]>
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 | |
4a55c101 JK |
157 | status = mnt_want_write_file(rec_file); |
158 | if (status) | |
159 | return; | |
160 | ||
e970a573 | 161 | dir = rec_file->f_path.dentry; |
c7b9a459 | 162 | /* lock the parent */ |
e970a573 | 163 | mutex_lock(&dir->d_inode->i_mutex); |
c7b9a459 | 164 | |
e970a573 | 165 | dentry = lookup_one_len(dname, dir, HEXDIR_LEN-1); |
c7b9a459 N |
166 | if (IS_ERR(dentry)) { |
167 | status = PTR_ERR(dentry); | |
168 | goto out_unlock; | |
169 | } | |
6577aac0 | 170 | if (dentry->d_inode) |
aec39680 BF |
171 | /* |
172 | * In the 4.1 case, where we're called from | |
173 | * reclaim_complete(), records from the previous reboot | |
174 | * may still be left, so this is OK. | |
175 | * | |
176 | * In the 4.0 case, we should never get here; but we may | |
177 | * as well be forgiving and just succeed silently. | |
178 | */ | |
c7b9a459 | 179 | goto out_put; |
e970a573 | 180 | status = vfs_mkdir(dir->d_inode, dentry, S_IRWXU); |
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); | |
4a55c101 | 192 | mnt_drop_write_file(rec_file); |
d84f4f99 | 193 | nfs4_reset_creds(original_cred); |
c7b9a459 N |
194 | } |
195 | ||
190e4fbf N |
196 | typedef int (recdir_func)(struct dentry *, struct dentry *); |
197 | ||
05f4f678 BF |
198 | struct name_list { |
199 | char name[HEXDIR_LEN]; | |
190e4fbf N |
200 | struct list_head list; |
201 | }; | |
202 | ||
190e4fbf | 203 | static int |
05f4f678 | 204 | nfsd4_build_namelist(void *arg, const char *name, int namlen, |
afefdbb2 | 205 | loff_t offset, u64 ino, unsigned int d_type) |
190e4fbf | 206 | { |
05f4f678 BF |
207 | struct list_head *names = arg; |
208 | struct name_list *entry; | |
190e4fbf | 209 | |
05f4f678 | 210 | if (namlen != HEXDIR_LEN - 1) |
b37ad28b | 211 | return 0; |
05f4f678 BF |
212 | entry = kmalloc(sizeof(struct name_list), GFP_KERNEL); |
213 | if (entry == NULL) | |
190e4fbf | 214 | return -ENOMEM; |
05f4f678 BF |
215 | memcpy(entry->name, name, HEXDIR_LEN - 1); |
216 | entry->name[HEXDIR_LEN - 1] = '\0'; | |
217 | list_add(&entry->list, names); | |
190e4fbf N |
218 | return 0; |
219 | } | |
220 | ||
221 | static int | |
5b4b299c | 222 | nfsd4_list_rec_dir(recdir_func *f) |
190e4fbf | 223 | { |
d84f4f99 | 224 | const struct cred *original_cred; |
5b4b299c | 225 | struct dentry *dir = rec_file->f_path.dentry; |
05f4f678 | 226 | LIST_HEAD(names); |
190e4fbf N |
227 | int status; |
228 | ||
d84f4f99 DH |
229 | status = nfs4_save_creds(&original_cred); |
230 | if (status < 0) | |
231 | return status; | |
190e4fbf | 232 | |
5b4b299c AV |
233 | status = vfs_llseek(rec_file, 0, SEEK_SET); |
234 | if (status < 0) { | |
235 | nfs4_reset_creds(original_cred); | |
236 | return status; | |
237 | } | |
238 | ||
239 | status = vfs_readdir(rec_file, nfsd4_build_namelist, &names); | |
8daed1e5 | 240 | mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT); |
05f4f678 | 241 | while (!list_empty(&names)) { |
5b4b299c | 242 | struct name_list *entry; |
05f4f678 | 243 | entry = list_entry(names.next, struct name_list, list); |
5b4b299c AV |
244 | if (!status) { |
245 | struct dentry *dentry; | |
246 | dentry = lookup_one_len(entry->name, dir, HEXDIR_LEN-1); | |
247 | if (IS_ERR(dentry)) { | |
248 | status = PTR_ERR(dentry); | |
249 | break; | |
250 | } | |
251 | status = f(dir, dentry); | |
252 | dput(dentry); | |
05f4f678 | 253 | } |
05f4f678 BF |
254 | list_del(&entry->list); |
255 | kfree(entry); | |
190e4fbf | 256 | } |
2f9092e1 | 257 | mutex_unlock(&dir->d_inode->i_mutex); |
d84f4f99 | 258 | nfs4_reset_creds(original_cred); |
190e4fbf N |
259 | return status; |
260 | } | |
261 | ||
c7b9a459 N |
262 | static int |
263 | nfsd4_unlink_clid_dir(char *name, int namlen) | |
264 | { | |
e970a573 | 265 | struct dentry *dir, *dentry; |
c7b9a459 N |
266 | int status; |
267 | ||
268 | dprintk("NFSD: nfsd4_unlink_clid_dir. name %.*s\n", namlen, name); | |
269 | ||
e970a573 CH |
270 | dir = rec_file->f_path.dentry; |
271 | mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT); | |
272 | dentry = lookup_one_len(name, dir, namlen); | |
c7b9a459 N |
273 | if (IS_ERR(dentry)) { |
274 | status = PTR_ERR(dentry); | |
2f9092e1 | 275 | goto out_unlock; |
c7b9a459 N |
276 | } |
277 | status = -ENOENT; | |
278 | if (!dentry->d_inode) | |
279 | goto out; | |
e970a573 | 280 | status = vfs_rmdir(dir->d_inode, dentry); |
c7b9a459 N |
281 | out: |
282 | dput(dentry); | |
2f9092e1 | 283 | out_unlock: |
e970a573 | 284 | mutex_unlock(&dir->d_inode->i_mutex); |
c7b9a459 N |
285 | return status; |
286 | } | |
287 | ||
2a4317c5 | 288 | static void |
c7b9a459 N |
289 | nfsd4_remove_clid_dir(struct nfs4_client *clp) |
290 | { | |
d84f4f99 | 291 | const struct cred *original_cred; |
c7b9a459 N |
292 | int status; |
293 | ||
a52d726b | 294 | if (!rec_file || !test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) |
c7b9a459 N |
295 | return; |
296 | ||
a561be71 | 297 | status = mnt_want_write_file(rec_file); |
0622753b DH |
298 | if (status) |
299 | goto out; | |
a52d726b | 300 | clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); |
d84f4f99 DH |
301 | |
302 | status = nfs4_save_creds(&original_cred); | |
303 | if (status < 0) | |
304 | goto out; | |
305 | ||
c7b9a459 | 306 | status = nfsd4_unlink_clid_dir(clp->cl_recdir, HEXDIR_LEN-1); |
d84f4f99 | 307 | nfs4_reset_creds(original_cred); |
c7b9a459 | 308 | if (status == 0) |
8018ab05 | 309 | vfs_fsync(rec_file, 0); |
2a79f17e | 310 | mnt_drop_write_file(rec_file); |
0622753b | 311 | out: |
c7b9a459 N |
312 | if (status) |
313 | printk("NFSD: Failed to remove expired client state directory" | |
314 | " %.*s\n", HEXDIR_LEN, clp->cl_recdir); | |
c7b9a459 N |
315 | } |
316 | ||
317 | static int | |
318 | purge_old(struct dentry *parent, struct dentry *child) | |
319 | { | |
320 | int status; | |
321 | ||
a1bcecd2 | 322 | if (nfs4_has_reclaimed_state(child->d_name.name, false)) |
b37ad28b | 323 | return 0; |
c7b9a459 | 324 | |
2f9092e1 | 325 | status = vfs_rmdir(parent->d_inode, child); |
c7b9a459 N |
326 | if (status) |
327 | printk("failed to remove client recovery directory %s\n", | |
328 | child->d_name.name); | |
329 | /* Keep trying, success or failure: */ | |
b37ad28b | 330 | return 0; |
c7b9a459 N |
331 | } |
332 | ||
2a4317c5 JL |
333 | static void |
334 | nfsd4_recdir_purge_old(struct net *net, time_t boot_time) | |
335 | { | |
c7b9a459 N |
336 | int status; |
337 | ||
e970a573 | 338 | if (!rec_file) |
c7b9a459 | 339 | return; |
a561be71 | 340 | status = mnt_want_write_file(rec_file); |
0622753b DH |
341 | if (status) |
342 | goto out; | |
5b4b299c | 343 | status = nfsd4_list_rec_dir(purge_old); |
c7b9a459 | 344 | if (status == 0) |
8018ab05 | 345 | vfs_fsync(rec_file, 0); |
2a79f17e | 346 | mnt_drop_write_file(rec_file); |
0622753b | 347 | out: |
c7b9a459 N |
348 | if (status) |
349 | printk("nfsd4: failed to purge old clients from recovery" | |
e970a573 | 350 | " directory %s\n", rec_file->f_path.dentry->d_name.name); |
c7b9a459 N |
351 | } |
352 | ||
190e4fbf N |
353 | static int |
354 | load_recdir(struct dentry *parent, struct dentry *child) | |
355 | { | |
356 | if (child->d_name.len != HEXDIR_LEN - 1) { | |
357 | printk("nfsd4: illegal name %s in recovery directory\n", | |
358 | child->d_name.name); | |
359 | /* Keep trying; maybe the others are OK: */ | |
b37ad28b | 360 | return 0; |
190e4fbf N |
361 | } |
362 | nfs4_client_to_reclaim(child->d_name.name); | |
b37ad28b | 363 | return 0; |
190e4fbf N |
364 | } |
365 | ||
2a4317c5 | 366 | static int |
190e4fbf N |
367 | nfsd4_recdir_load(void) { |
368 | int status; | |
369 | ||
e970a573 CH |
370 | if (!rec_file) |
371 | return 0; | |
372 | ||
5b4b299c | 373 | status = nfsd4_list_rec_dir(load_recdir); |
190e4fbf N |
374 | if (status) |
375 | printk("nfsd4: failed loading clients from recovery" | |
e970a573 | 376 | " directory %s\n", rec_file->f_path.dentry->d_name.name); |
190e4fbf N |
377 | return status; |
378 | } | |
379 | ||
380 | /* | |
381 | * Hold reference to the recovery directory. | |
382 | */ | |
383 | ||
2a4317c5 JL |
384 | static int |
385 | nfsd4_init_recdir(void) | |
190e4fbf | 386 | { |
d84f4f99 DH |
387 | const struct cred *original_cred; |
388 | int status; | |
190e4fbf N |
389 | |
390 | printk("NFSD: Using %s as the NFSv4 state recovery directory\n", | |
48483bf2 | 391 | user_recovery_dirname); |
190e4fbf | 392 | |
e970a573 | 393 | BUG_ON(rec_file); |
190e4fbf | 394 | |
d84f4f99 DH |
395 | status = nfs4_save_creds(&original_cred); |
396 | if (status < 0) { | |
397 | printk("NFSD: Unable to change credentials to find recovery" | |
398 | " directory: error %d\n", | |
399 | status); | |
2a4317c5 | 400 | return status; |
d84f4f99 | 401 | } |
190e4fbf | 402 | |
48483bf2 | 403 | rec_file = filp_open(user_recovery_dirname, O_RDONLY | O_DIRECTORY, 0); |
e970a573 | 404 | if (IS_ERR(rec_file)) { |
c2642ab0 | 405 | printk("NFSD: unable to find recovery directory %s\n", |
48483bf2 | 406 | user_recovery_dirname); |
2a4317c5 | 407 | status = PTR_ERR(rec_file); |
e970a573 CH |
408 | rec_file = NULL; |
409 | } | |
190e4fbf | 410 | |
d84f4f99 | 411 | nfs4_reset_creds(original_cred); |
2a4317c5 | 412 | return status; |
190e4fbf N |
413 | } |
414 | ||
2a4317c5 JL |
415 | static int |
416 | nfsd4_load_reboot_recovery_data(struct net *net) | |
417 | { | |
418 | int status; | |
419 | ||
cc27e0d4 JL |
420 | /* XXX: The legacy code won't work in a container */ |
421 | if (net != &init_net) { | |
422 | WARN(1, KERN_ERR "NFSD: attempt to initialize legacy client " | |
423 | "tracking in a container!\n"); | |
424 | return -EINVAL; | |
425 | } | |
426 | ||
2a4317c5 JL |
427 | nfs4_lock_state(); |
428 | status = nfsd4_init_recdir(); | |
429 | if (!status) | |
430 | status = nfsd4_recdir_load(); | |
431 | nfs4_unlock_state(); | |
432 | if (status) | |
433 | printk(KERN_ERR "NFSD: Failure reading reboot recovery data\n"); | |
434 | return status; | |
435 | } | |
436 | ||
437 | static void | |
190e4fbf N |
438 | nfsd4_shutdown_recdir(void) |
439 | { | |
e970a573 | 440 | if (!rec_file) |
190e4fbf | 441 | return; |
e970a573 CH |
442 | fput(rec_file); |
443 | rec_file = NULL; | |
190e4fbf | 444 | } |
48483bf2 | 445 | |
2a4317c5 JL |
446 | static void |
447 | nfsd4_legacy_tracking_exit(struct net *net) | |
448 | { | |
449 | nfs4_release_reclaim(); | |
450 | nfsd4_shutdown_recdir(); | |
451 | } | |
452 | ||
48483bf2 BF |
453 | /* |
454 | * Change the NFSv4 recovery directory to recdir. | |
455 | */ | |
456 | int | |
457 | nfs4_reset_recoverydir(char *recdir) | |
458 | { | |
459 | int status; | |
460 | struct path path; | |
461 | ||
462 | status = kern_path(recdir, LOOKUP_FOLLOW, &path); | |
463 | if (status) | |
464 | return status; | |
465 | status = -ENOTDIR; | |
466 | if (S_ISDIR(path.dentry->d_inode->i_mode)) { | |
467 | strcpy(user_recovery_dirname, recdir); | |
468 | status = 0; | |
469 | } | |
470 | path_put(&path); | |
471 | return status; | |
472 | } | |
473 | ||
474 | char * | |
475 | nfs4_recoverydir(void) | |
476 | { | |
477 | return user_recovery_dirname; | |
478 | } | |
2a4317c5 JL |
479 | |
480 | static int | |
481 | nfsd4_check_legacy_client(struct nfs4_client *clp) | |
482 | { | |
483 | /* did we already find that this client is stable? */ | |
484 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
485 | return 0; | |
486 | ||
487 | /* look for it in the reclaim hashtable otherwise */ | |
488 | if (nfsd4_find_reclaim_client(clp)) { | |
489 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
490 | return 0; | |
491 | } | |
492 | ||
493 | return -ENOENT; | |
494 | } | |
495 | ||
496 | static struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = { | |
497 | .init = nfsd4_load_reboot_recovery_data, | |
498 | .exit = nfsd4_legacy_tracking_exit, | |
499 | .create = nfsd4_create_clid_dir, | |
500 | .remove = nfsd4_remove_clid_dir, | |
501 | .check = nfsd4_check_legacy_client, | |
502 | .grace_done = nfsd4_recdir_purge_old, | |
503 | }; | |
504 | ||
f3f80148 JL |
505 | /* Globals */ |
506 | #define NFSD_PIPE_DIR "nfsd" | |
507 | #define NFSD_CLD_PIPE "cld" | |
508 | ||
509 | /* per-net-ns structure for holding cld upcall info */ | |
510 | struct cld_net { | |
511 | struct rpc_pipe *cn_pipe; | |
512 | spinlock_t cn_lock; | |
513 | struct list_head cn_list; | |
514 | unsigned int cn_xid; | |
515 | }; | |
516 | ||
517 | struct cld_upcall { | |
518 | struct list_head cu_list; | |
519 | struct cld_net *cu_net; | |
520 | struct task_struct *cu_task; | |
521 | struct cld_msg cu_msg; | |
522 | }; | |
523 | ||
524 | static int | |
525 | __cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
526 | { | |
527 | int ret; | |
528 | struct rpc_pipe_msg msg; | |
529 | ||
530 | memset(&msg, 0, sizeof(msg)); | |
531 | msg.data = cmsg; | |
532 | msg.len = sizeof(*cmsg); | |
533 | ||
534 | /* | |
535 | * Set task state before we queue the upcall. That prevents | |
536 | * wake_up_process in the downcall from racing with schedule. | |
537 | */ | |
538 | set_current_state(TASK_UNINTERRUPTIBLE); | |
539 | ret = rpc_queue_upcall(pipe, &msg); | |
540 | if (ret < 0) { | |
541 | set_current_state(TASK_RUNNING); | |
542 | goto out; | |
543 | } | |
544 | ||
545 | schedule(); | |
546 | set_current_state(TASK_RUNNING); | |
547 | ||
548 | if (msg.errno < 0) | |
549 | ret = msg.errno; | |
550 | out: | |
551 | return ret; | |
552 | } | |
553 | ||
554 | static int | |
555 | cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) | |
556 | { | |
557 | int ret; | |
558 | ||
559 | /* | |
560 | * -EAGAIN occurs when pipe is closed and reopened while there are | |
561 | * upcalls queued. | |
562 | */ | |
563 | do { | |
564 | ret = __cld_pipe_upcall(pipe, cmsg); | |
565 | } while (ret == -EAGAIN); | |
566 | ||
567 | return ret; | |
568 | } | |
569 | ||
570 | static ssize_t | |
571 | cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) | |
572 | { | |
573 | struct cld_upcall *tmp, *cup; | |
bc1b542b | 574 | struct cld_msg __user *cmsg = (struct cld_msg __user *)src; |
f3f80148 JL |
575 | uint32_t xid; |
576 | struct nfsd_net *nn = net_generic(filp->f_dentry->d_sb->s_fs_info, | |
577 | nfsd_net_id); | |
578 | struct cld_net *cn = nn->cld_net; | |
579 | ||
580 | if (mlen != sizeof(*cmsg)) { | |
8a7dc4b0 | 581 | dprintk("%s: got %zu bytes, expected %zu\n", __func__, mlen, |
f3f80148 JL |
582 | sizeof(*cmsg)); |
583 | return -EINVAL; | |
584 | } | |
585 | ||
586 | /* copy just the xid so we can try to find that */ | |
587 | if (copy_from_user(&xid, &cmsg->cm_xid, sizeof(xid)) != 0) { | |
588 | dprintk("%s: error when copying xid from userspace", __func__); | |
589 | return -EFAULT; | |
590 | } | |
591 | ||
592 | /* walk the list and find corresponding xid */ | |
593 | cup = NULL; | |
594 | spin_lock(&cn->cn_lock); | |
595 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
596 | if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) { | |
597 | cup = tmp; | |
598 | list_del_init(&cup->cu_list); | |
599 | break; | |
600 | } | |
601 | } | |
602 | spin_unlock(&cn->cn_lock); | |
603 | ||
604 | /* couldn't find upcall? */ | |
605 | if (!cup) { | |
21f72c9f | 606 | dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, xid); |
f3f80148 JL |
607 | return -EINVAL; |
608 | } | |
609 | ||
610 | if (copy_from_user(&cup->cu_msg, src, mlen) != 0) | |
611 | return -EFAULT; | |
612 | ||
613 | wake_up_process(cup->cu_task); | |
614 | return mlen; | |
615 | } | |
616 | ||
617 | static void | |
618 | cld_pipe_destroy_msg(struct rpc_pipe_msg *msg) | |
619 | { | |
620 | struct cld_msg *cmsg = msg->data; | |
621 | struct cld_upcall *cup = container_of(cmsg, struct cld_upcall, | |
622 | cu_msg); | |
623 | ||
624 | /* errno >= 0 means we got a downcall */ | |
625 | if (msg->errno >= 0) | |
626 | return; | |
627 | ||
628 | wake_up_process(cup->cu_task); | |
629 | } | |
630 | ||
631 | static const struct rpc_pipe_ops cld_upcall_ops = { | |
632 | .upcall = rpc_pipe_generic_upcall, | |
633 | .downcall = cld_pipe_downcall, | |
634 | .destroy_msg = cld_pipe_destroy_msg, | |
635 | }; | |
636 | ||
637 | static struct dentry * | |
638 | nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe) | |
639 | { | |
640 | struct dentry *dir, *dentry; | |
641 | ||
642 | dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR); | |
643 | if (dir == NULL) | |
644 | return ERR_PTR(-ENOENT); | |
645 | dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe); | |
646 | dput(dir); | |
647 | return dentry; | |
648 | } | |
649 | ||
650 | static void | |
651 | nfsd4_cld_unregister_sb(struct rpc_pipe *pipe) | |
652 | { | |
653 | if (pipe->dentry) | |
654 | rpc_unlink(pipe->dentry); | |
655 | } | |
656 | ||
657 | static struct dentry * | |
658 | nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe) | |
659 | { | |
660 | struct super_block *sb; | |
661 | struct dentry *dentry; | |
662 | ||
663 | sb = rpc_get_sb_net(net); | |
664 | if (!sb) | |
665 | return NULL; | |
666 | dentry = nfsd4_cld_register_sb(sb, pipe); | |
667 | rpc_put_sb_net(net); | |
668 | return dentry; | |
669 | } | |
670 | ||
671 | static void | |
672 | nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe) | |
673 | { | |
674 | struct super_block *sb; | |
675 | ||
676 | sb = rpc_get_sb_net(net); | |
677 | if (sb) { | |
678 | nfsd4_cld_unregister_sb(pipe); | |
679 | rpc_put_sb_net(net); | |
680 | } | |
681 | } | |
682 | ||
683 | /* Initialize rpc_pipefs pipe for communication with client tracking daemon */ | |
684 | static int | |
685 | nfsd4_init_cld_pipe(struct net *net) | |
686 | { | |
687 | int ret; | |
688 | struct dentry *dentry; | |
689 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
690 | struct cld_net *cn; | |
691 | ||
692 | if (nn->cld_net) | |
693 | return 0; | |
694 | ||
695 | cn = kzalloc(sizeof(*cn), GFP_KERNEL); | |
696 | if (!cn) { | |
697 | ret = -ENOMEM; | |
698 | goto err; | |
699 | } | |
700 | ||
701 | cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); | |
702 | if (IS_ERR(cn->cn_pipe)) { | |
703 | ret = PTR_ERR(cn->cn_pipe); | |
704 | goto err; | |
705 | } | |
706 | spin_lock_init(&cn->cn_lock); | |
707 | INIT_LIST_HEAD(&cn->cn_list); | |
708 | ||
709 | dentry = nfsd4_cld_register_net(net, cn->cn_pipe); | |
710 | if (IS_ERR(dentry)) { | |
711 | ret = PTR_ERR(dentry); | |
712 | goto err_destroy_data; | |
713 | } | |
714 | ||
715 | cn->cn_pipe->dentry = dentry; | |
716 | nn->cld_net = cn; | |
717 | return 0; | |
718 | ||
719 | err_destroy_data: | |
720 | rpc_destroy_pipe_data(cn->cn_pipe); | |
721 | err: | |
722 | kfree(cn); | |
723 | printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n", | |
724 | ret); | |
725 | return ret; | |
726 | } | |
727 | ||
728 | static void | |
729 | nfsd4_remove_cld_pipe(struct net *net) | |
730 | { | |
731 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
732 | struct cld_net *cn = nn->cld_net; | |
733 | ||
734 | nfsd4_cld_unregister_net(net, cn->cn_pipe); | |
735 | rpc_destroy_pipe_data(cn->cn_pipe); | |
736 | kfree(nn->cld_net); | |
737 | nn->cld_net = NULL; | |
738 | } | |
739 | ||
740 | static struct cld_upcall * | |
741 | alloc_cld_upcall(struct cld_net *cn) | |
742 | { | |
743 | struct cld_upcall *new, *tmp; | |
744 | ||
745 | new = kzalloc(sizeof(*new), GFP_KERNEL); | |
746 | if (!new) | |
747 | return new; | |
748 | ||
749 | /* FIXME: hard cap on number in flight? */ | |
750 | restart_search: | |
751 | spin_lock(&cn->cn_lock); | |
752 | list_for_each_entry(tmp, &cn->cn_list, cu_list) { | |
753 | if (tmp->cu_msg.cm_xid == cn->cn_xid) { | |
754 | cn->cn_xid++; | |
755 | spin_unlock(&cn->cn_lock); | |
756 | goto restart_search; | |
757 | } | |
758 | } | |
759 | new->cu_task = current; | |
760 | new->cu_msg.cm_vers = CLD_UPCALL_VERSION; | |
761 | put_unaligned(cn->cn_xid++, &new->cu_msg.cm_xid); | |
762 | new->cu_net = cn; | |
763 | list_add(&new->cu_list, &cn->cn_list); | |
764 | spin_unlock(&cn->cn_lock); | |
765 | ||
766 | dprintk("%s: allocated xid %u\n", __func__, new->cu_msg.cm_xid); | |
767 | ||
768 | return new; | |
769 | } | |
770 | ||
771 | static void | |
772 | free_cld_upcall(struct cld_upcall *victim) | |
773 | { | |
774 | struct cld_net *cn = victim->cu_net; | |
775 | ||
776 | spin_lock(&cn->cn_lock); | |
777 | list_del(&victim->cu_list); | |
778 | spin_unlock(&cn->cn_lock); | |
779 | kfree(victim); | |
780 | } | |
781 | ||
782 | /* Ask daemon to create a new record */ | |
783 | static void | |
784 | nfsd4_cld_create(struct nfs4_client *clp) | |
785 | { | |
786 | int ret; | |
787 | struct cld_upcall *cup; | |
788 | /* FIXME: determine net from clp */ | |
789 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
790 | struct cld_net *cn = nn->cld_net; | |
791 | ||
792 | /* Don't upcall if it's already stored */ | |
793 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
794 | return; | |
795 | ||
796 | cup = alloc_cld_upcall(cn); | |
797 | if (!cup) { | |
798 | ret = -ENOMEM; | |
799 | goto out_err; | |
800 | } | |
801 | ||
802 | cup->cu_msg.cm_cmd = Cld_Create; | |
803 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
804 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
805 | clp->cl_name.len); | |
806 | ||
807 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
808 | if (!ret) { | |
809 | ret = cup->cu_msg.cm_status; | |
810 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
811 | } | |
812 | ||
813 | free_cld_upcall(cup); | |
814 | out_err: | |
815 | if (ret) | |
816 | printk(KERN_ERR "NFSD: Unable to create client " | |
817 | "record on stable storage: %d\n", ret); | |
818 | } | |
819 | ||
820 | /* Ask daemon to create a new record */ | |
821 | static void | |
822 | nfsd4_cld_remove(struct nfs4_client *clp) | |
823 | { | |
824 | int ret; | |
825 | struct cld_upcall *cup; | |
826 | /* FIXME: determine net from clp */ | |
827 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
828 | struct cld_net *cn = nn->cld_net; | |
829 | ||
830 | /* Don't upcall if it's already removed */ | |
831 | if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
832 | return; | |
833 | ||
834 | cup = alloc_cld_upcall(cn); | |
835 | if (!cup) { | |
836 | ret = -ENOMEM; | |
837 | goto out_err; | |
838 | } | |
839 | ||
840 | cup->cu_msg.cm_cmd = Cld_Remove; | |
841 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
842 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
843 | clp->cl_name.len); | |
844 | ||
845 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
846 | if (!ret) { | |
847 | ret = cup->cu_msg.cm_status; | |
848 | clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
849 | } | |
850 | ||
851 | free_cld_upcall(cup); | |
852 | out_err: | |
853 | if (ret) | |
854 | printk(KERN_ERR "NFSD: Unable to remove client " | |
855 | "record from stable storage: %d\n", ret); | |
856 | } | |
857 | ||
858 | /* Check for presence of a record, and update its timestamp */ | |
859 | static int | |
860 | nfsd4_cld_check(struct nfs4_client *clp) | |
861 | { | |
862 | int ret; | |
863 | struct cld_upcall *cup; | |
864 | /* FIXME: determine net from clp */ | |
865 | struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id); | |
866 | struct cld_net *cn = nn->cld_net; | |
867 | ||
868 | /* Don't upcall if one was already stored during this grace pd */ | |
869 | if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) | |
870 | return 0; | |
871 | ||
872 | cup = alloc_cld_upcall(cn); | |
873 | if (!cup) { | |
874 | printk(KERN_ERR "NFSD: Unable to check client record on " | |
875 | "stable storage: %d\n", -ENOMEM); | |
876 | return -ENOMEM; | |
877 | } | |
878 | ||
879 | cup->cu_msg.cm_cmd = Cld_Check; | |
880 | cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len; | |
881 | memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data, | |
882 | clp->cl_name.len); | |
883 | ||
884 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
885 | if (!ret) { | |
886 | ret = cup->cu_msg.cm_status; | |
887 | set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); | |
888 | } | |
889 | ||
890 | free_cld_upcall(cup); | |
891 | return ret; | |
892 | } | |
893 | ||
894 | static void | |
895 | nfsd4_cld_grace_done(struct net *net, time_t boot_time) | |
896 | { | |
897 | int ret; | |
898 | struct cld_upcall *cup; | |
899 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
900 | struct cld_net *cn = nn->cld_net; | |
901 | ||
902 | cup = alloc_cld_upcall(cn); | |
903 | if (!cup) { | |
904 | ret = -ENOMEM; | |
905 | goto out_err; | |
906 | } | |
907 | ||
908 | cup->cu_msg.cm_cmd = Cld_GraceDone; | |
909 | cup->cu_msg.cm_u.cm_gracetime = (int64_t)boot_time; | |
910 | ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); | |
911 | if (!ret) | |
912 | ret = cup->cu_msg.cm_status; | |
913 | ||
914 | free_cld_upcall(cup); | |
915 | out_err: | |
916 | if (ret) | |
917 | printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret); | |
918 | } | |
919 | ||
920 | static struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = { | |
921 | .init = nfsd4_init_cld_pipe, | |
922 | .exit = nfsd4_remove_cld_pipe, | |
923 | .create = nfsd4_cld_create, | |
924 | .remove = nfsd4_cld_remove, | |
925 | .check = nfsd4_cld_check, | |
926 | .grace_done = nfsd4_cld_grace_done, | |
927 | }; | |
928 | ||
2a4317c5 JL |
929 | int |
930 | nfsd4_client_tracking_init(struct net *net) | |
931 | { | |
932 | int status; | |
f3f80148 | 933 | struct path path; |
2a4317c5 | 934 | |
f3f80148 JL |
935 | if (!client_tracking_ops) { |
936 | client_tracking_ops = &nfsd4_cld_tracking_ops; | |
937 | status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path); | |
938 | if (!status) { | |
939 | if (S_ISDIR(path.dentry->d_inode->i_mode)) | |
940 | client_tracking_ops = | |
941 | &nfsd4_legacy_tracking_ops; | |
942 | path_put(&path); | |
943 | } | |
944 | } | |
2a4317c5 JL |
945 | |
946 | status = client_tracking_ops->init(net); | |
947 | if (status) { | |
948 | printk(KERN_WARNING "NFSD: Unable to initialize client " | |
949 | "recovery tracking! (%d)\n", status); | |
950 | client_tracking_ops = NULL; | |
951 | } | |
952 | return status; | |
953 | } | |
954 | ||
955 | void | |
956 | nfsd4_client_tracking_exit(struct net *net) | |
957 | { | |
958 | if (client_tracking_ops) { | |
959 | client_tracking_ops->exit(net); | |
960 | client_tracking_ops = NULL; | |
961 | } | |
962 | } | |
963 | ||
964 | void | |
965 | nfsd4_client_record_create(struct nfs4_client *clp) | |
966 | { | |
967 | if (client_tracking_ops) | |
968 | client_tracking_ops->create(clp); | |
969 | } | |
970 | ||
971 | void | |
972 | nfsd4_client_record_remove(struct nfs4_client *clp) | |
973 | { | |
974 | if (client_tracking_ops) | |
975 | client_tracking_ops->remove(clp); | |
976 | } | |
977 | ||
978 | int | |
979 | nfsd4_client_record_check(struct nfs4_client *clp) | |
980 | { | |
981 | if (client_tracking_ops) | |
982 | return client_tracking_ops->check(clp); | |
983 | ||
984 | return -EOPNOTSUPP; | |
985 | } | |
986 | ||
987 | void | |
988 | nfsd4_record_grace_done(struct net *net, time_t boot_time) | |
989 | { | |
990 | if (client_tracking_ops) | |
991 | client_tracking_ops->grace_done(net, boot_time); | |
992 | } | |
813fd320 JL |
993 | |
994 | static int | |
995 | rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr) | |
996 | { | |
997 | struct super_block *sb = ptr; | |
998 | struct net *net = sb->s_fs_info; | |
999 | struct nfsd_net *nn = net_generic(net, nfsd_net_id); | |
1000 | struct cld_net *cn = nn->cld_net; | |
1001 | struct dentry *dentry; | |
1002 | int ret = 0; | |
1003 | ||
1004 | if (!try_module_get(THIS_MODULE)) | |
1005 | return 0; | |
1006 | ||
1007 | if (!cn) { | |
1008 | module_put(THIS_MODULE); | |
1009 | return 0; | |
1010 | } | |
1011 | ||
1012 | switch (event) { | |
1013 | case RPC_PIPEFS_MOUNT: | |
1014 | dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe); | |
1015 | if (IS_ERR(dentry)) { | |
1016 | ret = PTR_ERR(dentry); | |
1017 | break; | |
1018 | } | |
1019 | cn->cn_pipe->dentry = dentry; | |
1020 | break; | |
1021 | case RPC_PIPEFS_UMOUNT: | |
1022 | if (cn->cn_pipe->dentry) | |
1023 | nfsd4_cld_unregister_sb(cn->cn_pipe); | |
1024 | break; | |
1025 | default: | |
1026 | ret = -ENOTSUPP; | |
1027 | break; | |
1028 | } | |
1029 | module_put(THIS_MODULE); | |
1030 | return ret; | |
1031 | } | |
1032 | ||
2355c596 | 1033 | static struct notifier_block nfsd4_cld_block = { |
813fd320 JL |
1034 | .notifier_call = rpc_pipefs_event, |
1035 | }; | |
797a9d79 JL |
1036 | |
1037 | int | |
1038 | register_cld_notifier(void) | |
1039 | { | |
1040 | return rpc_pipefs_notifier_register(&nfsd4_cld_block); | |
1041 | } | |
1042 | ||
1043 | void | |
1044 | unregister_cld_notifier(void) | |
1045 | { | |
1046 | rpc_pipefs_notifier_unregister(&nfsd4_cld_block); | |
1047 | } |