blob: d0199189924cee7d5ea49aafd52dcc81fab21df2 [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;
104 struct vfsmount *mnt = mntget(exp->ex_mnt);
105 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,
135 const char *name, int len,
136 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. */
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
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);
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
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,
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;
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;
265 int accmode = MAY_SATTR;
266 int ftype = 0;
267 int imode;
Al Viro6264d692006-10-19 23:28:58 -0700268 __be32 err;
269 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 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);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800279 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 goto out;
281
282 dentry = fhp->fh_dentry;
283 inode = dentry->d_inode;
284
NeilBrown15b7a1b2005-11-07 01:00:23 -0800285 /* 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
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000292 /*
293 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 * 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)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000306 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000315 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 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 }
328
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000329 /*
330 * The size case is special.
331 * It changes the file as well as the attributes.
332 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 if (iap->ia_valid & ATTR_SIZE) {
334 if (iap->ia_size < inode->i_size) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700335 err = nfsd_permission(rqstp, fhp->fh_export, dentry, MAY_TRUNC|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
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);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700367 /* if changing uid/gid revoke setuid/setgid in mode */
368 if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) {
369 iap->ia_valid |= ATTR_KILL_PRIV;
370 iap->ia_mode &= ~S_ISUID;
371 }
372 if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid)
373 iap->ia_mode &= ~S_ISGID;
374 } else {
375 /*
376 * Revoke setuid/setgid bit on chown/chgrp
377 */
378 if ((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid)
379 iap->ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV;
380 if ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid)
381 iap->ia_valid |= ATTR_KILL_SGID;
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 */
466 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, MAY_SATTR);
467 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[] = {
567 { NFS3_ACCESS_READ, MAY_READ },
568 { NFS3_ACCESS_EXECUTE, MAY_EXEC },
569 { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_TRUNC },
570 { NFS3_ACCESS_EXTEND, MAY_WRITE },
571
572 { 0, 0 }
573};
574
575static struct accessmap nfs3_diraccess[] = {
576 { NFS3_ACCESS_READ, MAY_READ },
577 { NFS3_ACCESS_LOOKUP, MAY_EXEC },
578 { NFS3_ACCESS_MODIFY, MAY_EXEC|MAY_WRITE|MAY_TRUNC },
579 { NFS3_ACCESS_EXTEND, MAY_EXEC|MAY_WRITE },
580 { NFS3_ACCESS_DELETE, MAY_REMOVE },
581
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 */
593 { NFS3_ACCESS_READ, MAY_READ },
594 { NFS3_ACCESS_EXECUTE, MAY_EXEC },
595 { NFS3_ACCESS_MODIFY, MAY_WRITE|MAY_LOCAL_ACCESS },
596 { NFS3_ACCESS_EXTEND, MAY_WRITE|MAY_LOCAL_ACCESS },
597
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
610 error = fh_verify(rqstp, fhp, 0, MAY_NOP);
611 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 */
682 err = fh_verify(rqstp, fhp, type, access | MAY_OWNER_OVERRIDE);
683 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;
693 if (IS_APPEND(inode) && (access & MAY_WRITE))
694 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 */
Al Viro6264d692006-10-19 23:28:58 -0700710 host_err = break_lease(inode, O_NONBLOCK | ((access & MAY_WRITE) ? FMODE_WRITE : 0));
711 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
716 if (access & MAY_WRITE) {
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700717 if (access & MAY_READ)
718 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 }
724 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_mnt), flags);
725 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700726 host_err = PTR_ERR(*filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700728 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729out:
730 return err;
731}
732
733/*
734 * Close a file.
735 */
736void
737nfsd_close(struct file *filp)
738{
739 fput(filp);
740}
741
742/*
743 * Sync a file
744 * As this calls fsync (not fdatasync) there is no need for a write_inode
745 * after it.
746 */
David Shawa334de22006-01-06 00:19:58 -0800747static inline int nfsd_dosync(struct file *filp, struct dentry *dp,
Arjan van de Ven99ac48f2006-03-28 01:56:41 -0800748 const struct file_operations *fop)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749{
750 struct inode *inode = dp->d_inode;
751 int (*fsync) (struct file *, struct dentry *, int);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800752 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800754 err = filemap_fdatawrite(inode->i_mapping);
755 if (err == 0 && fop && (fsync = fop->fsync))
756 err = fsync(filp, dp, 0);
757 if (err == 0)
758 err = filemap_fdatawait(inode->i_mapping);
David Shawa334de22006-01-06 00:19:58 -0800759
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800760 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761}
762
763
David Shawa334de22006-01-06 00:19:58 -0800764static int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765nfsd_sync(struct file *filp)
766{
David Shawa334de22006-01-06 00:19:58 -0800767 int err;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800768 struct inode *inode = filp->f_path.dentry->d_inode;
769 dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800770 mutex_lock(&inode->i_mutex);
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800771 err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800772 mutex_unlock(&inode->i_mutex);
David Shawa334de22006-01-06 00:19:58 -0800773
774 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775}
776
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800777int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778nfsd_sync_dir(struct dentry *dp)
779{
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800780 return nfsd_dosync(NULL, dp, dp->d_inode->i_fop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
783/*
784 * Obtain the readahead parameters for the file
785 * specified by (dev, ino).
786 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788static inline struct raparms *
789nfsd_get_raparms(dev_t dev, ino_t ino)
790{
791 struct raparms *ra, **rap, **frap = NULL;
792 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700793 unsigned int hash;
794 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Greg Banksfce14562006-10-04 02:15:49 -0700796 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
797 rab = &raparm_hash[hash];
798
799 spin_lock(&rab->pb_lock);
800 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 if (ra->p_ino == ino && ra->p_dev == dev)
802 goto found;
803 depth++;
804 if (ra->p_count == 0)
805 frap = rap;
806 }
807 depth = nfsdstats.ra_size*11/10;
808 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700809 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 return NULL;
811 }
812 rap = frap;
813 ra = *frap;
814 ra->p_dev = dev;
815 ra->p_ino = ino;
816 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700817 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818found:
Greg Banksfce14562006-10-04 02:15:49 -0700819 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700821 ra->p_next = rab->pb_head;
822 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 }
824 ra->p_count++;
825 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700826 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 return ra;
828}
829
830/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200831 * Grab and keep cached pages associated with a file in the svc_rqst
832 * so that they can be passed to the network sendmsg/sendpage routines
833 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 */
835static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200836nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
837 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
Jens Axboecf8208d2007-06-12 21:22:14 +0200839 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700840 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200841 struct page *page = buf->page;
842 size_t size;
843 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844
Jens Axboecac36bb02007-06-14 13:10:48 +0200845 ret = buf->ops->confirm(pipe, buf);
Jens Axboecf8208d2007-06-12 21:22:14 +0200846 if (unlikely(ret))
847 return ret;
848
849 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850
851 if (rqstp->rq_res.page_len == 0) {
852 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700853 put_page(*pp);
854 *pp = page;
855 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200856 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700858 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800860 if (*pp)
861 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700862 *pp = page;
863 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700865 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 return size;
869}
870
Jens Axboecf8208d2007-06-12 21:22:14 +0200871static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
872 struct splice_desc *sd)
873{
874 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
875}
876
Dave Hansena8754be2007-10-16 23:31:15 -0700877static inline int svc_msnfs(struct svc_fh *ffhp)
878{
879#ifdef MSNFS
880 return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS);
881#else
882 return 0;
883#endif
884}
885
Al Viro6264d692006-10-19 23:28:58 -0700886static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
888 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
889{
890 struct inode *inode;
891 struct raparms *ra;
892 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700893 __be32 err;
894 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895
896 err = nfserr_perm;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800897 inode = file->f_path.dentry->d_inode;
Dave Hansena8754be2007-10-16 23:31:15 -0700898
899 if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901
902 /* Get readahead parameters */
903 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
904
905 if (ra && ra->p_set)
906 file->f_ra = ra->p_ra;
907
Jens Axboecf8208d2007-06-12 21:22:14 +0200908 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
909 struct splice_desc sd = {
910 .len = 0,
911 .total_len = *count,
912 .pos = offset,
913 .u.data = rqstp,
914 };
915
Jens Axboe4fbef202007-07-13 22:42:20 +0200916 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200917 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 } else {
919 oldfs = get_fs();
920 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700921 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 set_fs(oldfs);
923 }
924
925 /* Write back readahead params */
926 if (ra) {
Greg Banksfce14562006-10-04 02:15:49 -0700927 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
928 spin_lock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 ra->p_ra = file->f_ra;
930 ra->p_set = 1;
931 ra->p_count--;
Greg Banksfce14562006-10-04 02:15:49 -0700932 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 }
934
Al Viro6264d692006-10-19 23:28:58 -0700935 if (host_err >= 0) {
936 nfsdstats.io_read += host_err;
937 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 err = 0;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800939 fsnotify_access(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 } else
Al Viro6264d692006-10-19 23:28:58 -0700941 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942out:
943 return err;
944}
945
Neil Brown9f708e42006-01-06 00:19:59 -0800946static void kill_suid(struct dentry *dentry)
947{
948 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700949 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800950
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800951 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800952 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800953 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800954}
955
Al Viro6264d692006-10-19 23:28:58 -0700956static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
958 loff_t offset, struct kvec *vec, int vlen,
959 unsigned long cnt, int *stablep)
960{
961 struct svc_export *exp;
962 struct dentry *dentry;
963 struct inode *inode;
964 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700965 __be32 err = 0;
966 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 int stable = *stablep;
968
NeilBrown45bd3b32006-01-18 17:43:50 -0800969#ifdef MSNFS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 err = nfserr_perm;
971
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800973 (!lock_may_write(file->f_path.dentry->d_inode, offset, cnt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 goto out;
975#endif
976
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800977 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 inode = dentry->d_inode;
979 exp = fhp->fh_export;
980
981 /*
982 * Request sync writes if
983 * - the sync export option has been set, or
984 * - the client requested O_SYNC behavior (NFSv3 feature).
985 * - The file system doesn't support fsync().
986 * When gathered writes have been configured for this volume,
987 * flushing the data to disk is handled separately below.
988 */
989
990 if (file->f_op->fsync == 0) {/* COMMIT3 cannot work */
991 stable = 2;
992 *stablep = 2; /* FILE_SYNC */
993 }
994
995 if (!EX_ISSYNC(exp))
996 stable = 0;
997 if (stable && !EX_WGATHER(exp))
998 file->f_flags |= O_SYNC;
999
1000 /* Write the data. */
1001 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001002 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 set_fs(oldfs);
Al Viro6264d692006-10-19 23:28:58 -07001004 if (host_err >= 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 nfsdstats.io_write += cnt;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001006 fsnotify_modify(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 }
1008
1009 /* clear setuid/setgid flag after write */
Al Viro6264d692006-10-19 23:28:58 -07001010 if (host_err >= 0 && (inode->i_mode & (S_ISUID | S_ISGID)))
Neil Brown9f708e42006-01-06 00:19:59 -08001011 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012
Al Viro6264d692006-10-19 23:28:58 -07001013 if (host_err >= 0 && stable) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 static ino_t last_ino;
1015 static dev_t last_dev;
1016
1017 /*
1018 * Gathered writes: If another process is currently
1019 * writing to the file, there's a high chance
1020 * this is another nfsd (triggered by a bulk write
1021 * from a client's biod). Rather than syncing the
1022 * file with each write request, we sleep for 10 msec.
1023 *
1024 * I don't know if this roughly approximates
1025 * C. Juszak's idea of gathered writes, but it's a
1026 * nice and simple solution (IMHO), and it seems to
1027 * work:-)
1028 */
1029 if (EX_WGATHER(exp)) {
1030 if (atomic_read(&inode->i_writecount) > 1
1031 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001032 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 msleep(10);
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001034 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 }
1036
1037 if (inode->i_state & I_DIRTY) {
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -07001038 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Al Viro6264d692006-10-19 23:28:58 -07001039 host_err=nfsd_sync(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 }
1041#if 0
1042 wake_up(&inode->i_wait);
1043#endif
1044 }
1045 last_ino = inode->i_ino;
1046 last_dev = inode->i_sb->s_dev;
1047 }
1048
Al Viro6264d692006-10-19 23:28:58 -07001049 dprintk("nfsd: write complete host_err=%d\n", host_err);
1050 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 err = 0;
1052 else
Al Viro6264d692006-10-19 23:28:58 -07001053 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054out:
1055 return err;
1056}
1057
1058/*
1059 * Read data from a file. count must contain the requested read count
1060 * on entry. On return, *count contains the number of bytes actually read.
1061 * N.B. After this call fhp needs an fh_put
1062 */
Al Viro6264d692006-10-19 23:28:58 -07001063__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1065 loff_t offset, struct kvec *vec, int vlen,
1066 unsigned long *count)
1067{
Al Viro6264d692006-10-19 23:28:58 -07001068 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
1070 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001071 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 MAY_READ|MAY_OWNER_OVERRIDE);
1073 if (err)
1074 goto out;
1075 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1076 } else {
1077 err = nfsd_open(rqstp, fhp, S_IFREG, MAY_READ, &file);
1078 if (err)
1079 goto out;
1080 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1081 nfsd_close(file);
1082 }
1083out:
1084 return err;
1085}
1086
1087/*
1088 * Write data to a file.
1089 * The stable flag requests synchronous writes.
1090 * N.B. After this call fhp needs an fh_put
1091 */
Al Viro6264d692006-10-19 23:28:58 -07001092__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1094 loff_t offset, struct kvec *vec, int vlen, unsigned long cnt,
1095 int *stablep)
1096{
Al Viro6264d692006-10-19 23:28:58 -07001097 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
1099 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001100 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 MAY_WRITE|MAY_OWNER_OVERRIDE);
1102 if (err)
1103 goto out;
1104 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1105 stablep);
1106 } else {
1107 err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file);
1108 if (err)
1109 goto out;
1110
1111 if (cnt)
1112 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1113 cnt, stablep);
1114 nfsd_close(file);
1115 }
1116out:
1117 return err;
1118}
1119
1120#ifdef CONFIG_NFSD_V3
1121/*
1122 * Commit all pending writes to stable storage.
1123 * Strictly speaking, we could sync just the indicated file region here,
1124 * but there's currently no way we can ask the VFS to do so.
1125 *
1126 * Unfortunately we cannot lock the file to make sure we return full WCC
1127 * data to the client, as locking happens lower down in the filesystem.
1128 */
Al Viro6264d692006-10-19 23:28:58 -07001129__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1131 loff_t offset, unsigned long count)
1132{
1133 struct file *file;
Al Viro6264d692006-10-19 23:28:58 -07001134 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 if ((u64)count > ~(u64)offset)
1137 return nfserr_inval;
1138
1139 if ((err = nfsd_open(rqstp, fhp, S_IFREG, MAY_WRITE, &file)) != 0)
1140 return err;
1141 if (EX_ISSYNC(fhp->fh_export)) {
1142 if (file->f_op && file->f_op->fsync) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001143 err = nfserrno(nfsd_sync(file));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 } else {
1145 err = nfserr_notsupp;
1146 }
1147 }
1148
1149 nfsd_close(file);
1150 return err;
1151}
1152#endif /* CONFIG_NFSD_V3 */
1153
1154/*
1155 * Create a file (regular, directory, device, fifo); UNIX sockets
1156 * not yet implemented.
1157 * If the response fh has been verified, the parent directory should
1158 * already be locked. Note that the parent directory is left locked.
1159 *
1160 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1161 */
Al Viro6264d692006-10-19 23:28:58 -07001162__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1164 char *fname, int flen, struct iattr *iap,
1165 int type, dev_t rdev, struct svc_fh *resfhp)
1166{
1167 struct dentry *dentry, *dchild = NULL;
1168 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001169 __be32 err;
1170 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171
1172 err = nfserr_perm;
1173 if (!flen)
1174 goto out;
1175 err = nfserr_exist;
1176 if (isdotent(fname, flen))
1177 goto out;
1178
1179 err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE);
1180 if (err)
1181 goto out;
1182
1183 dentry = fhp->fh_dentry;
1184 dirp = dentry->d_inode;
1185
1186 err = nfserr_notdir;
1187 if(!dirp->i_op || !dirp->i_op->lookup)
1188 goto out;
1189 /*
1190 * Check whether the response file handle has been verified yet.
1191 * If it has, the parent directory should already be locked.
1192 */
1193 if (!resfhp->fh_dentry) {
1194 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001195 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001197 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 if (IS_ERR(dchild))
1199 goto out_nfserr;
1200 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1201 if (err)
1202 goto out;
1203 } else {
1204 /* called from nfsd_proc_create */
1205 dchild = dget(resfhp->fh_dentry);
1206 if (!fhp->fh_locked) {
1207 /* not actually possible */
1208 printk(KERN_ERR
1209 "nfsd_create: parent %s/%s not locked!\n",
1210 dentry->d_parent->d_name.name,
1211 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001212 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 goto out;
1214 }
1215 }
1216 /*
1217 * Make sure the child dentry is still negative ...
1218 */
1219 err = nfserr_exist;
1220 if (dchild->d_inode) {
1221 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1222 dentry->d_name.name, dchild->d_name.name);
1223 goto out;
1224 }
1225
1226 if (!(iap->ia_valid & ATTR_MODE))
1227 iap->ia_mode = 0;
1228 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1229
1230 /*
1231 * Get the dir op function pointer.
1232 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001233 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 switch (type) {
1235 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001236 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 break;
1238 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001239 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 break;
1241 case S_IFCHR:
1242 case S_IFBLK:
1243 case S_IFIFO:
1244 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001245 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 break;
1247 default:
1248 printk("nfsd: bad file type %o in nfsd_create\n", type);
Al Viro6264d692006-10-19 23:28:58 -07001249 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 }
Al Viro6264d692006-10-19 23:28:58 -07001251 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 goto out_nfserr;
1253
1254 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001255 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 write_inode_now(dchild->d_inode, 1);
1257 }
1258
1259
1260 /* Set file attributes. Mode has already been set and
1261 * setting uid/gid works only for root. Irix appears to
1262 * send along the gid when it tries to implement setgid
1263 * directories via NFS.
1264 */
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001265 if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) {
Al Viro6264d692006-10-19 23:28:58 -07001266 __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001267 if (err2)
1268 err = err2;
1269 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 /*
1271 * Update the file handle to get the new inode info.
1272 */
1273 if (!err)
1274 err = fh_update(resfhp);
1275out:
1276 if (dchild && !IS_ERR(dchild))
1277 dput(dchild);
1278 return err;
1279
1280out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001281 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 goto out;
1283}
1284
1285#ifdef CONFIG_NFSD_V3
1286/*
1287 * NFSv3 version of nfsd_create
1288 */
Al Viro6264d692006-10-19 23:28:58 -07001289__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
1291 char *fname, int flen, struct iattr *iap,
1292 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001293 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294{
1295 struct dentry *dentry, *dchild = NULL;
1296 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001297 __be32 err;
1298 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
1301 err = nfserr_perm;
1302 if (!flen)
1303 goto out;
1304 err = nfserr_exist;
1305 if (isdotent(fname, flen))
1306 goto out;
1307 if (!(iap->ia_valid & ATTR_MODE))
1308 iap->ia_mode = 0;
1309 err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE);
1310 if (err)
1311 goto out;
1312
1313 dentry = fhp->fh_dentry;
1314 dirp = dentry->d_inode;
1315
1316 /* Get all the sanity checks out of the way before
1317 * we lock the parent. */
1318 err = nfserr_notdir;
1319 if(!dirp->i_op || !dirp->i_op->lookup)
1320 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001321 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322
1323 /*
1324 * Compose the response file handle.
1325 */
1326 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001327 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 if (IS_ERR(dchild))
1329 goto out_nfserr;
1330
1331 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1332 if (err)
1333 goto out;
1334
1335 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001336 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001337 * the high bit set, so just clear the high bits. If this is
1338 * ever changed to use different attrs for storing the
1339 * verifier, then do_open_lookup() will also need to be fixed
1340 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 */
1342 v_mtime = verifier[0]&0x7fffffff;
1343 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 }
1345
1346 if (dchild->d_inode) {
1347 err = 0;
1348
1349 switch (createmode) {
1350 case NFS3_CREATE_UNCHECKED:
1351 if (! S_ISREG(dchild->d_inode->i_mode))
1352 err = nfserr_exist;
1353 else if (truncp) {
1354 /* in nfsv4, we need to treat this case a little
1355 * differently. we don't want to truncate the
1356 * file now; this would be wrong if the OPEN
1357 * fails for some other reason. furthermore,
1358 * if the size is nonzero, we should ignore it
1359 * according to spec!
1360 */
1361 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1362 }
1363 else {
1364 iap->ia_valid &= ATTR_SIZE;
1365 goto set_attr;
1366 }
1367 break;
1368 case NFS3_CREATE_EXCLUSIVE:
1369 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1370 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 && dchild->d_inode->i_size == 0 )
1372 break;
1373 /* fallthru */
1374 case NFS3_CREATE_GUARDED:
1375 err = nfserr_exist;
1376 }
1377 goto out;
1378 }
1379
Al Viro6264d692006-10-19 23:28:58 -07001380 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
1381 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 goto out_nfserr;
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001383 if (created)
1384 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385
1386 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001387 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 /* setattr will sync the child (or not) */
1389 }
1390
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001392 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001394 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 /* XXX someone who knows this better please fix it for nsec */
1396 iap->ia_mtime.tv_sec = v_mtime;
1397 iap->ia_atime.tv_sec = v_atime;
1398 iap->ia_mtime.tv_nsec = 0;
1399 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 }
1401
1402 /* Set file attributes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 * Irix appears to send along the gid when it tries to
1404 * implement setgid directories via NFS. Clear out all that cruft.
1405 */
1406 set_attr:
Peter Staubachc3978522007-01-26 00:57:00 -08001407 if ((iap->ia_valid &= ~(ATTR_UID|ATTR_GID|ATTR_MODE)) != 0) {
Al Viro6264d692006-10-19 23:28:58 -07001408 __be32 err2 = nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001409 if (err2)
NeilBrown45bd3b32006-01-18 17:43:50 -08001410 err = err2;
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001411 }
1412
1413 /*
1414 * Update the filehandle to get the new inode info.
1415 */
1416 if (!err)
1417 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 out:
1420 fh_unlock(fhp);
1421 if (dchild && !IS_ERR(dchild))
1422 dput(dchild);
1423 return err;
1424
1425 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001426 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 goto out;
1428}
1429#endif /* CONFIG_NFSD_V3 */
1430
1431/*
1432 * Read a symlink. On entry, *lenp must contain the maximum path length that
1433 * fits into the buffer. On return, it contains the true length.
1434 * N.B. After this call fhp needs an fh_put
1435 */
Al Viro6264d692006-10-19 23:28:58 -07001436__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1438{
1439 struct dentry *dentry;
1440 struct inode *inode;
1441 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001442 __be32 err;
1443 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
1445 err = fh_verify(rqstp, fhp, S_IFLNK, MAY_NOP);
1446 if (err)
1447 goto out;
1448
1449 dentry = fhp->fh_dentry;
1450 inode = dentry->d_inode;
1451
1452 err = nfserr_inval;
1453 if (!inode->i_op || !inode->i_op->readlink)
1454 goto out;
1455
1456 touch_atime(fhp->fh_export->ex_mnt, dentry);
1457 /* N.B. Why does this call need a get_fs()??
1458 * Remove the set_fs and watch the fireworks:-) --okir
1459 */
1460
1461 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001462 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 set_fs(oldfs);
1464
Al Viro6264d692006-10-19 23:28:58 -07001465 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001467 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 err = 0;
1469out:
1470 return err;
1471
1472out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001473 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 goto out;
1475}
1476
1477/*
1478 * Create a symlink and look up its inode
1479 * N.B. After this call _both_ fhp and resfhp need an fh_put
1480 */
Al Viro6264d692006-10-19 23:28:58 -07001481__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1483 char *fname, int flen,
1484 char *path, int plen,
1485 struct svc_fh *resfhp,
1486 struct iattr *iap)
1487{
1488 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001489 __be32 err, cerr;
1490 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 umode_t mode;
1492
1493 err = nfserr_noent;
1494 if (!flen || !plen)
1495 goto out;
1496 err = nfserr_exist;
1497 if (isdotent(fname, flen))
1498 goto out;
1499
1500 err = fh_verify(rqstp, fhp, S_IFDIR, MAY_CREATE);
1501 if (err)
1502 goto out;
1503 fh_lock(fhp);
1504 dentry = fhp->fh_dentry;
1505 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001506 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 if (IS_ERR(dnew))
1508 goto out_nfserr;
1509
1510 mode = S_IALLUGO;
1511 /* Only the MODE ATTRibute is even vaguely meaningful */
1512 if (iap && (iap->ia_valid & ATTR_MODE))
1513 mode = iap->ia_mode & S_IALLUGO;
1514
1515 if (unlikely(path[plen] != 0)) {
1516 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1517 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001518 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 else {
1520 strncpy(path_alloced, path, plen);
1521 path_alloced[plen] = 0;
Al Viro6264d692006-10-19 23:28:58 -07001522 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 kfree(path_alloced);
1524 }
1525 } else
Al Viro6264d692006-10-19 23:28:58 -07001526 host_err = vfs_symlink(dentry->d_inode, dnew, path, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Al Viro6264d692006-10-19 23:28:58 -07001528 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 if (EX_ISSYNC(fhp->fh_export))
Al Viro6264d692006-10-19 23:28:58 -07001530 host_err = nfsd_sync_dir(dentry);
1531 }
1532 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 fh_unlock(fhp);
1534
1535 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1536 dput(dnew);
1537 if (err==0) err = cerr;
1538out:
1539 return err;
1540
1541out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001542 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 goto out;
1544}
1545
1546/*
1547 * Create a hardlink
1548 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1549 */
Al Viro6264d692006-10-19 23:28:58 -07001550__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1552 char *name, int len, struct svc_fh *tfhp)
1553{
1554 struct dentry *ddir, *dnew, *dold;
1555 struct inode *dirp, *dest;
Al Viro6264d692006-10-19 23:28:58 -07001556 __be32 err;
1557 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
1559 err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_CREATE);
1560 if (err)
1561 goto out;
1562 err = fh_verify(rqstp, tfhp, -S_IFDIR, MAY_NOP);
1563 if (err)
1564 goto out;
1565
1566 err = nfserr_perm;
1567 if (!len)
1568 goto out;
1569 err = nfserr_exist;
1570 if (isdotent(name, len))
1571 goto out;
1572
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001573 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 ddir = ffhp->fh_dentry;
1575 dirp = ddir->d_inode;
1576
1577 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001578 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 if (IS_ERR(dnew))
1580 goto out_nfserr;
1581
1582 dold = tfhp->fh_dentry;
1583 dest = dold->d_inode;
1584
Al Viro6264d692006-10-19 23:28:58 -07001585 host_err = vfs_link(dold, dirp, dnew);
1586 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 if (EX_ISSYNC(ffhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001588 err = nfserrno(nfsd_sync_dir(ddir));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 write_inode_now(dest, 1);
1590 }
Al Viro6264d692006-10-19 23:28:58 -07001591 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 } else {
Al Viro6264d692006-10-19 23:28:58 -07001593 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 err = nfserr_acces;
1595 else
Al Viro6264d692006-10-19 23:28:58 -07001596 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 }
1598
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001600out_unlock:
1601 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602out:
1603 return err;
1604
1605out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001606 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001607 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608}
1609
1610/*
1611 * Rename a file
1612 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1613 */
Al Viro6264d692006-10-19 23:28:58 -07001614__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1616 struct svc_fh *tfhp, char *tname, int tlen)
1617{
1618 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1619 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001620 __be32 err;
1621 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623 err = fh_verify(rqstp, ffhp, S_IFDIR, MAY_REMOVE);
1624 if (err)
1625 goto out;
1626 err = fh_verify(rqstp, tfhp, S_IFDIR, MAY_CREATE);
1627 if (err)
1628 goto out;
1629
1630 fdentry = ffhp->fh_dentry;
1631 fdir = fdentry->d_inode;
1632
1633 tdentry = tfhp->fh_dentry;
1634 tdir = tdentry->d_inode;
1635
1636 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001637 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 goto out;
1639
1640 err = nfserr_perm;
1641 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1642 goto out;
1643
1644 /* cannot use fh_lock as we need deadlock protective ordering
1645 * so do it by hand */
1646 trap = lock_rename(tdentry, fdentry);
1647 ffhp->fh_locked = tfhp->fh_locked = 1;
1648 fill_pre_wcc(ffhp);
1649 fill_pre_wcc(tfhp);
1650
1651 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001652 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 if (IS_ERR(odentry))
1654 goto out_nfserr;
1655
Al Viro6264d692006-10-19 23:28:58 -07001656 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 if (!odentry->d_inode)
1658 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001659 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 if (odentry == trap)
1661 goto out_dput_old;
1662
1663 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001664 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 if (IS_ERR(ndentry))
1666 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001667 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 if (ndentry == trap)
1669 goto out_dput_new;
1670
1671#ifdef MSNFS
1672 if ((ffhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1673 ((atomic_read(&odentry->d_count) > 1)
1674 || (atomic_read(&ndentry->d_count) > 1))) {
Al Viro6264d692006-10-19 23:28:58 -07001675 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 } else
1677#endif
Al Viro6264d692006-10-19 23:28:58 -07001678 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
1679 if (!host_err && EX_ISSYNC(tfhp->fh_export)) {
1680 host_err = nfsd_sync_dir(tdentry);
1681 if (!host_err)
1682 host_err = nfsd_sync_dir(fdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 }
1684
1685 out_dput_new:
1686 dput(ndentry);
1687 out_dput_old:
1688 dput(odentry);
1689 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001690 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
1692 /* we cannot reply on fh_unlock on the two filehandles,
1693 * as that would do the wrong thing if the two directories
1694 * were the same, so again we do it by hand
1695 */
1696 fill_post_wcc(ffhp);
1697 fill_post_wcc(tfhp);
1698 unlock_rename(tdentry, fdentry);
1699 ffhp->fh_locked = tfhp->fh_locked = 0;
1700
1701out:
1702 return err;
1703}
1704
1705/*
1706 * Unlink a file or directory
1707 * N.B. After this call fhp needs an fh_put
1708 */
Al Viro6264d692006-10-19 23:28:58 -07001709__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1711 char *fname, int flen)
1712{
1713 struct dentry *dentry, *rdentry;
1714 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001715 __be32 err;
1716 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717
1718 err = nfserr_acces;
1719 if (!flen || isdotent(fname, flen))
1720 goto out;
1721 err = fh_verify(rqstp, fhp, S_IFDIR, MAY_REMOVE);
1722 if (err)
1723 goto out;
1724
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001725 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 dentry = fhp->fh_dentry;
1727 dirp = dentry->d_inode;
1728
1729 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001730 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 if (IS_ERR(rdentry))
1732 goto out_nfserr;
1733
1734 if (!rdentry->d_inode) {
1735 dput(rdentry);
1736 err = nfserr_noent;
1737 goto out;
1738 }
1739
1740 if (!type)
1741 type = rdentry->d_inode->i_mode & S_IFMT;
1742
1743 if (type != S_IFDIR) { /* It's UNLINK */
1744#ifdef MSNFS
1745 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1746 (atomic_read(&rdentry->d_count) > 1)) {
Al Viro6264d692006-10-19 23:28:58 -07001747 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 } else
1749#endif
Al Viro6264d692006-10-19 23:28:58 -07001750 host_err = vfs_unlink(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 } else { /* It's RMDIR */
Al Viro6264d692006-10-19 23:28:58 -07001752 host_err = vfs_rmdir(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 }
1754
1755 dput(rdentry);
1756
Al Viro6264d692006-10-19 23:28:58 -07001757 if (host_err)
1758 goto out_nfserr;
1759 if (EX_ISSYNC(fhp->fh_export))
1760 host_err = nfsd_sync_dir(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
1762out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001763 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001764out:
1765 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766}
1767
1768/*
1769 * Read entries from a directory.
1770 * The NFSv3/4 verifier we ignore for now.
1771 */
Al Viro6264d692006-10-19 23:28:58 -07001772__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001774 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775{
Al Viro6264d692006-10-19 23:28:58 -07001776 __be32 err;
1777 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 struct file *file;
1779 loff_t offset = *offsetp;
1780
1781 err = nfsd_open(rqstp, fhp, S_IFDIR, MAY_READ, &file);
1782 if (err)
1783 goto out;
1784
1785 offset = vfs_llseek(file, offset, 0);
1786 if (offset < 0) {
1787 err = nfserrno((int)offset);
1788 goto out_close;
1789 }
1790
1791 /*
1792 * Read the directory entries. This silly loop is necessary because
1793 * readdir() is not guaranteed to fill up the entire buffer, but
1794 * may choose to do less.
1795 */
1796
1797 do {
1798 cdp->err = nfserr_eof; /* will be cleared on successful read */
NeilBrowna0ad13e2007-01-26 00:57:10 -08001799 host_err = vfs_readdir(file, func, cdp);
Al Viro6264d692006-10-19 23:28:58 -07001800 } while (host_err >=0 && cdp->err == nfs_ok);
1801 if (host_err)
1802 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 else
1804 err = cdp->err;
1805 *offsetp = vfs_llseek(file, 0, 1);
1806
1807 if (err == nfserr_eof || err == nfserr_toosmall)
1808 err = nfs_ok; /* can still be found in ->err */
1809out_close:
1810 nfsd_close(file);
1811out:
1812 return err;
1813}
1814
1815/*
1816 * Get file system stats
1817 * N.B. After this call fhp needs an fh_put
1818 */
Al Viro6264d692006-10-19 23:28:58 -07001819__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat)
1821{
Al Viro6264d692006-10-19 23:28:58 -07001822 __be32 err = fh_verify(rqstp, fhp, 0, MAY_NOP);
David Howells726c3342006-06-23 02:02:58 -07001823 if (!err && vfs_statfs(fhp->fh_dentry,stat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 err = nfserr_io;
1825 return err;
1826}
1827
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07001828static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07001829{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07001830 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07001831}
1832
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833/*
1834 * Check for a user's access permissions to this inode.
1835 */
Al Viro6264d692006-10-19 23:28:58 -07001836__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001837nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
1838 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839{
1840 struct inode *inode = dentry->d_inode;
1841 int err;
1842
1843 if (acc == MAY_NOP)
1844 return 0;
1845#if 0
1846 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
1847 acc,
1848 (acc & MAY_READ)? " read" : "",
1849 (acc & MAY_WRITE)? " write" : "",
1850 (acc & MAY_EXEC)? " exec" : "",
1851 (acc & MAY_SATTR)? " sattr" : "",
1852 (acc & MAY_TRUNC)? " trunc" : "",
1853 (acc & MAY_LOCK)? " lock" : "",
1854 (acc & MAY_OWNER_OVERRIDE)? " owneroverride" : "",
1855 inode->i_mode,
1856 IS_IMMUTABLE(inode)? " immut" : "",
1857 IS_APPEND(inode)? " append" : "",
1858 IS_RDONLY(inode)? " ro" : "");
1859 dprintk(" owner %d/%d user %d/%d\n",
1860 inode->i_uid, inode->i_gid, current->fsuid, current->fsgid);
1861#endif
1862
1863 /* Normally we reject any write/sattr etc access on a read-only file
1864 * system. But if it is IRIX doing check on write-access for a
1865 * device special file, we ignore rofs.
1866 */
1867 if (!(acc & MAY_LOCAL_ACCESS))
1868 if (acc & (MAY_WRITE | MAY_SATTR | MAY_TRUNC)) {
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07001869 if (exp_rdonly(rqstp, exp) || IS_RDONLY(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 return nfserr_rofs;
1871 if (/* (acc & MAY_WRITE) && */ IS_IMMUTABLE(inode))
1872 return nfserr_perm;
1873 }
1874 if ((acc & MAY_TRUNC) && IS_APPEND(inode))
1875 return nfserr_perm;
1876
1877 if (acc & MAY_LOCK) {
1878 /* If we cannot rely on authentication in NLM requests,
1879 * just allow locks, otherwise require read permission, or
1880 * ownership
1881 */
1882 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
1883 return 0;
1884 else
1885 acc = MAY_READ | MAY_OWNER_OVERRIDE;
1886 }
1887 /*
1888 * The file owner always gets access permission for accesses that
1889 * would normally be checked at open time. This is to make
1890 * file access work even when the client has done a fchmod(fd, 0).
1891 *
1892 * However, `cp foo bar' should fail nevertheless when bar is
1893 * readonly. A sensible way to do this might be to reject all
1894 * attempts to truncate a read-only file, because a creat() call
1895 * always implies file truncation.
1896 * ... but this isn't really fair. A process may reasonably call
1897 * ftruncate on an open file descriptor on a file with perm 000.
1898 * We must trust the client to do permission checking - using "ACCESS"
1899 * with NFSv3.
1900 */
1901 if ((acc & MAY_OWNER_OVERRIDE) &&
1902 inode->i_uid == current->fsuid)
1903 return 0;
1904
1905 err = permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC), NULL);
1906
1907 /* Allow read access to binaries even when mode 111 */
1908 if (err == -EACCES && S_ISREG(inode->i_mode) &&
1909 acc == (MAY_READ | MAY_OWNER_OVERRIDE))
1910 err = permission(inode, MAY_EXEC, NULL);
1911
1912 return err? nfserrno(err) : 0;
1913}
1914
1915void
1916nfsd_racache_shutdown(void)
1917{
Greg Banksfce14562006-10-04 02:15:49 -07001918 if (!raparml)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 return;
1920 dprintk("nfsd: freeing readahead buffers.\n");
1921 kfree(raparml);
Greg Banksfce14562006-10-04 02:15:49 -07001922 raparml = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923}
1924/*
1925 * Initialize readahead param cache
1926 */
1927int
1928nfsd_racache_init(int cache_size)
1929{
1930 int i;
Greg Banksfce14562006-10-04 02:15:49 -07001931 int j = 0;
1932 int nperbucket;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
Greg Banksfce14562006-10-04 02:15:49 -07001934
1935 if (raparml)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 return 0;
Greg Banksfce14562006-10-04 02:15:49 -07001937 if (cache_size < 2*RAPARM_HASH_SIZE)
1938 cache_size = 2*RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08001939 raparml = kcalloc(cache_size, sizeof(struct raparms), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
Yan Burman4b3bb062006-12-08 02:39:41 -08001941 if (!raparml) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 printk(KERN_WARNING
Yan Burman4b3bb062006-12-08 02:39:41 -08001943 "nfsd: Could not allocate memory read-ahead cache.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 return -ENOMEM;
1945 }
Yan Burman4b3bb062006-12-08 02:39:41 -08001946
1947 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
1948 for (i = 0 ; i < RAPARM_HASH_SIZE ; i++) {
1949 raparm_hash[i].pb_head = NULL;
1950 spin_lock_init(&raparm_hash[i].pb_lock);
1951 }
J. Bruce Fieldsa280df32007-07-19 01:49:18 -07001952 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
Yan Burman4b3bb062006-12-08 02:39:41 -08001953 for (i = 0; i < cache_size - 1; i++) {
1954 if (i % nperbucket == 0)
1955 raparm_hash[j++].pb_head = raparml + i;
1956 if (i % nperbucket < nperbucket-1)
1957 raparml[i].p_next = raparml + i + 1;
1958 }
1959
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 nfsdstats.ra_size = cache_size;
1961 return 0;
1962}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001963
1964#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
1965struct posix_acl *
1966nfsd_get_posix_acl(struct svc_fh *fhp, int type)
1967{
1968 struct inode *inode = fhp->fh_dentry->d_inode;
1969 char *name;
1970 void *value = NULL;
1971 ssize_t size;
1972 struct posix_acl *acl;
1973
Christoph Hellwig5be196e2006-01-09 20:51:55 -08001974 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001975 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08001976
1977 switch (type) {
1978 case ACL_TYPE_ACCESS:
1979 name = POSIX_ACL_XATTR_ACCESS;
1980 break;
1981 case ACL_TYPE_DEFAULT:
1982 name = POSIX_ACL_XATTR_DEFAULT;
1983 break;
1984 default:
1985 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001986 }
1987
Christoph Hellwig5be196e2006-01-09 20:51:55 -08001988 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
1989 if (size < 0)
1990 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001991
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001992 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00001993 kfree(value);
1994 return acl;
1995}
1996
1997int
1998nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
1999{
2000 struct inode *inode = fhp->fh_dentry->d_inode;
2001 char *name;
2002 void *value = NULL;
2003 size_t size;
2004 int error;
2005
2006 if (!IS_POSIXACL(inode) || !inode->i_op ||
2007 !inode->i_op->setxattr || !inode->i_op->removexattr)
2008 return -EOPNOTSUPP;
2009 switch(type) {
2010 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002011 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002012 break;
2013 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002014 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002015 break;
2016 default:
2017 return -EOPNOTSUPP;
2018 }
2019
2020 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002021 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002022 value = kmalloc(size, GFP_KERNEL);
2023 if (!value)
2024 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002025 error = posix_acl_to_xattr(acl, value, size);
2026 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002027 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002028 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002029 } else
2030 size = 0;
2031
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002032 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002033 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002034 else {
2035 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2036 error = 0;
2037 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002038 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002039 if (error == -ENODATA)
2040 error = 0;
2041 }
2042 }
2043
2044getout:
2045 kfree(value);
2046 return error;
2047}
2048#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */