blob: fd0acca5370a5da699fe62a5d94d0b6e81f67b78 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * File operations used by nfsd. Some of these have been ripped from
3 * other parts of the kernel because they weren't exported, others
4 * are partial duplicates with added or changed functionality.
5 *
6 * Note that several functions dget() the dentry upon which they want
7 * to act, most notably those that create directory entries. Response
8 * dentry's are dput()'d if necessary in the release callback.
9 * So if you notice code paths that apparently fail to dput() the
10 * dentry, don't worry--they have been taken care of.
11 *
12 * Copyright (C) 1995-1999 Olaf Kirch <okir@monad.swb.de>
13 * Zerocpy NFS support (C) 2002 Hirokazu Takahashi <taka@valinux.co.jp>
14 */
15
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/fs.h>
17#include <linux/file.h>
Jens Axboed6b29d72007-06-04 09:59:47 +020018#include <linux/splice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/fcntl.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/namei.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/delay.h>
Robert Love0eeca282005-07-12 17:06:03 -040022#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/posix_acl_xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/xattr.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020025#include <linux/jhash.h>
26#include <linux/ima.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090027#include <linux/slab.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020028#include <asm/uaccess.h>
Ben Myersf5019122010-02-17 14:05:11 -060029#include <linux/exportfs.h>
30#include <linux/writeback.h>
Boaz Harrosh9a74af22009-12-03 20:30:56 +020031
32#ifdef CONFIG_NFSD_V3
33#include "xdr3.h"
34#endif /* CONFIG_NFSD_V3 */
35
Christoph Hellwig5be196e2006-01-09 20:51:55 -080036#ifdef CONFIG_NFSD_V4
J. Bruce Fields2ca72e12011-01-04 17:37:15 -050037#include "acl.h"
38#include "idmap.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#endif /* CONFIG_NFSD_V4 */
40
Boaz Harrosh9a74af22009-12-03 20:30:56 +020041#include "nfsd.h"
42#include "vfs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44#define NFSDDBG_FACILITY NFSDDBG_FILEOP
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
46
Linus Torvalds1da177e2005-04-16 15:20:36 -070047/*
48 * This is a cache of readahead params that help us choose the proper
49 * readahead strategy. Initially, we set all readahead parameters to 0
50 * and let the VFS handle things.
51 * If you increase the number of cached files very much, you'll need to
52 * add a hash table here.
53 */
54struct raparms {
55 struct raparms *p_next;
56 unsigned int p_count;
57 ino_t p_ino;
58 dev_t p_dev;
59 int p_set;
60 struct file_ra_state p_ra;
Greg Banksfce14562006-10-04 02:15:49 -070061 unsigned int p_hindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -070062};
63
Greg Banksfce14562006-10-04 02:15:49 -070064struct raparm_hbucket {
65 struct raparms *pb_head;
66 spinlock_t pb_lock;
67} ____cacheline_aligned_in_smp;
68
Greg Banksfce14562006-10-04 02:15:49 -070069#define RAPARM_HASH_BITS 4
70#define RAPARM_HASH_SIZE (1<<RAPARM_HASH_BITS)
71#define RAPARM_HASH_MASK (RAPARM_HASH_SIZE-1)
72static struct raparm_hbucket raparm_hash[RAPARM_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74/*
75 * Called from nfsd_lookup and encode_dirent. Check if we have crossed
76 * a mount point.
J.Bruce Fieldse0bb89e2006-12-13 00:35:25 -080077 * Returns -EAGAIN or -ETIMEDOUT leaving *dpp and *expp unchanged,
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 * or nfs_ok having possibly changed *dpp and *expp
79 */
80int
81nfsd_cross_mnt(struct svc_rqst *rqstp, struct dentry **dpp,
82 struct svc_export **expp)
83{
84 struct svc_export *exp = *expp, *exp2 = NULL;
85 struct dentry *dentry = *dpp;
Al Viro91c9fa82009-04-18 02:42:05 -040086 struct path path = {.mnt = mntget(exp->ex_path.mnt),
87 .dentry = dget(dentry)};
Al Viro6264d692006-10-19 23:28:58 -070088 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Al Viro7cc90cc2011-03-18 09:04:20 -040090 err = follow_down(&path);
David Howellscc53ce52011-01-14 18:45:26 +000091 if (err < 0)
92 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Al Viro91c9fa82009-04-18 02:42:05 -040094 exp2 = rqst_exp_get_by_name(rqstp, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095 if (IS_ERR(exp2)) {
J. Bruce Fields3b6cee7bc2009-10-25 21:18:19 -040096 err = PTR_ERR(exp2);
97 /*
98 * We normally allow NFS clients to continue
99 * "underneath" a mountpoint that is not exported.
100 * The exception is V4ROOT, where no traversal is ever
101 * allowed without an explicit export of the new
102 * directory.
103 */
104 if (err == -ENOENT && !(exp->ex_flags & NFSEXP_V4ROOT))
105 err = 0;
Al Viro91c9fa82009-04-18 02:42:05 -0400106 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 goto out;
108 }
Steve Dickson3c394dd2009-09-09 15:02:40 -0400109 if (nfsd_v4client(rqstp) ||
110 (exp->ex_flags & NFSEXP_CROSSMOUNT) || EX_NOHIDE(exp2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 /* successfully crossed mount point */
Al Viro1644ccc2009-04-18 02:32:31 -0400112 /*
Al Viro91c9fa82009-04-18 02:42:05 -0400113 * This is subtle: path.dentry is *not* on path.mnt
114 * at this point. The only reason we are safe is that
115 * original mnt is pinned down by exp, so we should
116 * put path *before* putting exp
Al Viro1644ccc2009-04-18 02:32:31 -0400117 */
Al Viro91c9fa82009-04-18 02:42:05 -0400118 *dpp = path.dentry;
119 path.dentry = dentry;
Al Viro1644ccc2009-04-18 02:32:31 -0400120 *expp = exp2;
Al Viro91c9fa82009-04-18 02:42:05 -0400121 exp2 = exp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 }
Al Viro91c9fa82009-04-18 02:42:05 -0400123 path_put(&path);
124 exp_put(exp2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125out:
126 return err;
127}
128
J. Bruce Fields289ede42009-09-26 20:32:24 -0400129static void follow_to_parent(struct path *path)
130{
131 struct dentry *dp;
132
133 while (path->dentry == path->mnt->mnt_root && follow_up(path))
134 ;
135 dp = dget_parent(path->dentry);
136 dput(path->dentry);
137 path->dentry = dp;
138}
139
140static int nfsd_lookup_parent(struct svc_rqst *rqstp, struct dentry *dparent, struct svc_export **exp, struct dentry **dentryp)
141{
142 struct svc_export *exp2;
143 struct path path = {.mnt = mntget((*exp)->ex_path.mnt),
144 .dentry = dget(dparent)};
145
146 follow_to_parent(&path);
147
148 exp2 = rqst_exp_parent(rqstp, &path);
149 if (PTR_ERR(exp2) == -ENOENT) {
150 *dentryp = dget(dparent);
151 } else if (IS_ERR(exp2)) {
152 path_put(&path);
153 return PTR_ERR(exp2);
154 } else {
155 *dentryp = dget(path.dentry);
156 exp_put(*exp);
157 *exp = exp2;
158 }
159 path_put(&path);
160 return 0;
161}
162
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400163/*
164 * For nfsd purposes, we treat V4ROOT exports as though there was an
165 * export at *every* directory.
166 */
J. Bruce Fields3227fa42009-10-25 21:43:01 -0400167int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp)
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400168{
169 if (d_mountpoint(dentry))
170 return 1;
171 if (!(exp->ex_flags & NFSEXP_V4ROOT))
172 return 0;
173 return dentry->d_inode != NULL;
174}
175
Al Viro6264d692006-10-19 23:28:58 -0700176__be32
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700177nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400178 const char *name, unsigned int len,
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700179 struct svc_export **exp_ret, struct dentry **dentry_ret)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180{
181 struct svc_export *exp;
182 struct dentry *dparent;
183 struct dentry *dentry;
Al Viro6264d692006-10-19 23:28:58 -0700184 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185
186 dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name);
187
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 dparent = fhp->fh_dentry;
189 exp = fhp->fh_export;
190 exp_get(exp);
191
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 /* Lookup the name, but don't follow links */
193 if (isdotent(name, len)) {
194 if (len==1)
195 dentry = dget(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800196 else if (dparent != exp->ex_path.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 dentry = dget_parent(dparent);
J. Bruce Fieldsfed83812009-09-26 16:53:01 -0400198 else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 dentry = dget(dparent); /* .. == . just like at / */
200 else {
201 /* checking mountpoint crossing is very different when stepping up */
J. Bruce Fields289ede42009-09-26 20:32:24 -0400202 host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry);
203 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 }
206 } else {
207 fh_lock(fhp);
208 dentry = lookup_one_len(name, dparent, len);
Al Viro6264d692006-10-19 23:28:58 -0700209 host_err = PTR_ERR(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 if (IS_ERR(dentry))
211 goto out_nfserr;
212 /*
213 * check if we have crossed a mount point ...
