]>
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 | ||
1da177e4 | 86 | static struct raparms * raparml; |
fce1456a GB |
87 | #define RAPARM_HASH_BITS 4 |
88 | #define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS) | |
89 | #define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1) | |
90 | static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE]; | |
1da177e4 LT |
91 | |
92 | /* | |
93 | * Called from nfsd_lookup and encode_dirent. Check if we have crossed | |
94 | * a mount point. | |
e0bb89ef | 95 | * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged, |
1da177e4 LT |
96 | * or nfs_ok having possibly changed *dpp and *expp |
97 | */ | |
98 | int | |
99 | nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp, | |
100 | struct svc_export **expp) | |
101 | { | |
102 | struct svc_export *exp = *expp, *exp2 = NULL; | |
103 | struct dentry *dentry = *dpp; | |
104 | struct vfsmount *mnt = mntget(exp->ex_mnt); | |
105 | struct dentry *mounts = dget(dentry); | |
6264d69d | 106 | int err = 0; |
1da177e4 LT |
107 | |
108 | while (follow_down(&mnt,&mounts)&&d_mountpoint(mounts)); | |
109 | ||
0989a788 | 110 | exp2 = rqst_exp_get_by_name(rqstp, mnt, mounts); |
1da177e4 | 111 | if (IS_ERR(exp2)) { |
a1033be7 NB |
112 | if (PTR_ERR(exp2) != -ENOENT) |
113 | err = PTR_ERR(exp2); | |
1da177e4 LT |
114 | dput(mounts); |
115 | mntput(mnt); | |
116 | goto out; | |
117 | } | |
5d3dbbea | 118 | if ((exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) { |
1da177e4 LT |
119 | /* successfully crossed mount point */ |
120 | exp_put(exp); | |
121 | *expp = exp2; | |
122 | dput(dentry); | |
123 | *dpp = mounts; | |
124 | } else { | |
5d3dbbea | 125 | exp_put(exp2); |
1da177e4 LT |
126 | dput(mounts); |
127 | } | |
128 | mntput(mnt); | |
129 | out: | |
130 | return err; | |
131 | } | |
132 | ||
6264d69d | 133 | __be32 |
6c0a654d BF |
134 | nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp, |
135 | const char *name, int len, | |
136 | struct svc_export **exp_ret, struct dentry **dentry_ret) | |
1da177e4 LT |
137 | { |
138 | struct svc_export *exp; | |
139 | struct dentry *dparent; | |
140 | struct dentry *dentry; | |
6264d69d AV |
141 | __be32 err; |
142 | int host_err; | |
1da177e4 LT |
143 | |
144 | dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name); | |
145 | ||
146 | /* Obtain dentry and export. */ | |
147 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_EXEC); | |
148 | if (err) | |
149 | return err; | |
150 | ||
151 | dparent = fhp->fh_dentry; | |
152 | exp = fhp->fh_export; | |
153 | exp_get(exp); | |
154 | ||
1da177e4 LT |
155 | /* Lookup the name, but don't follow links */ |
156 | if (isdotent(name, len)) { | |
157 | if (len==1) | |
158 | dentry = dget(dparent); | |
159 | else if (dparent != exp->ex_dentry) { | |
160 | dentry = dget_parent(dparent); | |
161 | } else if (!EX_NOHIDE(exp)) | |
162 | dentry = dget(dparent); /* .. == . just like at / */ | |
163 | else { | |
164 | /* checking mountpoint crossing is very different when stepping up */ | |
165 | struct svc_export *exp2 = NULL; | |
166 | struct dentry *dp; | |
167 | struct vfsmount *mnt = mntget(exp->ex_mnt); | |
168 | dentry = dget(dparent); | |
169 | while(dentry == mnt->mnt_root && follow_up(&mnt, &dentry)) | |
170 | ; | |
171 | dp = dget_parent(dentry); | |
172 | dput(dentry); | |
173 | dentry = dp; | |
174 | ||
0989a788 | 175 | exp2 = rqst_exp_parent(rqstp, mnt, dentry); |
2d3bb252 BF |
176 | if (PTR_ERR(exp2) == -ENOENT) { |
177 | dput(dentry); | |
178 | dentry = dget(dparent); | |
179 | } else if (IS_ERR(exp2)) { | |
6264d69d | 180 | host_err = PTR_ERR(exp2); |
1da177e4 LT |
181 | dput(dentry); |
182 | mntput(mnt); | |
183 | goto out_nfserr; | |
1da177e4 LT |
184 | } else { |
185 | exp_put(exp); | |
186 | exp = exp2; | |
187 | } | |
188 | mntput(mnt); | |
189 | } | |
190 | } else { | |
191 | fh_lock(fhp); | |
192 | dentry = lookup_one_len(name, dparent, len); | |
6264d69d | 193 | host_err = PTR_ERR(dentry); |
1da177e4 LT |
194 | if (IS_ERR(dentry)) |
195 | goto out_nfserr; | |
196 | /* | |
197 | * check if we have crossed a mount point ... | |
198 | */ | |
199 | if (d_mountpoint(dentry)) { | |
6264d69d | 200 | if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) { |
1da177e4 LT |
201 | dput(dentry); |
202 | goto out_nfserr; | |
203 | } | |
204 | } | |
205 | } | |
6c0a654d BF |
206 | *dentry_ret = dentry; |
207 | *exp_ret = exp; | |
208 | return 0; | |
209 | ||
210 | out_nfserr: | |
211 | exp_put(exp); | |
212 | return nfserrno(host_err); | |
213 | } | |
214 | ||
215 | /* | |
216 | * Look up one component of a pathname. | |
217 | * N.B. After this call _both_ fhp and resfh need an fh_put | |
218 | * | |
219 | * If the lookup would cross a mountpoint, and the mounted filesystem | |
220 | * is exported to the client with NFSEXP_NOHIDE, then the lookup is | |
221 | * accepted as it stands and the mounted directory is | |
222 | * returned. Otherwise the covered directory is returned. | |
223 | * NOTE: this mountpoint crossing is not supported properly by all | |
224 | * clients and is explicitly disallowed for NFSv3 | |
225 | * NeilBrown <neilb@cse.unsw.edu.au> | |
226 | */ | |
227 | __be32 | |
228 | nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name, | |
229 | int len, struct svc_fh *resfh) | |
230 | { | |
231 | struct svc_export *exp; | |
232 | struct dentry *dentry; | |
233 | __be32 err; | |
234 | ||
235 | err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry); | |
236 | if (err) | |
237 | return err; | |
32c1eb0c AA |
238 | err = check_nfsd_access(exp, rqstp); |
239 | if (err) | |
240 | goto out; | |
1da177e4 LT |
241 | /* |
242 | * Note: we compose the file handle now, but as the | |
243 | * dentry may be negative, it may need to be updated. | |
244 | */ | |
245 | err = fh_compose(resfh, exp, dentry, fhp); | |
246 | if (!err && !dentry->d_inode) | |
247 | err = nfserr_noent; | |
32c1eb0c | 248 | out: |
1da177e4 | 249 | dput(dentry); |
1da177e4 LT |
250 | exp_put(exp); |
251 | return err; | |
1da177e4 LT |
252 | } |
253 | ||
6c0a654d | 254 | |
1da177e4 LT |
255 | /* |
256 | * Set various file attributes. | |
257 | * N.B. After this call fhp needs an fh_put | |
258 | */ | |
6264d69d | 259 | __be32 |
1da177e4 LT |
260 | nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap, |
261 | int check_guard, time_t guardtime) | |
262 | { | |
263 | struct dentry *dentry; | |
264 | struct inode *inode; | |
265 | int accmode = MAY_SATTR; | |
266 | int ftype = 0; | |
267 | int imode; | |
6264d69d AV |
268 | __be32 err; |
269 | int host_err; | |
1da177e4 LT |
270 | int size_change = 0; |
271 | ||
272 | if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE)) | |
273 | accmode |= MAY_WRITE|MAY_OWNER_OVERRIDE; | |
274 | if (iap->ia_valid & ATTR_SIZE) | |
275 | ftype = S_IFREG; | |
276 | ||
277 | /* Get inode */ | |
278 | err = fh_verify(rqstp, fhp, ftype, accmode); | |
15b7a1b8 | 279 | if (err) |
1da177e4 LT |
280 | goto out; |
281 | ||
282 | dentry = fhp->fh_dentry; | |
283 | inode = dentry->d_inode; | |
284 | ||
15b7a1b8 N |
285 | /* Ignore any mode updates on symlinks */ |
286 | if (S_ISLNK(inode->i_mode)) | |
287 | iap->ia_valid &= ~ATTR_MODE; | |
288 | ||
289 | if (!iap->ia_valid) | |
290 | goto out; | |
291 | ||
9c85fca5 CH |
292 | /* |
293 | * NFSv2 does not differentiate between "set-[ac]time-to-now" | |
1da177e4 LT |
294 | * which only requires access, and "set-[ac]time-to-X" which |
295 | * requires ownership. | |
296 | * So if it looks like it might be "set both to the same time which | |
297 | * is close to now", and if inode_change_ok fails, then we | |
298 | * convert to "set to now" instead of "set to explicit time" | |
299 | * | |
300 | * We only call inode_change_ok as the last test as technically | |
301 | * it is not an interface that we should be using. It is only | |
302 | * valid if the filesystem does not define it's own i_op->setattr. | |
303 | */ | |
304 | #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET) | |
305 | #define MAX_TOUCH_TIME_ERROR (30*60) | |
9c85fca5 CH |
306 | if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET && |
307 | iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) { | |
308 | /* | |
309 | * Looks probable. | |
310 | * | |
311 | * Now just make sure time is in the right ballpark. | |
312 | * Solaris, at least, doesn't seem to care what the time | |
313 | * request is. We require it be within 30 minutes of now. | |
1da177e4 | 314 | */ |
9c85fca5 CH |
315 | time_t delta = iap->ia_atime.tv_sec - get_seconds(); |
316 | if (delta < 0) | |
317 | delta = -delta; | |
318 | if (delta < MAX_TOUCH_TIME_ERROR && | |
319 | inode_change_ok(inode, iap) != 0) { | |
320 | /* | |
321 | * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. | |
322 | * This will cause notify_change to set these times | |
323 | * to "now" | |
324 | */ | |
325 | iap->ia_valid &= ~BOTH_TIME_SET; | |
326 | } | |
1da177e4 LT |
327 | } |
328 | ||
9c85fca5 CH |
329 | /* |
330 | * The size case is special. | |
331 | * It changes the file as well as the attributes. | |
332 | */ | |
1da177e4 LT |
333 | if (iap->ia_valid & ATTR_SIZE) { |
334 | if (iap->ia_size < inode->i_size) { | |
0ec757df | 335 | err = nfsd_permission(rqstp, fhp->fh_export, dentry, MAY_TRUNC|MAY_OWNER_OVERRIDE); |
1da177e4 LT |
336 | if (err) |
337 | goto out; | |
338 | } | |
339 | ||
340 | /* | |
341 | * If we are changing the size of the file, then | |
342 | * we need to break all leases. | |
343 | */ | |
6264d69d AV |
344 | host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK); |
345 | if (host_err == -EWOULDBLOCK) | |
346 | host_err = -ETIMEDOUT; | |
347 | if (host_err) /* ENOMEM or EWOULDBLOCK */ | |
1da177e4 LT |
348 | goto out_nfserr; |
349 | ||
6264d69d AV |
350 | host_err = get_write_access(inode); |
351 | if (host_err) | |
1da177e4 LT |
352 | goto out_nfserr; |
353 | ||
354 | size_change = 1; | |
6264d69d AV |
355 | host_err = locks_verify_truncate(inode, NULL, iap->ia_size); |
356 | if (host_err) { | |
1da177e4 LT |
357 | put_write_access(inode); |
358 | goto out_nfserr; | |
359 | } | |
360 | DQUOT_INIT(inode); | |
361 | } | |
362 | ||
363 | imode = inode->i_mode; | |
364 | if (iap->ia_valid & ATTR_MODE) { | |
365 | iap->ia_mode &= S_IALLUGO; | |
366 | imode = iap->ia_mode |= (imode & ~S_IALLUGO); | |
367 | } | |
368 | ||
369 | /* Revoke setuid/setgid bit on chown/chgrp */ | |
370 | if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) | |
371 | iap->ia_valid |= ATTR_KILL_SUID; | |
372 | if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid) | |
373 | iap->ia_valid |= ATTR_KILL_SGID; | |
374 | ||
375 | /* Change the attributes. */ | |
376 | ||
377 | iap->ia_valid |= ATTR_CTIME; | |
378 | ||
379 | err = nfserr_notsync; | |
380 | if (!check_guard || guardtime == inode->i_ctime.tv_sec) { | |
381 | fh_lock(fhp); | |
6264d69d AV |
382 | host_err = notify_change(dentry, iap); |
383 | err = nfserrno(host_err); | |
1da177e4 LT |
384 | fh_unlock(fhp); |
385 | } | |
386 | if (size_change) | |
387 | put_write_access(inode); | |
388 | if (!err) | |
389 | if (EX_ISSYNC(fhp->fh_export)) | |
390 | write_inode_now(inode, 1); | |
391 | out: | |
392 | return err; | |
393 | ||
394 | out_nfserr: | |
6264d69d | 395 | err = nfserrno(host_err); |
1da177e4 LT |
396 | goto out; |
397 | } | |
398 | ||
5be196e5 CH |
399 | #if defined(CONFIG_NFSD_V2_ACL) || \ |
400 | defined(CONFIG_NFSD_V3_ACL) || \ | |
401 | defined(CONFIG_NFSD_V4) | |
402 | static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf) | |
403 | { | |
404 | ssize_t buflen; | |
5be196e5 CH |
405 | |
406 | buflen = vfs_getxattr(dentry, key, NULL, 0); | |
407 | if (buflen <= 0) | |
408 | return buflen; | |
1da177e4 | 409 | |
5be196e5 CH |
410 | *buf = kmalloc(buflen, GFP_KERNEL); |
411 | if (!*buf) | |
412 | return -ENOMEM; | |
413 | ||
b5872b0d | 414 | return vfs_getxattr(dentry, key, *buf, buflen); |
5be196e5 CH |
415 | } |
416 | #endif | |
417 | ||
418 | #if defined(CONFIG_NFSD_V4) | |
1da177e4 LT |
419 | static int |
420 | set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key) | |
421 | { | |
422 | int len; | |
423 | size_t buflen; | |
424 | char *buf = NULL; | |
425 | int error = 0; | |
1da177e4 LT |
426 | |
427 | buflen = posix_acl_xattr_size(pacl->a_count); | |
428 | buf = kmalloc(buflen, GFP_KERNEL); | |
429 | error = -ENOMEM; | |
430 | if (buf == NULL) | |
431 | goto out; | |
432 | ||
433 | len = posix_acl_to_xattr(pacl, buf, buflen); | |
434 | if (len < 0) { | |
435 | error = len; | |
436 | goto out; | |
437 | } | |
438 | ||
5be196e5 | 439 | error = vfs_setxattr(dentry, key, buf, len, 0); |
1da177e4 LT |
440 | out: |
441 | kfree(buf); | |
442 | return error; | |
443 | } | |
444 | ||
6264d69d | 445 | __be32 |
1da177e4 LT |
446 | nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp, |
447 | struct nfs4_acl *acl) | |
448 | { | |
6264d69d AV |
449 | __be32 error; |
450 | int host_error; | |
1da177e4 LT |
451 | struct dentry *dentry; |
452 | struct inode *inode; | |
453 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
454 | unsigned int flags = 0; | |
455 | ||
456 | /* Get inode */ | |
457 | error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, MAY_SATTR); | |
458 | if (error) | |
4b2ca38a | 459 | return error; |
1da177e4 LT |
460 | |
461 | dentry = fhp->fh_dentry; | |
462 | inode = dentry->d_inode; | |
463 | if (S_ISDIR(inode->i_mode)) | |
464 | flags = NFS4_ACL_DIR; | |
465 | ||
6264d69d AV |
466 | host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags); |
467 | if (host_error == -EINVAL) { | |
4b2ca38a | 468 | return nfserr_attrnotsupp; |
6264d69d | 469 | } else if (host_error < 0) |
1da177e4 LT |
470 | goto out_nfserr; |
471 | ||
6264d69d AV |
472 | host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS); |
473 | if (host_error < 0) | |
4b2ca38a | 474 | goto out_release; |
1da177e4 | 475 | |
4b2ca38a | 476 | if (S_ISDIR(inode->i_mode)) |
6264d69d | 477 | host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 | 478 | |
4b2ca38a | 479 | out_release: |
1da177e4 LT |
480 | posix_acl_release(pacl); |
481 | posix_acl_release(dpacl); | |
1da177e4 | 482 | out_nfserr: |
f34f9242 | 483 | if (host_error == -EOPNOTSUPP) |
4b2ca38a | 484 | return nfserr_attrnotsupp; |
f34f9242 | 485 | else |
4b2ca38a | 486 | return nfserrno(host_error); |
1da177e4 LT |
487 | } |
488 | ||
489 | static struct posix_acl * | |
490 | _get_posix_acl(struct dentry *dentry, char *key) | |
491 | { | |
5be196e5 | 492 | void *buf = NULL; |
1da177e4 | 493 | struct posix_acl *pacl = NULL; |
5be196e5 | 494 | int buflen; |
1da177e4 | 495 | |
5be196e5 CH |
496 | buflen = nfsd_getxattr(dentry, key, &buf); |
497 | if (!buflen) | |
498 | buflen = -ENODATA; | |
499 | if (buflen <= 0) | |
500 | return ERR_PTR(buflen); | |
1da177e4 LT |
501 | |
502 | pacl = posix_acl_from_xattr(buf, buflen); | |
1da177e4 LT |
503 | kfree(buf); |
504 | return pacl; | |
1da177e4 LT |
505 | } |
506 | ||
507 | int | |
508 | nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl) | |
509 | { | |
510 | struct inode *inode = dentry->d_inode; | |
511 | int error = 0; | |
512 | struct posix_acl *pacl = NULL, *dpacl = NULL; | |
513 | unsigned int flags = 0; | |
514 | ||
9a59f452 | 515 | pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS); |
1da177e4 LT |
516 | if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA) |
517 | pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL); | |
518 | if (IS_ERR(pacl)) { | |
519 | error = PTR_ERR(pacl); | |
520 | pacl = NULL; | |
521 | goto out; | |
522 | } | |
523 | ||
524 | if (S_ISDIR(inode->i_mode)) { | |
9a59f452 | 525 | dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT); |
1da177e4 LT |
526 | if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA) |
527 | dpacl = NULL; | |
528 | else if (IS_ERR(dpacl)) { | |
529 | error = PTR_ERR(dpacl); | |
530 | dpacl = NULL; | |
531 | goto out; | |
532 | } | |
533 | flags = NFS4_ACL_DIR; | |
534 | } | |
535 | ||
536 | *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags); | |
537 | if (IS_ERR(*acl)) { | |
538 | error = PTR_ERR(*acl); | |
539 | *acl = NULL; | |
540 | } | |
541 | out: | |
542 | posix_acl_release(pacl); | |
543 | posix_acl_release(dpacl); | |
544 | return error; | |
545 | } | |
546 | ||
547 | #endif /* defined(CONFIG_NFS_V4) */ | |
548 | ||
549 | #ifdef CONFIG_NFSD_V3 | |
550 | /* | |
551 | * Check server access rights to a file system object | |
552 | */ | |
553 | struct accessmap { | |
554 | u32 access; | |
555 | int how; | |
556 | }; | |
557 | static struct accessmap nfs3_regaccess[] = { | |
558 | { NFS3_ACCESS_READ, MAY_READ }, | |
559 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
560 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_TRUNC }, | |
561 | { NFS3_ACCESS_EXTEND, MAY_WRITE }, | |
562 | ||
563 | { 0, 0 } | |
564 | }; | |
565 | ||
566 | static struct accessmap nfs3_diraccess[] = { | |
567 | { NFS3_ACCESS_READ, MAY_READ }, | |
568 | { NFS3_ACCESS_LOOKUP, MAY_EXEC }, | |
569 | { NFS3_ACCESS_MODIFY, MAY_EXEC|MAY_WRITE|MAY_TRUNC }, | |
570 | { NFS3_ACCESS_EXTEND, MAY_EXEC|MAY_WRITE }, | |
571 | { NFS3_ACCESS_DELETE, MAY_REMOVE }, | |
572 | ||
573 | { 0, 0 } | |
574 | }; | |
575 | ||
576 | static struct accessmap nfs3_anyaccess[] = { | |
577 | /* Some clients - Solaris 2.6 at least, make an access call | |
578 | * to the server to check for access for things like /dev/null | |
579 | * (which really, the server doesn't care about). So | |
580 | * We provide simple access checking for them, looking | |
581 | * mainly at mode bits, and we make sure to ignore read-only | |
582 | * filesystem checks | |
583 | */ | |
584 | { NFS3_ACCESS_READ, MAY_READ }, | |
585 | { NFS3_ACCESS_EXECUTE, MAY_EXEC }, | |
586 | { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
587 | { NFS3_ACCESS_EXTEND, MAY_WRITE|MAY_LOCAL_ACCESS }, | |
588 | ||
589 | { 0, 0 } | |
590 | }; | |
591 | ||
6264d69d | 592 | __be32 |
1da177e4 LT |
593 | nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported) |
594 | { | |
595 | struct accessmap *map; | |
596 | struct svc_export *export; | |
597 | struct dentry *dentry; | |
598 | u32 query, result = 0, sresult = 0; | |
6264d69d | 599 | __be32 error; |
1da177e4 LT |
600 | |
601 | error = fh_verify(rqstp, fhp, 0, MAY_NOP); | |
602 | if (error) | |
603 | goto out; | |
604 | ||
605 | export = fhp->fh_export; | |
606 | dentry = fhp->fh_dentry; | |
607 | ||
608 | if (S_ISREG(dentry->d_inode->i_mode)) | |
609 | map = nfs3_regaccess; | |
610 | else if (S_ISDIR(dentry->d_inode->i_mode)) | |
611 | map = nfs3_diraccess; | |
612 | else | |
613 | map = nfs3_anyaccess; | |
614 | ||
615 | ||
616 | query = *access; | |
617 | for (; map->access; map++) { | |
618 | if (map->access & query) { | |
6264d69d | 619 | __be32 err2; |
1da177e4 LT |
620 | |
621 | sresult |= map->access; | |
622 | ||
0ec757df | 623 | err2 = nfsd_permission(rqstp, export, dentry, map->how); |
1da177e4 LT |
624 | switch (err2) { |
625 | case nfs_ok: | |
626 | result |= map->access; | |
627 | break; | |
628 | ||
629 | /* the following error codes just mean the access was not allowed, | |
630 | * rather than an error occurred */ | |
631 | case nfserr_rofs: | |
632 | case nfserr_acces: | |
633 | case nfserr_perm: | |
634 | /* simply don't "or" in the access bit. */ | |
635 | break; | |
636 | default: | |
637 | error = err2; | |
638 | goto out; | |
639 | } | |
640 | } | |
641 | } | |
642 | *access = result; | |
643 | if (supported) | |
644 | *supported = sresult; | |
645 | ||
646 | out: | |
647 | return error; | |
648 | } | |
649 | #endif /* CONFIG_NFSD_V3 */ | |
650 | ||
651 | ||
652 | ||
653 | /* | |
654 | * Open an existing file or directory. | |
655 | * The access argument indicates the type of open (read/write/lock) | |
656 | * N.B. After this call fhp needs an fh_put | |
657 | */ | |
6264d69d | 658 | __be32 |
1da177e4 LT |
659 | nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
660 | int access, struct file **filp) | |
661 | { | |
662 | struct dentry *dentry; | |
663 | struct inode *inode; | |
6264d69d AV |
664 | int flags = O_RDONLY|O_LARGEFILE; |
665 | __be32 err; | |
666 | int host_err; | |
1da177e4 LT |
667 | |
668 | /* | |
669 | * If we get here, then the client has already done an "open", | |
670 | * and (hopefully) checked permission - so allow OWNER_OVERRIDE | |
671 | * in case a chmod has now revoked permission. | |
672 | */ | |
673 | err = fh_verify(rqstp, fhp, type, access | MAY_OWNER_OVERRIDE); | |
674 | if (err) | |
675 | goto out; | |
676 | ||
677 | dentry = fhp->fh_dentry; | |
678 | inode = dentry->d_inode; | |
679 | ||
680 | /* Disallow write access to files with the append-only bit set | |
681 | * or any access when mandatory locking enabled | |
682 | */ | |
683 | err = nfserr_perm; | |
684 | if (IS_APPEND(inode) && (access & MAY_WRITE)) | |
685 | goto out; | |
5e7fc436 BF |
686 | /* |
687 | * We must ignore files (but only files) which might have mandatory | |
688 | * locks on them because there is no way to know if the accesser has | |
689 | * the lock. | |
690 | */ | |
691 | if (S_ISREG((inode)->i_mode) && mandatory_lock(inode)) | |
1da177e4 LT |
692 | goto out; |
693 | ||
694 | if (!inode->i_fop) | |
695 | goto out; | |
696 | ||
697 | /* | |
698 | * Check to see if there are any leases on this file. | |
699 | * This may block while leases are broken. | |
700 | */ | |
6264d69d AV |
701 | host_err = break_lease(inode, O_NONBLOCK | ((access & MAY_WRITE) ? FMODE_WRITE : 0)); |
702 | if (host_err == -EWOULDBLOCK) | |
703 | host_err = -ETIMEDOUT; | |
704 | if (host_err) /* NOMEM or WOULDBLOCK */ | |
1da177e4 LT |
705 | goto out_nfserr; |
706 | ||
707 | if (access & MAY_WRITE) { | |
9ecb6a08 BF |
708 | if (access & MAY_READ) |
709 | flags = O_RDWR|O_LARGEFILE; | |
710 | else | |
711 | flags = O_WRONLY|O_LARGEFILE; | |
1da177e4 LT |
712 | |
713 | DQUOT_INIT(inode); | |
714 | } | |
715 | *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_mnt), flags); | |
716 | if (IS_ERR(*filp)) | |
6264d69d | 717 | host_err = PTR_ERR(*filp); |
1da177e4 | 718 | out_nfserr: |
6264d69d | 719 | err = nfserrno(host_err); |
1da177e4 LT |
720 | out: |
721 | return err; | |
722 | } | |
723 | ||
724 | /* | |
725 | * Close a file. | |
726 | */ | |
727 | void | |
728 | nfsd_close(struct file *filp) | |
729 | { | |
730 | fput(filp); | |
731 | } | |
732 | ||
733 | /* | |
734 | * Sync a file | |
735 | * As this calls fsync (not fdatasync) there is no need for a write_inode | |
736 | * after it. | |
737 | */ | |
a334de28 | 738 | static inline int nfsd_dosync(struct file *filp, struct dentry *dp, |
99ac48f5 | 739 | const struct file_operations *fop) |
1da177e4 LT |
740 | { |
741 | struct inode *inode = dp->d_inode; | |
742 | int (*fsync) (struct file *, struct dentry *, int); | |
f193fbab | 743 | int err; |
1da177e4 | 744 | |
f193fbab YT |
745 | err = filemap_fdatawrite(inode->i_mapping); |
746 | if (err == 0 && fop && (fsync = fop->fsync)) | |
747 | err = fsync(filp, dp, 0); | |
748 | if (err == 0) | |
749 | err = filemap_fdatawait(inode->i_mapping); | |
a334de28 | 750 | |
f193fbab | 751 | return err; |
1da177e4 LT |
752 | } |
753 | ||
754 | ||
a334de28 | 755 | static int |
1da177e4 LT |
756 | nfsd_sync(struct file *filp) |
757 | { | |
a334de28 | 758 | int err; |
7eaa36e2 JJS |
759 | struct inode *inode = filp->f_path.dentry->d_inode; |
760 | dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name); | |
1b1dcc1b | 761 | mutex_lock(&inode->i_mutex); |
7eaa36e2 | 762 | err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op); |
1b1dcc1b | 763 | mutex_unlock(&inode->i_mutex); |
a334de28 DS |
764 | |
765 | return err; | |
1da177e4 LT |
766 | } |
767 | ||
f193fbab | 768 | int |
1da177e4 LT |
769 | nfsd_sync_dir(struct dentry *dp) |
770 | { | |
f193fbab | 771 | return nfsd_dosync(NULL, dp, dp->d_inode->i_fop); |
1da177e4 LT |
772 | } |
773 | ||
774 | /* | |
775 | * Obtain the readahead parameters for the file | |
776 | * specified by (dev, ino). | |
777 | */ | |
1da177e4 LT |
778 | |
779 | static inline struct raparms * | |
780 | nfsd_get_raparms(dev_t dev, ino_t ino) | |
781 | { | |
782 | struct raparms *ra, **rap, **frap = NULL; | |
783 | int depth = 0; | |
fce1456a GB |
784 | unsigned int hash; |
785 | struct raparm_hbucket *rab; | |
786 | ||
787 | hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK; | |
788 | rab = &raparm_hash[hash]; | |
1da177e4 | 789 | |
fce1456a GB |
790 | spin_lock(&rab->pb_lock); |
791 | for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) { | |
1da177e4 LT |
792 | if (ra->p_ino == ino && ra->p_dev == dev) |
793 | goto found; | |
794 | depth++; | |
795 | if (ra->p_count == 0) | |
796 | frap = rap; | |
797 | } | |
798 | depth = nfsdstats.ra_size*11/10; | |
799 | if (!frap) { | |
fce1456a | 800 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
801 | return NULL; |
802 | } | |
803 | rap = frap; | |
804 | ra = *frap; | |
805 | ra->p_dev = dev; | |
806 | ra->p_ino = ino; | |
807 | ra->p_set = 0; | |
fce1456a | 808 | ra->p_hindex = hash; |
1da177e4 | 809 | found: |
fce1456a | 810 | if (rap != &rab->pb_head) { |
1da177e4 | 811 | *rap = ra->p_next; |
fce1456a GB |
812 | ra->p_next = rab->pb_head; |
813 | rab->pb_head = ra; | |
1da177e4 LT |
814 | } |
815 | ra->p_count++; | |
816 | nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++; | |
fce1456a | 817 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
818 | return ra; |
819 | } | |
820 | ||
821 | /* | |
cf8208d0 JA |
822 | * Grab and keep cached pages associated with a file in the svc_rqst |
823 | * so that they can be passed to the network sendmsg/sendpage routines | |
824 | * directly. They will be released after the sending has completed. | |
1da177e4 LT |
825 | */ |
826 | static int | |
cf8208d0 JA |
827 | nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf, |
828 | struct splice_desc *sd) | |
1da177e4 | 829 | { |
cf8208d0 | 830 | struct svc_rqst *rqstp = sd->u.data; |
44524359 | 831 | struct page **pp = rqstp->rq_respages + rqstp->rq_resused; |
cf8208d0 JA |
832 | struct page *page = buf->page; |
833 | size_t size; | |
834 | int ret; | |
1da177e4 | 835 | |
cac36bb0 | 836 | ret = buf->ops->confirm(pipe, buf); |
cf8208d0 JA |
837 | if (unlikely(ret)) |
838 | return ret; | |
839 | ||
840 | size = sd->len; | |
1da177e4 LT |
841 | |
842 | if (rqstp->rq_res.page_len == 0) { | |
843 | get_page(page); | |
44524359 N |
844 | put_page(*pp); |
845 | *pp = page; | |
846 | rqstp->rq_resused++; | |
cf8208d0 | 847 | rqstp->rq_res.page_base = buf->offset; |
1da177e4 | 848 | rqstp->rq_res.page_len = size; |
44524359 | 849 | } else if (page != pp[-1]) { |
1da177e4 | 850 | get_page(page); |
250f3915 N |
851 | if (*pp) |
852 | put_page(*pp); | |
44524359 N |
853 | *pp = page; |
854 | rqstp->rq_resused++; | |
1da177e4 | 855 | rqstp->rq_res.page_len += size; |
44524359 | 856 | } else |
1da177e4 | 857 | rqstp->rq_res.page_len += size; |
1da177e4 | 858 | |
1da177e4 LT |
859 | return size; |
860 | } | |
861 | ||
cf8208d0 JA |
862 | static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe, |
863 | struct splice_desc *sd) | |
864 | { | |
865 | return __splice_from_pipe(pipe, sd, nfsd_splice_actor); | |
866 | } | |
867 | ||
a8754bee DH |
868 | static inline int svc_msnfs(struct svc_fh *ffhp) |
869 | { | |
870 | #ifdef MSNFS | |
871 | return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS); | |
872 | #else | |
873 | return 0; | |
874 | #endif | |
875 | } | |
876 | ||
6264d69d | 877 | static __be32 |
1da177e4 LT |
878 | nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
879 | loff_t offset, struct kvec *vec, int vlen, unsigned long *count) | |
880 | { | |
881 | struct inode *inode; | |
882 | struct raparms *ra; | |
883 | mm_segment_t oldfs; | |
6264d69d AV |
884 | __be32 err; |
885 | int host_err; | |
1da177e4 LT |
886 | |
887 | err = nfserr_perm; | |
7eaa36e2 | 888 | inode = file->f_path.dentry->d_inode; |
a8754bee DH |
889 | |
890 | if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count)) | |
1da177e4 | 891 | goto out; |
1da177e4 LT |
892 | |
893 | /* Get readahead parameters */ | |
894 | ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino); | |
895 | ||
896 | if (ra && ra->p_set) | |
897 | file->f_ra = ra->p_ra; | |
898 | ||
cf8208d0 JA |
899 | if (file->f_op->splice_read && rqstp->rq_splice_ok) { |
900 | struct splice_desc sd = { | |
901 | .len = 0, | |
902 | .total_len = *count, | |
903 | .pos = offset, | |
904 | .u.data = rqstp, | |
905 | }; | |
906 | ||
4fbef206 | 907 | rqstp->rq_resused = 1; |
cf8208d0 | 908 | host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor); |
1da177e4 LT |
909 | } else { |
910 | oldfs = get_fs(); | |
911 | set_fs(KERNEL_DS); | |
6264d69d | 912 | host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 LT |
913 | set_fs(oldfs); |
914 | } | |
915 | ||
916 | /* Write back readahead params */ | |
917 | if (ra) { | |
fce1456a GB |
918 | struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex]; |
919 | spin_lock(&rab->pb_lock); | |
1da177e4 LT |
920 | ra->p_ra = file->f_ra; |
921 | ra->p_set = 1; | |
922 | ra->p_count--; | |
fce1456a | 923 | spin_unlock(&rab->pb_lock); |
1da177e4 LT |
924 | } |
925 | ||
6264d69d AV |
926 | if (host_err >= 0) { |
927 | nfsdstats.io_read += host_err; | |
928 | *count = host_err; | |
1da177e4 | 929 | err = 0; |
7eaa36e2 | 930 | fsnotify_access(file->f_path.dentry); |
1da177e4 | 931 | } else |
6264d69d | 932 | err = nfserrno(host_err); |
1da177e4 LT |
933 | out: |
934 | return err; | |
935 | } | |
936 | ||
9f708e40 NB |
937 | static void kill_suid(struct dentry *dentry) |
938 | { | |
939 | struct iattr ia; | |
940 | ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID; | |
941 | ||
1b1dcc1b | 942 | mutex_lock(&dentry->d_inode->i_mutex); |
9f708e40 | 943 | notify_change(dentry, &ia); |
1b1dcc1b | 944 | mutex_unlock(&dentry->d_inode->i_mutex); |
9f708e40 NB |
945 | } |
946 | ||
6264d69d | 947 | static __be32 |
1da177e4 LT |
948 | nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
949 | loff_t offset, struct kvec *vec, int vlen, | |
950 | unsigned long cnt, int *stablep) | |
951 | { | |
952 | struct svc_export *exp; | |
953 | struct dentry *dentry; | |
954 | struct inode *inode; | |
955 | mm_segment_t oldfs; | |
6264d69d AV |
956 | __be32 err = 0; |
957 | int host_err; | |
1da177e4 LT |
958 | int stable = *stablep; |
959 | ||
45bd3b3d | 960 | #ifdef MSNFS |
1da177e4 LT |
961 | err = nfserr_perm; |
962 | ||
1da177e4 | 963 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && |
7eaa36e2 | 964 | (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt))) |
1da177e4 LT |
965 | goto out; |
966 | #endif | |
967 | ||
7eaa36e2 | 968 | dentry = file->f_path.dentry; |
1da177e4 LT |
969 | inode = dentry->d_inode; |
970 | exp = fhp->fh_export; | |
971 | ||
972 | /* | |
973 | * Request sync writes if | |
974 | * - the sync export option has been set, or | |
975 | * - the client requested O_SYNC behavior (NFSv3 feature). | |
976 | * - The file system doesn't support fsync(). | |
977 | * When gathered writes have been configured for this volume, | |
978 | * flushing the data to disk is handled separately below. | |
979 | */ | |
980 | ||
981 | if (file->f_op->fsync == 0) {/* COMMIT3 cannot work */ | |
982 | stable = 2; | |
983 | *stablep = 2; /* FILE_SYNC */ | |
984 | } | |
985 | ||
986 | if (!EX_ISSYNC(exp)) | |
987 | stable = 0; | |
988 | if (stable && !EX_WGATHER(exp)) | |
989 | file->f_flags |= O_SYNC; | |
990 | ||
991 | /* Write the data. */ | |
992 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 993 | host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset); |
1da177e4 | 994 | set_fs(oldfs); |
6264d69d | 995 | if (host_err >= 0) { |
1da177e4 | 996 | nfsdstats.io_write += cnt; |
7eaa36e2 | 997 | fsnotify_modify(file->f_path.dentry); |
1da177e4 LT |
998 | } |
999 | ||
1000 | /* clear setuid/setgid flag after write */ | |
6264d69d | 1001 | if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID))) |
9f708e40 | 1002 | kill_suid(dentry); |
1da177e4 | 1003 | |
6264d69d | 1004 | if (host_err >= 0 && stable) { |
1da177e4 LT |
1005 | static ino_t last_ino; |
1006 | static dev_t last_dev; | |
1007 | ||
1008 | /* | |
1009 | * Gathered writes: If another process is currently | |
1010 | * writing to the file, there's a high chance | |
1011 | * this is another nfsd (triggered by a bulk write | |
1012 | * from a client's biod). Rather than syncing the | |
1013 | * file with each write request, we sleep for 10 msec. | |
1014 | * | |
1015 | * I don't know if this roughly approximates | |
1016 | * C. Juszak's idea of gathered writes, but it's a | |
1017 | * nice and simple solution (IMHO), and it seems to | |
1018 | * work:-) | |
1019 | */ | |
1020 | if (EX_WGATHER(exp)) { | |
1021 | if (atomic_read(&inode->i_writecount) > 1 | |
1022 | || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) { | |
1023 | dprintk("nfsd: write defer %d\n", current->pid); | |
1024 | msleep(10); | |
1025 | dprintk("nfsd: write resume %d\n", current->pid); | |
1026 | } | |
1027 | ||
1028 | if (inode->i_state & I_DIRTY) { | |
1029 | dprintk("nfsd: write sync %d\n", current->pid); | |
6264d69d | 1030 | host_err=nfsd_sync(file); |
1da177e4 LT |
1031 | } |
1032 | #if 0 | |
1033 | wake_up(&inode->i_wait); | |
1034 | #endif | |
1035 | } | |
1036 | last_ino = inode->i_ino; | |
1037 | last_dev = inode->i_sb->s_dev; | |
1038 | } | |
1039 | ||
6264d69d AV |
1040 | dprintk("nfsd: write complete host_err=%d\n", host_err); |
1041 | if (host_err >= 0) | |
1da177e4 LT |
1042 | err = 0; |
1043 | else | |
6264d69d | 1044 | err = nfserrno(host_err); |
1da177e4 LT |
1045 | out: |
1046 | return err; | |
1047 | } | |
1048 | ||
1049 | /* | |
1050 | * Read data from a file. count must contain the requested read count | |
1051 | * on entry. On return, *count contains the number of bytes actually read. | |
1052 | * N.B. After this call fhp needs an fh_put | |
1053 | */ | |
6264d69d | 1054 | __be32 |
1da177e4 LT |
1055 | nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1056 | loff_t offset, struct kvec *vec, int vlen, | |
1057 | unsigned long *count) | |
1058 | { | |
6264d69d | 1059 | __be32 err; |
1da177e4 LT |
1060 | |
1061 | if (file) { | |
0ec757df | 1062 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
1da177e4 LT |
1063 | MAY_READ|MAY_OWNER_OVERRIDE); |
1064 | if (err) | |
1065 | goto out; | |
1066 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1067 | } else { | |
1068 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_READ, &file); | |
1069 | if (err) | |
1070 | goto out; | |
1071 | err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count); | |
1072 | nfsd_close(file); | |
1073 | } | |
1074 | out: | |
1075 | return err; | |
1076 | } | |
1077 | ||
1078 | /* | |
1079 | * Write data to a file. | |
1080 | * The stable flag requests synchronous writes. | |
1081 | * N.B. After this call fhp needs an fh_put | |
1082 | */ | |
6264d69d | 1083 | __be32 |
1da177e4 LT |
1084 | nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
1085 | loff_t offset, struct kvec *vec, int vlen, unsigned long cnt, | |
1086 | int *stablep) | |
1087 | { | |
6264d69d | 1088 | __be32 err = 0; |
1da177e4 LT |
1089 | |
1090 | if (file) { | |
0ec757df | 1091 | err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry, |
1da177e4 LT |
1092 | MAY_WRITE|MAY_OWNER_OVERRIDE); |
1093 | if (err) | |
1094 | goto out; | |
1095 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt, | |
1096 | stablep); | |
1097 | } else { | |
1098 | err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file); | |
1099 | if (err) | |
1100 | goto out; | |
1101 | ||
1102 | if (cnt) | |
1103 | err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, | |
1104 | cnt, stablep); | |
1105 | nfsd_close(file); | |
1106 | } | |
1107 | out: | |
1108 | return err; | |
1109 | } | |
1110 | ||
1111 | #ifdef CONFIG_NFSD_V3 | |
1112 | /* | |
1113 | * Commit all pending writes to stable storage. | |
1114 | * Strictly speaking, we could sync just the indicated file region here, | |
1115 | * but there's currently no way we can ask the VFS to do so. | |
1116 | * | |
1117 | * Unfortunately we cannot lock the file to make sure we return full WCC | |
1118 | * data to the client, as locking happens lower down in the filesystem. | |
1119 | */ | |
6264d69d | 1120 | __be32 |
1da177e4 LT |
1121 | nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1122 | loff_t offset, unsigned long count) | |
1123 | { | |
1124 | struct file *file; | |
6264d69d | 1125 | __be32 err; |
1da177e4 LT |
1126 | |
1127 | if ((u64)count > ~(u64)offset) | |
1128 | return nfserr_inval; | |
1129 | ||
1130 | if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0) | |
1131 | return err; | |
1132 | if (EX_ISSYNC(fhp->fh_export)) { | |
1133 | if (file->f_op && file->f_op->fsync) { | |
45bd3b3d | 1134 | err = nfserrno(nfsd_sync(file)); |
1da177e4 LT |
1135 | } else { |
1136 | err = nfserr_notsupp; | |
1137 | } | |
1138 | } | |
1139 | ||
1140 | nfsd_close(file); | |
1141 | return err; | |
1142 | } | |
1143 | #endif /* CONFIG_NFSD_V3 */ | |
1144 | ||
1145 | /* | |
1146 | * Create a file (regular, directory, device, fifo); UNIX sockets | |
1147 | * not yet implemented. | |
1148 | * If the response fh has been verified, the parent directory should | |
1149 | * already be locked. Note that the parent directory is left locked. | |
1150 | * | |
1151 | * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp | |
1152 | */ | |
6264d69d | 1153 | __be32 |
1da177e4 LT |
1154 | nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1155 | char *fname, int flen, struct iattr *iap, | |
1156 | int type, dev_t rdev, struct svc_fh *resfhp) | |
1157 | { | |
1158 | struct dentry *dentry, *dchild = NULL; | |
1159 | struct inode *dirp; | |
6264d69d AV |
1160 | __be32 err; |
1161 | int host_err; | |
1da177e4 LT |
1162 | |
1163 | err = nfserr_perm; | |
1164 | if (!flen) | |
1165 | goto out; | |
1166 | err = nfserr_exist; | |
1167 | if (isdotent(fname, flen)) | |
1168 | goto out; | |
1169 | ||
1170 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1171 | if (err) | |
1172 | goto out; | |
1173 | ||
1174 | dentry = fhp->fh_dentry; | |
1175 | dirp = dentry->d_inode; | |
1176 | ||
1177 | err = nfserr_notdir; | |
1178 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1179 | goto out; | |
1180 | /* | |
1181 | * Check whether the response file handle has been verified yet. | |
1182 | * If it has, the parent directory should already be locked. | |
1183 | */ | |
1184 | if (!resfhp->fh_dentry) { | |
1185 | /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ | |
12fd3520 | 1186 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 | 1187 | dchild = lookup_one_len(fname, dentry, flen); |
6264d69d | 1188 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1189 | if (IS_ERR(dchild)) |
1190 | goto out_nfserr; | |
1191 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1192 | if (err) | |
1193 | goto out; | |
1194 | } else { | |
1195 | /* called from nfsd_proc_create */ | |
1196 | dchild = dget(resfhp->fh_dentry); | |
1197 | if (!fhp->fh_locked) { | |
1198 | /* not actually possible */ | |
1199 | printk(KERN_ERR | |
1200 | "nfsd_create: parent %s/%s not locked!\n", | |
1201 | dentry->d_parent->d_name.name, | |
1202 | dentry->d_name.name); | |
d75f2b9f | 1203 | err = nfserr_io; |
1da177e4 LT |
1204 | goto out; |
1205 | } | |
1206 | } | |
1207 | /* | |
1208 | * Make sure the child dentry is still negative ... | |
1209 | */ | |
1210 | err = nfserr_exist; | |
1211 | if (dchild->d_inode) { | |
1212 | dprintk("nfsd_create: dentry %s/%s not negative!\n", | |
1213 | dentry->d_name.name, dchild->d_name.name); | |
1214 | goto out; | |
1215 | } | |
1216 | ||
1217 | if (!(iap->ia_valid & ATTR_MODE)) | |
1218 | iap->ia_mode = 0; | |
1219 | iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type; | |
1220 | ||
1221 | /* | |
1222 | * Get the dir op function pointer. | |
1223 | */ | |
088406bc | 1224 | err = 0; |
1da177e4 LT |
1225 | switch (type) { |
1226 | case S_IFREG: | |
6264d69d | 1227 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1da177e4 LT |
1228 | break; |
1229 | case S_IFDIR: | |
6264d69d | 1230 | host_err = vfs_mkdir(dirp, dchild, iap->ia_mode); |
1da177e4 LT |
1231 | break; |
1232 | case S_IFCHR: | |
1233 | case S_IFBLK: | |
1234 | case S_IFIFO: | |
1235 | case S_IFSOCK: | |
6264d69d | 1236 | host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); |
1da177e4 LT |
1237 | break; |
1238 | default: | |
1239 | printk("nfsd: bad file type %o in nfsd_create\n", type); | |
6264d69d | 1240 | host_err = -EINVAL; |
1da177e4 | 1241 | } |
6264d69d | 1242 | if (host_err < 0) |
1da177e4 LT |
1243 | goto out_nfserr; |
1244 | ||
1245 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1246 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1247 | write_inode_now(dchild->d_inode, 1); |
1248 | } | |
1249 | ||
1250 | ||
1251 | /* Set file attributes. Mode has already been set and | |
1252 | * setting uid/gid works only for root. Irix appears to | |
1253 | * send along the gid when it tries to implement setgid | |
1254 | * directories via NFS. | |
1255 | */ | |
f193fbab | 1256 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1257 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab YT |
1258 | if (err2) |
1259 | err = err2; | |
1260 | } | |
1da177e4 LT |
1261 | /* |
1262 | * Update the file handle to get the new inode info. | |
1263 | */ | |
1264 | if (!err) | |
1265 | err = fh_update(resfhp); | |
1266 | out: | |
1267 | if (dchild && !IS_ERR(dchild)) | |
1268 | dput(dchild); | |
1269 | return err; | |
1270 | ||
1271 | out_nfserr: | |
6264d69d | 1272 | err = nfserrno(host_err); |
1da177e4 LT |
1273 | goto out; |
1274 | } | |
1275 | ||
1276 | #ifdef CONFIG_NFSD_V3 | |
1277 | /* | |
1278 | * NFSv3 version of nfsd_create | |
1279 | */ | |
6264d69d | 1280 | __be32 |
1da177e4 LT |
1281 | nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1282 | char *fname, int flen, struct iattr *iap, | |
1283 | struct svc_fh *resfhp, int createmode, u32 *verifier, | |
81ac95c5 | 1284 | int *truncp, int *created) |
1da177e4 LT |
1285 | { |
1286 | struct dentry *dentry, *dchild = NULL; | |
1287 | struct inode *dirp; | |
6264d69d AV |
1288 | __be32 err; |
1289 | int host_err; | |
1da177e4 | 1290 | __u32 v_mtime=0, v_atime=0; |
1da177e4 LT |
1291 | |
1292 | err = nfserr_perm; | |
1293 | if (!flen) | |
1294 | goto out; | |
1295 | err = nfserr_exist; | |
1296 | if (isdotent(fname, flen)) | |
1297 | goto out; | |
1298 | if (!(iap->ia_valid & ATTR_MODE)) | |
1299 | iap->ia_mode = 0; | |
1300 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1301 | if (err) | |
1302 | goto out; | |
1303 | ||
1304 | dentry = fhp->fh_dentry; | |
1305 | dirp = dentry->d_inode; | |
1306 | ||
1307 | /* Get all the sanity checks out of the way before | |
1308 | * we lock the parent. */ | |
1309 | err = nfserr_notdir; | |
1310 | if(!dirp->i_op || !dirp->i_op->lookup) | |
1311 | goto out; | |
12fd3520 | 1312 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1313 | |
1314 | /* | |
1315 | * Compose the response file handle. | |
1316 | */ | |
1317 | dchild = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1318 | host_err = PTR_ERR(dchild); |
1da177e4 LT |
1319 | if (IS_ERR(dchild)) |
1320 | goto out_nfserr; | |
1321 | ||
1322 | err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); | |
1323 | if (err) | |
1324 | goto out; | |
1325 | ||
1326 | if (createmode == NFS3_CREATE_EXCLUSIVE) { | |
c397852c | 1327 | /* solaris7 gets confused (bugid 4218508) if these have |
749997e5 JL |
1328 | * the high bit set, so just clear the high bits. If this is |
1329 | * ever changed to use different attrs for storing the | |
1330 | * verifier, then do_open_lookup() will also need to be fixed | |
1331 | * accordingly. | |
1da177e4 LT |
1332 | */ |
1333 | v_mtime = verifier[0]&0x7fffffff; | |
1334 | v_atime = verifier[1]&0x7fffffff; | |
1da177e4 LT |
1335 | } |
1336 | ||
1337 | if (dchild->d_inode) { | |
1338 | err = 0; | |
1339 | ||
1340 | switch (createmode) { | |
1341 | case NFS3_CREATE_UNCHECKED: | |
1342 | if (! S_ISREG(dchild->d_inode->i_mode)) | |
1343 | err = nfserr_exist; | |
1344 | else if (truncp) { | |
1345 | /* in nfsv4, we need to treat this case a little | |
1346 | * differently. we don't want to truncate the | |
1347 | * file now; this would be wrong if the OPEN | |
1348 | * fails for some other reason. furthermore, | |
1349 | * if the size is nonzero, we should ignore it | |
1350 | * according to spec! | |
1351 | */ | |
1352 | *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size; | |
1353 | } | |
1354 | else { | |
1355 | iap->ia_valid &= ATTR_SIZE; | |
1356 | goto set_attr; | |
1357 | } | |
1358 | break; | |
1359 | case NFS3_CREATE_EXCLUSIVE: | |
1360 | if ( dchild->d_inode->i_mtime.tv_sec == v_mtime | |
1361 | && dchild->d_inode->i_atime.tv_sec == v_atime | |
1da177e4 LT |
1362 | && dchild->d_inode->i_size == 0 ) |
1363 | break; | |
1364 | /* fallthru */ | |
1365 | case NFS3_CREATE_GUARDED: | |
1366 | err = nfserr_exist; | |
1367 | } | |
1368 | goto out; | |
1369 | } | |
1370 | ||
6264d69d AV |
1371 | host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL); |
1372 | if (host_err < 0) | |
1da177e4 | 1373 | goto out_nfserr; |
81ac95c5 BF |
1374 | if (created) |
1375 | *created = 1; | |
1da177e4 LT |
1376 | |
1377 | if (EX_ISSYNC(fhp->fh_export)) { | |
45bd3b3d | 1378 | err = nfserrno(nfsd_sync_dir(dentry)); |
1da177e4 LT |
1379 | /* setattr will sync the child (or not) */ |
1380 | } | |
1381 | ||
1da177e4 | 1382 | if (createmode == NFS3_CREATE_EXCLUSIVE) { |
c397852c | 1383 | /* Cram the verifier into atime/mtime */ |
1da177e4 | 1384 | iap->ia_valid = ATTR_MTIME|ATTR_ATIME |
c397852c | 1385 | | ATTR_MTIME_SET|ATTR_ATIME_SET; |
1da177e4 LT |
1386 | /* XXX someone who knows this better please fix it for nsec */ |
1387 | iap->ia_mtime.tv_sec = v_mtime; | |
1388 | iap->ia_atime.tv_sec = v_atime; | |
1389 | iap->ia_mtime.tv_nsec = 0; | |
1390 | iap->ia_atime.tv_nsec = 0; | |
1da177e4 LT |
1391 | } |
1392 | ||
1393 | /* Set file attributes. | |
1da177e4 LT |
1394 | * Irix appears to send along the gid when it tries to |
1395 | * implement setgid directories via NFS. Clear out all that cruft. | |
1396 | */ | |
1397 | set_attr: | |
c397852c | 1398 | if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) { |
6264d69d | 1399 | __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0); |
f193fbab | 1400 | if (err2) |
45bd3b3d | 1401 | err = err2; |
f193fbab YT |
1402 | } |
1403 | ||
1404 | /* | |
1405 | * Update the filehandle to get the new inode info. | |
1406 | */ | |
1407 | if (!err) | |
1408 | err = fh_update(resfhp); | |
1da177e4 LT |
1409 | |
1410 | out: | |
1411 | fh_unlock(fhp); | |
1412 | if (dchild && !IS_ERR(dchild)) | |
1413 | dput(dchild); | |
1414 | return err; | |
1415 | ||
1416 | out_nfserr: | |
6264d69d | 1417 | err = nfserrno(host_err); |
1da177e4 LT |
1418 | goto out; |
1419 | } | |
1420 | #endif /* CONFIG_NFSD_V3 */ | |
1421 | ||
1422 | /* | |
1423 | * Read a symlink. On entry, *lenp must contain the maximum path length that | |
1424 | * fits into the buffer. On return, it contains the true length. | |
1425 | * N.B. After this call fhp needs an fh_put | |
1426 | */ | |
6264d69d | 1427 | __be32 |
1da177e4 LT |
1428 | nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp) |
1429 | { | |
1430 | struct dentry *dentry; | |
1431 | struct inode *inode; | |
1432 | mm_segment_t oldfs; | |
6264d69d AV |
1433 | __be32 err; |
1434 | int host_err; | |
1da177e4 LT |
1435 | |
1436 | err = fh_verify(rqstp, fhp, S_IFLNK, MAY_NOP); | |
1437 | if (err) | |
1438 | goto out; | |
1439 | ||
1440 | dentry = fhp->fh_dentry; | |
1441 | inode = dentry->d_inode; | |
1442 | ||
1443 | err = nfserr_inval; | |
1444 | if (!inode->i_op || !inode->i_op->readlink) | |
1445 | goto out; | |
1446 | ||
1447 | touch_atime(fhp->fh_export->ex_mnt, dentry); | |
1448 | /* N.B. Why does this call need a get_fs()?? | |
1449 | * Remove the set_fs and watch the fireworks:-) --okir | |
1450 | */ | |
1451 | ||
1452 | oldfs = get_fs(); set_fs(KERNEL_DS); | |
6264d69d | 1453 | host_err = inode->i_op->readlink(dentry, buf, *lenp); |
1da177e4 LT |
1454 | set_fs(oldfs); |
1455 | ||
6264d69d | 1456 | if (host_err < 0) |
1da177e4 | 1457 | goto out_nfserr; |
6264d69d | 1458 | *lenp = host_err; |
1da177e4 LT |
1459 | err = 0; |
1460 | out: | |
1461 | return err; | |
1462 | ||
1463 | out_nfserr: | |
6264d69d | 1464 | err = nfserrno(host_err); |
1da177e4 LT |
1465 | goto out; |
1466 | } | |
1467 | ||
1468 | /* | |
1469 | * Create a symlink and look up its inode | |
1470 | * N.B. After this call _both_ fhp and resfhp need an fh_put | |
1471 | */ | |
6264d69d | 1472 | __be32 |
1da177e4 LT |
1473 | nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp, |
1474 | char *fname, int flen, | |
1475 | char *path, int plen, | |
1476 | struct svc_fh *resfhp, | |
1477 | struct iattr *iap) | |
1478 | { | |
1479 | struct dentry *dentry, *dnew; | |
6264d69d AV |
1480 | __be32 err, cerr; |
1481 | int host_err; | |
1da177e4 LT |
1482 | umode_t mode; |
1483 | ||
1484 | err = nfserr_noent; | |
1485 | if (!flen || !plen) | |
1486 | goto out; | |
1487 | err = nfserr_exist; | |
1488 | if (isdotent(fname, flen)) | |
1489 | goto out; | |
1490 | ||
1491 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE); | |
1492 | if (err) | |
1493 | goto out; | |
1494 | fh_lock(fhp); | |
1495 | dentry = fhp->fh_dentry; | |
1496 | dnew = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1497 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1498 | if (IS_ERR(dnew)) |
1499 | goto out_nfserr; | |
1500 | ||
1501 | mode = S_IALLUGO; | |
1502 | /* Only the MODE ATTRibute is even vaguely meaningful */ | |
1503 | if (iap && (iap->ia_valid & ATTR_MODE)) | |
1504 | mode = iap->ia_mode & S_IALLUGO; | |
1505 | ||
1506 | if (unlikely(path[plen] != 0)) { | |
1507 | char *path_alloced = kmalloc(plen+1, GFP_KERNEL); | |
1508 | if (path_alloced == NULL) | |
6264d69d | 1509 | host_err = -ENOMEM; |
1da177e4 LT |
1510 | else { |
1511 | strncpy(path_alloced, path, plen); | |
1512 | path_alloced[plen] = 0; | |
6264d69d | 1513 | host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode); |
1da177e4 LT |
1514 | kfree(path_alloced); |
1515 | } | |
1516 | } else | |
6264d69d | 1517 | host_err = vfs_symlink(dentry->d_inode, dnew, path, mode); |
1da177e4 | 1518 | |
6264d69d | 1519 | if (!host_err) { |
1da177e4 | 1520 | if (EX_ISSYNC(fhp->fh_export)) |
6264d69d AV |
1521 | host_err = nfsd_sync_dir(dentry); |
1522 | } | |
1523 | err = nfserrno(host_err); | |
1da177e4 LT |
1524 | fh_unlock(fhp); |
1525 | ||
1526 | cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp); | |
1527 | dput(dnew); | |
1528 | if (err==0) err = cerr; | |
1529 | out: | |
1530 | return err; | |
1531 | ||
1532 | out_nfserr: | |
6264d69d | 1533 | err = nfserrno(host_err); |
1da177e4 LT |
1534 | goto out; |
1535 | } | |
1536 | ||
1537 | /* | |
1538 | * Create a hardlink | |
1539 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1540 | */ | |
6264d69d | 1541 | __be32 |
1da177e4 LT |
1542 | nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp, |
1543 | char *name, int len, struct svc_fh *tfhp) | |
1544 | { | |
1545 | struct dentry *ddir, *dnew, *dold; | |
1546 | struct inode *dirp, *dest; | |
6264d69d AV |
1547 | __be32 err; |
1548 | int host_err; | |
1da177e4 LT |
1549 | |
1550 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_CREATE); | |
1551 | if (err) | |
1552 | goto out; | |
1553 | err = fh_verify(rqstp, tfhp, -S_IFDIR, MAY_NOP); | |
1554 | if (err) | |
1555 | goto out; | |
1556 | ||
1557 | err = nfserr_perm; | |
1558 | if (!len) | |
1559 | goto out; | |
1560 | err = nfserr_exist; | |
1561 | if (isdotent(name, len)) | |
1562 | goto out; | |
1563 | ||
12fd3520 | 1564 | fh_lock_nested(ffhp, I_MUTEX_PARENT); |
1da177e4 LT |
1565 | ddir = ffhp->fh_dentry; |
1566 | dirp = ddir->d_inode; | |
1567 | ||
1568 | dnew = lookup_one_len(name, ddir, len); | |
6264d69d | 1569 | host_err = PTR_ERR(dnew); |
1da177e4 LT |
1570 | if (IS_ERR(dnew)) |
1571 | goto out_nfserr; | |
1572 | ||
1573 | dold = tfhp->fh_dentry; | |
1574 | dest = dold->d_inode; | |
1575 | ||
6264d69d AV |
1576 | host_err = vfs_link(dold, dirp, dnew); |
1577 | if (!host_err) { | |
1da177e4 | 1578 | if (EX_ISSYNC(ffhp->fh_export)) { |
45bd3b3d | 1579 | err = nfserrno(nfsd_sync_dir(ddir)); |
1da177e4 LT |
1580 | write_inode_now(dest, 1); |
1581 | } | |
6264d69d | 1582 | err = 0; |
1da177e4 | 1583 | } else { |
6264d69d | 1584 | if (host_err == -EXDEV && rqstp->rq_vers == 2) |
1da177e4 LT |
1585 | err = nfserr_acces; |
1586 | else | |
6264d69d | 1587 | err = nfserrno(host_err); |
1da177e4 LT |
1588 | } |
1589 | ||
1da177e4 | 1590 | dput(dnew); |
270d56e5 DR |
1591 | out_unlock: |
1592 | fh_unlock(ffhp); | |
1da177e4 LT |
1593 | out: |
1594 | return err; | |
1595 | ||
1596 | out_nfserr: | |
6264d69d | 1597 | err = nfserrno(host_err); |
270d56e5 | 1598 | goto out_unlock; |
1da177e4 LT |
1599 | } |
1600 | ||
1601 | /* | |
1602 | * Rename a file | |
1603 | * N.B. After this call _both_ ffhp and tfhp need an fh_put | |
1604 | */ | |
6264d69d | 1605 | __be32 |
1da177e4 LT |
1606 | nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen, |
1607 | struct svc_fh *tfhp, char *tname, int tlen) | |
1608 | { | |
1609 | struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap; | |
1610 | struct inode *fdir, *tdir; | |
6264d69d AV |
1611 | __be32 err; |
1612 | int host_err; | |
1da177e4 LT |
1613 | |
1614 | err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_REMOVE); | |
1615 | if (err) | |
1616 | goto out; | |
1617 | err = fh_verify(rqstp, tfhp, S_IFDIR, MAY_CREATE); | |
1618 | if (err) | |
1619 | goto out; | |
1620 | ||
1621 | fdentry = ffhp->fh_dentry; | |
1622 | fdir = fdentry->d_inode; | |
1623 | ||
1624 | tdentry = tfhp->fh_dentry; | |
1625 | tdir = tdentry->d_inode; | |
1626 | ||
1627 | err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev; | |
a56f3937 | 1628 | if (ffhp->fh_export != tfhp->fh_export) |
1da177e4 LT |
1629 | goto out; |
1630 | ||
1631 | err = nfserr_perm; | |
1632 | if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen)) | |
1633 | goto out; | |
1634 | ||
1635 | /* cannot use fh_lock as we need deadlock protective ordering | |
1636 | * so do it by hand */ | |
1637 | trap = lock_rename(tdentry, fdentry); | |
1638 | ffhp->fh_locked = tfhp->fh_locked = 1; | |
1639 | fill_pre_wcc(ffhp); | |
1640 | fill_pre_wcc(tfhp); | |
1641 | ||
1642 | odentry = lookup_one_len(fname, fdentry, flen); | |
6264d69d | 1643 | host_err = PTR_ERR(odentry); |
1da177e4 LT |
1644 | if (IS_ERR(odentry)) |
1645 | goto out_nfserr; | |
1646 | ||
6264d69d | 1647 | host_err = -ENOENT; |
1da177e4 LT |
1648 | if (!odentry->d_inode) |
1649 | goto out_dput_old; | |
6264d69d | 1650 | host_err = -EINVAL; |
1da177e4 LT |
1651 | if (odentry == trap) |
1652 | goto out_dput_old; | |
1653 | ||
1654 | ndentry = lookup_one_len(tname, tdentry, tlen); | |
6264d69d | 1655 | host_err = PTR_ERR(ndentry); |
1da177e4 LT |
1656 | if (IS_ERR(ndentry)) |
1657 | goto out_dput_old; | |
6264d69d | 1658 | host_err = -ENOTEMPTY; |
1da177e4 LT |
1659 | if (ndentry == trap) |
1660 | goto out_dput_new; | |
1661 | ||
1662 | #ifdef MSNFS | |
1663 | if ((ffhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1664 | ((atomic_read(&odentry->d_count) > 1) | |
1665 | || (atomic_read(&ndentry->d_count) > 1))) { | |
6264d69d | 1666 | host_err = -EPERM; |
1da177e4 LT |
1667 | } else |
1668 | #endif | |
6264d69d AV |
1669 | host_err = vfs_rename(fdir, odentry, tdir, ndentry); |
1670 | if (!host_err && EX_ISSYNC(tfhp->fh_export)) { | |
1671 | host_err = nfsd_sync_dir(tdentry); | |
1672 | if (!host_err) | |
1673 | host_err = nfsd_sync_dir(fdentry); | |
1da177e4 LT |
1674 | } |
1675 | ||
1676 | out_dput_new: | |
1677 | dput(ndentry); | |
1678 | out_dput_old: | |
1679 | dput(odentry); | |
1680 | out_nfserr: | |
6264d69d | 1681 | err = nfserrno(host_err); |
1da177e4 LT |
1682 | |
1683 | /* we cannot reply on fh_unlock on the two filehandles, | |
1684 | * as that would do the wrong thing if the two directories | |
1685 | * were the same, so again we do it by hand | |
1686 | */ | |
1687 | fill_post_wcc(ffhp); | |
1688 | fill_post_wcc(tfhp); | |
1689 | unlock_rename(tdentry, fdentry); | |
1690 | ffhp->fh_locked = tfhp->fh_locked = 0; | |
1691 | ||
1692 | out: | |
1693 | return err; | |
1694 | } | |
1695 | ||
1696 | /* | |
1697 | * Unlink a file or directory | |
1698 | * N.B. After this call fhp needs an fh_put | |
1699 | */ | |
6264d69d | 1700 | __be32 |
1da177e4 LT |
1701 | nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type, |
1702 | char *fname, int flen) | |
1703 | { | |
1704 | struct dentry *dentry, *rdentry; | |
1705 | struct inode *dirp; | |
6264d69d AV |
1706 | __be32 err; |
1707 | int host_err; | |
1da177e4 LT |
1708 | |
1709 | err = nfserr_acces; | |
1710 | if (!flen || isdotent(fname, flen)) | |
1711 | goto out; | |
1712 | err = fh_verify(rqstp, fhp, S_IFDIR, MAY_REMOVE); | |
1713 | if (err) | |
1714 | goto out; | |
1715 | ||
12fd3520 | 1716 | fh_lock_nested(fhp, I_MUTEX_PARENT); |
1da177e4 LT |
1717 | dentry = fhp->fh_dentry; |
1718 | dirp = dentry->d_inode; | |
1719 | ||
1720 | rdentry = lookup_one_len(fname, dentry, flen); | |
6264d69d | 1721 | host_err = PTR_ERR(rdentry); |
1da177e4 LT |
1722 | if (IS_ERR(rdentry)) |
1723 | goto out_nfserr; | |
1724 | ||
1725 | if (!rdentry->d_inode) { | |
1726 | dput(rdentry); | |
1727 | err = nfserr_noent; | |
1728 | goto out; | |
1729 | } | |
1730 | ||
1731 | if (!type) | |
1732 | type = rdentry->d_inode->i_mode & S_IFMT; | |
1733 | ||
1734 | if (type != S_IFDIR) { /* It's UNLINK */ | |
1735 | #ifdef MSNFS | |
1736 | if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) && | |
1737 | (atomic_read(&rdentry->d_count) > 1)) { | |
6264d69d | 1738 | host_err = -EPERM; |
1da177e4 LT |
1739 | } else |
1740 | #endif | |
6264d69d | 1741 | host_err = vfs_unlink(dirp, rdentry); |
1da177e4 | 1742 | } else { /* It's RMDIR */ |
6264d69d | 1743 | host_err = vfs_rmdir(dirp, rdentry); |
1da177e4 LT |
1744 | } |
1745 | ||
1746 | dput(rdentry); | |
1747 | ||
6264d69d AV |
1748 | if (host_err) |
1749 | goto out_nfserr; | |
1750 | if (EX_ISSYNC(fhp->fh_export)) | |
1751 | host_err = nfsd_sync_dir(dentry); | |
1da177e4 LT |
1752 | |
1753 | out_nfserr: | |
6264d69d | 1754 | err = nfserrno(host_err); |
f193fbab YT |
1755 | out: |
1756 | return err; | |
1da177e4 LT |
1757 | } |
1758 | ||
1759 | /* | |
1760 | * Read entries from a directory. | |
1761 | * The NFSv3/4 verifier we ignore for now. | |
1762 | */ | |
6264d69d | 1763 | __be32 |
1da177e4 | 1764 | nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp, |
a0ad13ef | 1765 | struct readdir_cd *cdp, filldir_t func) |
1da177e4 | 1766 | { |
6264d69d AV |
1767 | __be32 err; |
1768 | int host_err; | |
1da177e4 LT |
1769 | struct file *file; |
1770 | loff_t offset = *offsetp; | |
1771 | ||
1772 | err = nfsd_open(rqstp, fhp, S_IFDIR, MAY_READ, &file); | |
1773 | if (err) | |
1774 | goto out; | |
1775 | ||
1776 | offset = vfs_llseek(file, offset, 0); | |
1777 | if (offset < 0) { | |
1778 | err = nfserrno((int)offset); | |
1779 | goto out_close; | |
1780 | } | |
1781 | ||
1782 | /* | |
1783 | * Read the directory entries. This silly loop is necessary because | |
1784 | * readdir() is not guaranteed to fill up the entire buffer, but | |
1785 | * may choose to do less. | |
1786 | */ | |
1787 | ||
1788 | do { | |
1789 | cdp->err = nfserr_eof; /* will be cleared on successful read */ | |
a0ad13ef | 1790 | host_err = vfs_readdir(file, func, cdp); |
6264d69d AV |
1791 | } while (host_err >=0 && cdp->err == nfs_ok); |
1792 | if (host_err) | |
1793 | err = nfserrno(host_err); | |
1da177e4 LT |
1794 | else |
1795 | err = cdp->err; | |
1796 | *offsetp = vfs_llseek(file, 0, 1); | |
1797 | ||
1798 | if (err == nfserr_eof || err == nfserr_toosmall) | |
1799 | err = nfs_ok; /* can still be found in ->err */ | |
1800 | out_close: | |
1801 | nfsd_close(file); | |
1802 | out: | |
1803 | return err; | |
1804 | } | |
1805 | ||
1806 | /* | |
1807 | * Get file system stats | |
1808 | * N.B. After this call fhp needs an fh_put | |
1809 | */ | |
6264d69d | 1810 | __be32 |
1da177e4 LT |
1811 | nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat) |
1812 | { | |
6264d69d | 1813 | __be32 err = fh_verify(rqstp, fhp, 0, MAY_NOP); |
726c3342 | 1814 | if (!err && vfs_statfs(fhp->fh_dentry,stat)) |
1da177e4 LT |
1815 | err = nfserr_io; |
1816 | return err; | |
1817 | } | |
1818 | ||
c7d51402 | 1819 | static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp) |
e22841c6 | 1820 | { |
c7d51402 | 1821 | return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY; |
e22841c6 BF |
1822 | } |
1823 | ||
1da177e4 LT |
1824 | /* |
1825 | * Check for a user's access permissions to this inode. | |
1826 | */ | |
6264d69d | 1827 | __be32 |
0ec757df BF |
1828 | nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp, |
1829 | struct dentry *dentry, int acc) | |
1da177e4 LT |
1830 | { |
1831 | struct inode *inode = dentry->d_inode; | |
1832 | int err; | |
1833 | ||
1834 | if (acc == MAY_NOP) | |
1835 | return 0; | |
1836 | #if 0 | |
1837 | dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n", | |
1838 | acc, | |
1839 | (acc & MAY_READ)? " read" : "", | |
1840 | (acc & MAY_WRITE)? " write" : "", | |
1841 | (acc & MAY_EXEC)? " exec" : "", | |
1842 | (acc & MAY_SATTR)? " sattr" : "", | |
1843 | (acc & MAY_TRUNC)? " trunc" : "", | |
1844 | (acc & MAY_LOCK)? " lock" : "", | |
1845 | (acc & MAY_OWNER_OVERRIDE)? " owneroverride" : "", | |
1846 | inode->i_mode, | |
1847 | IS_IMMUTABLE(inode)? " immut" : "", | |
1848 | IS_APPEND(inode)? " append" : "", | |
1849 | IS_RDONLY(inode)? " ro" : ""); | |
1850 | dprintk(" owner %d/%d user %d/%d\n", | |
1851 | inode->i_uid, inode->i_gid, current->fsuid, current->fsgid); | |
1852 | #endif | |
1853 | ||
1854 | /* Normally we reject any write/sattr etc access on a read-only file | |
1855 | * system. But if it is IRIX doing check on write-access for a | |
1856 | * device special file, we ignore rofs. | |
1857 | */ | |
1858 | if (!(acc & MAY_LOCAL_ACCESS)) | |
1859 | if (acc & (MAY_WRITE | MAY_SATTR | MAY_TRUNC)) { | |
c7d51402 | 1860 | if (exp_rdonly(rqstp, exp) || IS_RDONLY(inode)) |
1da177e4 LT |
1861 | return nfserr_rofs; |
1862 | if (/* (acc & MAY_WRITE) && */ IS_IMMUTABLE(inode)) | |
1863 | return nfserr_perm; | |
1864 | } | |
1865 | if ((acc & MAY_TRUNC) && IS_APPEND(inode)) | |
1866 | return nfserr_perm; | |
1867 | ||
1868 | if (acc & MAY_LOCK) { | |
1869 | /* If we cannot rely on authentication in NLM requests, | |
1870 | * just allow locks, otherwise require read permission, or | |
1871 | * ownership | |
1872 | */ | |
1873 | if (exp->ex_flags & NFSEXP_NOAUTHNLM) | |
1874 | return 0; | |
1875 | else | |
1876 | acc = MAY_READ | MAY_OWNER_OVERRIDE; | |
1877 | } | |
1878 | /* | |
1879 | * The file owner always gets access permission for accesses that | |
1880 | * would normally be checked at open time. This is to make | |
1881 | * file access work even when the client has done a fchmod(fd, 0). | |
1882 | * | |
1883 | * However, `cp foo bar' should fail nevertheless when bar is | |
1884 | * readonly. A sensible way to do this might be to reject all | |
1885 | * attempts to truncate a read-only file, because a creat() call | |
1886 | * always implies file truncation. | |
1887 | * ... but this isn't really fair. A process may reasonably call | |
1888 | * ftruncate on an open file descriptor on a file with perm 000. | |
1889 | * We must trust the client to do permission checking - using "ACCESS" | |
1890 | * with NFSv3. | |
1891 | */ | |
1892 | if ((acc & MAY_OWNER_OVERRIDE) && | |
1893 | inode->i_uid == current->fsuid) | |
1894 | return 0; | |
1895 | ||
1896 | err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL); | |
1897 | ||
1898 | /* Allow read access to binaries even when mode 111 */ | |
1899 | if (err == -EACCES && S_ISREG(inode->i_mode) && | |
1900 | acc == (MAY_READ | MAY_OWNER_OVERRIDE)) | |
1901 | err = permission(inode, MAY_EXEC, NULL); | |
1902 | ||
1903 | return err? nfserrno(err) : 0; | |
1904 | } | |
1905 | ||
1906 | void | |
1907 | nfsd_racache_shutdown(void) | |
1908 | { | |
fce1456a | 1909 | if (!raparml) |
1da177e4 LT |
1910 | return; |
1911 | dprintk("nfsd: freeing readahead buffers.\n"); | |
1912 | kfree(raparml); | |
fce1456a | 1913 | raparml = NULL; |
1da177e4 LT |
1914 | } |
1915 | /* | |
1916 | * Initialize readahead param cache | |
1917 | */ | |
1918 | int | |
1919 | nfsd_racache_init(int cache_size) | |
1920 | { | |
1921 | int i; | |
fce1456a GB |
1922 | int j = 0; |
1923 | int nperbucket; | |
1da177e4 | 1924 | |
fce1456a GB |
1925 | |
1926 | if (raparml) | |
1da177e4 | 1927 | return 0; |
fce1456a GB |
1928 | if (cache_size < 2*RAPARM_HASH_SIZE) |
1929 | cache_size = 2*RAPARM_HASH_SIZE; | |
4b3bb06b YB |
1930 | raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL); |
1931 | ||
1932 | if (!raparml) { | |
1da177e4 | 1933 | printk(KERN_WARNING |
4b3bb06b | 1934 | "nfsd: Could not allocate memory read-ahead cache.\n"); |
1da177e4 LT |
1935 | return -ENOMEM; |
1936 | } | |
4b3bb06b YB |
1937 | |
1938 | dprintk("nfsd: allocating %d readahead buffers.\n", cache_size); | |
1939 | for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) { | |
1940 | raparm_hash[i].pb_head = NULL; | |
1941 | spin_lock_init(&raparm_hash[i].pb_lock); | |
1942 | } | |
a280df32 | 1943 | nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE); |
4b3bb06b YB |
1944 | for (i = 0; i < cache_size - 1; i++) { |
1945 | if (i % nperbucket == 0) | |
1946 | raparm_hash[j++].pb_head = raparml + i; | |
1947 | if (i % nperbucket < nperbucket-1) | |
1948 | raparml[i].p_next = raparml + i + 1; | |
1949 | } | |
1950 | ||
1da177e4 LT |
1951 | nfsdstats.ra_size = cache_size; |
1952 | return 0; | |
1953 | } | |
a257cdd0 AG |
1954 | |
1955 | #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) | |
1956 | struct posix_acl * | |
1957 | nfsd_get_posix_acl(struct svc_fh *fhp, int type) | |
1958 | { | |
1959 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1960 | char *name; | |
1961 | void *value = NULL; | |
1962 | ssize_t size; | |
1963 | struct posix_acl *acl; | |
1964 | ||
5be196e5 CH |
1965 | if (!IS_POSIXACL(inode)) |
1966 | return ERR_PTR(-EOPNOTSUPP); | |
1967 | ||
1968 | switch (type) { | |
1969 | case ACL_TYPE_ACCESS: | |
1970 | name = POSIX_ACL_XATTR_ACCESS; | |
1971 | break; | |
1972 | case ACL_TYPE_DEFAULT: | |
1973 | name = POSIX_ACL_XATTR_DEFAULT; | |
1974 | break; | |
1975 | default: | |
a257cdd0 | 1976 | return ERR_PTR(-EOPNOTSUPP); |
a257cdd0 AG |
1977 | } |
1978 | ||
5be196e5 CH |
1979 | size = nfsd_getxattr(fhp->fh_dentry, name, &value); |
1980 | if (size < 0) | |
1981 | return ERR_PTR(size); | |
a257cdd0 | 1982 | |
a257cdd0 | 1983 | acl = posix_acl_from_xattr(value, size); |
a257cdd0 AG |
1984 | kfree(value); |
1985 | return acl; | |
1986 | } | |
1987 | ||
1988 | int | |
1989 | nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl) | |
1990 | { | |
1991 | struct inode *inode = fhp->fh_dentry->d_inode; | |
1992 | char *name; | |
1993 | void *value = NULL; | |
1994 | size_t size; | |
1995 | int error; | |
1996 | ||
1997 | if (!IS_POSIXACL(inode) || !inode->i_op || | |
1998 | !inode->i_op->setxattr || !inode->i_op->removexattr) | |
1999 | return -EOPNOTSUPP; | |
2000 | switch(type) { | |
2001 | case ACL_TYPE_ACCESS: | |
334a13ec | 2002 | name = POSIX_ACL_XATTR_ACCESS; |
a257cdd0 AG |
2003 | break; |
2004 | case ACL_TYPE_DEFAULT: | |
334a13ec | 2005 | name = POSIX_ACL_XATTR_DEFAULT; |
a257cdd0 AG |
2006 | break; |
2007 | default: | |
2008 | return -EOPNOTSUPP; | |
2009 | } | |
2010 | ||
2011 | if (acl && acl->a_count) { | |
334a13ec | 2012 | size = posix_acl_xattr_size(acl->a_count); |
a257cdd0 AG |
2013 | value = kmalloc(size, GFP_KERNEL); |
2014 | if (!value) | |
2015 | return -ENOMEM; | |
9ccfc29c FM |
2016 | error = posix_acl_to_xattr(acl, value, size); |
2017 | if (error < 0) | |
a257cdd0 | 2018 | goto getout; |
9ccfc29c | 2019 | size = error; |
a257cdd0 AG |
2020 | } else |
2021 | size = 0; | |
2022 | ||
a257cdd0 | 2023 | if (size) |
5be196e5 | 2024 | error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0); |
a257cdd0 AG |
2025 | else { |
2026 | if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT) | |
2027 | error = 0; | |
2028 | else { | |
5be196e5 | 2029 | error = vfs_removexattr(fhp->fh_dentry, name); |
a257cdd0 AG |
2030 | if (error == -ENODATA) |
2031 | error = 0; | |
2032 | } | |
2033 | } | |
2034 | ||
2035 | getout: | |
2036 | kfree(value); | |
2037 | return error; | |
2038 | } | |
2039 | #endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */ |