blob: 8715d194561aa626bff1d18991a2ac667be00715 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/quotaops.h>
Robert Love0eeca282005-07-12 17:06:03 -040024#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/posix_acl_xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/xattr.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020027#include <linux/jhash.h>
28#include <linux/ima.h>
29#include <asm/uaccess.h>
30
31#ifdef CONFIG_NFSD_V3
32#include "xdr3.h"
33#endif /* CONFIG_NFSD_V3 */
34
Christoph Hellwig5be196e2006-01-09 20:51:55 -080035#ifdef CONFIG_NFSD_V4
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/nfs4_acl.h>
37#include <linux/nfsd_idmap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#endif /* CONFIG_NFSD_V4 */
39
Boaz Harrosh9a74af22009-12-03 20:30:56 +020040#include "nfsd.h"
41#include "vfs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43#define NFSDDBG_FACILITY NFSDDBG_FILEOP
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45
Linus Torvalds1da177e2005-04-16 15:20:36 -070046/*
47 * This is a cache of readahead params that help us choose the proper
48 * readahead strategy. Initially, we set all readahead parameters to 0
49 * and let the VFS handle things.
50 * If you increase the number of cached files very much, you'll need to
51 * add a hash table here.
52 */
53struct raparms {
54 struct raparms *p_next;
55 unsigned int p_count;
56 ino_t p_ino;
57 dev_t p_dev;
58 int p_set;
59 struct file_ra_state p_ra;
Greg Banksfce14562006-10-04 02:15:49 -070060 unsigned int p_hindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061};
62
Greg Banksfce14562006-10-04 02:15:49 -070063struct raparm_hbucket {
64 struct raparms *pb_head;
65 spinlock_t pb_lock;
66} ____cacheline_aligned_in_smp;
67
Greg Banksfce14562006-10-04 02:15:49 -070068#define RAPARM_HASH_BITS 4
69#define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS)
70#define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1)
71static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
73/*
74 * Called from nfsd_lookup and encode_dirent. Check if we have crossed
75 * a mount point.
J.Bruce Fieldse0bb89e2006-12-13 00:35:25 -080076 * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged,
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 * or nfs_ok having possibly changed *dpp and *expp
78 */
79int
80nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp,
81 struct svc_export **expp)
82{
83 struct svc_export *exp = *expp, *exp2 = NULL;
84 struct dentry *dentry = *dpp;
Al Viro91c9fa82009-04-18 02:42:05 -040085 struct path path = {.mnt = mntget(exp->ex_path.mnt),
86 .dentry = dget(dentry)};
Al Viro6264d692006-10-19 23:28:58 -070087 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
Al Viro9393bd02009-04-18 13:58:15 -040089 while (d_mountpoint(path.dentry) && follow_down(&path))
Al Viro91c9fa82009-04-18 02:42:05 -040090 ;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
Al Viro91c9fa82009-04-18 02:42:05 -040092 exp2 = rqst_exp_get_by_name(rqstp, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 if (IS_ERR(exp2)) {
J. Bruce Fields3b6cee7bc2009-10-25 21:18:19 -040094 err = PTR_ERR(exp2);
95 /*
96 * We normally allow NFS clients to continue
97 * "underneath" a mountpoint that is not exported.
98 * The exception is V4ROOT, where no traversal is ever
99 * allowed without an explicit export of the new
100 * directory.
101 */
102 if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT))
103 err = 0;
Al Viro91c9fa82009-04-18 02:42:05 -0400104 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 goto out;
106 }
Steve Dickson3c394dd2009-09-09 15:02:40 -0400107 if (nfsd_v4client(rqstp) ||
108 (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 /* successfully crossed mount point */
Al Viro1644ccc2009-04-18 02:32:31 -0400110 /*
Al Viro91c9fa82009-04-18 02:42:05 -0400111 * This is subtle: path.dentry is *not* on path.mnt
112 * at this point. The only reason we are safe is that
113 * original mnt is pinned down by exp, so we should
114 * put path *before* putting exp
Al Viro1644ccc2009-04-18 02:32:31 -0400115 */
Al Viro91c9fa82009-04-18 02:42:05 -0400116 *dpp = path.dentry;
117 path.dentry = dentry;
Al Viro1644ccc2009-04-18 02:32:31 -0400118 *expp = exp2;
Al Viro91c9fa82009-04-18 02:42:05 -0400119 exp2 = exp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 }
Al Viro91c9fa82009-04-18 02:42:05 -0400121 path_put(&path);
122 exp_put(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123out:
124 return err;
125}
126
J. Bruce Fields289ede42009-09-26 20:32:24 -0400127static void follow_to_parent(struct path *path)
128{
129 struct dentry *dp;
130
131 while (path->dentry == path->mnt->mnt_root && follow_up(path))
132 ;
133 dp = dget_parent(path->dentry);
134 dput(path->dentry);
135 path->dentry = dp;
136}
137
138static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp)
139{
140 struct svc_export *exp2;
141 struct path path = {.mnt = mntget((*exp)->ex_path.mnt),
142 .dentry = dget(dparent)};
143
144 follow_to_parent(&path);
145
146 exp2 = rqst_exp_parent(rqstp, &path);
147 if (PTR_ERR(exp2) == -ENOENT) {
148 *dentryp = dget(dparent);
149 } else if (IS_ERR(exp2)) {
150 path_put(&path);
151 return PTR_ERR(exp2);
152 } else {
153 *dentryp = dget(path.dentry);
154 exp_put(*exp);
155 *exp = exp2;
156 }
157 path_put(&path);
158 return 0;
159}
160
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400161/*
162 * For nfsd purposes, we treat V4ROOT exports as though there was an
163 * export at *every* directory.
164 */
J. Bruce Fields3227fa42009-10-25 21:43:01 -0400165int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp)
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400166{
167 if (d_mountpoint(dentry))
168 return 1;
169 if (!(exp->ex_flags & NFSEXP_V4ROOT))
170 return 0;
171 return dentry->d_inode != NULL;
172}
173
Al Viro6264d692006-10-19 23:28:58 -0700174__be32
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700175nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400176 const char *name, unsigned int len,
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700177 struct svc_export **exp_ret, struct dentry **dentry_ret)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178{
179 struct svc_export *exp;
180 struct dentry *dparent;
181 struct dentry *dentry;
Al Viro6264d692006-10-19 23:28:58 -0700182 __be32 err;
183 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184
185 dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name);
186
187 /* Obtain dentry and export. */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200188 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 if (err)
190 return err;
191
192 dparent = fhp->fh_dentry;
193 exp = fhp->fh_export;
194 exp_get(exp);
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 /* Lookup the name, but don't follow links */
197 if (isdotent(name, len)) {
198 if (len==1)
199 dentry = dget(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800200 else if (dparent != exp->ex_path.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 dentry = dget_parent(dparent);
J. Bruce Fieldsfed83812009-09-26 16:53:01 -0400202 else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 dentry = dget(dparent); /* .. == . just like at / */
204 else {
205 /* checking mountpoint crossing is very different when stepping up */
J. Bruce Fields289ede42009-09-26 20:32:24 -0400206 host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry);
207 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 }
210 } else {
211 fh_lock(fhp);
212 dentry = lookup_one_len(name, dparent, len);
Al Viro6264d692006-10-19 23:28:58 -0700213 host_err = PTR_ERR(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 if (IS_ERR(dentry))
215 goto out_nfserr;
216 /*
217 * check if we have crossed a mount point ...
218 */
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400219 if (nfsd_mountpoint(dentry, exp)) {
Al Viro6264d692006-10-19 23:28:58 -0700220 if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 dput(dentry);
222 goto out_nfserr;
223 }
224 }
225 }
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700226 *dentry_ret = dentry;
227 *exp_ret = exp;
228 return 0;
229
230out_nfserr:
231 exp_put(exp);
232 return nfserrno(host_err);
233}
234
235/*
236 * Look up one component of a pathname.
237 * N.B. After this call _both_ fhp and resfh need an fh_put
238 *
239 * If the lookup would cross a mountpoint, and the mounted filesystem
240 * is exported to the client with NFSEXP_NOHIDE, then the lookup is
241 * accepted as it stands and the mounted directory is
242 * returned. Otherwise the covered directory is returned.
243 * NOTE: this mountpoint crossing is not supported properly by all
244 * clients and is explicitly disallowed for NFSv3
245 * NeilBrown <neilb@cse.unsw.edu.au>
246 */
247__be32
248nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400249 unsigned int len, struct svc_fh *resfh)
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700250{
251 struct svc_export *exp;
252 struct dentry *dentry;
253 __be32 err;
254
255 err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry);
256 if (err)
257 return err;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700258 err = check_nfsd_access(exp, rqstp);
259 if (err)
260 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 /*
262 * Note: we compose the file handle now, but as the
263 * dentry may be negative, it may need to be updated.
264 */
265 err = fh_compose(resfh, exp, dentry, fhp);
266 if (!err && !dentry->d_inode)
267 err = nfserr_noent;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700268out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 exp_put(exp);
271 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272}
273
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275/*
276 * Set various file attributes.
277 * N.B. After this call fhp needs an fh_put
278 */
Al Viro6264d692006-10-19 23:28:58 -0700279__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
281 int check_guard, time_t guardtime)
282{
283 struct dentry *dentry;
284 struct inode *inode;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200285 int accmode = NFSD_MAY_SATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 int ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700287 __be32 err;
288 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 int size_change = 0;
290
291 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200292 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 if (iap->ia_valid & ATTR_SIZE)
294 ftype = S_IFREG;
295
296 /* Get inode */
297 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800298 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 goto out;
300
301 dentry = fhp->fh_dentry;
302 inode = dentry->d_inode;
303
NeilBrown15b7a1b2005-11-07 01:00:23 -0800304 /* Ignore any mode updates on symlinks */
305 if (S_ISLNK(inode->i_mode))
306 iap->ia_valid &= ~ATTR_MODE;
307
308 if (!iap->ia_valid)
309 goto out;
310
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000311 /*
312 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 * which only requires access, and "set-[ac]time-to-X" which
314 * requires ownership.
315 * So if it looks like it might be "set both to the same time which
316 * is close to now", and if inode_change_ok fails, then we
317 * convert to "set to now" instead of "set to explicit time"
318 *
319 * We only call inode_change_ok as the last test as technically
320 * it is not an interface that we should be using. It is only
321 * valid if the filesystem does not define it's own i_op->setattr.
322 */
323#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
324#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000325 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
326 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
327 /*
328 * Looks probable.
329 *
330 * Now just make sure time is in the right ballpark.
331 * Solaris, at least, doesn't seem to care what the time
332 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000334 time_t delta = iap->ia_atime.tv_sec - get_seconds();
335 if (delta < 0)
336 delta = -delta;
337 if (delta < MAX_TOUCH_TIME_ERROR &&
338 inode_change_ok(inode, iap) != 0) {
339 /*
340 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
341 * This will cause notify_change to set these times
342 * to "now"
343 */
344 iap->ia_valid &= ~BOTH_TIME_SET;
345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 }
347
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000348 /*
349 * The size case is special.
350 * It changes the file as well as the attributes.
351 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 if (iap->ia_valid & ATTR_SIZE) {
353 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200354 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
355 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 if (err)
357 goto out;
358 }
359
360 /*
361 * If we are changing the size of the file, then
362 * we need to break all leases.
363 */
Al Viro6264d692006-10-19 23:28:58 -0700364 host_err = break_lease(inode, FMODE_WRITE | O_NONBLOCK);
365 if (host_err == -EWOULDBLOCK)
366 host_err = -ETIMEDOUT;
367 if (host_err) /* ENOMEM or EWOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 goto out_nfserr;
369
Al Viro6264d692006-10-19 23:28:58 -0700370 host_err = get_write_access(inode);
371 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 goto out_nfserr;
373
374 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700375 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
376 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 put_write_access(inode);
378 goto out_nfserr;
379 }
Jan Kara90c0af02009-02-09 22:22:21 +0100380 vfs_dq_init(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 }
382
Jeff Laytonca456252008-04-16 16:28:47 -0400383 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 if (iap->ia_valid & ATTR_MODE) {
385 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400386 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400387 }
388
389 /* Revoke setuid/setgid on chown */
Sachin S. Prabhu0953e622009-02-23 16:22:03 +0000390 if (!S_ISDIR(inode->i_mode) &&
391 (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
392 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) {
Jeff Laytonca456252008-04-16 16:28:47 -0400393 iap->ia_valid |= ATTR_KILL_PRIV;
394 if (iap->ia_valid & ATTR_MODE) {
395 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700396 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400397 if (iap->ia_mode & S_IXGRP)
398 iap->ia_mode &= ~S_ISGID;
399 } else {
400 /* set ATTR_KILL_* bits and let VFS handle it */
401 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700402 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 }
404
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 /* Change the attributes. */
406
407 iap->ia_valid |= ATTR_CTIME;
408
409 err = nfserr_notsync;
410 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
411 fh_lock(fhp);
Al Viro6264d692006-10-19 23:28:58 -0700412 host_err = notify_change(dentry, iap);
413 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 fh_unlock(fhp);
415 }
416 if (size_change)
417 put_write_access(inode);
418 if (!err)
419 if (EX_ISSYNC(fhp->fh_export))
420 write_inode_now(inode, 1);
421out:
422 return err;
423
424out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700425 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 goto out;
427}
428
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800429#if defined(CONFIG_NFSD_V2_ACL) || \
430 defined(CONFIG_NFSD_V3_ACL) || \
431 defined(CONFIG_NFSD_V4)
432static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
433{
434 ssize_t buflen;
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530435 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800437 buflen = vfs_getxattr(dentry, key, NULL, 0);
438 if (buflen <= 0)
439 return buflen;
440
441 *buf = kmalloc(buflen, GFP_KERNEL);
442 if (!*buf)
443 return -ENOMEM;
444
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530445 ret = vfs_getxattr(dentry, key, *buf, buflen);
446 if (ret < 0)
447 kfree(*buf);
448 return ret;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800449}
450#endif
451
452#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453static int
454set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
455{
456 int len;
457 size_t buflen;
458 char *buf = NULL;
459 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
461 buflen = posix_acl_xattr_size(pacl->a_count);
462 buf = kmalloc(buflen, GFP_KERNEL);
463 error = -ENOMEM;
464 if (buf == NULL)
465 goto out;
466
467 len = posix_acl_to_xattr(pacl, buf, buflen);
468 if (len < 0) {
469 error = len;
470 goto out;
471 }
472
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800473 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474out:
475 kfree(buf);
476 return error;
477}
478
Al Viro6264d692006-10-19 23:28:58 -0700479__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
481 struct nfs4_acl *acl)
482{
Al Viro6264d692006-10-19 23:28:58 -0700483 __be32 error;
484 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 struct dentry *dentry;
486 struct inode *inode;
487 struct posix_acl *pacl = NULL, *dpacl = NULL;
488 unsigned int flags = 0;
489
490 /* Get inode */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200491 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700493 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494
495 dentry = fhp->fh_dentry;
496 inode = dentry->d_inode;
497 if (S_ISDIR(inode->i_mode))
498 flags = NFS4_ACL_DIR;
499
Al Viro6264d692006-10-19 23:28:58 -0700500 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
501 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700502 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700503 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 goto out_nfserr;
505
Al Viro6264d692006-10-19 23:28:58 -0700506 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
507 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700508 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700510 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700511 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700513out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 posix_acl_release(pacl);
515 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800517 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700518 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800519 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700520 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521}
522
523static struct posix_acl *
524_get_posix_acl(struct dentry *dentry, char *key)
525{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800526 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800528 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800530 buflen = nfsd_getxattr(dentry, key, &buf);
531 if (!buflen)
532 buflen = -ENODATA;
533 if (buflen <= 0)
534 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535
536 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 kfree(buf);
538 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539}
540
541int
542nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
543{
544 struct inode *inode = dentry->d_inode;
545 int error = 0;
546 struct posix_acl *pacl = NULL, *dpacl = NULL;
547 unsigned int flags = 0;
548
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700549 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
551 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
552 if (IS_ERR(pacl)) {
553 error = PTR_ERR(pacl);
554 pacl = NULL;
555 goto out;
556 }
557
558 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700559 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
561 dpacl = NULL;
562 else if (IS_ERR(dpacl)) {
563 error = PTR_ERR(dpacl);
564 dpacl = NULL;
565 goto out;
566 }
567 flags = NFS4_ACL_DIR;
568 }
569
570 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
571 if (IS_ERR(*acl)) {
572 error = PTR_ERR(*acl);
573 *acl = NULL;
574 }
575 out:
576 posix_acl_release(pacl);
577 posix_acl_release(dpacl);
578 return error;
579}
580
581#endif /* defined(CONFIG_NFS_V4) */
582
583#ifdef CONFIG_NFSD_V3
584/*
585 * Check server access rights to a file system object
586 */
587struct accessmap {
588 u32 access;
589 int how;
590};
591static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200592 { NFS3_ACCESS_READ, NFSD_MAY_READ },
593 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
594 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
595 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596
597 { 0, 0 }
598};
599
600static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200601 { NFS3_ACCESS_READ, NFSD_MAY_READ },
602 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
603 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
604 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
605 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
607 { 0, 0 }
608};
609
610static struct accessmap nfs3_anyaccess[] = {
611 /* Some clients - Solaris 2.6 at least, make an access call
612 * to the server to check for access for things like /dev/null
613 * (which really, the server doesn't care about). So
614 * We provide simple access checking for them, looking
615 * mainly at mode bits, and we make sure to ignore read-only
616 * filesystem checks
617 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200618 { NFS3_ACCESS_READ, NFSD_MAY_READ },
619 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
620 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
621 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623 { 0, 0 }
624};
625
Al Viro6264d692006-10-19 23:28:58 -0700626__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
628{
629 struct accessmap *map;
630 struct svc_export *export;
631 struct dentry *dentry;
632 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700633 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200635 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 if (error)
637 goto out;
638
639 export = fhp->fh_export;
640 dentry = fhp->fh_dentry;
641
642 if (S_ISREG(dentry->d_inode->i_mode))
643 map = nfs3_regaccess;
644 else if (S_ISDIR(dentry->d_inode->i_mode))
645 map = nfs3_diraccess;
646 else
647 map = nfs3_anyaccess;
648
649
650 query = *access;
651 for (; map->access; map++) {
652 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700653 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
655 sresult |= map->access;
656
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700657 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 switch (err2) {
659 case nfs_ok:
660 result |= map->access;
661 break;
662
663 /* the following error codes just mean the access was not allowed,
664 * rather than an error occurred */
665 case nfserr_rofs:
666 case nfserr_acces:
667 case nfserr_perm:
668 /* simply don't "or" in the access bit. */
669 break;
670 default:
671 error = err2;
672 goto out;
673 }
674 }
675 }
676 *access = result;
677 if (supported)
678 *supported = sresult;
679
680 out:
681 return error;
682}
683#endif /* CONFIG_NFSD_V3 */
684
685
686
687/*
688 * Open an existing file or directory.
689 * The access argument indicates the type of open (read/write/lock)
690 * N.B. After this call fhp needs an fh_put
691 */
Al Viro6264d692006-10-19 23:28:58 -0700692__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
694 int access, struct file **filp)
695{
696 struct dentry *dentry;
697 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700698 int flags = O_RDONLY|O_LARGEFILE;
699 __be32 err;
700 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
David Howellse0e81732009-09-02 09:13:40 +0100702 validate_process_creds();
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 /*
705 * If we get here, then the client has already done an "open",
706 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
707 * in case a chmod has now revoked permission.
708 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200709 err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 if (err)
711 goto out;
712
713 dentry = fhp->fh_dentry;
714 inode = dentry->d_inode;
715
716 /* Disallow write access to files with the append-only bit set
717 * or any access when mandatory locking enabled
718 */
719 err = nfserr_perm;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200720 if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400722 /*
723 * We must ignore files (but only files) which might have mandatory
724 * locks on them because there is no way to know if the accesser has
725 * the lock.
726 */
727 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 goto out;
729
730 if (!inode->i_fop)
731 goto out;
732
733 /*
734 * Check to see if there are any leases on this file.
735 * This may block while leases are broken.
736 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200737 host_err = break_lease(inode, O_NONBLOCK | ((access & NFSD_MAY_WRITE) ? FMODE_WRITE : 0));
Al Viro6264d692006-10-19 23:28:58 -0700738 if (host_err == -EWOULDBLOCK)
739 host_err = -ETIMEDOUT;
740 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 goto out_nfserr;
742
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200743 if (access & NFSD_MAY_WRITE) {
744 if (access & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700745 flags = O_RDWR|O_LARGEFILE;
746 else
747 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
Jan Kara90c0af02009-02-09 22:22:21 +0100749 vfs_dq_init(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 }
Jan Blunck54775492008-02-14 19:38:39 -0800751 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
David Howells033a6662009-07-02 14:35:32 +0100752 flags, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700754 host_err = PTR_ERR(*filp);
Chuck Ebbertaeaa5cc2010-02-15 18:07:39 -0500755 else
756 host_err = ima_file_check(*filp, access);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700758 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759out:
David Howellse0e81732009-09-02 09:13:40 +0100760 validate_process_creds();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 return err;
762}
763
764/*
765 * Close a file.
766 */
767void
768nfsd_close(struct file *filp)
769{
770 fput(filp);
771}
772
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500773/*
774 * Sync a file
775 * As this calls fsync (not fdatasync) there is no need for a write_inode
776 * after it.
777 */
778static inline int nfsd_dosync(struct file *filp, struct dentry *dp,
779 const struct file_operations *fop)
780{
781 struct inode *inode = dp->d_inode;
782 int (*fsync) (struct file *, struct dentry *, int);
783 int err;
784
Christoph Hellwig7211a4e2009-12-25 17:44:45 +0100785 err = filemap_write_and_wait(inode->i_mapping);
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500786 if (err == 0 && fop && (fsync = fop->fsync))
787 err = fsync(filp, dp, 0);
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500788 return err;
789}
790
David Shawa334de22006-01-06 00:19:58 -0800791static int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792nfsd_sync(struct file *filp)
793{
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500794 int err;
795 struct inode *inode = filp->f_path.dentry->d_inode;
796 dprintk("nfsd: sync file %s\n", filp->f_path.dentry->d_name.name);
797 mutex_lock(&inode->i_mutex);
798 err=nfsd_dosync(filp, filp->f_path.dentry, filp->f_op);
799 mutex_unlock(&inode->i_mutex);
800
801 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802}
803
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -0800804int
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500805nfsd_sync_dir(struct dentry *dp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500807 return nfsd_dosync(NULL, dp, dp->d_inode->i_fop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808}
809
810/*
811 * Obtain the readahead parameters for the file
812 * specified by (dev, ino).
813 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815static inline struct raparms *
816nfsd_get_raparms(dev_t dev, ino_t ino)
817{
818 struct raparms *ra, **rap, **frap = NULL;
819 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700820 unsigned int hash;
821 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
Greg Banksfce14562006-10-04 02:15:49 -0700823 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
824 rab = &raparm_hash[hash];
825
826 spin_lock(&rab->pb_lock);
827 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 if (ra->p_ino == ino && ra->p_dev == dev)
829 goto found;
830 depth++;
831 if (ra->p_count == 0)
832 frap = rap;
833 }
834 depth = nfsdstats.ra_size*11/10;
835 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700836 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 return NULL;
838 }
839 rap = frap;
840 ra = *frap;
841 ra->p_dev = dev;
842 ra->p_ino = ino;
843 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700844 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845found:
Greg Banksfce14562006-10-04 02:15:49 -0700846 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700848 ra->p_next = rab->pb_head;
849 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 }
851 ra->p_count++;
852 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700853 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 return ra;
855}
856
857/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200858 * Grab and keep cached pages associated with a file in the svc_rqst
859 * so that they can be passed to the network sendmsg/sendpage routines
860 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 */
862static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200863nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
864 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865{
Jens Axboecf8208d2007-06-12 21:22:14 +0200866 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700867 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200868 struct page *page = buf->page;
869 size_t size;
870 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
Jens Axboecac36bb02007-06-14 13:10:48 +0200872 ret = buf->ops->confirm(pipe, buf);
Jens Axboecf8208d2007-06-12 21:22:14 +0200873 if (unlikely(ret))
874 return ret;
875
876 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
878 if (rqstp->rq_res.page_len == 0) {
879 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700880 put_page(*pp);
881 *pp = page;
882 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200883 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700885 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800887 if (*pp)
888 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700889 *pp = page;
890 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700892 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 return size;
896}
897
Jens Axboecf8208d2007-06-12 21:22:14 +0200898static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
899 struct splice_desc *sd)
900{
901 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
902}
903
Dave Hansena8754be2007-10-16 23:31:15 -0700904static inline int svc_msnfs(struct svc_fh *ffhp)
905{
906#ifdef MSNFS
907 return (ffhp->fh_export->ex_flags & NFSEXP_MSNFS);
908#else
909 return 0;
910#endif
911}
912
Al Viro6264d692006-10-19 23:28:58 -0700913static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
915 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
916{
917 struct inode *inode;
918 struct raparms *ra;
919 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700920 __be32 err;
921 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
923 err = nfserr_perm;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800924 inode = file->f_path.dentry->d_inode;
Dave Hansena8754be2007-10-16 23:31:15 -0700925
926 if (svc_msnfs(fhp) && !lock_may_read(inode, offset, *count))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
929 /* Get readahead parameters */
930 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
931
932 if (ra && ra->p_set)
933 file->f_ra = ra->p_ra;
934
Jens Axboecf8208d2007-06-12 21:22:14 +0200935 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
936 struct splice_desc sd = {
937 .len = 0,
938 .total_len = *count,
939 .pos = offset,
940 .u.data = rqstp,
941 };
942
Jens Axboe4fbef202007-07-13 22:42:20 +0200943 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200944 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 } else {
946 oldfs = get_fs();
947 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700948 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 set_fs(oldfs);
950 }
951
952 /* Write back readahead params */
953 if (ra) {
Greg Banksfce14562006-10-04 02:15:49 -0700954 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
955 spin_lock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 ra->p_ra = file->f_ra;
957 ra->p_set = 1;
958 ra->p_count--;
Greg Banksfce14562006-10-04 02:15:49 -0700959 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 }
961
Al Viro6264d692006-10-19 23:28:58 -0700962 if (host_err >= 0) {
963 nfsdstats.io_read += host_err;
964 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 err = 0;
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800966 fsnotify_access(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 } else
Al Viro6264d692006-10-19 23:28:58 -0700968 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969out:
970 return err;
971}
972
Neil Brown9f708e42006-01-06 00:19:59 -0800973static void kill_suid(struct dentry *dentry)
974{
975 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700976 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800977
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800978 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800979 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800980 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800981}
982
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700983/*
984 * Gathered writes: If another process is currently writing to the file,
985 * there's a high chance this is another nfsd (triggered by a bulk write
986 * from a client's biod). Rather than syncing the file with each write
987 * request, we sleep for 10 msec.
988 *
989 * I don't know if this roughly approximates C. Juszak's idea of
990 * gathered writes, but it's a nice and simple solution (IMHO), and it
991 * seems to work:-)
992 *
993 * Note: we do this only in the NFSv2 case, since v3 and higher have a
994 * better tool (separate unstable writes and commits) for solving this
995 * problem.
996 */
997static int wait_for_concurrent_writes(struct file *file)
998{
999 struct inode *inode = file->f_path.dentry->d_inode;
1000 static ino_t last_ino;
1001 static dev_t last_dev;
1002 int err = 0;
1003
1004 if (atomic_read(&inode->i_writecount) > 1
1005 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
1006 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
1007 msleep(10);
1008 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
1009 }
1010
1011 if (inode->i_state & I_DIRTY) {
1012 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
1013 err = nfsd_sync(file);
1014 }
1015 last_ino = inode->i_ino;
1016 last_dev = inode->i_sb->s_dev;
1017 return err;
1018}
1019
Al Viro6264d692006-10-19 23:28:58 -07001020static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1022 loff_t offset, struct kvec *vec, int vlen,
David Shaw31dec252009-03-05 20:16:14 -05001023 unsigned long *cnt, int *stablep)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024{
1025 struct svc_export *exp;
1026 struct dentry *dentry;
1027 struct inode *inode;
1028 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001029 __be32 err = 0;
1030 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 int stable = *stablep;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001032 int use_wgather;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
NeilBrown45bd3b32006-01-18 17:43:50 -08001034#ifdef MSNFS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 err = nfserr_perm;
1036
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
David Shaw31dec252009-03-05 20:16:14 -05001038 (!lock_may_write(file->f_path.dentry->d_inode, offset, *cnt)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 goto out;
1040#endif
1041
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001042 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 inode = dentry->d_inode;
1044 exp = fhp->fh_export;
1045
1046 /*
1047 * Request sync writes if
1048 * - the sync export option has been set, or
1049 * - the client requested O_SYNC behavior (NFSv3 feature).
1050 * - The file system doesn't support fsync().
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001051 * When NFSv2 gathered writes have been configured for this volume,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 * flushing the data to disk is handled separately below.
1053 */
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001054 use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
Harvey Harrison3ba15142008-02-20 12:49:02 -08001056 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 stable = 2;
1058 *stablep = 2; /* FILE_SYNC */
1059 }
1060
1061 if (!EX_ISSYNC(exp))
1062 stable = 0;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001063 if (stable && !use_wgather) {
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001064 spin_lock(&file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 file->f_flags |= O_SYNC;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001066 spin_unlock(&file->f_lock);
1067 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
1069 /* Write the data. */
1070 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001071 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 set_fs(oldfs);
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001073 if (host_err < 0)
1074 goto out_nfserr;
1075 *cnt = host_err;
1076 nfsdstats.io_write += host_err;
1077 fsnotify_modify(file->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
1079 /* clear setuid/setgid flag after write */
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001080 if (inode->i_mode & (S_ISUID | S_ISGID))
Neil Brown9f708e42006-01-06 00:19:59 -08001081 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001083 if (stable && use_wgather)
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001084 host_err = wait_for_concurrent_writes(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001086out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001087 dprintk("nfsd: write complete host_err=%d\n", host_err);
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001088 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 err = 0;
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001090 else
Al Viro6264d692006-10-19 23:28:58 -07001091 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092out:
1093 return err;
1094}
1095
1096/*
1097 * Read data from a file. count must contain the requested read count
1098 * on entry. On return, *count contains the number of bytes actually read.
1099 * N.B. After this call fhp needs an fh_put
1100 */
Al Viro6264d692006-10-19 23:28:58 -07001101__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
1103 loff_t offset, struct kvec *vec, int vlen,
1104 unsigned long *count)
1105{
Al Viro6264d692006-10-19 23:28:58 -07001106 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
1108 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001109 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001110 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 if (err)
1112 goto out;
1113 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1114 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001115 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 if (err)
1117 goto out;
1118 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1119 nfsd_close(file);
1120 }
1121out:
1122 return err;
1123}
1124
1125/*
1126 * Write data to a file.
1127 * The stable flag requests synchronous writes.
1128 * N.B. After this call fhp needs an fh_put
1129 */
Al Viro6264d692006-10-19 23:28:58 -07001130__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
David Shaw31dec252009-03-05 20:16:14 -05001132 loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 int *stablep)
1134{
Al Viro6264d692006-10-19 23:28:58 -07001135 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001138 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001139 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 if (err)
1141 goto out;
1142 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1143 stablep);
1144 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001145 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 if (err)
1147 goto out;
1148
1149 if (cnt)
1150 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1151 cnt, stablep);
1152 nfsd_close(file);
1153 }
1154out:
1155 return err;
1156}
1157
1158#ifdef CONFIG_NFSD_V3
1159/*
1160 * Commit all pending writes to stable storage.
1161 * Strictly speaking, we could sync just the indicated file region here,
1162 * but there's currently no way we can ask the VFS to do so.
1163 *
1164 * Unfortunately we cannot lock the file to make sure we return full WCC
1165 * data to the client, as locking happens lower down in the filesystem.
1166 */
Al Viro6264d692006-10-19 23:28:58 -07001167__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1169 loff_t offset, unsigned long count)
1170{
1171 struct file *file;
Al Viro6264d692006-10-19 23:28:58 -07001172 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173
1174 if ((u64)count > ~(u64)offset)
1175 return nfserr_inval;
1176
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001177 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
1178 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 return err;
1180 if (EX_ISSYNC(fhp->fh_export)) {
1181 if (file->f_op && file->f_op->fsync) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001182 err = nfserrno(nfsd_sync(file));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 } else {
1184 err = nfserr_notsupp;
1185 }
1186 }
1187
1188 nfsd_close(file);
1189 return err;
1190}
1191#endif /* CONFIG_NFSD_V3 */
1192
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001193static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001194nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1195 struct iattr *iap)
1196{
1197 /*
1198 * Mode has already been set earlier in create:
1199 */
1200 iap->ia_valid &= ~ATTR_MODE;
1201 /*
1202 * Setting uid/gid works only for root. Irix appears to
1203 * send along the gid on create when it tries to implement
1204 * setgid directories via NFS:
1205 */
David Howells5cc0a842008-11-14 10:38:58 +11001206 if (current_fsuid() != 0)
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001207 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1208 if (iap->ia_valid)
1209 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1210 return 0;
1211}
1212
wengang wang4ac35c22009-02-10 11:27:51 +08001213/* HPUX client sometimes creates a file in mode 000, and sets size to 0.
1214 * setting size to 0 may fail for some specific file systems by the permission
1215 * checking which requires WRITE permission but the mode is 000.
1216 * we ignore the resizing(to 0) on the just new created file, since the size is
1217 * 0 after file created.
1218 *
1219 * call this only after vfs_create() is called.
1220 * */
1221static void
1222nfsd_check_ignore_resizing(struct iattr *iap)
1223{
1224 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
1225 iap->ia_valid &= ~ATTR_SIZE;
1226}
1227
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228/*
1229 * Create a file (regular, directory, device, fifo); UNIX sockets
1230 * not yet implemented.
1231 * If the response fh has been verified, the parent directory should
1232 * already be locked. Note that the parent directory is left locked.
1233 *
1234 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1235 */
Al Viro6264d692006-10-19 23:28:58 -07001236__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1238 char *fname, int flen, struct iattr *iap,
1239 int type, dev_t rdev, struct svc_fh *resfhp)
1240{
1241 struct dentry *dentry, *dchild = NULL;
1242 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001243 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001244 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001245 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
1247 err = nfserr_perm;
1248 if (!flen)
1249 goto out;
1250 err = nfserr_exist;
1251 if (isdotent(fname, flen))
1252 goto out;
1253
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001254 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 if (err)
1256 goto out;
1257
1258 dentry = fhp->fh_dentry;
1259 dirp = dentry->d_inode;
1260
1261 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001262 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 goto out;
1264 /*
1265 * Check whether the response file handle has been verified yet.
1266 * If it has, the parent directory should already be locked.
1267 */
1268 if (!resfhp->fh_dentry) {
1269 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001270 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001272 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 if (IS_ERR(dchild))
1274 goto out_nfserr;
1275 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1276 if (err)
1277 goto out;
1278 } else {
1279 /* called from nfsd_proc_create */
1280 dchild = dget(resfhp->fh_dentry);
1281 if (!fhp->fh_locked) {
1282 /* not actually possible */
1283 printk(KERN_ERR
1284 "nfsd_create: parent %s/%s not locked!\n",
1285 dentry->d_parent->d_name.name,
1286 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001287 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 goto out;
1289 }
1290 }
1291 /*
1292 * Make sure the child dentry is still negative ...
1293 */
1294 err = nfserr_exist;
1295 if (dchild->d_inode) {
1296 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1297 dentry->d_name.name, dchild->d_name.name);
1298 goto out;
1299 }
1300
1301 if (!(iap->ia_valid & ATTR_MODE))
1302 iap->ia_mode = 0;
1303 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1304
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001305 err = nfserr_inval;
1306 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1307 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1308 type);
1309 goto out;
1310 }
1311
1312 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1313 if (host_err)
1314 goto out_nfserr;
1315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 /*
1317 * Get the dir op function pointer.
1318 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001319 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 switch (type) {
1321 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001322 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
wengang wang4ac35c22009-02-10 11:27:51 +08001323 if (!host_err)
1324 nfsd_check_ignore_resizing(iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 break;
1326 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001327 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 break;
1329 case S_IFCHR:
1330 case S_IFBLK:
1331 case S_IFIFO:
1332 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001333 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001335 }
1336 if (host_err < 0) {
1337 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1338 goto out_nfserr;
1339 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
1341 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001342 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 write_inode_now(dchild->d_inode, 1);
1344 }
1345
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001346 err2 = nfsd_create_setattr(rqstp, resfhp, iap);
1347 if (err2)
1348 err = err2;
Dave Hansen463c3192008-02-15 14:37:57 -08001349 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 /*
1351 * Update the file handle to get the new inode info.
1352 */
1353 if (!err)
1354 err = fh_update(resfhp);
1355out:
1356 if (dchild && !IS_ERR(dchild))
1357 dput(dchild);
1358 return err;
1359
1360out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001361 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 goto out;
1363}
1364
1365#ifdef CONFIG_NFSD_V3
1366/*
1367 * NFSv3 version of nfsd_create
1368 */
Al Viro6264d692006-10-19 23:28:58 -07001369__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp,
1371 char *fname, int flen, struct iattr *iap,
1372 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001373 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374{
1375 struct dentry *dentry, *dchild = NULL;
1376 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001377 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001378 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001379 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381
1382 err = nfserr_perm;
1383 if (!flen)
1384 goto out;
1385 err = nfserr_exist;
1386 if (isdotent(fname, flen))
1387 goto out;
1388 if (!(iap->ia_valid & ATTR_MODE))
1389 iap->ia_mode = 0;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001390 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 if (err)
1392 goto out;
1393
1394 dentry = fhp->fh_dentry;
1395 dirp = dentry->d_inode;
1396
1397 /* Get all the sanity checks out of the way before
1398 * we lock the parent. */
1399 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001400 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001402 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
1404 /*
1405 * Compose the response file handle.
1406 */
1407 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001408 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 if (IS_ERR(dchild))
1410 goto out_nfserr;
1411
1412 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1413 if (err)
1414 goto out;
1415
1416 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001417 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001418 * the high bit set, so just clear the high bits. If this is
1419 * ever changed to use different attrs for storing the
1420 * verifier, then do_open_lookup() will also need to be fixed
1421 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 */
1423 v_mtime = verifier[0]&0x7fffffff;
1424 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 }
1426
Dave Hansen463c3192008-02-15 14:37:57 -08001427 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1428 if (host_err)
1429 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 if (dchild->d_inode) {
1431 err = 0;
1432
1433 switch (createmode) {
1434 case NFS3_CREATE_UNCHECKED:
1435 if (! S_ISREG(dchild->d_inode->i_mode))
1436 err = nfserr_exist;
1437 else if (truncp) {
1438 /* in nfsv4, we need to treat this case a little
1439 * differently. we don't want to truncate the
1440 * file now; this would be wrong if the OPEN
1441 * fails for some other reason. furthermore,
1442 * if the size is nonzero, we should ignore it
1443 * according to spec!
1444 */
1445 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1446 }
1447 else {
1448 iap->ia_valid &= ATTR_SIZE;
1449 goto set_attr;
1450 }
1451 break;
1452 case NFS3_CREATE_EXCLUSIVE:
1453 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1454 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 && dchild->d_inode->i_size == 0 )
1456 break;
1457 /* fallthru */
1458 case NFS3_CREATE_GUARDED:
1459 err = nfserr_exist;
1460 }
Dave Hansen463c3192008-02-15 14:37:57 -08001461 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 goto out;
1463 }
1464
Al Viro6264d692006-10-19 23:28:58 -07001465 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001466 if (host_err < 0) {
1467 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001469 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001470 if (created)
1471 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
1473 if (EX_ISSYNC(fhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001474 err = nfserrno(nfsd_sync_dir(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 /* setattr will sync the child (or not) */
1476 }
1477
wengang wang4ac35c22009-02-10 11:27:51 +08001478 nfsd_check_ignore_resizing(iap);
1479
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 if (createmode == NFS3_CREATE_EXCLUSIVE) {
Peter Staubachc3978522007-01-26 00:57:00 -08001481 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001483 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 /* XXX someone who knows this better please fix it for nsec */
1485 iap->ia_mtime.tv_sec = v_mtime;
1486 iap->ia_atime.tv_sec = v_atime;
1487 iap->ia_mtime.tv_nsec = 0;
1488 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 }
1490
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 set_attr:
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001492 err2 = nfsd_create_setattr(rqstp, resfhp, iap);
1493 if (err2)
1494 err = err2;
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001495
Dave Hansen463c3192008-02-15 14:37:57 -08001496 mnt_drop_write(fhp->fh_export->ex_path.mnt);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001497 /*
1498 * Update the filehandle to get the new inode info.
1499 */
1500 if (!err)
1501 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
1503 out:
1504 fh_unlock(fhp);
1505 if (dchild && !IS_ERR(dchild))
1506 dput(dchild);
1507 return err;
1508
1509 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001510 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 goto out;
1512}
1513#endif /* CONFIG_NFSD_V3 */
1514
1515/*
1516 * Read a symlink. On entry, *lenp must contain the maximum path length that
1517 * fits into the buffer. On return, it contains the true length.
1518 * N.B. After this call fhp needs an fh_put
1519 */
Al Viro6264d692006-10-19 23:28:58 -07001520__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1522{
1523 struct dentry *dentry;
1524 struct inode *inode;
1525 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001526 __be32 err;
1527 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001529 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 if (err)
1531 goto out;
1532
1533 dentry = fhp->fh_dentry;
1534 inode = dentry->d_inode;
1535
1536 err = nfserr_inval;
Al Viroacfa4382008-12-04 10:06:33 -05001537 if (!inode->i_op->readlink)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 goto out;
1539
Jan Blunck54775492008-02-14 19:38:39 -08001540 touch_atime(fhp->fh_export->ex_path.mnt, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 /* N.B. Why does this call need a get_fs()??
1542 * Remove the set_fs and watch the fireworks:-) --okir
1543 */
1544
1545 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001546 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 set_fs(oldfs);
1548
Al Viro6264d692006-10-19 23:28:58 -07001549 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001551 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 err = 0;
1553out:
1554 return err;
1555
1556out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001557 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 goto out;
1559}
1560
1561/*
1562 * Create a symlink and look up its inode
1563 * N.B. After this call _both_ fhp and resfhp need an fh_put
1564 */
Al Viro6264d692006-10-19 23:28:58 -07001565__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1567 char *fname, int flen,
1568 char *path, int plen,
1569 struct svc_fh *resfhp,
1570 struct iattr *iap)
1571{
1572 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001573 __be32 err, cerr;
1574 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575
1576 err = nfserr_noent;
1577 if (!flen || !plen)
1578 goto out;
1579 err = nfserr_exist;
1580 if (isdotent(fname, flen))
1581 goto out;
1582
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001583 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 if (err)
1585 goto out;
1586 fh_lock(fhp);
1587 dentry = fhp->fh_dentry;
1588 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001589 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 if (IS_ERR(dnew))
1591 goto out_nfserr;
1592
Dave Hansen75c3f292008-02-15 14:37:45 -08001593 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1594 if (host_err)
1595 goto out_nfserr;
1596
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 if (unlikely(path[plen] != 0)) {
1598 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1599 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001600 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 else {
1602 strncpy(path_alloced, path, plen);
1603 path_alloced[plen] = 0;
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001604 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 kfree(path_alloced);
1606 }
1607 } else
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001608 host_err = vfs_symlink(dentry->d_inode, dnew, path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
Al Viro6264d692006-10-19 23:28:58 -07001610 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 if (EX_ISSYNC(fhp->fh_export))
Al Viro6264d692006-10-19 23:28:58 -07001612 host_err = nfsd_sync_dir(dentry);
1613 }
1614 err = nfserrno(host_err);
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;
1639 struct inode *dirp, *dest;
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;
1667 dest = dold->d_inode;
1668
Dave Hansen75c3f292008-02-15 14:37:45 -08001669 host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt);
1670 if (host_err) {
1671 err = nfserrno(host_err);
1672 goto out_dput;
1673 }
Al Viro6264d692006-10-19 23:28:58 -07001674 host_err = vfs_link(dold, dirp, dnew);
1675 if (!host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 if (EX_ISSYNC(ffhp->fh_export)) {
NeilBrown45bd3b32006-01-18 17:43:50 -08001677 err = nfserrno(nfsd_sync_dir(ddir));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 write_inode_now(dest, 1);
1679 }
Al Viro6264d692006-10-19 23:28:58 -07001680 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 } else {
Al Viro6264d692006-10-19 23:28:58 -07001682 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 err = nfserr_acces;
1684 else
Al Viro6264d692006-10-19 23:28:58 -07001685 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 }
Dave Hansen75c3f292008-02-15 14:37:45 -08001687 mnt_drop_write(tfhp->fh_export->ex_path.mnt);
1688out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001690out_unlock:
1691 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692out:
1693 return err;
1694
1695out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001696 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001697 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698}
1699
1700/*
1701 * Rename a file
1702 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1703 */
Al Viro6264d692006-10-19 23:28:58 -07001704__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1706 struct svc_fh *tfhp, char *tname, int tlen)
1707{
1708 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1709 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001710 __be32 err;
1711 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001713 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (err)
1715 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001716 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 if (err)
1718 goto out;
1719
1720 fdentry = ffhp->fh_dentry;
1721 fdir = fdentry->d_inode;
1722
1723 tdentry = tfhp->fh_dentry;
1724 tdir = tdentry->d_inode;
1725
1726 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001727 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 goto out;
1729
1730 err = nfserr_perm;
1731 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1732 goto out;
1733
1734 /* cannot use fh_lock as we need deadlock protective ordering
1735 * so do it by hand */
1736 trap = lock_rename(tdentry, fdentry);
1737 ffhp->fh_locked = tfhp->fh_locked = 1;
1738 fill_pre_wcc(ffhp);
1739 fill_pre_wcc(tfhp);
1740
1741 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001742 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (IS_ERR(odentry))
1744 goto out_nfserr;
1745
Al Viro6264d692006-10-19 23:28:58 -07001746 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 if (!odentry->d_inode)
1748 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001749 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (odentry == trap)
1751 goto out_dput_old;
1752
1753 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001754 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 if (IS_ERR(ndentry))
1756 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001757 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 if (ndentry == trap)
1759 goto out_dput_new;
1760
Dave Hansen9079b1e2008-02-15 14:37:49 -08001761 if (svc_msnfs(ffhp) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 ((atomic_read(&odentry->d_count) > 1)
1763 || (atomic_read(&ndentry->d_count) > 1))) {
Al Viro6264d692006-10-19 23:28:58 -07001764 host_err = -EPERM;
Dave Hansen9079b1e2008-02-15 14:37:49 -08001765 goto out_dput_new;
1766 }
1767
1768 host_err = -EXDEV;
1769 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1770 goto out_dput_new;
1771 host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt);
1772 if (host_err)
1773 goto out_dput_new;
1774
Al Viro6264d692006-10-19 23:28:58 -07001775 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
1776 if (!host_err && EX_ISSYNC(tfhp->fh_export)) {
1777 host_err = nfsd_sync_dir(tdentry);
1778 if (!host_err)
1779 host_err = nfsd_sync_dir(fdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 }
1781
Dave Hansen9079b1e2008-02-15 14:37:49 -08001782 mnt_drop_write(ffhp->fh_export->ex_path.mnt);
1783
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 out_dput_new:
1785 dput(ndentry);
1786 out_dput_old:
1787 dput(odentry);
1788 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001789 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
1791 /* we cannot reply on fh_unlock on the two filehandles,
1792 * as that would do the wrong thing if the two directories
1793 * were the same, so again we do it by hand
1794 */
1795 fill_post_wcc(ffhp);
1796 fill_post_wcc(tfhp);
1797 unlock_rename(tdentry, fdentry);
1798 ffhp->fh_locked = tfhp->fh_locked = 0;
1799
1800out:
1801 return err;
1802}
1803
1804/*
1805 * Unlink a file or directory
1806 * N.B. After this call fhp needs an fh_put
1807 */
Al Viro6264d692006-10-19 23:28:58 -07001808__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1810 char *fname, int flen)
1811{
1812 struct dentry *dentry, *rdentry;
1813 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001814 __be32 err;
1815 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816
1817 err = nfserr_acces;
1818 if (!flen || isdotent(fname, flen))
1819 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001820 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 if (err)
1822 goto out;
1823
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001824 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 dentry = fhp->fh_dentry;
1826 dirp = dentry->d_inode;
1827
1828 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001829 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 if (IS_ERR(rdentry))
1831 goto out_nfserr;
1832
1833 if (!rdentry->d_inode) {
1834 dput(rdentry);
1835 err = nfserr_noent;
1836 goto out;
1837 }
1838
1839 if (!type)
1840 type = rdentry->d_inode->i_mode & S_IFMT;
1841
Dave Hansen06227532008-02-15 14:37:34 -08001842 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1843 if (host_err)
1844 goto out_nfserr;
1845
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 if (type != S_IFDIR) { /* It's UNLINK */
1847#ifdef MSNFS
1848 if ((fhp->fh_export->ex_flags & NFSEXP_MSNFS) &&
1849 (atomic_read(&rdentry->d_count) > 1)) {
Al Viro6264d692006-10-19 23:28:58 -07001850 host_err = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 } else
1852#endif
Al Viro6264d692006-10-19 23:28:58 -07001853 host_err = vfs_unlink(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 } else { /* It's RMDIR */
Al Viro6264d692006-10-19 23:28:58 -07001855 host_err = vfs_rmdir(dirp, rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 }
1857
1858 dput(rdentry);
1859
Al Viro6264d692006-10-19 23:28:58 -07001860 if (host_err)
Dave Hansen06227532008-02-15 14:37:34 -08001861 goto out_drop;
Al Viro6264d692006-10-19 23:28:58 -07001862 if (EX_ISSYNC(fhp->fh_export))
1863 host_err = nfsd_sync_dir(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864
Dave Hansen06227532008-02-15 14:37:34 -08001865out_drop:
1866 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001868 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001869out:
1870 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871}
1872
1873/*
David Woodhouse14f7dd62008-07-31 20:29:12 +01001874 * We do this buffering because we must not call back into the file
1875 * system's ->lookup() method from the filldir callback. That may well
1876 * deadlock a number of file systems.
1877 *
1878 * This is based heavily on the implementation of same in XFS.
1879 */
1880struct buffered_dirent {
1881 u64 ino;
1882 loff_t offset;
1883 int namlen;
1884 unsigned int d_type;
1885 char name[];
1886};
1887
1888struct readdir_data {
1889 char *dirent;
1890 size_t used;
Al Viro53c9c5c2008-08-24 07:29:52 -04001891 int full;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001892};
1893
1894static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen,
1895 loff_t offset, u64 ino, unsigned int d_type)
David Woodhouse2628b762008-07-31 17:16:51 +01001896{
David Woodhouse14f7dd62008-07-31 20:29:12 +01001897 struct readdir_data *buf = __buf;
1898 struct buffered_dirent *de = (void *)(buf->dirent + buf->used);
1899 unsigned int reclen;
1900
1901 reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64));
Al Viro53c9c5c2008-08-24 07:29:52 -04001902 if (buf->used + reclen > PAGE_SIZE) {
1903 buf->full = 1;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001904 return -EINVAL;
Al Viro53c9c5c2008-08-24 07:29:52 -04001905 }
David Woodhouse14f7dd62008-07-31 20:29:12 +01001906
1907 de->namlen = namlen;
1908 de->offset = offset;
1909 de->ino = ino;
1910 de->d_type = d_type;
1911 memcpy(de->name, name, namlen);
1912 buf->used += reclen;
1913
1914 return 0;
1915}
1916
David Woodhouse2f9092e2009-04-20 23:18:37 +01001917static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func,
1918 struct readdir_cd *cdp, loff_t *offsetp)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001919{
1920 struct readdir_data buf;
1921 struct buffered_dirent *de;
David Woodhouse2628b762008-07-31 17:16:51 +01001922 int host_err;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001923 int size;
1924 loff_t offset;
David Woodhouse2628b762008-07-31 17:16:51 +01001925
David Woodhouse14f7dd62008-07-31 20:29:12 +01001926 buf.dirent = (void *)__get_free_page(GFP_KERNEL);
1927 if (!buf.dirent)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001928 return nfserrno(-ENOMEM);
David Woodhouse2628b762008-07-31 17:16:51 +01001929
David Woodhouse14f7dd62008-07-31 20:29:12 +01001930 offset = *offsetp;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001931
1932 while (1) {
David Woodhouse2f9092e2009-04-20 23:18:37 +01001933 struct inode *dir_inode = file->f_path.dentry->d_inode;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001934 unsigned int reclen;
1935
Doug Nazarb726e922008-11-05 06:16:28 -05001936 cdp->err = nfserr_eof; /* will be cleared on successful read */
David Woodhouse14f7dd62008-07-31 20:29:12 +01001937 buf.used = 0;
Al Viro53c9c5c2008-08-24 07:29:52 -04001938 buf.full = 0;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001939
1940 host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf);
Al Viro53c9c5c2008-08-24 07:29:52 -04001941 if (buf.full)
1942 host_err = 0;
1943
1944 if (host_err < 0)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001945 break;
1946
1947 size = buf.used;
1948
1949 if (!size)
1950 break;
1951
David Woodhouse2f9092e2009-04-20 23:18:37 +01001952 /*
1953 * Various filldir functions may end up calling back into
1954 * lookup_one_len() and the file system's ->lookup() method.
1955 * These expect i_mutex to be held, as it would within readdir.
1956 */
1957 host_err = mutex_lock_killable(&dir_inode->i_mutex);
1958 if (host_err)
1959 break;
1960
David Woodhouse14f7dd62008-07-31 20:29:12 +01001961 de = (struct buffered_dirent *)buf.dirent;
1962 while (size > 0) {
1963 offset = de->offset;
1964
1965 if (func(cdp, de->name, de->namlen, de->offset,
1966 de->ino, de->d_type))
David Woodhouse2f9092e2009-04-20 23:18:37 +01001967 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001968
1969 if (cdp->err != nfs_ok)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001970 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001971
1972 reclen = ALIGN(sizeof(*de) + de->namlen,
1973 sizeof(u64));
1974 size -= reclen;
1975 de = (struct buffered_dirent *)((char *)de + reclen);
1976 }
David Woodhouse2f9092e2009-04-20 23:18:37 +01001977 mutex_unlock(&dir_inode->i_mutex);
1978 if (size > 0) /* We bailed out early */
1979 break;
1980
David Woodhousec002a6c2008-08-17 17:21:18 +01001981 offset = vfs_llseek(file, 0, SEEK_CUR);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001982 }
1983
David Woodhouse14f7dd62008-07-31 20:29:12 +01001984 free_page((unsigned long)(buf.dirent));
David Woodhouse2628b762008-07-31 17:16:51 +01001985
1986 if (host_err)
1987 return nfserrno(host_err);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001988
1989 *offsetp = offset;
1990 return cdp->err;
David Woodhouse2628b762008-07-31 17:16:51 +01001991}
1992
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993/*
1994 * Read entries from a directory.
1995 * The NFSv3/4 verifier we ignore for now.
1996 */
Al Viro6264d692006-10-19 23:28:58 -07001997__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001999 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000{
Al Viro6264d692006-10-19 23:28:58 -07002001 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 struct file *file;
2003 loff_t offset = *offsetp;
2004
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002005 err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 if (err)
2007 goto out;
2008
2009 offset = vfs_llseek(file, offset, 0);
2010 if (offset < 0) {
2011 err = nfserrno((int)offset);
2012 goto out_close;
2013 }
2014
David Woodhouse14f7dd62008-07-31 20:29:12 +01002015 err = nfsd_buffered_readdir(file, func, cdp, offsetp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017 if (err == nfserr_eof || err == nfserr_toosmall)
2018 err = nfs_ok; /* can still be found in ->err */
2019out_close:
2020 nfsd_close(file);
2021out:
2022 return err;
2023}
2024
2025/*
2026 * Get file system stats
2027 * N.B. After this call fhp needs an fh_put
2028 */
Al Viro6264d692006-10-19 23:28:58 -07002029__be32
J. Bruce Fields04716e62008-08-07 13:00:20 -04002030nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031{
J. Bruce Fields04716e62008-08-07 13:00:20 -04002032 __be32 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access);
David Howells726c3342006-06-23 02:02:58 -07002033 if (!err && vfs_statfs(fhp->fh_dentry,stat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 err = nfserr_io;
2035 return err;
2036}
2037
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002038static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002039{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002040 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002041}
2042
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043/*
2044 * Check for a user's access permissions to this inode.
2045 */
Al Viro6264d692006-10-19 23:28:58 -07002046__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07002047nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
2048 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049{
2050 struct inode *inode = dentry->d_inode;
Mimi Zohar14dba532009-05-27 09:31:52 -04002051 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 int err;
2053
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002054 if (acc == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 return 0;
2056#if 0
2057 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
2058 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002059 (acc & NFSD_MAY_READ)? " read" : "",
2060 (acc & NFSD_MAY_WRITE)? " write" : "",
2061 (acc & NFSD_MAY_EXEC)? " exec" : "",
2062 (acc & NFSD_MAY_SATTR)? " sattr" : "",
2063 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
2064 (acc & NFSD_MAY_LOCK)? " lock" : "",
2065 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 inode->i_mode,
2067 IS_IMMUTABLE(inode)? " immut" : "",
2068 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08002069 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 dprintk(" owner %d/%d user %d/%d\n",
David Howells5cc0a842008-11-14 10:38:58 +11002071 inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072#endif
2073
2074 /* Normally we reject any write/sattr etc access on a read-only file
2075 * system. But if it is IRIX doing check on write-access for a
2076 * device special file, we ignore rofs.
2077 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002078 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
2079 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08002080 if (exp_rdonly(rqstp, exp) ||
2081 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002083 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 return nfserr_perm;
2085 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002086 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087 return nfserr_perm;
2088
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002089 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 /* If we cannot rely on authentication in NLM requests,
2091 * just allow locks, otherwise require read permission, or
2092 * ownership
2093 */
2094 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
2095 return 0;
2096 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002097 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098 }
2099 /*
2100 * The file owner always gets access permission for accesses that
2101 * would normally be checked at open time. This is to make
2102 * file access work even when the client has done a fchmod(fd, 0).
2103 *
2104 * However, `cp foo bar' should fail nevertheless when bar is
2105 * readonly. A sensible way to do this might be to reject all
2106 * attempts to truncate a read-only file, because a creat() call
2107 * always implies file truncation.
2108 * ... but this isn't really fair. A process may reasonably call
2109 * ftruncate on an open file descriptor on a file with perm 000.
2110 * We must trust the client to do permission checking - using "ACCESS"
2111 * with NFSv3.
2112 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002113 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
David Howells5cc0a842008-11-14 10:38:58 +11002114 inode->i_uid == current_fsuid())
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 return 0;
2116
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002117 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Al Virof419a2e2008-07-22 00:07:17 -04002118 err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119
2120 /* Allow read access to binaries even when mode 111 */
2121 if (err == -EACCES && S_ISREG(inode->i_mode) &&
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002122 acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
Al Virof419a2e2008-07-22 00:07:17 -04002123 err = inode_permission(inode, MAY_EXEC);
Mimi Zohar14dba532009-05-27 09:31:52 -04002124 if (err)
2125 goto nfsd_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126
Mimi Zohar14dba532009-05-27 09:31:52 -04002127 /* Do integrity (permission) checking now, but defer incrementing
2128 * IMA counts to the actual file open.
2129 */
2130 path.mnt = exp->ex_path.mnt;
2131 path.dentry = dentry;
Mimi Zohar14dba532009-05-27 09:31:52 -04002132nfsd_out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 return err? nfserrno(err) : 0;
2134}
2135
2136void
2137nfsd_racache_shutdown(void)
2138{
Jeff Layton54a66e52008-08-13 22:03:27 -04002139 struct raparms *raparm, *last_raparm;
2140 unsigned int i;
2141
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 dprintk("nfsd: freeing readahead buffers.\n");
Jeff Layton54a66e52008-08-13 22:03:27 -04002143
2144 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
2145 raparm = raparm_hash[i].pb_head;
2146 while(raparm) {
2147 last_raparm = raparm;
2148 raparm = raparm->p_next;
2149 kfree(last_raparm);
2150 }
2151 raparm_hash[i].pb_head = NULL;
2152 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153}
2154/*
2155 * Initialize readahead param cache
2156 */
2157int
2158nfsd_racache_init(int cache_size)
2159{
2160 int i;
Greg Banksfce14562006-10-04 02:15:49 -07002161 int j = 0;
2162 int nperbucket;
Jeff Layton54a66e52008-08-13 22:03:27 -04002163 struct raparms **raparm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164
Greg Banksfce14562006-10-04 02:15:49 -07002165
Jeff Layton54a66e52008-08-13 22:03:27 -04002166 if (raparm_hash[0].pb_head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002168 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
2169 if (nperbucket < 2)
2170 nperbucket = 2;
2171 cache_size = nperbucket * RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08002172
2173 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
Jeff Layton54a66e52008-08-13 22:03:27 -04002174
2175 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
Yan Burman4b3bb062006-12-08 02:39:41 -08002176 spin_lock_init(&raparm_hash[i].pb_lock);
Jeff Layton54a66e52008-08-13 22:03:27 -04002177
2178 raparm = &raparm_hash[i].pb_head;
2179 for (j = 0; j < nperbucket; j++) {
2180 *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL);
2181 if (!*raparm)
2182 goto out_nomem;
2183 raparm = &(*raparm)->p_next;
2184 }
2185 *raparm = NULL;
Yan Burman4b3bb062006-12-08 02:39:41 -08002186 }
2187
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 nfsdstats.ra_size = cache_size;
2189 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002190
2191out_nomem:
2192 dprintk("nfsd: kmalloc failed, freeing readahead buffers\n");
2193 nfsd_racache_shutdown();
2194 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002196
2197#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2198struct posix_acl *
2199nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2200{
2201 struct inode *inode = fhp->fh_dentry->d_inode;
2202 char *name;
2203 void *value = NULL;
2204 ssize_t size;
2205 struct posix_acl *acl;
2206
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002207 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002208 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002209
2210 switch (type) {
2211 case ACL_TYPE_ACCESS:
2212 name = POSIX_ACL_XATTR_ACCESS;
2213 break;
2214 case ACL_TYPE_DEFAULT:
2215 name = POSIX_ACL_XATTR_DEFAULT;
2216 break;
2217 default:
2218 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002219 }
2220
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002221 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2222 if (size < 0)
2223 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002224
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002225 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002226 kfree(value);
2227 return acl;
2228}
2229
2230int
2231nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2232{
2233 struct inode *inode = fhp->fh_dentry->d_inode;
2234 char *name;
2235 void *value = NULL;
2236 size_t size;
2237 int error;
2238
Al Viroacfa4382008-12-04 10:06:33 -05002239 if (!IS_POSIXACL(inode) ||
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002240 !inode->i_op->setxattr || !inode->i_op->removexattr)
2241 return -EOPNOTSUPP;
2242 switch(type) {
2243 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002244 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002245 break;
2246 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002247 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002248 break;
2249 default:
2250 return -EOPNOTSUPP;
2251 }
2252
2253 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002254 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002255 value = kmalloc(size, GFP_KERNEL);
2256 if (!value)
2257 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002258 error = posix_acl_to_xattr(acl, value, size);
2259 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002260 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002261 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002262 } else
2263 size = 0;
2264
Dave Hansen18f335a2008-02-15 14:37:38 -08002265 error = mnt_want_write(fhp->fh_export->ex_path.mnt);
2266 if (error)
2267 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002268 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002269 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002270 else {
2271 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2272 error = 0;
2273 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002274 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002275 if (error == -ENODATA)
2276 error = 0;
2277 }
2278 }
Dave Hansen18f335a2008-02-15 14:37:38 -08002279 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002280
2281getout:
2282 kfree(value);
2283 return error;
2284}
2285#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */