blob: 184938fcff04d5dff712b8bee01f5d7e41bb20ad [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;
Ben Myersf5019122010-02-17 14:05:11 -0600284
285 if (!EX_ISSYNC(fhp->fh_export))
286 return 0;
287
Christoph Hellwigc37650162010-10-06 10:48:20 +0200288 if (export_ops->commit_metadata)
289 return export_ops->commit_metadata(inode);
290 return sync_inode_metadata(inode, 1);
Ben Myersf5019122010-02-17 14:05:11 -0600291}
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293/*
294 * Set various file attributes.
295 * N.B. After this call fhp needs an fh_put
296 */
Al Viro6264d692006-10-19 23:28:58 -0700297__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
299 int check_guard, time_t guardtime)
300{
301 struct dentry *dentry;
302 struct inode *inode;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200303 int accmode = NFSD_MAY_SATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 int ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700305 __be32 err;
306 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 int size_change = 0;
308
309 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200310 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 if (iap->ia_valid & ATTR_SIZE)
312 ftype = S_IFREG;
313
314 /* Get inode */
315 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800316 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 goto out;
318
319 dentry = fhp->fh_dentry;
320 inode = dentry->d_inode;
321
NeilBrown15b7a1b2005-11-07 01:00:23 -0800322 /* Ignore any mode updates on symlinks */
323 if (S_ISLNK(inode->i_mode))
324 iap->ia_valid &= ~ATTR_MODE;
325
326 if (!iap->ia_valid)
327 goto out;
328
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000329 /*
330 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 * which only requires access, and "set-[ac]time-to-X" which
332 * requires ownership.
333 * So if it looks like it might be "set both to the same time which
334 * is close to now", and if inode_change_ok fails, then we
335 * convert to "set to now" instead of "set to explicit time"
336 *
337 * We only call inode_change_ok as the last test as technically
338 * it is not an interface that we should be using. It is only
339 * valid if the filesystem does not define it's own i_op->setattr.
340 */
341#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
342#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000343 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
344 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
345 /*
346 * Looks probable.
347 *
348 * Now just make sure time is in the right ballpark.
349 * Solaris, at least, doesn't seem to care what the time
350 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000352 time_t delta = iap->ia_atime.tv_sec - get_seconds();
353 if (delta < 0)
354 delta = -delta;
355 if (delta < MAX_TOUCH_TIME_ERROR &&
356 inode_change_ok(inode, iap) != 0) {
357 /*
358 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
359 * This will cause notify_change to set these times
360 * to "now"
361 */
362 iap->ia_valid &= ~BOTH_TIME_SET;
363 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 }
365
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000366 /*
367 * The size case is special.
368 * It changes the file as well as the attributes.
369 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 if (iap->ia_valid & ATTR_SIZE) {
371 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200372 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
373 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 if (err)
375 goto out;
376 }
377
378 /*
379 * If we are changing the size of the file, then
380 * we need to break all leases.
381 */
Al Viro8737c932009-12-24 06:47:55 -0500382 host_err = break_lease(inode, O_WRONLY | O_NONBLOCK);
Al Viro6264d692006-10-19 23:28:58 -0700383 if (host_err == -EWOULDBLOCK)
384 host_err = -ETIMEDOUT;
385 if (host_err) /* ENOMEM or EWOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 goto out_nfserr;
387
Al Viro6264d692006-10-19 23:28:58 -0700388 host_err = get_write_access(inode);
389 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 goto out_nfserr;
391
392 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700393 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
394 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 put_write_access(inode);
396 goto out_nfserr;
397 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 }
399
Jeff Laytonca456252008-04-16 16:28:47 -0400400 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 if (iap->ia_valid & ATTR_MODE) {
402 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400403 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400404 }
405
406 /* Revoke setuid/setgid on chown */
Sachin S. Prabhu0953e622009-02-23 16:22:03 +0000407 if (!S_ISDIR(inode->i_mode) &&
408 (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
409 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) {
Jeff Laytonca456252008-04-16 16:28:47 -0400410 iap->ia_valid |= ATTR_KILL_PRIV;
411 if (iap->ia_valid & ATTR_MODE) {
412 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700413 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400414 if (iap->ia_mode & S_IXGRP)
415 iap->ia_mode &= ~S_ISGID;
416 } else {
417 /* set ATTR_KILL_* bits and let VFS handle it */
418 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 }
421
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 /* Change the attributes. */
423
424 iap->ia_valid |= ATTR_CTIME;
425
426 err = nfserr_notsync;
427 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
428 fh_lock(fhp);
Al Viro6264d692006-10-19 23:28:58 -0700429 host_err = notify_change(dentry, iap);
430 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 fh_unlock(fhp);
432 }
433 if (size_change)
434 put_write_access(inode);
435 if (!err)
Christoph Hellwigb160fda2010-06-01 21:59:18 +0200436 commit_metadata(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437out:
438 return err;
439
440out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700441 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 goto out;
443}
444
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800445#if defined(CONFIG_NFSD_V2_ACL) || \
446 defined(CONFIG_NFSD_V3_ACL) || \
447 defined(CONFIG_NFSD_V4)
448static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
449{
450 ssize_t buflen;
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530451 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800453 buflen = vfs_getxattr(dentry, key, NULL, 0);
454 if (buflen <= 0)
455 return buflen;
456
457 *buf = kmalloc(buflen, GFP_KERNEL);
458 if (!*buf)
459 return -ENOMEM;
460
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530461 ret = vfs_getxattr(dentry, key, *buf, buflen);
462 if (ret < 0)
463 kfree(*buf);
464 return ret;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800465}
466#endif
467
468#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469static int
470set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
471{
472 int len;
473 size_t buflen;
474 char *buf = NULL;
475 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477 buflen = posix_acl_xattr_size(pacl->a_count);
478 buf = kmalloc(buflen, GFP_KERNEL);
479 error = -ENOMEM;
480 if (buf == NULL)
481 goto out;
482
483 len = posix_acl_to_xattr(pacl, buf, buflen);
484 if (len < 0) {
485 error = len;
486 goto out;
487 }
488
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800489 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490out:
491 kfree(buf);
492 return error;
493}
494
Al Viro6264d692006-10-19 23:28:58 -0700495__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
497 struct nfs4_acl *acl)
498{
Al Viro6264d692006-10-19 23:28:58 -0700499 __be32 error;
500 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 struct dentry *dentry;
502 struct inode *inode;
503 struct posix_acl *pacl = NULL, *dpacl = NULL;
504 unsigned int flags = 0;
505
506 /* Get inode */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200507 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700509 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
511 dentry = fhp->fh_dentry;
512 inode = dentry->d_inode;
513 if (S_ISDIR(inode->i_mode))
514 flags = NFS4_ACL_DIR;
515
Al Viro6264d692006-10-19 23:28:58 -0700516 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
517 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700518 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700519 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 goto out_nfserr;
521
Al Viro6264d692006-10-19 23:28:58 -0700522 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
523 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700524 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700526 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700527 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700529out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 posix_acl_release(pacl);
531 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800533 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700534 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800535 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700536 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537}
538
539static struct posix_acl *
540_get_posix_acl(struct dentry *dentry, char *key)
541{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800542 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800544 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800546 buflen = nfsd_getxattr(dentry, key, &buf);
547 if (!buflen)
548 buflen = -ENODATA;
549 if (buflen <= 0)
550 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
552 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 kfree(buf);
554 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555}
556
557int
558nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
559{
560 struct inode *inode = dentry->d_inode;
561 int error = 0;
562 struct posix_acl *pacl = NULL, *dpacl = NULL;
563 unsigned int flags = 0;
564
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700565 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
567 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
568 if (IS_ERR(pacl)) {
569 error = PTR_ERR(pacl);
570 pacl = NULL;
571 goto out;
572 }
573
574 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700575 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
577 dpacl = NULL;
578 else if (IS_ERR(dpacl)) {
579 error = PTR_ERR(dpacl);
580 dpacl = NULL;
581 goto out;
582 }
583 flags = NFS4_ACL_DIR;
584 }
585
586 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
587 if (IS_ERR(*acl)) {
588 error = PTR_ERR(*acl);
589 *acl = NULL;
590 }
591 out:
592 posix_acl_release(pacl);
593 posix_acl_release(dpacl);
594 return error;
595}
596
J. Bruce Fields6a85d6c2010-07-06 12:39:12 -0400597#endif /* defined(CONFIG_NFSD_V4) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
599#ifdef CONFIG_NFSD_V3
600/*
601 * Check server access rights to a file system object
602 */
603struct accessmap {
604 u32 access;
605 int how;
606};
607static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200608 { NFS3_ACCESS_READ, NFSD_MAY_READ },
609 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
610 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
611 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 { 0, 0 }
614};
615
616static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200617 { NFS3_ACCESS_READ, NFSD_MAY_READ },
618 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
619 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
620 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
621 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623 { 0, 0 }
624};
625
626static struct accessmap nfs3_anyaccess[] = {
627 /* Some clients - Solaris 2.6 at least, make an access call
628 * to the server to check for access for things like /dev/null
629 * (which really, the server doesn't care about). So
630 * We provide simple access checking for them, looking
631 * mainly at mode bits, and we make sure to ignore read-only
632 * filesystem checks
633 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200634 { NFS3_ACCESS_READ, NFSD_MAY_READ },
635 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
636 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
637 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
639 { 0, 0 }
640};
641
Al Viro6264d692006-10-19 23:28:58 -0700642__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
644{
645 struct accessmap *map;
646 struct svc_export *export;
647 struct dentry *dentry;
648 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700649 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200651 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 if (error)
653 goto out;
654
655 export = fhp->fh_export;
656 dentry = fhp->fh_dentry;
657
658 if (S_ISREG(dentry->d_inode->i_mode))
659 map = nfs3_regaccess;
660 else if (S_ISDIR(dentry->d_inode->i_mode))
661 map = nfs3_diraccess;
662 else
663 map = nfs3_anyaccess;
664
665
666 query = *access;
667 for (; map->access; map++) {
668 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700669 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
671 sresult |= map->access;
672
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700673 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 switch (err2) {
675 case nfs_ok:
676 result |= map->access;
677 break;
678
679 /* the following error codes just mean the access was not allowed,
680 * rather than an error occurred */
681 case nfserr_rofs:
682 case nfserr_acces:
683 case nfserr_perm:
684 /* simply don't "or" in the access bit. */
685 break;
686 default:
687 error = err2;
688 goto out;
689 }
690 }
691 }
692 *access = result;
693 if (supported)
694 *supported = sresult;
695
696 out:
697 return error;
698}
699#endif /* CONFIG_NFSD_V3 */
700
701
702
703/*
704 * Open an existing file or directory.
705 * The access argument indicates the type of open (read/write/lock)
706 * N.B. After this call fhp needs an fh_put
707 */
Al Viro6264d692006-10-19 23:28:58 -0700708__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
710 int access, struct file **filp)
711{
712 struct dentry *dentry;
713 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700714 int flags = O_RDONLY|O_LARGEFILE;
715 __be32 err;
Jeff Layton91885252010-03-19 08:06:28 -0400716 int host_err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
David Howellse0e81732009-09-02 09:13:40 +0100718 validate_process_creds();
719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 /*
721 * If we get here, then the client has already done an "open",
722 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
723 * in case a chmod has now revoked permission.
724 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200725 err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 if (err)
727 goto out;
728
729 dentry = fhp->fh_dentry;
730 inode = dentry->d_inode;
731
732 /* Disallow write access to files with the append-only bit set
733 * or any access when mandatory locking enabled
734 */
735 err = nfserr_perm;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200736 if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400738 /*
739 * We must ignore files (but only files) which might have mandatory
740 * locks on them because there is no way to know if the accesser has
741 * the lock.
742 */
743 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 goto out;
745
746 if (!inode->i_fop)
747 goto out;
748
749 /*
750 * Check to see if there are any leases on this file.
751 * This may block while leases are broken.
752 */
Jeff Layton91885252010-03-19 08:06:28 -0400753 if (!(access & NFSD_MAY_NOT_BREAK_LEASE))
754 host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? O_WRONLY : 0));
Al Viro6264d692006-10-19 23:28:58 -0700755 if (host_err == -EWOULDBLOCK)
756 host_err = -ETIMEDOUT;
757 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 goto out_nfserr;
759
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200760 if (access & NFSD_MAY_WRITE) {
761 if (access & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700762 flags = O_RDWR|O_LARGEFILE;
763 else
764 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 }
Jan Blunck54775492008-02-14 19:38:39 -0800766 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
David Howells033a6662009-07-02 14:35:32 +0100767 flags, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700769 host_err = PTR_ERR(*filp);
Chuck Ebbertaeaa5cc2010-02-15 18:07:39 -0500770 else
771 host_err = ima_file_check(*filp, access);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700773 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774out:
David Howellse0e81732009-09-02 09:13:40 +0100775 validate_process_creds();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 return err;
777}
778
779/*
780 * Close a file.
781 */
782void
783nfsd_close(struct file *filp)
784{
785 fput(filp);
786}
787
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500788/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 * Obtain the readahead parameters for the file
790 * specified by (dev, ino).
791 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793static inline struct raparms *
794nfsd_get_raparms(dev_t dev, ino_t ino)
795{
796 struct raparms *ra, **rap, **frap = NULL;
797 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700798 unsigned int hash;
799 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
Greg Banksfce14562006-10-04 02:15:49 -0700801 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
802 rab = &raparm_hash[hash];
803
804 spin_lock(&rab->pb_lock);
805 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 if (ra->p_ino == ino && ra->p_dev == dev)
807 goto found;
808 depth++;
809 if (ra->p_count == 0)
810 frap = rap;
811 }
812 depth = nfsdstats.ra_size*11/10;
813 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700814 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 return NULL;
816 }
817 rap = frap;
818 ra = *frap;
819 ra->p_dev = dev;
820 ra->p_ino = ino;
821 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700822 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823found:
Greg Banksfce14562006-10-04 02:15:49 -0700824 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700826 ra->p_next = rab->pb_head;
827 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 }
829 ra->p_count++;
830 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700831 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 return ra;
833}
834
835/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200836 * Grab and keep cached pages associated with a file in the svc_rqst
837 * so that they can be passed to the network sendmsg/sendpage routines
838 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 */
840static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200841nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
842 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843{
Jens Axboecf8208d2007-06-12 21:22:14 +0200844 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700845 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200846 struct page *page = buf->page;
847 size_t size;
848 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Jens Axboecac36bb02007-06-14 13:10:48 +0200850 ret = buf->ops->confirm(pipe, buf);
Jens Axboecf8208d2007-06-12 21:22:14 +0200851 if (unlikely(ret))
852 return ret;
853
854 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856 if (rqstp->rq_res.page_len == 0) {
857 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700858 put_page(*pp);
859 *pp = page;
860 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200861 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700863 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800865 if (*pp)
866 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700867 *pp = page;
868 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700870 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 return size;
874}
875
Jens Axboecf8208d2007-06-12 21:22:14 +0200876static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
877 struct splice_desc *sd)
878{
879 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
880}
881
Dave Hansena8754be2007-10-16 23:31:15 -0700882static inline int svc_msnfs(struct svc_fh *ffhp)
883{
884#ifdef MSNFS
885 return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS);
886#else
887 return 0;
888#endif
889}
890
Al Viro6264d692006-10-19 23:28:58 -0700891static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
893 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
894{
895 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700897 __be32 err;
898 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
900 err = nfserr_perm;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800901 inode = file->f_path.dentry->d_inode;
Dave Hansena8754be2007-10-16 23:31:15 -0700902
903 if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
Jens Axboecf8208d2007-06-12 21:22:14 +0200906 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
907 struct splice_desc sd = {
908 .len = 0,
909 .total_len = *count,
910 .pos = offset,
911 .u.data = rqstp,
912 };
913
Jens Axboe4fbef202007-07-13 22:42:20 +0200914 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200915 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 } else {
917 oldfs = get_fs();
918 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700919 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 set_fs(oldfs);
921 }
922
Al Viro6264d692006-10-19 23:28:58 -0700923 if (host_err >= 0) {
924 nfsdstats.io_read += host_err;
925 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 err = 0;
Eric Paris2a12a9d2009-12-17 21:24:21 -0500927 fsnotify_access(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 } else
Al Viro6264d692006-10-19 23:28:58 -0700929 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930out:
931 return err;
932}
933
Neil Brown9f708e42006-01-06 00:19:59 -0800934static void kill_suid(struct dentry *dentry)
935{
936 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700937 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800938
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800939 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800940 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800941 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800942}
943
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700944/*
945 * Gathered writes: If another process is currently writing to the file,
946 * there's a high chance this is another nfsd (triggered by a bulk write
947 * from a client's biod). Rather than syncing the file with each write
948 * request, we sleep for 10 msec.
949 *
950 * I don't know if this roughly approximates C. Juszak's idea of
951 * gathered writes, but it's a nice and simple solution (IMHO), and it
952 * seems to work:-)
953 *
954 * Note: we do this only in the NFSv2 case, since v3 and higher have a
955 * better tool (separate unstable writes and commits) for solving this
956 * problem.
957 */
958static int wait_for_concurrent_writes(struct file *file)
959{
960 struct inode *inode = file->f_path.dentry->d_inode;
961 static ino_t last_ino;
962 static dev_t last_dev;
963 int err = 0;
964
965 if (atomic_read(&inode->i_writecount) > 1
966 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
967 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
968 msleep(10);
969 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
970 }
971
972 if (inode->i_state & I_DIRTY) {
973 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Christoph Hellwig8018ab02010-03-22 17:32:25 +0100974 err = vfs_fsync(file, 0);
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700975 }
976 last_ino = inode->i_ino;
977 last_dev = inode->i_sb->s_dev;
978 return err;
979}
980
Al Viro6264d692006-10-19 23:28:58 -0700981static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
983 loff_t offset, struct kvec *vec, int vlen,
David Shaw31dec252009-03-05 20:16:14 -0500984 unsigned long *cnt, int *stablep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985{
986 struct svc_export *exp;
987 struct dentry *dentry;
988 struct inode *inode;
989 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700990 __be32 err = 0;
991 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 int stable = *stablep;
Trond Myklebust48e03bc2009-06-05 12:35:15 -0400993 int use_wgather;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
NeilBrown45bd3b32006-01-18 17:43:50 -0800995#ifdef MSNFS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 err = nfserr_perm;
997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
David Shaw31dec252009-03-05 20:16:14 -0500999 (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 goto out;
1001#endif
1002
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001003 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 inode = dentry->d_inode;
1005 exp = fhp->fh_export;
1006
1007 /*
1008 * Request sync writes if
1009 * - the sync export option has been set, or
1010 * - the client requested O_SYNC behavior (NFSv3 feature).
1011 * - The file system doesn't support fsync().
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001012 * When NFSv2 gathered writes have been configured for this volume,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 * flushing the data to disk is handled separately below.
1014 */
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001015 use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
Harvey Harrison3ba15142008-02-20 12:49:02 -08001017 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 stable = 2;
1019 *stablep = 2; /* FILE_SYNC */
1020 }
1021
1022 if (!EX_ISSYNC(exp))
1023 stable = 0;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001024 if (stable && !use_wgather) {
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001025 spin_lock(&file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 file->f_flags |= O_SYNC;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001027 spin_unlock(&file->f_lock);
1028 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029
1030 /* Write the data. */
1031 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001032 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 set_fs(oldfs);
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001034 if (host_err < 0)
1035 goto out_nfserr;
1036 *cnt = host_err;
1037 nfsdstats.io_write += host_err;
Eric Paris2a12a9d2009-12-17 21:24:21 -05001038 fsnotify_modify(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
1040 /* clear setuid/setgid flag after write */
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001041 if (inode->i_mode & (S_ISUID | S_ISGID))
Neil Brown9f708e42006-01-06 00:19:59 -08001042 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001044 if (stable && use_wgather)
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001045 host_err = wait_for_concurrent_writes(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001047out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001048 dprintk("nfsd: write complete host_err=%d\n", host_err);
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001049 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 err = 0;
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001051 else
Al Viro6264d692006-10-19 23:28:58 -07001052 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053out:
1054 return err;
1055}
1056
1057/*
1058 * Read data from a file. count must contain the requested read count
1059 * on entry. On return, *count contains the number of bytes actually read.
1060 * N.B. After this call fhp needs an fh_put
1061 */
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001062__be32 nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
J. Bruce Fieldsfa0a2122010-07-27 16:48:54 -04001063 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
1064{
1065 struct file *file;
1066 struct inode *inode;
1067 struct raparms *ra;
1068 __be32 err;
1069
1070 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
1071 if (err)
1072 return err;
1073
1074 inode = file->f_path.dentry->d_inode;
1075
1076 /* Get readahead parameters */
1077 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
1078
1079 if (ra && ra->p_set)
1080 file->f_ra = ra->p_ra;
1081
1082 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1083
1084 /* Write back readahead params */
1085 if (ra) {
1086 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
1087 spin_lock(&rab->pb_lock);
1088 ra->p_ra = file->f_ra;
1089 ra->p_set = 1;
1090 ra->p_count--;
1091 spin_unlock(&rab->pb_lock);
1092 }
1093
1094 nfsd_close(file);
1095 return err;
1096}
1097
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001098/* As above, but use the provided file descriptor. */
Al Viro6264d692006-10-19 23:28:58 -07001099__be32
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001100nfsd_read_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 loff_t offset, struct kvec *vec, int vlen,
1102 unsigned long *count)
1103{
Al Viro6264d692006-10-19 23:28:58 -07001104 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
1106 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001107 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001108 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 if (err)
1110 goto out;
1111 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001112 } else /* Note file may still be NULL in NFSv4 special stateid case: */
1113 err = nfsd_read(rqstp, fhp, offset, vec, vlen, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114out:
1115 return err;
1116}
1117
1118/*
1119 * Write data to a file.
1120 * The stable flag requests synchronous writes.
1121 * N.B. After this call fhp needs an fh_put
1122 */
Al Viro6264d692006-10-19 23:28:58 -07001123__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
David Shaw31dec252009-03-05 20:16:14 -05001125 loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 int *stablep)
1127{
Al Viro6264d692006-10-19 23:28:58 -07001128 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129
1130 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001131 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001132 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 if (err)
1134 goto out;
1135 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1136 stablep);
1137 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001138 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 if (err)
1140 goto out;
1141
1142 if (cnt)
1143 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1144 cnt, stablep);
1145 nfsd_close(file);
1146 }
1147out:
1148 return err;
1149}
1150
1151#ifdef CONFIG_NFSD_V3
1152/*
1153 * Commit all pending writes to stable storage.
Trond Myklebustaa696a62010-01-29 16:44:25 -05001154 *
1155 * Note: we only guarantee that data that lies within the range specified
1156 * by the 'offset' and 'count' parameters will be synced.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 *
1158 * Unfortunately we cannot lock the file to make sure we return full WCC
1159 * data to the client, as locking happens lower down in the filesystem.
1160 */
Al Viro6264d692006-10-19 23:28:58 -07001161__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1163 loff_t offset, unsigned long count)
1164{
1165 struct file *file;
Trond Myklebustaa696a62010-01-29 16:44:25 -05001166 loff_t end = LLONG_MAX;
1167 __be32 err = nfserr_inval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Trond Myklebustaa696a62010-01-29 16:44:25 -05001169 if (offset < 0)
1170 goto out;
1171 if (count != 0) {
1172 end = offset + (loff_t)count - 1;
1173 if (end < offset)
1174 goto out;
1175 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176
Jeff Layton91885252010-03-19 08:06:28 -04001177 err = nfsd_open(rqstp, fhp, S_IFREG,
1178 NFSD_MAY_WRITE|NFSD_MAY_NOT_BREAK_LEASE, &file);
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001179 if (err)
Trond Myklebustaa696a62010-01-29 16:44:25 -05001180 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 if (EX_ISSYNC(fhp->fh_export)) {
Christoph Hellwig8018ab02010-03-22 17:32:25 +01001182 int err2 = vfs_fsync_range(file, offset, end, 0);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001183
1184 if (err2 != -EINVAL)
1185 err = nfserrno(err2);
1186 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 err = nfserr_notsupp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 }
1189
1190 nfsd_close(file);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001191out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 return err;
1193}
1194#endif /* CONFIG_NFSD_V3 */
1195
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001196static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001197nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1198 struct iattr *iap)
1199{
1200 /*
1201 * Mode has already been set earlier in create:
1202 */
1203 iap->ia_valid &= ~ATTR_MODE;
1204 /*
1205 * Setting uid/gid works only for root. Irix appears to
1206 * send along the gid on create when it tries to implement
1207 * setgid directories via NFS:
1208 */
David Howells5cc0a842008-11-14 10:38:58 +11001209 if (current_fsuid() != 0)
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001210 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1211 if (iap->ia_valid)
1212 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1213 return 0;
1214}
1215
wengang wang4ac35c22009-02-10 11:27:51 +08001216/* HPUX client sometimes creates a file in mode 000, and sets size to 0.
1217 * setting size to 0 may fail for some specific file systems by the permission
1218 * checking which requires WRITE permission but the mode is 000.
1219 * we ignore the resizing(to 0) on the just new created file, since the size is
1220 * 0 after file created.
1221 *
1222 * call this only after vfs_create() is called.
1223 * */
1224static void
1225nfsd_check_ignore_resizing(struct iattr *iap)
1226{
1227 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
1228 iap->ia_valid &= ~ATTR_SIZE;
1229}
1230
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231/*
1232 * Create a file (regular, directory, device, fifo); UNIX sockets
1233 * not yet implemented.
1234 * If the response fh has been verified, the parent directory should
1235 * already be locked. Note that the parent directory is left locked.
1236 *
1237 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1238 */
Al Viro6264d692006-10-19 23:28:58 -07001239__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1241 char *fname, int flen, struct iattr *iap,
1242 int type, dev_t rdev, struct svc_fh *resfhp)
1243{
1244 struct dentry *dentry, *dchild = NULL;
1245 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001246 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001247 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001248 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249
1250 err = nfserr_perm;
1251 if (!flen)
1252 goto out;
1253 err = nfserr_exist;
1254 if (isdotent(fname, flen))
1255 goto out;
1256
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001257 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 if (err)
1259 goto out;
1260
1261 dentry = fhp->fh_dentry;
1262 dirp = dentry->d_inode;
1263
1264 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001265 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 goto out;
1267 /*
1268 * Check whether the response file handle has been verified yet.
1269 * If it has, the parent directory should already be locked.
1270 */
1271 if (!resfhp->fh_dentry) {
1272 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001273 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001275 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 if (IS_ERR(dchild))
1277 goto out_nfserr;
1278 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1279 if (err)
1280 goto out;
1281 } else {
1282 /* called from nfsd_proc_create */
1283 dchild = dget(resfhp->fh_dentry);
1284 if (!fhp->fh_locked) {
1285 /* not actually possible */
1286 printk(KERN_ERR
1287 "nfsd_create: parent %s/%s not locked!\n",
1288 dentry->d_parent->d_name.name,
1289 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001290 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 goto out;
1292 }
1293 }
1294 /*
1295 * Make sure the child dentry is still negative ...
1296 */
1297 err = nfserr_exist;
1298 if (dchild->d_inode) {
1299 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1300 dentry->d_name.name, dchild->d_name.name);
1301 goto out;
1302 }
1303
1304 if (!(iap->ia_valid & ATTR_MODE))
1305 iap->ia_mode = 0;
1306 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1307
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001308 err = nfserr_inval;
1309 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1310 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1311 type);
1312 goto out;
1313 }
1314
1315 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1316 if (host_err)
1317 goto out_nfserr;
1318
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 /*
1320 * Get the dir op function pointer.
1321 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001322 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 switch (type) {
1324 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001325 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
wengang wang4ac35c22009-02-10 11:27:51 +08001326 if (!host_err)
1327 nfsd_check_ignore_resizing(iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 break;
1329 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001330 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 break;
1332 case S_IFCHR:
1333 case S_IFBLK:
1334 case S_IFIFO:
1335 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001336 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001338 }
1339 if (host_err < 0) {
1340 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1341 goto out_nfserr;
1342 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
Ben Myersf5019122010-02-17 14:05:11 -06001344 err = nfsd_create_setattr(rqstp, resfhp, iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
Ben Myersf5019122010-02-17 14:05:11 -06001346 /*
1347 * nfsd_setattr already committed the child. Transactional filesystems
1348 * had a chance to commit changes for both parent and child
1349 * simultaneously making the following commit_metadata a noop.
1350 */
1351 err2 = nfserrno(commit_metadata(fhp));
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001352 if (err2)
1353 err = err2;
Dave Hansen463c3192008-02-15 14:37:57 -08001354 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 /*
1356 * Update the file handle to get the new inode info.
1357 */
1358 if (!err)
1359 err = fh_update(resfhp);
1360out:
1361 if (dchild && !IS_ERR(dchild))
1362 dput(dchild);
1363 return err;
1364
1365out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001366 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 goto out;
1368}
1369
1370#ifdef CONFIG_NFSD_V3
1371/*
1372 * NFSv3 version of nfsd_create
1373 */
Al Viro6264d692006-10-19 23:28:58 -07001374__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
1376 char *fname, int flen, struct iattr *iap,
1377 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001378 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379{
1380 struct dentry *dentry, *dchild = NULL;
1381 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001382 __be32 err;
1383 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385
1386 err = nfserr_perm;
1387 if (!flen)
1388 goto out;
1389 err = nfserr_exist;
1390 if (isdotent(fname, flen))
1391 goto out;
1392 if (!(iap->ia_valid & ATTR_MODE))
1393 iap->ia_mode = 0;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001394 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 if (err)
1396 goto out;
1397
1398 dentry = fhp->fh_dentry;
1399 dirp = dentry->d_inode;
1400
1401 /* Get all the sanity checks out of the way before
1402 * we lock the parent. */
1403 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001404 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001406 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407
1408 /*
1409 * Compose the response file handle.
1410 */
1411 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001412 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 if (IS_ERR(dchild))
1414 goto out_nfserr;
1415
1416 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1417 if (err)
1418 goto out;
1419
1420 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001421 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001422 * the high bit set, so just clear the high bits. If this is
1423 * ever changed to use different attrs for storing the
1424 * verifier, then do_open_lookup() will also need to be fixed
1425 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 */
1427 v_mtime = verifier[0]&0x7fffffff;
1428 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 }
1430
Dave Hansen463c3192008-02-15 14:37:57 -08001431 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1432 if (host_err)
1433 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 if (dchild->d_inode) {
1435 err = 0;
1436
1437 switch (createmode) {
1438 case NFS3_CREATE_UNCHECKED:
1439 if (! S_ISREG(dchild->d_inode->i_mode))
1440 err = nfserr_exist;
1441 else if (truncp) {
1442 /* in nfsv4, we need to treat this case a little
1443 * differently. we don't want to truncate the
1444 * file now; this would be wrong if the OPEN
1445 * fails for some other reason. furthermore,
1446 * if the size is nonzero, we should ignore it
1447 * according to spec!
1448 */
1449 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1450 }
1451 else {
1452 iap->ia_valid &= ATTR_SIZE;
1453 goto set_attr;
1454 }
1455 break;
1456 case NFS3_CREATE_EXCLUSIVE:
1457 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1458 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 && dchild->d_inode->i_size == 0 )
1460 break;
1461 /* fallthru */
1462 case NFS3_CREATE_GUARDED:
1463 err = nfserr_exist;
1464 }
Dave Hansen463c3192008-02-15 14:37:57 -08001465 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 goto out;
1467 }
1468
Al Viro6264d692006-10-19 23:28:58 -07001469 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001470 if (host_err < 0) {
1471 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001473 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001474 if (created)
1475 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476
wengang wang4ac35c22009-02-10 11:27:51 +08001477 nfsd_check_ignore_resizing(iap);
1478
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001480 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001482 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 /* XXX someone who knows this better please fix it for nsec */
1484 iap->ia_mtime.tv_sec = v_mtime;
1485 iap->ia_atime.tv_sec = v_atime;
1486 iap->ia_mtime.tv_nsec = 0;
1487 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 }
1489
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 set_attr:
Ben Myersf5019122010-02-17 14:05:11 -06001491 err = nfsd_create_setattr(rqstp, resfhp, iap);
1492
1493 /*
1494 * nfsd_setattr already committed the child (and possibly also the parent).
1495 */
1496 if (!err)
1497 err = nfserrno(commit_metadata(fhp));
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001498
Dave Hansen463c3192008-02-15 14:37:57 -08001499 mnt_drop_write(fhp->fh_export->ex_path.mnt);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001500 /*
1501 * Update the filehandle to get the new inode info.
1502 */
1503 if (!err)
1504 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505
1506 out:
1507 fh_unlock(fhp);
1508 if (dchild && !IS_ERR(dchild))
1509 dput(dchild);
1510 return err;
1511
1512 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001513 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 goto out;
1515}
1516#endif /* CONFIG_NFSD_V3 */
1517
1518/*
1519 * Read a symlink. On entry, *lenp must contain the maximum path length that
1520 * fits into the buffer. On return, it contains the true length.
1521 * N.B. After this call fhp needs an fh_put
1522 */
Al Viro6264d692006-10-19 23:28:58 -07001523__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1525{
1526 struct dentry *dentry;
1527 struct inode *inode;
1528 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001529 __be32 err;
1530 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001532 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 if (err)
1534 goto out;
1535
1536 dentry = fhp->fh_dentry;
1537 inode = dentry->d_inode;
1538
1539 err = nfserr_inval;
Al Viroacfa4382008-12-04 10:06:33 -05001540 if (!inode->i_op->readlink)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 goto out;
1542
Jan Blunck54775492008-02-14 19:38:39 -08001543 touch_atime(fhp->fh_export->ex_path.mnt, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 /* N.B. Why does this call need a get_fs()??
1545 * Remove the set_fs and watch the fireworks:-) --okir
1546 */
1547
1548 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001549 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 set_fs(oldfs);
1551
Al Viro6264d692006-10-19 23:28:58 -07001552 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001554 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 err = 0;
1556out:
1557 return err;
1558
1559out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001560 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 goto out;
1562}
1563
1564/*
1565 * Create a symlink and look up its inode
1566 * N.B. After this call _both_ fhp and resfhp need an fh_put
1567 */
Al Viro6264d692006-10-19 23:28:58 -07001568__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1570 char *fname, int flen,
1571 char *path, int plen,
1572 struct svc_fh *resfhp,
1573 struct iattr *iap)
1574{
1575 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001576 __be32 err, cerr;
1577 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578
1579 err = nfserr_noent;
1580 if (!flen || !plen)
1581 goto out;
1582 err = nfserr_exist;
1583 if (isdotent(fname, flen))
1584 goto out;
1585
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001586 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 if (err)
1588 goto out;
1589 fh_lock(fhp);
1590 dentry = fhp->fh_dentry;
1591 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001592 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 if (IS_ERR(dnew))
1594 goto out_nfserr;
1595
Dave Hansen75c3f292008-02-15 14:37:45 -08001596 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1597 if (host_err)
1598 goto out_nfserr;
1599
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 if (unlikely(path[plen] != 0)) {
1601 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1602 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001603 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 else {
1605 strncpy(path_alloced, path, plen);
1606 path_alloced[plen] = 0;
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001607 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 kfree(path_alloced);
1609 }
1610 } else
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001611 host_err = vfs_symlink(dentry->d_inode, dnew, path);
Al Viro6264d692006-10-19 23:28:58 -07001612 err = nfserrno(host_err);
Ben Myersf5019122010-02-17 14:05:11 -06001613 if (!err)
1614 err = nfserrno(commit_metadata(fhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 fh_unlock(fhp);
1616
Dave Hansen75c3f292008-02-15 14:37:45 -08001617 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1620 dput(dnew);
1621 if (err==0) err = cerr;
1622out:
1623 return err;
1624
1625out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001626 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 goto out;
1628}
1629
1630/*
1631 * Create a hardlink
1632 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1633 */
Al Viro6264d692006-10-19 23:28:58 -07001634__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1636 char *name, int len, struct svc_fh *tfhp)
1637{
1638 struct dentry *ddir, *dnew, *dold;
J. Bruce Fields55b13352010-07-19 16:38:24 -04001639 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001640 __be32 err;
1641 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001643 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 if (err)
1645 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001646 err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 if (err)
1648 goto out;
1649
1650 err = nfserr_perm;
1651 if (!len)
1652 goto out;
1653 err = nfserr_exist;
1654 if (isdotent(name, len))
1655 goto out;
1656
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001657 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 ddir = ffhp->fh_dentry;
1659 dirp = ddir->d_inode;
1660
1661 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001662 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 if (IS_ERR(dnew))
1664 goto out_nfserr;
1665
1666 dold = tfhp->fh_dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Dave Hansen75c3f292008-02-15 14:37:45 -08001668 host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt);
1669 if (host_err) {
1670 err = nfserrno(host_err);
1671 goto out_dput;
1672 }
Al Viro6264d692006-10-19 23:28:58 -07001673 host_err = vfs_link(dold, dirp, dnew);
1674 if (!host_err) {
Ben Myersf5019122010-02-17 14:05:11 -06001675 err = nfserrno(commit_metadata(ffhp));
1676 if (!err)
1677 err = nfserrno(commit_metadata(tfhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 } else {
Al Viro6264d692006-10-19 23:28:58 -07001679 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 err = nfserr_acces;
1681 else
Al Viro6264d692006-10-19 23:28:58 -07001682 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 }
Dave Hansen75c3f292008-02-15 14:37:45 -08001684 mnt_drop_write(tfhp->fh_export->ex_path.mnt);
1685out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001687out_unlock:
1688 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689out:
1690 return err;
1691
1692out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001693 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001694 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695}
1696
1697/*
1698 * Rename a file
1699 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1700 */
Al Viro6264d692006-10-19 23:28:58 -07001701__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1703 struct svc_fh *tfhp, char *tname, int tlen)
1704{
1705 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1706 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001707 __be32 err;
1708 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001710 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 if (err)
1712 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001713 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (err)
1715 goto out;
1716
1717 fdentry = ffhp->fh_dentry;
1718 fdir = fdentry->d_inode;
1719
1720 tdentry = tfhp->fh_dentry;
1721 tdir = tdentry->d_inode;
1722
1723 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001724 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 goto out;
1726
1727 err = nfserr_perm;
1728 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1729 goto out;
1730
1731 /* cannot use fh_lock as we need deadlock protective ordering
1732 * so do it by hand */
1733 trap = lock_rename(tdentry, fdentry);
1734 ffhp->fh_locked = tfhp->fh_locked = 1;
1735 fill_pre_wcc(ffhp);
1736 fill_pre_wcc(tfhp);
1737
1738 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001739 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 if (IS_ERR(odentry))
1741 goto out_nfserr;
1742
Al Viro6264d692006-10-19 23:28:58 -07001743 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 if (!odentry->d_inode)
1745 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001746 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 if (odentry == trap)
1748 goto out_dput_old;
1749
1750 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001751 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 if (IS_ERR(ndentry))
1753 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001754 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 if (ndentry == trap)
1756 goto out_dput_new;
1757
Dave Hansen9079b1e2008-02-15 14:37:49 -08001758 if (svc_msnfs(ffhp) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 ((atomic_read(&odentry->d_count) > 1)
1760 || (atomic_read(&ndentry->d_count) > 1))) {
Al Viro6264d692006-10-19 23:28:58 -07001761 host_err = -EPERM;
Dave Hansen9079b1e2008-02-15 14:37:49 -08001762 goto out_dput_new;
1763 }
1764
1765 host_err = -EXDEV;
1766 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1767 goto out_dput_new;
1768 host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt);
1769 if (host_err)
1770 goto out_dput_new;
1771
Al Viro6264d692006-10-19 23:28:58 -07001772 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
Ben Myersf5019122010-02-17 14:05:11 -06001773 if (!host_err) {
1774 host_err = commit_metadata(tfhp);
Al Viro6264d692006-10-19 23:28:58 -07001775 if (!host_err)
Ben Myersf5019122010-02-17 14:05:11 -06001776 host_err = commit_metadata(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 }
1778
Dave Hansen9079b1e2008-02-15 14:37:49 -08001779 mnt_drop_write(ffhp->fh_export->ex_path.mnt);
1780
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 out_dput_new:
1782 dput(ndentry);
1783 out_dput_old:
1784 dput(odentry);
1785 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001786 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
1788 /* we cannot reply on fh_unlock on the two filehandles,
1789 * as that would do the wrong thing if the two directories
1790 * were the same, so again we do it by hand
1791 */
1792 fill_post_wcc(ffhp);
1793 fill_post_wcc(tfhp);
1794 unlock_rename(tdentry, fdentry);
1795 ffhp->fh_locked = tfhp->fh_locked = 0;
1796
1797out:
1798 return err;
1799}
1800
1801/*
1802 * Unlink a file or directory
1803 * N.B. After this call fhp needs an fh_put
1804 */
Al Viro6264d692006-10-19 23:28:58 -07001805__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1807 char *fname, int flen)
1808{
1809 struct dentry *dentry, *rdentry;
1810 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001811 __be32 err;
1812 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813
1814 err = nfserr_acces;
1815 if (!flen || isdotent(fname, flen))
1816 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001817 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 if (err)
1819 goto out;
1820
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001821 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 dentry = fhp->fh_dentry;
1823 dirp = dentry->d_inode;
1824
1825 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001826 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 if (IS_ERR(rdentry))
1828 goto out_nfserr;
1829
1830 if (!rdentry->d_inode) {
1831 dput(rdentry);
1832 err = nfserr_noent;
1833 goto out;
1834 }
1835
1836 if (!type)
1837 type = rdentry->d_inode->i_mode & S_IFMT;
1838
Dave Hansen06227532008-02-15 14:37:34 -08001839 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1840 if (host_err)
1841 goto out_nfserr;
1842
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 if (type != S_IFDIR) { /* It's UNLINK */
1844#ifdef MSNFS
1845 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1846 (atomic_read(&rdentry->d_count) > 1)) {
Al Viro6264d692006-10-19 23:28:58 -07001847 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 } else
1849#endif
Al Viro6264d692006-10-19 23:28:58 -07001850 host_err = vfs_unlink(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 } else { /* It's RMDIR */
Al Viro6264d692006-10-19 23:28:58 -07001852 host_err = vfs_rmdir(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 }
1854
1855 dput(rdentry);
1856
Ben Myersf5019122010-02-17 14:05:11 -06001857 if (!host_err)
1858 host_err = commit_metadata(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859
Dave Hansen06227532008-02-15 14:37:34 -08001860 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001862 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001863out:
1864 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865}
1866
1867/*
David Woodhouse14f7dd62008-07-31 20:29:12 +01001868 * We do this buffering because we must not call back into the file
1869 * system's ->lookup() method from the filldir callback. That may well
1870 * deadlock a number of file systems.
1871 *
1872 * This is based heavily on the implementation of same in XFS.
1873 */
1874struct buffered_dirent {
1875 u64 ino;
1876 loff_t offset;
1877 int namlen;
1878 unsigned int d_type;
1879 char name[];
1880};
1881
1882struct readdir_data {
1883 char *dirent;
1884 size_t used;
Al Viro53c9c5c2008-08-24 07:29:52 -04001885 int full;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001886};
1887
1888static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen,
1889 loff_t offset, u64 ino, unsigned int d_type)
David Woodhouse2628b762008-07-31 17:16:51 +01001890{
David Woodhouse14f7dd62008-07-31 20:29:12 +01001891 struct readdir_data *buf = __buf;
1892 struct buffered_dirent *de = (void *)(buf->dirent + buf->used);
1893 unsigned int reclen;
1894
1895 reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64));
Al Viro53c9c5c2008-08-24 07:29:52 -04001896 if (buf->used + reclen > PAGE_SIZE) {
1897 buf->full = 1;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001898 return -EINVAL;
Al Viro53c9c5c2008-08-24 07:29:52 -04001899 }
David Woodhouse14f7dd62008-07-31 20:29:12 +01001900
1901 de->namlen = namlen;
1902 de->offset = offset;
1903 de->ino = ino;
1904 de->d_type = d_type;
1905 memcpy(de->name, name, namlen);
1906 buf->used += reclen;
1907
1908 return 0;
1909}
1910
David Woodhouse2f9092e2009-04-20 23:18:37 +01001911static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func,
1912 struct readdir_cd *cdp, loff_t *offsetp)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001913{
1914 struct readdir_data buf;
1915 struct buffered_dirent *de;
David Woodhouse2628b762008-07-31 17:16:51 +01001916 int host_err;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001917 int size;
1918 loff_t offset;
David Woodhouse2628b762008-07-31 17:16:51 +01001919
David Woodhouse14f7dd62008-07-31 20:29:12 +01001920 buf.dirent = (void *)__get_free_page(GFP_KERNEL);
1921 if (!buf.dirent)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001922 return nfserrno(-ENOMEM);
David Woodhouse2628b762008-07-31 17:16:51 +01001923
David Woodhouse14f7dd62008-07-31 20:29:12 +01001924 offset = *offsetp;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001925
1926 while (1) {
David Woodhouse2f9092e2009-04-20 23:18:37 +01001927 struct inode *dir_inode = file->f_path.dentry->d_inode;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001928 unsigned int reclen;
1929
Doug Nazarb726e922008-11-05 06:16:28 -05001930 cdp->err = nfserr_eof; /* will be cleared on successful read */
David Woodhouse14f7dd62008-07-31 20:29:12 +01001931 buf.used = 0;
Al Viro53c9c5c2008-08-24 07:29:52 -04001932 buf.full = 0;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001933
1934 host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf);
Al Viro53c9c5c2008-08-24 07:29:52 -04001935 if (buf.full)
1936 host_err = 0;
1937
1938 if (host_err < 0)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001939 break;
1940
1941 size = buf.used;
1942
1943 if (!size)
1944 break;
1945
David Woodhouse2f9092e2009-04-20 23:18:37 +01001946 /*
1947 * Various filldir functions may end up calling back into
1948 * lookup_one_len() and the file system's ->lookup() method.
1949 * These expect i_mutex to be held, as it would within readdir.
1950 */
1951 host_err = mutex_lock_killable(&dir_inode->i_mutex);
1952 if (host_err)
1953 break;
1954
David Woodhouse14f7dd62008-07-31 20:29:12 +01001955 de = (struct buffered_dirent *)buf.dirent;
1956 while (size > 0) {
1957 offset = de->offset;
1958
1959 if (func(cdp, de->name, de->namlen, de->offset,
1960 de->ino, de->d_type))
David Woodhouse2f9092e2009-04-20 23:18:37 +01001961 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001962
1963 if (cdp->err != nfs_ok)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001964 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001965
1966 reclen = ALIGN(sizeof(*de) + de->namlen,
1967 sizeof(u64));
1968 size -= reclen;
1969 de = (struct buffered_dirent *)((char *)de + reclen);
1970 }
David Woodhouse2f9092e2009-04-20 23:18:37 +01001971 mutex_unlock(&dir_inode->i_mutex);
1972 if (size > 0) /* We bailed out early */
1973 break;
1974
David Woodhousec002a6c2008-08-17 17:21:18 +01001975 offset = vfs_llseek(file, 0, SEEK_CUR);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001976 }
1977
David Woodhouse14f7dd62008-07-31 20:29:12 +01001978 free_page((unsigned long)(buf.dirent));
David Woodhouse2628b762008-07-31 17:16:51 +01001979
1980 if (host_err)
1981 return nfserrno(host_err);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001982
1983 *offsetp = offset;
1984 return cdp->err;
David Woodhouse2628b762008-07-31 17:16:51 +01001985}
1986
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987/*
1988 * Read entries from a directory.
1989 * The NFSv3/4 verifier we ignore for now.
1990 */
Al Viro6264d692006-10-19 23:28:58 -07001991__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001993 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994{
Al Viro6264d692006-10-19 23:28:58 -07001995 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 struct file *file;
1997 loff_t offset = *offsetp;
1998
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001999 err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 if (err)
2001 goto out;
2002
2003 offset = vfs_llseek(file, offset, 0);
2004 if (offset < 0) {
2005 err = nfserrno((int)offset);
2006 goto out_close;
2007 }
2008
David Woodhouse14f7dd62008-07-31 20:29:12 +01002009 err = nfsd_buffered_readdir(file, func, cdp, offsetp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010
2011 if (err == nfserr_eof || err == nfserr_toosmall)
2012 err = nfs_ok; /* can still be found in ->err */
2013out_close:
2014 nfsd_close(file);
2015out:
2016 return err;
2017}
2018
2019/*
2020 * Get file system stats
2021 * N.B. After this call fhp needs an fh_put
2022 */
Al Viro6264d692006-10-19 23:28:58 -07002023__be32
J. Bruce Fields04716e62008-08-07 13:00:20 -04002024nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025{
Christoph Hellwigebabe9a2010-07-07 18:53:11 +02002026 __be32 err;
2027
2028 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access);
Takashi Iwaif6360ef2010-08-13 15:53:49 +02002029 if (!err) {
2030 struct path path = {
2031 .mnt = fhp->fh_export->ex_path.mnt,
2032 .dentry = fhp->fh_dentry,
2033 };
2034 if (vfs_statfs(&path, stat))
2035 err = nfserr_io;
2036 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 return err;
2038}
2039
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002040static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002041{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002042 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002043}
2044
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045/*
2046 * Check for a user's access permissions to this inode.
2047 */
Al Viro6264d692006-10-19 23:28:58 -07002048__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07002049nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
2050 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051{
2052 struct inode *inode = dentry->d_inode;
2053 int err;
2054
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002055 if (acc == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 return 0;
2057#if 0
2058 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
2059 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002060 (acc & NFSD_MAY_READ)? " read" : "",
2061 (acc & NFSD_MAY_WRITE)? " write" : "",
2062 (acc & NFSD_MAY_EXEC)? " exec" : "",
2063 (acc & NFSD_MAY_SATTR)? " sattr" : "",
2064 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
2065 (acc & NFSD_MAY_LOCK)? " lock" : "",
2066 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067 inode->i_mode,
2068 IS_IMMUTABLE(inode)? " immut" : "",
2069 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08002070 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 dprintk(" owner %d/%d user %d/%d\n",
David Howells5cc0a842008-11-14 10:38:58 +11002072 inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073#endif
2074
2075 /* Normally we reject any write/sattr etc access on a read-only file
2076 * system. But if it is IRIX doing check on write-access for a
2077 * device special file, we ignore rofs.
2078 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002079 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
2080 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08002081 if (exp_rdonly(rqstp, exp) ||
2082 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002084 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 return nfserr_perm;
2086 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002087 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 return nfserr_perm;
2089
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002090 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 /* If we cannot rely on authentication in NLM requests,
2092 * just allow locks, otherwise require read permission, or
2093 * ownership
2094 */
2095 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
2096 return 0;
2097 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002098 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 }
2100 /*
2101 * The file owner always gets access permission for accesses that
2102 * would normally be checked at open time. This is to make
2103 * file access work even when the client has done a fchmod(fd, 0).
2104 *
2105 * However, `cp foo bar' should fail nevertheless when bar is
2106 * readonly. A sensible way to do this might be to reject all
2107 * attempts to truncate a read-only file, because a creat() call
2108 * always implies file truncation.
2109 * ... but this isn't really fair. A process may reasonably call
2110 * ftruncate on an open file descriptor on a file with perm 000.
2111 * We must trust the client to do permission checking - using "ACCESS"
2112 * with NFSv3.
2113 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002114 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
David Howells5cc0a842008-11-14 10:38:58 +11002115 inode->i_uid == current_fsuid())
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 return 0;
2117
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002118 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Al Virof419a2e2008-07-22 00:07:17 -04002119 err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120
2121 /* Allow read access to binaries even when mode 111 */
2122 if (err == -EACCES && S_ISREG(inode->i_mode) &&
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002123 acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
Al Virof419a2e2008-07-22 00:07:17 -04002124 err = inode_permission(inode, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125
2126 return err? nfserrno(err) : 0;
2127}
2128
2129void
2130nfsd_racache_shutdown(void)
2131{
Jeff Layton54a66e52008-08-13 22:03:27 -04002132 struct raparms *raparm, *last_raparm;
2133 unsigned int i;
2134
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135 dprintk("nfsd: freeing readahead buffers.\n");
Jeff Layton54a66e52008-08-13 22:03:27 -04002136
2137 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
2138 raparm = raparm_hash[i].pb_head;
2139 while(raparm) {
2140 last_raparm = raparm;
2141 raparm = raparm->p_next;
2142 kfree(last_raparm);
2143 }
2144 raparm_hash[i].pb_head = NULL;
2145 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146}
2147/*
2148 * Initialize readahead param cache
2149 */
2150int
2151nfsd_racache_init(int cache_size)
2152{
2153 int i;
Greg Banksfce14562006-10-04 02:15:49 -07002154 int j = 0;
2155 int nperbucket;
Jeff Layton54a66e52008-08-13 22:03:27 -04002156 struct raparms **raparm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157
Greg Banksfce14562006-10-04 02:15:49 -07002158
Jeff Layton54a66e52008-08-13 22:03:27 -04002159 if (raparm_hash[0].pb_head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002161 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
2162 if (nperbucket < 2)
2163 nperbucket = 2;
2164 cache_size = nperbucket * RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08002165
2166 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
Jeff Layton54a66e52008-08-13 22:03:27 -04002167
2168 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
Yan Burman4b3bb062006-12-08 02:39:41 -08002169 spin_lock_init(&raparm_hash[i].pb_lock);
Jeff Layton54a66e52008-08-13 22:03:27 -04002170
2171 raparm = &raparm_hash[i].pb_head;
2172 for (j = 0; j < nperbucket; j++) {
2173 *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL);
2174 if (!*raparm)
2175 goto out_nomem;
2176 raparm = &(*raparm)->p_next;
2177 }
2178 *raparm = NULL;
Yan Burman4b3bb062006-12-08 02:39:41 -08002179 }
2180
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 nfsdstats.ra_size = cache_size;
2182 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002183
2184out_nomem:
2185 dprintk("nfsd: kmalloc failed, freeing readahead buffers\n");
2186 nfsd_racache_shutdown();
2187 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002189
2190#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2191struct posix_acl *
2192nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2193{
2194 struct inode *inode = fhp->fh_dentry->d_inode;
2195 char *name;
2196 void *value = NULL;
2197 ssize_t size;
2198 struct posix_acl *acl;
2199
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002200 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002201 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002202
2203 switch (type) {
2204 case ACL_TYPE_ACCESS:
2205 name = POSIX_ACL_XATTR_ACCESS;
2206 break;
2207 case ACL_TYPE_DEFAULT:
2208 name = POSIX_ACL_XATTR_DEFAULT;
2209 break;
2210 default:
2211 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002212 }
2213
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002214 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2215 if (size < 0)
2216 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002217
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002218 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002219 kfree(value);
2220 return acl;
2221}
2222
2223int
2224nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2225{
2226 struct inode *inode = fhp->fh_dentry->d_inode;
2227 char *name;
2228 void *value = NULL;
2229 size_t size;
2230 int error;
2231
Al Viroacfa4382008-12-04 10:06:33 -05002232 if (!IS_POSIXACL(inode) ||
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002233 !inode->i_op->setxattr || !inode->i_op->removexattr)
2234 return -EOPNOTSUPP;
2235 switch(type) {
2236 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002237 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002238 break;
2239 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002240 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002241 break;
2242 default:
2243 return -EOPNOTSUPP;
2244 }
2245
2246 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002247 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002248 value = kmalloc(size, GFP_KERNEL);
2249 if (!value)
2250 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002251 error = posix_acl_to_xattr(acl, value, size);
2252 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002253 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002254 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002255 } else
2256 size = 0;
2257
Dave Hansen18f335a2008-02-15 14:37:38 -08002258 error = mnt_want_write(fhp->fh_export->ex_path.mnt);
2259 if (error)
2260 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002261 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002262 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002263 else {
2264 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2265 error = 0;
2266 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002267 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002268 if (error == -ENODATA)
2269 error = 0;
2270 }
2271 }
Dave Hansen18f335a2008-02-15 14:37:38 -08002272 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002273
2274getout:
2275 kfree(value);
2276 return error;
2277}
2278#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */