Commit | Line | Data |
---|---|---|
4bac07c9 JF |
1 | /****************************************************************************** |
2 | * xenbus_xs.c | |
3 | * | |
4 | * This is the kernel equivalent of the "xs" library. We don't need everything | |
5 | * and we use xenbus_comms for communication. | |
6 | * | |
7 | * Copyright (C) 2005 Rusty Russell, IBM Corporation | |
8 | * | |
9 | * This program is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU General Public License version 2 | |
11 | * as published by the Free Software Foundation; or, when distributed | |
12 | * separately from the Linux kernel or incorporated into other | |
13 | * software packages, subject to the following license: | |
14 | * | |
15 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
16 | * of this source file (the "Software"), to deal in the Software without | |
17 | * restriction, including without limitation the rights to use, copy, modify, | |
18 | * merge, publish, distribute, sublicense, and/or sell copies of the Software, | |
19 | * and to permit persons to whom the Software is furnished to do so, subject to | |
20 | * the following conditions: | |
21 | * | |
22 | * The above copyright notice and this permission notice shall be included in | |
23 | * all copies or substantial portions of the Software. | |
24 | * | |
25 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
26 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
27 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
28 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
29 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | |
30 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS | |
31 | * IN THE SOFTWARE. | |
32 | */ | |
33 | ||
34 | #include <linux/unistd.h> | |
35 | #include <linux/errno.h> | |
36 | #include <linux/types.h> | |
37 | #include <linux/uio.h> | |
38 | #include <linux/kernel.h> | |
39 | #include <linux/string.h> | |
40 | #include <linux/err.h> | |
41 | #include <linux/slab.h> | |
42 | #include <linux/fcntl.h> | |
43 | #include <linux/kthread.h> | |
44 | #include <linux/rwsem.h> | |
45 | #include <linux/module.h> | |
46 | #include <linux/mutex.h> | |
ecc635f9 | 47 | #include <asm/xen/hypervisor.h> |
4bac07c9 | 48 | #include <xen/xenbus.h> |
5b25d89e | 49 | #include <xen/xen.h> |
4bac07c9 JF |
50 | #include "xenbus_comms.h" |
51 | ||
52 | struct xs_stored_msg { | |
53 | struct list_head list; | |
54 | ||
55 | struct xsd_sockmsg hdr; | |
56 | ||
57 | union { | |
58 | /* Queued replies. */ | |
59 | struct { | |
60 | char *body; | |
61 | } reply; | |
62 | ||
63 | /* Queued watch events. */ | |
64 | struct { | |
65 | struct xenbus_watch *handle; | |
66 | char **vec; | |
67 | unsigned int vec_size; | |
68 | } watch; | |
69 | } u; | |
70 | }; | |
71 | ||
72 | struct xs_handle { | |
73 | /* A list of replies. Currently only one will ever be outstanding. */ | |
74 | struct list_head reply_list; | |
75 | spinlock_t reply_lock; | |
76 | wait_queue_head_t reply_waitq; | |
77 | ||
78 | /* | |
79 | * Mutex ordering: transaction_mutex -> watch_mutex -> request_mutex. | |
80 | * response_mutex is never taken simultaneously with the other three. | |
4c31a781 IC |
81 | * |
82 | * transaction_mutex must be held before incrementing | |
83 | * transaction_count. The mutex is held when a suspend is in | |
84 | * progress to prevent new transactions starting. | |
85 | * | |
86 | * When decrementing transaction_count to zero the wait queue | |
87 | * should be woken up, the suspend code waits for count to | |
88 | * reach zero. | |
4bac07c9 JF |
89 | */ |
90 | ||
91 | /* One request at a time. */ | |
92 | struct mutex request_mutex; | |
93 | ||
94 | /* Protect xenbus reader thread against save/restore. */ | |
