blob: 0f4481e0502d132b624e952f3c2cbad50163c4ab [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#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
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#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>
Jens Axboed6b29d72007-06-04 09:59:47 +020026#include <linux/splice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#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>
Robert Love0eeca282005-07-12 17:06:03 -040047#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/posix_acl.h>
49#include <linux/posix_acl_xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/xattr.h>
Christoph Hellwig5be196e2006-01-09 20:51:55 -080051#ifdef CONFIG_NFSD_V4
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#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 */
Greg Banksfce14562006-10-04 02:15:49 -070057#include <linux/jhash.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
59#include <asm/uaccess.h>
60
61#define NFSDDBG_FACILITY NFSDDBG_FILEOP
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63
Linus Torvalds1da177e2005-04-16 15:20:36 -070064/*
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 */
71struct 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;
Greg Banksfce14562006-10-04 02:15:49 -070078 unsigned int p_hindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079};
80
Greg Banksfce14562006-10-04 02:15:49 -070081struct raparm_hbucket {
82 struct raparms *pb_head;
83 spinlock_t pb_lock;
84} ____cacheline_aligned_in_smp;
85
Linus Torvalds1da177e2005-04-16 15:20:36 -070086static struct raparms * raparml;
Greg Banksfce14562006-10-04 02:15:49 -070087#define RAPARM_HASH_BITS 4
88#define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS)
89#define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1)
90static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92/*
93 * Called from nfsd_lookup and encode_dirent. Check if we have crossed
94 * a mount point.
J.Bruce Fieldse0bb89e2006-12-13 00:35:25 -080095 * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged,
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 * or nfs_ok having possibly changed *dpp and *expp
97 */
98int
99nfsd_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;
Jan Blunck54775492008-02-14 19:38:39 -0800104 struct vfsmount *mnt = mntget(exp->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 struct dentry *mounts = dget(dentry);
Al Viro6264d692006-10-19 23:28:58 -0700106 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107
108 while (follow_down(&mnt,&mounts)&&d_mountpoint(mounts));
109
J. Bruce Fields0989a782007-07-17 04:04:44 -0700110 exp2 = rqst_exp_get_by_name(rqstp, mnt, mounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 if (IS_ERR(exp2)) {
Neil Browna1033be2007-09-05 17:22:12 -0400112 if (PTR_ERR(exp2) != -ENOENT)
113 err = PTR_ERR(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 dput(mounts);
115 mntput(mnt);
116 goto out;
117 }
J. Bruce Fields5d3dbbe2007-07-19 01:49:19 -0700118 if ((exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 /* successfully crossed mount point */
120 exp_put(exp);
121 *expp = exp2;
122 dput(dentry);
123 *dpp = mounts;
124 } else {
J. Bruce Fields5d3dbbe2007-07-19 01:49:19 -0700125 exp_put(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 dput(mounts);
127 }
128 mntput(mnt);
129out:
130 return err;
131}
132
Al Viro6264d692006-10-19 23:28:58 -0700133__be32
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700134nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400135 const char *name, unsigned int len,
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700136 struct svc_export **exp_ret, struct dentry **dentry_ret)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137{
138 struct svc_export *exp;
139 struct dentry *dparent;
140 struct dentry *dentry;
Al Viro6264d692006-10-19 23:28:58 -0700141 __be32 err;
142 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144 dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name);
145
146 /* Obtain dentry and export. */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200147 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 if (err)
149 return err;
150
151 dparent = fhp->fh_dentry;
152 exp = fhp->fh_export;
153 exp_get(exp);
154
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 /* Lookup the name, but don't follow links */
156 if (isdotent(name, len)) {
157 if (len==1)
158 dentry = dget(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800159 else if (dparent != exp->ex_path.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 dentry = dget_parent(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800161 else if (!EX_NOHIDE(exp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 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;
Jan Blunck54775492008-02-14 19:38:39 -0800167 struct vfsmount *mnt = mntget(exp->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 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
J. Bruce Fields0989a782007-07-17 04:04:44 -0700175 exp2 = rqst_exp_parent(rqstp, mnt, dentry);
J. Bruce Fields2d3bb252007-07-17 04:04:40 -0700176 if (PTR_ERR(exp2) == -ENOENT) {
177 dput(dentry);
178 dentry = dget(dparent);
179 } else if (IS_ERR(exp2)) {
Al Viro6264d692006-10-19 23:28:58 -0700180 host_err = PTR_ERR(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 dput(dentry);
182 mntput(mnt);
183 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 } 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);
Al Viro6264d692006-10-19 23:28:58 -0700193 host_err = PTR_ERR(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 if (IS_ERR(dentry))
195 goto out_nfserr;
196 /*
197 * check if we have crossed a mount point ...
198 */
199 if (d_mountpoint(dentry)) {
Al Viro6264d692006-10-19 23:28:58 -0700200 if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 dput(dentry);
202 goto out_nfserr;
203 }
204 }
205 }
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700206 *dentry_ret = dentry;
207 *exp_ret = exp;
208 return 0;
209
210out_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
228nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400229 unsigned int len, struct svc_fh *resfh)
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700230{
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;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700238 err = check_nfsd_access(exp, rqstp);
239 if (err)
240 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 /*
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;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700248out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 exp_put(exp);
251 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252}
253
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255/*
256 * Set various file attributes.
257 * N.B. After this call fhp needs an fh_put
258 */
Al Viro6264d692006-10-19 23:28:58 -0700259__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260nfsd_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;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200265 int accmode = NFSD_MAY_SATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 int ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700267 __be32 err;
268 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 int size_change = 0;
270
271 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200272 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 if (iap->ia_valid & ATTR_SIZE)
274 ftype = S_IFREG;
275
276 /* Get inode */
277 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800278 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 goto out;
280
281 dentry = fhp->fh_dentry;
282 inode = dentry->d_inode;
283
NeilBrown15b7a1b2005-11-07 01:00:23 -0800284 /* Ignore any mode updates on symlinks */
285 if (S_ISLNK(inode->i_mode))
286 iap->ia_valid &= ~ATTR_MODE;
287
288 if (!iap->ia_valid)
289 goto out;
290
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000291 /*
292 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 * which only requires access, and "set-[ac]time-to-X" which
294 * requires ownership.
295 * So if it looks like it might be "set both to the same time which
296 * is close to now", and if inode_change_ok fails, then we
297 * convert to "set to now" instead of "set to explicit time"
298 *
299 * We only call inode_change_ok as the last test as technically
300 * it is not an interface that we should be using. It is only
301 * valid if the filesystem does not define it's own i_op->setattr.
302 */
303#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
304#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000305 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
306 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
307 /*
308 * Looks probable.
309 *
310 * Now just make sure time is in the right ballpark.
311 * Solaris, at least, doesn't seem to care what the time
312 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000314 time_t delta = iap->ia_atime.tv_sec - get_seconds();
315 if (delta < 0)
316 delta = -delta;
317 if (delta < MAX_TOUCH_TIME_ERROR &&
318 inode_change_ok(inode, iap) != 0) {
319 /*
320 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
321 * This will cause notify_change to set these times
322 * to "now"
323 */
324 iap->ia_valid &= ~BOTH_TIME_SET;
325 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 }
327
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000328 /*
329 * The size case is special.
330 * It changes the file as well as the attributes.
331 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 if (iap->ia_valid & ATTR_SIZE) {
333 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200334 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
335 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 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 */
Al Viro6264d692006-10-19 23:28:58 -0700344 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 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 goto out_nfserr;
349
Al Viro6264d692006-10-19 23:28:58 -0700350 host_err = get_write_access(inode);
351 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 goto out_nfserr;
353
354 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700355 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
356 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 put_write_access(inode);
358 goto out_nfserr;
359 }
360 DQUOT_INIT(inode);
361 }
362
Jeff Laytonca456252008-04-16 16:28:47 -0400363 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 if (iap->ia_valid & ATTR_MODE) {
365 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400366 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400367 }
368
369 /* Revoke setuid/setgid on chown */
370 if (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
371 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid)) {
372 iap->ia_valid |= ATTR_KILL_PRIV;
373 if (iap->ia_valid & ATTR_MODE) {
374 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700375 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400376 if (iap->ia_mode & S_IXGRP)
377 iap->ia_mode &= ~S_ISGID;
378 } else {
379 /* set ATTR_KILL_* bits and let VFS handle it */
380 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700381 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 }
383
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 /* Change the attributes. */
385
386 iap->ia_valid |= ATTR_CTIME;
387
388 err = nfserr_notsync;
389 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
390 fh_lock(fhp);
Al Viro6264d692006-10-19 23:28:58 -0700391 host_err = notify_change(dentry, iap);
392 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 fh_unlock(fhp);
394 }
395 if (size_change)
396 put_write_access(inode);
397 if (!err)
398 if (EX_ISSYNC(fhp->fh_export))
399 write_inode_now(inode, 1);
400out:
401 return err;
402
403out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700404 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 goto out;
406}
407
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800408#if defined(CONFIG_NFSD_V2_ACL) || \
409 defined(CONFIG_NFSD_V3_ACL) || \
410 defined(CONFIG_NFSD_V4)
411static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
412{
413 ssize_t buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800415 buflen = vfs_getxattr(dentry, key, NULL, 0);
416 if (buflen <= 0)
417 return buflen;
418
419 *buf = kmalloc(buflen, GFP_KERNEL);
420 if (!*buf)
421 return -ENOMEM;
422
NeilBrownb5872b02006-04-10 22:55:26 -0700423 return vfs_getxattr(dentry, key, *buf, buflen);
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800424}
425#endif
426
427#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428static int
429set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
430{
431 int len;
432 size_t buflen;
433 char *buf = NULL;
434 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436 buflen = posix_acl_xattr_size(pacl->a_count);
437 buf = kmalloc(buflen, GFP_KERNEL);
438 error = -ENOMEM;
439 if (buf == NULL)
440 goto out;
441
442 len = posix_acl_to_xattr(pacl, buf, buflen);
443 if (len < 0) {
444 error = len;
445 goto out;
446 }
447
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800448 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449out:
450 kfree(buf);
451 return error;
452}
453
Al Viro6264d692006-10-19 23:28:58 -0700454__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
456 struct nfs4_acl *acl)
457{
Al Viro6264d692006-10-19 23:28:58 -0700458 __be32 error;
459 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 struct dentry *dentry;
461 struct inode *inode;
462 struct posix_acl *pacl = NULL, *dpacl = NULL;
463 unsigned int flags = 0;
464
465 /* Get inode */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200466 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700468 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470 dentry = fhp->fh_dentry;
471 inode = dentry->d_inode;
472 if (S_ISDIR(inode->i_mode))
473 flags = NFS4_ACL_DIR;
474
Al Viro6264d692006-10-19 23:28:58 -0700475 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
476 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700477 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700478 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 goto out_nfserr;
480
Al Viro6264d692006-10-19 23:28:58 -0700481 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
482 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700483 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700485 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700486 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700488out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 posix_acl_release(pacl);
490 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800492 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700493 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800494 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700495 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496}
497
498static struct posix_acl *
499_get_posix_acl(struct dentry *dentry, char *key)
500{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800501 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800503 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800505 buflen = nfsd_getxattr(dentry, key, &buf);
506 if (!buflen)
507 buflen = -ENODATA;
508 if (buflen <= 0)
509 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
511 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 kfree(buf);
513 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514}
515
516int
517nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
518{
519 struct inode *inode = dentry->d_inode;
520 int error = 0;
521 struct posix_acl *pacl = NULL, *dpacl = NULL;
522 unsigned int flags = 0;
523
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700524 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
526 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
527 if (IS_ERR(pacl)) {
528 error = PTR_ERR(pacl);
529 pacl = NULL;
530 goto out;
531 }
532
533 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700534 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
536 dpacl = NULL;
537 else if (IS_ERR(dpacl)) {
538 error = PTR_ERR(dpacl);
539 dpacl = NULL;
540 goto out;
541 }
542 flags = NFS4_ACL_DIR;
543 }
544
545 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
546 if (IS_ERR(*acl)) {
547 error = PTR_ERR(*acl);
548 *acl = NULL;
549 }
550 out:
551 posix_acl_release(pacl);
552 posix_acl_release(dpacl);
553 return error;
554}
555
556#endif /* defined(CONFIG_NFS_V4) */
557
558#ifdef CONFIG_NFSD_V3
559/*
560 * Check server access rights to a file system object
561 */
562struct accessmap {
563 u32 access;
564 int how;
565};
566static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200567 { NFS3_ACCESS_READ, NFSD_MAY_READ },
568 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
569 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
570 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 { 0, 0 }
573};
574
575static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200576 { NFS3_ACCESS_READ, NFSD_MAY_READ },
577 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
578 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
579 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
580 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
582 { 0, 0 }
583};
584
585static struct accessmap nfs3_anyaccess[] = {
586 /* Some clients - Solaris 2.6 at least, make an access call
587 * to the server to check for access for things like /dev/null
588 * (which really, the server doesn't care about). So
589 * We provide simple access checking for them, looking
590 * mainly at mode bits, and we make sure to ignore read-only
591 * filesystem checks
592 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200593 { NFS3_ACCESS_READ, NFSD_MAY_READ },
594 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
595 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
596 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
598 { 0, 0 }
599};
600
Al Viro6264d692006-10-19 23:28:58 -0700601__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
603{
604 struct accessmap *map;
605 struct svc_export *export;
606 struct dentry *dentry;
607 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700608 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200610 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 if (error)
612 goto out;
613
614 export = fhp->fh_export;
615 dentry = fhp->fh_dentry;
616
617 if (S_ISREG(dentry->d_inode->i_mode))
618 map = nfs3_regaccess;
619 else if (S_ISDIR(dentry->d_inode->i_mode))
620 map = nfs3_diraccess;
621 else
622 map = nfs3_anyaccess;
623
624
625 query = *access;
626 for (; map->access; map++) {
627 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700628 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 sresult |= map->access;
631
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700632 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 switch (err2) {
634 case nfs_ok:
635 result |= map->access;
636 break;
637
638 /* the following error codes just mean the access was not allowed,
639 * rather than an error occurred */
640 case nfserr_rofs:
641 case nfserr_acces:
642 case nfserr_perm:
643 /* simply don't "or" in the access bit. */
644 break;
645 default:
646 error = err2;
647 goto out;
648 }
649 }
650 }
651 *access = result;
652 if (supported)
653 *supported = sresult;
654
655 out:
656 return error;
657}
658#endif /* CONFIG_NFSD_V3 */
659
660
661
662/*
663 * Open an existing file or directory.
664 * The access argument indicates the type of open (read/write/lock)
665 * N.B. After this call fhp needs an fh_put
666 */
Al Viro6264d692006-10-19 23:28:58 -0700667__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
669 int access, struct file **filp)
670{
671 struct dentry *dentry;
672 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700673 int flags = O_RDONLY|O_LARGEFILE;
674 __be32 err;
675 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
677 /*
678 * If we get here, then the client has already done an "open",
679 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
680 * in case a chmod has now revoked permission.
681 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200682 err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 if (err)
684 goto out;
685
686 dentry = fhp->fh_dentry;
687 inode = dentry->d_inode;
688
689 /* Disallow write access to files with the append-only bit set
690 * or any access when mandatory locking enabled
691 */
692 err = nfserr_perm;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200693 if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400695 /*
696 * We must ignore files (but only files) which might have mandatory
697 * locks on them because there is no way to know if the accesser has
698 * the lock.
699 */
700 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 goto out;
702
703 if (!inode->i_fop)
704 goto out;
705
706 /*
707 * Check to see if there are any leases on this file.
708 * This may block while leases are broken.
709 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200710 host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0));
Al Viro6264d692006-10-19 23:28:58 -0700711 if (host_err == -EWOULDBLOCK)
712 host_err = -ETIMEDOUT;
713 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 goto out_nfserr;
715
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200716 if (access & NFSD_MAY_WRITE) {
717 if (access & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700718 flags = O_RDWR|O_LARGEFILE;
719 else
720 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
722 DQUOT_INIT(inode);
723 }
Jan Blunck54775492008-02-14 19:38:39 -0800724 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
725 flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700727 host_err = PTR_ERR(*filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700729 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730out:
731 return err;
732}
733
734/*
735 * Close a file.
736 */
737void
738nfsd_close(struct file *filp)
739{
740 fput(filp);
741}
742
743/*
744 * Sync a file
745 * As this calls fsync (not fdatasync) there is no need for a write_inode
746 * after it.
747 */
David Shawa334de22006-01-06 00:19:58 -0800748static inline int nfsd_dosync(struct file *filp, struct dentry *dp,
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800749 const struct file_operations *fop)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750{
751 struct inode *inode = dp->d_inode;
752 int (*fsync) (struct file *, struct dentry *, int);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800753 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800755 err = filemap_fdatawrite(inode->i_mapping);
756 if (err == 0 && fop && (fsync = fop->fsync))
757 err = fsync(filp, dp, 0);
758 if (err == 0)
759 err = filemap_fdatawait(inode->i_mapping);
David Shawa334de22006-01-06 00:19:58 -0800760
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800761 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762}
763
764
David Shawa334de22006-01-06 00:19:58 -0800765static int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766nfsd_sync(struct file *filp)
767{
David Shawa334de22006-01-06 00:19:58 -0800768 int err;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800769 struct inode *inode = filp->f_path.dentry->d_inode;
770 dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800771 mutex_lock(&inode->i_mutex);
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800772 err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800773 mutex_unlock(&inode->i_mutex);
David Shawa334de22006-01-06 00:19:58 -0800774
775 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776}
777
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800778int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779nfsd_sync_dir(struct dentry *dp)
780{
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800781 return nfsd_dosync(NULL, dp, dp->d_inode->i_fop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
783
784/*
785 * Obtain the readahead parameters for the file
786 * specified by (dev, ino).
787 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789static inline struct raparms *
790nfsd_get_raparms(dev_t dev, ino_t ino)
791{
792 struct raparms *ra, **rap, **frap = NULL;
793 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700794 unsigned int hash;
795 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Greg Banksfce14562006-10-04 02:15:49 -0700797 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
798 rab = &raparm_hash[hash];
799
800 spin_lock(&rab->pb_lock);
801 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 if (ra->p_ino == ino && ra->p_dev == dev)
803 goto found;
804 depth++;
805 if (ra->p_count == 0)
806 frap = rap;
807 }
808 depth = nfsdstats.ra_size*11/10;
809 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700810 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 return NULL;
812 }
813 rap = frap;
814 ra = *frap;
815 ra->p_dev = dev;
816 ra->p_ino = ino;
817 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700818 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819found:
Greg Banksfce14562006-10-04 02:15:49 -0700820 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700822 ra->p_next = rab->pb_head;
823 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 }
825 ra->p_count++;
826 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700827 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 return ra;
829}
830
831/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200832 * Grab and keep cached pages associated with a file in the svc_rqst
833 * so that they can be passed to the network sendmsg/sendpage routines
834 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 */
836static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200837nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
838 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839{
Jens Axboecf8208d2007-06-12 21:22:14 +0200840 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700841 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200842 struct page *page = buf->page;
843 size_t size;
844 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845
Jens Axboecac36bb02007-06-14 13:10:48 +0200846 ret = buf->ops->confirm(pipe, buf);
Jens Axboecf8208d2007-06-12 21:22:14 +0200847 if (unlikely(ret))
848 return ret;
849
850 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
852 if (rqstp->rq_res.page_len == 0) {
853 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700854 put_page(*pp);
855 *pp = page;
856 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200857 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700859 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800861 if (*pp)
862 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700863 *pp = page;
864 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700866 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 return size;
870}
871
Jens Axboecf8208d2007-06-12 21:22:14 +0200872static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
873 struct splice_desc *sd)
874{
875 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
876}
877
Dave Hansena8754be2007-10-16 23:31:15 -0700878static inline int svc_msnfs(struct svc_fh *ffhp)
879{
880#ifdef MSNFS
881 return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS);
882#else
883 return 0;
884#endif
885}
886
Al Viro6264d692006-10-19 23:28:58 -0700887static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
889 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
890{
891 struct inode *inode;
892 struct raparms *ra;
893 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700894 __be32 err;
895 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896
897 err = nfserr_perm;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800898 inode = file->f_path.dentry->d_inode;
Dave Hansena8754be2007-10-16 23:31:15 -0700899
900 if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
903 /* Get readahead parameters */
904 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
905
906 if (ra && ra->p_set)
907 file->f_ra = ra->p_ra;
908
Jens Axboecf8208d2007-06-12 21:22:14 +0200909 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
910 struct splice_desc sd = {
911 .len = 0,
912 .total_len = *count,
913 .pos = offset,
914 .u.data = rqstp,
915 };
916
Jens Axboe4fbef202007-07-13 22:42:20 +0200917 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200918 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 } else {
920 oldfs = get_fs();
921 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700922 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 set_fs(oldfs);
924 }
925
926 /* Write back readahead params */
927 if (ra) {
Greg Banksfce14562006-10-04 02:15:49 -0700928 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
929 spin_lock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 ra->p_ra = file->f_ra;
931 ra->p_set = 1;
932 ra->p_count--;
Greg Banksfce14562006-10-04 02:15:49 -0700933 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 }
935
Al Viro6264d692006-10-19 23:28:58 -0700936 if (host_err >= 0) {
937 nfsdstats.io_read += host_err;
938 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 err = 0;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800940 fsnotify_access(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 } else
Al Viro6264d692006-10-19 23:28:58 -0700942 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943out:
944 return err;
945}
946
Neil Brown9f708e42006-01-06 00:19:59 -0800947static void kill_suid(struct dentry *dentry)
948{
949 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700950 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800951
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800952 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800953 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800954 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800955}
956
Al Viro6264d692006-10-19 23:28:58 -0700957static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
959 loff_t offset, struct kvec *vec, int vlen,
960 unsigned long cnt, int *stablep)
961{
962 struct svc_export *exp;
963 struct dentry *dentry;
964 struct inode *inode;
965 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700966 __be32 err = 0;
967 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 int stable = *stablep;
969
NeilBrown45bd3b32006-01-18 17:43:50 -0800970#ifdef MSNFS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 err = nfserr_perm;
972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800974 (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 goto out;
976#endif
977
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800978 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 inode = dentry->d_inode;
980 exp = fhp->fh_export;
981
982 /*
983 * Request sync writes if
984 * - the sync export option has been set, or
985 * - the client requested O_SYNC behavior (NFSv3 feature).
986 * - The file system doesn't support fsync().
987 * When gathered writes have been configured for this volume,
988 * flushing the data to disk is handled separately below.
989 */
990
Harvey Harrison3ba15142008-02-20 12:49:02 -0800991 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 stable = 2;
993 *stablep = 2; /* FILE_SYNC */
994 }
995
996 if (!EX_ISSYNC(exp))
997 stable = 0;
998 if (stable && !EX_WGATHER(exp))
999 file->f_flags |= O_SYNC;
1000
1001 /* Write the data. */
1002 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001003 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 set_fs(oldfs);
Al Viro6264d692006-10-19 23:28:58 -07001005 if (host_err >= 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 nfsdstats.io_write += cnt;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001007 fsnotify_modify(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 }
1009
1010 /* clear setuid/setgid flag after write */
Al Viro6264d692006-10-19 23:28:58 -07001011 if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID)))
Neil Brown9f708e42006-01-06 00:19:59 -08001012 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013
Al Viro6264d692006-10-19 23:28:58 -07001014 if (host_err >= 0 && stable) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 static ino_t last_ino;
1016 static dev_t last_dev;
1017
1018 /*
1019 * Gathered writes: If another process is currently
1020 * writing to the file, there's a high chance
1021 * this is another nfsd (triggered by a bulk write
1022 * from a client's biod). Rather than syncing the
1023 * file with each write request, we sleep for 10 msec.
1024 *
1025 * I don't know if this roughly approximates
1026 * C. Juszak's idea of gathered writes, but it's a
1027 * nice and simple solution (IMHO), and it seems to
1028 * work:-)
1029 */
1030 if (EX_WGATHER(exp)) {
1031 if (atomic_read(&inode->i_writecount) > 1
1032 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001033 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 msleep(10);
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001035 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 }
1037
1038 if (inode->i_state & I_DIRTY) {
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001039 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Al Viro6264d692006-10-19 23:28:58 -07001040 host_err=nfsd_sync(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 }
1042#if 0
1043 wake_up(&inode->i_wait);
1044#endif
1045 }
1046 last_ino = inode->i_ino;
1047 last_dev = inode->i_sb->s_dev;
1048 }
1049
Al Viro6264d692006-10-19 23:28:58 -07001050 dprintk("nfsd: write complete host_err=%d\n", host_err);
1051 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 err = 0;
1053 else
Al Viro6264d692006-10-19 23:28:58 -07001054 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055out:
1056 return err;
1057}
1058
1059/*
1060 * Read data from a file. count must contain the requested read count
1061 * on entry. On return, *count contains the number of bytes actually read.
1062 * N.B. After this call fhp needs an fh_put
1063 */
Al Viro6264d692006-10-19 23:28:58 -07001064__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1066 loff_t offset, struct kvec *vec, int vlen,
1067 unsigned long *count)
1068{
Al Viro6264d692006-10-19 23:28:58 -07001069 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
1071 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001072 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001073 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 if (err)
1075 goto out;
1076 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1077 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001078 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 if (err)
1080 goto out;
1081 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1082 nfsd_close(file);
1083 }
1084out:
1085 return err;
1086}
1087
1088/*
1089 * Write data to a file.
1090 * The stable flag requests synchronous writes.
1091 * N.B. After this call fhp needs an fh_put
1092 */
Al Viro6264d692006-10-19 23:28:58 -07001093__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1095 loff_t offset, struct kvec *vec, int vlen, unsigned long cnt,
1096 int *stablep)
1097{
Al Viro6264d692006-10-19 23:28:58 -07001098 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
1100 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001101 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001102 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 if (err)
1104 goto out;
1105 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1106 stablep);
1107 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001108 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 if (err)
1110 goto out;
1111
1112 if (cnt)
1113 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1114 cnt, stablep);
1115 nfsd_close(file);
1116 }
1117out:
1118 return err;
1119}
1120
1121#ifdef CONFIG_NFSD_V3
1122/*
1123 * Commit all pending writes to stable storage.
1124 * Strictly speaking, we could sync just the indicated file region here,
1125 * but there's currently no way we can ask the VFS to do so.
1126 *
1127 * Unfortunately we cannot lock the file to make sure we return full WCC
1128 * data to the client, as locking happens lower down in the filesystem.
1129 */
Al Viro6264d692006-10-19 23:28:58 -07001130__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1132 loff_t offset, unsigned long count)
1133{
1134 struct file *file;
Al Viro6264d692006-10-19 23:28:58 -07001135 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137 if ((u64)count > ~(u64)offset)
1138 return nfserr_inval;
1139
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001140 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
1141 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 return err;
1143 if (EX_ISSYNC(fhp->fh_export)) {
1144 if (file->f_op && file->f_op->fsync) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001145 err = nfserrno(nfsd_sync(file));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 } else {
1147 err = nfserr_notsupp;
1148 }
1149 }
1150
1151 nfsd_close(file);
1152 return err;
1153}
1154#endif /* CONFIG_NFSD_V3 */
1155
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001156static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001157nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1158 struct iattr *iap)
1159{
1160 /*
1161 * Mode has already been set earlier in create:
1162 */
1163 iap->ia_valid &= ~ATTR_MODE;
1164 /*
1165 * Setting uid/gid works only for root. Irix appears to
1166 * send along the gid on create when it tries to implement
1167 * setgid directories via NFS:
1168 */
1169 if (current->fsuid != 0)
1170 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1171 if (iap->ia_valid)
1172 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1173 return 0;
1174}
1175
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176/*
1177 * Create a file (regular, directory, device, fifo); UNIX sockets
1178 * not yet implemented.
1179 * If the response fh has been verified, the parent directory should
1180 * already be locked. Note that the parent directory is left locked.
1181 *
1182 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1183 */
Al Viro6264d692006-10-19 23:28:58 -07001184__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1186 char *fname, int flen, struct iattr *iap,
1187 int type, dev_t rdev, struct svc_fh *resfhp)
1188{
1189 struct dentry *dentry, *dchild = NULL;
1190 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001191 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001192 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001193 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
1195 err = nfserr_perm;
1196 if (!flen)
1197 goto out;
1198 err = nfserr_exist;
1199 if (isdotent(fname, flen))
1200 goto out;
1201
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001202 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 if (err)
1204 goto out;
1205
1206 dentry = fhp->fh_dentry;
1207 dirp = dentry->d_inode;
1208
1209 err = nfserr_notdir;
1210 if(!dirp->i_op || !dirp->i_op->lookup)
1211 goto out;
1212 /*
1213 * Check whether the response file handle has been verified yet.
1214 * If it has, the parent directory should already be locked.
1215 */
1216 if (!resfhp->fh_dentry) {
1217 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001218 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001220 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 if (IS_ERR(dchild))
1222 goto out_nfserr;
1223 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1224 if (err)
1225 goto out;
1226 } else {
1227 /* called from nfsd_proc_create */
1228 dchild = dget(resfhp->fh_dentry);
1229 if (!fhp->fh_locked) {
1230 /* not actually possible */
1231 printk(KERN_ERR
1232 "nfsd_create: parent %s/%s not locked!\n",
1233 dentry->d_parent->d_name.name,
1234 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001235 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 goto out;
1237 }
1238 }
1239 /*
1240 * Make sure the child dentry is still negative ...
1241 */
1242 err = nfserr_exist;
1243 if (dchild->d_inode) {
1244 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1245 dentry->d_name.name, dchild->d_name.name);
1246 goto out;
1247 }
1248
1249 if (!(iap->ia_valid & ATTR_MODE))
1250 iap->ia_mode = 0;
1251 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1252
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001253 err = nfserr_inval;
1254 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1255 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1256 type);
1257 goto out;
1258 }
1259
1260 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1261 if (host_err)
1262 goto out_nfserr;
1263
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 /*
1265 * Get the dir op function pointer.
1266 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001267 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 switch (type) {
1269 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001270 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 break;
1272 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001273 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 break;
1275 case S_IFCHR:
1276 case S_IFBLK:
1277 case S_IFIFO:
1278 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001279 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001281 }
1282 if (host_err < 0) {
1283 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1284 goto out_nfserr;
1285 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286
1287 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001288 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 write_inode_now(dchild->d_inode, 1);
1290 }
1291
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001292 err2 = nfsd_create_setattr(rqstp, resfhp, iap);
1293 if (err2)
1294 err = err2;
Dave Hansen463c3192008-02-15 14:37:57 -08001295 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 /*
1297 * Update the file handle to get the new inode info.
1298 */
1299 if (!err)
1300 err = fh_update(resfhp);
1301out:
1302 if (dchild && !IS_ERR(dchild))
1303 dput(dchild);
1304 return err;
1305
1306out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001307 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 goto out;
1309}
1310
1311#ifdef CONFIG_NFSD_V3
1312/*
1313 * NFSv3 version of nfsd_create
1314 */
Al Viro6264d692006-10-19 23:28:58 -07001315__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
1317 char *fname, int flen, struct iattr *iap,
1318 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001319 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320{
1321 struct dentry *dentry, *dchild = NULL;
1322 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001323 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001324 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001325 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
1328 err = nfserr_perm;
1329 if (!flen)
1330 goto out;
1331 err = nfserr_exist;
1332 if (isdotent(fname, flen))
1333 goto out;
1334 if (!(iap->ia_valid & ATTR_MODE))
1335 iap->ia_mode = 0;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001336 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 if (err)
1338 goto out;
1339
1340 dentry = fhp->fh_dentry;
1341 dirp = dentry->d_inode;
1342
1343 /* Get all the sanity checks out of the way before
1344 * we lock the parent. */
1345 err = nfserr_notdir;
1346 if(!dirp->i_op || !dirp->i_op->lookup)
1347 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001348 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
1350 /*
1351 * Compose the response file handle.
1352 */
1353 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001354 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 if (IS_ERR(dchild))
1356 goto out_nfserr;
1357
1358 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1359 if (err)
1360 goto out;
1361
1362 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001363 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001364 * the high bit set, so just clear the high bits. If this is
1365 * ever changed to use different attrs for storing the
1366 * verifier, then do_open_lookup() will also need to be fixed
1367 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 */
1369 v_mtime = verifier[0]&0x7fffffff;
1370 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 }
1372
Dave Hansen463c3192008-02-15 14:37:57 -08001373 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1374 if (host_err)
1375 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 if (dchild->d_inode) {
1377 err = 0;
1378
1379 switch (createmode) {
1380 case NFS3_CREATE_UNCHECKED:
1381 if (! S_ISREG(dchild->d_inode->i_mode))
1382 err = nfserr_exist;
1383 else if (truncp) {
1384 /* in nfsv4, we need to treat this case a little
1385 * differently. we don't want to truncate the
1386 * file now; this would be wrong if the OPEN
1387 * fails for some other reason. furthermore,
1388 * if the size is nonzero, we should ignore it
1389 * according to spec!
1390 */
1391 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1392 }
1393 else {
1394 iap->ia_valid &= ATTR_SIZE;
1395 goto set_attr;
1396 }
1397 break;
1398 case NFS3_CREATE_EXCLUSIVE:
1399 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1400 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 && dchild->d_inode->i_size == 0 )
1402 break;
1403 /* fallthru */
1404 case NFS3_CREATE_GUARDED:
1405 err = nfserr_exist;
1406 }
Dave Hansen463c3192008-02-15 14:37:57 -08001407 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 goto out;
1409 }
1410
Al Viro6264d692006-10-19 23:28:58 -07001411 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001412 if (host_err < 0) {
1413 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001415 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001416 if (created)
1417 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001420 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 /* setattr will sync the child (or not) */
1422 }
1423
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001425 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001427 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 /* XXX someone who knows this better please fix it for nsec */
1429 iap->ia_mtime.tv_sec = v_mtime;
1430 iap->ia_atime.tv_sec = v_atime;
1431 iap->ia_mtime.tv_nsec = 0;
1432 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 }
1434
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 set_attr:
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001436 err2 = nfsd_create_setattr(rqstp, resfhp, iap);
1437 if (err2)
1438 err = err2;
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001439
Dave Hansen463c3192008-02-15 14:37:57 -08001440 mnt_drop_write(fhp->fh_export->ex_path.mnt);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001441 /*
1442 * Update the filehandle to get the new inode info.
1443 */
1444 if (!err)
1445 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
1447 out:
1448 fh_unlock(fhp);
1449 if (dchild && !IS_ERR(dchild))
1450 dput(dchild);
1451 return err;
1452
1453 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001454 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 goto out;
1456}
1457#endif /* CONFIG_NFSD_V3 */
1458
1459/*
1460 * Read a symlink. On entry, *lenp must contain the maximum path length that
1461 * fits into the buffer. On return, it contains the true length.
1462 * N.B. After this call fhp needs an fh_put
1463 */
Al Viro6264d692006-10-19 23:28:58 -07001464__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1466{
1467 struct dentry *dentry;
1468 struct inode *inode;
1469 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001470 __be32 err;
1471 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001473 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 if (err)
1475 goto out;
1476
1477 dentry = fhp->fh_dentry;
1478 inode = dentry->d_inode;
1479
1480 err = nfserr_inval;
1481 if (!inode->i_op || !inode->i_op->readlink)
1482 goto out;
1483
Jan Blunck54775492008-02-14 19:38:39 -08001484 touch_atime(fhp->fh_export->ex_path.mnt, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 /* N.B. Why does this call need a get_fs()??
1486 * Remove the set_fs and watch the fireworks:-) --okir
1487 */
1488
1489 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001490 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 set_fs(oldfs);
1492
Al Viro6264d692006-10-19 23:28:58 -07001493 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001495 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 err = 0;
1497out:
1498 return err;
1499
1500out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001501 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 goto out;
1503}
1504
1505/*
1506 * Create a symlink and look up its inode
1507 * N.B. After this call _both_ fhp and resfhp need an fh_put
1508 */
Al Viro6264d692006-10-19 23:28:58 -07001509__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1511 char *fname, int flen,
1512 char *path, int plen,
1513 struct svc_fh *resfhp,
1514 struct iattr *iap)
1515{
1516 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001517 __be32 err, cerr;
1518 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 umode_t mode;
1520
1521 err = nfserr_noent;
1522 if (!flen || !plen)
1523 goto out;
1524 err = nfserr_exist;
1525 if (isdotent(fname, flen))
1526 goto out;
1527
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001528 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 if (err)
1530 goto out;
1531 fh_lock(fhp);
1532 dentry = fhp->fh_dentry;
1533 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001534 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 if (IS_ERR(dnew))
1536 goto out_nfserr;
1537
1538 mode = S_IALLUGO;
1539 /* Only the MODE ATTRibute is even vaguely meaningful */
1540 if (iap && (iap->ia_valid & ATTR_MODE))
1541 mode = iap->ia_mode & S_IALLUGO;
1542
Dave Hansen75c3f292008-02-15 14:37:45 -08001543 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1544 if (host_err)
1545 goto out_nfserr;
1546
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 if (unlikely(path[plen] != 0)) {
1548 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1549 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001550 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 else {
1552 strncpy(path_alloced, path, plen);
1553 path_alloced[plen] = 0;
Al Viro6264d692006-10-19 23:28:58 -07001554 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 kfree(path_alloced);
1556 }
1557 } else
Al Viro6264d692006-10-19 23:28:58 -07001558 host_err = vfs_symlink(dentry->d_inode, dnew, path, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559
Al Viro6264d692006-10-19 23:28:58 -07001560 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 if (EX_ISSYNC(fhp->fh_export))
Al Viro6264d692006-10-19 23:28:58 -07001562 host_err = nfsd_sync_dir(dentry);
1563 }
1564 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 fh_unlock(fhp);
1566
Dave Hansen75c3f292008-02-15 14:37:45 -08001567 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1570 dput(dnew);
1571 if (err==0) err = cerr;
1572out:
1573 return err;
1574
1575out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001576 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 goto out;
1578}
1579
1580/*
1581 * Create a hardlink
1582 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1583 */
Al Viro6264d692006-10-19 23:28:58 -07001584__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1586 char *name, int len, struct svc_fh *tfhp)
1587{
1588 struct dentry *ddir, *dnew, *dold;
1589 struct inode *dirp, *dest;
Al Viro6264d692006-10-19 23:28:58 -07001590 __be32 err;
1591 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001593 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 if (err)
1595 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001596 err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 if (err)
1598 goto out;
1599
1600 err = nfserr_perm;
1601 if (!len)
1602 goto out;
1603 err = nfserr_exist;
1604 if (isdotent(name, len))
1605 goto out;
1606
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001607 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 ddir = ffhp->fh_dentry;
1609 dirp = ddir->d_inode;
1610
1611 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001612 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 if (IS_ERR(dnew))
1614 goto out_nfserr;
1615
1616 dold = tfhp->fh_dentry;
1617 dest = dold->d_inode;
1618
Dave Hansen75c3f292008-02-15 14:37:45 -08001619 host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt);
1620 if (host_err) {
1621 err = nfserrno(host_err);
1622 goto out_dput;
1623 }
Al Viro6264d692006-10-19 23:28:58 -07001624 host_err = vfs_link(dold, dirp, dnew);
1625 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 if (EX_ISSYNC(ffhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001627 err = nfserrno(nfsd_sync_dir(ddir));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 write_inode_now(dest, 1);
1629 }
Al Viro6264d692006-10-19 23:28:58 -07001630 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 } else {
Al Viro6264d692006-10-19 23:28:58 -07001632 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 err = nfserr_acces;
1634 else
Al Viro6264d692006-10-19 23:28:58 -07001635 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 }
Dave Hansen75c3f292008-02-15 14:37:45 -08001637 mnt_drop_write(tfhp->fh_export->ex_path.mnt);
1638out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001640out_unlock:
1641 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642out:
1643 return err;
1644
1645out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001646 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001647 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648}
1649
1650/*
1651 * Rename a file
1652 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1653 */
Al Viro6264d692006-10-19 23:28:58 -07001654__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1656 struct svc_fh *tfhp, char *tname, int tlen)
1657{
1658 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1659 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001660 __be32 err;
1661 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001663 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 if (err)
1665 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001666 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 if (err)
1668 goto out;
1669
1670 fdentry = ffhp->fh_dentry;
1671 fdir = fdentry->d_inode;
1672
1673 tdentry = tfhp->fh_dentry;
1674 tdir = tdentry->d_inode;
1675
1676 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001677 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 goto out;
1679
1680 err = nfserr_perm;
1681 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1682 goto out;
1683
1684 /* cannot use fh_lock as we need deadlock protective ordering
1685 * so do it by hand */
1686 trap = lock_rename(tdentry, fdentry);
1687 ffhp->fh_locked = tfhp->fh_locked = 1;
1688 fill_pre_wcc(ffhp);
1689 fill_pre_wcc(tfhp);
1690
1691 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001692 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (IS_ERR(odentry))
1694 goto out_nfserr;
1695
Al Viro6264d692006-10-19 23:28:58 -07001696 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 if (!odentry->d_inode)
1698 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001699 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 if (odentry == trap)
1701 goto out_dput_old;
1702
1703 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001704 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 if (IS_ERR(ndentry))
1706 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001707 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 if (ndentry == trap)
1709 goto out_dput_new;
1710
Dave Hansen9079b1e2008-02-15 14:37:49 -08001711 if (svc_msnfs(ffhp) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 ((atomic_read(&odentry->d_count) > 1)
1713 || (atomic_read(&ndentry->d_count) > 1))) {
Al Viro6264d692006-10-19 23:28:58 -07001714 host_err = -EPERM;
Dave Hansen9079b1e2008-02-15 14:37:49 -08001715 goto out_dput_new;
1716 }
1717
1718 host_err = -EXDEV;
1719 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1720 goto out_dput_new;
1721 host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt);
1722 if (host_err)
1723 goto out_dput_new;
1724
Al Viro6264d692006-10-19 23:28:58 -07001725 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
1726 if (!host_err && EX_ISSYNC(tfhp->fh_export)) {
1727 host_err = nfsd_sync_dir(tdentry);
1728 if (!host_err)
1729 host_err = nfsd_sync_dir(fdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 }
1731
Dave Hansen9079b1e2008-02-15 14:37:49 -08001732 mnt_drop_write(ffhp->fh_export->ex_path.mnt);
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 out_dput_new:
1735 dput(ndentry);
1736 out_dput_old:
1737 dput(odentry);
1738 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001739 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
1741 /* we cannot reply on fh_unlock on the two filehandles,
1742 * as that would do the wrong thing if the two directories
1743 * were the same, so again we do it by hand
1744 */
1745 fill_post_wcc(ffhp);
1746 fill_post_wcc(tfhp);
1747 unlock_rename(tdentry, fdentry);
1748 ffhp->fh_locked = tfhp->fh_locked = 0;
1749
1750out:
1751 return err;
1752}
1753
1754/*
1755 * Unlink a file or directory
1756 * N.B. After this call fhp needs an fh_put
1757 */
Al Viro6264d692006-10-19 23:28:58 -07001758__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1760 char *fname, int flen)
1761{
1762 struct dentry *dentry, *rdentry;
1763 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001764 __be32 err;
1765 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
1767 err = nfserr_acces;
1768 if (!flen || isdotent(fname, flen))
1769 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001770 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 if (err)
1772 goto out;
1773
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001774 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 dentry = fhp->fh_dentry;
1776 dirp = dentry->d_inode;
1777
1778 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001779 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 if (IS_ERR(rdentry))
1781 goto out_nfserr;
1782
1783 if (!rdentry->d_inode) {
1784 dput(rdentry);
1785 err = nfserr_noent;
1786 goto out;
1787 }
1788
1789 if (!type)
1790 type = rdentry->d_inode->i_mode & S_IFMT;
1791
Dave Hansen06227532008-02-15 14:37:34 -08001792 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1793 if (host_err)
1794 goto out_nfserr;
1795
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 if (type != S_IFDIR) { /* It's UNLINK */
1797#ifdef MSNFS
1798 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1799 (atomic_read(&rdentry->d_count) > 1)) {
Al Viro6264d692006-10-19 23:28:58 -07001800 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 } else
1802#endif
Al Viro6264d692006-10-19 23:28:58 -07001803 host_err = vfs_unlink(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 } else { /* It's RMDIR */
Al Viro6264d692006-10-19 23:28:58 -07001805 host_err = vfs_rmdir(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 }
1807
1808 dput(rdentry);
1809
Al Viro6264d692006-10-19 23:28:58 -07001810 if (host_err)
Dave Hansen06227532008-02-15 14:37:34 -08001811 goto out_drop;
Al Viro6264d692006-10-19 23:28:58 -07001812 if (EX_ISSYNC(fhp->fh_export))
1813 host_err = nfsd_sync_dir(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814
Dave Hansen06227532008-02-15 14:37:34 -08001815out_drop:
1816 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001818 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001819out:
1820 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821}
1822
1823/*
1824 * Read entries from a directory.
1825 * The NFSv3/4 verifier we ignore for now.
1826 */
Al Viro6264d692006-10-19 23:28:58 -07001827__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001829 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830{
Al Viro6264d692006-10-19 23:28:58 -07001831 __be32 err;
1832 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 struct file *file;
1834 loff_t offset = *offsetp;
1835
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001836 err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 if (err)
1838 goto out;
1839
1840 offset = vfs_llseek(file, offset, 0);
1841 if (offset < 0) {
1842 err = nfserrno((int)offset);
1843 goto out_close;
1844 }
1845
1846 /*
1847 * Read the directory entries. This silly loop is necessary because
1848 * readdir() is not guaranteed to fill up the entire buffer, but
1849 * may choose to do less.
1850 */
1851
1852 do {
1853 cdp->err = nfserr_eof; /* will be cleared on successful read */
NeilBrowna0ad13e2007-01-26 00:57:10 -08001854 host_err = vfs_readdir(file, func, cdp);
Al Viro6264d692006-10-19 23:28:58 -07001855 } while (host_err >=0 && cdp->err == nfs_ok);
1856 if (host_err)
1857 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 else
1859 err = cdp->err;
1860 *offsetp = vfs_llseek(file, 0, 1);
1861
1862 if (err == nfserr_eof || err == nfserr_toosmall)
1863 err = nfs_ok; /* can still be found in ->err */
1864out_close:
1865 nfsd_close(file);
1866out:
1867 return err;
1868}
1869
1870/*
1871 * Get file system stats
1872 * N.B. After this call fhp needs an fh_put
1873 */
Al Viro6264d692006-10-19 23:28:58 -07001874__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat)
1876{
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001877 __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
David Howells726c3342006-06-23 02:02:58 -07001878 if (!err && vfs_statfs(fhp->fh_dentry,stat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 err = nfserr_io;
1880 return err;
1881}
1882
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07001883static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07001884{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07001885 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07001886}
1887
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888/*
1889 * Check for a user's access permissions to this inode.
1890 */
Al Viro6264d692006-10-19 23:28:58 -07001891__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001892nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
1893 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894{
1895 struct inode *inode = dentry->d_inode;
1896 int err;
1897
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001898 if (acc == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 return 0;
1900#if 0
1901 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
1902 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001903 (acc & NFSD_MAY_READ)? " read" : "",
1904 (acc & NFSD_MAY_WRITE)? " write" : "",
1905 (acc & NFSD_MAY_EXEC)? " exec" : "",
1906 (acc & NFSD_MAY_SATTR)? " sattr" : "",
1907 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
1908 (acc & NFSD_MAY_LOCK)? " lock" : "",
1909 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 inode->i_mode,
1911 IS_IMMUTABLE(inode)? " immut" : "",
1912 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08001913 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 dprintk(" owner %d/%d user %d/%d\n",
1915 inode->i_uid, inode->i_gid, current->fsuid, current->fsgid);
1916#endif
1917
1918 /* Normally we reject any write/sattr etc access on a read-only file
1919 * system. But if it is IRIX doing check on write-access for a
1920 * device special file, we ignore rofs.
1921 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001922 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
1923 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08001924 if (exp_rdonly(rqstp, exp) ||
1925 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001927 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928 return nfserr_perm;
1929 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001930 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 return nfserr_perm;
1932
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001933 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 /* If we cannot rely on authentication in NLM requests,
1935 * just allow locks, otherwise require read permission, or
1936 * ownership
1937 */
1938 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
1939 return 0;
1940 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001941 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 }
1943 /*
1944 * The file owner always gets access permission for accesses that
1945 * would normally be checked at open time. This is to make
1946 * file access work even when the client has done a fchmod(fd, 0).
1947 *
1948 * However, `cp foo bar' should fail nevertheless when bar is
1949 * readonly. A sensible way to do this might be to reject all
1950 * attempts to truncate a read-only file, because a creat() call
1951 * always implies file truncation.
1952 * ... but this isn't really fair. A process may reasonably call
1953 * ftruncate on an open file descriptor on a file with perm 000.
1954 * We must trust the client to do permission checking - using "ACCESS"
1955 * with NFSv3.
1956 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001957 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 inode->i_uid == current->fsuid)
1959 return 0;
1960
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001961 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL);
1963
1964 /* Allow read access to binaries even when mode 111 */
1965 if (err == -EACCES && S_ISREG(inode->i_mode) &&
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001966 acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 err = permission(inode, MAY_EXEC, NULL);
1968
1969 return err? nfserrno(err) : 0;
1970}
1971
1972void
1973nfsd_racache_shutdown(void)
1974{
Greg Banksfce14562006-10-04 02:15:49 -07001975 if (!raparml)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 return;
1977 dprintk("nfsd: freeing readahead buffers.\n");
1978 kfree(raparml);
Greg Banksfce14562006-10-04 02:15:49 -07001979 raparml = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980}
1981/*
1982 * Initialize readahead param cache
1983 */
1984int
1985nfsd_racache_init(int cache_size)
1986{
1987 int i;
Greg Banksfce14562006-10-04 02:15:49 -07001988 int j = 0;
1989 int nperbucket;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Greg Banksfce14562006-10-04 02:15:49 -07001991
1992 if (raparml)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 return 0;
Greg Banksfce14562006-10-04 02:15:49 -07001994 if (cache_size < 2*RAPARM_HASH_SIZE)
1995 cache_size = 2*RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08001996 raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997
Yan Burman4b3bb062006-12-08 02:39:41 -08001998 if (!raparml) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 printk(KERN_WARNING
Yan Burman4b3bb062006-12-08 02:39:41 -08002000 "nfsd: Could not allocate memory read-ahead cache.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 return -ENOMEM;
2002 }
Yan Burman4b3bb062006-12-08 02:39:41 -08002003
2004 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
2005 for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) {
2006 raparm_hash[i].pb_head = NULL;
2007 spin_lock_init(&raparm_hash[i].pb_lock);
2008 }
J. Bruce Fieldsa280df32007-07-19 01:49:18 -07002009 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
Yan Burman4b3bb062006-12-08 02:39:41 -08002010 for (i = 0; i < cache_size - 1; i++) {
2011 if (i % nperbucket == 0)
2012 raparm_hash[j++].pb_head = raparml + i;
2013 if (i % nperbucket < nperbucket-1)
2014 raparml[i].p_next = raparml + i + 1;
2015 }
2016
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 nfsdstats.ra_size = cache_size;
2018 return 0;
2019}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002020
2021#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2022struct posix_acl *
2023nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2024{
2025 struct inode *inode = fhp->fh_dentry->d_inode;
2026 char *name;
2027 void *value = NULL;
2028 ssize_t size;
2029 struct posix_acl *acl;
2030
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002031 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002032 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002033
2034 switch (type) {
2035 case ACL_TYPE_ACCESS:
2036 name = POSIX_ACL_XATTR_ACCESS;
2037 break;
2038 case ACL_TYPE_DEFAULT:
2039 name = POSIX_ACL_XATTR_DEFAULT;
2040 break;
2041 default:
2042 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002043 }
2044
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002045 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2046 if (size < 0)
2047 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002048
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002049 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002050 kfree(value);
2051 return acl;
2052}
2053
2054int
2055nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2056{
2057 struct inode *inode = fhp->fh_dentry->d_inode;
2058 char *name;
2059 void *value = NULL;
2060 size_t size;
2061 int error;
2062
2063 if (!IS_POSIXACL(inode) || !inode->i_op ||
2064 !inode->i_op->setxattr || !inode->i_op->removexattr)
2065 return -EOPNOTSUPP;
2066 switch(type) {
2067 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002068 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002069 break;
2070 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002071 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002072 break;
2073 default:
2074 return -EOPNOTSUPP;
2075 }
2076
2077 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002078 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002079 value = kmalloc(size, GFP_KERNEL);
2080 if (!value)
2081 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002082 error = posix_acl_to_xattr(acl, value, size);
2083 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002084 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002085 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002086 } else
2087 size = 0;
2088
Dave Hansen18f335a2008-02-15 14:37:38 -08002089 error = mnt_want_write(fhp->fh_export->ex_path.mnt);
2090 if (error)
2091 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002092 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002093 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002094 else {
2095 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2096 error = 0;
2097 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002098 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002099 if (error == -ENODATA)
2100 error = 0;
2101 }
2102 }
Dave Hansen18f335a2008-02-15 14:37:38 -08002103 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002104
2105getout:
2106 kfree(value);
2107 return error;
2108}
2109#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */