blob: c8bd9c3be7f747410622fd1172b2c7243886f838 [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;
Trond Myklebust11fcee022011-09-12 19:37:26 -0400171 if (nfsd4_is_junction(dentry))
172 return 1;
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400173 if (!(exp->ex_flags & NFSEXP_V4ROOT))
174 return 0;
175 return dentry->d_inode != NULL;
176}
177
Al Viro6264d692006-10-19 23:28:58 -0700178__be32
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700179nfsd_lookup_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400180 const char *name, unsigned int len,
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700181 struct svc_export **exp_ret, struct dentry **dentry_ret)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182{
183 struct svc_export *exp;
184 struct dentry *dparent;
185 struct dentry *dentry;
Al Viro6264d692006-10-19 23:28:58 -0700186 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
188 dprintk("nfsd: nfsd_lookup(fh %s, %.*s)\n", SVCFH_fmt(fhp), len,name);
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 dparent = fhp->fh_dentry;
191 exp = fhp->fh_export;
192 exp_get(exp);
193
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 /* Lookup the name, but don't follow links */
195 if (isdotent(name, len)) {
196 if (len==1)
197 dentry = dget(dparent);
Jan Blunck54775492008-02-14 19:38:39 -0800198 else if (dparent != exp->ex_path.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 dentry = dget_parent(dparent);
J. Bruce Fieldsfed83812009-09-26 16:53:01 -0400200 else if (!EX_NOHIDE(exp) && !nfsd_v4client(rqstp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 dentry = dget(dparent); /* .. == . just like at / */
202 else {
203 /* checking mountpoint crossing is very different when stepping up */
J. Bruce Fields289ede42009-09-26 20:32:24 -0400204 host_err = nfsd_lookup_parent(rqstp, dparent, &exp, &dentry);
205 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 }
208 } else {
209 fh_lock(fhp);
210 dentry = lookup_one_len(name, dparent, len);
Al Viro6264d692006-10-19 23:28:58 -0700211 host_err = PTR_ERR(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 if (IS_ERR(dentry))
213 goto out_nfserr;
214 /*
215 * check if we have crossed a mount point ...
216 */
J. Bruce Fields82ead7f2009-10-25 21:33:15 -0400217 if (nfsd_mountpoint(dentry, exp)) {
Al Viro6264d692006-10-19 23:28:58 -0700218 if ((host_err = nfsd_cross_mnt(rqstp, &dentry, &exp))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 dput(dentry);
220 goto out_nfserr;
221 }
222 }
223 }
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700224 *dentry_ret = dentry;
225 *exp_ret = exp;
226 return 0;
227
228out_nfserr:
229 exp_put(exp);
230 return nfserrno(host_err);
231}
232
233/*
234 * Look up one component of a pathname.
235 * N.B. After this call _both_ fhp and resfh need an fh_put
236 *
237 * If the lookup would cross a mountpoint, and the mounted filesystem
238 * is exported to the client with NFSEXP_NOHIDE, then the lookup is
239 * accepted as it stands and the mounted directory is
240 * returned. Otherwise the covered directory is returned.
241 * NOTE: this mountpoint crossing is not supported properly by all
242 * clients and is explicitly disallowed for NFSv3
243 * NeilBrown <neilb@cse.unsw.edu.au>
244 */
245__be32
246nfsd_lookup(struct svc_rqst *rqstp, struct svc_fh *fhp, const char *name,
Chuck Lever5a022fc2007-11-01 16:57:09 -0400247 unsigned int len, struct svc_fh *resfh)
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700248{
249 struct svc_export *exp;
250 struct dentry *dentry;
251 __be32 err;
252
J. Bruce Fields29a78a32011-04-09 11:28:53 -0400253 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
254 if (err)
255 return err;
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700256 err = nfsd_lookup_dentry(rqstp, fhp, name, len, &exp, &dentry);
257 if (err)
258 return err;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700259 err = check_nfsd_access(exp, rqstp);
260 if (err)
261 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 /*
263 * Note: we compose the file handle now, but as the
264 * dentry may be negative, it may need to be updated.
265 */
266 err = fh_compose(resfh, exp, dentry, fhp);
267 if (!err && !dentry->d_inode)
268 err = nfserr_noent;
Andy Adamson32c1eb02007-07-17 04:04:48 -0700269out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 exp_put(exp);
272 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273}
274
J. Bruce Fields4795bb32011-01-11 13:55:46 -0500275static int nfsd_break_lease(struct inode *inode)
276{
277 if (!S_ISREG(inode->i_mode))
278 return 0;
279 return break_lease(inode, O_WRONLY | O_NONBLOCK);
280}
281
Ben Myersf5019122010-02-17 14:05:11 -0600282/*
283 * Commit metadata changes to stable storage.
284 */
285static int
286commit_metadata(struct svc_fh *fhp)
287{
288 struct inode *inode = fhp->fh_dentry->d_inode;
289 const struct export_operations *export_ops = inode->i_sb->s_export_op;
Ben Myersf5019122010-02-17 14:05:11 -0600290
291 if (!EX_ISSYNC(fhp->fh_export))
292 return 0;
293
Christoph Hellwigc37650162010-10-06 10:48:20 +0200294 if (export_ops->commit_metadata)
295 return export_ops->commit_metadata(inode);
296 return sync_inode_metadata(inode, 1);
Ben Myersf5019122010-02-17 14:05:11 -0600297}
J. Bruce Fields6c0a6542007-07-17 04:04:47 -0700298
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299/*
300 * Set various file attributes.
301 * N.B. After this call fhp needs an fh_put
302 */
Al Viro6264d692006-10-19 23:28:58 -0700303__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
305 int check_guard, time_t guardtime)
306{
307 struct dentry *dentry;
308 struct inode *inode;
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200309 int accmode = NFSD_MAY_SATTR;
Al Viro175a4eb2011-07-26 03:30:54 -0400310 umode_t ftype = 0;
Al Viro6264d692006-10-19 23:28:58 -0700311 __be32 err;
312 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 int size_change = 0;
314
315 if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200316 accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 if (iap->ia_valid & ATTR_SIZE)
318 ftype = S_IFREG;
319
320 /* Get inode */
321 err = fh_verify(rqstp, fhp, ftype, accmode);
NeilBrown15b7a1b2005-11-07 01:00:23 -0800322 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 goto out;
324
325 dentry = fhp->fh_dentry;
326 inode = dentry->d_inode;
327
NeilBrown15b7a1b2005-11-07 01:00:23 -0800328 /* Ignore any mode updates on symlinks */
329 if (S_ISLNK(inode->i_mode))
330 iap->ia_valid &= ~ATTR_MODE;
331
332 if (!iap->ia_valid)
333 goto out;
334
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000335 /*
336 * NFSv2 does not differentiate between "set-[ac]time-to-now"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 * which only requires access, and "set-[ac]time-to-X" which
338 * requires ownership.
339 * So if it looks like it might be "set both to the same time which
340 * is close to now", and if inode_change_ok fails, then we
341 * convert to "set to now" instead of "set to explicit time"
342 *
343 * We only call inode_change_ok as the last test as technically
344 * it is not an interface that we should be using. It is only
345 * valid if the filesystem does not define it's own i_op->setattr.
346 */
347#define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
348#define MAX_TOUCH_TIME_ERROR (30*60)
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000349 if ((iap->ia_valid & BOTH_TIME_SET) == BOTH_TIME_SET &&
350 iap->ia_mtime.tv_sec == iap->ia_atime.tv_sec) {
351 /*
352 * Looks probable.
353 *
354 * Now just make sure time is in the right ballpark.
355 * Solaris, at least, doesn't seem to care what the time
356 * request is. We require it be within 30 minutes of now.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 */
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000358 time_t delta = iap->ia_atime.tv_sec - get_seconds();
359 if (delta < 0)
360 delta = -delta;
361 if (delta < MAX_TOUCH_TIME_ERROR &&
362 inode_change_ok(inode, iap) != 0) {
363 /*
364 * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME.
365 * This will cause notify_change to set these times
366 * to "now"
367 */
368 iap->ia_valid &= ~BOTH_TIME_SET;
369 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 }
371
Christoph Hellwig9c85fca2007-03-07 15:26:25 +0000372 /*
373 * The size case is special.
374 * It changes the file as well as the attributes.
375 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 if (iap->ia_valid & ATTR_SIZE) {
377 if (iap->ia_size < inode->i_size) {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200378 err = nfsd_permission(rqstp, fhp->fh_export, dentry,
379 NFSD_MAY_TRUNC|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 if (err)
381 goto out;
382 }
383
Al Viro6264d692006-10-19 23:28:58 -0700384 host_err = get_write_access(inode);
385 if (host_err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 goto out_nfserr;
387
388 size_change = 1;
Al Viro6264d692006-10-19 23:28:58 -0700389 host_err = locks_verify_truncate(inode, NULL, iap->ia_size);
390 if (host_err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 put_write_access(inode);
392 goto out_nfserr;
393 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 }
395
Jeff Laytonca456252008-04-16 16:28:47 -0400396 /* sanitize the mode change */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 if (iap->ia_valid & ATTR_MODE) {
398 iap->ia_mode &= S_IALLUGO;
Jeff Laytondee32092008-04-16 16:28:46 -0400399 iap->ia_mode |= (inode->i_mode & ~S_IALLUGO);
Jeff Laytonca456252008-04-16 16:28:47 -0400400 }
401
402 /* Revoke setuid/setgid on chown */
Sachin S. Prabhu0953e622009-02-23 16:22:03 +0000403 if (!S_ISDIR(inode->i_mode) &&
404 (((iap->ia_valid & ATTR_UID) && iap->ia_uid != inode->i_uid) ||
405 ((iap->ia_valid & ATTR_GID) && iap->ia_gid != inode->i_gid))) {
Jeff Laytonca456252008-04-16 16:28:47 -0400406 iap->ia_valid |= ATTR_KILL_PRIV;
407 if (iap->ia_valid & ATTR_MODE) {
408 /* we're setting mode too, just clear the s*id bits */
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700409 iap->ia_mode &= ~S_ISUID;
Jeff Laytonca456252008-04-16 16:28:47 -0400410 if (iap->ia_mode & S_IXGRP)
411 iap->ia_mode &= ~S_ISGID;
412 } else {
413 /* set ATTR_KILL_* bits and let VFS handle it */
414 iap->ia_valid |= (ATTR_KILL_SUID | ATTR_KILL_SGID);
Jeff Layton8a0ce7d2007-10-18 03:05:19 -0700415 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 }
417
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 /* Change the attributes. */
419
420 iap->ia_valid |= ATTR_CTIME;
421
422 err = nfserr_notsync;
423 if (!check_guard || guardtime == inode->i_ctime.tv_sec) {
J. Bruce Fields4795bb32011-01-11 13:55:46 -0500424 host_err = nfsd_break_lease(inode);
J. Bruce Fields6a76beb2011-01-11 12:54:39 -0500425 if (host_err)
426 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 fh_lock(fhp);
J. Bruce Fields6a76beb2011-01-11 12:54:39 -0500428
Al Viro6264d692006-10-19 23:28:58 -0700429 host_err = notify_change(dentry, iap);
430 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 fh_unlock(fhp);
432 }
433 if (size_change)
434 put_write_access(inode);
435 if (!err)
Christoph Hellwigb160fda2010-06-01 21:59:18 +0200436 commit_metadata(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437out:
438 return err;
439
440out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700441 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 goto out;
443}
444
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800445#if defined(CONFIG_NFSD_V2_ACL) || \
446 defined(CONFIG_NFSD_V3_ACL) || \
447 defined(CONFIG_NFSD_V4)
448static ssize_t nfsd_getxattr(struct dentry *dentry, char *key, void **buf)
449{
450 ssize_t buflen;
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530451 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800453 buflen = vfs_getxattr(dentry, key, NULL, 0);
454 if (buflen <= 0)
455 return buflen;
456
457 *buf = kmalloc(buflen, GFP_KERNEL);
458 if (!*buf)
459 return -ENOMEM;
460
Krishna Kumar6c6a4262008-10-22 14:48:36 +0530461 ret = vfs_getxattr(dentry, key, *buf, buflen);
462 if (ret < 0)
463 kfree(*buf);
464 return ret;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800465}
466#endif
467
468#if defined(CONFIG_NFSD_V4)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469static int
470set_nfsv4_acl_one(struct dentry *dentry, struct posix_acl *pacl, char *key)
471{
472 int len;
473 size_t buflen;
474 char *buf = NULL;
475 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477 buflen = posix_acl_xattr_size(pacl->a_count);
478 buf = kmalloc(buflen, GFP_KERNEL);
479 error = -ENOMEM;
480 if (buf == NULL)
481 goto out;
482
483 len = posix_acl_to_xattr(pacl, buf, buflen);
484 if (len < 0) {
485 error = len;
486 goto out;
487 }
488
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800489 error = vfs_setxattr(dentry, key, buf, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490out:
491 kfree(buf);
492 return error;
493}
494
Al Viro6264d692006-10-19 23:28:58 -0700495__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496nfsd4_set_nfs4_acl(struct svc_rqst *rqstp, struct svc_fh *fhp,
497 struct nfs4_acl *acl)
498{
Al Viro6264d692006-10-19 23:28:58 -0700499 __be32 error;
500 int host_error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 struct dentry *dentry;
502 struct inode *inode;
503 struct posix_acl *pacl = NULL, *dpacl = NULL;
504 unsigned int flags = 0;
505
506 /* Get inode */
J. Bruce Fieldse281d812011-08-15 16:57:07 -0400507 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_SATTR);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 if (error)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700509 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
511 dentry = fhp->fh_dentry;
512 inode = dentry->d_inode;
513 if (S_ISDIR(inode->i_mode))
514 flags = NFS4_ACL_DIR;
515
Al Viro6264d692006-10-19 23:28:58 -0700516 host_error = nfs4_acl_nfsv4_to_posix(acl, &pacl, &dpacl, flags);
517 if (host_error == -EINVAL) {
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700518 return nfserr_attrnotsupp;
Al Viro6264d692006-10-19 23:28:58 -0700519 } else if (host_error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 goto out_nfserr;
521
Al Viro6264d692006-10-19 23:28:58 -0700522 host_error = set_nfsv4_acl_one(dentry, pacl, POSIX_ACL_XATTR_ACCESS);
523 if (host_error < 0)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700524 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700526 if (S_ISDIR(inode->i_mode))
Al Viro6264d692006-10-19 23:28:58 -0700527 host_error = set_nfsv4_acl_one(dentry, dpacl, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700529out_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 posix_acl_release(pacl);
531 posix_acl_release(dpacl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532out_nfserr:
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800533 if (host_error == -EOPNOTSUPP)
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700534 return nfserr_attrnotsupp;
J. Bruce Fieldsf34f9242007-02-16 01:28:34 -0800535 else
J. Bruce Fields4b2ca382007-07-17 04:04:37 -0700536 return nfserrno(host_error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537}
538
539static struct posix_acl *
540_get_posix_acl(struct dentry *dentry, char *key)
541{
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800542 void *buf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct posix_acl *pacl = NULL;
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800544 int buflen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Christoph Hellwig5be196e2006-01-09 20:51:55 -0800546 buflen = nfsd_getxattr(dentry, key, &buf);
547 if (!buflen)
548 buflen = -ENODATA;
549 if (buflen <= 0)
550 return ERR_PTR(buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
552 pacl = posix_acl_from_xattr(buf, buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 kfree(buf);
554 return pacl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555}
556
557int
558nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_acl **acl)
559{
560 struct inode *inode = dentry->d_inode;
561 int error = 0;
562 struct posix_acl *pacl = NULL, *dpacl = NULL;
563 unsigned int flags = 0;
564
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700565 pacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_ACCESS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 if (IS_ERR(pacl) && PTR_ERR(pacl) == -ENODATA)
567 pacl = posix_acl_from_mode(inode->i_mode, GFP_KERNEL);
568 if (IS_ERR(pacl)) {
569 error = PTR_ERR(pacl);
570 pacl = NULL;
571 goto out;
572 }
573
574 if (S_ISDIR(inode->i_mode)) {
Christoph Hellwig9a59f452005-06-23 00:10:19 -0700575 dpacl = _get_posix_acl(dentry, POSIX_ACL_XATTR_DEFAULT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 if (IS_ERR(dpacl) && PTR_ERR(dpacl) == -ENODATA)
577 dpacl = NULL;
578 else if (IS_ERR(dpacl)) {
579 error = PTR_ERR(dpacl);
580 dpacl = NULL;
581 goto out;
582 }
583 flags = NFS4_ACL_DIR;
584 }
585
586 *acl = nfs4_acl_posix_to_nfsv4(pacl, dpacl, flags);
587 if (IS_ERR(*acl)) {
588 error = PTR_ERR(*acl);
589 *acl = NULL;
590 }
591 out:
592 posix_acl_release(pacl);
593 posix_acl_release(dpacl);
594 return error;
595}
596
Chuck Lever9b4146e2012-01-04 16:26:43 -0500597/*
598 * NFS junction information is stored in an extended attribute.
599 */
600#define NFSD_JUNCTION_XATTR_NAME XATTR_TRUSTED_PREFIX "junction.nfs"
601
602/**
603 * nfsd4_is_junction - Test if an object could be an NFS junction
604 *
605 * @dentry: object to test
606 *
607 * Returns 1 if "dentry" appears to contain NFS junction information.
608 * Otherwise 0 is returned.
609 */
Trond Myklebust11fcee022011-09-12 19:37:26 -0400610int nfsd4_is_junction(struct dentry *dentry)
611{
612 struct inode *inode = dentry->d_inode;
613
614 if (inode == NULL)
615 return 0;
616 if (inode->i_mode & S_IXUGO)
617 return 0;
618 if (!(inode->i_mode & S_ISVTX))
619 return 0;
Chuck Lever9b4146e2012-01-04 16:26:43 -0500620 if (vfs_getxattr(dentry, NFSD_JUNCTION_XATTR_NAME, NULL, 0) <= 0)
Trond Myklebust11fcee022011-09-12 19:37:26 -0400621 return 0;
622 return 1;
623}
J. Bruce Fields6a85d6c2010-07-06 12:39:12 -0400624#endif /* defined(CONFIG_NFSD_V4) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626#ifdef CONFIG_NFSD_V3
627/*
628 * Check server access rights to a file system object
629 */
630struct accessmap {
631 u32 access;
632 int how;
633};
634static struct accessmap nfs3_regaccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200635 { NFS3_ACCESS_READ, NFSD_MAY_READ },
636 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
637 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_TRUNC },
638 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
640 { 0, 0 }
641};
642
643static struct accessmap nfs3_diraccess[] = {
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200644 { NFS3_ACCESS_READ, NFSD_MAY_READ },
645 { NFS3_ACCESS_LOOKUP, NFSD_MAY_EXEC },
646 { NFS3_ACCESS_MODIFY, NFSD_MAY_EXEC|NFSD_MAY_WRITE|NFSD_MAY_TRUNC},
647 { NFS3_ACCESS_EXTEND, NFSD_MAY_EXEC|NFSD_MAY_WRITE },
648 { NFS3_ACCESS_DELETE, NFSD_MAY_REMOVE },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
650 { 0, 0 }
651};
652
653static struct accessmap nfs3_anyaccess[] = {
654 /* Some clients - Solaris 2.6 at least, make an access call
655 * to the server to check for access for things like /dev/null
656 * (which really, the server doesn't care about). So
657 * We provide simple access checking for them, looking
658 * mainly at mode bits, and we make sure to ignore read-only
659 * filesystem checks
660 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200661 { NFS3_ACCESS_READ, NFSD_MAY_READ },
662 { NFS3_ACCESS_EXECUTE, NFSD_MAY_EXEC },
663 { NFS3_ACCESS_MODIFY, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
664 { NFS3_ACCESS_EXTEND, NFSD_MAY_WRITE|NFSD_MAY_LOCAL_ACCESS },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
666 { 0, 0 }
667};
668
Al Viro6264d692006-10-19 23:28:58 -0700669__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670nfsd_access(struct svc_rqst *rqstp, struct svc_fh *fhp, u32 *access, u32 *supported)
671{
672 struct accessmap *map;
673 struct svc_export *export;
674 struct dentry *dentry;
675 u32 query, result = 0, sresult = 0;
Al Viro6264d692006-10-19 23:28:58 -0700676 __be32 error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Miklos Szeredi8837abc2008-06-16 13:20:29 +0200678 error = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 if (error)
680 goto out;
681
682 export = fhp->fh_export;
683 dentry = fhp->fh_dentry;
684
685 if (S_ISREG(dentry->d_inode->i_mode))
686 map = nfs3_regaccess;
687 else if (S_ISDIR(dentry->d_inode->i_mode))
688 map = nfs3_diraccess;
689 else
690 map = nfs3_anyaccess;
691
692
693 query = *access;
694 for (; map->access; map++) {
695 if (map->access & query) {
Al Viro6264d692006-10-19 23:28:58 -0700696 __be32 err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
698 sresult |= map->access;
699
J. Bruce Fields0ec757d2007-07-17 04:04:48 -0700700 err2 = nfsd_permission(rqstp, export, dentry, map->how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 switch (err2) {
702 case nfs_ok:
703 result |= map->access;
704 break;
705
706 /* the following error codes just mean the access was not allowed,
707 * rather than an error occurred */
708 case nfserr_rofs:
709 case nfserr_acces:
710 case nfserr_perm:
711 /* simply don't "or" in the access bit. */
712 break;
713 default:
714 error = err2;
715 goto out;
716 }
717 }
718 }
719 *access = result;
720 if (supported)
721 *supported = sresult;
722
723 out:
724 return error;
725}
726#endif /* CONFIG_NFSD_V3 */
727
J. Bruce Fields105f4622011-06-07 11:50:23 -0400728static int nfsd_open_break_lease(struct inode *inode, int access)
729{
730 unsigned int mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
J. Bruce Fields105f4622011-06-07 11:50:23 -0400732 if (access & NFSD_MAY_NOT_BREAK_LEASE)
733 return 0;
734 mode = (access & NFSD_MAY_WRITE) ? O_WRONLY : O_RDONLY;
735 return break_lease(inode, mode | O_NONBLOCK);
736}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737
738/*
739 * Open an existing file or directory.
Bernd Schubert999448a2012-03-18 22:44:49 -0400740 * The may_flags argument indicates the type of open (read/write/lock)
741 * and additional flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 * N.B. After this call fhp needs an fh_put
743 */
Al Viro6264d692006-10-19 23:28:58 -0700744__be32
Al Viro175a4eb2011-07-26 03:30:54 -0400745nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, umode_t type,
Bernd Schubert999448a2012-03-18 22:44:49 -0400746 int may_flags, struct file **filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747{
748 struct dentry *dentry;
749 struct inode *inode;
Al Viro6264d692006-10-19 23:28:58 -0700750 int flags = O_RDONLY|O_LARGEFILE;
751 __be32 err;
Jeff Layton91885252010-03-19 08:06:28 -0400752 int host_err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
David Howellse0e81732009-09-02 09:13:40 +0100754 validate_process_creds();
755
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 /*
757 * If we get here, then the client has already done an "open",
758 * and (hopefully) checked permission - so allow OWNER_OVERRIDE
759 * in case a chmod has now revoked permission.
760 */
Bernd Schubert999448a2012-03-18 22:44:49 -0400761 err = fh_verify(rqstp, fhp, type, may_flags | NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 if (err)
763 goto out;
764
765 dentry = fhp->fh_dentry;
766 inode = dentry->d_inode;
767
768 /* Disallow write access to files with the append-only bit set
769 * or any access when mandatory locking enabled
770 */
771 err = nfserr_perm;
Bernd Schubert999448a2012-03-18 22:44:49 -0400772 if (IS_APPEND(inode) && (may_flags & NFSD_MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 goto out;
J. Bruce Fields5e7fc432007-10-02 14:18:12 -0400774 /*
775 * We must ignore files (but only files) which might have mandatory
776 * locks on them because there is no way to know if the accesser has
777 * the lock.
778 */
779 if (S_ISREG((inode)->i_mode) && mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 goto out;
781
782 if (!inode->i_fop)
783 goto out;
784
Bernd Schubert999448a2012-03-18 22:44:49 -0400785 host_err = nfsd_open_break_lease(inode, may_flags);
Al Viro6264d692006-10-19 23:28:58 -0700786 if (host_err) /* NOMEM or WOULDBLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 goto out_nfserr;
788
Bernd Schubert999448a2012-03-18 22:44:49 -0400789 if (may_flags & NFSD_MAY_WRITE) {
790 if (may_flags & NFSD_MAY_READ)
J. Bruce Fields9ecb6a02006-06-30 01:56:17 -0700791 flags = O_RDWR|O_LARGEFILE;
792 else
793 flags = O_WRONLY|O_LARGEFILE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 }
Jan Blunck54775492008-02-14 19:38:39 -0800795 *filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
David Howells033a6662009-07-02 14:35:32 +0100796 flags, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 if (IS_ERR(*filp))
Al Viro6264d692006-10-19 23:28:58 -0700798 host_err = PTR_ERR(*filp);
Bernd Schubert06effdb2012-03-18 22:44:50 -0400799 else {
Bernd Schubert999448a2012-03-18 22:44:49 -0400800 host_err = ima_file_check(*filp, may_flags);
801
Bernd Schubert06effdb2012-03-18 22:44:50 -0400802 if (may_flags & NFSD_MAY_64BIT_COOKIE)
803 (*filp)->f_mode |= FMODE_64BITHASH;
804 else
805 (*filp)->f_mode |= FMODE_32BITHASH;
806 }
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -0700809 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810out:
David Howellse0e81732009-09-02 09:13:40 +0100811 validate_process_creds();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 return err;
813}
814
815/*
816 * Close a file.
817 */
818void
819nfsd_close(struct file *filp)
820{
821 fput(filp);
822}
823
J. Bruce Fields9a8d2482009-01-06 13:37:03 -0500824/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 * Obtain the readahead parameters for the file
826 * specified by (dev, ino).
827 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
829static inline struct raparms *
830nfsd_get_raparms(dev_t dev, ino_t ino)
831{
832 struct raparms *ra, **rap, **frap = NULL;
833 int depth = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700834 unsigned int hash;
835 struct raparm_hbucket *rab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
Greg Banksfce14562006-10-04 02:15:49 -0700837 hash = jhash_2words(dev, ino, 0xfeedbeef) & RAPARM_HASH_MASK;
838 rab = &raparm_hash[hash];
839
840 spin_lock(&rab->pb_lock);
841 for (rap = &rab->pb_head; (ra = *rap); rap = &ra->p_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 if (ra->p_ino == ino && ra->p_dev == dev)
843 goto found;
844 depth++;
845 if (ra->p_count == 0)
846 frap = rap;
847 }
Konstantin Khorenko3aa6e0a2011-02-01 17:16:29 +0300848 depth = nfsdstats.ra_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 if (!frap) {
Greg Banksfce14562006-10-04 02:15:49 -0700850 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 return NULL;
852 }
853 rap = frap;
854 ra = *frap;
855 ra->p_dev = dev;
856 ra->p_ino = ino;
857 ra->p_set = 0;
Greg Banksfce14562006-10-04 02:15:49 -0700858 ra->p_hindex = hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859found:
Greg Banksfce14562006-10-04 02:15:49 -0700860 if (rap != &rab->pb_head) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 *rap = ra->p_next;
Greg Banksfce14562006-10-04 02:15:49 -0700862 ra->p_next = rab->pb_head;
863 rab->pb_head = ra;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 }
865 ra->p_count++;
866 nfsdstats.ra_depth[depth*10/nfsdstats.ra_size]++;
Greg Banksfce14562006-10-04 02:15:49 -0700867 spin_unlock(&rab->pb_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 return ra;
869}
870
871/*
Jens Axboecf8208d2007-06-12 21:22:14 +0200872 * Grab and keep cached pages associated with a file in the svc_rqst
873 * so that they can be passed to the network sendmsg/sendpage routines
874 * directly. They will be released after the sending has completed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 */
876static int
Jens Axboecf8208d2007-06-12 21:22:14 +0200877nfsd_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
878 struct splice_desc *sd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879{
Jens Axboecf8208d2007-06-12 21:22:14 +0200880 struct svc_rqst *rqstp = sd->u.data;
NeilBrown44524352006-10-04 02:15:46 -0700881 struct page **pp = rqstp->rq_respages + rqstp->rq_resused;
Jens Axboecf8208d2007-06-12 21:22:14 +0200882 struct page *page = buf->page;
883 size_t size;
Jens Axboecf8208d2007-06-12 21:22:14 +0200884
885 size = sd->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886
887 if (rqstp->rq_res.page_len == 0) {
888 get_page(page);
NeilBrown44524352006-10-04 02:15:46 -0700889 put_page(*pp);
890 *pp = page;
891 rqstp->rq_resused++;
Jens Axboecf8208d2007-06-12 21:22:14 +0200892 rqstp->rq_res.page_base = buf->offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 rqstp->rq_res.page_len = size;
NeilBrown44524352006-10-04 02:15:46 -0700894 } else if (page != pp[-1]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 get_page(page);
NeilBrown250f3912007-01-26 00:56:59 -0800896 if (*pp)
897 put_page(*pp);
NeilBrown44524352006-10-04 02:15:46 -0700898 *pp = page;
899 rqstp->rq_resused++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 rqstp->rq_res.page_len += size;
NeilBrown44524352006-10-04 02:15:46 -0700901 } else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 rqstp->rq_res.page_len += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 return size;
905}
906
Jens Axboecf8208d2007-06-12 21:22:14 +0200907static int nfsd_direct_splice_actor(struct pipe_inode_info *pipe,
908 struct splice_desc *sd)
909{
910 return __splice_from_pipe(pipe, sd, nfsd_splice_actor);
911}
912
Al Viro6264d692006-10-19 23:28:58 -0700913static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
915 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
916{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -0700918 __be32 err;
919 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
921 err = nfserr_perm;
Dave Hansena8754be2007-10-16 23:31:15 -0700922
Jens Axboecf8208d2007-06-12 21:22:14 +0200923 if (file->f_op->splice_read && rqstp->rq_splice_ok) {
924 struct splice_desc sd = {
925 .len = 0,
926 .total_len = *count,
927 .pos = offset,
928 .u.data = rqstp,
929 };
930
Jens Axboe4fbef202007-07-13 22:42:20 +0200931 rqstp->rq_resused = 1;
Jens Axboecf8208d2007-06-12 21:22:14 +0200932 host_err = splice_direct_to_actor(file, &sd, nfsd_direct_splice_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 } else {
934 oldfs = get_fs();
935 set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -0700936 host_err = vfs_readv(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 set_fs(oldfs);
938 }
939
Al Viro6264d692006-10-19 23:28:58 -0700940 if (host_err >= 0) {
941 nfsdstats.io_read += host_err;
942 *count = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 err = 0;
Eric Paris2a12a9d2009-12-17 21:24:21 -0500944 fsnotify_access(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 } else
Al Viro6264d692006-10-19 23:28:58 -0700946 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 return err;
948}
949
Neil Brown9f708e42006-01-06 00:19:59 -0800950static void kill_suid(struct dentry *dentry)
951{
952 struct iattr ia;
Serge E. Hallynb5376772007-10-16 23:31:36 -0700953 ia.ia_valid = ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Neil Brown9f708e42006-01-06 00:19:59 -0800954
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800955 mutex_lock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800956 notify_change(dentry, &ia);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800957 mutex_unlock(&dentry->d_inode->i_mutex);
Neil Brown9f708e42006-01-06 00:19:59 -0800958}
959
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700960/*
961 * Gathered writes: If another process is currently writing to the file,
962 * there's a high chance this is another nfsd (triggered by a bulk write
963 * from a client's biod). Rather than syncing the file with each write
964 * request, we sleep for 10 msec.
965 *
966 * I don't know if this roughly approximates C. Juszak's idea of
967 * gathered writes, but it's a nice and simple solution (IMHO), and it
968 * seems to work:-)
969 *
970 * Note: we do this only in the NFSv2 case, since v3 and higher have a
971 * better tool (separate unstable writes and commits) for solving this
972 * problem.
973 */
974static int wait_for_concurrent_writes(struct file *file)
975{
976 struct inode *inode = file->f_path.dentry->d_inode;
977 static ino_t last_ino;
978 static dev_t last_dev;
979 int err = 0;
980
981 if (atomic_read(&inode->i_writecount) > 1
982 || (last_ino == inode->i_ino && last_dev == inode->i_sb->s_dev)) {
983 dprintk("nfsd: write defer %d\n", task_pid_nr(current));
984 msleep(10);
985 dprintk("nfsd: write resume %d\n", task_pid_nr(current));
986 }
987
988 if (inode->i_state & I_DIRTY) {
989 dprintk("nfsd: write sync %d\n", task_pid_nr(current));
Christoph Hellwig8018ab02010-03-22 17:32:25 +0100990 err = vfs_fsync(file, 0);
J. Bruce Fieldsd911df72009-06-15 16:03:53 -0700991 }
992 last_ino = inode->i_ino;
993 last_dev = inode->i_sb->s_dev;
994 return err;
995}
996
Al Viro6264d692006-10-19 23:28:58 -0700997static __be32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
999 loff_t offset, struct kvec *vec, int vlen,
David Shaw31dec252009-03-05 20:16:14 -05001000 unsigned long *cnt, int *stablep)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
1002 struct svc_export *exp;
1003 struct dentry *dentry;
1004 struct inode *inode;
1005 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001006 __be32 err = 0;
1007 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 int stable = *stablep;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001009 int use_wgather;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Josef "Jeff" Sipek7eaa36e2006-12-08 02:36:41 -08001011 dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 inode = dentry->d_inode;
1013 exp = fhp->fh_export;
1014
1015 /*
1016 * Request sync writes if
1017 * - the sync export option has been set, or
1018 * - the client requested O_SYNC behavior (NFSv3 feature).
1019 * - The file system doesn't support fsync().
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001020 * When NFSv2 gathered writes have been configured for this volume,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 * flushing the data to disk is handled separately below.
1022 */
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001023 use_wgather = (rqstp->rq_vers == 2) && EX_WGATHER(exp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
Harvey Harrison3ba15142008-02-20 12:49:02 -08001025 if (!file->f_op->fsync) {/* COMMIT3 cannot work */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 stable = 2;
1027 *stablep = 2; /* FILE_SYNC */
1028 }
1029
1030 if (!EX_ISSYNC(exp))
1031 stable = 0;
Trond Myklebust48e03bc2009-06-05 12:35:15 -04001032 if (stable && !use_wgather) {
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001033 spin_lock(&file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 file->f_flags |= O_SYNC;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001035 spin_unlock(&file->f_lock);
1036 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
1038 /* Write the data. */
1039 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro6264d692006-10-19 23:28:58 -07001040 host_err = vfs_writev(file, (struct iovec __user *)vec, vlen, &offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 set_fs(oldfs);
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001042 if (host_err < 0)
1043 goto out_nfserr;
1044 *cnt = host_err;
1045 nfsdstats.io_write += host_err;
Eric Paris2a12a9d2009-12-17 21:24:21 -05001046 fsnotify_modify(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
1048 /* clear setuid/setgid flag after write */
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001049 if (inode->i_mode & (S_ISUID | S_ISGID))
Neil Brown9f708e42006-01-06 00:19:59 -08001050 kill_suid(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001052 if (stable && use_wgather)
J. Bruce Fieldsd911df72009-06-15 16:03:53 -07001053 host_err = wait_for_concurrent_writes(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
J. Bruce Fieldse4636d52009-06-15 19:07:13 -07001055out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001056 dprintk("nfsd: write complete host_err=%d\n", host_err);
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001057 if (host_err >= 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 err = 0;
Wei Yongjuna0d24b22009-05-19 12:03:15 +08001059 else
Al Viro6264d692006-10-19 23:28:58 -07001060 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 return err;
1062}
1063
1064/*
1065 * Read data from a file. count must contain the requested read count
1066 * on entry. On return, *count contains the number of bytes actually read.
1067 * N.B. After this call fhp needs an fh_put
1068 */
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001069__be32 nfsd_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
J. Bruce Fieldsfa0a2122010-07-27 16:48:54 -04001070 loff_t offset, struct kvec *vec, int vlen, unsigned long *count)
1071{
1072 struct file *file;
1073 struct inode *inode;
1074 struct raparms *ra;
1075 __be32 err;
1076
1077 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
1078 if (err)
1079 return err;
1080
1081 inode = file->f_path.dentry->d_inode;
1082
1083 /* Get readahead parameters */
1084 ra = nfsd_get_raparms(inode->i_sb->s_dev, inode->i_ino);
1085
1086 if (ra && ra->p_set)
1087 file->f_ra = ra->p_ra;
1088
1089 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
1090
1091 /* Write back readahead params */
1092 if (ra) {
1093 struct raparm_hbucket *rab = &raparm_hash[ra->p_hindex];
1094 spin_lock(&rab->pb_lock);
1095 ra->p_ra = file->f_ra;
1096 ra->p_set = 1;
1097 ra->p_count--;
1098 spin_unlock(&rab->pb_lock);
1099 }
1100
1101 nfsd_close(file);
1102 return err;
1103}
1104
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001105/* As above, but use the provided file descriptor. */
Al Viro6264d692006-10-19 23:28:58 -07001106__be32
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001107nfsd_read_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 loff_t offset, struct kvec *vec, int vlen,
1109 unsigned long *count)
1110{
Al Viro6264d692006-10-19 23:28:58 -07001111 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001114 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001115 NFSD_MAY_READ|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 if (err)
1117 goto out;
1118 err = nfsd_vfs_read(rqstp, fhp, file, offset, vec, vlen, count);
J. Bruce Fields039a87c2010-07-30 11:33:32 -04001119 } else /* Note file may still be NULL in NFSv4 special stateid case: */
1120 err = nfsd_read(rqstp, fhp, offset, vec, vlen, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121out:
1122 return err;
1123}
1124
1125/*
1126 * Write data to a file.
1127 * The stable flag requests synchronous writes.
1128 * N.B. After this call fhp needs an fh_put
1129 */
Al Viro6264d692006-10-19 23:28:58 -07001130__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131nfsd_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
David Shaw31dec252009-03-05 20:16:14 -05001132 loff_t offset, struct kvec *vec, int vlen, unsigned long *cnt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 int *stablep)
1134{
Al Viro6264d692006-10-19 23:28:58 -07001135 __be32 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137 if (file) {
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07001138 err = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001139 NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 if (err)
1141 goto out;
1142 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen, cnt,
1143 stablep);
1144 } else {
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001145 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_WRITE, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 if (err)
1147 goto out;
1148
1149 if (cnt)
1150 err = nfsd_vfs_write(rqstp, fhp, file, offset, vec, vlen,
1151 cnt, stablep);
1152 nfsd_close(file);
1153 }
1154out:
1155 return err;
1156}
1157
1158#ifdef CONFIG_NFSD_V3
1159/*
1160 * Commit all pending writes to stable storage.
Trond Myklebustaa696a62010-01-29 16:44:25 -05001161 *
1162 * Note: we only guarantee that data that lies within the range specified
1163 * by the 'offset' and 'count' parameters will be synced.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 *
1165 * Unfortunately we cannot lock the file to make sure we return full WCC
1166 * data to the client, as locking happens lower down in the filesystem.
1167 */
Al Viro6264d692006-10-19 23:28:58 -07001168__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp,
1170 loff_t offset, unsigned long count)
1171{
1172 struct file *file;
Trond Myklebustaa696a62010-01-29 16:44:25 -05001173 loff_t end = LLONG_MAX;
1174 __be32 err = nfserr_inval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
Trond Myklebustaa696a62010-01-29 16:44:25 -05001176 if (offset < 0)
1177 goto out;
1178 if (count != 0) {
1179 end = offset + (loff_t)count - 1;
1180 if (end < offset)
1181 goto out;
1182 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Jeff Layton91885252010-03-19 08:06:28 -04001184 err = nfsd_open(rqstp, fhp, S_IFREG,
1185 NFSD_MAY_WRITE|NFSD_MAY_NOT_BREAK_LEASE, &file);
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001186 if (err)
Trond Myklebustaa696a62010-01-29 16:44:25 -05001187 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 if (EX_ISSYNC(fhp->fh_export)) {
Christoph Hellwig8018ab02010-03-22 17:32:25 +01001189 int err2 = vfs_fsync_range(file, offset, end, 0);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001190
1191 if (err2 != -EINVAL)
1192 err = nfserrno(err2);
1193 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 err = nfserr_notsupp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 }
1196
1197 nfsd_close(file);
Trond Myklebustaa696a62010-01-29 16:44:25 -05001198out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 return err;
1200}
1201#endif /* CONFIG_NFSD_V3 */
1202
Adrian Bunkf2b0dee2008-02-13 23:30:26 +02001203static __be32
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001204nfsd_create_setattr(struct svc_rqst *rqstp, struct svc_fh *resfhp,
1205 struct iattr *iap)
1206{
1207 /*
1208 * Mode has already been set earlier in create:
1209 */
1210 iap->ia_valid &= ~ATTR_MODE;
1211 /*
1212 * Setting uid/gid works only for root. Irix appears to
1213 * send along the gid on create when it tries to implement
1214 * setgid directories via NFS:
1215 */
David Howells5cc0a842008-11-14 10:38:58 +11001216 if (current_fsuid() != 0)
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001217 iap->ia_valid &= ~(ATTR_UID|ATTR_GID);
1218 if (iap->ia_valid)
1219 return nfsd_setattr(rqstp, resfhp, iap, 0, (time_t)0);
1220 return 0;
1221}
1222
wengang wang4ac35c22009-02-10 11:27:51 +08001223/* HPUX client sometimes creates a file in mode 000, and sets size to 0.
1224 * setting size to 0 may fail for some specific file systems by the permission
1225 * checking which requires WRITE permission but the mode is 000.
1226 * we ignore the resizing(to 0) on the just new created file, since the size is
1227 * 0 after file created.
1228 *
1229 * call this only after vfs_create() is called.
1230 * */
1231static void
1232nfsd_check_ignore_resizing(struct iattr *iap)
1233{
1234 if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0))
1235 iap->ia_valid &= ~ATTR_SIZE;
1236}
1237
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238/*
1239 * Create a file (regular, directory, device, fifo); UNIX sockets
1240 * not yet implemented.
1241 * If the response fh has been verified, the parent directory should
1242 * already be locked. Note that the parent directory is left locked.
1243 *
1244 * N.B. Every call to nfsd_create needs an fh_put for _both_ fhp and resfhp
1245 */
Al Viro6264d692006-10-19 23:28:58 -07001246__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
1248 char *fname, int flen, struct iattr *iap,
1249 int type, dev_t rdev, struct svc_fh *resfhp)
1250{
1251 struct dentry *dentry, *dchild = NULL;
1252 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001253 __be32 err;
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001254 __be32 err2;
Al Viro6264d692006-10-19 23:28:58 -07001255 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
1257 err = nfserr_perm;
1258 if (!flen)
1259 goto out;
1260 err = nfserr_exist;
1261 if (isdotent(fname, flen))
1262 goto out;
1263
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001264 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 if (err)
1266 goto out;
1267
1268 dentry = fhp->fh_dentry;
1269 dirp = dentry->d_inode;
1270
1271 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001272 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 goto out;
1274 /*
1275 * Check whether the response file handle has been verified yet.
1276 * If it has, the parent directory should already be locked.
1277 */
1278 if (!resfhp->fh_dentry) {
1279 /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001280 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001282 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 if (IS_ERR(dchild))
1284 goto out_nfserr;
1285 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1286 if (err)
1287 goto out;
1288 } else {
1289 /* called from nfsd_proc_create */
1290 dchild = dget(resfhp->fh_dentry);
1291 if (!fhp->fh_locked) {
1292 /* not actually possible */
1293 printk(KERN_ERR
1294 "nfsd_create: parent %s/%s not locked!\n",
1295 dentry->d_parent->d_name.name,
1296 dentry->d_name.name);
Al Virod75f2b92006-01-18 17:43:44 -08001297 err = nfserr_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 goto out;
1299 }
1300 }
1301 /*
1302 * Make sure the child dentry is still negative ...
1303 */
1304 err = nfserr_exist;
1305 if (dchild->d_inode) {
1306 dprintk("nfsd_create: dentry %s/%s not negative!\n",
1307 dentry->d_name.name, dchild->d_name.name);
1308 goto out;
1309 }
1310
1311 if (!(iap->ia_valid & ATTR_MODE))
1312 iap->ia_mode = 0;
1313 iap->ia_mode = (iap->ia_mode & S_IALLUGO) | type;
1314
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001315 err = nfserr_inval;
1316 if (!S_ISREG(type) && !S_ISDIR(type) && !special_file(type)) {
1317 printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n",
1318 type);
1319 goto out;
1320 }
1321
Al Virobad0dcf2011-11-23 12:03:18 -05001322 host_err = fh_want_write(fhp);
Miklos Szeredi07cad1d2008-07-01 15:38:35 +02001323 if (host_err)
1324 goto out_nfserr;
1325
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 /*
1327 * Get the dir op function pointer.
1328 */
J. Bruce Fields088406b2006-11-08 17:44:59 -08001329 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 switch (type) {
1331 case S_IFREG:
Al Viro6264d692006-10-19 23:28:58 -07001332 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
wengang wang4ac35c22009-02-10 11:27:51 +08001333 if (!host_err)
1334 nfsd_check_ignore_resizing(iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 break;
1336 case S_IFDIR:
Al Viro6264d692006-10-19 23:28:58 -07001337 host_err = vfs_mkdir(dirp, dchild, iap->ia_mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 break;
1339 case S_IFCHR:
1340 case S_IFBLK:
1341 case S_IFIFO:
1342 case S_IFSOCK:
Al Viro6264d692006-10-19 23:28:58 -07001343 host_err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 break;
Dave Hansen463c3192008-02-15 14:37:57 -08001345 }
1346 if (host_err < 0) {
Al Virobad0dcf2011-11-23 12:03:18 -05001347 fh_drop_write(fhp);
Dave Hansen463c3192008-02-15 14:37:57 -08001348 goto out_nfserr;
1349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
Ben Myersf5019122010-02-17 14:05:11 -06001351 err = nfsd_create_setattr(rqstp, resfhp, iap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Ben Myersf5019122010-02-17 14:05:11 -06001353 /*
1354 * nfsd_setattr already committed the child. Transactional filesystems
1355 * had a chance to commit changes for both parent and child
1356 * simultaneously making the following commit_metadata a noop.
1357 */
1358 err2 = nfserrno(commit_metadata(fhp));
J. Bruce Fields5c002b32007-11-30 16:55:23 -05001359 if (err2)
1360 err = err2;
Al Virobad0dcf2011-11-23 12:03:18 -05001361 fh_drop_write(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 /*
1363 * Update the file handle to get the new inode info.
1364 */
1365 if (!err)
1366 err = fh_update(resfhp);
1367out:
1368 if (dchild && !IS_ERR(dchild))
1369 dput(dchild);
1370 return err;
1371
1372out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001373 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 goto out;
1375}
1376
1377#ifdef CONFIG_NFSD_V3
Mi Jinlongac6721a2011-04-20 17:06:25 +08001378
1379static inline int nfsd_create_is_exclusive(int createmode)
1380{
1381 return createmode == NFS3_CREATE_EXCLUSIVE
1382 || createmode == NFS4_CREATE_EXCLUSIVE4_1;
1383}
1384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385/*
Mi Jinlongac6721a2011-04-20 17:06:25 +08001386 * NFSv3 and NFSv4 version of nfsd_create
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 */
Al Viro6264d692006-10-19 23:28:58 -07001388__be32
Mi Jinlongac6721a2011-04-20 17:06:25 +08001389do_nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 char *fname, int flen, struct iattr *iap,
1391 struct svc_fh *resfhp, int createmode, u32 *verifier,
J. Bruce Fields856121b2011-10-13 11:37:11 -04001392 bool *truncp, bool *created)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393{
1394 struct dentry *dentry, *dchild = NULL;
1395 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001396 __be32 err;
1397 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 __u32 v_mtime=0, v_atime=0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399
1400 err = nfserr_perm;
1401 if (!flen)
1402 goto out;
1403 err = nfserr_exist;
1404 if (isdotent(fname, flen))
1405 goto out;
1406 if (!(iap->ia_valid & ATTR_MODE))
1407 iap->ia_mode = 0;
Sachin Prabhu1574dff2011-04-20 13:09:35 +01001408 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 if (err)
1410 goto out;
1411
1412 dentry = fhp->fh_dentry;
1413 dirp = dentry->d_inode;
1414
1415 /* Get all the sanity checks out of the way before
1416 * we lock the parent. */
1417 err = nfserr_notdir;
Al Viroacfa4382008-12-04 10:06:33 -05001418 if (!dirp->i_op->lookup)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 goto out;
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001420 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
1422 /*
1423 * Compose the response file handle.
1424 */
1425 dchild = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001426 host_err = PTR_ERR(dchild);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 if (IS_ERR(dchild))
1428 goto out_nfserr;
1429
Sachin Prabhu1574dff2011-04-20 13:09:35 +01001430 /* If file doesn't exist, check for permissions to create one */
1431 if (!dchild->d_inode) {
1432 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
1433 if (err)
1434 goto out;
1435 }
1436
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 err = fh_compose(resfhp, fhp->fh_export, dchild, fhp);
1438 if (err)
1439 goto out;
1440
Mi Jinlongac6721a2011-04-20 17:06:25 +08001441 if (nfsd_create_is_exclusive(createmode)) {
Peter Staubachc3978522007-01-26 00:57:00 -08001442 /* solaris7 gets confused (bugid 4218508) if these have
Jeff Layton749997e2007-07-31 00:37:51 -07001443 * the high bit set, so just clear the high bits. If this is
1444 * ever changed to use different attrs for storing the
1445 * verifier, then do_open_lookup() will also need to be fixed
1446 * accordingly.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 */
1448 v_mtime = verifier[0]&0x7fffffff;
1449 v_atime = verifier[1]&0x7fffffff;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 }
1451
Al Virobad0dcf2011-11-23 12:03:18 -05001452 host_err = fh_want_write(fhp);
Dave Hansen463c3192008-02-15 14:37:57 -08001453 if (host_err)
1454 goto out_nfserr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 if (dchild->d_inode) {
1456 err = 0;
1457
1458 switch (createmode) {
1459 case NFS3_CREATE_UNCHECKED:
1460 if (! S_ISREG(dchild->d_inode->i_mode))
J. Bruce Fields9dc4e6c2012-04-09 18:06:49 -04001461 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 else if (truncp) {
1463 /* in nfsv4, we need to treat this case a little
1464 * differently. we don't want to truncate the
1465 * file now; this would be wrong if the OPEN
1466 * fails for some other reason. furthermore,
1467 * if the size is nonzero, we should ignore it
1468 * according to spec!
1469 */
1470 *truncp = (iap->ia_valid & ATTR_SIZE) && !iap->ia_size;
1471 }
1472 else {
1473 iap->ia_valid &= ATTR_SIZE;
1474 goto set_attr;
1475 }
1476 break;
1477 case NFS3_CREATE_EXCLUSIVE:
1478 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1479 && dchild->d_inode->i_atime.tv_sec == v_atime
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 && dchild->d_inode->i_size == 0 )
1481 break;
Mi Jinlongac6721a2011-04-20 17:06:25 +08001482 case NFS4_CREATE_EXCLUSIVE4_1:
1483 if ( dchild->d_inode->i_mtime.tv_sec == v_mtime
1484 && dchild->d_inode->i_atime.tv_sec == v_atime
1485 && dchild->d_inode->i_size == 0 )
1486 goto set_attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 /* fallthru */
1488 case NFS3_CREATE_GUARDED:
1489 err = nfserr_exist;
1490 }
Al Virobad0dcf2011-11-23 12:03:18 -05001491 fh_drop_write(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 goto out;
1493 }
1494
Al Viro6264d692006-10-19 23:28:58 -07001495 host_err = vfs_create(dirp, dchild, iap->ia_mode, NULL);
Dave Hansen463c3192008-02-15 14:37:57 -08001496 if (host_err < 0) {
Al Virobad0dcf2011-11-23 12:03:18 -05001497 fh_drop_write(fhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 goto out_nfserr;
Dave Hansen463c3192008-02-15 14:37:57 -08001499 }
J. Bruce Fields81ac95c2006-11-08 17:44:40 -08001500 if (created)
1501 *created = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
wengang wang4ac35c22009-02-10 11:27:51 +08001503 nfsd_check_ignore_resizing(iap);
1504
Mi Jinlongac6721a2011-04-20 17:06:25 +08001505 if (nfsd_create_is_exclusive(createmode)) {
Peter Staubachc3978522007-01-26 00:57:00 -08001506 /* Cram the verifier into atime/mtime */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 iap->ia_valid = ATTR_MTIME|ATTR_ATIME
Peter Staubachc3978522007-01-26 00:57:00 -08001508 | ATTR_MTIME_SET|ATTR_ATIME_SET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 /* XXX someone who knows this better please fix it for nsec */
1510 iap->ia_mtime.tv_sec = v_mtime;
1511 iap->ia_atime.tv_sec = v_atime;
1512 iap->ia_mtime.tv_nsec = 0;
1513 iap->ia_atime.tv_nsec = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 }
1515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 set_attr:
Ben Myersf5019122010-02-17 14:05:11 -06001517 err = nfsd_create_setattr(rqstp, resfhp, iap);
1518
1519 /*
1520 * nfsd_setattr already committed the child (and possibly also the parent).
1521 */
1522 if (!err)
1523 err = nfserrno(commit_metadata(fhp));
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001524
Al Virobad0dcf2011-11-23 12:03:18 -05001525 fh_drop_write(fhp);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001526 /*
1527 * Update the filehandle to get the new inode info.
1528 */
1529 if (!err)
1530 err = fh_update(resfhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
1532 out:
1533 fh_unlock(fhp);
1534 if (dchild && !IS_ERR(dchild))
1535 dput(dchild);
1536 return err;
1537
1538 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001539 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 goto out;
1541}
1542#endif /* CONFIG_NFSD_V3 */
1543
1544/*
1545 * Read a symlink. On entry, *lenp must contain the maximum path length that
1546 * fits into the buffer. On return, it contains the true length.
1547 * N.B. After this call fhp needs an fh_put
1548 */
Al Viro6264d692006-10-19 23:28:58 -07001549__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550nfsd_readlink(struct svc_rqst *rqstp, struct svc_fh *fhp, char *buf, int *lenp)
1551{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 struct inode *inode;
1553 mm_segment_t oldfs;
Al Viro6264d692006-10-19 23:28:58 -07001554 __be32 err;
1555 int host_err;
Al Viro68ac1232012-03-15 08:21:57 -04001556 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001558 err = fh_verify(rqstp, fhp, S_IFLNK, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 if (err)
1560 goto out;
1561
Al Viro68ac1232012-03-15 08:21:57 -04001562 path.mnt = fhp->fh_export->ex_path.mnt;
1563 path.dentry = fhp->fh_dentry;
1564 inode = path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
1566 err = nfserr_inval;
Al Viroacfa4382008-12-04 10:06:33 -05001567 if (!inode->i_op->readlink)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 goto out;
1569
Al Viro68ac1232012-03-15 08:21:57 -04001570 touch_atime(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 /* N.B. Why does this call need a get_fs()??
1572 * Remove the set_fs and watch the fireworks:-) --okir
1573 */
1574
1575 oldfs = get_fs(); set_fs(KERNEL_DS);
Al Viro68ac1232012-03-15 08:21:57 -04001576 host_err = inode->i_op->readlink(path.dentry, buf, *lenp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 set_fs(oldfs);
1578
Al Viro6264d692006-10-19 23:28:58 -07001579 if (host_err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 goto out_nfserr;
Al Viro6264d692006-10-19 23:28:58 -07001581 *lenp = host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 err = 0;
1583out:
1584 return err;
1585
1586out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001587 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 goto out;
1589}
1590
1591/*
1592 * Create a symlink and look up its inode
1593 * N.B. After this call _both_ fhp and resfhp need an fh_put
1594 */
Al Viro6264d692006-10-19 23:28:58 -07001595__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596nfsd_symlink(struct svc_rqst *rqstp, struct svc_fh *fhp,
1597 char *fname, int flen,
1598 char *path, int plen,
1599 struct svc_fh *resfhp,
1600 struct iattr *iap)
1601{
1602 struct dentry *dentry, *dnew;
Al Viro6264d692006-10-19 23:28:58 -07001603 __be32 err, cerr;
1604 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
1606 err = nfserr_noent;
1607 if (!flen || !plen)
1608 goto out;
1609 err = nfserr_exist;
1610 if (isdotent(fname, flen))
1611 goto out;
1612
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001613 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 if (err)
1615 goto out;
1616 fh_lock(fhp);
1617 dentry = fhp->fh_dentry;
1618 dnew = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001619 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 if (IS_ERR(dnew))
1621 goto out_nfserr;
1622
Al Virobad0dcf2011-11-23 12:03:18 -05001623 host_err = fh_want_write(fhp);
Dave Hansen75c3f292008-02-15 14:37:45 -08001624 if (host_err)
1625 goto out_nfserr;
1626
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 if (unlikely(path[plen] != 0)) {
1628 char *path_alloced = kmalloc(plen+1, GFP_KERNEL);
1629 if (path_alloced == NULL)
Al Viro6264d692006-10-19 23:28:58 -07001630 host_err = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 else {
1632 strncpy(path_alloced, path, plen);
1633 path_alloced[plen] = 0;
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001634 host_err = vfs_symlink(dentry->d_inode, dnew, path_alloced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 kfree(path_alloced);
1636 }
1637 } else
Miklos Szeredidb2e7472008-06-24 16:50:16 +02001638 host_err = vfs_symlink(dentry->d_inode, dnew, path);
Al Viro6264d692006-10-19 23:28:58 -07001639 err = nfserrno(host_err);
Ben Myersf5019122010-02-17 14:05:11 -06001640 if (!err)
1641 err = nfserrno(commit_metadata(fhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 fh_unlock(fhp);
1643
Al Virobad0dcf2011-11-23 12:03:18 -05001644 fh_drop_write(fhp);
Dave Hansen75c3f292008-02-15 14:37:45 -08001645
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 cerr = fh_compose(resfhp, fhp->fh_export, dnew, fhp);
1647 dput(dnew);
1648 if (err==0) err = cerr;
1649out:
1650 return err;
1651
1652out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001653 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 goto out;
1655}
1656
1657/*
1658 * Create a hardlink
1659 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1660 */
Al Viro6264d692006-10-19 23:28:58 -07001661__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662nfsd_link(struct svc_rqst *rqstp, struct svc_fh *ffhp,
1663 char *name, int len, struct svc_fh *tfhp)
1664{
1665 struct dentry *ddir, *dnew, *dold;
J. Bruce Fields55b13352010-07-19 16:38:24 -04001666 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001667 __be32 err;
1668 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001670 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 if (err)
1672 goto out;
J. Bruce Fields7d818a72011-08-15 16:59:55 -04001673 err = fh_verify(rqstp, tfhp, 0, NFSD_MAY_NOP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 if (err)
1675 goto out;
J. Bruce Fields7d818a72011-08-15 16:59:55 -04001676 err = nfserr_isdir;
1677 if (S_ISDIR(tfhp->fh_dentry->d_inode->i_mode))
1678 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 err = nfserr_perm;
1680 if (!len)
1681 goto out;
1682 err = nfserr_exist;
1683 if (isdotent(name, len))
1684 goto out;
1685
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001686 fh_lock_nested(ffhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 ddir = ffhp->fh_dentry;
1688 dirp = ddir->d_inode;
1689
1690 dnew = lookup_one_len(name, ddir, len);
Al Viro6264d692006-10-19 23:28:58 -07001691 host_err = PTR_ERR(dnew);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 if (IS_ERR(dnew))
1693 goto out_nfserr;
1694
1695 dold = tfhp->fh_dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696
Al Virobad0dcf2011-11-23 12:03:18 -05001697 host_err = fh_want_write(tfhp);
Dave Hansen75c3f292008-02-15 14:37:45 -08001698 if (host_err) {
1699 err = nfserrno(host_err);
1700 goto out_dput;
1701 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001702 err = nfserr_noent;
1703 if (!dold->d_inode)
1704 goto out_drop_write;
1705 host_err = nfsd_break_lease(dold->d_inode);
Casey Bodley7d751f62011-06-03 12:21:23 -04001706 if (host_err) {
1707 err = nfserrno(host_err);
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001708 goto out_drop_write;
Casey Bodley7d751f62011-06-03 12:21:23 -04001709 }
Al Viro6264d692006-10-19 23:28:58 -07001710 host_err = vfs_link(dold, dirp, dnew);
1711 if (!host_err) {
Ben Myersf5019122010-02-17 14:05:11 -06001712 err = nfserrno(commit_metadata(ffhp));
1713 if (!err)
1714 err = nfserrno(commit_metadata(tfhp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 } else {
Al Viro6264d692006-10-19 23:28:58 -07001716 if (host_err == -EXDEV && rqstp->rq_vers == 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 err = nfserr_acces;
1718 else
Al Viro6264d692006-10-19 23:28:58 -07001719 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001721out_drop_write:
Al Virobad0dcf2011-11-23 12:03:18 -05001722 fh_drop_write(tfhp);
Dave Hansen75c3f292008-02-15 14:37:45 -08001723out_dput:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 dput(dnew);
David M. Richter270d56e2006-06-30 01:56:15 -07001725out_unlock:
1726 fh_unlock(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727out:
1728 return err;
1729
1730out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001731 err = nfserrno(host_err);
David M. Richter270d56e2006-06-30 01:56:15 -07001732 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733}
1734
1735/*
1736 * Rename a file
1737 * N.B. After this call _both_ ffhp and tfhp need an fh_put
1738 */
Al Viro6264d692006-10-19 23:28:58 -07001739__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740nfsd_rename(struct svc_rqst *rqstp, struct svc_fh *ffhp, char *fname, int flen,
1741 struct svc_fh *tfhp, char *tname, int tlen)
1742{
1743 struct dentry *fdentry, *tdentry, *odentry, *ndentry, *trap;
1744 struct inode *fdir, *tdir;
Al Viro6264d692006-10-19 23:28:58 -07001745 __be32 err;
1746 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001748 err = fh_verify(rqstp, ffhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 if (err)
1750 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001751 err = fh_verify(rqstp, tfhp, S_IFDIR, NFSD_MAY_CREATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 if (err)
1753 goto out;
1754
1755 fdentry = ffhp->fh_dentry;
1756 fdir = fdentry->d_inode;
1757
1758 tdentry = tfhp->fh_dentry;
1759 tdir = tdentry->d_inode;
1760
1761 err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
NeilBrowna56f3932006-06-30 01:56:10 -07001762 if (ffhp->fh_export != tfhp->fh_export)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 goto out;
1764
1765 err = nfserr_perm;
1766 if (!flen || isdotent(fname, flen) || !tlen || isdotent(tname, tlen))
1767 goto out;
1768
1769 /* cannot use fh_lock as we need deadlock protective ordering
1770 * so do it by hand */
1771 trap = lock_rename(tdentry, fdentry);
1772 ffhp->fh_locked = tfhp->fh_locked = 1;
1773 fill_pre_wcc(ffhp);
1774 fill_pre_wcc(tfhp);
1775
1776 odentry = lookup_one_len(fname, fdentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001777 host_err = PTR_ERR(odentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 if (IS_ERR(odentry))
1779 goto out_nfserr;
1780
Al Viro6264d692006-10-19 23:28:58 -07001781 host_err = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 if (!odentry->d_inode)
1783 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001784 host_err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 if (odentry == trap)
1786 goto out_dput_old;
1787
1788 ndentry = lookup_one_len(tname, tdentry, tlen);
Al Viro6264d692006-10-19 23:28:58 -07001789 host_err = PTR_ERR(ndentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 if (IS_ERR(ndentry))
1791 goto out_dput_old;
Al Viro6264d692006-10-19 23:28:58 -07001792 host_err = -ENOTEMPTY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 if (ndentry == trap)
1794 goto out_dput_new;
1795
Dave Hansen9079b1e2008-02-15 14:37:49 -08001796 host_err = -EXDEV;
1797 if (ffhp->fh_export->ex_path.mnt != tfhp->fh_export->ex_path.mnt)
1798 goto out_dput_new;
Al Virobad0dcf2011-11-23 12:03:18 -05001799 host_err = fh_want_write(ffhp);
Dave Hansen9079b1e2008-02-15 14:37:49 -08001800 if (host_err)
1801 goto out_dput_new;
1802
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001803 host_err = nfsd_break_lease(odentry->d_inode);
1804 if (host_err)
1805 goto out_drop_write;
J. Bruce Fields83f6b0c2011-02-06 16:46:30 -05001806 if (ndentry->d_inode) {
1807 host_err = nfsd_break_lease(ndentry->d_inode);
1808 if (host_err)
1809 goto out_drop_write;
1810 }
Al Viro6264d692006-10-19 23:28:58 -07001811 host_err = vfs_rename(fdir, odentry, tdir, ndentry);
Ben Myersf5019122010-02-17 14:05:11 -06001812 if (!host_err) {
1813 host_err = commit_metadata(tfhp);
Al Viro6264d692006-10-19 23:28:58 -07001814 if (!host_err)
Ben Myersf5019122010-02-17 14:05:11 -06001815 host_err = commit_metadata(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 }
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001817out_drop_write:
Al Virobad0dcf2011-11-23 12:03:18 -05001818 fh_drop_write(ffhp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 out_dput_new:
1820 dput(ndentry);
1821 out_dput_old:
1822 dput(odentry);
1823 out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001824 err = nfserrno(host_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825
1826 /* we cannot reply on fh_unlock on the two filehandles,
1827 * as that would do the wrong thing if the two directories
1828 * were the same, so again we do it by hand
1829 */
1830 fill_post_wcc(ffhp);
1831 fill_post_wcc(tfhp);
1832 unlock_rename(tdentry, fdentry);
1833 ffhp->fh_locked = tfhp->fh_locked = 0;
1834
1835out:
1836 return err;
1837}
1838
1839/*
1840 * Unlink a file or directory
1841 * N.B. After this call fhp needs an fh_put
1842 */
Al Viro6264d692006-10-19 23:28:58 -07001843__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
1845 char *fname, int flen)
1846{
1847 struct dentry *dentry, *rdentry;
1848 struct inode *dirp;
Al Viro6264d692006-10-19 23:28:58 -07001849 __be32 err;
1850 int host_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851
1852 err = nfserr_acces;
1853 if (!flen || isdotent(fname, flen))
1854 goto out;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02001855 err = fh_verify(rqstp, fhp, S_IFDIR, NFSD_MAY_REMOVE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 if (err)
1857 goto out;
1858
Peter Zijlstra12fd3522006-10-02 02:18:03 -07001859 fh_lock_nested(fhp, I_MUTEX_PARENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 dentry = fhp->fh_dentry;
1861 dirp = dentry->d_inode;
1862
1863 rdentry = lookup_one_len(fname, dentry, flen);
Al Viro6264d692006-10-19 23:28:58 -07001864 host_err = PTR_ERR(rdentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865 if (IS_ERR(rdentry))
1866 goto out_nfserr;
1867
1868 if (!rdentry->d_inode) {
1869 dput(rdentry);
1870 err = nfserr_noent;
1871 goto out;
1872 }
1873
1874 if (!type)
1875 type = rdentry->d_inode->i_mode & S_IFMT;
1876
Al Virobad0dcf2011-11-23 12:03:18 -05001877 host_err = fh_want_write(fhp);
Dave Hansen06227532008-02-15 14:37:34 -08001878 if (host_err)
J. Bruce Fields541ce982011-01-14 20:00:02 -05001879 goto out_put;
Dave Hansen06227532008-02-15 14:37:34 -08001880
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001881 host_err = nfsd_break_lease(rdentry->d_inode);
1882 if (host_err)
J. Bruce Fields541ce982011-01-14 20:00:02 -05001883 goto out_drop_write;
J. Bruce Fields9ce137e2011-01-11 14:07:12 -05001884 if (type != S_IFDIR)
Al Viro6264d692006-10-19 23:28:58 -07001885 host_err = vfs_unlink(dirp, rdentry);
J. Bruce Fields9ce137e2011-01-11 14:07:12 -05001886 else
Al Viro6264d692006-10-19 23:28:58 -07001887 host_err = vfs_rmdir(dirp, rdentry);
J. Bruce Fields541ce982011-01-14 20:00:02 -05001888 if (!host_err)
1889 host_err = commit_metadata(fhp);
1890out_drop_write:
Al Virobad0dcf2011-11-23 12:03:18 -05001891 fh_drop_write(fhp);
J. Bruce Fields4795bb32011-01-11 13:55:46 -05001892out_put:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 dput(rdentry);
1894
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895out_nfserr:
Al Viro6264d692006-10-19 23:28:58 -07001896 err = nfserrno(host_err);
YAMAMOTO Takashif193fba2006-01-18 17:43:13 -08001897out:
1898 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899}
1900
1901/*
David Woodhouse14f7dd62008-07-31 20:29:12 +01001902 * We do this buffering because we must not call back into the file
1903 * system's ->lookup() method from the filldir callback. That may well
1904 * deadlock a number of file systems.
1905 *
1906 * This is based heavily on the implementation of same in XFS.
1907 */
1908struct buffered_dirent {
1909 u64 ino;
1910 loff_t offset;
1911 int namlen;
1912 unsigned int d_type;
1913 char name[];
1914};
1915
1916struct readdir_data {
1917 char *dirent;
1918 size_t used;
Al Viro53c9c5c2008-08-24 07:29:52 -04001919 int full;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001920};
1921
1922static int nfsd_buffered_filldir(void *__buf, const char *name, int namlen,
1923 loff_t offset, u64 ino, unsigned int d_type)
David Woodhouse2628b762008-07-31 17:16:51 +01001924{
David Woodhouse14f7dd62008-07-31 20:29:12 +01001925 struct readdir_data *buf = __buf;
1926 struct buffered_dirent *de = (void *)(buf->dirent + buf->used);
1927 unsigned int reclen;
1928
1929 reclen = ALIGN(sizeof(struct buffered_dirent) + namlen, sizeof(u64));
Al Viro53c9c5c2008-08-24 07:29:52 -04001930 if (buf->used + reclen > PAGE_SIZE) {
1931 buf->full = 1;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001932 return -EINVAL;
Al Viro53c9c5c2008-08-24 07:29:52 -04001933 }
David Woodhouse14f7dd62008-07-31 20:29:12 +01001934
1935 de->namlen = namlen;
1936 de->offset = offset;
1937 de->ino = ino;
1938 de->d_type = d_type;
1939 memcpy(de->name, name, namlen);
1940 buf->used += reclen;
1941
1942 return 0;
1943}
1944
David Woodhouse2f9092e2009-04-20 23:18:37 +01001945static __be32 nfsd_buffered_readdir(struct file *file, filldir_t func,
1946 struct readdir_cd *cdp, loff_t *offsetp)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001947{
1948 struct readdir_data buf;
1949 struct buffered_dirent *de;
David Woodhouse2628b762008-07-31 17:16:51 +01001950 int host_err;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001951 int size;
1952 loff_t offset;
David Woodhouse2628b762008-07-31 17:16:51 +01001953
David Woodhouse14f7dd62008-07-31 20:29:12 +01001954 buf.dirent = (void *)__get_free_page(GFP_KERNEL);
1955 if (!buf.dirent)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001956 return nfserrno(-ENOMEM);
David Woodhouse2628b762008-07-31 17:16:51 +01001957
David Woodhouse14f7dd62008-07-31 20:29:12 +01001958 offset = *offsetp;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001959
1960 while (1) {
David Woodhouse2f9092e2009-04-20 23:18:37 +01001961 struct inode *dir_inode = file->f_path.dentry->d_inode;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001962 unsigned int reclen;
1963
Doug Nazarb726e922008-11-05 06:16:28 -05001964 cdp->err = nfserr_eof; /* will be cleared on successful read */
David Woodhouse14f7dd62008-07-31 20:29:12 +01001965 buf.used = 0;
Al Viro53c9c5c2008-08-24 07:29:52 -04001966 buf.full = 0;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001967
1968 host_err = vfs_readdir(file, nfsd_buffered_filldir, &buf);
Al Viro53c9c5c2008-08-24 07:29:52 -04001969 if (buf.full)
1970 host_err = 0;
1971
1972 if (host_err < 0)
David Woodhouse14f7dd62008-07-31 20:29:12 +01001973 break;
1974
1975 size = buf.used;
1976
1977 if (!size)
1978 break;
1979
David Woodhouse2f9092e2009-04-20 23:18:37 +01001980 /*
1981 * Various filldir functions may end up calling back into
1982 * lookup_one_len() and the file system's ->lookup() method.
1983 * These expect i_mutex to be held, as it would within readdir.
1984 */
1985 host_err = mutex_lock_killable(&dir_inode->i_mutex);
1986 if (host_err)
1987 break;
1988
David Woodhouse14f7dd62008-07-31 20:29:12 +01001989 de = (struct buffered_dirent *)buf.dirent;
1990 while (size > 0) {
1991 offset = de->offset;
1992
1993 if (func(cdp, de->name, de->namlen, de->offset,
1994 de->ino, de->d_type))
David Woodhouse2f9092e2009-04-20 23:18:37 +01001995 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001996
1997 if (cdp->err != nfs_ok)
David Woodhouse2f9092e2009-04-20 23:18:37 +01001998 break;
David Woodhouse14f7dd62008-07-31 20:29:12 +01001999
2000 reclen = ALIGN(sizeof(*de) + de->namlen,
2001 sizeof(u64));
2002 size -= reclen;
2003 de = (struct buffered_dirent *)((char *)de + reclen);
2004 }
David Woodhouse2f9092e2009-04-20 23:18:37 +01002005 mutex_unlock(&dir_inode->i_mutex);
2006 if (size > 0) /* We bailed out early */
2007 break;
2008
David Woodhousec002a6c2008-08-17 17:21:18 +01002009 offset = vfs_llseek(file, 0, SEEK_CUR);
David Woodhouse14f7dd62008-07-31 20:29:12 +01002010 }
2011
David Woodhouse14f7dd62008-07-31 20:29:12 +01002012 free_page((unsigned long)(buf.dirent));
David Woodhouse2628b762008-07-31 17:16:51 +01002013
2014 if (host_err)
2015 return nfserrno(host_err);
David Woodhouse14f7dd62008-07-31 20:29:12 +01002016
2017 *offsetp = offset;
2018 return cdp->err;
David Woodhouse2628b762008-07-31 17:16:51 +01002019}
2020
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021/*
2022 * Read entries from a directory.
2023 * The NFSv3/4 verifier we ignore for now.
2024 */
Al Viro6264d692006-10-19 23:28:58 -07002025__be32
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026nfsd_readdir(struct svc_rqst *rqstp, struct svc_fh *fhp, loff_t *offsetp,
NeilBrowna0ad13e2007-01-26 00:57:10 -08002027 struct readdir_cd *cdp, filldir_t func)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028{
Al Viro6264d692006-10-19 23:28:58 -07002029 __be32 err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 struct file *file;
2031 loff_t offset = *offsetp;
Bernd Schubert06effdb2012-03-18 22:44:50 -04002032 int may_flags = NFSD_MAY_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Bernd Schubert06effdb2012-03-18 22:44:50 -04002034 /* NFSv2 only supports 32 bit cookies */
2035 if (rqstp->rq_vers > 2)
2036 may_flags |= NFSD_MAY_64BIT_COOKIE;
2037
2038 err = nfsd_open(rqstp, fhp, S_IFDIR, may_flags, &file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 if (err)
2040 goto out;
2041
Jeff Laytonb108fe62012-04-25 15:30:00 -04002042 offset = vfs_llseek(file, offset, SEEK_SET);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 if (offset < 0) {
2044 err = nfserrno((int)offset);
2045 goto out_close;
2046 }
2047
David Woodhouse14f7dd62008-07-31 20:29:12 +01002048 err = nfsd_buffered_readdir(file, func, cdp, offsetp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
2050 if (err == nfserr_eof || err == nfserr_toosmall)
2051 err = nfs_ok; /* can still be found in ->err */
2052out_close:
2053 nfsd_close(file);
2054out:
2055 return err;
2056}
2057
2058/*
2059 * Get file system stats
2060 * N.B. After this call fhp needs an fh_put
2061 */
Al Viro6264d692006-10-19 23:28:58 -07002062__be32
J. Bruce Fields04716e62008-08-07 13:00:20 -04002063nfsd_statfs(struct svc_rqst *rqstp, struct svc_fh *fhp, struct kstatfs *stat, int access)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064{
Christoph Hellwigebabe9a2010-07-07 18:53:11 +02002065 __be32 err;
2066
2067 err = fh_verify(rqstp, fhp, 0, NFSD_MAY_NOP | access);
Takashi Iwaif6360ef2010-08-13 15:53:49 +02002068 if (!err) {
2069 struct path path = {
2070 .mnt = fhp->fh_export->ex_path.mnt,
2071 .dentry = fhp->fh_dentry,
2072 };
2073 if (vfs_statfs(&path, stat))
2074 err = nfserr_io;
2075 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 return err;
2077}
2078
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002079static int exp_rdonly(struct svc_rqst *rqstp, struct svc_export *exp)
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002080{
J. Bruce Fieldsc7d51402007-07-19 01:49:20 -07002081 return nfsexp_flags(rqstp, exp) & NFSEXP_READONLY;
J. Bruce Fieldse22841c2007-07-19 01:49:20 -07002082}
2083
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084/*
2085 * Check for a user's access permissions to this inode.
2086 */
Al Viro6264d692006-10-19 23:28:58 -07002087__be32
J. Bruce Fields0ec757d2007-07-17 04:04:48 -07002088nfsd_permission(struct svc_rqst *rqstp, struct svc_export *exp,
2089 struct dentry *dentry, int acc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
2091 struct inode *inode = dentry->d_inode;
2092 int err;
2093
J. Bruce Fieldsaea93392011-04-10 10:35:12 -04002094 if ((acc & NFSD_MAY_MASK) == NFSD_MAY_NOP)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 return 0;
2096#if 0
2097 dprintk("nfsd: permission 0x%x%s%s%s%s%s%s%s mode 0%o%s%s%s\n",
2098 acc,
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002099 (acc & NFSD_MAY_READ)? " read" : "",
2100 (acc & NFSD_MAY_WRITE)? " write" : "",
2101 (acc & NFSD_MAY_EXEC)? " exec" : "",
2102 (acc & NFSD_MAY_SATTR)? " sattr" : "",
2103 (acc & NFSD_MAY_TRUNC)? " trunc" : "",
2104 (acc & NFSD_MAY_LOCK)? " lock" : "",
2105 (acc & NFSD_MAY_OWNER_OVERRIDE)? " owneroverride" : "",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 inode->i_mode,
2107 IS_IMMUTABLE(inode)? " immut" : "",
2108 IS_APPEND(inode)? " append" : "",
Dave Hansen2c463e92008-02-15 14:37:56 -08002109 __mnt_is_readonly(exp->ex_path.mnt)? " ro" : "");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 dprintk(" owner %d/%d user %d/%d\n",
David Howells5cc0a842008-11-14 10:38:58 +11002111 inode->i_uid, inode->i_gid, current_fsuid(), current_fsgid());
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112#endif
2113
2114 /* Normally we reject any write/sattr etc access on a read-only file
2115 * system. But if it is IRIX doing check on write-access for a
2116 * device special file, we ignore rofs.
2117 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002118 if (!(acc & NFSD_MAY_LOCAL_ACCESS))
2119 if (acc & (NFSD_MAY_WRITE | NFSD_MAY_SATTR | NFSD_MAY_TRUNC)) {
Dave Hansen2c463e92008-02-15 14:37:56 -08002120 if (exp_rdonly(rqstp, exp) ||
2121 __mnt_is_readonly(exp->ex_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 return nfserr_rofs;
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002123 if (/* (acc & NFSD_MAY_WRITE) && */ IS_IMMUTABLE(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124 return nfserr_perm;
2125 }
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002126 if ((acc & NFSD_MAY_TRUNC) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 return nfserr_perm;
2128
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002129 if (acc & NFSD_MAY_LOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 /* If we cannot rely on authentication in NLM requests,
2131 * just allow locks, otherwise require read permission, or
2132 * ownership
2133 */
2134 if (exp->ex_flags & NFSEXP_NOAUTHNLM)
2135 return 0;
2136 else
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002137 acc = NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 }
2139 /*
2140 * The file owner always gets access permission for accesses that
2141 * would normally be checked at open time. This is to make
2142 * file access work even when the client has done a fchmod(fd, 0).
2143 *
2144 * However, `cp foo bar' should fail nevertheless when bar is
2145 * readonly. A sensible way to do this might be to reject all
2146 * attempts to truncate a read-only file, because a creat() call
2147 * always implies file truncation.
2148 * ... but this isn't really fair. A process may reasonably call
2149 * ftruncate on an open file descriptor on a file with perm 000.
2150 * We must trust the client to do permission checking - using "ACCESS"
2151 * with NFSv3.
2152 */
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002153 if ((acc & NFSD_MAY_OWNER_OVERRIDE) &&
David Howells5cc0a842008-11-14 10:38:58 +11002154 inode->i_uid == current_fsuid())
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 return 0;
2156
Miklos Szeredi8837abc2008-06-16 13:20:29 +02002157 /* This assumes NFSD_MAY_{READ,WRITE,EXEC} == MAY_{READ,WRITE,EXEC} */
Al Virof419a2e2008-07-22 00:07:17 -04002158 err = inode_permission(inode, acc & (MAY_READ|MAY_WRITE|MAY_EXEC));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159
2160 /* Allow read access to binaries even when mode 111 */
2161 if (err == -EACCES && S_ISREG(inode->i_mode) &&
J. Bruce Fieldsa0432262011-08-25 10:48:39 -04002162 (acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE) ||
2163 acc == (NFSD_MAY_READ | NFSD_MAY_READ_IF_EXEC)))
Al Virof419a2e2008-07-22 00:07:17 -04002164 err = inode_permission(inode, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165
2166 return err? nfserrno(err) : 0;
2167}
2168
2169void
2170nfsd_racache_shutdown(void)
2171{
Jeff Layton54a66e52008-08-13 22:03:27 -04002172 struct raparms *raparm, *last_raparm;
2173 unsigned int i;
2174
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175 dprintk("nfsd: freeing readahead buffers.\n");
Jeff Layton54a66e52008-08-13 22:03:27 -04002176
2177 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
2178 raparm = raparm_hash[i].pb_head;
2179 while(raparm) {
2180 last_raparm = raparm;
2181 raparm = raparm->p_next;
2182 kfree(last_raparm);
2183 }
2184 raparm_hash[i].pb_head = NULL;
2185 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186}
2187/*
2188 * Initialize readahead param cache
2189 */
2190int
2191nfsd_racache_init(int cache_size)
2192{
2193 int i;
Greg Banksfce14562006-10-04 02:15:49 -07002194 int j = 0;
2195 int nperbucket;
Jeff Layton54a66e52008-08-13 22:03:27 -04002196 struct raparms **raparm = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197
Greg Banksfce14562006-10-04 02:15:49 -07002198
Jeff Layton54a66e52008-08-13 22:03:27 -04002199 if (raparm_hash[0].pb_head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002201 nperbucket = DIV_ROUND_UP(cache_size, RAPARM_HASH_SIZE);
2202 if (nperbucket < 2)
2203 nperbucket = 2;
2204 cache_size = nperbucket * RAPARM_HASH_SIZE;
Yan Burman4b3bb062006-12-08 02:39:41 -08002205
2206 dprintk("nfsd: allocating %d readahead buffers.\n", cache_size);
Jeff Layton54a66e52008-08-13 22:03:27 -04002207
2208 for (i = 0; i < RAPARM_HASH_SIZE; i++) {
Yan Burman4b3bb062006-12-08 02:39:41 -08002209 spin_lock_init(&raparm_hash[i].pb_lock);
Jeff Layton54a66e52008-08-13 22:03:27 -04002210
2211 raparm = &raparm_hash[i].pb_head;
2212 for (j = 0; j < nperbucket; j++) {
2213 *raparm = kzalloc(sizeof(struct raparms), GFP_KERNEL);
2214 if (!*raparm)
2215 goto out_nomem;
2216 raparm = &(*raparm)->p_next;
2217 }
2218 *raparm = NULL;
Yan Burman4b3bb062006-12-08 02:39:41 -08002219 }
2220
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 nfsdstats.ra_size = cache_size;
2222 return 0;
Jeff Layton54a66e52008-08-13 22:03:27 -04002223
2224out_nomem:
2225 dprintk("nfsd: kmalloc failed, freeing readahead buffers\n");
2226 nfsd_racache_shutdown();
2227 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228}
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002229
2230#if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
2231struct posix_acl *
2232nfsd_get_posix_acl(struct svc_fh *fhp, int type)
2233{
2234 struct inode *inode = fhp->fh_dentry->d_inode;
2235 char *name;
2236 void *value = NULL;
2237 ssize_t size;
2238 struct posix_acl *acl;
2239
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002240 if (!IS_POSIXACL(inode))
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002241 return ERR_PTR(-EOPNOTSUPP);
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002242
2243 switch (type) {
2244 case ACL_TYPE_ACCESS:
2245 name = POSIX_ACL_XATTR_ACCESS;
2246 break;
2247 case ACL_TYPE_DEFAULT:
2248 name = POSIX_ACL_XATTR_DEFAULT;
2249 break;
2250 default:
2251 return ERR_PTR(-EOPNOTSUPP);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002252 }
2253
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002254 size = nfsd_getxattr(fhp->fh_dentry, name, &value);
2255 if (size < 0)
2256 return ERR_PTR(size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002257
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002258 acl = posix_acl_from_xattr(value, size);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002259 kfree(value);
2260 return acl;
2261}
2262
2263int
2264nfsd_set_posix_acl(struct svc_fh *fhp, int type, struct posix_acl *acl)
2265{
2266 struct inode *inode = fhp->fh_dentry->d_inode;
2267 char *name;
2268 void *value = NULL;
2269 size_t size;
2270 int error;
2271
Al Viroacfa4382008-12-04 10:06:33 -05002272 if (!IS_POSIXACL(inode) ||
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002273 !inode->i_op->setxattr || !inode->i_op->removexattr)
2274 return -EOPNOTSUPP;
2275 switch(type) {
2276 case ACL_TYPE_ACCESS:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002277 name = POSIX_ACL_XATTR_ACCESS;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002278 break;
2279 case ACL_TYPE_DEFAULT:
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002280 name = POSIX_ACL_XATTR_DEFAULT;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002281 break;
2282 default:
2283 return -EOPNOTSUPP;
2284 }
2285
2286 if (acl && acl->a_count) {
Christoph Hellwig334a13e2005-06-28 20:44:58 -07002287 size = posix_acl_xattr_size(acl->a_count);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002288 value = kmalloc(size, GFP_KERNEL);
2289 if (!value)
2290 return -ENOMEM;
Florin Malita9ccfc292006-05-20 14:59:58 -07002291 error = posix_acl_to_xattr(acl, value, size);
2292 if (error < 0)
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002293 goto getout;
Florin Malita9ccfc292006-05-20 14:59:58 -07002294 size = error;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002295 } else
2296 size = 0;
2297
Al Virobad0dcf2011-11-23 12:03:18 -05002298 error = fh_want_write(fhp);
Dave Hansen18f335a2008-02-15 14:37:38 -08002299 if (error)
2300 goto getout;
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002301 if (size)
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002302 error = vfs_setxattr(fhp->fh_dentry, name, value, size, 0);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002303 else {
2304 if (!S_ISDIR(inode->i_mode) && type == ACL_TYPE_DEFAULT)
2305 error = 0;
2306 else {
Christoph Hellwig5be196e2006-01-09 20:51:55 -08002307 error = vfs_removexattr(fhp->fh_dentry, name);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002308 if (error == -ENODATA)
2309 error = 0;
2310 }
2311 }
Al Virobad0dcf2011-11-23 12:03:18 -05002312 fh_drop_write(fhp);
Andreas Gruenbachera257cdd2005-06-22 17:16:26 +00002313
2314getout:
2315 kfree(value);
2316 return error;
2317}
2318#endif /* defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL) */