95 | struct mutex response_mutex; | |
96 | ||
97 | /* Protect transactions against save/restore. */ | |
4c31a781 IC |
98 | struct mutex transaction_mutex; |
99 | atomic_t transaction_count; | |
100 | wait_queue_head_t transaction_wq; | |
4bac07c9 JF |
101 | |
102 | /* Protect watch (de)register against save/restore. */ | |
103 | struct rw_semaphore watch_mutex; | |
104 | }; | |
105 | ||
106 | static struct xs_handle xs_state; | |
107 | ||
108 | /* List of registered watches, and a lock to protect it. */ | |
109 | static LIST_HEAD(watches); | |
110 | static DEFINE_SPINLOCK(watches_lock); | |
111 | ||
112 | /* List of pending watch callback events, and a lock to protect it. */ | |
113 | static LIST_HEAD(watch_events); | |
114 | static DEFINE_SPINLOCK(watch_events_lock); | |
115 | ||
116 | /* | |
117 | * Details of the xenwatch callback kernel thread. The thread waits on the | |
118 | * watch_events_waitq for work to do (queued on watch_events list). When it | |
119 | * wakes up it acquires the xenwatch_mutex before reading the list and | |
120 | * carrying out work. | |
121 | */ | |
122 | static pid_t xenwatch_pid; | |
123 | static DEFINE_MUTEX(xenwatch_mutex); | |
124 | static DECLARE_WAIT_QUEUE_HEAD(watch_events_waitq); | |
125 | ||
126 | static int get_error(const char *errorstring) | |
127 | { | |
128 | unsigned int i; | |
129 | ||
130 | for (i = 0; strcmp(errorstring, xsd_errors[i].errstring) != 0; i++) { | |
131 | if (i == ARRAY_SIZE(xsd_errors) - 1) { | |
132 | printk(KERN_WARNING | |
133 | "XENBUS xen store gave: unknown error %s", | |
134 | errorstring); | |
135 | return EINVAL; | |
136 | } | |
137 | } | |
138 | return xsd_errors[i].errnum; | |
139 | } | |
140 | ||
141 | static void *read_reply(enum xsd_sockmsg_type *type, unsigned int *len) | |
142 | { | |
143 | struct xs_stored_msg *msg; | |
144 | char *body; | |
145 | ||
146 | spin_lock(&xs_state.reply_lock); | |
147 | ||
148 | while (list_empty(&xs_state.reply_list)) { | |
149 | spin_unlock(&xs_state.reply_lock); | |
150 | /* XXX FIXME: Avoid synchronous wait for response here. */ | |
151 | wait_event(xs_state.reply_waitq, | |
152 | !list_empty(&xs_state.reply_list)); | |
153 | spin_lock(&xs_state.reply_lock); | |
154 | } | |
155 | ||
156 | msg = list_entry(xs_state.reply_list.next, | |
157 | struct xs_stored_msg, list); | |
158 | list_del(&msg->list); | |
159 | ||
160 | spin_unlock(&xs_state.reply_lock); | |
161 | ||
162 | *type = msg->hdr.type; | |
163 | if (len) | |
164 | *len = msg->hdr.len; | |
165 | body = msg->u.reply.body; | |
166 | ||
167 | kfree(msg); | |
168 | ||
169 | return body; | |
170 | } | |
171 | ||
4c31a781 IC |
172 | static void transaction_start(void) |
173 | { | |
174 | mutex_lock(&xs_state.transaction_mutex); | |
175 | atomic_inc(&xs_state.transaction_count); | |
176 | mutex_unlock(&xs_state.transaction_mutex); | |
177 | } | |
178 | ||
179 | static void transaction_end(void) | |
180 | { | |
181 | if (atomic_dec_and_test(&xs_state.transaction_count)) | |
182 | wake_up(&xs_state.transaction_wq); | |
183 | } | |
184 | ||
185 | static void transaction_suspend(void) | |
186 | { | |
187 | mutex_lock(&xs_state.transaction_mutex); | |
188 | wait_event(xs_state.transaction_wq, | |
189 | atomic_read(&xs_state.transaction_count) == 0); | |
190 | } | |
191 | ||
192 | static void transaction_resume(void) | |
193 | { | |
194 | mutex_unlock(&xs_state.transaction_mutex); | |
195 | } | |
196 | ||
4bac07c9 JF |
197 | void *xenbus_dev_request_and_reply(struct xsd_sockmsg *msg) |
198 | { | |
199 | void *ret; | |
200 | struct xsd_sockmsg req_msg = *msg; | |
201 | int err; | |
202 | ||
203 | if (req_msg.type == XS_TRANSACTION_START) | |
4c31a781 | 204 | transaction_start(); |
4bac07c9 JF |
205 | |
206 | mutex_lock(&xs_state.request_mutex); | |
207 | ||
208 | err = xb_write(msg, sizeof(*msg) + msg->len); | |
209 | if (err) { | |
210 | msg->type = XS_ERROR; | |
211 | ret = ERR_PTR(err); | |
212 | } else | |
213 | ret = read_reply(&msg->type, &msg->len); | |
214 | ||
215 | mutex_unlock(&xs_state.request_mutex); | |
216 | ||
217 | if ((msg->type == XS_TRANSACTION_END) || | |
218 | ((req_msg.type == XS_TRANSACTION_START) && | |
219 | (msg->type == XS_ERROR))) | |
4c31a781 | 220 | transaction_end(); |
4bac07c9 JF |
221 | |
222 | return ret; | |
223 | } | |
1107ba88 | 224 | EXPORT_SYMBOL(xenbus_dev_request_and_reply); |
4bac07c9 JF |
225 | |
226 | /* Send message to xs, get kmalloc'ed reply. ERR_PTR() on error. */ | |
227 | static void *xs_talkv(struct xenbus_transaction t, | |
228 | enum xsd_sockmsg_type type, | |
229 | const struct kvec *iovec, | |
230 | unsigned int num_vecs, | |
231 | unsigned int *len) | |
232 | { | |
233 | struct xsd_sockmsg msg; | |
234 | void *ret = NULL; | |
235 | unsigned int i; | |
236 | int err; | |
237 | ||
238 | msg.tx_id = t.id; | |
239 | msg.req_id = 0; | |
240 | msg.type = type; | |
241 | msg.len = 0; | |
242 | for (i = 0; i < num_vecs; i++) | |
243 | msg.len += iovec[i].iov_len; | |
244 | ||
245 | mutex_lock(&xs_state.request_mutex); | |
246 | ||
247 | err = xb_write(&msg, sizeof(msg)); | |
248 | if (err) { | |
249 | mutex_unlock(&xs_state.request_mutex); | |
250 | return ERR_PTR(err); | |
251 | } | |
252 | ||
253 | for (i = 0; i < num_vecs; i++) { | |
254 | err = xb_write(iovec[i].iov_base, iovec[i].iov_len); | |
255 | if (err) { | |
256 | mutex_unlock(&xs_state.request_mutex); | |
257 | return ERR_PTR(err); | |
258 | } | |
259 | } | |
260 | ||
261 | ret = read_reply(&msg.type, len); | |
262 | ||
263 | mutex_unlock(&xs_state.request_mutex); | |
264 | ||
265 | if (IS_ERR(ret)) | |
266 | return ret; | |
267 | ||
268 | if (msg.type == XS_ERROR) { | |
269 | err = get_error(ret); | |
270 | kfree(ret); | |
271 | return ERR_PTR(-err); | |
272 | } | |
273 | ||
274 | if (msg.type != type) { | |
275 | if (printk_ratelimit()) | |
276 | printk(KERN_WARNING | |
277 | "XENBUS unexpected type [%d], expected [%d]\n", | |
278 | msg.type, type); | |
279 | kfree(ret); | |
280 | return ERR_PTR(-EINVAL); | |
281 | } | |
282 | return ret; | |
283 | } | |
284 | ||
285 | /* Simplified version of xs_talkv: single message. */ | |
286 | static void *xs_single(struct xenbus_transaction t, | |
287 | enum xsd_sockmsg_type type, | |
288 | const char *string, | |
289 | unsigned int *len) | |
290 | { | |
291 | struct kvec iovec; | |
292 | ||
293 | iovec.iov_base = (void *)string; | |
294 | iovec.iov_len = strlen(string) + 1; | |
295 | return xs_talkv(t, type, &iovec, 1, len); | |
296 | } | |
297 | ||
298 | /* Many commands only need an ack, don't care what it says. */ | |
299 | static int xs_error(char *reply) | |
300 | { | |
301 | if (IS_ERR(reply)) | |
302 | return PTR_ERR(reply); | |
303 | kfree(reply); | |
304 | return 0; | |
305 | } | |
306 | ||
307 | static unsigned int count_strings(const char *strings, unsigned int len) | |
308 | { | |
309 | unsigned int num; | |
310 | const char *p; | |
311 | ||
312 | for (p = strings, num = 0; p < strings + len; p += strlen(p) + 1) | |
313 | num++; | |
314 | ||
315 | return num; | |
316 | } | |
317 | ||
318 | /* Return the path to dir with /name appended. Buffer must be kfree()'ed. */ | |
319 | static char *join(const char *dir, const char *name) | |
320 | { | |
321 | char *buffer; | |
322 | ||
323 | if (strlen(name) == 0) | |
a144ff09 | 324 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s", dir); |
4bac07c9 | 325 | else |
a144ff09 | 326 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s/%s", dir, name); |
4bac07c9 JF |
327 | return (!buffer) ? ERR_PTR(-ENOMEM) : buffer; |
328 | } | |
329 | ||
330 | static char **split(char *strings, unsigned int len, unsigned int *num) | |
331 | { | |
332 | char *p, **ret; | |
333 | ||
334 | /* Count the strings. */ | |
335 | *num = count_strings(strings, len); | |
336 | ||
337 | /* Transfer to one big alloc for easy freeing. */ | |
a144ff09 | 338 | ret = kmalloc(*num * sizeof(char *) + len, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
339 | if (!ret) { |
340 | kfree(strings); | |
341 | return ERR_PTR(-ENOMEM); | |
342 | } | |
343 | memcpy(&ret[*num], strings, len); | |
344 | kfree(strings); | |
345 | ||
346 | strings = (char *)&ret[*num]; | |
347 | for (p = strings, *num = 0; p < strings + len; p += strlen(p) + 1) | |
348 | ret[(*num)++] = p; | |
349 | ||
350 | return ret; | |
351 | } | |
352 | ||
353 | char **xenbus_directory(struct xenbus_transaction t, | |
354 | const char *dir, const char *node, unsigned int *num) | |
355 | { | |
356 | char *strings, *path; | |
357 | unsigned int len; | |
358 | ||
359 | path = join(dir, node); | |
360 | if (IS_ERR(path)) | |
361 | return (char **)path; | |
362 | ||
363 | strings = xs_single(t, XS_DIRECTORY, path, &len); | |
364 | kfree(path); | |
365 | if (IS_ERR(strings)) | |
366 | return (char **)strings; | |
367 | ||
368 | return split(strings, len, num); | |
369 | } | |
370 | EXPORT_SYMBOL_GPL(xenbus_directory); | |
371 | ||
372 | /* Check if a path exists. Return 1 if it does. */ | |
373 | int xenbus_exists(struct xenbus_transaction t, | |
374 | const char *dir, const char *node) | |
375 | { | |
376 | char **d; | |
377 | int dir_n; | |
378 | ||
379 | d = xenbus_directory(t, dir, node, &dir_n); | |
380 | if (IS_ERR(d)) | |
381 | return 0; | |
382 | kfree(d); | |
383 | return 1; | |
384 | } | |
385 | EXPORT_SYMBOL_GPL(xenbus_exists); | |
386 | ||
387 | /* Get the value of a single file. | |
388 | * Returns a kmalloced value: call free() on it after use. | |
389 | * len indicates length in bytes. | |
390 | */ | |
391 | void *xenbus_read(struct xenbus_transaction t, | |
392 | const char *dir, const char *node, unsigned int *len) | |
393 | { | |
394 | char *path; | |
395 | void *ret; | |
396 | ||
397 | path = join(dir, node); | |
398 | if (IS_ERR(path)) | |
399 | return (void *)path; | |
400 | ||
401 | ret = xs_single(t, XS_READ, path, len); | |
402 | kfree(path); | |
403 | return ret; | |
404 | } | |
405 | EXPORT_SYMBOL_GPL(xenbus_read); | |
406 | ||
407 | /* Write the value of a single file. | |
408 | * Returns -err on failure. | |
409 | */ | |
410 | int xenbus_write(struct xenbus_transaction t, | |
411 | const char *dir, const char *node, const char *string) | |
412 | { | |
413 | const char *path; | |
414 | struct kvec iovec[2]; | |
415 | int ret; | |
416 | ||
417 | path = join(dir, node); | |
418 | if (IS_ERR(path)) | |
419 | return PTR_ERR(path); | |
420 | ||
421 | iovec[0].iov_base = (void *)path; | |
422 | iovec[0].iov_len = strlen(path) + 1; | |
423 | iovec[1].iov_base = (void *)string; | |
424 | iovec[1].iov_len = strlen(string); | |
425 | ||
426 | ret = xs_error(xs_talkv(t, XS_WRITE, iovec, ARRAY_SIZE(iovec), NULL)); | |
427 | kfree(path); | |
428 | return ret; | |
429 | } | |
430 | EXPORT_SYMBOL_GPL(xenbus_write); | |
431 | ||
432 | /* Create a new directory. */ | |
433 | int xenbus_mkdir(struct xenbus_transaction t, | |
434 | const char *dir, const char *node) | |
435 | { | |
436 | char *path; | |
437 | int ret; | |
438 | ||
439 | path = join(dir, node); | |
440 | if (IS_ERR(path)) | |
441 | return PTR_ERR(path); | |
442 | ||
443 | ret = xs_error(xs_single(t, XS_MKDIR, path, NULL)); | |
444 | kfree(path); | |
445 | return ret; | |
446 | } | |
447 | EXPORT_SYMBOL_GPL(xenbus_mkdir); | |
448 | ||
449 | /* Destroy a file or directory (directories must be empty). */ | |
450 | int xenbus_rm(struct xenbus_transaction t, const char *dir, const char *node) | |
451 | { | |
452 | char *path; | |
453 | int ret; | |
454 | ||
455 | path = join(dir, node); | |
456 | if (IS_ERR(path)) | |
457 | return PTR_ERR(path); | |
458 | ||
459 | ret = xs_error(xs_single(t, XS_RM, path, NULL)); | |
460 | kfree(path); | |
461 | return ret; | |
462 | } | |
463 | EXPORT_SYMBOL_GPL(xenbus_rm); | |
464 | ||
465 | /* Start a transaction: changes by others will not be seen during this | |
466 | * transaction, and changes will not be visible to others until end. | |
467 | */ | |
468 | int xenbus_transaction_start(struct xenbus_transaction *t) | |
469 | { | |
470 | char *id_str; | |
471 | ||
4c31a781 | 472 | transaction_start(); |
4bac07c9 JF |
473 | |
474 | id_str = xs_single(XBT_NIL, XS_TRANSACTION_START, "", NULL); | |
475 | if (IS_ERR(id_str)) { | |
4c31a781 | 476 | transaction_end(); |
4bac07c9 JF |
477 | return PTR_ERR(id_str); |
478 | } | |
479 | ||
480 | t->id = simple_strtoul(id_str, NULL, 0); | |
481 | kfree(id_str); | |
482 | return 0; | |
483 | } | |
484 | EXPORT_SYMBOL_GPL(xenbus_transaction_start); | |
485 | ||
486 | /* End a transaction. | |
487 | * If abandon is true, transaction is discarded instead of committed. | |
488 | */ | |
489 | int xenbus_transaction_end(struct xenbus_transaction t, int abort) | |
490 | { | |
491 | char abortstr[2]; | |
492 | int err; | |
493 | ||
494 | if (abort) | |
495 | strcpy(abortstr, "F"); | |
496 | else | |
497 | strcpy(abortstr, "T"); | |
498 | ||
499 | err = xs_error(xs_single(t, XS_TRANSACTION_END, abortstr, NULL)); | |
500 | ||
4c31a781 | 501 | transaction_end(); |
4bac07c9 JF |
502 | |
503 | return err; | |
504 | } | |
505 | EXPORT_SYMBOL_GPL(xenbus_transaction_end); | |
506 | ||
507 | /* Single read and scanf: returns -errno or num scanned. */ | |
508 | int xenbus_scanf(struct xenbus_transaction t, | |
509 | const char *dir, const char *node, const char *fmt, ...) | |
510 | { | |
511 | va_list ap; | |
512 | int ret; | |
513 | char *val; | |
514 | ||
515 | val = xenbus_read(t, dir, node, NULL); | |
516 | if (IS_ERR(val)) | |
517 | return PTR_ERR(val); | |
518 | ||
519 | va_start(ap, fmt); | |
520 | ret = vsscanf(val, fmt, ap); | |
521 | va_end(ap); | |
522 | kfree(val); | |
523 | /* Distinctive errno. */ | |
524 | if (ret == 0) | |
525 | return -ERANGE; | |
526 | return ret; | |
527 | } | |
528 | EXPORT_SYMBOL_GPL(xenbus_scanf); | |
529 | ||
530 | /* Single printf and write: returns -errno or 0. */ | |
531 | int xenbus_printf(struct xenbus_transaction t, | |
532 | const char *dir, const char *node, const char *fmt, ...) | |
533 | { | |
534 | va_list ap; | |
535 | int ret; | |
a800651e | 536 | char *buf; |
4bac07c9 JF |
537 | |
538 | va_start(ap, fmt); | |
a800651e | 539 | buf = kvasprintf(GFP_NOIO | __GFP_HIGH, fmt, ap); |
4bac07c9 JF |
540 | va_end(ap); |
541 | ||
a800651e IC |
542 | if (!buf) |
543 | return -ENOMEM; | |
544 | ||
545 | ret = xenbus_write(t, dir, node, buf); | |
4bac07c9 | 546 | |
a800651e | 547 | kfree(buf); |
4bac07c9 JF |
548 | |
549 | return ret; | |
550 | } | |
551 | EXPORT_SYMBOL_GPL(xenbus_printf); | |
552 | ||
553 | /* Takes tuples of names, scanf-style args, and void **, NULL terminated. */ | |
554 | int xenbus_gather(struct xenbus_transaction t, const char *dir, ...) | |
555 | { | |
556 | va_list ap; | |
557 | const char *name; | |
558 | int ret = 0; | |
559 | ||
560 | va_start(ap, dir); | |
561 | while (ret == 0 && (name = va_arg(ap, char *)) != NULL) { | |
562 | const char *fmt = va_arg(ap, char *); | |
563 | void *result = va_arg(ap, void *); | |
564 | char *p; | |
565 | ||
566 | p = xenbus_read(t, dir, name, NULL); | |
567 | if (IS_ERR(p)) { | |
568 | ret = PTR_ERR(p); | |
569 | break; | |
570 | } | |
571 | if (fmt) { | |
572 | if (sscanf(p, fmt, result) == 0) | |
573 | ret = -EINVAL; | |
574 | kfree(p); | |
575 | } else | |
576 | *(char **)result = p; | |
577 | } | |
578 | va_end(ap); | |
579 | return ret; | |
580 | } | |
581 | EXPORT_SYMBOL_GPL(xenbus_gather); | |
582 | ||
583 | static int xs_watch(const char *path, const char *token) | |
584 | { | |
585 | struct kvec iov[2]; | |
586 | ||
587 | iov[0].iov_base = (void *)path; | |
588 | iov[0].iov_len = strlen(path) + 1; | |
589 | iov[1].iov_base = (void *)token; | |
590 | iov[1].iov_len = strlen(token) + 1; | |
591 | ||
592 | return xs_error(xs_talkv(XBT_NIL, XS_WATCH, iov, | |
593 | ARRAY_SIZE(iov), NULL)); | |
594 | } | |
595 | ||
596 | static int xs_unwatch(const char *path, const char *token) | |
597 | { | |
598 | struct kvec iov[2]; | |
599 | ||
600 | iov[0].iov_base = (char *)path; | |
601 | iov[0].iov_len = strlen(path) + 1; | |
602 | iov[1].iov_base = (char *)token; | |
603 | iov[1].iov_len = strlen(token) + 1; | |
604 | ||
605 | return xs_error(xs_talkv(XBT_NIL, XS_UNWATCH, iov, | |
606 | ARRAY_SIZE(iov), NULL)); | |
607 | } | |
608 | ||
609 | static struct xenbus_watch *find_watch(const char *token) | |
610 | { | |
611 | struct xenbus_watch *i, *cmp; | |
612 | ||
613 | cmp = (void *)simple_strtoul(token, NULL, 16); | |
614 | ||
615 | list_for_each_entry(i, &watches, list) | |
616 | if (i == cmp) | |
617 | return i; | |
618 | ||
619 | return NULL; | |
620 | } | |
621 | ||
254d1a3f OH |
622 | static void xs_reset_watches(void) |
623 | { | |
624 | int err, supported = 0; | |
625 | ||
ecc635f9 | 626 | if (!xen_hvm_domain() || xen_initial_domain()) |
254d1a3f OH |
627 | return; |
628 | ||
629 | err = xenbus_scanf(XBT_NIL, "control", | |
630 | "platform-feature-xs_reset_watches", "%d", &supported); | |
631 | if (err != 1 || !supported) | |
632 | return; | |
633 | ||
634 | err = xs_error(xs_single(XBT_NIL, XS_RESET_WATCHES, "", NULL)); | |
635 | if (err && err != -EEXIST) | |
636 | printk(KERN_WARNING "xs_reset_watches failed: %d\n", err); | |
637 | } | |
638 | ||
4bac07c9 JF |
639 | /* Register callback to watch this node. */ |
640 | int register_xenbus_watch(struct xenbus_watch *watch) | |
641 | { | |
642 | /* Pointer in ascii is the token. */ | |
643 | char token[sizeof(watch) * 2 + 1]; | |
644 | int err; | |
645 | ||
646 | sprintf(token, "%lX", (long)watch); | |
647 | ||
648 | down_read(&xs_state.watch_mutex); | |
649 | ||
650 | spin_lock(&watches_lock); | |
651 | BUG_ON(find_watch(token)); | |
652 | list_add(&watch->list, &watches); | |
653 | spin_unlock(&watches_lock); | |
654 | ||
655 | err = xs_watch(watch->node, token); | |
656 | ||
c4c303c7 | 657 | if (err) { |
4bac07c9 JF |
658 | spin_lock(&watches_lock); |
659 | list_del(&watch->list); | |
660 | spin_unlock(&watches_lock); | |
661 | } | |
662 | ||
663 | up_read(&xs_state.watch_mutex); | |
664 | ||
665 | return err; | |
666 | } | |
667 | EXPORT_SYMBOL_GPL(register_xenbus_watch); | |
668 | ||
669 | void unregister_xenbus_watch(struct xenbus_watch *watch) | |
670 | { | |
671 | struct xs_stored_msg *msg, *tmp; | |
672 | char token[sizeof(watch) * 2 + 1]; | |
673 | int err; | |
674 | ||
675 | sprintf(token, "%lX", (long)watch); | |
676 | ||
677 | down_read(&xs_state.watch_mutex); | |
678 | ||
679 | spin_lock(&watches_lock); | |
680 | BUG_ON(!find_watch(token)); | |
681 | list_del(&watch->list); | |
682 | spin_unlock(&watches_lock); | |
683 | ||
684 | err = xs_unwatch(watch->node, token); | |
685 | if (err) | |
686 | printk(KERN_WARNING | |
687 | "XENBUS Failed to release watch %s: %i\n", | |
688 | watch->node, err); | |
689 | ||
690 | up_read(&xs_state.watch_mutex); | |
691 | ||
692 | /* Make sure there are no callbacks running currently (unless | |
693 | its us) */ | |
694 | if (current->pid != xenwatch_pid) | |
695 | mutex_lock(&xenwatch_mutex); | |
696 | ||
697 | /* Cancel pending watch events. */ | |
698 | spin_lock(&watch_events_lock); | |
699 | list_for_each_entry_safe(msg, tmp, &watch_events, list) { | |
700 | if (msg->u.watch.handle != watch) | |
701 | continue; | |
702 | list_del(&msg->list); | |
703 | kfree(msg->u.watch.vec); | |
704 | kfree(msg); | |
705 | } | |
706 | spin_unlock(&watch_events_lock); | |
707 | ||
708 | if (current->pid != xenwatch_pid) | |
709 | mutex_unlock(&xenwatch_mutex); | |
710 | } | |
711 | EXPORT_SYMBOL_GPL(unregister_xenbus_watch); | |
712 | ||
713 | void xs_suspend(void) | |
714 | { | |
4c31a781 | 715 | transaction_suspend(); |
4bac07c9 JF |
716 | down_write(&xs_state.watch_mutex); |
717 | mutex_lock(&xs_state.request_mutex); | |
718 | mutex_lock(&xs_state.response_mutex); | |
719 | } | |
720 | ||
721 | void xs_resume(void) | |
722 | { | |
723 | struct xenbus_watch *watch; | |
724 | char token[sizeof(watch) * 2 + 1]; | |
725 | ||
de5b31bd IC |
726 | xb_init_comms(); |
727 | ||
4bac07c9 JF |
728 | mutex_unlock(&xs_state.response_mutex); |
729 | mutex_unlock(&xs_state.request_mutex); | |
4c31a781 | 730 | transaction_resume(); |
4bac07c9 JF |
731 | |
732 | /* No need for watches_lock: the watch_mutex is sufficient. */ | |
733 | list_for_each_entry(watch, &watches, list) { | |
734 | sprintf(token, "%lX", (long)watch); | |
735 | xs_watch(watch->node, token); | |
736 | } | |
737 | ||
738 | up_write(&xs_state.watch_mutex); | |
739 | } | |
740 | ||
741 | void xs_suspend_cancel(void) | |
742 | { | |
743 | mutex_unlock(&xs_state.response_mutex); | |
744 | mutex_unlock(&xs_state.request_mutex); | |
745 | up_write(&xs_state.watch_mutex); | |
4c31a781 | 746 | mutex_unlock(&xs_state.transaction_mutex); |
4bac07c9 JF |
747 | } |
748 | ||
749 | static int xenwatch_thread(void *unused) | |
750 | { | |
751 | struct list_head *ent; | |
752 | struct xs_stored_msg *msg; | |
753 | ||
754 | for (;;) { | |
755 | wait_event_interruptible(watch_events_waitq, | |
756 | !list_empty(&watch_events)); | |
757 | ||
758 | if (kthread_should_stop()) | |
759 | break; | |
760 | ||
761 | mutex_lock(&xenwatch_mutex); | |
762 | ||
763 | spin_lock(&watch_events_lock); | |
764 | ent = watch_events.next; | |
765 | if (ent != &watch_events) | |
766 | list_del(ent); | |
767 | spin_unlock(&watch_events_lock); | |
768 | ||
769 | if (ent != &watch_events) { | |
770 | msg = list_entry(ent, struct xs_stored_msg, list); | |
771 | msg->u.watch.handle->callback( | |
772 | msg->u.watch.handle, | |
773 | (const char **)msg->u.watch.vec, | |
774 | msg->u.watch.vec_size); | |
775 | kfree(msg->u.watch.vec); | |
776 | kfree(msg); | |
777 | } | |
778 | ||
779 | mutex_unlock(&xenwatch_mutex); | |
780 | } | |
781 | ||
782 | return 0; | |
783 | } | |
784 | ||
785 | static int process_msg(void) | |
786 | { | |
787 | struct xs_stored_msg *msg; | |
788 | char *body; | |
789 | int err; | |
790 | ||
791 | /* | |
792 | * We must disallow save/restore while reading a xenstore message. | |
793 | * A partial read across s/r leaves us out of sync with xenstored. | |
794 | */ | |
795 | for (;;) { | |
796 | err = xb_wait_for_data_to_read(); | |
797 | if (err) | |
798 | return err; | |
799 | mutex_lock(&xs_state.response_mutex); | |
800 | if (xb_data_to_read()) | |
801 | break; | |
802 | /* We raced with save/restore: pending data 'disappeared'. */ | |
803 | mutex_unlock(&xs_state.response_mutex); | |
804 | } | |
805 | ||
806 | ||
a144ff09 | 807 | msg = kmalloc(sizeof(*msg), GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
808 | if (msg == NULL) { |
809 | err = -ENOMEM; | |
810 | goto out; | |
811 | } | |
812 | ||
813 | err = xb_read(&msg->hdr, sizeof(msg->hdr)); | |
814 | if (err) { | |
815 | kfree(msg); | |
816 | goto out; | |
817 | } | |
818 | ||
9e7860ce IC |
819 | if (msg->hdr.len > XENSTORE_PAYLOAD_MAX) { |
820 | kfree(msg); | |
821 | err = -EINVAL; | |
822 | goto out; | |
823 | } | |
824 | ||
a144ff09 | 825 | body = kmalloc(msg->hdr.len + 1, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
826 | if (body == NULL) { |
827 | kfree(msg); | |
828 | err = -ENOMEM; | |
829 | goto out; | |
830 | } | |
831 | ||
832 | err = xb_read(body, msg->hdr.len); | |
833 | if (err) { | |
834 | kfree(body); | |
835 | kfree(msg); | |
836 | goto out; | |
837 | } | |
838 | body[msg->hdr.len] = '\0'; | |
839 | ||
840 | if (msg->hdr.type == XS_WATCH_EVENT) { | |
841 | msg->u.watch.vec = split(body, msg->hdr.len, | |
842 | &msg->u.watch.vec_size); | |
843 | if (IS_ERR(msg->u.watch.vec)) { | |
4bac07c9 | 844 | err = PTR_ERR(msg->u.watch.vec); |
98ac0e53 | 845 | kfree(msg); |
4bac07c9 JF |
846 | goto out; |
847 | } | |
848 | ||
849 | spin_lock(&watches_lock); | |
850 | msg->u.watch.handle = find_watch( | |
851 | msg->u.watch.vec[XS_WATCH_TOKEN]); | |
852 | if (msg->u.watch.handle != NULL) { | |
853 | spin_lock(&watch_events_lock); | |
854 | list_add_tail(&msg->list, &watch_events); | |
855 | wake_up(&watch_events_waitq); | |
856 | spin_unlock(&watch_events_lock); | |
857 | } else { | |
858 | kfree(msg->u.watch.vec); | |
859 | kfree(msg); | |
860 | } | |
861 | spin_unlock(&watches_lock); | |
862 | } else { | |
863 | msg->u.reply.body = body; | |
864 | spin_lock(&xs_state.reply_lock); | |
865 | list_add_tail(&msg->list, &xs_state.reply_list); | |
866 | spin_unlock(&xs_state.reply_lock); | |
867 | wake_up(&xs_state.reply_waitq); | |
868 | } | |
869 | ||
870 | out: | |
871 | mutex_unlock(&xs_state.response_mutex); | |
872 | return err; | |
873 | } | |
874 | ||
875 | static int xenbus_thread(void *unused) | |
876 | { | |
877 | int err; | |
878 | ||
879 | for (;;) { | |
880 | err = process_msg(); | |
881 | if (err) | |
882 | printk(KERN_WARNING "XENBUS error %d while reading " | |
883 | "message\n", err); | |
884 | if (kthread_should_stop()) | |
885 | break; | |
886 | } | |
887 | ||
888 | return 0; | |
889 | } | |
890 | ||
891 | int xs_init(void) | |
892 | { | |
893 | int err; | |
894 | struct task_struct *task; | |
895 | ||
896 | INIT_LIST_HEAD(&xs_state.reply_list); | |
897 | spin_lock_init(&xs_state.reply_lock); | |
898 | init_waitqueue_head(&xs_state.reply_waitq); | |
899 | ||
900 | mutex_init(&xs_state.request_mutex); | |
901 | mutex_init(&xs_state.response_mutex); | |
4c31a781 | 902 | mutex_init(&xs_state.transaction_mutex); |
4bac07c9 | 903 | init_rwsem(&xs_state.watch_mutex); |
4c31a781 IC |
904 | atomic_set(&xs_state.transaction_count, 0); |
905 | init_waitqueue_head(&xs_state.transaction_wq); | |
4bac07c9 JF |
906 | |
907 | /* Initialize the shared memory rings to talk to xenstored */ | |
908 | err = xb_init_comms(); | |
909 | if (err) | |
910 | return err; | |
911 | ||
912 | task = kthread_run(xenwatch_thread, NULL, "xenwatch"); | |
913 | if (IS_ERR(task)) | |
914 | return PTR_ERR(task); | |
915 | xenwatch_pid = task->pid; | |
916 | ||
917 | task = kthread_run(xenbus_thread, NULL, "xenbus"); | |
918 | if (IS_ERR(task)) | |
919 | return PTR_ERR(task); | |
920 | ||
254d1a3f OH |
921 | /* shutdown watches for kexec boot */ |
922 | xs_reset_watches(); | |
923 | ||
4bac07c9 JF |
924 | return 0; |
925 | } |