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 | ||
cc27e0d4 JL |
419 | /* XXX: The legacy code won't work in a container */ |
420 | if (net != &init_net) { | |
421 | WARN(1, KERN_ERR "NFSD: attempt to initialize legacy client " | |
422 | "tracking in a container!\n"); | |
423 | return -EINVAL; | |
424 | } | |
425 | ||
2a4317c5 JL |
426 | nfs4_lock_state(); |
427 | status = nfsd4_init_recdir(); | |
428 | if (!status) | |
429 | status = nfsd4_recdir_load(); | |
430 | nfs4_unlock_state(); | |
431 | if (status) | |
432 | printk(KERN_ERR "NFSD: Failure reading reboot recovery data\n"); | |
433 | return status; | |
434 | } | |
435 | ||
436 | static void | |
190e4fbf N |
437 | nfsd4_shutdown_recdir(void) |
438 | { | |
e970a573 | 439 | if (!rec_file) |
190e4fbf | 440 | return; |
e970a573 CH |
441 | fput(rec_file); |
442 | rec_file = NULL; | |
190e4fbf | 443 | } |
48483bf2 | 444 | |
2a4317c5 JL |
445 | static void |
446 | nfsd4_legacy_tracking_exit(struct net *net) | |
447 | { | |
448 | nfs4_release_reclaim(); | |
449 | nfsd4_shutdown_recdir(); | |
450 | } | |
451 | ||
48483bf2 BF |
452 | /* |
453 | * Change the NFSv4 recovery directory to recdir. | |
454 | */ | |
455 | int | |
456 | nfs4_reset_recoverydir(char *recdir) | |
457 | { | |
458 | int status; | |
459 | struct path path; | |
460 | ||
461 | status = kern_path(recdir, LOOKUP_FOLLOW, &path); | |
462 | if (status) | |
463 | return status; | |
464 | status = -ENOTDIR; | |
465 | if (S_ISDIR(path.dentry->d_inode->i_mode)) { | |
466 | strcpy(user_recovery_dirname, recdir); | |
467 | status = 0; | |
468 | } | |
469 | path_put(&path); | |
470 | return status; | |
471 | } | |
472 | ||
473 | char * | |
474 | nfs4_recoverydir(void) | |
475 | { | |
476 | return user_recovery_dirname; | |
477 | } | |
2a4317c5 JL |
478 | |
479 | static int | |
480 | nfsd4_check_legacy_client(struct nfs4_client *clp) | |
481 | { | |
482 | /* did we already find that this client is stable? */ | |
483 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
484 | return 0; | |
485 | ||
486 | /* look for it in the reclaim hashtable otherwise */ | |
487 | if (nfsd4_find_reclaim_client(clp)) { | |
488 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
489 | return 0; | |
490 | } | |
491 | ||
492 | return -ENOENT; | |
493 | } | |
494 | ||
495 | static struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = { | |
496 | .init = nfsd4_load_reboot_recovery_data, | |
497 | .exit = nfsd4_legacy_tracking_exit, | |
498 | .create = nfsd4_create_clid_dir, | |
499 | .remove = nfsd4_remove_clid_dir, | |
500 | .check = nfsd4_check_legacy_client, | |
501 | .grace_done = nfsd4_recdir_purge_old, | |
502 | }; | |
503 | ||
f3f80148 JL |
504 | /* Globals */ |
505 | #define NFSD_PIPE_DIR "nfsd" | |
506 | #define NFSD_CLD_PIPE "cld" | |
507 | ||
508 | /* per-net-ns structure for holding cld upcall info */ | |
509 | struct cld_net { | |
510 | struct rpc_pipe *cn_pipe; | |
511 | spinlock_t cn_lock; | |
512 | struct list_head cn_list; | |
513 | unsigned int cn_xid; | |
514 | }; | |
515 | ||
516 | struct cld_upcall { | |
517 | struct list_head cu_list; | |
518 | struct cld_net *cu_net; | |
519 | struct task_struct *cu_task; | |
520 | struct cld_msg cu_msg; | |
521 | }; | |
522 | ||
523 | static int | |
524 | __cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
525 | { | |
526 | int ret; | |
527 | struct rpc_pipe_msg msg; | |
528 | ||
529 | memset(&msg, 0, sizeof(msg)); | |
530 | msg.data = cmsg; | |
531 | msg.len = sizeof(*cmsg); | |
532 | ||
533 | /* | |
534 | * Set task state before we queue the upcall. That prevents | |
535 | * wake_up_process in the downcall from racing with schedule. | |
536 | */ | |
537 | set_current_state(TASK_UNINTERRUPTIBLE); | |
538 | ret = rpc_queue_upcall(pipe, &msg); | |
539 | if (ret < 0) { | |
540 | set_current_state(TASK_RUNNING); | |
541 | goto out; | |
542 | } | |
543 | ||
544 | schedule(); | |
545 | set_current_state(TASK_RUNNING); | |
546 | ||
547 | if (msg.errno < 0) | |
548 | ret = msg.errno; | |
549 | out: | |
550 | return ret; | |
551 | } | |
552 | ||
553 | static int | |
554 | cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
555 | { | |
556 | int ret; | |
557 | ||
558 | /* | |
559 | * -EAGAIN occurs when pipe is closed and reopened while there are | |
560 | * upcalls queued. | |
561 | */ | |
562 | do { | |
563 | ret = __cld_pipe_upcall(pipe, cmsg); | |
564 | } while (ret == -EAGAIN); | |
565 | ||
566 | return ret; | |
567 | } | |
568 | ||
569 | static ssize_t | |
570 | cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) | |
571 | { | |
572 | struct cld_upcall *tmp, *cup; | |
573 | struct cld_msg *cmsg = (struct cld_msg *)src; | |
574 | uint32_t xid; | |
575 | struct nfsd_net *nn = net_generic(filp->f_dentry->d_sb->s_fs_info, | |
576 | nfsd_net_id); | |
577 | struct cld_net *cn = nn->cld_net; | |
578 | ||
579 | if (mlen != sizeof(*cmsg)) { | |
580 | dprintk("%s: got %lu bytes, expected %lu\n", __func__, mlen, | |
581 | sizeof(*cmsg)); | |
582 | return -EINVAL; | |
583 | } | |
584 | ||
585 | /* copy just the xid so we can try to find that */ | |
586 | if (copy_from_user(&xid, &cmsg->cm_xid, sizeof(xid)) != 0) { | |
587 | dprintk("%s: error when copying xid from userspace", __func__); | |
588 | return -EFAULT; | |
589 | } | |
590 | ||
591 | /* walk the list and find corresponding xid */ | |
592 | cup = NULL; | |
593 | spin_lock(&cn->cn_lock); | |
594 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
595 | if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) { | |
596 | cup = tmp; | |
597 | list_del_init(&cup->cu_list); | |
598 | break; | |
599 | } | |
600 | } | |
601 | spin_unlock(&cn->cn_lock); | |
602 | ||
603 | /* couldn't find upcall? */ | |
604 | if (!cup) { | |
21f72c9f | 605 | dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, xid); |
f3f80148 JL |
606 | return -EINVAL; |
607 | } | |
608 | ||
609 | if (copy_from_user(&cup->cu_msg, src, mlen) != 0) | |
610 | return -EFAULT; | |
611 | ||
612 | wake_up_process(cup->cu_task); | |
613 | return mlen; | |
614 | } | |
615 | ||
616 | static void | |
617 | cld_pipe_destroy_msg(struct rpc_pipe_msg *msg) | |
618 | { | |
619 | struct cld_msg *cmsg = msg->data; | |
620 | struct cld_upcall *cup = container_of(cmsg, struct cld_upcall, | |
621 | cu_msg); | |
622 | ||
623 | /* errno >= 0 means we got a downcall */ | |
624 | if (msg->errno >= 0) | |
625 | return; | |
626 | ||
627 | wake_up_process(cup->cu_task); | |
628 | } | |
629 | ||
630 | static const struct rpc_pipe_ops cld_upcall_ops = { | |
631 | .upcall = rpc_pipe_generic_upcall, | |
632 | .downcall = cld_pipe_downcall, | |
633 | .destroy_msg = cld_pipe_destroy_msg, | |
634 | }; | |
635 | ||
636 | static struct dentry * | |
637 | nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe) | |
638 | { | |
639 | struct dentry *dir, *dentry; | |
640 | ||
641 | dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR); | |
642 | if (dir == NULL) | |
643 | return ERR_PTR(-ENOENT); | |
644 | dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe); | |
645 | dput(dir); | |
646 | return dentry; | |
647 | } | |
648 | ||
649 | static void | |
650 | nfsd4_cld_unregister_sb(struct rpc_pipe *pipe) | |
651 | { | |
652 | if (pipe->dentry) | |
653 | rpc_unlink(pipe->dentry); | |
654 | } | |
655 | ||
656 | static struct dentry * | |
657 | nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe) | |
658 | { | |
659 | struct super_block *sb; | |
660 | struct dentry *dentry; | |
661 | ||
662 | sb = rpc_get_sb_net(net); | |
663 | if (!sb) | |
664 | return NULL; | |
665 | dentry = nfsd4_cld_register_sb(sb, pipe); | |
666 | rpc_put_sb_net(net); | |
667 | return dentry; | |
668 | } | |
669 | ||
670 | static void | |
671 | nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe) | |
672 | { | |
673 | struct super_block *sb; | |
674 | ||
675 | sb = rpc_get_sb_net(net); | |
676 | if (sb) { | |
677 | nfsd4_cld_unregister_sb(pipe); | |
678 | rpc_put_sb_net(net); | |
679 | } | |
680 | } | |
681 | ||
682 | /* Initialize rpc_pipefs pipe for communication with client tracking daemon */ | |
683 | static int | |
684 | nfsd4_init_cld_pipe(struct net *net) | |
685 | { | |
686 | int ret; | |
687 | struct dentry *dentry; | |
688 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
689 | struct cld_net *cn; | |
690 | ||
691 | if (nn->cld_net) | |
692 | return 0; | |
693 | ||
694 | cn = kzalloc(sizeof(*cn), GFP_KERNEL); | |
695 | if (!cn) { | |
696 | ret = -ENOMEM; | |
697 | goto err; | |
698 | } | |
699 | ||
700 | cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); | |
701 | if (IS_ERR(cn->cn_pipe)) { | |
702 | ret = PTR_ERR(cn->cn_pipe); | |
703 | goto err; | |
704 | } | |
705 | spin_lock_init(&cn->cn_lock); | |
706 | INIT_LIST_HEAD(&cn->cn_list); | |
707 | ||
708 | dentry = nfsd4_cld_register_net(net, cn->cn_pipe); | |
709 | if (IS_ERR(dentry)) { | |
710 | ret = PTR_ERR(dentry); | |
711 | goto err_destroy_data; | |
712 | } | |
713 | ||
714 | cn->cn_pipe->dentry = dentry; | |
715 | nn->cld_net = cn; | |
716 | return 0; | |
717 | ||
718 | err_destroy_data: | |
719 | rpc_destroy_pipe_data(cn->cn_pipe); | |
720 | err: | |
721 | kfree(cn); | |
722 | printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n", | |
723 | ret); | |
724 | return ret; | |
725 | } | |
726 | ||
727 | static void | |
728 | nfsd4_remove_cld_pipe(struct net *net) | |
729 | { | |
730 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
731 | struct cld_net *cn = nn->cld_net; | |
732 | ||
733 | nfsd4_cld_unregister_net(net, cn->cn_pipe); | |
734 | rpc_destroy_pipe_data(cn->cn_pipe); | |
735 | kfree(nn->cld_net); | |
736 | nn->cld_net = NULL; | |
737 | } | |
738 | ||
739 | static struct cld_upcall * | |
740 | alloc_cld_upcall(struct cld_net *cn) | |
741 | { | |
742 | struct cld_upcall *new, *tmp; | |
743 | ||
744 | new = kzalloc(sizeof(*new), GFP_KERNEL); | |
745 | if (!new) | |
746 | return new; | |
747 | ||
748 | /* FIXME: hard cap on number in flight? */ | |
749 | restart_search: | |
750 | spin_lock(&cn->cn_lock); | |
751 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
752 | if (tmp->cu_msg.cm_xid == cn->cn_xid) { | |
753 | cn->cn_xid++; | |
754 | spin_unlock(&cn->cn_lock); | |
755 | goto restart_search; | |
756 | } | |
757 | } | |
758 | new->cu_task = current; | |
759 | new->cu_msg.cm_vers = CLD_UPCALL_VERSION; | |
760 | put_unaligned(cn->cn_xid++, &new->cu_msg.cm_xid); | |
761 | new->cu_net = cn; | |
762 | list_add(&new->cu_list, &cn->cn_list); | |
763 | spin_unlock(&cn->cn_lock); | |
764 | ||
765 | dprintk("%s: allocated xid %u\n", __func__, new->cu_msg.cm_xid); | |
766 | ||
767 | return new; | |
768 | } | |
769 | ||
770 | static void | |
771 | free_cld_upcall(struct cld_upcall *victim) | |
772 | { | |
773 | struct cld_net *cn = victim->cu_net; | |
774 | ||
775 | spin_lock(&cn->cn_lock); | |
776 | list_del(&victim->cu_list); | |
777 | spin_unlock(&cn->cn_lock); | |
778 | kfree(victim); | |
779 | } | |
780 | ||
781 | /* Ask daemon to create a new record */ | |
782 | static void | |
783 | nfsd4_cld_create(struct nfs4_client *clp) | |
784 | { | |
785 | int ret; | |
786 | struct cld_upcall *cup; | |
787 | /* FIXME: determine net from clp */ | |
788 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
789 | struct cld_net *cn = nn->cld_net; | |
790 | ||
791 | /* Don't upcall if it's already stored */ | |
792 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
793 | return; | |
794 | ||
795 | cup = alloc_cld_upcall(cn); | |
796 | if (!cup) { | |
797 | ret = -ENOMEM; | |
798 | goto out_err; | |
799 | } | |
800 | ||
801 | cup->cu_msg.cm_cmd = Cld_Create; | |
802 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
803 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
804 | clp->cl_name.len); | |
805 | ||
806 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
807 | if (!ret) { | |
808 | ret = cup->cu_msg.cm_status; | |
809 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
810 | } | |
811 | ||
812 | free_cld_upcall(cup); | |
813 | out_err: | |
814 | if (ret) | |
815 | printk(KERN_ERR "NFSD: Unable to create client " | |
816 | "record on stable storage: %d\n", ret); | |
817 | } | |
818 | ||
819 | /* Ask daemon to create a new record */ | |
820 | static void | |
821 | nfsd4_cld_remove(struct nfs4_client *clp) | |
822 | { | |
823 | int ret; | |
824 | struct cld_upcall *cup; | |
825 | /* FIXME: determine net from clp */ | |
826 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
827 | struct cld_net *cn = nn->cld_net; | |
828 | ||
829 | /* Don't upcall if it's already removed */ | |
830 | if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
831 | return; | |
832 | ||
833 | cup = alloc_cld_upcall(cn); | |
834 | if (!cup) { | |
835 | ret = -ENOMEM; | |
836 | goto out_err; | |
837 | } | |
838 | ||
839 | cup->cu_msg.cm_cmd = Cld_Remove; | |
840 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
841 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
842 | clp->cl_name.len); | |
843 | ||
844 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
845 | if (!ret) { | |
846 | ret = cup->cu_msg.cm_status; | |
847 | clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
848 | } | |
849 | ||
850 | free_cld_upcall(cup); | |
851 | out_err: | |
852 | if (ret) | |
853 | printk(KERN_ERR "NFSD: Unable to remove client " | |
854 | "record from stable storage: %d\n", ret); | |
855 | } | |
856 | ||
857 | /* Check for presence of a record, and update its timestamp */ | |
858 | static int | |
859 | nfsd4_cld_check(struct nfs4_client *clp) | |
860 | { | |
861 | int ret; | |
862 | struct cld_upcall *cup; | |
863 | /* FIXME: determine net from clp */ | |
864 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
865 | struct cld_net *cn = nn->cld_net; | |
866 | ||
867 | /* Don't upcall if one was already stored during this grace pd */ | |
868 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
869 | return 0; | |
870 | ||
871 | cup = alloc_cld_upcall(cn); | |
872 | if (!cup) { | |
873 | printk(KERN_ERR "NFSD: Unable to check client record on " | |
874 | "stable storage: %d\n", -ENOMEM); | |
875 | return -ENOMEM; | |
876 | } | |
877 | ||
878 | cup->cu_msg.cm_cmd = Cld_Check; | |
879 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
880 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
881 | clp->cl_name.len); | |
882 | ||
883 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
884 | if (!ret) { | |
885 | ret = cup->cu_msg.cm_status; | |
886 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
887 | } | |
888 | ||
889 | free_cld_upcall(cup); | |
890 | return ret; | |
891 | } | |
892 | ||
893 | static void | |
894 | nfsd4_cld_grace_done(struct net *net, time_t boot_time) | |
895 | { | |
896 | int ret; | |
897 | struct cld_upcall *cup; | |
898 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
899 | struct cld_net *cn = nn->cld_net; | |
900 | ||
901 | cup = alloc_cld_upcall(cn); | |
902 | if (!cup) { | |
903 | ret = -ENOMEM; | |
904 | goto out_err; | |
905 | } | |
906 | ||
907 | cup->cu_msg.cm_cmd = Cld_GraceDone; | |
908 | cup->cu_msg.cm_u.cm_gracetime = (int64_t)boot_time; | |
909 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
910 | if (!ret) | |
911 | ret = cup->cu_msg.cm_status; | |
912 | ||
913 | free_cld_upcall(cup); | |
914 | out_err: | |
915 | if (ret) | |
916 | printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret); | |
917 | } | |
918 | ||
919 | static struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = { | |
920 | .init = nfsd4_init_cld_pipe, | |
921 | .exit = nfsd4_remove_cld_pipe, | |
922 | .create = nfsd4_cld_create, | |
923 | .remove = nfsd4_cld_remove, | |
924 | .check = nfsd4_cld_check, | |
925 | .grace_done = nfsd4_cld_grace_done, | |
926 | }; | |
927 | ||
2a4317c5 JL |
928 | int |
929 | nfsd4_client_tracking_init(struct net *net) | |
930 | { | |
931 | int status; | |
f3f80148 | 932 | struct path path; |
2a4317c5 | 933 | |
f3f80148 JL |
934 | if (!client_tracking_ops) { |
935 | client_tracking_ops = &nfsd4_cld_tracking_ops; | |
936 | status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path); | |
937 | if (!status) { | |
938 | if (S_ISDIR(path.dentry->d_inode->i_mode)) | |
939 | client_tracking_ops = | |
940 | &nfsd4_legacy_tracking_ops; | |
941 | path_put(&path); | |
942 | } | |
943 | } | |
2a4317c5 JL |
944 | |
945 | status = client_tracking_ops->init(net); | |
946 | if (status) { | |
947 | printk(KERN_WARNING "NFSD: Unable to initialize client " | |
948 | "recovery tracking! (%d)\n", status); | |
949 | client_tracking_ops = NULL; | |
950 | } | |
951 | return status; | |
952 | } | |
953 | ||
954 | void | |
955 | nfsd4_client_tracking_exit(struct net *net) | |
956 | { | |
957 | if (client_tracking_ops) { | |
958 | client_tracking_ops->exit(net); | |
959 | client_tracking_ops = NULL; | |
960 | } | |
961 | } | |
962 | ||
963 | void | |
964 | nfsd4_client_record_create(struct nfs4_client *clp) | |
965 | { | |
966 | if (client_tracking_ops) | |
967 | client_tracking_ops->create(clp); | |
968 | } | |
969 | ||
970 | void | |
971 | nfsd4_client_record_remove(struct nfs4_client *clp) | |
972 | { | |
973 | if (client_tracking_ops) | |
974 | client_tracking_ops->remove(clp); | |
975 | } | |
976 | ||
977 | int | |
978 | nfsd4_client_record_check(struct nfs4_client *clp) | |
979 | { | |
980 | if (client_tracking_ops) | |
981 | return client_tracking_ops->check(clp); | |
982 | ||
983 | return -EOPNOTSUPP; | |
984 | } | |
985 | ||
986 | void | |
987 | nfsd4_record_grace_done(struct net *net, time_t boot_time) | |
988 | { | |
989 | if (client_tracking_ops) | |
990 | client_tracking_ops->grace_done(net, boot_time); | |
991 | } | |
813fd320 JL |
992 | |
993 | static int | |
994 | rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr) | |
995 | { | |
996 | struct super_block *sb = ptr; | |
997 | struct net *net = sb->s_fs_info; | |
998 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
999 | struct cld_net *cn = nn->cld_net; | |
1000 | struct dentry *dentry; | |
1001 | int ret = 0; | |
1002 | ||
1003 | if (!try_module_get(THIS_MODULE)) | |
1004 | return 0; | |
1005 | ||
1006 | if (!cn) { | |
1007 | module_put(THIS_MODULE); | |
1008 | return 0; | |
1009 | } | |
1010 | ||
1011 | switch (event) { | |
1012 | case RPC_PIPEFS_MOUNT: | |
1013 | dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe); | |
1014 | if (IS_ERR(dentry)) { | |
1015 | ret = PTR_ERR(dentry); | |
1016 | break; | |
1017 | } | |
1018 | cn->cn_pipe->dentry = dentry; | |
1019 | break; | |
1020 | case RPC_PIPEFS_UMOUNT: | |
1021 | if (cn->cn_pipe->dentry) | |
1022 | nfsd4_cld_unregister_sb(cn->cn_pipe); | |
1023 | break; | |
1024 | default: | |
1025 | ret = -ENOTSUPP; | |
1026 | break; | |
1027 | } | |
1028 | module_put(THIS_MODULE); | |
1029 | return ret; | |
1030 | } | |
1031 | ||
1032 | struct notifier_block nfsd4_cld_block = { | |
1033 | .notifier_call = rpc_pipefs_event, | |
1034 | }; | |
797a9d79 JL |
1035 | |
1036 | int | |
1037 | register_cld_notifier(void) | |
1038 | { | |
1039 | return rpc_pipefs_notifier_register(&nfsd4_cld_block); | |
1040 | } | |
1041 | ||
1042 | void | |
1043 | unregister_cld_notifier(void) | |
1044 | { | |
1045 | rpc_pipefs_notifier_unregister(&nfsd4_cld_block); | |
1046 | } |