214 */
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400215 if (nfsd_mountpoint(dentry, exp)) {
Al Viro6264d692006-10-19 23:28:58 -0700216 if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 dput(dentry);
218 goto out_nfserr;
219 }
220 }
221 }
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700222 *dentry_ret = dentry;
223 *exp_ret = exp;
224 return 0;
225
226out_nfserr:
227 exp_put(exp);
228 return nfserrno(host_err);
229}
230
231/*
232 * Look up one component of a pathname.
233 * N.B. After this call _both_ fhp and resfh need an fh_put
234 *
235 * If the lookup would cross a mountpoint, and the mounted filesystem
236 * is exported to the client with NFSEXP_NOHIDE, then the lookup is
237 * accepted as it stands and the mounted directory is
238 * returned. Otherwise the covered directory is returned.
239 * NOTE: this mountpoint crossing is not supported properly by all
240 * clients and is explicitly disallowed for NFSv3
241 * NeilBrown <neilb@cse.unsw.edu.au>
242 */
243__be32
244nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400245 unsigned int len, struct svc_fh *resfh)
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700246{
247 struct svc_export *exp;
248 struct dentry *dentry;
249 __be32 err;
250
J. Bruce Fields29a78a32011-04-09 11:28:53 -0400251 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
252 if (err)
253 return err;
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700254 err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry);
255 if (err)
256 return err;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700257 err = check_nfsd_access(exp, rqstp);
258 if (err)
259 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 /*
261 * Note: we compose the file handle now, but as the
262 * dentry may be negative, it may need to be updated.
263 */
264 err = fh_compose(resfh, exp, dentry, fhp);
265 if (!err && !dentry->d_inode)
266 err = nfserr_noent;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700267out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 exp_put(exp);
270 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271}
272
J. Bruce Fields4795bb32011-01-11 13:55:46 -0500273static int nfsd_break_lease(struct inode *inode)
274{
275 if (!S_ISREG(inode->i_mode))
276 return 0;
277 return break_lease(inode, O_WRONLY | O_NONBLOCK);
278}
279
Ben Myersf5019122010-02-17 14:05:11 -0600280/*
281 * Commit metadata changes to stable storage.
282 */
283static int
284commit_metadata(struct svc_fh *fhp)
285{
286 struct inode *inode = fhp->fh_dentry->d_inode;
287 const struct export_operations *export_ops = inode->i_sb->s_export_op;
Ben Myersf5019122010-02-17 14:05:11 -0600288
289 if (!EX_ISSYNC(fhp->fh_export))
290 return 0;
291
Christoph Hellwigc37650162010-10-06 10:48:20 +0200292 if (export_ops->commit_metadata)
293 return export_ops->commit_metadata(inode);
294 return sync_inode_metadata(inode, 1);
Ben Myersf5019122010-02-17 14:05:11 -0600295}
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700296
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297/*
298 * Set various file attributes.
299 * N.B. After this call fhp needs an fh_put
300 */
Al Viro6264d692006-10-19 23:28:58 -0700301__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
303 int check_guard, time_t guardtime)
304{
305 struct dentry *dentry;
306 struct inode *inode;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200307 int accmode = NFSD_MAY_SATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 int ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700309 __be32 err;
310 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 int size_change = 0;
312
313 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200314 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 if (iap->ia_valid & ATTR_SIZE)
316 ftype = S_IFREG;
317
318 /* Get inode */
319 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800320 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 goto out;
322
323 dentry = fhp->fh_dentry;
324 inode = dentry->d_inode;
325
NeilBrown15b7a1b2005-11-07 01:00:23 -0800326 /* Ignore any mode updates on symlinks */
327 if (S_ISLNK(inode->i_mode))
328 iap->ia_valid &= ~ATTR_MODE;
329
330 if (!iap->ia_valid)
331 goto out;
332
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000333 /*
334 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 * which only requires access, and "set-[ac]time-to-X" which
336 * requires ownership.
337 * So if it looks like it might be "set both to the same time which
338 * is close to now", and if inode_change_ok fails, then we
339 * convert to "set to now" instead of "set to explicit time"
340 *
341 * We only call inode_change_ok as the last test as technically
342 * it is not an interface that we should be using. It is only
343 * valid if the filesystem does not define it's own i_op->setattr.
344 */
345#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
346#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000347 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
348 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
349 /*
350 * Looks probable.
351 *
352 * Now just make sure time is in the right ballpark.
353 * Solaris, at least, doesn't seem to care what the time
354 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000356 time_t delta = iap->ia_atime.tv_sec - get_seconds();
357 if (delta < 0)
358 delta = -delta;
359 if (delta < MAX_TOUCH_TIME_ERROR &&
360 inode_change_ok(inode, iap) != 0) {
361 /*
362 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
363 * This will cause notify_change to set these times
364 * to "now"
365 */
366 iap->ia_valid &= ~BOTH_TIME_SET;
367 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 }
369
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000370 /*
371 * The size case is special.
372 * It changes the file as well as the attributes.
373 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 if (iap->ia_valid & ATTR_SIZE) {
375 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200376 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
377 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 if (err)
379 goto out;
380 }
381
Al Viro6264d692006-10-19 23:28:58 -0700382 host_err = get_write_access(inode);
383 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 goto out_nfserr;
385
386 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700387 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
388 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 put_write_access(inode);
390 goto out_nfserr;
391 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 }
393
Jeff Laytonca456252008-04-16 16:28:47 -0400394 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 if (iap->ia_valid & ATTR_MODE) {
396 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400397 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400398 }
399
400 /* Revoke setuid/setgid on chown */
Sachin S. Prabhu0953e622009-02-23 16:22:03 +0000401 if (!S_ISDIR(inode->i_mode) &&
402 (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
403 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) {
Jeff Laytonca456252008-04-16 16:28:47 -0400404 iap->ia_valid |= ATTR_KILL_PRIV;
405 if (iap->ia_valid & ATTR_MODE) {
406 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700407 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400408 if (iap->ia_mode & S_IXGRP)
409 iap->ia_mode &= ~S_ISGID;
410 } else {
411 /* set ATTR_KILL_* bits and let VFS handle it */
412 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700413 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 }
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 /* Change the attributes. */
417
418 iap->ia_valid |= ATTR_CTIME;
419
420 err = nfserr_notsync;
421 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
J. Bruce Fields4795bb32011-01-11 13:55:46 -0500422 host_err = nfsd_break_lease(inode);
J. Bruce Fields6a76beb2011-01-11 12:54:39 -0500423 if (host_err)
424 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 fh_lock(fhp);
J. Bruce Fields6a76beb2011-01-11 12:54:39 -0500426
Al Viro6264d692006-10-19 23:28:58 -0700427 host_err = notify_change(dentry, iap);
428 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 fh_unlock(fhp);
430 }
431 if (size_change)
432 put_write_access(inode);
433 if (!err)
Christoph Hellwigb160fda2010-06-01 21:59:18 +0200434 commit_metadata(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435out:
436 return err;
437
438out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700439 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 goto out;
441}
442
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800443#if defined(CONFIG_NFSD_V2_ACL) || \
444 defined(CONFIG_NFSD_V3_ACL) || \
445 defined(CONFIG_NFSD_V4)
446static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
447{
448 ssize_t buflen;
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530449 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800451 buflen = vfs_getxattr(dentry, key, NULL, 0);
452 if (buflen <= 0)
453 return buflen;
454
455 *buf = kmalloc(buflen, GFP_KERNEL);
456 if (!*buf)
457 return -ENOMEM;
458
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530459 ret = vfs_getxattr(dentry, key, *buf, buflen);
460 if (ret < 0)
461 kfree(*buf);
462 return ret;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800463}
464#endif
465
466#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467static int
468set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
469{
470 int len;
471 size_t buflen;
472 char *buf = NULL;
473 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
475 buflen = posix_acl_xattr_size(pacl->a_count);
476 buf = kmalloc(buflen, GFP_KERNEL);
477 error = -ENOMEM;
478 if (buf == NULL)
479 goto out;
480
481 len = posix_acl_to_xattr(pacl, buf, buflen);
482 if (len < 0) {
483 error = len;
484 goto out;
485 }
486
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800487 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488out:
489 kfree(buf);
490 return error;
491}
492
Al Viro6264d692006-10-19 23:28:58 -0700493__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
495 struct nfs4_acl *acl)
496{
Al Viro6264d692006-10-19 23:28:58 -0700497 __be32 error;
498 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 struct dentry *dentry;
500 struct inode *inode;
501 struct posix_acl *pacl = NULL, *dpacl = NULL;
502 unsigned int flags = 0;
503
504 /* Get inode */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200505 error = fh_verify(rqstp, fhp, 0 /* S_IFREG */, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700507 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
509 dentry = fhp->fh_dentry;
510 inode = dentry->d_inode;
511 if (S_ISDIR(inode->i_mode))
512 flags = NFS4_ACL_DIR;
513
Al Viro6264d692006-10-19 23:28:58 -0700514 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
515 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700516 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700517 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 goto out_nfserr;
519
Al Viro6264d692006-10-19 23:28:58 -0700520 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
521 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700522 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700524 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700525 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700527out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 posix_acl_release(pacl);
529 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800531 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700532 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800533 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700534 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}
536
537static struct posix_acl *
538_get_posix_acl(struct dentry *dentry, char *key)
539{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800540 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800542 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800544 buflen = nfsd_getxattr(dentry, key, &buf);
545 if (!buflen)
546 buflen = -ENODATA;
547 if (buflen <= 0)
548 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 kfree(buf);
552 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553}
554
555int
556nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
557{
558 struct inode *inode = dentry->d_inode;
559 int error = 0;
560 struct posix_acl *pacl = NULL, *dpacl = NULL;
561 unsigned int flags = 0;
562
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700563 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
565 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
566 if (IS_ERR(pacl)) {
567 error = PTR_ERR(pacl);
568 pacl = NULL;
569 goto out;
570 }
571
572 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700573 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
575 dpacl = NULL;
576 else if (IS_ERR(dpacl)) {
577 error = PTR_ERR(dpacl);
578 dpacl = NULL;
579 goto out;
580 }
581 flags = NFS4_ACL_DIR;
582 }
583
584 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
585 if (IS_ERR(*acl)) {
586 error = PTR_ERR(*acl);
587 *acl = NULL;
588 }
589 out:
590 posix_acl_release(pacl);
591 posix_acl_release(dpacl);
592 return error;
593}
594
J. Bruce Fields6a85d6c2010-07-06 12:39:12 -0400595#endif /* defined(CONFIG_NFSD_V4) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596
597#ifdef CONFIG_NFSD_V3
598/*
599 * Check server access rights to a file system object
600 */
601struct accessmap {
602 u32 access;
603 int how;
604};
605static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200606 { NFS3_ACCESS_READ, NFSD_MAY_READ },
607 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
608 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
609 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610
611 { 0, 0 }
612};
613
614static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200615 { NFS3_ACCESS_READ, NFSD_MAY_READ },
616 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
617 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
618 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
619 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
621 { 0, 0 }
622};
623
624static struct accessmap nfs3_anyaccess[] = {
625 /* Some clients - Solaris 2.6 at least, make an access call
626 * to the server to check for access for things like /dev/null
627 * (which really, the server doesn't care about). So
628 * We provide simple access checking for them, looking
629 * mainly at mode bits, and we make sure to ignore read-only
630 * filesystem checks
631 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200632 { NFS3_ACCESS_READ, NFSD_MAY_READ },
633 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
634 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
635 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
637 { 0, 0 }
638};
639
Al Viro6264d692006-10-19 23:28:58 -0700640__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
642{
643 struct accessmap *map;
644 struct svc_export *export;
645 struct dentry *dentry;
646 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700647 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200649 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 if (error)
651 goto out;
652
653 export = fhp->fh_export;
654 dentry = fhp->fh_dentry;
655
656 if (S_ISREG(dentry->d_inode->i_mode))
657 map = nfs3_regaccess;
658 else if (S_ISDIR(dentry->d_inode->i_mode))
659 map = nfs3_diraccess;
660 else
661 map = nfs3_anyaccess;
662
663
664 query = *access;
665 for (; map->access; map++) {
666 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700667 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
669 sresult |= map->access;
670
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700671 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 switch (err2) {
673 case nfs_ok:
674 result |= map->access;
675 break;
676
677 /* the following error codes just mean the access was not allowed,
678 * rather than an error occurred */
679 case nfserr_rofs:
680 case nfserr_acces:
681 case nfserr_perm:
682 /* simply don't "or" in the access bit. */
683 break;
684 default:
685 error = err2;
686 goto out;
687 }
688 }
689 }
690 *access = result;
691 if (supported)
692 *supported = sresult;
693
694 out:
695 return error;
696}
697#endif /* CONFIG_NFSD_V3 */
698
J. Bruce Fields105f4622011-06-07 11:50:23 -0400699static int nfsd_open_break_lease(struct inode *inode, int access)
700{
701 unsigned int mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
J. Bruce Fields105f4622011-06-07 11:50:23 -0400703 if (access & NFSD_MAY_NOT_BREAK_LEASE)
704 return 0;
705 mode = (access & NFSD_MAY_WRITE) ? O_WRONLY : O_RDONLY;
706 return break_lease(inode, mode | O_NONBLOCK);
707}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
709/*
710 * Open an existing file or directory.
711 * The access argument indicates the type of open (read/write/lock)
712 * N.B. After this call fhp needs an fh_put
713 */
Al Viro6264d692006-10-19 23:28:58 -0700714__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
716 int access, struct file **filp)
717{
718 struct dentry *dentry;
719 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700720 int flags = O_RDONLY|O_LARGEFILE;
721 __be32 err;
Jeff Layton91885252010-03-19 08:06:28 -0400722 int host_err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
David Howellse0e81732009-09-02 09:13:40 +0100724 validate_process_creds();
725
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 /*
727 * If we get here, then the client has already done an "open",
728 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
729 * in case a chmod has now revoked permission.
730 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200731 err = fh_verify(rqstp, fhp, type, access | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 if (err)
733 goto out;
734
735 dentry = fhp->fh_dentry;
736 inode = dentry->d_inode;
737
738 /* Disallow write access to files with the append-only bit set
739 * or any access when mandatory locking enabled
740 */
741 err = nfserr_perm;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200742 if (IS_APPEND(inode) && (access & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400744 /*
745 * We must ignore files (but only files) which might have mandatory
746 * locks on them because there is no way to know if the accesser has
747 * the lock.
748 */
749 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 goto out;
751
752 if (!inode->i_fop)
753 goto out;
754
J. Bruce Fields105f4622011-06-07 11:50:23 -0400755 host_err = nfsd_open_break_lease(inode, access);
Al Viro6264d692006-10-19 23:28:58 -0700756 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 goto out_nfserr;
758
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200759 if (access & NFSD_MAY_WRITE) {
760 if (access & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700761 flags = O_RDWR|O_LARGEFILE;
762 else
763 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 }
Jan Blunck54775492008-02-14 19:38:39 -0800765 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
David Howells033a6662009-07-02 14:35:32 +0100766 flags, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700768 host_err = PTR_ERR(*filp);
Chuck Ebbertaeaa5cc2010-02-15 18:07:39 -0500769 else
770 host_err = ima_file_check(*filp, access);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700772 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773out:
David Howellse0e81732009-09-02 09:13:40 +0100774 validate_process_creds();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 return err;
776}
777
778/*
779 * Close a file.
780 */
781void
782nfsd_close(struct file *filp)
783{
784 fput(filp);
785}
786
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500787/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 * Obtain the readahead parameters for the file
789 * specified by (dev, ino).
790 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
792static inline struct raparms *
793nfsd_get_raparms(dev_t dev, ino_t ino)
794{
795 struct raparms *ra, **rap, **frap = NULL;
796 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700797 unsigned int hash;
798 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799
Greg Banksfce14562006-10-04 02:15:49 -0700800 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
801 rab = &raparm_hash[hash];
802
803 spin_lock(&rab->pb_lock);
804 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 if (ra->p_ino == ino && ra->p_dev == dev)
806 goto found;
807 depth++;
808 if (ra->p_count == 0)
809 frap = rap;
810 }
Konstantin Khorenko3aa6e0a2011-02-01 17:16:29 +0300811 depth = nfsdstats.ra_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700813 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 return NULL;
815 }
816 rap = frap;
817 ra = *frap;
818 ra->p_dev = dev;
819 ra->p_ino = ino;
820 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700821 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822found:
Greg Banksfce14562006-10-04 02:15:49 -0700823 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700825 ra->p_next = rab->pb_head;
826 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 }
828 ra->p_count++;
829 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700830 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 return ra;
832}
833
834/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200835 * Grab and keep cached pages associated with a file in the svc_rqst
836 * so that they can be passed to the network sendmsg/sendpage routines
837 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 */
839static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200840nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
841 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842{
Jens Axboecf8208d2007-06-12 21:22:14 +0200843 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700844 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200845 struct page *page = buf->page;
846 size_t size;
Jens Axboecf8208d2007-06-12 21:22:14 +0200847
848 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
850 if (rqstp->rq_res.page_len == 0) {
851 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700852 put_page(*pp);
853 *pp = page;
854 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200855 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700857 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800859 if (*pp)
860 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700861 *pp = page;
862 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700864 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 return size;
868}
869
Jens Axboecf8208d2007-06-12 21:22:14 +0200870static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
871 struct splice_desc *sd)
872{
873 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
874}
875
Al Viro6264d692006-10-19 23:28:58 -0700876static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
878 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
879{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700881 __be32 err;
882 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
884 err = nfserr_perm;
Dave Hansena8754be2007-10-16 23:31:15 -0700885
Jens Axboecf8208d2007-06-12 21:22:14 +0200886 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
887 struct splice_desc sd = {
888 .len = 0,
889 .total_len = *count,
890 .pos = offset,
891 .u.data = rqstp,
892 };
893
Jens Axboe4fbef202007-07-13 22:42:20 +0200894 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200895 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 } else {
897 oldfs = get_fs();
898 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700899 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 set_fs(oldfs);
901 }
902
Al Viro6264d692006-10-19 23:28:58 -0700903 if (host_err >= 0) {
904 nfsdstats.io_read += host_err;
905 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 err = 0;
Eric Paris2a12a9d2009-12-17 21:24:21 -0500907 fsnotify_access(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 } else
Al Viro6264d692006-10-19 23:28:58 -0700909 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 return err;
911}
912
Neil Brown9f708e42006-01-06 00:19:59 -0800913static void kill_suid(struct dentry *dentry)
914{
915 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700916 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800917
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800918 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800919 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800920 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800921}
922
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700923/*
924 * Gathered writes: If another process is currently writing to the file,
925 * there's a high chance this is another nfsd (triggered by a bulk write
926 * from a client's biod). Rather than syncing the file with each write
927 * request, we sleep for 10 msec.
928 *
929 * I don't know if this roughly approximates C. Juszak's idea of
930 * gathered writes, but it's a nice and simple solution (IMHO), and it
931 * seems to work:-)
932 *
933 * Note: we do this only in the NFSv2 case, since v3 and higher have a
934 * better tool (separate unstable writes and commits) for solving this
935 * problem.
936 */
937static int wait_for_concurrent_writes(struct file *file)
938{
939 struct inode *inode = file->f_path.dentry->d_inode;
940 static ino_t last_ino;
941 static dev_t last_dev;
942 int err = 0;
943
944 if (atomic_read(&inode->i_writecount) > 1
945 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
946 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
947 msleep(10);
948 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
949 }
950
951 if (inode->i_state & I_DIRTY) {
952 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Christoph Hellwig8018ab02010-03-22 17:32:25 +0100953 err = vfs_fsync(file, 0);
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700954 }
955 last_ino = inode->i_ino;
956 last_dev = inode->i_sb->s_dev;
957 return err;
958}
959
Al Viro6264d692006-10-19 23:28:58 -0700960static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
962 loff_t offset, struct kvec *vec, int vlen,
David Shaw31dec252009-03-05 20:16:14 -0500963 unsigned long *cnt, int *stablep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964{
965 struct svc_export *exp;
966 struct dentry *dentry;
967 struct inode *inode;
968 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700969 __be32 err = 0;
970 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 int stable = *stablep;
Trond Myklebust48e03bc2009-06-05 12:35:15 -0400972 int use_wgather;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -0800974 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 inode = dentry->d_inode;
976 exp = fhp->fh_export;
977
978 /*
979 * Request sync writes if
980 * - the sync export option has been set, or
981 * - the client requested O_SYNC behavior (NFSv3 feature).
982 * - The file system doesn't support fsync().
Trond Myklebust48e03bc2009-06-05 12:35:15 -0400983 * When NFSv2 gathered writes have been configured for this volume,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 * flushing the data to disk is handled separately below.
985 */
Trond Myklebust48e03bc2009-06-05 12:35:15 -0400986 use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
Harvey Harrison3ba15142008-02-20 12:49:02 -0800988 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 stable = 2;
990 *stablep = 2; /* FILE_SYNC */
991 }
992
993 if (!EX_ISSYNC(exp))
994 stable = 0;
Trond Myklebust48e03bc2009-06-05 12:35:15 -0400995 if (stable && !use_wgather) {
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -0700996 spin_lock(&file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 file->f_flags |= O_SYNC;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -0700998 spin_unlock(&file->f_lock);
999 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
1001 /* Write the data. */
1002 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001003 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 set_fs(oldfs);
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001005 if (host_err < 0)
1006 goto out_nfserr;
1007 *cnt = host_err;
1008 nfsdstats.io_write += host_err;
Eric Paris2a12a9d2009-12-17 21:24:21 -05001009 fsnotify_modify(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
1011 /* clear setuid/setgid flag after write */
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001012 if (inode->i_mode & (S_ISUID | S_ISGID))
Neil Brown9f708e42006-01-06 00:19:59 -08001013 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001015 if (stable && use_wgather)
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001016 host_err = wait_for_concurrent_writes(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001018out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001019 dprintk("nfsd: write complete host_err=%d\n", host_err);
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001020 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 err = 0;
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001022 else
Al Viro6264d692006-10-19 23:28:58 -07001023 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 return err;
1025}
1026
1027/*
1028 * Read data from a file. count must contain the requested read count
1029 * on entry. On return, *count contains the number of bytes actually read.
1030 * N.B. After this call fhp needs an fh_put
1031 */
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001032__be32 nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
J. Bruce Fieldsfa0a2122010-07-27 16:48:54 -04001033 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
1034{
1035 struct file *file;
1036 struct inode *inode;
1037 struct raparms *ra;
1038 __be32 err;
1039
1040 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
1041 if (err)
1042 return err;
1043
1044 inode = file->f_path.dentry->d_inode;
1045
1046 /* Get readahead parameters */
1047 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
1048
1049 if (ra && ra->p_set)
1050 file->f_ra = ra->p_ra;
1051
1052 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1053
1054 /* Write back readahead params */
1055 if (ra) {
1056 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
1057 spin_lock(&rab->pb_lock);
1058 ra->p_ra = file->f_ra;
1059 ra->p_set = 1;
1060 ra->p_count--;
1061 spin_unlock(&rab->pb_lock);
1062 }
1063
1064 nfsd_close(file);
1065 return err;
1066}
1067
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001068/* As above, but use the provided file descriptor. */
Al Viro6264d692006-10-19 23:28:58 -07001069__be32
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001070nfsd_read_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 loff_t offset, struct kvec *vec, int vlen,
1072 unsigned long *count)
1073{
Al Viro6264d692006-10-19 23:28:58 -07001074 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
1076 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001077 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001078 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 if (err)
1080 goto out;
1081 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001082 } else /* Note file may still be NULL in NFSv4 special stateid case: */
1083 err = nfsd_read(rqstp, fhp, offset, vec, vlen, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084out:
1085 return err;
1086}
1087
1088/*
1089 * Write data to a file.
1090 * The stable flag requests synchronous writes.
1091 * N.B. After this call fhp needs an fh_put
1092 */
Al Viro6264d692006-10-19 23:28:58 -07001093__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
David Shaw31dec252009-03-05 20:16:14 -05001095 loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 int *stablep)
1097{
Al Viro6264d692006-10-19 23:28:58 -07001098 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
1100 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001101 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001102 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 if (err)
1104 goto out;
1105 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1106 stablep);
1107 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001108 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 if (err)
1110 goto out;
1111
1112 if (cnt)
1113 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1114 cnt, stablep);
1115 nfsd_close(file);
1116 }
1117out:
1118 return err;
1119}
1120
1121#ifdef CONFIG_NFSD_V3
1122/*
1123 * Commit all pending writes to stable storage.
Trond Myklebustaa696a62010-01-29 16:44:25 -05001124 *
1125 * Note: we only guarantee that data that lies within the range specified
1126 * by the 'offset' and 'count' parameters will be synced.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 *
1128 * Unfortunately we cannot lock the file to make sure we return full WCC
1129 * data to the client, as locking happens lower down in the filesystem.
1130 */
Al Viro6264d692006-10-19 23:28:58 -07001131__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1133 loff_t offset, unsigned long count)
1134{
1135 struct file *file;
Trond Myklebustaa696a62010-01-29 16:44:25 -05001136 loff_t end = LLONG_MAX;
1137 __be32 err = nfserr_inval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
Trond Myklebustaa696a62010-01-29 16:44:25 -05001139 if (offset < 0)
1140 goto out;
1141 if (count != 0) {
1142 end = offset + (loff_t)count - 1;
1143 if (end < offset)
1144 goto out;
1145 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Jeff Layton91885252010-03-19 08:06:28 -04001147 err = nfsd_open(rqstp, fhp, S_IFREG,
1148 NFSD_MAY_WRITE|NFSD_MAY_NOT_BREAK_LEASE, &file);
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001149 if (err)
Trond Myklebustaa696a62010-01-29 16:44:25 -05001150 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 if (EX_ISSYNC(fhp->fh_export)) {
Christoph Hellwig8018ab02010-03-22 17:32:25 +01001152 int err2 = vfs_fsync_range(file, offset, end, 0);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001153
1154 if (err2 != -EINVAL)
1155 err = nfserrno(err2);
1156 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 err = nfserr_notsupp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 }
1159
1160 nfsd_close(file);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001161out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 return err;
1163}
1164#endif /* CONFIG_NFSD_V3 */
1165
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001166static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001167nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1168 struct iattr *iap)
1169{
1170 /*
1171 * Mode has already been set earlier in create:
1172 */
1173 iap->ia_valid &= ~ATTR_MODE;
1174 /*
1175 * Setting uid/gid works only for root. Irix appears to
1176 * send along the gid on create when it tries to implement
1177 * setgid directories via NFS:
1178 */
David Howells5cc0a842008-11-14 10:38:58 +11001179 if (current_fsuid() != 0)
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001180 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1181 if (iap->ia_valid)
1182 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1183 return 0;
1184}
1185
wengang wang4ac35c22009-02-10 11:27:51 +08001186/* HPUX client sometimes creates a file in mode 000, and sets size to 0.
1187 * setting size to 0 may fail for some specific file systems by the permission
1188 * checking which requires WRITE permission but the mode is 000.
1189 * we ignore the resizing(to 0) on the just new created file, since the size is
1190 * 0 after file created.
1191 *
1192 * call this only after vfs_create() is called.
1193 * */
1194static void
1195nfsd_check_ignore_resizing(struct iattr *iap)
1196{
1197 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
1198 iap->ia_valid &= ~ATTR_SIZE;
1199}
1200
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201/*
1202 * Create a file (regular, directory, device, fifo); UNIX sockets
1203 * not yet implemented.
1204 * If the response fh has been verified, the parent directory should
1205 * already be locked. Note that the parent directory is left locked.
1206 *
1207 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1208 */
Al Viro6264d692006-10-19 23:28:58 -07001209__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1211 char *fname, int flen, struct iattr *iap,
1212 int type, dev_t rdev, struct svc_fh *resfhp)
1213{
1214 struct dentry *dentry, *dchild = NULL;
1215 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001216 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001217 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001218 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219
1220 err = nfserr_perm;
1221 if (!flen)
1222 goto out;
1223 err = nfserr_exist;
1224 if (isdotent(fname, flen))
1225 goto out;
1226
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001227 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 if (err)
1229 goto out;
1230
1231 dentry = fhp->fh_dentry;
1232 dirp = dentry->d_inode;
1233
1234 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001235 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 goto out;
1237 /*
1238 * Check whether the response file handle has been verified yet.
1239 * If it has, the parent directory should already be locked.
1240 */
1241 if (!resfhp->fh_dentry) {
1242 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001243 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001245 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 if (IS_ERR(dchild))
1247 goto out_nfserr;
1248 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1249 if (err)
1250 goto out;
1251 } else {
1252 /* called from nfsd_proc_create */
1253 dchild = dget(resfhp->fh_dentry);
1254 if (!fhp->fh_locked) {
1255 /* not actually possible */
1256 printk(KERN_ERR
1257 "nfsd_create: parent %s/%s not locked!\n",
1258 dentry->d_parent->d_name.name,
1259 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001260 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 goto out;
1262 }
1263 }
1264 /*
1265 * Make sure the child dentry is still negative ...
1266 */
1267 err = nfserr_exist;
1268 if (dchild->d_inode) {
1269 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1270 dentry->d_name.name, dchild->d_name.name);
1271 goto out;
1272 }
1273
1274 if (!(iap->ia_valid & ATTR_MODE))
1275 iap->ia_mode = 0;
1276 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1277
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001278 err = nfserr_inval;
1279 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1280 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1281 type);
1282 goto out;
1283 }
1284
1285 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1286 if (host_err)
1287 goto out_nfserr;
1288
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 /*
1290 * Get the dir op function pointer.
1291 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001292 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 switch (type) {
1294 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001295 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
wengang wang4ac35c22009-02-10 11:27:51 +08001296 if (!host_err)
1297 nfsd_check_ignore_resizing(iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 break;
1299 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001300 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 break;
1302 case S_IFCHR:
1303 case S_IFBLK:
1304 case S_IFIFO:
1305 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001306 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001308 }
1309 if (host_err < 0) {
1310 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1311 goto out_nfserr;
1312 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
Ben Myersf5019122010-02-17 14:05:11 -06001314 err = nfsd_create_setattr(rqstp, resfhp, iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
Ben Myersf5019122010-02-17 14:05:11 -06001316 /*
1317 * nfsd_setattr already committed the child. Transactional filesystems
1318 * had a chance to commit changes for both parent and child
1319 * simultaneously making the following commit_metadata a noop.
1320 */
1321 err2 = nfserrno(commit_metadata(fhp));
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001322 if (err2)
1323 err = err2;
Dave Hansen463c3192008-02-15 14:37:57 -08001324 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 /*
1326 * Update the file handle to get the new inode info.
1327 */
1328 if (!err)
1329 err = fh_update(resfhp);
1330out:
1331 if (dchild && !IS_ERR(dchild))
1332 dput(dchild);
1333 return err;
1334
1335out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001336 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 goto out;
1338}
1339
1340#ifdef CONFIG_NFSD_V3
Mi Jinlongac6721a2011-04-20 17:06:25 +08001341
1342static inline int nfsd_create_is_exclusive(int createmode)
1343{
1344 return createmode == NFS3_CREATE_EXCLUSIVE
1345 || createmode == NFS4_CREATE_EXCLUSIVE4_1;
1346}
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348/*
Mi Jinlongac6721a2011-04-20 17:06:25 +08001349 * NFSv3 and NFSv4 version of nfsd_create
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 */
Al Viro6264d692006-10-19 23:28:58 -07001351__be32
Mi Jinlongac6721a2011-04-20 17:06:25 +08001352do_nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 char *fname, int flen, struct iattr *iap,
1354 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001355 int *truncp, int *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356{
1357 struct dentry *dentry, *dchild = NULL;
1358 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001359 __be32 err;
1360 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
1363 err = nfserr_perm;
1364 if (!flen)
1365 goto out;
1366 err = nfserr_exist;
1367 if (isdotent(fname, flen))
1368 goto out;
1369 if (!(iap->ia_valid & ATTR_MODE))
1370 iap->ia_mode = 0;
Sachin Prabhu1574dff2011-04-20 13:09:35 +01001371 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 if (err)
1373 goto out;
1374
1375 dentry = fhp->fh_dentry;
1376 dirp = dentry->d_inode;
1377
1378 /* Get all the sanity checks out of the way before
1379 * we lock the parent. */
1380 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001381 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001383 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384
1385 /*
1386 * Compose the response file handle.
1387 */
1388 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001389 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (IS_ERR(dchild))
1391 goto out_nfserr;
1392
Sachin Prabhu1574dff2011-04-20 13:09:35 +01001393 /* If file doesn't exist, check for permissions to create one */
1394 if (!dchild->d_inode) {
1395 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
1396 if (err)
1397 goto out;
1398 }
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1401 if (err)
1402 goto out;
1403
Mi Jinlongac6721a2011-04-20 17:06:25 +08001404 if (nfsd_create_is_exclusive(createmode)) {
Peter Staubachc3978522007-01-26 00:57:00 -08001405 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001406 * the high bit set, so just clear the high bits. If this is
1407 * ever changed to use different attrs for storing the
1408 * verifier, then do_open_lookup() will also need to be fixed
1409 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 */
1411 v_mtime = verifier[0]&0x7fffffff;
1412 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 }
1414
Dave Hansen463c3192008-02-15 14:37:57 -08001415 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1416 if (host_err)
1417 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 if (dchild->d_inode) {
1419 err = 0;
1420
1421 switch (createmode) {
1422 case NFS3_CREATE_UNCHECKED:
1423 if (! S_ISREG(dchild->d_inode->i_mode))
1424 err = nfserr_exist;
1425 else if (truncp) {
1426 /* in nfsv4, we need to treat this case a little
1427 * differently. we don't want to truncate the
1428 * file now; this would be wrong if the OPEN
1429 * fails for some other reason. furthermore,
1430 * if the size is nonzero, we should ignore it
1431 * according to spec!
1432 */
1433 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1434 }
1435 else {
1436 iap->ia_valid &= ATTR_SIZE;
1437 goto set_attr;
1438 }
1439 break;
1440 case NFS3_CREATE_EXCLUSIVE:
1441 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1442 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 && dchild->d_inode->i_size == 0 )
1444 break;
Mi Jinlongac6721a2011-04-20 17:06:25 +08001445 case NFS4_CREATE_EXCLUSIVE4_1:
1446 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1447 && dchild->d_inode->i_atime.tv_sec == v_atime
1448 && dchild->d_inode->i_size == 0 )
1449 goto set_attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 /* fallthru */
1451 case NFS3_CREATE_GUARDED:
1452 err = nfserr_exist;
1453 }
Dave Hansen463c3192008-02-15 14:37:57 -08001454 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 goto out;
1456 }
1457
Al Viro6264d692006-10-19 23:28:58 -07001458 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001459 if (host_err < 0) {
1460 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001462 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001463 if (created)
1464 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
wengang wang4ac35c22009-02-10 11:27:51 +08001466 nfsd_check_ignore_resizing(iap);
1467
Mi Jinlongac6721a2011-04-20 17:06:25 +08001468 if (nfsd_create_is_exclusive(createmode)) {
Peter Staubachc3978522007-01-26 00:57:00 -08001469 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001471 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 /* XXX someone who knows this better please fix it for nsec */
1473 iap->ia_mtime.tv_sec = v_mtime;
1474 iap->ia_atime.tv_sec = v_atime;
1475 iap->ia_mtime.tv_nsec = 0;
1476 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 }
1478
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 set_attr:
Ben Myersf5019122010-02-17 14:05:11 -06001480 err = nfsd_create_setattr(rqstp, resfhp, iap);
1481
1482 /*
1483 * nfsd_setattr already committed the child (and possibly also the parent).
1484 */
1485 if (!err)
1486 err = nfserrno(commit_metadata(fhp));
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001487
Dave Hansen463c3192008-02-15 14:37:57 -08001488 mnt_drop_write(fhp->fh_export->ex_path.mnt);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001489 /*
1490 * Update the filehandle to get the new inode info.
1491 */
1492 if (!err)
1493 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494
1495 out:
1496 fh_unlock(fhp);
1497 if (dchild && !IS_ERR(dchild))
1498 dput(dchild);
1499 return err;
1500
1501 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001502 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 goto out;
1504}
1505#endif /* CONFIG_NFSD_V3 */
1506
1507/*
1508 * Read a symlink. On entry, *lenp must contain the maximum path length that
1509 * fits into the buffer. On return, it contains the true length.
1510 * N.B. After this call fhp needs an fh_put
1511 */
Al Viro6264d692006-10-19 23:28:58 -07001512__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1514{
1515 struct dentry *dentry;
1516 struct inode *inode;
1517 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001518 __be32 err;
1519 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001521 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 if (err)
1523 goto out;
1524
1525 dentry = fhp->fh_dentry;
1526 inode = dentry->d_inode;
1527
1528 err = nfserr_inval;
Al Viroacfa4382008-12-04 10:06:33 -05001529 if (!inode->i_op->readlink)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 goto out;
1531
Jan Blunck54775492008-02-14 19:38:39 -08001532 touch_atime(fhp->fh_export->ex_path.mnt, dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 /* N.B. Why does this call need a get_fs()??
1534 * Remove the set_fs and watch the fireworks:-) --okir
1535 */
1536
1537 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001538 host_err = inode->i_op->readlink(dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 set_fs(oldfs);
1540
Al Viro6264d692006-10-19 23:28:58 -07001541 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001543 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 err = 0;
1545out:
1546 return err;
1547
1548out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001549 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 goto out;
1551}
1552
1553/*
1554 * Create a symlink and look up its inode
1555 * N.B. After this call _both_ fhp and resfhp need an fh_put
1556 */
Al Viro6264d692006-10-19 23:28:58 -07001557__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1559 char *fname, int flen,
1560 char *path, int plen,
1561 struct svc_fh *resfhp,
1562 struct iattr *iap)
1563{
1564 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001565 __be32 err, cerr;
1566 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
1568 err = nfserr_noent;
1569 if (!flen || !plen)
1570 goto out;
1571 err = nfserr_exist;
1572 if (isdotent(fname, flen))
1573 goto out;
1574
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001575 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (err)
1577 goto out;
1578 fh_lock(fhp);
1579 dentry = fhp->fh_dentry;
1580 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001581 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (IS_ERR(dnew))
1583 goto out_nfserr;
1584
Dave Hansen75c3f292008-02-15 14:37:45 -08001585 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1586 if (host_err)
1587 goto out_nfserr;
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if (unlikely(path[plen] != 0)) {
1590 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1591 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001592 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 else {
1594 strncpy(path_alloced, path, plen);
1595 path_alloced[plen] = 0;
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001596 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 kfree(path_alloced);
1598 }
1599 } else
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001600 host_err = vfs_symlink(dentry->d_inode, dnew, path);
Al Viro6264d692006-10-19 23:28:58 -07001601 err = nfserrno(host_err);
Ben Myersf5019122010-02-17 14:05:11 -06001602 if (!err)
1603 err = nfserrno(commit_metadata(fhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 fh_unlock(fhp);
1605
Dave Hansen75c3f292008-02-15 14:37:45 -08001606 mnt_drop_write(fhp->fh_export->ex_path.mnt);
1607
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1609 dput(dnew);
1610 if (err==0) err = cerr;
1611out:
1612 return err;
1613
1614out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001615 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 goto out;
1617}
1618
1619/*
1620 * Create a hardlink
1621 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1622 */
Al Viro6264d692006-10-19 23:28:58 -07001623__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1625 char *name, int len, struct svc_fh *tfhp)
1626{
1627 struct dentry *ddir, *dnew, *dold;
J. Bruce Fields55b13352010-07-19 16:38:24 -04001628 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001629 __be32 err;
1630 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001632 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 if (err)
1634 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001635 err = fh_verify(rqstp, tfhp, -S_IFDIR, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 if (err)
1637 goto out;
1638
1639 err = nfserr_perm;
1640 if (!len)
1641 goto out;
1642 err = nfserr_exist;
1643 if (isdotent(name, len))
1644 goto out;
1645
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001646 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 ddir = ffhp->fh_dentry;
1648 dirp = ddir->d_inode;
1649
1650 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001651 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 if (IS_ERR(dnew))
1653 goto out_nfserr;
1654
1655 dold = tfhp->fh_dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
Dave Hansen75c3f292008-02-15 14:37:45 -08001657 host_err = mnt_want_write(tfhp->fh_export->ex_path.mnt);
1658 if (host_err) {
1659 err = nfserrno(host_err);
1660 goto out_dput;
1661 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001662 err = nfserr_noent;
1663 if (!dold->d_inode)
1664 goto out_drop_write;
1665 host_err = nfsd_break_lease(dold->d_inode);
Casey Bodley7d751f62011-06-03 12:21:23 -04001666 if (host_err) {
1667 err = nfserrno(host_err);
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001668 goto out_drop_write;
Casey Bodley7d751f62011-06-03 12:21:23 -04001669 }
Al Viro6264d692006-10-19 23:28:58 -07001670 host_err = vfs_link(dold, dirp, dnew);
1671 if (!host_err) {
Ben Myersf5019122010-02-17 14:05:11 -06001672 err = nfserrno(commit_metadata(ffhp));
1673 if (!err)
1674 err = nfserrno(commit_metadata(tfhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 } else {
Al Viro6264d692006-10-19 23:28:58 -07001676 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 err = nfserr_acces;
1678 else
Al Viro6264d692006-10-19 23:28:58 -07001679 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001681out_drop_write:
Dave Hansen75c3f292008-02-15 14:37:45 -08001682 mnt_drop_write(tfhp->fh_export->ex_path.mnt);
1683out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001685out_unlock:
1686 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687out:
1688 return err;
1689
1690out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001691 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001692 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693}
1694
1695/*
1696 * Rename a file
1697 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1698 */
Al Viro6264d692006-10-19 23:28:58 -07001699__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1701 struct svc_fh *tfhp, char *tname, int tlen)
1702{
1703 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1704 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001705 __be32 err;
1706 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001708 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 if (err)
1710 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001711 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 if (err)
1713 goto out;
1714
1715 fdentry = ffhp->fh_dentry;
1716 fdir = fdentry->d_inode;
1717
1718 tdentry = tfhp->fh_dentry;
1719 tdir = tdentry->d_inode;
1720
1721 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001722 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 goto out;
1724
1725 err = nfserr_perm;
1726 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1727 goto out;
1728
1729 /* cannot use fh_lock as we need deadlock protective ordering
1730 * so do it by hand */
1731 trap = lock_rename(tdentry, fdentry);
1732 ffhp->fh_locked = tfhp->fh_locked = 1;
1733 fill_pre_wcc(ffhp);
1734 fill_pre_wcc(tfhp);
1735
1736 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001737 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 if (IS_ERR(odentry))
1739 goto out_nfserr;
1740
Al Viro6264d692006-10-19 23:28:58 -07001741 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 if (!odentry->d_inode)
1743 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001744 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 if (odentry == trap)
1746 goto out_dput_old;
1747
1748 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001749 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (IS_ERR(ndentry))
1751 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001752 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (ndentry == trap)
1754 goto out_dput_new;
1755
Dave Hansen9079b1e2008-02-15 14:37:49 -08001756 host_err = -EXDEV;
1757 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1758 goto out_dput_new;
1759 host_err = mnt_want_write(ffhp->fh_export->ex_path.mnt);
1760 if (host_err)
1761 goto out_dput_new;
1762
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001763 host_err = nfsd_break_lease(odentry->d_inode);
1764 if (host_err)
1765 goto out_drop_write;
J. Bruce Fields83f6b0c2011-02-06 16:46:30 -05001766 if (ndentry->d_inode) {
1767 host_err = nfsd_break_lease(ndentry->d_inode);
1768 if (host_err)
1769 goto out_drop_write;
1770 }
Al Viro6264d692006-10-19 23:28:58 -07001771 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
Ben Myersf5019122010-02-17 14:05:11 -06001772 if (!host_err) {
1773 host_err = commit_metadata(tfhp);
Al Viro6264d692006-10-19 23:28:58 -07001774 if (!host_err)
Ben Myersf5019122010-02-17 14:05:11 -06001775 host_err = commit_metadata(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001777out_drop_write:
Dave Hansen9079b1e2008-02-15 14:37:49 -08001778 mnt_drop_write(ffhp->fh_export->ex_path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 out_dput_new:
1780 dput(ndentry);
1781 out_dput_old:
1782 dput(odentry);
1783 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001784 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785
1786 /* we cannot reply on fh_unlock on the two filehandles,
1787 * as that would do the wrong thing if the two directories
1788 * were the same, so again we do it by hand
1789 */
1790 fill_post_wcc(ffhp);
1791 fill_post_wcc(tfhp);
1792 unlock_rename(tdentry, fdentry);
1793 ffhp->fh_locked = tfhp->fh_locked = 0;
1794
1795out:
1796 return err;
1797}
1798
1799/*
1800 * Unlink a file or directory
1801 * N.B. After this call fhp needs an fh_put
1802 */
Al Viro6264d692006-10-19 23:28:58 -07001803__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1805 char *fname, int flen)
1806{
1807 struct dentry *dentry, *rdentry;
1808 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001809 __be32 err;
1810 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811
1812 err = nfserr_acces;
1813 if (!flen || isdotent(fname, flen))
1814 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001815 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 if (err)
1817 goto out;
1818
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001819 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 dentry = fhp->fh_dentry;
1821 dirp = dentry->d_inode;
1822
1823 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001824 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 if (IS_ERR(rdentry))
1826 goto out_nfserr;
1827
1828 if (!rdentry->d_inode) {
1829 dput(rdentry);
1830 err = nfserr_noent;
1831 goto out;
1832 }
1833
1834 if (!type)
1835 type = rdentry->d_inode->i_mode & S_IFMT;
1836
Dave Hansen06227532008-02-15 14:37:34 -08001837 host_err = mnt_want_write(fhp->fh_export->ex_path.mnt);
1838 if (host_err)
J. Bruce Fields541ce982011-01-14 20:00:02 -05001839 goto out_put;
Dave Hansen06227532008-02-15 14:37:34 -08001840
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001841 host_err = nfsd_break_lease(rdentry->d_inode);
1842 if (host_err)
J. Bruce Fields541ce982011-01-14 20:00:02 -05001843 goto out_drop_write;
J. Bruce Fields9ce137e2011-01-11 14:07:12 -05001844 if (type != S_IFDIR)
Al Viro6264d692006-10-19 23:28:58 -07001845 host_err = vfs_unlink(dirp, rdentry);
J. Bruce Fields9ce137e2011-01-11 14:07:12 -05001846 else
Al Viro6264d692006-10-19 23:28:58 -07001847 host_err = vfs_rmdir(dirp, rdentry);
J. Bruce Fields541ce982011-01-14 20:00:02 -05001848 if (!host_err)
1849 host_err = commit_metadata(fhp);
1850out_drop_write:
1851 mnt_drop_write(fhp->fh_export->ex_path.mnt);
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001852out_put:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 dput(rdentry);
1854
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001856 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001857out:
1858 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859}
1860
1861/*
David Woodhouse14f7dd62008-07-31 20:29:12 +01001862 * We do this buffering because we must not call back into the file
1863 * system's ->lookup() method from the filldir callback. That may well
1864 * deadlock a number of file systems.
1865 *
1866 * This is based heavily on the implementation of same in XFS.
1867 */
1868struct buffered_dirent {
1869 u64 ino;
1870 loff_t offset;
1871 int namlen;
1872 unsigned int d_type;
1873 char name[];
1874};
1875
1876struct readdir_data {
1877 char *dirent;
1878 size_t used;
Al Viro53c9c5c2008-08-24 07:29:52 -04001879 int full;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001880};
1881
1882static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen,
1883 loff_t offset, u64 ino, unsigned int d_type)
David Woodhouse2628b762008-07-31 17:16:51 +01001884{
David Woodhouse14f7dd62008-07-31 20:29:12 +01001885 struct readdir_data *buf = __buf;
1886 struct buffered_dirent *de = (void *)(buf->dirent + buf->used);
1887 unsigned int reclen;
1888
1889 reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64));
Al Viro53c9c5c2008-08-24 07:29:52 -04001890 if (buf->used + reclen > PAGE_SIZE) {
1891 buf->full = 1;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001892 return -EINVAL;
Al Viro53c9c5c2008-08-24 07:29:52 -04001893 }
David Woodhouse14f7dd62008-07-31 20:29:12 +01001894
1895 de->namlen = namlen;
1896 de->offset = offset;
1897 de->ino = ino;
1898 de->d_type = d_type;
1899 memcpy(de->name, name, namlen);
1900 buf->used += reclen;
1901
1902 return 0;
1903}
1904
David Woodhouse2f9092e2009-04-20 23:18:37 +01001905static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func,
1906 struct readdir_cd *cdp, loff_t *offsetp)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001907{
1908 struct readdir_data buf;
1909 struct buffered_dirent *de;
David Woodhouse2628b762008-07-31 17:16:51 +01001910 int host_err;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001911 int size;
1912 loff_t offset;
David Woodhouse2628b762008-07-31 17:16:51 +01001913
David Woodhouse14f7dd62008-07-31 20:29:12 +01001914 buf.dirent = (void *)__get_free_page(GFP_KERNEL);
1915 if (!buf.dirent)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001916 return nfserrno(-ENOMEM);
David Woodhouse2628b762008-07-31 17:16:51 +01001917
David Woodhouse14f7dd62008-07-31 20:29:12 +01001918 offset = *offsetp;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001919
1920 while (1) {
David Woodhouse2f9092e2009-04-20 23:18:37 +01001921 struct inode *dir_inode = file->f_path.dentry->d_inode;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001922 unsigned int reclen;
1923
Doug Nazarb726e922008-11-05 06:16:28 -05001924 cdp->err = nfserr_eof; /* will be cleared on successful read */
David Woodhouse14f7dd62008-07-31 20:29:12 +01001925 buf.used = 0;
Al Viro53c9c5c2008-08-24 07:29:52 -04001926 buf.full = 0;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001927
1928 host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf);
Al Viro53c9c5c2008-08-24 07:29:52 -04001929 if (buf.full)
1930 host_err = 0;
1931
1932 if (host_err < 0)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001933 break;
1934
1935 size = buf.used;
1936
1937 if (!size)
1938 break;
1939
David Woodhouse2f9092e2009-04-20 23:18:37 +01001940 /*
1941 * Various filldir functions may end up calling back into
1942 * lookup_one_len() and the file system's ->lookup() method.
1943 * These expect i_mutex to be held, as it would within readdir.
1944 */
1945 host_err = mutex_lock_killable(&dir_inode->i_mutex);
1946 if (host_err)
1947 break;
1948
David Woodhouse14f7dd62008-07-31 20:29:12 +01001949 de = (struct buffered_dirent *)buf.dirent;
1950 while (size > 0) {
1951 offset = de->offset;
1952
1953 if (func(cdp, de->name, de->namlen, de->offset,
1954 de->ino, de->d_type))
David Woodhouse2f9092e2009-04-20 23:18:37 +01001955 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001956
1957 if (cdp->err != nfs_ok)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001958 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001959
1960 reclen = ALIGN(sizeof(*de) + de->namlen,
1961 sizeof(u64));
1962 size -= reclen;
1963 de = (struct buffered_dirent *)((char *)de + reclen);
1964 }
David Woodhouse2f9092e2009-04-20 23:18:37 +01001965 mutex_unlock(&dir_inode->i_mutex);
1966 if (size > 0) /* We bailed out early */
1967 break;
1968
David Woodhousec002a6c2008-08-17 17:21:18 +01001969 offset = vfs_llseek(file, 0, SEEK_CUR);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001970 }
1971
David Woodhouse14f7dd62008-07-31 20:29:12 +01001972 free_page((unsigned long)(buf.dirent));
David Woodhouse2628b762008-07-31 17:16:51 +01001973
1974 if (host_err)
1975 return nfserrno(host_err);
David Woodhouse14f7dd62008-07-31 20:29:12 +01001976
1977 *offsetp = offset;
1978 return cdp->err;
David Woodhouse2628b762008-07-31 17:16:51 +01001979}
1980
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981/*
1982 * Read entries from a directory.
1983 * The NFSv3/4 verifier we ignore for now.
1984 */
Al Viro6264d692006-10-19 23:28:58 -07001985__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08001987 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988{
Al Viro6264d692006-10-19 23:28:58 -07001989 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 struct file *file;
1991 loff_t offset = *offsetp;
1992
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001993 err = nfsd_open(rqstp, fhp, S_IFDIR, NFSD_MAY_READ, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 if (err)
1995 goto out;
1996
1997 offset = vfs_llseek(file, offset, 0);
1998 if (offset < 0) {
1999 err = nfserrno((int)offset);
2000 goto out_close;
2001 }
2002
David Woodhouse14f7dd62008-07-31 20:29:12 +01002003 err = nfsd_buffered_readdir(file, func, cdp, offsetp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004
2005 if (err == nfserr_eof || err == nfserr_toosmall)
2006 err = nfs_ok; /* can still be found in ->err */
2007out_close:
2008 nfsd_close(file);
2009out:
2010 return err;
2011}
2012
2013/*
2014 * Get file system stats
2015 * N.B. After this call fhp needs an fh_put
2016 */
Al Viro6264d692006-10-19 23:28:58 -07002017__be32
J. Bruce Fields04716e62008-08-07 13:00:20 -04002018nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019{
Christoph Hellwigebabe9a2010-07-07 18:53:11 +02002020 __be32 err;
2021
2022 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access);
Takashi Iwaif6360ef2010-08-13 15:53:49 +02002023 if (!err) {
2024 struct path path = {
2025 .mnt = fhp->fh_export->ex_path.mnt,
2026 .dentry = fhp->fh_dentry,
2027 };
2028 if (vfs_statfs(&path, stat))
2029 err = nfserr_io;
2030 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 return err;
2032}
2033
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002034static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002035{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002036 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002037}
2038
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039/*
2040 * Check for a user's access permissions to this inode.
2041 */
Al Viro6264d692006-10-19 23:28:58 -07002042__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07002043nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
2044 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045{
2046 struct inode *inode = dentry->d_inode;
2047 int err;
2048
J. Bruce Fieldsaea93392011-04-10 10:35:12 -04002049 if ((acc & NFSD_MAY_MASK) == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 return 0;
2051#if 0
2052 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
2053 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002054 (acc & NFSD_MAY_READ)? " read" : "",
2055 (acc & NFSD_MAY_WRITE)? " write" : "",
2056 (acc & NFSD_MAY_EXEC)? " exec" : "",
2057 (acc & NFSD_MAY_SATTR)? " sattr" : "",
2058 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
2059 (acc & NFSD_MAY_LOCK)? " lock" : "",
2060 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 inode->i_mode,
2062 IS_IMMUTABLE(inode)? " immut" : "",
2063 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08002064 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 dprintk(" owner %d/%d user %d/%d\n",
David Howells5cc0a842008-11-14 10:38:58 +11002066 inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067#endif
2068
2069 /* Normally we reject any write/sattr etc access on a read-only file
2070 * system. But if it is IRIX doing check on write-access for a
2071 * device special file, we ignore rofs.
2072 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002073 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
2074 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08002075 if (exp_rdonly(rqstp, exp) ||
2076 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002078 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 return nfserr_perm;
2080 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002081 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082 return nfserr_perm;
2083
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002084 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 /* If we cannot rely on authentication in NLM requests,
2086 * just allow locks, otherwise require read permission, or
2087 * ownership
2088 */
2089 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
2090 return 0;
2091 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002092 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 }
2094 /*
2095 * The file owner always gets access permission for accesses that
2096 * would normally be checked at open time. This is to make
2097 * file access work even when the client has done a fchmod(fd, 0).
2098 *
2099 * However, `cp foo bar' should fail nevertheless when bar is
2100 * readonly. A sensible way to do this might be to reject all
2101 * attempts to truncate a read-only file, because a creat() call
2102 * always implies file truncation.
2103 * ... but this isn't really fair. A process may reasonably call
2104 * ftruncate on an open file descriptor on a file with perm 000.
2105 * We must trust the client to do permission checking - using "ACCESS"
2106 * with NFSv3.
2107 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002108 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
David Howells5cc0a842008-11-14 10:38:58 +11002109 inode->i_uid == current_fsuid())
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 return 0;
2111
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002112 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Al Virof419a2e2008-07-22 00:07:17 -04002113 err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114
2115 /* Allow read access to binaries even when mode 111 */
2116 if (err == -EACCES && S_ISREG(inode->i_mode) &&
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002117 acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
Al Virof419a2e2008-07-22 00:07:17 -04002118 err = inode_permission(inode, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119
2120 return err? nfserrno(err) : 0;
2121}
2122
2123void
2124nfsd_racache_shutdown(void)
2125{
Jeff Layton54a66e52008-08-13 22:03:27 -04002126 struct raparms *raparm, *last_raparm;
2127 unsigned int i;
2128
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 dprintk("nfsd: freeing readahead buffers.\n");
Jeff Layton54a66e52008-08-13 22:03:27 -04002130
2131 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
2132 raparm = raparm_hash[i].pb_head;
2133 while(raparm) {
2134 last_raparm = raparm;
2135 raparm = raparm->p_next;
2136 kfree(last_raparm);
2137 }
2138 raparm_hash[i].pb_head = NULL;
2139 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140}
2141/*
2142 * Initialize readahead param cache
2143 */
2144int
2145nfsd_racache_init(int cache_size)
2146{
2147 int i;
Greg Banksfce14562006-10-04 02:15:49 -07002148 int j = 0;
2149 int nperbucket;
Jeff Layton54a66e52008-08-13 22:03:27 -04002150 struct raparms **raparm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151
Greg Banksfce14562006-10-04 02:15:49 -07002152
Jeff Layton54a66e52008-08-13 22:03:27 -04002153 if (raparm_hash[0].pb_head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002155 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
2156 if (nperbucket < 2)
2157 nperbucket = 2;
2158 cache_size = nperbucket * RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08002159
2160 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
Jeff Layton54a66e52008-08-13 22:03:27 -04002161
2162 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
Yan Burman4b3bb062006-12-08 02:39:41 -08002163 spin_lock_init(&raparm_hash[i].pb_lock);
Jeff Layton54a66e52008-08-13 22:03:27 -04002164
2165 raparm = &raparm_hash[i].pb_head;
2166 for (j = 0; j < nperbucket; j++) {
2167 *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL);
2168 if (!*raparm)
2169 goto out_nomem;
2170 raparm = &(*raparm)->p_next;
2171 }
2172 *raparm = NULL;
Yan Burman4b3bb062006-12-08 02:39:41 -08002173 }
2174
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175 nfsdstats.ra_size = cache_size;
2176 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002177
2178out_nomem:
2179 dprintk("nfsd: kmalloc failed, freeing readahead buffers\n");
2180 nfsd_racache_shutdown();
2181 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002183
2184#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2185struct posix_acl *
2186nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2187{
2188 struct inode *inode = fhp->fh_dentry->d_inode;
2189 char *name;
2190 void *value = NULL;
2191 ssize_t size;
2192 struct posix_acl *acl;
2193
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002194 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002195 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002196
2197 switch (type) {
2198 case ACL_TYPE_ACCESS:
2199 name = POSIX_ACL_XATTR_ACCESS;
2200 break;
2201 case ACL_TYPE_DEFAULT:
2202 name = POSIX_ACL_XATTR_DEFAULT;
2203 break;
2204 default:
2205 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002206 }
2207
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002208 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2209 if (size < 0)
2210 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002211
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002212 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002213 kfree(value);
2214 return acl;
2215}
2216
2217int
2218nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2219{
2220 struct inode *inode = fhp->fh_dentry->d_inode;
2221 char *name;
2222 void *value = NULL;
2223 size_t size;
2224 int error;
2225
Al Viroacfa4382008-12-04 10:06:33 -05002226 if (!IS_POSIXACL(inode) ||
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002227 !inode->i_op->setxattr || !inode->i_op->removexattr)
2228 return -EOPNOTSUPP;
2229 switch(type) {
2230 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002231 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002232 break;
2233 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002234 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002235 break;
2236 default:
2237 return -EOPNOTSUPP;
2238 }
2239
2240 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002241 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002242 value = kmalloc(size, GFP_KERNEL);
2243 if (!value)
2244 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002245 error = posix_acl_to_xattr(acl, value, size);
2246 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002247 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002248 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002249 } else
2250 size = 0;
2251
Dave Hansen18f335a2008-02-15 14:37:38 -08002252 error = mnt_want_write(fhp->fh_export->ex_path.mnt);
2253 if (error)
2254 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002255 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002256 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002257 else {
2258 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2259 error = 0;
2260 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002261 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002262 if (error == -ENODATA)
2263 error = 0;
2264 }
2265 }
Dave Hansen18f335a2008-02-15 14:37:38 -08002266 mnt_drop_write(fhp->fh_export->ex_path.mnt);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002267
2268getout:
2269 kfree(value);
2270 return error;
2271}
2272#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */