 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
807 | rab = &raparm_hash[hash]; | |
1da177e4 | 808 | |
fce1456a GB |
809 | spin_lock(&rab->pb_lock); |
810 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
811 | if (ra->p_ino == ino && ra->p_dev == dev) |
812 | goto found; | |
813 | depth++; | |
814 | if (ra->p_count == 0) | |
815 | frap = rap; | |
816 | } | |
817 | depth = nfsdstats.ra_size*11/10; | |
818 | if (!frap) { | |
fce1456a | 819 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
820 | return NULL; |
821 | } | |
822 | rap = frap; | |
823 | ra = *frap; | |
824 | ra->p_dev = dev; | |
825 | ra->p_ino = ino; | |
826 | ra->p_set = 0; | |
fce1456a | 827 | ra->p_hindex = hash; |
1da177e4 | 828 | found: |
fce1456a | 829 | if (rap != &rab->pb_head) { |
1da177e4 | 830 | *rap = ra->p_next; |
fce1456a GB |
831 | ra->p_next = rab->pb_head; |
832 | rab->pb_head = ra; | |
1da177e4 LT |
833 | } |
834 | ra->p_count++; | |
835 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 836 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
837 | return ra; |
838 | } | |
839 | ||
840 | /* | |
cf8208d0 JA |
841 | * Grab and keep cached pages associated with a file in the svc_rqst |
842 | * so that they can be passed to the network sendmsg/sendpage routines | |
843 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
844 | */ |
845 | static int | |
cf8208d0 JA |
846 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
847 | struct splice_desc *sd) | |
1da177e4 | 848 | { |
cf8208d0 | 849 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 850 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
851 | struct page *page = buf->page; |
852 | size_t size; | |
853 | int ret; | |
1da177e4 | 854 | |
cac36bb0 | 855 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
856 | if (unlikely(ret)) |
857 | return ret; | |
858 | ||
859 | size = sd->len; | |
1da177e4 LT |
860 | |
861 | if (rqstp->rq_res.page_len == 0) { | |
862 | get_page(page); | |
44524359 N |
863 | put_page(*pp); |
864 | *pp = page; | |
865 | rqstp->rq_resused++; | |
cf8208d0 | 866 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 867 | rqstp->rq_res.page_len = size; |
44524359 | 868 | } else if (page != pp[-1]) { |
1da177e4 | 869 | get_page(page); |
250f3915 N |
870 | if (*pp) |
871 | put_page(*pp); | |
44524359 N |
872 | *pp = page; |
873 | rqstp->rq_resused++; | |
1da177e4 | 874 | rqstp->rq_res.page_len += size; |
44524359 | 875 | } else |
1da177e4 | 876 | rqstp->rq_res.page_len += size; |
1da177e4 | 877 | |
1da177e4 LT |
878 | return size; |
879 | } | |
880 | ||
cf8208d0 JA |
881 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
882 | struct splice_desc *sd) | |
883 | { | |
884 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
885 | } | |
886 | ||
a8754bee DH |
887 | static inline int svc_msnfs(struct svc_fh *ffhp) |
888 | { | |
889 | #ifdef MSNFS | |
890 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
891 | #else | |
892 | return 0; | |
893 | #endif | |
894 | } | |
895 | ||
6264d69d | 896 | static __be32 |
1da177e4 LT |
897 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
898 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
899 | { | |
900 | struct inode *inode; | |
901 | struct raparms *ra; | |
902 | mm_segment_t oldfs; | |
6264d69d AV |
903 | __be32 err; |
904 | int host_err; | |
1da177e4 LT |
905 | |
906 | err = nfserr_perm; | |
7eaa36e2 | 907 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
908 | |
909 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 910 | goto out; |
1da177e4 LT |
911 | |
912 | /* Get readahead parameters */ | |
913 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
914 | ||
915 | if (ra && ra->p_set) | |
916 | file->f_ra = ra->p_ra; | |
917 | ||
cf8208d0 JA |
918 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
919 | struct splice_desc sd = { | |
920 | .len = 0, | |
921 | .total_len = *count, | |
922 | .pos = offset, | |
923 | .u.data = rqstp, | |
924 | }; | |
925 | ||
4fbef206 | 926 | rqstp->rq_resused = 1; |
cf8208d0 | 927 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
928 | } else { |
929 | oldfs = get_fs(); | |
930 | set_fs(KERNEL_DS); | |
6264d69d | 931 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
932 | set_fs(oldfs); |
933 | } | |
934 | ||
935 | /* Write back readahead params */ | |
936 | if (ra) { | |
fce1456a GB |
937 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
938 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
939 | ra->p_ra = file->f_ra; |
940 | ra->p_set = 1; | |
941 | ra->p_count--; | |
fce1456a | 942 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
943 | } |
944 | ||
6264d69d AV |
945 | if (host_err >= 0) { |
946 | nfsdstats.io_read += host_err; | |
947 | *count = host_err; | |
1da177e4 | 948 | err = 0; |
7eaa36e2 | 949 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 950 | } else |
6264d69d | 951 | err = nfserrno(host_err); |
1da177e4 LT |
952 | out: |
953 | return err; | |
954 | } | |
955 | ||
9f708e40 NB |
956 | static void kill_suid(struct dentry *dentry) |
957 | { | |
958 | struct iattr ia; | |
b5376771 | 959 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; |
9f708e40 | 960 | |
1b1dcc1b | 961 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 962 | notify_change(dentry, &ia); |
1b1dcc1b | 963 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
964 | } |
965 | ||
d911df7b BF |
966 | /* |
967 | * Gathered writes: If another process is currently writing to the file, | |
968 | * there's a high chance this is another nfsd (triggered by a bulk write | |
969 | * from a client's biod). Rather than syncing the file with each write | |
970 | * request, we sleep for 10 msec. | |
971 | * | |
972 | * I don't know if this roughly approximates C. Juszak's idea of | |
973 | * gathered writes, but it's a nice and simple solution (IMHO), and it | |
974 | * seems to work:-) | |
975 | * | |
976 | * Note: we do this only in the NFSv2 case, since v3 and higher have a | |
977 | * better tool (separate unstable writes and commits) for solving this | |
978 | * problem. | |
979 | */ | |
980 | static int wait_for_concurrent_writes(struct file *file) | |
981 | { | |
982 | struct inode *inode = file->f_path.dentry->d_inode; | |
983 | static ino_t last_ino; | |
984 | static dev_t last_dev; | |
985 | int err = 0; | |
986 | ||
987 | if (atomic_read(&inode->i_writecount) > 1 | |
988 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
989 | dprintk("nfsd: write defer %d\n", task_pid_nr(current)); | |
990 | msleep(10); | |
991 | dprintk("nfsd: write resume %d\n", task_pid_nr(current)); | |
992 | } | |
993 | ||
994 | if (inode->i_state & I_DIRTY) { | |
995 | dprintk("nfsd: write sync %d\n", task_pid_nr(current)); | |
996 | err = nfsd_sync(file); | |
997 | } | |
998 | last_ino = inode->i_ino; | |
999 | last_dev = inode->i_sb->s_dev; | |
1000 | return err; | |
1001 | } | |
1002 | ||
6264d69d | 1003 | static __be32 |
1da177e4 LT |
1004 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1005 | loff_t offset, struct kvec *vec, int vlen, | |
31dec253 | 1006 | unsigned long *cnt, int *stablep) |
1da177e4 LT |
1007 | { |
1008 | struct svc_export *exp; | |
1009 | struct dentry *dentry; | |
1010 | struct inode *inode; | |
1011 | mm_segment_t oldfs; | |
6264d69d AV |
1012 | __be32 err = 0; |
1013 | int host_err; | |
1da177e4 | 1014 | int stable = *stablep; |
48e03bc5 | 1015 | int use_wgather; |
1da177e4 | 1016 | |
45bd3b3d | 1017 | #ifdef MSNFS |
1da177e4 LT |
1018 | err = nfserr_perm; |
1019 | ||
1da177e4 | 1020 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
31dec253 | 1021 | (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt))) |
1da177e4 LT |
1022 | goto out; |
1023 | #endif | |
1024 | ||
7eaa36e2 | 1025 | dentry = file->f_path.dentry; |
1da177e4 LT |
1026 | inode = dentry->d_inode; |
1027 | exp = fhp->fh_export; | |
1028 | ||
1029 | /* | |
1030 | * Request sync writes if | |
1031 | * - the sync export option has been set, or | |
1032 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
1033 | * - The file system doesn't support fsync(). | |
48e03bc5 | 1034 | * When NFSv2 gathered writes have been configured for this volume, |
1da177e4 LT |
1035 | * flushing the data to disk is handled separately below. |
1036 | */ | |
48e03bc5 | 1037 | use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp); |
1da177e4 | 1038 | |
3ba15148 | 1039 | if (!file->f_op->fsync) {/* COMMIT3 cannot work */ |
1da177e4 LT |
1040 | stable = 2; |
1041 | *stablep = 2; /* FILE_SYNC */ | |
1042 | } | |
1043 | ||
1044 | if (!EX_ISSYNC(exp)) | |
1045 | stable = 0; | |
48e03bc5 | 1046 | if (stable && !use_wgather) { |
db1dd4d3 | 1047 | spin_lock(&file->f_lock); |
1da177e4 | 1048 | file->f_flags |= O_SYNC; |
db1dd4d3 JC |
1049 | spin_unlock(&file->f_lock); |
1050 | } | |
1da177e4 LT |
1051 | |
1052 | /* Write the data. */ | |
1053 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1054 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 1055 | set_fs(oldfs); |
e4636d53 BF |
1056 | if (host_err < 0) |
1057 | goto out_nfserr; | |
1058 | *cnt = host_err; | |
1059 | nfsdstats.io_write += host_err; | |
1060 | fsnotify_modify(file->f_path.dentry); | |
1da177e4 LT |
1061 | |
1062 | /* clear setuid/setgid flag after write */ | |
e4636d53 | 1063 | if (inode->i_mode & (S_ISUID | S_ISGID)) |
9f708e40 | 1064 | kill_suid(dentry); |
1da177e4 | 1065 | |
e4636d53 | 1066 | if (stable && use_wgather) |
d911df7b | 1067 | host_err = wait_for_concurrent_writes(file); |
1da177e4 | 1068 | |
e4636d53 | 1069 | out_nfserr: |
6264d69d | 1070 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
a0d24b29 | 1071 | if (host_err >= 0) |
1da177e4 | 1072 | err = 0; |
a0d24b29 | 1073 | else |
6264d69d | 1074 | err = nfserrno(host_err); |
1da177e4 LT |
1075 | out: |
1076 | return err; | |
1077 | } | |
1078 | ||
1079 | /* | |
1080 | * Read data from a file. count must contain the requested read count | |
1081 | * on entry. On return, *count contains the number of bytes actually read. | |
1082 | * N.B. After this call fhp needs an fh_put | |
1083 | */ | |
6264d69d | 1084 | __be32 |
1da177e4 LT |
1085 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1086 | loff_t offset, struct kvec *vec, int vlen, | |
1087 | unsigned long *count) | |
1088 | { | |
6264d69d | 1089 | __be32 err; |
1da177e4 LT |
1090 | |
1091 | if (file) { | |
0ec757df | 1092 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1093 | NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1094 | if (err) |
1095 | goto out; | |
1096 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1097 | } else { | |
8837abca | 1098 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file); |
1da177e4 LT |
1099 | if (err) |
1100 | goto out; | |
1101 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1102 | nfsd_close(file); | |
1103 | } | |
1104 | out: | |
1105 | return err; | |
1106 | } | |
1107 | ||
1108 | /* | |
1109 | * Write data to a file. | |
1110 | * The stable flag requests synchronous writes. | |
1111 | * N.B. After this call fhp needs an fh_put | |
1112 | */ | |
6264d69d | 1113 | __be32 |
1da177e4 | 1114 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
31dec253 | 1115 | loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt, |
1da177e4 LT |
1116 | int *stablep) |
1117 | { | |
6264d69d | 1118 | __be32 err = 0; |
1da177e4 LT |
1119 | |
1120 | if (file) { | |
0ec757df | 1121 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
8837abca | 1122 | NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE); |
1da177e4 LT |
1123 | if (err) |
1124 | goto out; | |
1125 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1126 | stablep); | |
1127 | } else { | |
8837abca | 1128 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1da177e4 LT |
1129 | if (err) |
1130 | goto out; | |
1131 | ||
1132 | if (cnt) | |
1133 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1134 | cnt, stablep); | |
1135 | nfsd_close(file); | |
1136 | } | |
1137 | out: | |
1138 | return err; | |
1139 | } | |
1140 | ||
1141 | #ifdef CONFIG_NFSD_V3 | |
1142 | /* | |
1143 | * Commit all pending writes to stable storage. | |
1144 | * Strictly speaking, we could sync just the indicated file region here, | |
1145 | * but there's currently no way we can ask the VFS to do so. | |
1146 | * | |
1147 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1148 | * data to the client, as locking happens lower down in the filesystem. | |
1149 | */ | |
6264d69d | 1150 | __be32 |
1da177e4 LT |
1151 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1152 | loff_t offset, unsigned long count) | |
1153 | { | |
1154 | struct file *file; | |
6264d69d | 1155 | __be32 err; |
1da177e4 LT |
1156 | |
1157 | if ((u64)count > ~(u64)offset) | |
1158 | return nfserr_inval; | |
1159 | ||
8837abca MS |
1160 | err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file); |
1161 | if (err) | |
1da177e4 LT |
1162 | return err; |
1163 | if (EX_ISSYNC(fhp->fh_export)) { | |
1164 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1165 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1166 | } else { |
1167 | err = nfserr_notsupp; | |
1168 | } | |
1169 | } | |
1170 | ||
1171 | nfsd_close(file); | |
1172 | return err; | |
1173 | } | |
1174 | #endif /* CONFIG_NFSD_V3 */ | |
1175 | ||
f2b0dee2 | 1176 | static __be32 |
5c002b3b BF |
1177 | nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp, |
1178 | struct iattr *iap) | |
1179 | { | |
1180 | /* | |
1181 | * Mode has already been set earlier in create: | |
1182 | */ | |
1183 | iap->ia_valid &= ~ATTR_MODE; | |
1184 | /* | |
1185 | * Setting uid/gid works only for root. Irix appears to | |
1186 | * send along the gid on create when it tries to implement | |
1187 | * setgid directories via NFS: | |
1188 | */ | |
5cc0a840 | 1189 | if (current_fsuid() != 0) |
5c002b3b BF |
1190 | iap->ia_valid &= ~(ATTR_UID|ATTR_GID); |
1191 | if (iap->ia_valid) | |
1192 | return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); | |
1193 | return 0; | |
1194 | } | |
1195 | ||
4ac35c2f | 1196 | /* HPUX client sometimes creates a file in mode 000, and sets size to 0. |
1197 | * setting size to 0 may fail for some specific file systems by the permission | |
1198 | * checking which requires WRITE permission but the mode is 000. | |
1199 | * we ignore the resizing(to 0) on the just new created file, since the size is | |
1200 | * 0 after file created. | |
1201 | * | |
1202 | * call this only after vfs_create() is called. | |
1203 | * */ | |
1204 | static void | |
1205 | nfsd_check_ignore_resizing(struct iattr *iap) | |
1206 | { | |
1207 | if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) | |
1208 | iap->ia_valid &= ~ATTR_SIZE; | |
1209 | } | |
1210 | ||
1da177e4 LT |
1211 | /* |
1212 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1213 | * not yet implemented. | |
1214 | * If the response fh has been verified, the parent directory should | |
1215 | * already be locked. Note that the parent directory is left locked. | |
1216 | * | |
1217 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1218 | */ | |
6264d69d | 1219 | __be32 |
1da177e4 LT |
1220 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1221 | char *fname, int flen, struct iattr *iap, | |
1222 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1223 | { | |
1224 | struct dentry *dentry, *dchild = NULL; | |
1225 | struct inode *dirp; | |
6264d69d | 1226 | __be32 err; |
5c002b3b | 1227 | __be32 err2; |
6264d69d | 1228 | int host_err; |
1da177e4 LT |
1229 | |
1230 | err = nfserr_perm; | |
1231 | if (!flen) | |
1232 | goto out; | |
1233 | err = nfserr_exist; | |
1234 | if (isdotent(fname, flen)) | |
1235 | goto out; | |
1236 | ||
8837abca | 1237 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1238 | if (err) |
1239 | goto out; | |
1240 | ||
1241 | dentry = fhp->fh_dentry; | |
1242 | dirp = dentry->d_inode; | |
1243 | ||
1244 | err = nfserr_notdir; | |
acfa4380 | 1245 | if (!dirp->i_op->lookup) |
1da177e4 LT |
1246 | goto out; |
1247 | /* | |
1248 | * Check whether the response file handle has been verified yet. | |
1249 | * If it has, the parent directory should already be locked. | |
1250 | */ | |
1251 | if (!resfhp->fh_dentry) { | |
1252 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1253 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1254 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1255 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1256 | if (IS_ERR(dchild)) |
1257 | goto out_nfserr; | |
1258 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1259 | if (err) | |
1260 | goto out; | |
1261 | } else { | |
1262 | /* called from nfsd_proc_create */ | |
1263 | dchild = dget(resfhp->fh_dentry); | |
1264 | if (!fhp->fh_locked) { | |
1265 | /* not actually possible */ | |
1266 | printk(KERN_ERR | |
1267 | "nfsd_create: parent %s/%s not locked!\n", | |
1268 | dentry->d_parent->d_name.name, | |
1269 | dentry->d_name.name); | |
d75f2b9f | 1270 | err = nfserr_io; |
1da177e4 LT |
1271 | goto out; |
1272 | } | |
1273 | } | |
1274 | /* | |
1275 | * Make sure the child dentry is still negative ... | |
1276 | */ | |
1277 | err = nfserr_exist; | |
1278 | if (dchild->d_inode) { | |
1279 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1280 | dentry->d_name.name, dchild->d_name.name); | |
1281 | goto out; | |
1282 | } | |
1283 | ||
1284 | if (!(iap->ia_valid & ATTR_MODE)) | |
1285 | iap->ia_mode = 0; | |
1286 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1287 | ||
07cad1d2 MS |
1288 | err = nfserr_inval; |
1289 | if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) { | |
1290 | printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", | |
1291 | type); | |
1292 | goto out; | |
1293 | } | |
1294 | ||
1295 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); | |
1296 | if (host_err) | |
1297 | goto out_nfserr; | |
1298 | ||
1da177e4 LT |
1299 | /* |
1300 | * Get the dir op function pointer. | |
1301 | */ | |
088406bc | 1302 | err = 0; |
1da177e4 LT |
1303 | switch (type) { |
1304 | case S_IFREG: | |
6264d69d | 1305 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
4ac35c2f | 1306 | if (!host_err) |
1307 | nfsd_check_ignore_resizing(iap); | |
1da177e4 LT |
1308 | break; |
1309 | case S_IFDIR: | |
6264d69d | 1310 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1311 | break; |
1312 | case S_IFCHR: | |
1313 | case S_IFBLK: | |
1314 | case S_IFIFO: | |
1315 | case S_IFSOCK: | |
6264d69d | 1316 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 | 1317 | break; |
1da177e4 | 1318 | } |
463c3197 DH |
1319 | if (host_err < 0) { |
1320 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1321 | goto out_nfserr; |
463c3197 | 1322 | } |
1da177e4 LT |
1323 | |
1324 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1325 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1326 | write_inode_now(dchild->d_inode, 1); |
1327 | } | |
1328 | ||
5c002b3b BF |
1329 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1330 | if (err2) | |
1331 | err = err2; | |
463c3197 | 1332 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1333 | /* |
1334 | * Update the file handle to get the new inode info. | |
1335 | */ | |
1336 | if (!err) | |
1337 | err = fh_update(resfhp); | |
1338 | out: | |
1339 | if (dchild && !IS_ERR(dchild)) | |
1340 | dput(dchild); | |
1341 | return err; | |
1342 | ||
1343 | out_nfserr: | |
6264d69d | 1344 | err = nfserrno(host_err); |
1da177e4 LT |
1345 | goto out; |
1346 | } | |
1347 | ||
1348 | #ifdef CONFIG_NFSD_V3 | |
1349 | /* | |
1350 | * NFSv3 version of nfsd_create | |
1351 | */ | |
6264d69d | 1352 | __be32 |
1da177e4 LT |
1353 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1354 | char *fname, int flen, struct iattr *iap, | |
1355 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1356 | int *truncp, int *created) |
1da177e4 LT |
1357 | { |
1358 | struct dentry *dentry, *dchild = NULL; | |
1359 | struct inode *dirp; | |
6264d69d | 1360 | __be32 err; |
5c002b3b | 1361 | __be32 err2; |
6264d69d | 1362 | int host_err; |
1da177e4 | 1363 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1364 | |
1365 | err = nfserr_perm; | |
1366 | if (!flen) | |
1367 | goto out; | |
1368 | err = nfserr_exist; | |
1369 | if (isdotent(fname, flen)) | |
1370 | goto out; | |
1371 | if (!(iap->ia_valid & ATTR_MODE)) | |
1372 | iap->ia_mode = 0; | |
8837abca | 1373 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1374 | if (err) |
1375 | goto out; | |
1376 | ||
1377 | dentry = fhp->fh_dentry; | |
1378 | dirp = dentry->d_inode; | |
1379 | ||
1380 | /* Get all the sanity checks out of the way before | |
1381 | * we lock the parent. */ | |
1382 | err = nfserr_notdir; | |
acfa4380 | 1383 | if (!dirp->i_op->lookup) |
1da177e4 | 1384 | goto out; |
12fd3520 | 1385 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1386 | |
1387 | /* | |
1388 | * Compose the response file handle. | |
1389 | */ | |
1390 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1391 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1392 | if (IS_ERR(dchild)) |
1393 | goto out_nfserr; | |
1394 | ||
1395 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1396 | if (err) | |
1397 | goto out; | |
1398 | ||
1399 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1400 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1401 | * the high bit set, so just clear the high bits. If this is |
1402 | * ever changed to use different attrs for storing the | |
1403 | * verifier, then do_open_lookup() will also need to be fixed | |
1404 | * accordingly. | |
1da177e4 LT |
1405 | */ |
1406 | v_mtime = verifier[0]&0x7fffffff; | |
1407 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1408 | } |
1409 | ||
463c3197 DH |
1410 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1411 | if (host_err) | |
1412 | goto out_nfserr; | |
1da177e4 LT |
1413 | if (dchild->d_inode) { |
1414 | err = 0; | |
1415 | ||
1416 | switch (createmode) { | |
1417 | case NFS3_CREATE_UNCHECKED: | |
1418 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1419 | err = nfserr_exist; | |
1420 | else if (truncp) { | |
1421 | /* in nfsv4, we need to treat this case a little | |
1422 | * differently. we don't want to truncate the | |
1423 | * file now; this would be wrong if the OPEN | |
1424 | * fails for some other reason. furthermore, | |
1425 | * if the size is nonzero, we should ignore it | |
1426 | * according to spec! | |
1427 | */ | |
1428 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1429 | } | |
1430 | else { | |
1431 | iap->ia_valid &= ATTR_SIZE; | |
1432 | goto set_attr; | |
1433 | } | |
1434 | break; | |
1435 | case NFS3_CREATE_EXCLUSIVE: | |
1436 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1437 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1438 | && dchild->d_inode->i_size == 0 ) |
1439 | break; | |
1440 | /* fallthru */ | |
1441 | case NFS3_CREATE_GUARDED: | |
1442 | err = nfserr_exist; | |
1443 | } | |
463c3197 | 1444 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1da177e4 LT |
1445 | goto out; |
1446 | } | |
1447 | ||
6264d69d | 1448 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
463c3197 DH |
1449 | if (host_err < 0) { |
1450 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1451 | goto out_nfserr; |
463c3197 | 1452 | } |
81ac95c5 BF |
1453 | if (created) |
1454 | *created = 1; | |
1da177e4 LT |
1455 | |
1456 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1457 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1458 | /* setattr will sync the child (or not) */ |
1459 | } | |
1460 | ||
4ac35c2f | 1461 | nfsd_check_ignore_resizing(iap); |
1462 | ||
1da177e4 | 1463 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1464 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1465 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1466 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1467 | /* XXX someone who knows this better please fix it for nsec */ |
1468 | iap->ia_mtime.tv_sec = v_mtime; | |
1469 | iap->ia_atime.tv_sec = v_atime; | |
1470 | iap->ia_mtime.tv_nsec = 0; | |
1471 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1472 | } |
1473 | ||
1da177e4 | 1474 | set_attr: |
5c002b3b BF |
1475 | err2 = nfsd_create_setattr(rqstp, resfhp, iap); |
1476 | if (err2) | |
1477 | err = err2; | |
f193fbab | 1478 | |
463c3197 | 1479 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
f193fbab YT |
1480 | /* |
1481 | * Update the filehandle to get the new inode info. | |
1482 | */ | |
1483 | if (!err) | |
1484 | err = fh_update(resfhp); | |
1da177e4 LT |
1485 | |
1486 | out: | |
1487 | fh_unlock(fhp); | |
1488 | if (dchild && !IS_ERR(dchild)) | |
1489 | dput(dchild); | |
1490 | return err; | |
1491 | ||
1492 | out_nfserr: | |
6264d69d | 1493 | err = nfserrno(host_err); |
1da177e4 LT |
1494 | goto out; |
1495 | } | |
1496 | #endif /* CONFIG_NFSD_V3 */ | |
1497 | ||
1498 | /* | |
1499 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1500 | * fits into the buffer. On return, it contains the true length. | |
1501 | * N.B. After this call fhp needs an fh_put | |
1502 | */ | |
6264d69d | 1503 | __be32 |
1da177e4 LT |
1504 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1505 | { | |
1506 | struct dentry *dentry; | |
1507 | struct inode *inode; | |
1508 | mm_segment_t oldfs; | |
6264d69d AV |
1509 | __be32 err; |
1510 | int host_err; | |
1da177e4 | 1511 | |
8837abca | 1512 | err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP); |
1da177e4 LT |
1513 | if (err) |
1514 | goto out; | |
1515 | ||
1516 | dentry = fhp->fh_dentry; | |
1517 | inode = dentry->d_inode; | |
1518 | ||
1519 | err = nfserr_inval; | |
acfa4380 | 1520 | if (!inode->i_op->readlink) |
1da177e4 LT |
1521 | goto out; |
1522 | ||
54775491 | 1523 | touch_atime(fhp->fh_export->ex_path.mnt, dentry); |
1da177e4 LT |
1524 | /* N.B. Why does this call need a get_fs()?? |
1525 | * Remove the set_fs and watch the fireworks:-) --okir | |
1526 | */ | |
1527 | ||
1528 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1529 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1530 | set_fs(oldfs); |
1531 | ||
6264d69d | 1532 | if (host_err < 0) |
1da177e4 | 1533 | goto out_nfserr; |
6264d69d | 1534 | *lenp = host_err; |
1da177e4 LT |
1535 | err = 0; |
1536 | out: | |
1537 | return err; | |
1538 | ||
1539 | out_nfserr: | |
6264d69d | 1540 | err = nfserrno(host_err); |
1da177e4 LT |
1541 | goto out; |
1542 | } | |
1543 | ||
1544 | /* | |
1545 | * Create a symlink and look up its inode | |
1546 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1547 | */ | |
6264d69d | 1548 | __be32 |
1da177e4 LT |
1549 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1550 | char *fname, int flen, | |
1551 | char *path, int plen, | |
1552 | struct svc_fh *resfhp, | |
1553 | struct iattr *iap) | |
1554 | { | |
1555 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1556 | __be32 err, cerr; |
1557 | int host_err; | |
1da177e4 LT |
1558 | |
1559 | err = nfserr_noent; | |
1560 | if (!flen || !plen) | |
1561 | goto out; | |
1562 | err = nfserr_exist; | |
1563 | if (isdotent(fname, flen)) | |
1564 | goto out; | |
1565 | ||
8837abca | 1566 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1567 | if (err) |
1568 | goto out; | |
1569 | fh_lock(fhp); | |
1570 | dentry = fhp->fh_dentry; | |
1571 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1572 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1573 | if (IS_ERR(dnew)) |
1574 | goto out_nfserr; | |
1575 | ||
75c3f29d DH |
1576 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1577 | if (host_err) | |
1578 | goto out_nfserr; | |
1579 | ||
1da177e4 LT |
1580 | if (unlikely(path[plen] != 0)) { |
1581 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1582 | if (path_alloced == NULL) | |
6264d69d | 1583 | host_err = -ENOMEM; |
1da177e4 LT |
1584 | else { |
1585 | strncpy(path_alloced, path, plen); | |
1586 | path_alloced[plen] = 0; | |
db2e747b | 1587 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced); |
1da177e4 LT |
1588 | kfree(path_alloced); |
1589 | } | |
1590 | } else | |
db2e747b | 1591 | host_err = vfs_symlink(dentry->d_inode, dnew, path); |
1da177e4 | 1592 | |
6264d69d | 1593 | if (!host_err) { |
1da177e4 | 1594 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1595 | host_err = nfsd_sync_dir(dentry); |
1596 | } | |
1597 | err = nfserrno(host_err); | |
1da177e4 LT |
1598 | fh_unlock(fhp); |
1599 | ||
75c3f29d DH |
1600 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
1601 | ||
1da177e4 LT |
1602 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); |
1603 | dput(dnew); | |
1604 | if (err==0) err = cerr; | |
1605 | out: | |
1606 | return err; | |
1607 | ||
1608 | out_nfserr: | |
6264d69d | 1609 | err = nfserrno(host_err); |
1da177e4 LT |
1610 | goto out; |
1611 | } | |
1612 | ||
1613 | /* | |
1614 | * Create a hardlink | |
1615 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1616 | */ | |
6264d69d | 1617 | __be32 |
1da177e4 LT |
1618 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1619 | char *name, int len, struct svc_fh *tfhp) | |
1620 | { | |
1621 | struct dentry *ddir, *dnew, *dold; | |
1622 | struct inode *dirp, *dest; | |
6264d69d AV |
1623 | __be32 err; |
1624 | int host_err; | |
1da177e4 | 1625 | |
8837abca | 1626 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1627 | if (err) |
1628 | goto out; | |
8837abca | 1629 | err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP); |
1da177e4 LT |
1630 | if (err) |
1631 | goto out; | |
1632 | ||
1633 | err = nfserr_perm; | |
1634 | if (!len) | |
1635 | goto out; | |
1636 | err = nfserr_exist; | |
1637 | if (isdotent(name, len)) | |
1638 | goto out; | |
1639 | ||
12fd3520 | 1640 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1641 | ddir = ffhp->fh_dentry; |
1642 | dirp = ddir->d_inode; | |
1643 | ||
1644 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1645 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1646 | if (IS_ERR(dnew)) |
1647 | goto out_nfserr; | |
1648 | ||
1649 | dold = tfhp->fh_dentry; | |
1650 | dest = dold->d_inode; | |
1651 | ||
75c3f29d DH |
1652 | host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt); |
1653 | if (host_err) { | |
1654 | err = nfserrno(host_err); | |
1655 | goto out_dput; | |
1656 | } | |
6264d69d AV |
1657 | host_err = vfs_link(dold, dirp, dnew); |
1658 | if (!host_err) { | |
1da177e4 | 1659 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1660 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1661 | write_inode_now(dest, 1); |
1662 | } | |
6264d69d | 1663 | err = 0; |
1da177e4 | 1664 | } else { |
6264d69d | 1665 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1666 | err = nfserr_acces; |
1667 | else | |
6264d69d | 1668 | err = nfserrno(host_err); |
1da177e4 | 1669 | } |
75c3f29d DH |
1670 | mnt_drop_write(tfhp->fh_export->ex_path.mnt); |
1671 | out_dput: | |
1da177e4 | 1672 | dput(dnew); |
270d56e5 DR |
1673 | out_unlock: |
1674 | fh_unlock(ffhp); | |
1da177e4 LT |
1675 | out: |
1676 | return err; | |
1677 | ||
1678 | out_nfserr: | |
6264d69d | 1679 | err = nfserrno(host_err); |
270d56e5 | 1680 | goto out_unlock; |
1da177e4 LT |
1681 | } |
1682 | ||
1683 | /* | |
1684 | * Rename a file | |
1685 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1686 | */ | |
6264d69d | 1687 | __be32 |
1da177e4 LT |
1688 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1689 | struct svc_fh *tfhp, char *tname, int tlen) | |
1690 | { | |
1691 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1692 | struct inode *fdir, *tdir; | |
6264d69d AV |
1693 | __be32 err; |
1694 | int host_err; | |
1da177e4 | 1695 | |
8837abca | 1696 | err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1697 | if (err) |
1698 | goto out; | |
8837abca | 1699 | err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE); |
1da177e4 LT |
1700 | if (err) |
1701 | goto out; | |
1702 | ||
1703 | fdentry = ffhp->fh_dentry; | |
1704 | fdir = fdentry->d_inode; | |
1705 | ||
1706 | tdentry = tfhp->fh_dentry; | |
1707 | tdir = tdentry->d_inode; | |
1708 | ||
1709 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1710 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1711 | goto out; |
1712 | ||
1713 | err = nfserr_perm; | |
1714 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1715 | goto out; | |
1716 | ||
1717 | /* cannot use fh_lock as we need deadlock protective ordering | |
1718 | * so do it by hand */ | |
1719 | trap = lock_rename(tdentry, fdentry); | |
1720 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1721 | fill_pre_wcc(ffhp); | |
1722 | fill_pre_wcc(tfhp); | |
1723 | ||
1724 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1725 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1726 | if (IS_ERR(odentry)) |
1727 | goto out_nfserr; | |
1728 | ||
6264d69d | 1729 | host_err = -ENOENT; |
1da177e4 LT |
1730 | if (!odentry->d_inode) |
1731 | goto out_dput_old; | |
6264d69d | 1732 | host_err = -EINVAL; |
1da177e4 LT |
1733 | if (odentry == trap) |
1734 | goto out_dput_old; | |
1735 | ||
1736 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1737 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1738 | if (IS_ERR(ndentry)) |
1739 | goto out_dput_old; | |
6264d69d | 1740 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1741 | if (ndentry == trap) |
1742 | goto out_dput_new; | |
1743 | ||
9079b1eb | 1744 | if (svc_msnfs(ffhp) && |
1da177e4 LT |
1745 | ((atomic_read(&odentry->d_count) > 1) |
1746 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1747 | host_err = -EPERM; |
9079b1eb DH |
1748 | goto out_dput_new; |
1749 | } | |
1750 | ||
1751 | host_err = -EXDEV; | |
1752 | if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt) | |
1753 | goto out_dput_new; | |
1754 | host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt); | |
1755 | if (host_err) | |
1756 | goto out_dput_new; | |
1757 | ||
6264d69d AV |
1758 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1759 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1760 | host_err = nfsd_sync_dir(tdentry); | |
1761 | if (!host_err) | |
1762 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1763 | } |
1764 | ||
9079b1eb DH |
1765 | mnt_drop_write(ffhp->fh_export->ex_path.mnt); |
1766 | ||
1da177e4 LT |
1767 | out_dput_new: |
1768 | dput(ndentry); | |
1769 | out_dput_old: | |
1770 | dput(odentry); | |
1771 | out_nfserr: | |
6264d69d | 1772 | err = nfserrno(host_err); |
1da177e4 LT |
1773 | |
1774 | /* we cannot reply on fh_unlock on the two filehandles, | |
1775 | * as that would do the wrong thing if the two directories | |
1776 | * were the same, so again we do it by hand | |
1777 | */ | |
1778 | fill_post_wcc(ffhp); | |
1779 | fill_post_wcc(tfhp); | |
1780 | unlock_rename(tdentry, fdentry); | |
1781 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1782 | ||
1783 | out: | |
1784 | return err; | |
1785 | } | |
1786 | ||
1787 | /* | |
1788 | * Unlink a file or directory | |
1789 | * N.B. After this call fhp needs an fh_put | |
1790 | */ | |
6264d69d | 1791 | __be32 |
1da177e4 LT |
1792 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1793 | char *fname, int flen) | |
1794 | { | |
1795 | struct dentry *dentry, *rdentry; | |
1796 | struct inode *dirp; | |
6264d69d AV |
1797 | __be32 err; |
1798 | int host_err; | |
1da177e4 LT |
1799 | |
1800 | err = nfserr_acces; | |
1801 | if (!flen || isdotent(fname, flen)) | |
1802 | goto out; | |
8837abca | 1803 | err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE); |
1da177e4 LT |
1804 | if (err) |
1805 | goto out; | |
1806 | ||
12fd3520 | 1807 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1808 | dentry = fhp->fh_dentry; |
1809 | dirp = dentry->d_inode; | |
1810 | ||
1811 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1812 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1813 | if (IS_ERR(rdentry)) |
1814 | goto out_nfserr; | |
1815 | ||
1816 | if (!rdentry->d_inode) { | |
1817 | dput(rdentry); | |
1818 | err = nfserr_noent; | |
1819 | goto out; | |
1820 | } | |
1821 | ||
1822 | if (!type) | |
1823 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1824 | ||
0622753b DH |
1825 | host_err = mnt_want_write(fhp->fh_export->ex_path.mnt); |
1826 | if (host_err) | |
1827 | goto out_nfserr; | |
1828 | ||
1da177e4 LT |
1829 | if (type != S_IFDIR) { /* It's UNLINK */ |
1830 | #ifdef MSNFS | |
1831 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1832 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1833 | host_err = -EPERM; |
1da177e4 LT |
1834 | } else |
1835 | #endif | |
6264d69d | 1836 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1837 | } else { /* It's RMDIR */ |
6264d69d | 1838 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1839 | } |
1840 | ||
1841 | dput(rdentry); | |
1842 | ||
6264d69d | 1843 | if (host_err) |
0622753b | 1844 | goto out_drop; |
6264d69d AV |
1845 | if (EX_ISSYNC(fhp->fh_export)) |
1846 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 | 1847 | |
0622753b DH |
1848 | out_drop: |
1849 | mnt_drop_write(fhp->fh_export->ex_path.mnt); | |
1da177e4 | 1850 | out_nfserr: |
6264d69d | 1851 | err = nfserrno(host_err); |
f193fbab YT |
1852 | out: |
1853 | return err; | |
1da177e4 LT |
1854 | } |
1855 | ||
14f7dd63 DW |
1856 | /* |
1857 | * We do this buffering because we must not call back into the file | |
1858 | * system's ->lookup() method from the filldir callback. That may well | |
1859 | * deadlock a number of file systems. | |
1860 | * | |
1861 | * This is based heavily on the implementation of same in XFS. | |
1862 | */ | |
1863 | struct buffered_dirent { | |
1864 | u64 ino; | |
1865 | loff_t offset; | |
1866 | int namlen; | |
1867 | unsigned int d_type; | |
1868 | char name[]; | |
1869 | }; | |
1870 | ||
1871 | struct readdir_data { | |
1872 | char *dirent; | |
1873 | size_t used; | |
53c9c5c0 | 1874 | int full; |
14f7dd63 DW |
1875 | }; |
1876 | ||
1877 | static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen, | |
1878 | loff_t offset, u64 ino, unsigned int d_type) | |
1879 | { | |
1880 | struct readdir_data *buf = __buf; | |
1881 | struct buffered_dirent *de = (void *)(buf->dirent + buf->used); | |
1882 | unsigned int reclen; | |
1883 | ||
1884 | reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64)); | |
53c9c5c0 AV |
1885 | if (buf->used + reclen > PAGE_SIZE) { |
1886 | buf->full = 1; | |
14f7dd63 | 1887 | return -EINVAL; |
53c9c5c0 | 1888 | } |
14f7dd63 DW |
1889 | |
1890 | de->namlen = namlen; | |
1891 | de->offset = offset; | |
1892 | de->ino = ino; | |
1893 | de->d_type = d_type; | |
1894 | memcpy(de->name, name, namlen); | |
1895 | buf->used += reclen; | |
1896 | ||
1897 | return 0; | |
1898 | } | |
1899 | ||
2f9092e1 DW |
1900 | static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func, |
1901 | struct readdir_cd *cdp, loff_t *offsetp) | |
2628b766 | 1902 | { |
14f7dd63 DW |
1903 | struct readdir_data buf; |
1904 | struct buffered_dirent *de; | |
2628b766 | 1905 | int host_err; |
14f7dd63 DW |
1906 | int size; |
1907 | loff_t offset; | |
2628b766 | 1908 | |
14f7dd63 DW |
1909 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); |
1910 | if (!buf.dirent) | |
2f9092e1 | 1911 | return nfserrno(-ENOMEM); |
14f7dd63 DW |
1912 | |
1913 | offset = *offsetp; | |
2628b766 | 1914 | |
14f7dd63 | 1915 | while (1) { |
2f9092e1 | 1916 | struct inode *dir_inode = file->f_path.dentry->d_inode; |
14f7dd63 DW |
1917 | unsigned int reclen; |
1918 | ||
b726e923 | 1919 | cdp->err = nfserr_eof; /* will be cleared on successful read */ |
14f7dd63 | 1920 | buf.used = 0; |
53c9c5c0 | 1921 | buf.full = 0; |
14f7dd63 DW |
1922 | |
1923 | host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf); | |
53c9c5c0 AV |
1924 | if (buf.full) |
1925 | host_err = 0; | |
1926 | ||
1927 | if (host_err < 0) | |
14f7dd63 DW |
1928 | break; |
1929 | ||
1930 | size = buf.used; | |
1931 | ||
1932 | if (!size) | |
1933 | break; | |
1934 | ||
2f9092e1 DW |
1935 | /* |
1936 | * Various filldir functions may end up calling back into | |
1937 | * lookup_one_len() and the file system's ->lookup() method. | |
1938 | * These expect i_mutex to be held, as it would within readdir. | |
1939 | */ | |
1940 | host_err = mutex_lock_killable(&dir_inode->i_mutex); | |
1941 | if (host_err) | |
1942 | break; | |
1943 | ||
14f7dd63 DW |
1944 | de = (struct buffered_dirent *)buf.dirent; |
1945 | while (size > 0) { | |
1946 | offset = de->offset; | |
1947 | ||
1948 | if (func(cdp, de->name, de->namlen, de->offset, | |
1949 | de->ino, de->d_type)) | |
2f9092e1 | 1950 | break; |
14f7dd63 DW |
1951 | |
1952 | if (cdp->err != nfs_ok) | |
2f9092e1 | 1953 | break; |
14f7dd63 DW |
1954 | |
1955 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
1956 | sizeof(u64)); | |
1957 | size -= reclen; | |
1958 | de = (struct buffered_dirent *)((char *)de + reclen); | |
1959 | } | |
2f9092e1 DW |
1960 | mutex_unlock(&dir_inode->i_mutex); |
1961 | if (size > 0) /* We bailed out early */ | |
1962 | break; | |
1963 | ||
c002a6c7 | 1964 | offset = vfs_llseek(file, 0, SEEK_CUR); |
14f7dd63 DW |
1965 | } |
1966 | ||
14f7dd63 | 1967 | free_page((unsigned long)(buf.dirent)); |
2628b766 DW |
1968 | |
1969 | if (host_err) | |
1970 | return nfserrno(host_err); | |
14f7dd63 DW |
1971 | |
1972 | *offsetp = offset; | |
1973 | return cdp->err; | |
2628b766 DW |
1974 | } |
1975 | ||
1da177e4 LT |
1976 | /* |
1977 | * Read entries from a directory. | |
1978 | * The NFSv3/4 verifier we ignore for now. | |
1979 | */ | |
6264d69d | 1980 | __be32 |
1da177e4 | 1981 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1982 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1983 | { |
6264d69d | 1984 | __be32 err; |
1da177e4 LT |
1985 | struct file *file; |
1986 | loff_t offset = *offsetp; | |
1987 | ||
8837abca | 1988 | err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file); |
1da177e4 LT |
1989 | if (err) |
1990 | goto out; | |
1991 | ||
1992 | offset = vfs_llseek(file, offset, 0); | |
1993 | if (offset < 0) { | |
1994 | err = nfserrno((int)offset); | |
1995 | goto out_close; | |
1996 | } | |
1997 | ||
14f7dd63 | 1998 | err = nfsd_buffered_readdir(file, func, cdp, offsetp); |
1da177e4 LT |
1999 | |
2000 | if (err == nfserr_eof || err == nfserr_toosmall) | |
2001 | err = nfs_ok; /* can still be found in ->err */ | |
2002 | out_close: | |
2003 | nfsd_close(file); | |
2004 | out: | |
2005 | return err; | |
2006 | } | |
2007 | ||
2008 | /* | |
2009 | * Get file system stats | |
2010 | * N.B. After this call fhp needs an fh_put | |
2011 | */ | |
6264d69d | 2012 | __be32 |
04716e66 | 2013 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access) |
1da177e4 | 2014 | { |
04716e66 | 2015 | __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access); |
726c3342 | 2016 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
2017 | err = nfserr_io; |
2018 | return err; | |
2019 | } | |
2020 | ||
c7d51402 | 2021 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 2022 | { |
c7d51402 | 2023 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
2024 | } |
2025 | ||
1da177e4 LT |
2026 | /* |
2027 | * Check for a user's access permissions to this inode. | |
2028 | */ | |
6264d69d | 2029 | __be32 |
0ec757df BF |
2030 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
2031 | struct dentry *dentry, int acc) | |
1da177e4 LT |
2032 | { |
2033 | struct inode *inode = dentry->d_inode; | |
2034 | int err; | |
2035 | ||
8837abca | 2036 | if (acc == NFSD_MAY_NOP) |
1da177e4 LT |
2037 | return 0; |
2038 | #if 0 | |
2039 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
2040 | acc, | |
8837abca MS |
2041 | (acc & NFSD_MAY_READ)? " read" : "", |
2042 | (acc & NFSD_MAY_WRITE)? " write" : "", | |
2043 | (acc & NFSD_MAY_EXEC)? " exec" : "", | |
2044 | (acc & NFSD_MAY_SATTR)? " sattr" : "", | |
2045 | (acc & NFSD_MAY_TRUNC)? " trunc" : "", | |
2046 | (acc & NFSD_MAY_LOCK)? " lock" : "", | |
2047 | (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1da177e4 LT |
2048 | inode->i_mode, |
2049 | IS_IMMUTABLE(inode)? " immut" : "", | |
2050 | IS_APPEND(inode)? " append" : "", | |
2c463e95 | 2051 | __mnt_is_readonly(exp->ex_path.mnt)? " ro" : ""); |
1da177e4 | 2052 | dprintk(" owner %d/%d user %d/%d\n", |
5cc0a840 | 2053 | inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid()); |
1da177e4 LT |
2054 | #endif |
2055 | ||
2056 | /* Normally we reject any write/sattr etc access on a read-only file | |
2057 | * system. But if it is IRIX doing check on write-access for a | |
2058 | * device special file, we ignore rofs. | |
2059 | */ | |
8837abca MS |
2060 | if (!(acc & NFSD_MAY_LOCAL_ACCESS)) |
2061 | if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) { | |
2c463e95 DH |
2062 | if (exp_rdonly(rqstp, exp) || |
2063 | __mnt_is_readonly(exp->ex_path.mnt)) | |
1da177e4 | 2064 | return nfserr_rofs; |
8837abca | 2065 | if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode)) |
1da177e4 LT |
2066 | return nfserr_perm; |
2067 | } | |
8837abca | 2068 | if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode)) |
1da177e4 LT |
2069 | return nfserr_perm; |
2070 | ||
8837abca | 2071 | if (acc & NFSD_MAY_LOCK) { |
1da177e4 LT |
2072 | /* If we cannot rely on authentication in NLM requests, |
2073 | * just allow locks, otherwise require read permission, or | |
2074 | * ownership | |
2075 | */ | |
2076 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
2077 | return 0; | |
2078 | else | |
8837abca | 2079 | acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE; |
1da177e4 LT |
2080 | } |
2081 | /* | |
2082 | * The file owner always gets access permission for accesses that | |
2083 | * would normally be checked at open time. This is to make | |
2084 | * file access work even when the client has done a fchmod(fd, 0). | |
2085 | * | |
2086 | * However, `cp foo bar' should fail nevertheless when bar is | |
2087 | * readonly. A sensible way to do this might be to reject all | |
2088 | * attempts to truncate a read-only file, because a creat() call | |
2089 | * always implies file truncation. | |
2090 | * ... but this isn't really fair. A process may reasonably call | |
2091 | * ftruncate on an open file descriptor on a file with perm 000. | |
2092 | * We must trust the client to do permission checking - using "ACCESS" | |
2093 | * with NFSv3. | |
2094 | */ | |
8837abca | 2095 | if ((acc & NFSD_MAY_OWNER_OVERRIDE) && |
5cc0a840 | 2096 | inode->i_uid == current_fsuid()) |
1da177e4 LT |
2097 | return 0; |
2098 | ||
8837abca | 2099 | /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */ |
f419a2e3 | 2100 | err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC)); |
1da177e4 LT |
2101 | |
2102 | /* Allow read access to binaries even when mode 111 */ | |
2103 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
8837abca | 2104 | acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE)) |
f419a2e3 | 2105 | err = inode_permission(inode, MAY_EXEC); |
1da177e4 LT |
2106 | |
2107 | return err? nfserrno(err) : 0; | |
2108 | } | |
2109 | ||
2110 | void | |
2111 | nfsd_racache_shutdown(void) | |
2112 | { | |
54a66e54 JL |
2113 | struct raparms *raparm, *last_raparm; |
2114 | unsigned int i; | |
2115 | ||
1da177e4 | 2116 | dprintk("nfsd: freeing readahead buffers.\n"); |
54a66e54 JL |
2117 | |
2118 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
2119 | raparm = raparm_hash[i].pb_head; | |
2120 | while(raparm) { | |
2121 | last_raparm = raparm; | |
2122 | raparm = raparm->p_next; | |
2123 | kfree(last_raparm); | |
2124 | } | |
2125 | raparm_hash[i].pb_head = NULL; | |
2126 | } | |
1da177e4 LT |
2127 | } |
2128 | /* | |
2129 | * Initialize readahead param cache | |
2130 | */ | |
2131 | int | |
2132 | nfsd_racache_init(int cache_size) | |
2133 | { | |
2134 | int i; | |
fce1456a GB |
2135 | int j = 0; |
2136 | int nperbucket; | |
54a66e54 | 2137 | struct raparms **raparm = NULL; |
1da177e4 | 2138 | |
fce1456a | 2139 | |
54a66e54 | 2140 | if (raparm_hash[0].pb_head) |
1da177e4 | 2141 | return 0; |
54a66e54 JL |
2142 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
2143 | if (nperbucket < 2) | |
2144 | nperbucket = 2; | |
2145 | cache_size = nperbucket * RAPARM_HASH_SIZE; | |
4b3bb06b YB |
2146 | |
2147 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
54a66e54 JL |
2148 | |
2149 | for (i = 0; i < RAPARM_HASH_SIZE; i++) { | |
4b3bb06b | 2150 | spin_lock_init(&raparm_hash[i].pb_lock); |
54a66e54 JL |
2151 | |
2152 | raparm = &raparm_hash[i].pb_head; | |
2153 | for (j = 0; j < nperbucket; j++) { | |
2154 | *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL); | |
2155 | if (!*raparm) | |
2156 | goto out_nomem; | |
2157 | raparm = &(*raparm)->p_next; | |
2158 | } | |
2159 | *raparm = NULL; | |
4b3bb06b YB |
2160 | } |
2161 | ||
1da177e4 LT |
2162 | nfsdstats.ra_size = cache_size; |
2163 | return 0; | |
54a66e54 JL |
2164 | |
2165 | out_nomem: | |
2166 | dprintk("nfsd: kmalloc failed, freeing readahead buffers\n"); | |
2167 | nfsd_racache_shutdown(); | |
2168 | return -ENOMEM; | |
1da177e4 | 2169 | } |
a257cdd0 AG |
2170 | |
2171 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
2172 | struct posix_acl * | |
2173 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
2174 | { | |
2175 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2176 | char *name; | |
2177 | void *value = NULL; | |
2178 | ssize_t size; | |
2179 | struct posix_acl *acl; | |
2180 | ||
5be196e5 CH |
2181 | if (!IS_POSIXACL(inode)) |
2182 | return ERR_PTR(-EOPNOTSUPP); | |
2183 | ||
2184 | switch (type) { | |
2185 | case ACL_TYPE_ACCESS: | |
2186 | name = POSIX_ACL_XATTR_ACCESS; | |
2187 | break; | |
2188 | case ACL_TYPE_DEFAULT: | |
2189 | name = POSIX_ACL_XATTR_DEFAULT; | |
2190 | break; | |
2191 | default: | |
a257cdd0 | 2192 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
2193 | } |
2194 | ||
5be196e5 CH |
2195 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
2196 | if (size < 0) | |
2197 | return ERR_PTR(size); | |
a257cdd0 | 2198 | |
a257cdd0 | 2199 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
2200 | kfree(value); |
2201 | return acl; | |
2202 | } | |
2203 | ||
2204 | int | |
2205 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
2206 | { | |
2207 | struct inode *inode = fhp->fh_dentry->d_inode; | |
2208 | char *name; | |
2209 | void *value = NULL; | |
2210 | size_t size; | |
2211 | int error; | |
2212 | ||
acfa4380 | 2213 | if (!IS_POSIXACL(inode) || |
a257cdd0 AG |
2214 | !inode->i_op->setxattr || !inode->i_op->removexattr) |
2215 | return -EOPNOTSUPP; | |
2216 | switch(type) { | |
2217 | case ACL_TYPE_ACCESS: | |
334a13ec | 2218 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2219 | break; |
2220 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2221 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2222 | break; |
2223 | default: | |
2224 | return -EOPNOTSUPP; | |
2225 | } | |
2226 | ||
2227 | if (acl && acl->a_count) { | |
334a13ec | 2228 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2229 | value = kmalloc(size, GFP_KERNEL); |
2230 | if (!value) | |
2231 | return -ENOMEM; | |
9ccfc29c FM |
2232 | error = posix_acl_to_xattr(acl, value, size); |
2233 | if (error < 0) | |
a257cdd0 | 2234 | goto getout; |
9ccfc29c | 2235 | size = error; |
a257cdd0 AG |
2236 | } else |
2237 | size = 0; | |
2238 | ||
18f335af DH |
2239 | error = mnt_want_write(fhp->fh_export->ex_path.mnt); |
2240 | if (error) | |
2241 | goto getout; | |
a257cdd0 | 2242 | if (size) |
5be196e5 | 2243 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2244 | else { |
2245 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2246 | error = 0; | |
2247 | else { | |
5be196e5 | 2248 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2249 | if (error == -ENODATA) |
2250 | error = 0; | |
2251 | } | |
2252 | } | |
18f335af | 2253 | mnt_drop_write(fhp->fh_export->ex_path.mnt); |
a257cdd0 AG |
2254 | |
2255 | getout: | |
2256 | kfree(value); | |
2257 | return error; | |
2258 | } | |
2259 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |