Commit | Line | Data |
---|---|---|
a55370a3 | 1 | /* |
a55370a3 | 2 | * Copyright (c) 2004 The Regents of the University of Michigan. |
f3f80148 | 3 | * Copyright (c) 2012 Jeff Layton <jlayton@redhat.com> |
a55370a3 N |
4 | * All rights reserved. |
5 | * | |
6 | * Andy Adamson <andros@citi.umich.edu> | |
7 | * | |
8 | * Redistribution and use in source and binary forms, with or without | |
9 | * modification, are permitted provided that the following conditions | |
10 | * are met: | |
11 | * | |
12 | * 1. Redistributions of source code must retain the above copyright | |
13 | * notice, this list of conditions and the following disclaimer. | |
14 | * 2. Redistributions in binary form must reproduce the above copyright | |
15 | * notice, this list of conditions and the following disclaimer in the | |
16 | * documentation and/or other materials provided with the distribution. | |
17 | * 3. Neither the name of the University nor the names of its | |
18 | * contributors may be used to endorse or promote products derived | |
19 | * from this software without specific prior written permission. | |
20 | * | |
21 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED | |
22 | * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF | |
23 | * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE | |
24 | * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
25 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | |
26 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | |
27 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR | |
28 | * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF | |
29 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING | |
30 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS | |
31 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
32 | * | |
33 | */ | |
34 | ||
190e4fbf | 35 | #include <linux/file.h> |
5a0e3ad6 | 36 | #include <linux/slab.h> |
190e4fbf | 37 | #include <linux/namei.h> |
a55370a3 | 38 | #include <linux/crypto.h> |
e8edc6e0 | 39 | #include <linux/sched.h> |
f3f80148 | 40 | #include <linux/fs.h> |
813fd320 | 41 | #include <linux/module.h> |
f3f80148 JL |
42 | #include <net/net_namespace.h> |
43 | #include <linux/sunrpc/rpc_pipe_fs.h> | |
44 | #include <linux/sunrpc/clnt.h> | |
45 | #include <linux/nfsd/cld.h> | |
9a74af21 BH |
46 | |
47 | #include "nfsd.h" | |
48 | #include "state.h" | |
0a3adade | 49 | #include "vfs.h" |
f3f80148 | 50 | #include "netns.h" |
a55370a3 N |
51 | |
52 | #define NFSDDBG_FACILITY NFSDDBG_PROC | |
53 | ||
2a4317c5 JL |
54 | /* Declarations */ |
55 | struct nfsd4_client_tracking_ops { | |
56 | int (*init)(struct net *); | |
57 | void (*exit)(struct net *); | |
58 | void (*create)(struct nfs4_client *); | |
59 | void (*remove)(struct nfs4_client *); | |
60 | int (*check)(struct nfs4_client *); | |
61 | void (*grace_done)(struct net *, time_t); | |
62 | }; | |
63 | ||
190e4fbf | 64 | /* Globals */ |
e970a573 | 65 | static struct file *rec_file; |
48483bf2 | 66 | static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery"; |
2a4317c5 | 67 | static struct nfsd4_client_tracking_ops *client_tracking_ops; |
190e4fbf | 68 | |
d84f4f99 DH |
69 | static int |
70 | nfs4_save_creds(const struct cred **original_creds) | |
190e4fbf | 71 | { |
d84f4f99 DH |
72 | struct cred *new; |
73 | ||
74 | new = prepare_creds(); | |
75 | if (!new) | |
76 | return -ENOMEM; | |
77 | ||
78 | new->fsuid = 0; | |
79 | new->fsgid = 0; | |
80 | *original_creds = override_creds(new); | |
81 | put_cred(new); | |
82 | return 0; | |
190e4fbf N |
83 | } |
84 | ||
85 | static void | |
d84f4f99 | 86 | nfs4_reset_creds(const struct cred *original) |
190e4fbf | 87 | { |
d84f4f99 | 88 | revert_creds(original); |
190e4fbf N |
89 | } |
90 | ||
a55370a3 N |
91 | static void |
92 | md5_to_hex(char *out, char *md5) | |
93 | { | |
94 | int i; | |
95 | ||
96 | for (i=0; i<16; i++) { | |
97 | unsigned char c = md5[i]; | |
98 | ||
99 | *out++ = '0' + ((c&0xf0)>>4) + (c>=0xa0)*('a'-'9'-1); | |
100 | *out++ = '0' + (c&0x0f) + ((c&0x0f)>=0x0a)*('a'-'9'-1); | |
101 | } | |
102 | *out = '\0'; | |
103 | } | |
104 | ||
b37ad28b | 105 | __be32 |
a55370a3 N |
106 | nfs4_make_rec_clidname(char *dname, struct xdr_netobj *clname) |
107 | { | |
108 | struct xdr_netobj cksum; | |
35058687 | 109 | struct hash_desc desc; |
60c74f81 | 110 | struct scatterlist sg; |
3e772463 | 111 | __be32 status = nfserr_jukebox; |
a55370a3 N |
112 | |
113 | dprintk("NFSD: nfs4_make_rec_clidname for %.*s\n", | |
114 | clname->len, clname->data); | |
35058687 HX |
115 | desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP; |
116 | desc.tfm = crypto_alloc_hash("md5", 0, CRYPTO_ALG_ASYNC); | |
117 | if (IS_ERR(desc.tfm)) | |
118 | goto out_no_tfm; | |
119 | cksum.len = crypto_hash_digestsize(desc.tfm); | |
a55370a3 N |
120 | cksum.data = kmalloc(cksum.len, GFP_KERNEL); |
121 | if (cksum.data == NULL) | |
122 | goto out; | |
a55370a3 | 123 | |
60c74f81 | 124 | sg_init_one(&sg, clname->data, clname->len); |
a55370a3 | 125 | |
60c74f81 | 126 | if (crypto_hash_digest(&desc, &sg, sg.length, cksum.data)) |
35058687 | 127 | goto out; |
a55370a3 N |
128 | |
129 | md5_to_hex(dname, cksum.data); | |
130 | ||
a55370a3 N |
131 | status = nfs_ok; |
132 | out: | |
2bd9e7b6 | 133 | kfree(cksum.data); |
35058687 HX |
134 | crypto_free_hash(desc.tfm); |
135 | out_no_tfm: | |
a55370a3 N |
136 | return status; |
137 | } | |
190e4fbf | 138 | |
2a4317c5 JL |
139 | static void |
140 | nfsd4_create_clid_dir(struct nfs4_client *clp) | |
c7b9a459 | 141 | { |
d84f4f99 | 142 | const struct cred *original_cred; |
c7b9a459 | 143 | char *dname = clp->cl_recdir; |
e970a573 | 144 | struct dentry *dir, *dentry; |
c7b9a459 N |
145 | int status; |
146 | ||
147 | dprintk("NFSD: nfsd4_create_clid_dir for \"%s\"\n", dname); | |
148 | ||
a52d726b | 149 | if (test_and_set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) |
7a6ef8c7 | 150 | return; |
b8548894 | 151 | if (!rec_file) |
7a6ef8c7 | 152 | return; |
d84f4f99 DH |
153 | status = nfs4_save_creds(&original_cred); |
154 | if (status < 0) | |
7a6ef8c7 | 155 | return; |
c7b9a459 | 156 | |
e970a573 | 157 | dir = rec_file->f_path.dentry; |
c7b9a459 | 158 | /* lock the parent */ |
e970a573 | 159 | mutex_lock(&dir->d_inode->i_mutex); |
c7b9a459 | 160 | |
e970a573 | 161 | dentry = lookup_one_len(dname, dir, HEXDIR_LEN-1); |
c7b9a459 N |
162 | if (IS_ERR(dentry)) { |
163 | status = PTR_ERR(dentry); | |
164 | goto out_unlock; | |
165 | } | |
6577aac0 | 166 | if (dentry->d_inode) |
aec39680 BF |
167 | /* |
168 | * In the 4.1 case, where we're called from | |
169 | * reclaim_complete(), records from the previous reboot | |
170 | * may still be left, so this is OK. | |
171 | * | |
172 | * In the 4.0 case, we should never get here; but we may | |
173 | * as well be forgiving and just succeed silently. | |
174 | */ | |
c7b9a459 | 175 | goto out_put; |
a561be71 | 176 | status = mnt_want_write_file(rec_file); |
463c3197 DH |
177 | if (status) |
178 | goto out_put; | |
e970a573 | 179 | status = vfs_mkdir(dir->d_inode, dentry, S_IRWXU); |
2a79f17e | 180 | mnt_drop_write_file(rec_file); |
c7b9a459 N |
181 | out_put: |
182 | dput(dentry); | |
183 | out_unlock: | |
e970a573 | 184 | mutex_unlock(&dir->d_inode->i_mutex); |
6577aac0 | 185 | if (status == 0) |
8018ab05 | 186 | vfs_fsync(rec_file, 0); |
6577aac0 BH |
187 | else |
188 | printk(KERN_ERR "NFSD: failed to write recovery record" | |
189 | " (err %d); please check that %s exists" | |
190 | " and is writeable", status, | |
191 | user_recovery_dirname); | |
d84f4f99 | 192 | nfs4_reset_creds(original_cred); |
c7b9a459 N |
193 | } |
194 | ||
190e4fbf N |
195 | typedef int (recdir_func)(struct dentry *, struct dentry *); |
196 | ||
05f4f678 BF |
197 | struct name_list { |
198 | char name[HEXDIR_LEN]; | |
190e4fbf N |
199 | struct list_head list; |
200 | }; | |
201 | ||
190e4fbf | 202 | static int |
05f4f678 | 203 | nfsd4_build_namelist(void *arg, const char *name, int namlen, |
afefdbb2 | 204 | loff_t offset, u64 ino, unsigned int d_type) |
190e4fbf | 205 | { |
05f4f678 BF |
206 | struct list_head *names = arg; |
207 | struct name_list *entry; | |
190e4fbf | 208 | |
05f4f678 | 209 | if (namlen != HEXDIR_LEN - 1) |
b37ad28b | 210 | return 0; |
05f4f678 BF |
211 | entry = kmalloc(sizeof(struct name_list), GFP_KERNEL); |
212 | if (entry == NULL) | |
190e4fbf | 213 | return -ENOMEM; |
05f4f678 BF |
214 | memcpy(entry->name, name, HEXDIR_LEN - 1); |
215 | entry->name[HEXDIR_LEN - 1] = '\0'; | |
216 | list_add(&entry->list, names); | |
190e4fbf N |
217 | return 0; |
218 | } | |
219 | ||
220 | static int | |
5b4b299c | 221 | nfsd4_list_rec_dir(recdir_func *f) |
190e4fbf | 222 | { |
d84f4f99 | 223 | const struct cred *original_cred; |
5b4b299c | 224 | struct dentry *dir = rec_file->f_path.dentry; |
05f4f678 | 225 | LIST_HEAD(names); |
190e4fbf N |
226 | int status; |
227 | ||
d84f4f99 DH |
228 | status = nfs4_save_creds(&original_cred); |
229 | if (status < 0) | |
230 | return status; | |
190e4fbf | 231 | |
5b4b299c AV |
232 | status = vfs_llseek(rec_file, 0, SEEK_SET); |
233 | if (status < 0) { | |
234 | nfs4_reset_creds(original_cred); | |
235 | return status; | |
236 | } | |
237 | ||
238 | status = vfs_readdir(rec_file, nfsd4_build_namelist, &names); | |
8daed1e5 | 239 | mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT); |
05f4f678 | 240 | while (!list_empty(&names)) { |
5b4b299c | 241 | struct name_list *entry; |
05f4f678 | 242 | entry = list_entry(names.next, struct name_list, list); |
5b4b299c AV |
243 | if (!status) { |
244 | struct dentry *dentry; | |
245 | dentry = lookup_one_len(entry->name, dir, HEXDIR_LEN-1); | |
246 | if (IS_ERR(dentry)) { | |
247 | status = PTR_ERR(dentry); | |
248 | break; | |
249 | } | |
250 | status = f(dir, dentry); | |
251 | dput(dentry); | |
05f4f678 | 252 | } |
05f4f678 BF |
253 | list_del(&entry->list); |
254 | kfree(entry); | |
190e4fbf | 255 | } |
2f9092e1 | 256 | mutex_unlock(&dir->d_inode->i_mutex); |
d84f4f99 | 257 | nfs4_reset_creds(original_cred); |
190e4fbf N |
258 | return status; |
259 | } | |
260 | ||
c7b9a459 N |
261 | static int |
262 | nfsd4_unlink_clid_dir(char *name, int namlen) | |
263 | { | |
e970a573 | 264 | struct dentry *dir, *dentry; |
c7b9a459 N |
265 | int status; |
266 | ||
267 | dprintk("NFSD: nfsd4_unlink_clid_dir. name %.*s\n", namlen, name); | |
268 | ||
e970a573 CH |
269 | dir = rec_file->f_path.dentry; |
270 | mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT); | |
271 | dentry = lookup_one_len(name, dir, namlen); | |
c7b9a459 N |
272 | if (IS_ERR(dentry)) { |
273 | status = PTR_ERR(dentry); | |
2f9092e1 | 274 | goto out_unlock; |
c7b9a459 N |
275 | } |
276 | status = -ENOENT; | |
277 | if (!dentry->d_inode) | |
278 | goto out; | |
e970a573 | 279 | status = vfs_rmdir(dir->d_inode, dentry); |
c7b9a459 N |
280 | out: |
281 | dput(dentry); | |
2f9092e1 | 282 | out_unlock: |
e970a573 | 283 | mutex_unlock(&dir->d_inode->i_mutex); |
c7b9a459 N |
284 | return status; |
285 | } | |
286 | ||
2a4317c5 | 287 | static void |
c7b9a459 N |
288 | nfsd4_remove_clid_dir(struct nfs4_client *clp) |
289 | { | |
d84f4f99 | 290 | const struct cred *original_cred; |
c7b9a459 N |
291 | int status; |
292 | ||
a52d726b | 293 | if (!rec_file || !test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) |
c7b9a459 N |
294 | return; |
295 | ||
a561be71 | 296 | status = mnt_want_write_file(rec_file); |
0622753b DH |
297 | if (status) |
298 | goto out; | |
a52d726b | 299 | clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); |
d84f4f99 DH |
300 | |
301 | status = nfs4_save_creds(&original_cred); | |
302 | if (status < 0) | |
303 | goto out; | |
304 | ||
c7b9a459 | 305 | status = nfsd4_unlink_clid_dir(clp->cl_recdir, HEXDIR_LEN-1); |
d84f4f99 | 306 | nfs4_reset_creds(original_cred); |
c7b9a459 | 307 | if (status == 0) |
8018ab05 | 308 | vfs_fsync(rec_file, 0); |
2a79f17e | 309 | mnt_drop_write_file(rec_file); |
0622753b | 310 | out: |
c7b9a459 N |
311 | if (status) |
312 | printk("NFSD: Failed to remove expired client state directory" | |
313 | " %.*s\n", HEXDIR_LEN, clp->cl_recdir); | |
c7b9a459 N |
314 | } |
315 | ||
316 | static int | |
317 | purge_old(struct dentry *parent, struct dentry *child) | |
318 | { | |
319 | int status; | |
320 | ||
a1bcecd2 | 321 | if (nfs4_has_reclaimed_state(child->d_name.name, false)) |
b37ad28b | 322 | return 0; |
c7b9a459 | 323 | |
2f9092e1 | 324 | status = vfs_rmdir(parent->d_inode, child); |
c7b9a459 N |
325 | if (status) |
326 | printk("failed to remove client recovery directory %s\n", | |
327 | child->d_name.name); | |
328 | /* Keep trying, success or failure: */ | |
b37ad28b | 329 | return 0; |
c7b9a459 N |
330 | } |
331 | ||
2a4317c5 JL |
332 | static void |
333 | nfsd4_recdir_purge_old(struct net *net, time_t boot_time) | |
334 | { | |
c7b9a459 N |
335 | int status; |
336 | ||
e970a573 | 337 | if (!rec_file) |
c7b9a459 | 338 | return; |
a561be71 | 339 | status = mnt_want_write_file(rec_file); |
0622753b DH |
340 | if (status) |
341 | goto out; | |
5b4b299c | 342 | status = nfsd4_list_rec_dir(purge_old); |
c7b9a459 | 343 | if (status == 0) |
8018ab05 | 344 | vfs_fsync(rec_file, 0); |
2a79f17e | 345 | mnt_drop_write_file(rec_file); |
0622753b | 346 | out: |
c7b9a459 N |
347 | if (status) |
348 | printk("nfsd4: failed to purge old clients from recovery" | |
e970a573 | 349 | " directory %s\n", rec_file->f_path.dentry->d_name.name); |
c7b9a459 N |
350 | } |
351 | ||
190e4fbf N |
352 | static int |
353 | load_recdir(struct dentry *parent, struct dentry *child) | |
354 | { | |
355 | if (child->d_name.len != HEXDIR_LEN - 1) { | |
356 | printk("nfsd4: illegal name %s in recovery directory\n", | |
357 | child->d_name.name); | |
358 | /* Keep trying; maybe the others are OK: */ | |
b37ad28b | 359 | return 0; |
190e4fbf N |
360 | } |
361 | nfs4_client_to_reclaim(child->d_name.name); | |
b37ad28b | 362 | return 0; |
190e4fbf N |
363 | } |
364 | ||
2a4317c5 | 365 | static int |
190e4fbf N |
366 | nfsd4_recdir_load(void) { |
367 | int status; | |
368 | ||
e970a573 CH |
369 | if (!rec_file) |
370 | return 0; | |
371 | ||
5b4b299c | 372 | status = nfsd4_list_rec_dir(load_recdir); |
190e4fbf N |
373 | if (status) |
374 | printk("nfsd4: failed loading clients from recovery" | |
e970a573 | 375 | " directory %s\n", rec_file->f_path.dentry->d_name.name); |
190e4fbf N |
376 | return status; |
377 | } | |
378 | ||
379 | /* | |
380 | * Hold reference to the recovery directory. | |
381 | */ | |
382 | ||
2a4317c5 JL |
383 | static int |
384 | nfsd4_init_recdir(void) | |
190e4fbf | 385 | { |
d84f4f99 DH |
386 | const struct cred *original_cred; |
387 | int status; | |
190e4fbf N |
388 | |
389 | printk("NFSD: Using %s as the NFSv4 state recovery directory\n", | |
48483bf2 | 390 | user_recovery_dirname); |
190e4fbf | 391 | |
e970a573 | 392 | BUG_ON(rec_file); |
190e4fbf | 393 | |
d84f4f99 DH |
394 | status = nfs4_save_creds(&original_cred); |
395 | if (status < 0) { | |
396 | printk("NFSD: Unable to change credentials to find recovery" | |
397 | " directory: error %d\n", | |
398 | status); | |
2a4317c5 | 399 | return status; |
d84f4f99 | 400 | } |
190e4fbf | 401 | |
48483bf2 | 402 | rec_file = filp_open(user_recovery_dirname, O_RDONLY | O_DIRECTORY, 0); |
e970a573 | 403 | if (IS_ERR(rec_file)) { |
c2642ab0 | 404 | printk("NFSD: unable to find recovery directory %s\n", |
48483bf2 | 405 | user_recovery_dirname); |
2a4317c5 | 406 | status = PTR_ERR(rec_file); |
e970a573 CH |
407 | rec_file = NULL; |
408 | } | |
190e4fbf | 409 | |
d84f4f99 | 410 | nfs4_reset_creds(original_cred); |
2a4317c5 | 411 | return status; |
190e4fbf N |
412 | } |
413 | ||
2a4317c5 JL |
414 | static int |
415 | nfsd4_load_reboot_recovery_data(struct net *net) | |
416 | { | |
417 | int status; | |
418 | ||
419 | nfs4_lock_state(); | |
420 | status = nfsd4_init_recdir(); | |
421 | if (!status) | |
422 | status = nfsd4_recdir_load(); | |
423 | nfs4_unlock_state(); | |
424 | if (status) | |
425 | printk(KERN_ERR "NFSD: Failure reading reboot recovery data\n"); | |
426 | return status; | |
427 | } | |
428 | ||
429 | static void | |
190e4fbf N |
430 | nfsd4_shutdown_recdir(void) |
431 | { | |
e970a573 | 432 | if (!rec_file) |
190e4fbf | 433 | return; |
e970a573 CH |
434 | fput(rec_file); |
435 | rec_file = NULL; | |
190e4fbf | 436 | } |
48483bf2 | 437 | |
2a4317c5 JL |
438 | static void |
439 | nfsd4_legacy_tracking_exit(struct net *net) | |
440 | { | |
441 | nfs4_release_reclaim(); | |
442 | nfsd4_shutdown_recdir(); | |
443 | } | |
444 | ||
48483bf2 BF |
445 | /* |
446 | * Change the NFSv4 recovery directory to recdir. | |
447 | */ | |
448 | int | |
449 | nfs4_reset_recoverydir(char *recdir) | |
450 | { | |
451 | int status; | |
452 | struct path path; | |
453 | ||
454 | status = kern_path(recdir, LOOKUP_FOLLOW, &path); | |
455 | if (status) | |
456 | return status; | |
457 | status = -ENOTDIR; | |
458 | if (S_ISDIR(path.dentry->d_inode->i_mode)) { | |
459 | strcpy(user_recovery_dirname, recdir); | |
460 | status = 0; | |
461 | } | |
462 | path_put(&path); | |
463 | return status; | |
464 | } | |
465 | ||
466 | char * | |
467 | nfs4_recoverydir(void) | |
468 | { | |
469 | return user_recovery_dirname; | |
470 | } | |
2a4317c5 JL |
471 | |
472 | static int | |
473 | nfsd4_check_legacy_client(struct nfs4_client *clp) | |
474 | { | |
475 | /* did we already find that this client is stable? */ | |
476 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
477 | return 0; | |
478 | ||
479 | /* look for it in the reclaim hashtable otherwise */ | |
480 | if (nfsd4_find_reclaim_client(clp)) { | |
481 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
482 | return 0; | |
483 | } | |
484 | ||
485 | return -ENOENT; | |
486 | } | |
487 | ||
488 | static struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = { | |
489 | .init = nfsd4_load_reboot_recovery_data, | |
490 | .exit = nfsd4_legacy_tracking_exit, | |
491 | .create = nfsd4_create_clid_dir, | |
492 | .remove = nfsd4_remove_clid_dir, | |
493 | .check = nfsd4_check_legacy_client, | |
494 | .grace_done = nfsd4_recdir_purge_old, | |
495 | }; | |
496 | ||
f3f80148 JL |
497 | /* Globals */ |
498 | #define NFSD_PIPE_DIR "nfsd" | |
499 | #define NFSD_CLD_PIPE "cld" | |
500 | ||
501 | /* per-net-ns structure for holding cld upcall info */ | |
502 | struct cld_net { | |
503 | struct rpc_pipe *cn_pipe; | |
504 | spinlock_t cn_lock; | |
505 | struct list_head cn_list; | |
506 | unsigned int cn_xid; | |
507 | }; | |
508 | ||
509 | struct cld_upcall { | |
510 | struct list_head cu_list; | |
511 | struct cld_net *cu_net; | |
512 | struct task_struct *cu_task; | |
513 | struct cld_msg cu_msg; | |
514 | }; | |
515 | ||
516 | static int | |
517 | __cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
518 | { | |
519 | int ret; | |
520 | struct rpc_pipe_msg msg; | |
521 | ||
522 | memset(&msg, 0, sizeof(msg)); | |
523 | msg.data = cmsg; | |
524 | msg.len = sizeof(*cmsg); | |
525 | ||
526 | /* | |
527 | * Set task state before we queue the upcall. That prevents | |
528 | * wake_up_process in the downcall from racing with schedule. | |
529 | */ | |
530 | set_current_state(TASK_UNINTERRUPTIBLE); | |
531 | ret = rpc_queue_upcall(pipe, &msg); | |
532 | if (ret < 0) { | |
533 | set_current_state(TASK_RUNNING); | |
534 | goto out; | |
535 | } | |
536 | ||
537 | schedule(); | |
538 | set_current_state(TASK_RUNNING); | |
539 | ||
540 | if (msg.errno < 0) | |
541 | ret = msg.errno; | |
542 | out: | |
543 | return ret; | |
544 | } | |
545 | ||
546 | static int | |
547 | cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
548 | { | |
549 | int ret; | |
550 | ||
551 | /* | |
552 | * -EAGAIN occurs when pipe is closed and reopened while there are | |
553 | * upcalls queued. | |
554 | */ | |
555 | do { | |
556 | ret = __cld_pipe_upcall(pipe, cmsg); | |
557 | } while (ret == -EAGAIN); | |
558 | ||
559 | return ret; | |
560 | } | |
561 | ||
562 | static ssize_t | |
563 | cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) | |
564 | { | |
565 | struct cld_upcall *tmp, *cup; | |
566 | struct cld_msg *cmsg = (struct cld_msg *)src; | |
567 | uint32_t xid; | |
568 | struct nfsd_net *nn = net_generic(filp->f_dentry->d_sb->s_fs_info, | |
569 | nfsd_net_id); | |
570 | struct cld_net *cn = nn->cld_net; | |
571 | ||
572 | if (mlen != sizeof(*cmsg)) { | |
573 | dprintk("%s: got %lu bytes, expected %lu\n", __func__, mlen, | |
574 | sizeof(*cmsg)); | |
575 | return -EINVAL; | |
576 | } | |
577 | ||
578 | /* copy just the xid so we can try to find that */ | |
579 | if (copy_from_user(&xid, &cmsg->cm_xid, sizeof(xid)) != 0) { | |
580 | dprintk("%s: error when copying xid from userspace", __func__); | |
581 | return -EFAULT; | |
582 | } | |
583 | ||
584 | /* walk the list and find corresponding xid */ | |
585 | cup = NULL; | |
586 | spin_lock(&cn->cn_lock); | |
587 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
588 | if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) { | |
589 | cup = tmp; | |
590 | list_del_init(&cup->cu_list); | |
591 | break; | |
592 | } | |
593 | } | |
594 | spin_unlock(&cn->cn_lock); | |
595 | ||
596 | /* couldn't find upcall? */ | |
597 | if (!cup) { | |
598 | dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, | |
599 | cup->cu_msg.cm_xid); | |
600 | return -EINVAL; | |
601 | } | |
602 | ||
603 | if (copy_from_user(&cup->cu_msg, src, mlen) != 0) | |
604 | return -EFAULT; | |
605 | ||
606 | wake_up_process(cup->cu_task); | |
607 | return mlen; | |
608 | } | |
609 | ||
610 | static void | |
611 | cld_pipe_destroy_msg(struct rpc_pipe_msg *msg) | |
612 | { | |
613 | struct cld_msg *cmsg = msg->data; | |
614 | struct cld_upcall *cup = container_of(cmsg, struct cld_upcall, | |
615 | cu_msg); | |
616 | ||
617 | /* errno >= 0 means we got a downcall */ | |
618 | if (msg->errno >= 0) | |
619 | return; | |
620 | ||
621 | wake_up_process(cup->cu_task); | |
622 | } | |
623 | ||
624 | static const struct rpc_pipe_ops cld_upcall_ops = { | |
625 | .upcall = rpc_pipe_generic_upcall, | |
626 | .downcall = cld_pipe_downcall, | |
627 | .destroy_msg = cld_pipe_destroy_msg, | |
628 | }; | |
629 | ||
630 | static struct dentry * | |
631 | nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe) | |
632 | { | |
633 | struct dentry *dir, *dentry; | |
634 | ||
635 | dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR); | |
636 | if (dir == NULL) | |
637 | return ERR_PTR(-ENOENT); | |
638 | dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe); | |
639 | dput(dir); | |
640 | return dentry; | |
641 | } | |
642 | ||
643 | static void | |
644 | nfsd4_cld_unregister_sb(struct rpc_pipe *pipe) | |
645 | { | |
646 | if (pipe->dentry) | |
647 | rpc_unlink(pipe->dentry); | |
648 | } | |
649 | ||
650 | static struct dentry * | |
651 | nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe) | |
652 | { | |
653 | struct super_block *sb; | |
654 | struct dentry *dentry; | |
655 | ||
656 | sb = rpc_get_sb_net(net); | |
657 | if (!sb) | |
658 | return NULL; | |
659 | dentry = nfsd4_cld_register_sb(sb, pipe); | |
660 | rpc_put_sb_net(net); | |
661 | return dentry; | |
662 | } | |
663 | ||
664 | static void | |
665 | nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe) | |
666 | { | |
667 | struct super_block *sb; | |
668 | ||
669 | sb = rpc_get_sb_net(net); | |
670 | if (sb) { | |
671 | nfsd4_cld_unregister_sb(pipe); | |
672 | rpc_put_sb_net(net); | |
673 | } | |
674 | } | |
675 | ||
676 | /* Initialize rpc_pipefs pipe for communication with client tracking daemon */ | |
677 | static int | |
678 | nfsd4_init_cld_pipe(struct net *net) | |
679 | { | |
680 | int ret; | |
681 | struct dentry *dentry; | |
682 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
683 | struct cld_net *cn; | |
684 | ||
685 | if (nn->cld_net) | |
686 | return 0; | |
687 | ||
688 | cn = kzalloc(sizeof(*cn), GFP_KERNEL); | |
689 | if (!cn) { | |
690 | ret = -ENOMEM; | |
691 | goto err; | |
692 | } | |
693 | ||
694 | cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); | |
695 | if (IS_ERR(cn->cn_pipe)) { | |
696 | ret = PTR_ERR(cn->cn_pipe); | |
697 | goto err; | |
698 | } | |
699 | spin_lock_init(&cn->cn_lock); | |
700 | INIT_LIST_HEAD(&cn->cn_list); | |
701 | ||
702 | dentry = nfsd4_cld_register_net(net, cn->cn_pipe); | |
703 | if (IS_ERR(dentry)) { | |
704 | ret = PTR_ERR(dentry); | |
705 | goto err_destroy_data; | |
706 | } | |
707 | ||
708 | cn->cn_pipe->dentry = dentry; | |
709 | nn->cld_net = cn; | |
710 | return 0; | |
711 | ||
712 | err_destroy_data: | |
713 | rpc_destroy_pipe_data(cn->cn_pipe); | |
714 | err: | |
715 | kfree(cn); | |
716 | printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n", | |
717 | ret); | |
718 | return ret; | |
719 | } | |
720 | ||
721 | static void | |
722 | nfsd4_remove_cld_pipe(struct net *net) | |
723 | { | |
724 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
725 | struct cld_net *cn = nn->cld_net; | |
726 | ||
727 | nfsd4_cld_unregister_net(net, cn->cn_pipe); | |
728 | rpc_destroy_pipe_data(cn->cn_pipe); | |
729 | kfree(nn->cld_net); | |
730 | nn->cld_net = NULL; | |
731 | } | |
732 | ||
733 | static struct cld_upcall * | |
734 | alloc_cld_upcall(struct cld_net *cn) | |
735 | { | |
736 | struct cld_upcall *new, *tmp; | |
737 | ||
738 | new = kzalloc(sizeof(*new), GFP_KERNEL); | |
739 | if (!new) | |
740 | return new; | |
741 | ||
742 | /* FIXME: hard cap on number in flight? */ | |
743 | restart_search: | |
744 | spin_lock(&cn->cn_lock); | |
745 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
746 | if (tmp->cu_msg.cm_xid == cn->cn_xid) { | |
747 | cn->cn_xid++; | |
748 | spin_unlock(&cn->cn_lock); | |
749 | goto restart_search; | |
750 | } | |
751 | } | |
752 | new->cu_task = current; | |
753 | new->cu_msg.cm_vers = CLD_UPCALL_VERSION; | |
754 | put_unaligned(cn->cn_xid++, &new->cu_msg.cm_xid); | |
755 | new->cu_net = cn; | |
756 | list_add(&new->cu_list, &cn->cn_list); | |
757 | spin_unlock(&cn->cn_lock); | |
758 | ||
759 | dprintk("%s: allocated xid %u\n", __func__, new->cu_msg.cm_xid); | |
760 | ||
761 | return new; | |
762 | } | |
763 | ||
764 | static void | |
765 | free_cld_upcall(struct cld_upcall *victim) | |
766 | { | |
767 | struct cld_net *cn = victim->cu_net; | |
768 | ||
769 | spin_lock(&cn->cn_lock); | |
770 | list_del(&victim->cu_list); | |
771 | spin_unlock(&cn->cn_lock); | |
772 | kfree(victim); | |
773 | } | |
774 | ||
775 | /* Ask daemon to create a new record */ | |
776 | static void | |
777 | nfsd4_cld_create(struct nfs4_client *clp) | |
778 | { | |
779 | int ret; | |
780 | struct cld_upcall *cup; | |
781 | /* FIXME: determine net from clp */ | |
782 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
783 | struct cld_net *cn = nn->cld_net; | |
784 | ||
785 | /* Don't upcall if it's already stored */ | |
786 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
787 | return; | |
788 | ||
789 | cup = alloc_cld_upcall(cn); | |
790 | if (!cup) { | |
791 | ret = -ENOMEM; | |
792 | goto out_err; | |
793 | } | |
794 | ||
795 | cup->cu_msg.cm_cmd = Cld_Create; | |
796 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
797 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
798 | clp->cl_name.len); | |
799 | ||
800 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
801 | if (!ret) { | |
802 | ret = cup->cu_msg.cm_status; | |
803 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
804 | } | |
805 | ||
806 | free_cld_upcall(cup); | |
807 | out_err: | |
808 | if (ret) | |
809 | printk(KERN_ERR "NFSD: Unable to create client " | |
810 | "record on stable storage: %d\n", ret); | |
811 | } | |
812 | ||
813 | /* Ask daemon to create a new record */ | |
814 | static void | |
815 | nfsd4_cld_remove(struct nfs4_client *clp) | |
816 | { | |
817 | int ret; | |
818 | struct cld_upcall *cup; | |
819 | /* FIXME: determine net from clp */ | |
820 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
821 | struct cld_net *cn = nn->cld_net; | |
822 | ||
823 | /* Don't upcall if it's already removed */ | |
824 | if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
825 | return; | |
826 | ||
827 | cup = alloc_cld_upcall(cn); | |
828 | if (!cup) { | |
829 | ret = -ENOMEM; | |
830 | goto out_err; | |
831 | } | |
832 | ||
833 | cup->cu_msg.cm_cmd = Cld_Remove; | |
834 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
835 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
836 | clp->cl_name.len); | |
837 | ||
838 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
839 | if (!ret) { | |
840 | ret = cup->cu_msg.cm_status; | |
841 | clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
842 | } | |
843 | ||
844 | free_cld_upcall(cup); | |
845 | out_err: | |
846 | if (ret) | |
847 | printk(KERN_ERR "NFSD: Unable to remove client " | |
848 | "record from stable storage: %d\n", ret); | |
849 | } | |
850 | ||
851 | /* Check for presence of a record, and update its timestamp */ | |
852 | static int | |
853 | nfsd4_cld_check(struct nfs4_client *clp) | |
854 | { | |
855 | int ret; | |
856 | struct cld_upcall *cup; | |
857 | /* FIXME: determine net from clp */ | |
858 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
859 | struct cld_net *cn = nn->cld_net; | |
860 | ||
861 | /* Don't upcall if one was already stored during this grace pd */ | |
862 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
863 | return 0; | |
864 | ||
865 | cup = alloc_cld_upcall(cn); | |
866 | if (!cup) { | |
867 | printk(KERN_ERR "NFSD: Unable to check client record on " | |
868 | "stable storage: %d\n", -ENOMEM); | |
869 | return -ENOMEM; | |
870 | } | |
871 | ||
872 | cup->cu_msg.cm_cmd = Cld_Check; | |
873 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
874 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
875 | clp->cl_name.len); | |
876 | ||
877 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
878 | if (!ret) { | |
879 | ret = cup->cu_msg.cm_status; | |
880 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
881 | } | |
882 | ||
883 | free_cld_upcall(cup); | |
884 | return ret; | |
885 | } | |
886 | ||
887 | static void | |
888 | nfsd4_cld_grace_done(struct net *net, time_t boot_time) | |
889 | { | |
890 | int ret; | |
891 | struct cld_upcall *cup; | |
892 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
893 | struct cld_net *cn = nn->cld_net; | |
894 | ||
895 | cup = alloc_cld_upcall(cn); | |
896 | if (!cup) { | |
897 | ret = -ENOMEM; | |
898 | goto out_err; | |
899 | } | |
900 | ||
901 | cup->cu_msg.cm_cmd = Cld_GraceDone; | |
902 | cup->cu_msg.cm_u.cm_gracetime = (int64_t)boot_time; | |
903 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
904 | if (!ret) | |
905 | ret = cup->cu_msg.cm_status; | |
906 | ||
907 | free_cld_upcall(cup); | |
908 | out_err: | |
909 | if (ret) | |
910 | printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret); | |
911 | } | |
912 | ||
913 | static struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = { | |
914 | .init = nfsd4_init_cld_pipe, | |
915 | .exit = nfsd4_remove_cld_pipe, | |
916 | .create = nfsd4_cld_create, | |
917 | .remove = nfsd4_cld_remove, | |
918 | .check = nfsd4_cld_check, | |
919 | .grace_done = nfsd4_cld_grace_done, | |
920 | }; | |
921 | ||
2a4317c5 JL |
922 | int |
923 | nfsd4_client_tracking_init(struct net *net) | |
924 | { | |
925 | int status; | |
f3f80148 | 926 | struct path path; |
2a4317c5 | 927 | |
f3f80148 JL |
928 | if (!client_tracking_ops) { |
929 | client_tracking_ops = &nfsd4_cld_tracking_ops; | |
930 | status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path); | |
931 | if (!status) { | |
932 | if (S_ISDIR(path.dentry->d_inode->i_mode)) | |
933 | client_tracking_ops = | |
934 | &nfsd4_legacy_tracking_ops; | |
935 | path_put(&path); | |
936 | } | |
937 | } | |
2a4317c5 JL |
938 | |
939 | status = client_tracking_ops->init(net); | |
940 | if (status) { | |
941 | printk(KERN_WARNING "NFSD: Unable to initialize client " | |
942 | "recovery tracking! (%d)\n", status); | |
943 | client_tracking_ops = NULL; | |
944 | } | |
945 | return status; | |
946 | } | |
947 | ||
948 | void | |
949 | nfsd4_client_tracking_exit(struct net *net) | |
950 | { | |
951 | if (client_tracking_ops) { | |
952 | client_tracking_ops->exit(net); | |
953 | client_tracking_ops = NULL; | |
954 | } | |
955 | } | |
956 | ||
957 | void | |
958 | nfsd4_client_record_create(struct nfs4_client *clp) | |
959 | { | |
960 | if (client_tracking_ops) | |
961 | client_tracking_ops->create(clp); | |
962 | } | |
963 | ||
964 | void | |
965 | nfsd4_client_record_remove(struct nfs4_client *clp) | |
966 | { | |
967 | if (client_tracking_ops) | |
968 | client_tracking_ops->remove(clp); | |
969 | } | |
970 | ||
971 | int | |
972 | nfsd4_client_record_check(struct nfs4_client *clp) | |
973 | { | |
974 | if (client_tracking_ops) | |
975 | return client_tracking_ops->check(clp); | |
976 | ||
977 | return -EOPNOTSUPP; | |
978 | } | |
979 | ||
980 | void | |
981 | nfsd4_record_grace_done(struct net *net, time_t boot_time) | |
982 | { | |
983 | if (client_tracking_ops) | |
984 | client_tracking_ops->grace_done(net, boot_time); | |
985 | } | |
813fd320 JL |
986 | |
987 | static int | |
988 | rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr) | |
989 | { | |
990 | struct super_block *sb = ptr; | |
991 | struct net *net = sb->s_fs_info; | |
992 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
993 | struct cld_net *cn = nn->cld_net; | |
994 | struct dentry *dentry; | |
995 | int ret = 0; | |
996 | ||
997 | if (!try_module_get(THIS_MODULE)) | |
998 | return 0; | |
999 | ||
1000 | if (!cn) { | |
1001 | module_put(THIS_MODULE); | |
1002 | return 0; | |
1003 | } | |
1004 | ||
1005 | switch (event) { | |
1006 | case RPC_PIPEFS_MOUNT: | |
1007 | dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe); | |
1008 | if (IS_ERR(dentry)) { | |
1009 | ret = PTR_ERR(dentry); | |
1010 | break; | |
1011 | } | |
1012 | cn->cn_pipe->dentry = dentry; | |
1013 | break; | |
1014 | case RPC_PIPEFS_UMOUNT: | |
1015 | if (cn->cn_pipe->dentry) | |
1016 | nfsd4_cld_unregister_sb(cn->cn_pipe); | |
1017 | break; | |
1018 | default: | |
1019 | ret = -ENOTSUPP; | |
1020 | break; | |
1021 | } | |
1022 | module_put(THIS_MODULE); | |
1023 | return ret; | |
1024 | } | |
1025 | ||
1026 | struct notifier_block nfsd4_cld_block = { | |
1027 | .notifier_call = rpc_pipefs_event, | |
1028 | }; |