blob: 6dd5f1970e01dedc3da245d8d26d87236dfadaf4 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#define MSNFS /* HACK HACK */
2/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 * File operations used by nfsd. Some of these have been ripped from
4 * other parts of the kernel because they weren't exported, others
5 * are partial duplicates with added or changed functionality.
6 *
7 * Note that several functions dget() the dentry upon which they want
8 * to act, most notably those that create directory entries. Response
9 * dentry's are dput()'d if necessary in the release callback.
10 * So if you notice code paths that apparently fail to dput() the
11 * dentry, don't worry--they have been taken care of.
12 *
13 * Copyright (C) 1995-1999 Olaf Kirch <okir@monad.swb.de>
14 * Zerocpy NFS support (C) 2002 Hirokazu Takahashi <taka@valinux.co.jp>
15 */
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/fs.h>
18#include <linux/file.h>
Jens Axboed6b29d72007-06-04 09:59:47 +020019#include <linux/splice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/fcntl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/namei.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/delay.h>
Robert Love0eeca282005-07-12 17:06:03 -040023#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/posix_acl_xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/xattr.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020026#include <linux/jhash.h>
27#include <linux/ima.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090028#include <linux/slab.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020029#include <asm/uaccess.h>
Ben Myersf5019122010-02-17 14:05:11 -060030#include <linux/exportfs.h>
31#include <linux/writeback.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020032
33#ifdef CONFIG_NFSD_V3
34#include "xdr3.h"
35#endif /* CONFIG_NFSD_V3 */
36
Christoph Hellwig5be196e2006-01-09 20:51:55 -080037#ifdef CONFIG_NFSD_V4
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <linux/nfs4_acl.h>
39#include <linux/nfsd_idmap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#endif /* CONFIG_NFSD_V4 */
41
Boaz Harrosh9a74af22009-12-03 20:30:56 +020042#include "nfsd.h"
43#include "vfs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45#define NFSDDBG_FACILITY NFSDDBG_FILEOP
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47
Linus Torvalds1da177e2005-04-16 15:20:36 -070048/*
49 * This is a cache of readahead params that help us choose the proper
50 * readahead strategy. Initially, we set all readahead parameters to 0
51 * and let the VFS handle things.
52 * If you increase the number of cached files very much, you'll need to
53 * add a hash table here.
54 */
55struct raparms {
56 struct raparms *p_next;
57 unsigned int p_count;
58 ino_t p_ino;
59 dev_t p_dev;
60 int p_set;
61 struct file_ra_state p_ra;
Greg Banksfce14562006-10-04 02:15:49 -070062 unsigned int p_hindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -070063};
64
Greg Banksfce14562006-10-04 02:15:49 -070065struct raparm_hbucket {
66 struct raparms *pb_head;
67 spinlock_t pb_lock;
68} ____cacheline_aligned_in_smp;
69
Greg Banksfce14562006-10-04 02:15:49 -070070#define RAPARM_HASH_BITS 4
71#define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS)
72#define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1)
73static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
75/*
76 * Called from nfsd_lookup and encode_dirent. Check if we have crossed
77 * a mount point.
J.Bruce Fieldse0bb89e2006-12-13 00:35:25 -080078 * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged,
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 * or nfs_ok having possibly changed *dpp and *expp
80 */
81int
82nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp,
83 struct svc_export **expp)
84{
85 struct svc_export *exp = *expp, *exp2 = NULL;
86 struct dentry *dentry = *dpp;
Al Viro91c9fa82009-04-18 02:42:05 -040087 struct path path = {.mnt = mntget(exp->ex_path.mnt),
88 .dentry = dget(dentry)};
Al Viro6264d692006-10-19 23:28:58 -070089 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Al Viro9393bd02009-04-18 13:58:15 -040091 while (d_mountpoint(path.dentry) && follow_down(&path))
Al Viro91c9fa82009-04-18 02:42:05 -040092 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Al Viro91c9fa82009-04-18 02:42:05 -040094 exp2 = rqst_exp_get_by_name(rqstp, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095 if (IS_ERR(exp2)) {
J. Bruce Fields3b6cee7bc2009-10-25 21:18:19 -040096 err = PTR_ERR(exp2);
97 /*
98 * We normally allow NFS clients to continue
99 * "underneath" a mountpoint that is not exported.
100 * The exception is V4ROOT, where no traversal is ever
101 * allowed without an explicit export of the new
102 * directory.
103 */
104 if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT))
105 err = 0;
Al Viro91c9fa82009-04-18 02:42:05 -0400106 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 goto out;
108 }
Steve Dickson3c394dd2009-09-09 15:02:40 -0400109 if (nfsd_v4client(rqstp) ||
110 (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 /* successfully crossed mount point */
Al Viro1644ccc2009-04-18 02:32:31 -0400112 /*
Al Viro91c9fa82009-04-18 02:42:05 -0400113 * This is subtle: path.dentry is *not* on path.mnt
114 * at this point. The only reason we are safe is that
115 * original mnt is pinned down by exp, so we should
116 * put path *before* putting exp
Al Viro1644ccc2009-04-18 02:32:31 -0400117 */
Al Viro91c9fa82009-04-18 02:42:05 -0400118 *dpp = path.dentry;
119 path.dentry = dentry;
Al Viro1644ccc2009-04-18 02:32:31 -0400120 *expp = exp2;
Al Viro91c9fa82009-04-18 02:42:05 -0400121 exp2 = exp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 }
Al Viro91c9fa82009-04-18 02:42:05 -0400123 path_put(&path);
124 exp_put(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125out:
126 return err;
127}
128
J. Bruce Fields289ede42009-09-26 20:32:24 -0400129static void follow_to_parent(struct path *path)
130{
131 struct dentry *dp;
132
133 while (path->dentry == path->mnt->mnt_root && follow_up(path))
134 ;
135 dp = dget_parent(path->dentry);
136 dput(path->dentry);
137 path->dentry = dp;
138}
139
140static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp)
141{
142 struct svc_export *exp2;
143 struct path path = {.mnt = mntget((*exp)->ex_path.mnt),
144 .dentry = dget(dparent)};
145
146 follow_to_parent(&path);
147
148 exp2 = rqst_exp_parent(rqstp, &path);
149 if (PTR_ERR(exp2) == -ENOENT) {
150 *dentryp = dget(dparent);
151 } else if (IS_ERR(exp2)) {
152 path_put(&path);
153 return PTR_ERR(exp2);
154 } else {
155 *dentryp = dget(path.dentry);
156 exp_put(*exp);
157 *exp = exp2;
158 }
159 path_put(&path);
160 return 0;
161}
162
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400163/*
164 * For nfsd purposes, we treat V4ROOT exports as though there was an
165 * export at *every* directory.
166 */
J. Bruce Fields3227fa42009-10-25 21:43:01 -0400167int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp)
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400168{
169 if (d_mountpoint(dentry))
170 return 1;
171 if (!(exp->ex_flags & NFSEXP_V4ROOT))
172 return 0;
173 return dentry->d_inode != NULL;
174}
175
Al Viro6264d692006-10-19 23:28:58 -0700176__be32
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700177nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400178 const char *name, unsigned int len,
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700179 struct svc_export **exp_ret, struct dentry **dentry_ret)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180{
181 struct svc_export *exp;
182 struct dentry *dparent;
183 struct dentry *dentry;
Al Viro6264d692006-10-19 23:28:58 -0700184 __be32 err;
185 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187 dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name);
188
189 /* Obtain dentry and export. */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200190 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 if (err)
192 return err;
193
194 dparent = fhp->fh_dentry;
195 exp = fhp->fh_export;
196 exp_get(exp);
197
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 /* Lookup the name, but don't follow links */
199 if (isdotent(name, len)) {
200 if (len==1)
201 dentry = dget(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800202 else if (dparent != exp->ex_path.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 dentry = dget_parent(dparent);
J. Bruce Fieldsfed83812009-09-26 16:53:01 -0400204 else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 dentry = dget(dparent); /* .. == . just like at / */
206 else {
207 /* checking mountpoint crossing is very different when stepping up */
J. Bruce Fields289ede42009-09-26 20:32:24 -0400208 host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry);
209 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 }
212 } else {
213 fh_lock(fhp);
214 dentry = lookup_one_len(name, dparent, len);
Al Viro6264d692006-10-19 23:28:58 -0700215 host_err = PTR_ERR(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 if (IS_ERR(dentry))
217 goto out_nfserr;
218 /*
219 * check if we have crossed a mount point ...
220 */
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400221 if (nfsd_mountpoint(dentry, exp)) {
Al Viro6264d692006-10-19 23:28:58 -0700222 if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 dput(dentry);
224 goto out_nfserr;
225 }
226 }
227 }
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700228 *dentry_ret = dentry;
229 *exp_ret = exp;
230 return 0;
231
232out_nfserr:
233 exp_put(exp);
234 return nfserrno(host_err);
235}
236
237/*
238 * Look up one component of a pathname.
239 * N.B. After this call _both_ fhp and resfh need an fh_put
240 *
241 * If the lookup would cross a mountpoint, and the mounted filesystem
242 * is exported to the client with NFSEXP_NOHIDE, then the lookup is
243 * accepted as it stands and the mounted directory is
244 * returned. Otherwise the covered directory is returned.
245 * NOTE: this mountpoint crossing is not supported properly by all
246 * clients and is explicitly disallowed for NFSv3
247 * NeilBrown <neilb@cse.unsw.edu.au>
248 */
249__be32
250nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400251 unsigned int len, struct svc_fh *resfh)
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700252{
253 struct svc_export *exp;
254 struct dentry *dentry;
255 __be32 err;
256
257 err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry);
258 if (err)
259 return err;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700260 err = check_nfsd_access(exp, rqstp);
261 if (err)
262 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 /*
264 * Note: we compose the file handle now, but as the
265 * dentry may be negative, it may need to be updated.
266 */
267 err = fh_compose(resfh, exp, dentry, fhp);
268 if (!err && !dentry->d_inode)
269 err = nfserr_noent;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700270out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 exp_put(exp);
273 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274}
275
Ben Myersf5019122010-02-17 14:05:11 -0600276/*
277 * Commit metadata changes to stable storage.
278 */
279static int
280commit_metadata(struct svc_fh *fhp)
281{
282 struct inode *inode = fhp->fh_dentry->d_inode;
283 const struct export_operations *export_ops = inode->i_sb->s_export_op;
284 int error = 0;
285
286 if (!EX_ISSYNC(fhp->fh_export))
287 return 0;
288
289 if (export_ops->commit_metadata) {
290 error = export_ops->commit_metadata(inode);
291 } else {
292 struct writeback_control wbc = {
293 .sync_mode = WB_SYNC_ALL,
294 .nr_to_write = 0, /* metadata only */
295 };
296
297 error = sync_inode(inode, &wbc);
298 }
299
300 return error;
301}
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700302
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303/*
304 * Set various file attributes.
305 * N.B. After this call fhp needs an fh_put
306 */
Al Viro6264d692006-10-19 23:28:58 -0700307__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
309 int check_guard, time_t guardtime)
310{
311 struct dentry *dentry;
312 struct inode *inode;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200313 int accmode = NFSD_MAY_SATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 int ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700315 __be32 err;
316 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 int size_change = 0;
318
319 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200320 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 if (iap->ia_valid & ATTR_SIZE)
322 ftype = S_IFREG;
323
324 /* Get inode */
325 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800326 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 goto out;
328
329 dentry = fhp->fh_dentry;
330 inode = dentry->d_inode;
331
NeilBrown15b7a1b2005-11-07 01:00:23 -0800332 /* Ignore any mode updates on symlinks */
333 if (S_ISLNK(inode->i_mode))
334 iap->ia_valid &= ~ATTR_MODE;
335
336 if (!iap->ia_valid)
337 goto out;
338
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000339 /*
340 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 * which only requires access, and "set-[ac]time-to-X" which
342 * requires ownership.
343 * So if it looks like it might be "set both to the same time which
344 * is close to now", and if inode_change_ok fails, then we
345 * convert to "set to now" instead of "set to explicit time"
346 *
347 * We only call inode_change_ok as the last test as technically
348 * it is not an interface that we should be using. It is only
349 * valid if the filesystem does not define it's own i_op->setattr.
350 */
351#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
352#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000353 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
354 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
355 /*
356 * Looks probable.
357 *
358 * Now just make sure time is in the right ballpark.
359 * Solaris, at least, doesn't seem to care what the time
360 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000362 time_t delta = iap->ia_atime.tv_sec - get_seconds();
363 if (delta < 0)
364 delta = -delta;
365 if (delta < MAX_TOUCH_TIME_ERROR &&
366 inode_change_ok(inode, iap) != 0) {
367 /*
368 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
369 * This will cause notify_change to set these times
370 * to "now"
371 */
372 iap->ia_valid &= ~BOTH_TIME_SET;
373 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 }
375
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000376 /*
377 * The size case is special.
378 * It changes the file as well as the attributes.
379 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 if (iap->ia_valid & ATTR_SIZE) {
381 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200382 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
383 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 if (err)
385 goto out;
386 }
387
388 /*
389 * If we are changing the size of the file, then
390 * we need to break all leases.
391 */
Al Viro8737c932009-12-24 06:47:55 -0500392 host_err = break_lease(inode, O_WRONLY | O_NONBLOCK);
Al Viro6264d692006-10-19 23:28:58 -0700393 if (host_err == -EWOULDBLOCK)
394 host_err = -ETIMEDOUT;
395 if (host_err) /* ENOMEM or EWOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 goto out_nfserr;
397
Al Viro6264d692006-10-19 23:28:58 -0700398 host_err = get_write_access(inode);
399 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 goto out_nfserr;
401
402 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700403 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
404 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 put_write_access(inode);
406 goto out_nfserr;
407 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 }
409
Jeff Laytonca456252008-04-16 16:28:47 -0400410 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 if (iap->ia_valid & ATTR_MODE) {
412 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400413 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400414 }
415
416 /* Revoke setuid/setgid on chown */
Sachin S. Prabhu0953e622009-02-23 16:22:03 +0000417 if (!S_ISDIR(inode->i_mode) &&
418 (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
419 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) {
Jeff Laytonca456252008-04-16 16:28:47 -0400420 iap->ia_valid |= ATTR_KILL_PRIV;
421 if (iap->ia_valid & ATTR_MODE) {
422 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700423 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400424 if (iap->ia_mode & S_IXGRP)
425 iap->ia_mode &= ~S_ISGID;
426 } else {
427 /* set ATTR_KILL_* bits and let VFS handle it */
428 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700429 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 }
431
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 /* Change the attributes. */
433
434 iap->ia_valid |= ATTR_CTIME;
435
436 err = nfserr_notsync;
437 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
438 fh_lock(fhp);
Al Viro6264d692006-10-19 23:28:58 -0700439 host_err = notify_change(dentry, iap);
440 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 fh_unlock(fhp);
442 }
443 if (size_change)
444 put_write_access(inode);
445 if (!err)
446 if (EX_ISSYNC(fhp->fh_export))
447 write_inode_now(inode, 1);
448out:
449 return err;
450
451out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700452 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 goto out;
454}
455
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800456#if defined(CONFIG_NFSD_V2_ACL) || \
457 defined(CONFIG_NFSD_V3_ACL) || \
458 defined(CONFIG_NFSD_V4)
459static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
460{
461 ssize_t buflen;
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530462 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800464 buflen = vfs_getxattr(dentry, key, NULL, 0);
465 if (buflen <= 0)
466 return buflen;
467
468 *buf = kmalloc(buflen, GFP_KERNEL);
469 if (!*buf)
470 return -ENOMEM;
471
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530472 ret = vfs_getxattr(dentry, key, *buf, buflen);
473 if (ret < 0)
474 kfree(*buf);
475 return ret;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800476}
477#endif
478
479#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480static int
481set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
482{
483 int len;
484 size_t buflen;
485 char *buf = NULL;
486 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
488 buflen = posix_acl_xattr_size(pacl->a_count);
489 buf = kmalloc(buflen, GFP_KERNEL);
490 error = -ENOMEM;
491 if (buf == NULL)
492 goto out;
493
494 len = posix_acl_to_xattr(pacl, buf, buflen);
495 if (len < 0) {
496 error = len;
497 goto out;
498 }
499
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800500 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501out:
502 kfree(buf);
503 return error;
504}
505
Al Viro6264d692006-10-19 23:28:58 -0700506__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
508 struct nfs4_acl *acl)
509{
Al Viro6264d692006-10-19 23:28:58 -0700510 __be32 error;
511 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 struct dentry *dentry;
513 struct inode *inode;
514 struct posix_acl *pacl = NULL, *dpacl = NULL;
515 unsigned int flags = 0;
516
517 /* Get inode */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200518 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700520 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
522 dentry = fhp->fh_dentry;
523 inode = dentry->d_inode;
524 if (S_ISDIR(inode->i_mode))
525 flags = NFS4_ACL_DIR;
526
Al Viro6264d692006-10-19 23:28:58 -0700527 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
528 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700529 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700530 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 goto out_nfserr;
532
Al Viro6264d692006-10-19 23:28:58 -0700533 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
534 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700535 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700537 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700538 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700540out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 posix_acl_release(pacl);
542 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800544 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700545 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800546 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700547 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548}
549
550static struct posix_acl *
551_get_posix_acl(struct dentry *dentry, char *key)
552{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800553 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800555 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800557 buflen = nfsd_getxattr(dentry, key, &buf);
558 if (!buflen)
559 buflen = -ENODATA;
560 if (buflen <= 0)
561 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
563 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 kfree(buf);
565 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566}
567
568int
569nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
570{
571 struct inode *inode = dentry->d_inode;
572 int error = 0;
573 struct posix_acl *pacl = NULL, *dpacl = NULL;
574 unsigned int flags = 0;
575
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700576 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
578 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
579 if (IS_ERR(pacl)) {
580 error = PTR_ERR(pacl);
581 pacl = NULL;
582 goto out;
583 }
584
585 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700586 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
588 dpacl = NULL;
589 else if (IS_ERR(dpacl)) {
590 error = PTR_ERR(dpacl);
591 dpacl = NULL;
592 goto out;
593 }
594 flags = NFS4_ACL_DIR;
595 }
596
597 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
598 if (IS_ERR(*acl)) {
599 error = PTR_ERR(*acl);
600 *acl = NULL;
601 }
602 out:
603 posix_acl_release(pacl);
604 posix_acl_release(dpacl);
605 return error;
606}
607
608#endif /* defined(CONFIG_NFS_V4) */
609
610#ifdef CONFIG_NFSD_V3
611/*
612 * Check server access rights to a file system object
613 */
614struct accessmap {
615 u32 access;
616 int how;
617};
618static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200619 { NFS3_ACCESS_READ, NFSD_MAY_READ },
620 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
621 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
622 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
624 { 0, 0 }
625};
626
627static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200628 { NFS3_ACCESS_READ, NFSD_MAY_READ },
629 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
630 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
631 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
632 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
634 { 0, 0 }
635};
636
637static struct accessmap nfs3_anyaccess[] = {
638 /* Some clients - Solaris 2.6 at least, make an access call
639 * to the server to check for access for things like /dev/null
640 * (which really, the server doesn't care about). So
641 * We provide simple access checking for them, looking
642 * mainly at mode bits, and we make sure to ignore read-only
643 * filesystem checks
644 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200645 { NFS3_ACCESS_READ, NFSD_MAY_READ },
646 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
647 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
648 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
650 { 0, 0 }
651};
652
Al Viro6264d692006-10-19 23:28:58 -0700653__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
655{
656 struct accessmap *map;
657 struct svc_export *export;
658 struct dentry *dentry;
659 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700660 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200662 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 if (error)
664 goto out;
665
666 export = fhp->fh_export;
667 dentry = fhp->fh_dentry;
668
669 if (S_ISREG(dentry->d_inode->i_mode))
670 map = nfs3_regaccess;
671 else if (S_ISDIR(dentry->d_inode->i_mode))
672 map = nfs3_diraccess;
673 else
674 map = nfs3_anyaccess;
675
676
677 query = *access;
678 for (; map->access; map++) {
679 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700680 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681
682 sresult |= map->access;
683
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700684 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 switch (err2) {
686 case nfs_ok:
687 result |= map->access;
688 break;
689
690 /* the following error codes just mean the access was not allowed,
691 * rather than an error occurred */
692 case nfserr_rofs:
693 case nfserr_acces:
694 case nfserr_perm:
695 /* simply don't "or" in the access bit. */
696 break;
697 default:
698 error = err2;
699 goto out;
700 }
701 }
702 }
703 *access = result;
704 if (supported)
705 *supported = sresult;
706
707 out:
708 return error;
709}
710#endif /* CONFIG_NFSD_V3 */
711
712
713
714/*
715 * Open an existing file or directory.
716 * The access argument indicates the type of open (read/write/lock)
717 * N.B. After this call fhp needs an fh_put
718 */
Al Viro6264d692006-10-19 23:28:58 -0700719__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
721 int access, struct file **filp)
722{
723 struct dentry *dentry;
724 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700725 int flags = O_RDONLY|O_LARGEFILE;
726 __be32 err;
727 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
David Howellse0e81732009-09-02 09:13:40 +0100729 validate_process_creds();
730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 /*
732 * If we get here, then the client has already done an "open",
733 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
734 * in case a chmod has now revoked permission.
735 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200736 err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 if (err)
738 goto out;
739
740 dentry = fhp->fh_dentry;
741 inode = dentry->d_inode;
742
743 /* Disallow write access to files with the append-only bit set
744 * or any access when mandatory locking enabled
745 */
746 err = nfserr_perm;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200747 if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400749 /*
750 * We must ignore files (but only files) which might have mandatory
751 * locks on them because there is no way to know if the accesser has
752 * the lock.
753 */
754 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 goto out;
756
757 if (!inode->i_fop)
758 goto out;
759
760 /*
761 * Check to see if there are any leases on this file.
762 * This may block while leases are broken.
763 */
Al Viro8737c932009-12-24 06:47:55 -0500764 host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? O_WRONLY : 0));
Al Viro6264d692006-10-19 23:28:58 -0700765 if (host_err == -EWOULDBLOCK)
766 host_err = -ETIMEDOUT;
767 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 goto out_nfserr;
769
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200770 if (access & NFSD_MAY_WRITE) {
771 if (access & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700772 flags = O_RDWR|O_LARGEFILE;
773 else
774 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
Jan Blunck54775492008-02-14 19:38:39 -0800776 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
David Howells033a6662009-07-02 14:35:32 +0100777 flags, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700779 host_err = PTR_ERR(*filp);
Chuck Ebbertaeaa5cc2010-02-15 18:07:39 -0500780 else
781 host_err = ima_file_check(*filp, access);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700783 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784out:
David Howellse0e81732009-09-02 09:13:40 +0100785 validate_process_creds();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 return err;
787}
788
789/*
790 * Close a file.
791 */
792void
793nfsd_close(struct file *filp)
794{
795 fput(filp);
796}
797
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500798/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 * Obtain the readahead parameters for the file
800 * specified by (dev, ino).
801 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
803static inline struct raparms *
804nfsd_get_raparms(dev_t dev, ino_t ino)
805{
806 struct raparms *ra, **rap, **frap = NULL;
807 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700808 unsigned int hash;
809 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
Greg Banksfce14562006-10-04 02:15:49 -0700811 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
812 rab = &raparm_hash[hash];
813
814 spin_lock(&rab->pb_lock);
815 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 if (ra->p_ino == ino && ra->p_dev == dev)
817 goto found;
818 depth++;
819 if (ra->p_count == 0)
820 frap = rap;
821 }
822 depth = nfsdstats.ra_size*11/10;
823 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700824 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 return NULL;
826 }
827 rap = frap;
828 ra = *frap;
829 ra->p_dev = dev;
830 ra->p_ino = ino;
831 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700832 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833found:
Greg Banksfce14562006-10-04 02:15:49 -0700834 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700836 ra->p_next = rab->pb_head;
837 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 }
839 ra->p_count++;
840 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700841 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 return ra;
843}
844
845/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200846 * Grab and keep cached pages associated with a file in the svc_rqst
847 * so that they can be passed to the network sendmsg/sendpage routines
848 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 */
850static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200851nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
852 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853{
Jens Axboecf8208d2007-06-12 21:22:14 +0200854 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700855 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200856 struct page *page = buf->page;
857 size_t size;
858 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
Jens Axboecac36bb02007-06-14 13:10:48 +0200860 ret = buf->ops->confirm(pipe, buf);
Jens Axboecf8208d2007-06-12 21:22:14 +0200861 if (unlikely(ret))
862 return ret;
863
864 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866 if (rqstp->rq_res.page_len == 0) {
867 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700868 put_page(*pp);
869 *pp = page;
870 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200871 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700873 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800875 if (*pp)
876 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700877 *pp = page;
878 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700880 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 return size;
884}
885
Jens Axboecf8208d2007-06-12 21:22:14 +0200886static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
887 struct splice_desc *sd)
888{
889 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
890}
891
Dave Hansena8754be2007-10-16 23:31:15 -0700892static inline int svc_msnfs(struct svc_fh *ffhp)
893{
894#ifdef MSNFS
895 return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS);
896#else
897 return 0;
898#endif
899}
900
Al Viro6264d692006-10-19 23:28:58 -0700901static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
903 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
904{
905 struct inode *inode;
906 struct raparms *ra;
907 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700908 __be32 err;
909 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910
911 err = nfserr_perm;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800912 inode = file->f_path.dentry->d_inode;
Dave Hansena8754be2007-10-16 23:31:15 -0700913
914 if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
917 /* Get readahead parameters */
918 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
919
920 if (ra && ra->p_set)
921 file->f_ra = ra->p_ra;
922
Jens Axboecf8208d2007-06-12 21:22:14 +0200923 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
924 struct splice_desc sd = {
925 .len = 0,
926 .total_len = *count,
927 .pos = offset,
928 .u.data = rqstp,
929 };
930
Jens Axboe4fbef202007-07-13 22:42:20 +0200931 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200932 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 } else {
934 oldfs = get_fs();
935 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700936 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 set_fs(oldfs);
938 }
939
940 /* Write back readahead params */
941 if (ra) {
Greg Banksfce14562006-10-04 02:15:49 -0700942 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
943 spin_lock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 ra->p_ra = file->f_ra;
945 ra->p_set = 1;
946 ra->p_count--;
Greg Banksfce14562006-10-04 02:15:49 -0700947 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 }
949
Al Viro6264d692006-10-19 23:28:58 -0700950 if (host_err >= 0) {
951 nfsdstats.io_read += host_err;
952 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 err = 0;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800954 fsnotify_access(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 } else
Al Viro6264d692006-10-19 23:28:58 -0700956 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957out:
958 return err;
959}
960
Neil Brown9f708e42006-01-06 00:19:59 -0800961static void kill_suid(struct dentry *dentry)
962{
963 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700964 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800965
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800966 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800967 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800968 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800969}
970
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700971/*
972 * Gathered writes: If another process is currently writing to the file,
973 * there's a high chance this is another nfsd (triggered by a bulk write
974 * from a client's biod). Rather than syncing the file with each write
975 * request, we sleep for 10 msec.
976 *
977 * I don't know if this roughly approximates C. Juszak's idea of
978 * gathered writes, but it's a nice and simple solution (IMHO), and it
979 * seems to work:-)
980 *
981 * Note: we do this only in the NFSv2 case, since v3 and higher have a
982 * better tool (separate unstable writes and commits) for solving this
983 * problem.
984 */
985static int wait_for_concurrent_writes(struct file *file)
986{
987 struct inode *inode = file->f_path.dentry->d_inode;
988 static ino_t last_ino;
989 static dev_t last_dev;
990 int err = 0;
991
992 if (atomic_read(&inode->i_writecount) > 1
993 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
994 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
995 msleep(10);
996 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
997 }
998
999 if (inode->i_state & I_DIRTY) {
1000 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Christoph Hellwig6a68f892009-12-25 17:45:13 +01001001 err = vfs_fsync(file, file->f_path.dentry, 0);
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001002 }
1003 last_ino = inode->i_ino;
1004 last_dev = inode->i_sb->s_dev;
1005 return err;
1006}
1007
Al Viro6264d692006-10-19 23:28:58 -07001008static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1010 loff_t offset, struct kvec *vec, int vlen,
David Shaw31dec252009-03-05 20:16:14 -05001011 unsigned long *cnt, int *stablep)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
1013 struct svc_export *exp;
1014 struct dentry *dentry;
1015 struct inode *inode;
1016 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001017 __be32 err = 0;
1018 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 int stable = *stablep;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001020 int use_wgather;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
NeilBrown45bd3b32006-01-18 17:43:50 -08001022#ifdef MSNFS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 err = nfserr_perm;
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
David Shaw31dec252009-03-05 20:16:14 -05001026 (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 goto out;
1028#endif
1029
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001030 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 inode = dentry->d_inode;
1032 exp = fhp->fh_export;
1033
1034 /*
1035 * Request sync writes if
1036 * - the sync export option has been set, or
1037 * - the client requested O_SYNC behavior (NFSv3 feature).
1038 * - The file system doesn't support fsync().
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001039 * When NFSv2 gathered writes have been configured for this volume,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 * flushing the data to disk is handled separately below.
1041 */
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001042 use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
Harvey Harrison3ba15142008-02-20 12:49:02 -08001044 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 stable = 2;
1046 *stablep = 2; /* FILE_SYNC */
1047 }
1048
1049 if (!EX_ISSYNC(exp))
1050 stable = 0;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001051 if (stable && !use_wgather) {
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001052 spin_lock(&file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 file->f_flags |= O_SYNC;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001054 spin_unlock(&file->f_lock);
1055 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
1057 /* Write the data. */
1058 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001059 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 set_fs(oldfs);
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001061 if (host_err < 0)
1062 goto out_nfserr;
1063 *cnt = host_err;
1064 nfsdstats.io_write += host_err;
1065 fsnotify_modify(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067 /* clear setuid/setgid flag after write */
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001068 if (inode->i_mode & (S_ISUID | S_ISGID))
Neil Brown9f708e42006-01-06 00:19:59 -08001069 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001071 if (stable && use_wgather)
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001072 host_err = wait_for_concurrent_writes(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001074out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001075 dprintk("nfsd: write complete host_err=%d\n", host_err);
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001076 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 err = 0;
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001078 else
Al Viro6264d692006-10-19 23:28:58 -07001079 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080out:
1081 return err;
1082}
1083
1084/*
1085 * Read data from a file. count must contain the requested read count
1086 * on entry. On return, *count contains the number of bytes actually read.
1087 * N.B. After this call fhp needs an fh_put
1088 */
Al Viro6264d692006-10-19 23:28:58 -07001089__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1091 loff_t offset, struct kvec *vec, int vlen,
1092 unsigned long *count)
1093{
Al Viro6264d692006-10-19 23:28:58 -07001094 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
1096 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001097 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001098 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 if (err)
1100 goto out;
1101 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1102 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001103 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 if (err)
1105 goto out;
1106 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1107 nfsd_close(file);
1108 }
1109out:
1110 return err;
1111}
1112
1113/*
1114 * Write data to a file.
1115 * The stable flag requests synchronous writes.
1116 * N.B. After this call fhp needs an fh_put
1117 */
Al Viro6264d692006-10-19 23:28:58 -07001118__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
David Shaw31dec252009-03-05 20:16:14 -05001120 loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 int *stablep)
1122{
Al Viro6264d692006-10-19 23:28:58 -07001123 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
1125 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001126 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001127 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 if (err)
1129 goto out;
1130 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1131 stablep);
1132 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001133 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 if (err)
1135 goto out;
1136
1137 if (cnt)
1138 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1139 cnt, stablep);
1140 nfsd_close(file);
1141 }
1142out:
1143 return err;
1144}
1145
1146#ifdef CONFIG_NFSD_V3
1147/*
1148 * Commit all pending writes to stable storage.
Trond Myklebustaa696a62010-01-29 16:44:25 -05001149 *
1150 * Note: we only guarantee that data that lies within the range specified
1151 * by the 'offset' and 'count' parameters will be synced.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 *
1153 * Unfortunately we cannot lock the file to make sure we return full WCC
1154 * data to the client, as locking happens lower down in the filesystem.
1155 */
Al Viro6264d692006-10-19 23:28:58 -07001156__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1158 loff_t offset, unsigned long count)
1159{
1160 struct file *file;
Trond Myklebustaa696a62010-01-29 16:44:25 -05001161 loff_t end = LLONG_MAX;
1162 __be32 err = nfserr_inval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
Trond Myklebustaa696a62010-01-29 16:44:25 -05001164 if (offset < 0)
1165 goto out;
1166 if (count != 0) {
1167 end = offset + (loff_t)count - 1;
1168 if (end < offset)
1169 goto out;
1170 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001172 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
1173 if (err)
Trond Myklebustaa696a62010-01-29 16:44:25 -05001174 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 if (EX_ISSYNC(fhp->fh_export)) {
Trond Myklebustaa696a62010-01-29 16:44:25 -05001176 int err2 = vfs_fsync_range(file, file->f_path.dentry,
1177 offset, end, 0);
1178
1179 if (err2 != -EINVAL)
1180 err = nfserrno(err2);
1181 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 err = nfserr_notsupp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 }
1184
1185 nfsd_close(file);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001186out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 return err;
1188}
1189#endif /* CONFIG_NFSD_V3 */
1190
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001191static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001192nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1193 struct iattr *iap)
1194{
1195 /*
1196 * Mode has already been set earlier in create:
1197 */
1198 iap->ia_valid &= ~ATTR_MODE;
1199 /*
1200 * Setting uid/gid works only for root. Irix appears to
1201 * send along the gid on create when it tries to implement
1202 * setgid directories via NFS:
1203 */
David Howells5cc0a842008-11-14 10:38:58 +11001204 if (current_fsuid() != 0)
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001205 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1206 if (iap->ia_valid)
1207 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1208 return 0;
1209}
1210
wengang wang4ac35c22009-02-10 11:27:51 +08001211/* HPUX client sometimes creates a file in mode 000, and sets size to 0.
1212 * setting size to 0 may fail for some specific file systems by the permission
1213 * checking which requires WRITE permission but the mode is 000.
1214 * we ignore the resizing(to 0) on the just new created file, since the size is
1215 * 0 after file created.
1216 *
1217 * call this only after vfs_create() is called.
1218 * */
1219static void
1220nfsd_check_ignore_resizing(struct iattr *iap)
1221{
1222 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
1223 iap->ia_valid &= ~ATTR_SIZE;
1224}
1225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226/*
1227 * Create a file (regular, directory, device, fifo); UNIX sockets
1228 * not yet implemented.
1229 * If the response fh has been verified, the parent directory should
1230 * already be locked. Note that the parent directory is left locked.
1231 *
1232 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1233 */
Al Viro6264d692006-10-19 23:28:58 -07001234__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1236 char *fname, int flen, struct iattr *iap,
1237 int type, dev_t rdev, struct svc_fh *resfhp)
1238{
1239 struct dentry *dentry, *dchild = NULL;
1240 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001241 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001242 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001243 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244
1245 err = nfserr_perm;
1246 if (!flen)
1247 goto out;
1248 err = nfserr_exist;
1249 if (isdotent(fname, flen))
1250 goto out;
1251
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001252 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 if (err)
1254 goto out;
1255
1256 dentry = fhp->fh_dentry;
1257 dirp = dentry->d_inode;
1258
1259 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001260 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 goto out;
1262 /*
1263 * Check whether the response file handle has been verified yet.
1264 * If it has, the parent directory should already be locked.
1265 */
1266 if (!resfhp->fh_dentry) {
1267 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001268 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001270 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 if (IS_ERR(dchild))
1272 goto out_nfserr;
1273 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1274 if (err)
1275 goto out;
1276 } else {
1277 /* called from nfsd_proc_create */
1278 dchild = dget(resfhp->fh_dentry);
1279 if (!fhp->fh_locked) {
1280 /* not actually possible */
1281 printk(KERN_ERR
1282 "nfsd_create: parent %s/%s not locked!\n",
1283 dentry->d_parent->d_name.name,
1284 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001285 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 goto out;
1287 }
1288 }
1289 /*
1290 * Make sure the child dentry is still negative ...
1291 */
1292 err = nfserr_exist;
1293 if (dchild->d_inode) {
1294 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1295 dentry->d_name.name, dchild->d_name.name);
1296 goto out;
1297 }
1298
1299 if (!(iap->ia_valid & ATTR_MODE))
1300 iap->ia_mode = 0;
1301 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1302
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001303 err = nfserr_inval;
1304 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1305 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1306 type);
1307 goto out;
1308 }
1309
1310 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1311 if (host_err)
1312 goto out_nfserr;
1313
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 /*
1315 * Get the dir op function pointer.
1316 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001317 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 switch (type) {
1319 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001320 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
wengang wang4ac35c22009-02-10 11:27:51 +08001321 if (!host_err)
1322 nfsd_check_ignore_resizing(iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 break;
1324 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001325 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 break;
1327 case S_IFCHR:
1328 case S_IFBLK:
1329 case S_IFIFO:
1330 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001331 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001333 }
1334 if (host_err < 0) {
1335 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1336 goto out_nfserr;
1337 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Ben Myersf5019122010-02-17 14:05:11 -06001339 err = nfsd_create_setattr(rqstp, resfhp, iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Ben Myersf5019122010-02-17 14:05:11 -06001341 /*
1342 * nfsd_setattr already committed the child. Transactional filesystems
1343 * had a chance to commit changes for both parent and child
1344 * simultaneously making the following commit_metadata a noop.
1345 */
1346 err2 = nfserrno(commit_metadata(fhp));
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001347 if (err2)
1348 err = err2;
Dave Hansen463c3192008-02-15 14:37:57 -08001349 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 /*
1351 * Update the file handle to get the new inode info.
1352 */
1353 if (!err)
1354 err = fh_update(resfhp);
1355out:
1356 if (dchild && !IS_ERR(dchild))
1357 dput(dchild);
1358 return err;
1359
1360out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001361 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 goto out;
1363}
1364
1365#ifdef CONFIG_NFSD_V3
1366/*
1367 * NFSv3 version of nfsd_create
1368 */
Al Viro6264d692006-10-19 23:28:58 -07001369__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
1371 char *fname, int flen, struct iattr *iap,
1372 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001373 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374{
1375 struct dentry *dentry, *dchild = NULL;
1376 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001377 __be32 err;
1378 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
1381 err = nfserr_perm;
1382 if (!flen)
1383 goto out;
1384 err = nfserr_exist;
1385 if (isdotent(fname, flen))
1386 goto out;
1387 if (!(iap->ia_valid & ATTR_MODE))
1388 iap->ia_mode = 0;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001389 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (err)
1391 goto out;
1392
1393 dentry = fhp->fh_dentry;
1394 dirp = dentry->d_inode;
1395
1396 /* Get all the sanity checks out of the way before
1397 * we lock the parent. */
1398 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001399 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001401 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
1403 /*
1404 * Compose the response file handle.
1405 */
1406 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001407 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 if (IS_ERR(dchild))
1409 goto out_nfserr;
1410
1411 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1412 if (err)
1413 goto out;
1414
1415 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001416 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001417 * the high bit set, so just clear the high bits. If this is
1418 * ever changed to use different attrs for storing the
1419 * verifier, then do_open_lookup() will also need to be fixed
1420 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 */
1422 v_mtime = verifier[0]&0x7fffffff;
1423 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 }
1425
Dave Hansen463c3192008-02-15 14:37:57 -08001426 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1427 if (host_err)
1428 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 if (dchild->d_inode) {
1430 err = 0;
1431
1432 switch (createmode) {
1433 case NFS3_CREATE_UNCHECKED:
1434 if (! S_ISREG(dchild->d_inode->i_mode))
1435 err = nfserr_exist;
1436 else if (truncp) {
1437 /* in nfsv4, we need to treat this case a little
1438 * differently. we don't want to truncate the
1439 * file now; this would be wrong if the OPEN
1440 * fails for some other reason. furthermore,
1441 * if the size is nonzero, we should ignore it
1442 * according to spec!
1443 */
1444 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1445 }
1446 else {
1447 iap->ia_valid &= ATTR_SIZE;
1448 goto set_attr;
1449 }
1450 break;
1451 case NFS3_CREATE_EXCLUSIVE:
1452 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1453 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 && dchild->d_inode->i_size == 0 )
1455 break;
1456 /* fallthru */
1457 case NFS3_CREATE_GUARDED:
1458 err = nfserr_exist;
1459 }
Dave Hansen463c3192008-02-15 14:37:57 -08001460 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 goto out;
1462 }
1463
Al Viro6264d692006-10-19 23:28:58 -07001464 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001465 if (host_err < 0) {
1466 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001468 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001469 if (created)
1470 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
wengang wang4ac35c22009-02-10 11:27:51 +08001472 nfsd_check_ignore_resizing(iap);
1473
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001475 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001477 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 /* XXX someone who knows this better please fix it for nsec */
1479 iap->ia_mtime.tv_sec = v_mtime;
1480 iap->ia_atime.tv_sec = v_atime;
1481 iap->ia_mtime.tv_nsec = 0;
1482 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 }
1484
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 set_attr:
Ben Myersf5019122010-02-17 14:05:11 -06001486 err = nfsd_create_setattr(rqstp, resfhp, iap);
1487
1488 /*
1489 * nfsd_setattr already committed the child (and possibly also the parent).
1490 */
1491 if (!err)
1492 err = nfserrno(commit_metadata(fhp));
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001493
Dave Hansen463c3192008-02-15 14:37:57 -08001494 mnt_drop_write(fhp->fh_export->ex_path.mnt);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001495 /*
1496 * Update the filehandle to get the new inode info.
1497 */
1498 if (!err)
1499 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500
1501 out:
1502 fh_unlock(fhp);
1503 if (dchild && !IS_ERR(dchild))
1504 dput(dchild);
1505 return err;
1506
1507 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001508 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 goto out;
1510}
1511#endif /* CONFIG_NFSD_V3 */
1512
1513/*
1514 * Read a symlink. On entry, *lenp must contain the maximum path length that
1515 * fits into the buffer. On return, it contains the true length.
1516 * N.B. After this call fhp needs an fh_put
1517 */
Al Viro6264d692006-10-19 23:28:58 -07001518__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1520{
1521 struct dentry *dentry;
1522 struct inode *inode;
1523 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001524 __be32 err;
1525 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001527 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (err)
1529 goto out;
1530
1531 dentry = fhp->fh_dentry;
1532 inode = dentry->d_inode;
1533
1534 err = nfserr_inval;
Al Viroacfa4382008-12-04 10:06:33 -05001535 if (!inode->i_op->readlink)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 goto out;
1537
Jan Blunck54775492008-02-14 19:38:39 -08001538 touch_atime(fhp->fh_export->ex_path.mnt, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 /* N.B. Why does this call need a get_fs()??
1540 * Remove the set_fs and watch the fireworks:-) --okir
1541 */
1542
1543 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001544 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 set_fs(oldfs);
1546
Al Viro6264d692006-10-19 23:28:58 -07001547 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001549 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 err = 0;
1551out:
1552 return err;
1553
1554out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001555 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 goto out;
1557}
1558
1559/*
1560 * Create a symlink and look up its inode
1561 * N.B. After this call _both_ fhp and resfhp need an fh_put
1562 */
Al Viro6264d692006-10-19 23:28:58 -07001563__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1565 char *fname, int flen,
1566 char *path, int plen,
1567 struct svc_fh *resfhp,
1568 struct iattr *iap)
1569{
1570 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001571 __be32 err, cerr;
1572 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
1574 err = nfserr_noent;
1575 if (!flen || !plen)
1576 goto out;
1577 err = nfserr_exist;
1578 if (isdotent(fname, flen))
1579 goto out;
1580
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001581 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (err)
1583 goto out;
1584 fh_lock(fhp);
1585 dentry = fhp->fh_dentry;
1586 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001587 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 if (IS_ERR(dnew))
1589 goto out_nfserr;
1590
Dave Hansen75c3f292008-02-15 14:37:45 -08001591 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1592 if (host_err)
1593 goto out_nfserr;
1594
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 if (unlikely(path[plen] != 0)) {
1596 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1597 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001598 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 else {
1600 strncpy(path_alloced, path, plen);
1601 path_alloced[plen] = 0;
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001602 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 kfree(path_alloced);
1604 }
1605 } else
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001606 host_err = vfs_symlink(dentry->d_inode, dnew, path);
Al Viro6264d692006-10-19 23:28:58 -07001607 err = nfserrno(host_err);
Ben Myersf5019122010-02-17 14:05:11 -06001608 if (!err)
1609 err = nfserrno(commit_metadata(fhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 fh_unlock(fhp);
1611
Dave Hansen75c3f292008-02-15 14:37:45 -08001612 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1613
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1615 dput(dnew);
1616 if (err==0) err = cerr;
1617out:
1618 return err;
1619
1620out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001621 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 goto out;
1623}
1624
1625/*
1626 * Create a hardlink
1627 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1628 */
Al Viro6264d692006-10-19 23:28:58 -07001629__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1631 char *name, int len, struct svc_fh *tfhp)
1632{
1633 struct dentry *ddir, *dnew, *dold;
1634 struct inode *dirp, *dest;
Al Viro6264d692006-10-19 23:28:58 -07001635 __be32 err;
1636 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001638 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 if (err)
1640 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001641 err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 if (err)
1643 goto out;
1644
1645 err = nfserr_perm;
1646 if (!len)
1647 goto out;
1648 err = nfserr_exist;
1649 if (isdotent(name, len))
1650 goto out;
1651
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001652 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 ddir = ffhp->fh_dentry;
1654 dirp = ddir->d_inode;
1655
1656 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001657 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 if (IS_ERR(dnew))
1659 goto out_nfserr;
1660
1661 dold = tfhp->fh_dentry;
1662 dest = dold->d_inode;
1663
Dave Hansen75c3f292008-02-15 14:37:45 -08001664 host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt);
1665 if (host_err) {
1666 err = nfserrno(host_err);
1667 goto out_dput;
1668 }
Al Viro6264d692006-10-19 23:28:58 -07001669 host_err = vfs_link(dold, dirp, dnew);
1670 if (!host_err) {
Ben Myersf5019122010-02-17 14:05:11 -06001671 err = nfserrno(commit_metadata(ffhp));
1672 if (!err)
1673 err = nfserrno(commit_metadata(tfhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 } else {
Al Viro6264d692006-10-19 23:28:58 -07001675 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 err = nfserr_acces;
1677 else
Al Viro6264d692006-10-19 23:28:58 -07001678 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 }
Dave Hansen75c3f292008-02-15 14:37:45 -08001680 mnt_drop_write(tfhp->fh_export->ex_path.mnt);
1681out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001683out_unlock:
1684 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685out:
1686 return err;
1687
1688out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001689 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001690 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691}
1692
1693/*
1694 * Rename a file
1695 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1696 */
Al Viro6264d692006-10-19 23:28:58 -07001697__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1699 struct svc_fh *tfhp, char *tname, int tlen)
1700{
1701 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1702 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001703 __be32 err;
1704 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001706 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 if (err)
1708 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001709 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 if (err)
1711 goto out;
1712
1713 fdentry = ffhp->fh_dentry;
1714 fdir = fdentry->d_inode;
1715
1716 tdentry = tfhp->fh_dentry;
1717 tdir = tdentry->d_inode;
1718
1719 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001720 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 goto out;
1722
1723 err = nfserr_perm;
1724 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1725 goto out;
1726
1727 /* cannot use fh_lock as we need deadlock protective ordering
1728 * so do it by hand */
1729 trap = lock_rename(tdentry, fdentry);
1730 ffhp->fh_locked = tfhp->fh_locked = 1;
1731 fill_pre_wcc(ffhp);
1732 fill_pre_wcc(tfhp);
1733
1734 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001735 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 if (IS_ERR(odentry))
1737 goto out_nfserr;
1738
Al Viro6264d692006-10-19 23:28:58 -07001739 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 if (!odentry->d_inode)
1741 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001742 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (odentry == trap)
1744 goto out_dput_old;
1745
1746 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001747 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 if (IS_ERR(ndentry))
1749 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001750 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 if (ndentry == trap)
1752 goto out_dput_new;
1753
Dave Hansen9079b1e2008-02-15 14:37:49 -08001754 if (svc_msnfs(ffhp) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 ((atomic_read(&odentry->d_count) > 1)
1756 || (atomic_read(&ndentry->d_count) > 1))) {
Al Viro6264d692006-10-19 23:28:58 -07001757 host_err = -EPERM;
Dave Hansen9079b1e2008-02-15 14:37:49 -08001758 goto out_dput_new;
1759 }
1760
1761 host_err = -EXDEV;
1762 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1763 goto out_dput_new;
1764 host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt);
1765 if (host_err)
1766 goto out_dput_new;
1767
Al Viro6264d692006-10-19 23:28:58 -07001768 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
Ben Myersf5019122010-02-17 14:05:11 -06001769 if (!host_err) {
1770 host_err = commit_metadata(tfhp);
Al Viro6264d692006-10-19 23:28:58 -07001771 if (!host_err)
Ben Myersf5019122010-02-17 14:05:11 -06001772 host_err = commit_metadata(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 }
1774
Dave Hansen9079b1e2008-02-15 14:37:49 -08001775 mnt_drop_write(ffhp->fh_export->ex_path.mnt);
1776
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 out_dput_new:
1778 dput(ndentry);
1779 out_dput_old:
1780 dput(odentry);
1781 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001782 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
1784 /* we cannot reply on fh_unlock on the two filehandles,
1785 * as that would do the wrong thing if the two directories
1786 * were the same, so again we do it by hand
1787 */
1788 fill_post_wcc(ffhp);
1789 fill_post_wcc(tfhp);
1790 unlock_rename(tdentry, fdentry);
1791 ffhp->fh_locked = tfhp->fh_locked = 0;
1792
1793out:
1794 return err;
1795}
1796
1797/*
1798 * Unlink a file or directory
1799 * N.B. After this call fhp needs an fh_put
1800 */
Al Viro6264d692006-10-19 23:28:58 -07001801__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1803 char *fname, int flen)
1804{
1805 struct dentry *dentry, *rdentry;
1806 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001807 __be32 err;
1808 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809
1810 err = nfserr_acces;
1811 if (!flen || isdotent(fname, flen))
1812 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001813 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 if (err)
1815 goto out;
1816
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001817 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 dentry = fhp->fh_dentry;
1819 dirp = dentry->d_inode;
1820
1821 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001822 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 if (IS_ERR(rdentry))
1824 goto out_nfserr;
1825
1826 if (!rdentry->d_inode) {
1827 dput(rdentry);
1828 err = nfserr_noent;
1829 goto out;
1830 }
1831
1832 if (!type)
1833 type = rdentry->d_inode->i_mode & S_IFMT;
1834
Dave Hansen06227532008-02-15 14:37:34 -08001835 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1836 if (host_err)
1837 goto out_nfserr;
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 if (type != S_IFDIR) { /* It's UNLINK */
1840#ifdef MSNFS
1841 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1842 (atomic_read(&rdentry->d_count) > 1)) {
Al Viro6264d692006-10-19 23:28:58 -07001843 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 } else
1845#endif
Al Viro6264d692006-10-19 23:28:58 -07001846 host_err = vfs_unlink(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 } else { /* It's RMDIR */
Al Viro6264d692006-10-19 23:28:58 -07001848 host_err = vfs_rmdir(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 }
1850
1851 dput(rdentry);
1852
Ben Myersf5019122010-02-17 14:05:11 -06001853 if (!host_err)
1854 host_err = commit_metadata(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
Dave Hansen06227532008-02-15 14:37:34 -08001856 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001858 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001859out:
1860 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861}
1862
1863/*
David Woodhouse14f7dd62008-07-31 20:29:12 +01001864 * We do this buffering because we must not call back into the file
1865 * system's ->lookup() method from the filldir callback. That may well
1866 * deadlock a number of file systems.
1867 *
1868 * This is based heavily on the implementation of same in XFS.
1869 */
1870struct buffered_dirent {
1871 u64 ino;
1872 loff_t offset;
1873 int namlen;
1874 unsigned int d_type;
1875 char name[];
1876};
1877
1878struct readdir_data {
1879 char *dirent;
1880 size_t used;
Al Viro53c9c5c2008-08-24 07:29:52 -04001881 int full;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001882};
1883
1884static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen,
1885 loff_t offset, u64 ino, unsigned int d_type)
David Woodhouse2628b762008-07-31 17:16:51 +01001886{
David Woodhouse14f7dd62008-07-31 20:29:12 +01001887 struct readdir_data *buf = __buf;
1888 struct buffered_dirent *de = (void *)(buf->dirent + buf->used);
1889 unsigned int reclen;
1890
1891 reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64));
Al Viro53c9c5c2008-08-24 07:29:52 -04001892 if (buf->used + reclen > PAGE_SIZE) {
1893 buf->full = 1;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001894 return -EINVAL;
Al Viro53c9c5c2008-08-24 07:29:52 -04001895 }
David Woodhouse14f7dd62008-07-31 20:29:12 +01001896
1897 de->namlen = namlen;
1898 de->offset = offset;
1899 de->ino = ino;
1900 de->d_type = d_type;
1901 memcpy(de->name, name, namlen);
1902 buf->used += reclen;
1903
1904 return 0;
1905}
1906
David Woodhouse2f9092e2009-04-20 23:18:37 +01001907static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func,
1908 struct readdir_cd *cdp, loff_t *offsetp)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001909{
1910 struct readdir_data buf;
1911 struct buffered_dirent *de;
David Woodhouse2628b762008-07-31 17:16:51 +01001912 int host_err;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001913 int size;
1914 loff_t offset;
David Woodhouse2628b762008-07-31 17:16:51 +01001915
David Woodhouse14f7dd62008-07-31 20:29:12 +01001916 buf.dirent = (void *)__get_free_page(GFP_KERNEL);
1917 if (!buf.dirent)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001918 return nfserrno(-ENOMEM);
David Woodhouse2628b762008-07-31 17:16:51 +01001919
David Woodhouse14f7dd62008-07-31 20:29:12 +01001920 offset = *offsetp;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001921
1922 while (1) {
David Woodhouse2f9092e2009-04-20 23:18:37 +01001923 struct inode *dir_inode = file->f_path.dentry->d_inode;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001924 unsigned int reclen;
1925
Doug Nazarb726e922008-11-05 06:16:28 -05001926 cdp->err = nfserr_eof; /* will be cleared on successful read */
David Woodhouse14f7dd62008-07-31 20:29:12 +01001927 buf.used = 0;
Al Viro53c9c5c2008-08-24 07:29:52 -04001928 buf.full = 0;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001929
1930 host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf);
Al Viro53c9c5c2008-08-24 07:29:52 -04001931 if (buf.full)
1932 host_err = 0;
1933
1934 if (host_err < 0)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001935 break;
1936
1937 size = buf.used;
1938
1939 if (!size)
1940 break;
1941
David Woodhouse2f9092e2009-04-20 23:18:37 +01001942 /*
1943 * Various filldir functions may end up calling back into
1944 * lookup_one_len() and the file system's ->lookup() method.
1945 * These expect i_mutex to be held, as it would within readdir.
1946 */
1947 host_err = mutex_lock_killable(&dir_inode->i_mutex);
1948 if (host_err)
1949 break;
1950
David Woodhouse14f7dd62008-07-31 20:29:12 +01001951 de = (struct buffered_dirent *)buf.dirent;
1952 while (size > 0) {
1953 offset = de->offset;
1954
1955 if (func(cdp, de->name, de->namlen, de->offset,
1956 de->ino, de->d_type))
David Woodhouse2f9092e2009-04-20 23:18:37 +01001957 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001958
1959 if (cdp->err != nfs_ok)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001960 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001961
1962 reclen = ALIGN(sizeof(*de) + de->namlen,
1963 sizeof(u64));
1964 size -= reclen;
1965 de = (struct buffered_dirent *)((char *)de + reclen);
1966 }
David Woodhouse2f9092e2009-04-20 23:18:37 +01001967 mutex_unlock(&dir_inode->i_mutex);
1968 if (size > 0) /* We bailed out early */
1969 break;
1970
David Woodhousec002a6c2008-08-17 17:21:18 +01001971 offset = vfs_llseek(file, 0, SEEK_CUR);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001972 }
1973
David Woodhouse14f7dd62008-07-31 20:29:12 +01001974 free_page((unsigned long)(buf.dirent));
David Woodhouse2628b762008-07-31 17:16:51 +01001975
1976 if (host_err)
1977 return nfserrno(host_err);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001978
1979 *offsetp = offset;
1980 return cdp->err;
David Woodhouse2628b762008-07-31 17:16:51 +01001981}
1982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983/*
1984 * Read entries from a directory.
1985 * The NFSv3/4 verifier we ignore for now.
1986 */
Al Viro6264d692006-10-19 23:28:58 -07001987__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001989 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990{
Al Viro6264d692006-10-19 23:28:58 -07001991 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 struct file *file;
1993 loff_t offset = *offsetp;
1994
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001995 err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 if (err)
1997 goto out;
1998
1999 offset = vfs_llseek(file, offset, 0);
2000 if (offset < 0) {
2001 err = nfserrno((int)offset);
2002 goto out_close;
2003 }
2004
David Woodhouse14f7dd62008-07-31 20:29:12 +01002005 err = nfsd_buffered_readdir(file, func, cdp, offsetp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006
2007 if (err == nfserr_eof || err == nfserr_toosmall)
2008 err = nfs_ok; /* can still be found in ->err */
2009out_close:
2010 nfsd_close(file);
2011out:
2012 return err;
2013}
2014
2015/*
2016 * Get file system stats
2017 * N.B. After this call fhp needs an fh_put
2018 */
Al Viro6264d692006-10-19 23:28:58 -07002019__be32
J. Bruce Fields04716e62008-08-07 13:00:20 -04002020nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021{
J. Bruce Fields04716e62008-08-07 13:00:20 -04002022 __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access);
David Howells726c3342006-06-23 02:02:58 -07002023 if (!err && vfs_statfs(fhp->fh_dentry,stat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 err = nfserr_io;
2025 return err;
2026}
2027
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002028static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002029{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002030 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002031}
2032
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033/*
2034 * Check for a user's access permissions to this inode.
2035 */
Al Viro6264d692006-10-19 23:28:58 -07002036__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07002037nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
2038 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039{
2040 struct inode *inode = dentry->d_inode;
Mimi Zohar14dba532009-05-27 09:31:52 -04002041 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 int err;
2043
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002044 if (acc == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 return 0;
2046#if 0
2047 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
2048 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002049 (acc & NFSD_MAY_READ)? " read" : "",
2050 (acc & NFSD_MAY_WRITE)? " write" : "",
2051 (acc & NFSD_MAY_EXEC)? " exec" : "",
2052 (acc & NFSD_MAY_SATTR)? " sattr" : "",
2053 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
2054 (acc & NFSD_MAY_LOCK)? " lock" : "",
2055 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 inode->i_mode,
2057 IS_IMMUTABLE(inode)? " immut" : "",
2058 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08002059 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060 dprintk(" owner %d/%d user %d/%d\n",
David Howells5cc0a842008-11-14 10:38:58 +11002061 inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062#endif
2063
2064 /* Normally we reject any write/sattr etc access on a read-only file
2065 * system. But if it is IRIX doing check on write-access for a
2066 * device special file, we ignore rofs.
2067 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002068 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
2069 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08002070 if (exp_rdonly(rqstp, exp) ||
2071 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002073 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 return nfserr_perm;
2075 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002076 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077 return nfserr_perm;
2078
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002079 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 /* If we cannot rely on authentication in NLM requests,
2081 * just allow locks, otherwise require read permission, or
2082 * ownership
2083 */
2084 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
2085 return 0;
2086 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002087 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 }
2089 /*
2090 * The file owner always gets access permission for accesses that
2091 * would normally be checked at open time. This is to make
2092 * file access work even when the client has done a fchmod(fd, 0).
2093 *
2094 * However, `cp foo bar' should fail nevertheless when bar is
2095 * readonly. A sensible way to do this might be to reject all
2096 * attempts to truncate a read-only file, because a creat() call
2097 * always implies file truncation.
2098 * ... but this isn't really fair. A process may reasonably call
2099 * ftruncate on an open file descriptor on a file with perm 000.
2100 * We must trust the client to do permission checking - using "ACCESS"
2101 * with NFSv3.
2102 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002103 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
David Howells5cc0a842008-11-14 10:38:58 +11002104 inode->i_uid == current_fsuid())
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 return 0;
2106
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002107 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Al Virof419a2e2008-07-22 00:07:17 -04002108 err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109
2110 /* Allow read access to binaries even when mode 111 */
2111 if (err == -EACCES && S_ISREG(inode->i_mode) &&
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002112 acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
Al Virof419a2e2008-07-22 00:07:17 -04002113 err = inode_permission(inode, MAY_EXEC);
Mimi Zohar14dba532009-05-27 09:31:52 -04002114 if (err)
2115 goto nfsd_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116
Mimi Zohar14dba532009-05-27 09:31:52 -04002117 /* Do integrity (permission) checking now, but defer incrementing
2118 * IMA counts to the actual file open.
2119 */
2120 path.mnt = exp->ex_path.mnt;
2121 path.dentry = dentry;
Mimi Zohar14dba532009-05-27 09:31:52 -04002122nfsd_out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 return err? nfserrno(err) : 0;
2124}
2125
2126void
2127nfsd_racache_shutdown(void)
2128{
Jeff Layton54a66e52008-08-13 22:03:27 -04002129 struct raparms *raparm, *last_raparm;
2130 unsigned int i;
2131
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 dprintk("nfsd: freeing readahead buffers.\n");
Jeff Layton54a66e52008-08-13 22:03:27 -04002133
2134 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
2135 raparm = raparm_hash[i].pb_head;
2136 while(raparm) {
2137 last_raparm = raparm;
2138 raparm = raparm->p_next;
2139 kfree(last_raparm);
2140 }
2141 raparm_hash[i].pb_head = NULL;
2142 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143}
2144/*
2145 * Initialize readahead param cache
2146 */
2147int
2148nfsd_racache_init(int cache_size)
2149{
2150 int i;
Greg Banksfce14562006-10-04 02:15:49 -07002151 int j = 0;
2152 int nperbucket;
Jeff Layton54a66e52008-08-13 22:03:27 -04002153 struct raparms **raparm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154
Greg Banksfce14562006-10-04 02:15:49 -07002155
Jeff Layton54a66e52008-08-13 22:03:27 -04002156 if (raparm_hash[0].pb_head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002158 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
2159 if (nperbucket < 2)
2160 nperbucket = 2;
2161 cache_size = nperbucket * RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08002162
2163 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
Jeff Layton54a66e52008-08-13 22:03:27 -04002164
2165 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
Yan Burman4b3bb062006-12-08 02:39:41 -08002166 spin_lock_init(&raparm_hash[i].pb_lock);
Jeff Layton54a66e52008-08-13 22:03:27 -04002167
2168 raparm = &raparm_hash[i].pb_head;
2169 for (j = 0; j < nperbucket; j++) {
2170 *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL);
2171 if (!*raparm)
2172 goto out_nomem;
2173 raparm = &(*raparm)->p_next;
2174 }
2175 *raparm = NULL;
Yan Burman4b3bb062006-12-08 02:39:41 -08002176 }
2177
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 nfsdstats.ra_size = cache_size;
2179 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002180
2181out_nomem:
2182 dprintk("nfsd: kmalloc failed, freeing readahead buffers\n");
2183 nfsd_racache_shutdown();
2184 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002186
2187#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2188struct posix_acl *
2189nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2190{
2191 struct inode *inode = fhp->fh_dentry->d_inode;
2192 char *name;
2193 void *value = NULL;
2194 ssize_t size;
2195 struct posix_acl *acl;
2196
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002197 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002198 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002199
2200 switch (type) {
2201 case ACL_TYPE_ACCESS:
2202 name = POSIX_ACL_XATTR_ACCESS;
2203 break;
2204 case ACL_TYPE_DEFAULT:
2205 name = POSIX_ACL_XATTR_DEFAULT;
2206 break;
2207 default:
2208 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002209 }
2210
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002211 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2212 if (size < 0)
2213 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002214
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002215 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002216 kfree(value);
2217 return acl;
2218}
2219
2220int
2221nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2222{
2223 struct inode *inode = fhp->fh_dentry->d_inode;
2224 char *name;
2225 void *value = NULL;
2226 size_t size;
2227 int error;
2228
Al Viroacfa4382008-12-04 10:06:33 -05002229 if (!IS_POSIXACL(inode) ||
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002230 !inode->i_op->setxattr || !inode->i_op->removexattr)
2231 return -EOPNOTSUPP;
2232 switch(type) {
2233 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002234 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002235 break;
2236 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002237 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002238 break;
2239 default:
2240 return -EOPNOTSUPP;
2241 }
2242
2243 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002244 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002245 value = kmalloc(size, GFP_KERNEL);
2246 if (!value)
2247 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002248 error = posix_acl_to_xattr(acl, value, size);
2249 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002250 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002251 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002252 } else
2253 size = 0;
2254
Dave Hansen18f335a2008-02-15 14:37:38 -08002255 error = mnt_want_write(fhp->fh_export->ex_path.mnt);
2256 if (error)
2257 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002258 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002259 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002260 else {
2261 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2262 error = 0;
2263 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002264 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002265 if (error == -ENODATA)
2266 error = 0;
2267 }
2268 }
Dave Hansen18f335a2008-02-15 14:37:38 -08002269 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002270
2271getout:
2272 kfree(value);
2273 return error;
2274}
2275#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */