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 | ||
283c0972 JP |
34 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
35 | ||
4bac07c9 JF |
36 | #include <linux/unistd.h> |
37 | #include <linux/errno.h> | |
38 | #include <linux/types.h> | |
39 | #include <linux/uio.h> | |
40 | #include <linux/kernel.h> | |
41 | #include <linux/string.h> | |
42 | #include <linux/err.h> | |
43 | #include <linux/slab.h> | |
44 | #include <linux/fcntl.h> | |
45 | #include <linux/kthread.h> | |
46 | #include <linux/rwsem.h> | |
47 | #include <linux/module.h> | |
48 | #include <linux/mutex.h> | |
ecc635f9 | 49 | #include <asm/xen/hypervisor.h> |
4bac07c9 | 50 | #include <xen/xenbus.h> |
5b25d89e | 51 | #include <xen/xen.h> |
4bac07c9 JF |
52 | #include "xenbus_comms.h" |
53 | ||
54 | struct xs_stored_msg { | |
55 | struct list_head list; | |
56 | ||
57 | struct xsd_sockmsg hdr; | |
58 | ||
59 | union { | |
60 | /* Queued replies. */ | |
61 | struct { | |
62 | char *body; | |
63 | } reply; | |
64 | ||
65 | /* Queued watch events. */ | |
66 | struct { | |
67 | struct xenbus_watch *handle; | |
68 | char **vec; | |
69 | unsigned int vec_size; | |
70 | } watch; | |
71 | } u; | |
72 | }; | |
73 | ||
74 | struct xs_handle { | |
75 | /* A list of replies. Currently only one will ever be outstanding. */ | |
76 | struct list_head reply_list; | |
77 | spinlock_t reply_lock; | |
78 | wait_queue_head_t reply_waitq; | |
79 | ||
80 | /* | |
81 | * Mutex ordering: transaction_mutex -> watch_mutex -> request_mutex. | |
82 | * response_mutex is never taken simultaneously with the other three. | |
4c31a781 IC |
83 | * |
84 | * transaction_mutex must be held before incrementing | |
85 | * transaction_count. The mutex is held when a suspend is in | |
86 | * progress to prevent new transactions starting. | |
87 | * | |
88 | * When decrementing transaction_count to zero the wait queue | |
89 | * should be woken up, the suspend code waits for count to | |
90 | * reach zero. | |
4bac07c9 JF |
91 | */ |
92 | ||
93 | /* One request at a time. */ | |
94 | struct mutex request_mutex; | |
95 | ||
96 | /* Protect xenbus reader thread against save/restore. */ | |
97 | struct mutex response_mutex; | |
98 | ||
99 | /* Protect transactions against save/restore. */ | |
4c31a781 IC |
100 | struct mutex transaction_mutex; |
101 | atomic_t transaction_count; | |
102 | wait_queue_head_t transaction_wq; | |
4bac07c9 JF |
103 | |
104 | /* Protect watch (de)register against save/restore. */ | |
105 | struct rw_semaphore watch_mutex; | |
106 | }; | |
107 | ||
108 | static struct xs_handle xs_state; | |
109 | ||
110 | /* List of registered watches, and a lock to protect it. */ | |
111 | static LIST_HEAD(watches); | |
112 | static DEFINE_SPINLOCK(watches_lock); | |
113 | ||
114 | /* List of pending watch callback events, and a lock to protect it. */ | |
115 | static LIST_HEAD(watch_events); | |
116 | static DEFINE_SPINLOCK(watch_events_lock); | |
117 | ||
118 | /* | |
119 | * Details of the xenwatch callback kernel thread. The thread waits on the | |
120 | * watch_events_waitq for work to do (queued on watch_events list). When it | |
121 | * wakes up it acquires the xenwatch_mutex before reading the list and | |
122 | * carrying out work. | |
123 | */ | |
124 | static pid_t xenwatch_pid; | |
125 | static DEFINE_MUTEX(xenwatch_mutex); | |
126 | static DECLARE_WAIT_QUEUE_HEAD(watch_events_waitq); | |
127 | ||
128 | static int get_error(const char *errorstring) | |
129 | { | |
130 | unsigned int i; | |
131 | ||
132 | for (i = 0; strcmp(errorstring, xsd_errors[i].errstring) != 0; i++) { | |
133 | if (i == ARRAY_SIZE(xsd_errors) - 1) { | |
283c0972 JP |
134 | pr_warn("xen store gave: unknown error %s\n", |
135 | errorstring); | |
4bac07c9 JF |
136 | return EINVAL; |
137 | } | |
138 | } | |
139 | return xsd_errors[i].errnum; | |
140 | } | |
141 | ||
142 | static void *read_reply(enum xsd_sockmsg_type *type, unsigned int *len) | |
143 | { | |
144 | struct xs_stored_msg *msg; | |
145 | char *body; | |
146 | ||
147 | spin_lock(&xs_state.reply_lock); | |
148 | ||
149 | while (list_empty(&xs_state.reply_list)) { | |
150 | spin_unlock(&xs_state.reply_lock); | |
151 | /* XXX FIXME: Avoid synchronous wait for response here. */ | |
152 | wait_event(xs_state.reply_waitq, | |
153 | !list_empty(&xs_state.reply_list)); | |
154 | spin_lock(&xs_state.reply_lock); | |
155 | } | |
156 | ||
157 | msg = list_entry(xs_state.reply_list.next, | |
158 | struct xs_stored_msg, list); | |
159 | list_del(&msg->list); | |
160 | ||
161 | spin_unlock(&xs_state.reply_lock); | |
162 | ||
163 | *type = msg->hdr.type; | |
164 | if (len) | |
165 | *len = msg->hdr.len; | |
166 | body = msg->u.reply.body; | |
167 | ||
168 | kfree(msg); | |
169 | ||
170 | return body; | |
171 | } | |
172 | ||
4c31a781 IC |
173 | static void transaction_start(void) |
174 | { | |
175 | mutex_lock(&xs_state.transaction_mutex); | |
176 | atomic_inc(&xs_state.transaction_count); | |
177 | mutex_unlock(&xs_state.transaction_mutex); | |
178 | } | |
179 | ||
180 | static void transaction_end(void) | |
181 | { | |
182 | if (atomic_dec_and_test(&xs_state.transaction_count)) | |
183 | wake_up(&xs_state.transaction_wq); | |
184 | } | |
185 | ||
186 | static void transaction_suspend(void) | |
187 | { | |
188 | mutex_lock(&xs_state.transaction_mutex); | |
189 | wait_event(xs_state.transaction_wq, | |
190 | atomic_read(&xs_state.transaction_count) == 0); | |
191 | } | |
192 | ||
193 | static void transaction_resume(void) | |
194 | { | |
195 | mutex_unlock(&xs_state.transaction_mutex); | |
196 | } | |
197 | ||
4bac07c9 JF |
198 | void *xenbus_dev_request_and_reply(struct xsd_sockmsg *msg) |
199 | { | |
200 | void *ret; | |
201 | struct xsd_sockmsg req_msg = *msg; | |
202 | int err; | |
203 | ||
204 | if (req_msg.type == XS_TRANSACTION_START) | |
4c31a781 | 205 | transaction_start(); |
4bac07c9 JF |
206 | |
207 | mutex_lock(&xs_state.request_mutex); | |
208 | ||
209 | err = xb_write(msg, sizeof(*msg) + msg->len); | |
210 | if (err) { | |
211 | msg->type = XS_ERROR; | |
212 | ret = ERR_PTR(err); | |
213 | } else | |
214 | ret = read_reply(&msg->type, &msg->len); | |
215 | ||
216 | mutex_unlock(&xs_state.request_mutex); | |
217 | ||
218 | if ((msg->type == XS_TRANSACTION_END) || | |
219 | ((req_msg.type == XS_TRANSACTION_START) && | |
220 | (msg->type == XS_ERROR))) | |
4c31a781 | 221 | transaction_end(); |
4bac07c9 JF |
222 | |
223 | return ret; | |
224 | } | |
1107ba88 | 225 | EXPORT_SYMBOL(xenbus_dev_request_and_reply); |
4bac07c9 JF |
226 | |
227 | /* Send message to xs, get kmalloc'ed reply. ERR_PTR() on error. */ | |
228 | static void *xs_talkv(struct xenbus_transaction t, | |
229 | enum xsd_sockmsg_type type, | |
230 | const struct kvec *iovec, | |
231 | unsigned int num_vecs, | |
232 | unsigned int *len) | |
233 | { | |
234 | struct xsd_sockmsg msg; | |
235 | void *ret = NULL; | |
236 | unsigned int i; | |
237 | int err; | |
238 | ||
239 | msg.tx_id = t.id; | |
240 | msg.req_id = 0; | |
241 | msg.type = type; | |
242 | msg.len = 0; | |
243 | for (i = 0; i < num_vecs; i++) | |
244 | msg.len += iovec[i].iov_len; | |
245 | ||
246 | mutex_lock(&xs_state.request_mutex); | |
247 | ||
248 | err = xb_write(&msg, sizeof(msg)); | |
249 | if (err) { | |
250 | mutex_unlock(&xs_state.request_mutex); | |
251 | return ERR_PTR(err); | |
252 | } | |
253 | ||
254 | for (i = 0; i < num_vecs; i++) { | |
255 | err = xb_write(iovec[i].iov_base, iovec[i].iov_len); | |
256 | if (err) { | |
257 | mutex_unlock(&xs_state.request_mutex); | |
258 | return ERR_PTR(err); | |
259 | } | |
260 | } | |
261 | ||
262 | ret = read_reply(&msg.type, len); | |
263 | ||
264 | mutex_unlock(&xs_state.request_mutex); | |
265 | ||
266 | if (IS_ERR(ret)) | |
267 | return ret; | |
268 | ||
269 | if (msg.type == XS_ERROR) { | |
270 | err = get_error(ret); | |
271 | kfree(ret); | |
272 | return ERR_PTR(-err); | |
273 | } | |
274 | ||
275 | if (msg.type != type) { | |
283c0972 JP |
276 | pr_warn_ratelimited("unexpected type [%d], expected [%d]\n", |
277 | msg.type, type); | |
4bac07c9 JF |
278 | kfree(ret); |
279 | return ERR_PTR(-EINVAL); | |
280 | } | |
281 | return ret; | |
282 | } | |
283 | ||
284 | /* Simplified version of xs_talkv: single message. */ | |
285 | static void *xs_single(struct xenbus_transaction t, | |
286 | enum xsd_sockmsg_type type, | |
287 | const char *string, | |
288 | unsigned int *len) | |
289 | { | |
290 | struct kvec iovec; | |
291 | ||
292 | iovec.iov_base = (void *)string; | |
293 | iovec.iov_len = strlen(string) + 1; | |
294 | return xs_talkv(t, type, &iovec, 1, len); | |
295 | } | |
296 | ||
297 | /* Many commands only need an ack, don't care what it says. */ | |
298 | static int xs_error(char *reply) | |
299 | { | |
300 | if (IS_ERR(reply)) | |
301 | return PTR_ERR(reply); | |
302 | kfree(reply); | |
303 | return 0; | |
304 | } | |
305 | ||
306 | static unsigned int count_strings(const char *strings, unsigned int len) | |
307 | { | |
308 | unsigned int num; | |
309 | const char *p; | |
310 | ||
311 | for (p = strings, num = 0; p < strings + len; p += strlen(p) + 1) | |
312 | num++; | |
313 | ||
314 | return num; | |
315 | } | |
316 | ||
317 | /* Return the path to dir with /name appended. Buffer must be kfree()'ed. */ | |
318 | static char *join(const char *dir, const char *name) | |
319 | { | |
320 | char *buffer; | |
321 | ||
322 | if (strlen(name) == 0) | |
a144ff09 | 323 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s", dir); |
4bac07c9 | 324 | else |
a144ff09 | 325 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s/%s", dir, name); |
4bac07c9 JF |
326 | return (!buffer) ? ERR_PTR(-ENOMEM) : buffer; |
327 | } | |
328 | ||
329 | static char **split(char *strings, unsigned int len, unsigned int *num) | |
330 | { | |
331 | char *p, **ret; | |
332 | ||
333 | /* Count the strings. */ | |
334 | *num = count_strings(strings, len); | |
335 | ||
336 | /* Transfer to one big alloc for easy freeing. */ | |
a144ff09 | 337 | ret = kmalloc(*num * sizeof(char *) + len, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
338 | if (!ret) { |
339 | kfree(strings); | |
340 | return ERR_PTR(-ENOMEM); | |
341 | } | |
342 | memcpy(&ret[*num], strings, len); | |
343 | kfree(strings); | |
344 | ||
345 | strings = (char *)&ret[*num]; | |
346 | for (p = strings, *num = 0; p < strings + len; p += strlen(p) + 1) | |
347 | ret[(*num)++] = p; | |
348 | ||
349 | return ret; | |
350 | } | |
351 | ||
352 | char **xenbus_directory(struct xenbus_transaction t, | |
353 | const char *dir, const char *node, unsigned int *num) | |
354 | { | |
355 | char *strings, *path; | |
356 | unsigned int len; | |
357 | ||
358 | path = join(dir, node); | |
359 | if (IS_ERR(path)) | |
360 | return (char **)path; | |
361 | ||
362 | strings = xs_single(t, XS_DIRECTORY, path, &len); | |
363 | kfree(path); | |
364 | if (IS_ERR(strings)) | |
365 | return (char **)strings; | |
366 | ||
367 | return split(strings, len, num); | |
368 | } | |
369 | EXPORT_SYMBOL_GPL(xenbus_directory); | |
370 | ||
371 | /* Check if a path exists. Return 1 if it does. */ | |
372 | int xenbus_exists(struct xenbus_transaction t, | |
373 | const char *dir, const char *node) | |
374 | { | |
375 | char **d; | |
376 | int dir_n; | |
377 | ||
378 | d = xenbus_directory(t, dir, node, &dir_n); | |
379 | if (IS_ERR(d)) | |
380 | return 0; | |
381 | kfree(d); | |
382 | return 1; | |
383 | } | |
384 | EXPORT_SYMBOL_GPL(xenbus_exists); | |
385 | ||
386 | /* Get the value of a single file. | |
387 | * Returns a kmalloced value: call free() on it after use. | |
388 | * len indicates length in bytes. | |
389 | */ | |
390 | void *xenbus_read(struct xenbus_transaction t, | |
391 | const char *dir, const char *node, unsigned int *len) | |
392 | { | |
393 | char *path; | |
394 | void *ret; | |
395 | ||
396 | path = join(dir, node); | |
397 | if (IS_ERR(path)) | |
398 | return (void *)path; | |
399 | ||
400 | ret = xs_single(t, XS_READ, path, len); | |
401 | kfree(path); | |
402 | return ret; | |
403 | } | |
404 | EXPORT_SYMBOL_GPL(xenbus_read); | |
405 | ||
406 | /* Write the value of a single file. | |
407 | * Returns -err on failure. | |
408 | */ | |
409 | int xenbus_write(struct xenbus_transaction t, | |
410 | const char *dir, const char *node, const char *string) | |
411 | { | |
412 | const char *path; | |
413 | struct kvec iovec[2]; | |
414 | int ret; | |
415 | ||
416 | path = join(dir, node); | |
417 | if (IS_ERR(path)) | |
418 | return PTR_ERR(path); | |
419 | ||
420 | iovec[0].iov_base = (void *)path; | |
421 | iovec[0].iov_len = strlen(path) + 1; | |
422 | iovec[1].iov_base = (void *)string; | |
423 | iovec[1].iov_len = strlen(string); | |
424 | ||
425 | ret = xs_error(xs_talkv(t, XS_WRITE, iovec, ARRAY_SIZE(iovec), NULL)); | |
426 | kfree(path); | |
427 | return ret; | |
428 | } | |
429 | EXPORT_SYMBOL_GPL(xenbus_write); | |
430 | ||
431 | /* Create a new directory. */ | |
432 | int xenbus_mkdir(struct xenbus_transaction t, | |
433 | const char *dir, const char *node) | |
434 | { | |
435 | char *path; | |
436 | int ret; | |
437 | ||
438 | path = join(dir, node); | |
439 | if (IS_ERR(path)) | |
440 | return PTR_ERR(path); | |
441 | ||
442 | ret = xs_error(xs_single(t, XS_MKDIR, path, NULL)); | |
443 | kfree(path); | |
444 | return ret; | |
445 | } | |
446 | EXPORT_SYMBOL_GPL(xenbus_mkdir); | |
447 | ||
448 | /* Destroy a file or directory (directories must be empty). */ | |
449 | int xenbus_rm(struct xenbus_transaction t, const char *dir, const char *node) | |
450 | { | |
451 | char *path; | |
452 | int ret; | |
453 | ||
454 | path = join(dir, node); | |
455 | if (IS_ERR(path)) | |
456 | return PTR_ERR(path); | |
457 | ||
458 | ret = xs_error(xs_single(t, XS_RM, path, NULL)); | |
459 | kfree(path); | |
460 | return ret; | |
461 | } | |
462 | EXPORT_SYMBOL_GPL(xenbus_rm); | |
463 | ||
464 | /* Start a transaction: changes by others will not be seen during this | |
465 | * transaction, and changes will not be visible to others until end. | |
466 | */ | |
467 | int xenbus_transaction_start(struct xenbus_transaction *t) | |
468 | { | |
469 | char *id_str; | |
470 | ||
4c31a781 | 471 | transaction_start(); |
4bac07c9 JF |
472 | |
473 | id_str = xs_single(XBT_NIL, XS_TRANSACTION_START, "", NULL); | |
474 | if (IS_ERR(id_str)) { | |
4c31a781 | 475 | transaction_end(); |
4bac07c9 JF |
476 | return PTR_ERR(id_str); |
477 | } | |
478 | ||
479 | t->id = simple_strtoul(id_str, NULL, 0); | |
480 | kfree(id_str); | |
481 | return 0; | |
482 | } | |
483 | EXPORT_SYMBOL_GPL(xenbus_transaction_start); | |
484 | ||
485 | /* End a transaction. | |
486 | * If abandon is true, transaction is discarded instead of committed. | |
487 | */ | |
488 | int xenbus_transaction_end(struct xenbus_transaction t, int abort) | |
489 | { | |
490 | char abortstr[2]; | |
491 | int err; | |
492 | ||
493 | if (abort) | |
494 | strcpy(abortstr, "F"); | |
495 | else | |
496 | strcpy(abortstr, "T"); | |
497 | ||
498 | err = xs_error(xs_single(t, XS_TRANSACTION_END, abortstr, NULL)); | |
499 | ||
4c31a781 | 500 | transaction_end(); |
4bac07c9 JF |
501 | |
502 | return err; | |
503 | } | |
504 | EXPORT_SYMBOL_GPL(xenbus_transaction_end); | |
505 | ||
506 | /* Single read and scanf: returns -errno or num scanned. */ | |
507 | int xenbus_scanf(struct xenbus_transaction t, | |
508 | const char *dir, const char *node, const char *fmt, ...) | |
509 | { | |
510 | va_list ap; | |
511 | int ret; | |
512 | char *val; | |
513 | ||
514 | val = xenbus_read(t, dir, node, NULL); | |
515 | if (IS_ERR(val)) | |
516 | return PTR_ERR(val); | |
517 | ||
518 | va_start(ap, fmt); | |
519 | ret = vsscanf(val, fmt, ap); | |
520 | va_end(ap); | |
521 | kfree(val); | |
522 | /* Distinctive errno. */ | |
523 | if (ret == 0) | |
524 | return -ERANGE; | |
525 | return ret; | |
526 | } | |
527 | EXPORT_SYMBOL_GPL(xenbus_scanf); | |
528 | ||
529 | /* Single printf and write: returns -errno or 0. */ | |
530 | int xenbus_printf(struct xenbus_transaction t, | |
531 | const char *dir, const char *node, const char *fmt, ...) | |
532 | { | |
533 | va_list ap; | |
534 | int ret; | |
a800651e | 535 | char *buf; |
4bac07c9 JF |
536 | |
537 | va_start(ap, fmt); | |
a800651e | 538 | buf = kvasprintf(GFP_NOIO | __GFP_HIGH, fmt, ap); |
4bac07c9 JF |
539 | va_end(ap); |
540 | ||
a800651e IC |
541 | if (!buf) |
542 | return -ENOMEM; | |
543 | ||
544 | ret = xenbus_write(t, dir, node, buf); | |
4bac07c9 | 545 | |
a800651e | 546 | kfree(buf); |
4bac07c9 JF |
547 | |
548 | return ret; | |
549 | } | |
550 | EXPORT_SYMBOL_GPL(xenbus_printf); | |
551 | ||
552 | /* Takes tuples of names, scanf-style args, and void **, NULL terminated. */ | |
553 | int xenbus_gather(struct xenbus_transaction t, const char *dir, ...) | |
554 | { | |
555 | va_list ap; | |
556 | const char *name; | |
557 | int ret = 0; | |
558 | ||
559 | va_start(ap, dir); | |
560 | while (ret == 0 && (name = va_arg(ap, char *)) != NULL) { | |
561 | const char *fmt = va_arg(ap, char *); | |
562 | void *result = va_arg(ap, void *); | |
563 | char *p; | |
564 | ||
565 | p = xenbus_read(t, dir, name, NULL); | |
566 | if (IS_ERR(p)) { | |
567 | ret = PTR_ERR(p); | |
568 | break; | |
569 | } | |
570 | if (fmt) { | |
571 | if (sscanf(p, fmt, result) == 0) | |
572 | ret = -EINVAL; | |
573 | kfree(p); | |
574 | } else | |
575 | *(char **)result = p; | |
576 | } | |
577 | va_end(ap); | |
578 | return ret; | |
579 | } | |
580 | EXPORT_SYMBOL_GPL(xenbus_gather); | |
581 | ||
582 | static int xs_watch(const char *path, const char *token) | |
583 | { | |
584 | struct kvec iov[2]; | |
585 | ||
586 | iov[0].iov_base = (void *)path; | |
587 | iov[0].iov_len = strlen(path) + 1; | |
588 | iov[1].iov_base = (void *)token; | |
589 | iov[1].iov_len = strlen(token) + 1; | |
590 | ||
591 | return xs_error(xs_talkv(XBT_NIL, XS_WATCH, iov, | |
592 | ARRAY_SIZE(iov), NULL)); | |
593 | } | |
594 | ||
595 | static int xs_unwatch(const char *path, const char *token) | |
596 | { | |
597 | struct kvec iov[2]; | |
598 | ||
599 | iov[0].iov_base = (char *)path; | |
600 | iov[0].iov_len = strlen(path) + 1; | |
601 | iov[1].iov_base = (char *)token; | |
602 | iov[1].iov_len = strlen(token) + 1; | |
603 | ||
604 | return xs_error(xs_talkv(XBT_NIL, XS_UNWATCH, iov, | |
605 | ARRAY_SIZE(iov), NULL)); | |
606 | } | |
607 | ||
608 | static struct xenbus_watch *find_watch(const char *token) | |
609 | { | |
610 | struct xenbus_watch *i, *cmp; | |
611 | ||
612 | cmp = (void *)simple_strtoul(token, NULL, 16); | |
613 | ||
614 | list_for_each_entry(i, &watches, list) | |
615 | if (i == cmp) | |
616 | return i; | |
617 | ||
618 | return NULL; | |
619 | } | |
cb6b6df1 KRW |
620 | /* |
621 | * Certain older XenBus toolstack cannot handle reading values that are | |
622 | * not populated. Some Xen 3.4 installation are incapable of doing this | |
623 | * so if we are running on anything older than 4 do not attempt to read | |
624 | * control/platform-feature-xs_reset_watches. | |
625 | */ | |
e9d1aa05 | 626 | static bool xen_strict_xenbus_quirk(void) |
cb6b6df1 | 627 | { |
7644bdac | 628 | #ifdef CONFIG_X86 |
cb6b6df1 KRW |
629 | uint32_t eax, ebx, ecx, edx, base; |
630 | ||
631 | base = xen_cpuid_base(); | |
632 | cpuid(base + 1, &eax, &ebx, &ecx, &edx); | |
4bac07c9 | 633 | |
cb6b6df1 KRW |
634 | if ((eax >> 16) < 4) |
635 | return true; | |
7644bdac | 636 | #endif |
cb6b6df1 KRW |
637 | return false; |
638 | ||
639 | } | |
254d1a3f OH |
640 | static void xs_reset_watches(void) |
641 | { | |
642 | int err, supported = 0; | |
643 | ||
ecc635f9 | 644 | if (!xen_hvm_domain() || xen_initial_domain()) |
254d1a3f OH |
645 | return; |
646 | ||
cb6b6df1 KRW |
647 | if (xen_strict_xenbus_quirk()) |
648 | return; | |
649 | ||
254d1a3f OH |
650 | err = xenbus_scanf(XBT_NIL, "control", |
651 | "platform-feature-xs_reset_watches", "%d", &supported); | |
652 | if (err != 1 || !supported) | |
653 | return; | |
654 | ||
655 | err = xs_error(xs_single(XBT_NIL, XS_RESET_WATCHES, "", NULL)); | |
656 | if (err && err != -EEXIST) | |
283c0972 | 657 | pr_warn("xs_reset_watches failed: %d\n", err); |
254d1a3f OH |
658 | } |
659 | ||
4bac07c9 JF |
660 | /* Register callback to watch this node. */ |
661 | int register_xenbus_watch(struct xenbus_watch *watch) | |
662 | { | |
663 | /* Pointer in ascii is the token. */ | |
664 | char token[sizeof(watch) * 2 + 1]; | |
665 | int err; | |
666 | ||
667 | sprintf(token, "%lX", (long)watch); | |
668 | ||
669 | down_read(&xs_state.watch_mutex); | |
670 | ||
671 | spin_lock(&watches_lock); | |
672 | BUG_ON(find_watch(token)); | |
673 | list_add(&watch->list, &watches); | |
674 | spin_unlock(&watches_lock); | |
675 | ||
676 | err = xs_watch(watch->node, token); | |
677 | ||
c4c303c7 | 678 | if (err) { |
4bac07c9 JF |
679 | spin_lock(&watches_lock); |
680 | list_del(&watch->list); | |
681 | spin_unlock(&watches_lock); | |
682 | } | |
683 | ||
684 | up_read(&xs_state.watch_mutex); | |
685 | ||
686 | return err; | |
687 | } | |
688 | EXPORT_SYMBOL_GPL(register_xenbus_watch); | |
689 | ||
690 | void unregister_xenbus_watch(struct xenbus_watch *watch) | |
691 | { | |
692 | struct xs_stored_msg *msg, *tmp; | |
693 | char token[sizeof(watch) * 2 + 1]; | |
694 | int err; | |
695 | ||
696 | sprintf(token, "%lX", (long)watch); | |
697 | ||
698 | down_read(&xs_state.watch_mutex); | |
699 | ||
700 | spin_lock(&watches_lock); | |
701 | BUG_ON(!find_watch(token)); | |
702 | list_del(&watch->list); | |
703 | spin_unlock(&watches_lock); | |
704 | ||
705 | err = xs_unwatch(watch->node, token); | |
706 | if (err) | |
283c0972 | 707 | pr_warn("Failed to release watch %s: %i\n", watch->node, err); |
4bac07c9 JF |
708 | |
709 | up_read(&xs_state.watch_mutex); | |
710 | ||
711 | /* Make sure there are no callbacks running currently (unless | |
712 | its us) */ | |
713 | if (current->pid != xenwatch_pid) | |
714 | mutex_lock(&xenwatch_mutex); | |
715 | ||
716 | /* Cancel pending watch events. */ | |
717 | spin_lock(&watch_events_lock); | |
718 | list_for_each_entry_safe(msg, tmp, &watch_events, list) { | |
719 | if (msg->u.watch.handle != watch) | |
720 | continue; | |
721 | list_del(&msg->list); | |
722 | kfree(msg->u.watch.vec); | |
723 | kfree(msg); | |
724 | } | |
725 | spin_unlock(&watch_events_lock); | |
726 | ||
727 | if (current->pid != xenwatch_pid) | |
728 | mutex_unlock(&xenwatch_mutex); | |
729 | } | |
730 | EXPORT_SYMBOL_GPL(unregister_xenbus_watch); | |
731 | ||
732 | void xs_suspend(void) | |
733 | { | |
4c31a781 | 734 | transaction_suspend(); |
4bac07c9 JF |
735 | down_write(&xs_state.watch_mutex); |
736 | mutex_lock(&xs_state.request_mutex); | |
737 | mutex_lock(&xs_state.response_mutex); | |
738 | } | |
739 | ||
740 | void xs_resume(void) | |
741 | { | |
742 | struct xenbus_watch *watch; | |
743 | char token[sizeof(watch) * 2 + 1]; | |
744 | ||
de5b31bd IC |
745 | xb_init_comms(); |
746 | ||
4bac07c9 JF |
747 | mutex_unlock(&xs_state.response_mutex); |
748 | mutex_unlock(&xs_state.request_mutex); | |
4c31a781 | 749 | transaction_resume(); |
4bac07c9 JF |
750 | |
751 | /* No need for watches_lock: the watch_mutex is sufficient. */ | |
752 | list_for_each_entry(watch, &watches, list) { | |
753 | sprintf(token, "%lX", (long)watch); | |
754 | xs_watch(watch->node, token); | |
755 | } | |
756 | ||
757 | up_write(&xs_state.watch_mutex); | |
758 | } | |
759 | ||
760 | void xs_suspend_cancel(void) | |
761 | { | |
762 | mutex_unlock(&xs_state.response_mutex); | |
763 | mutex_unlock(&xs_state.request_mutex); | |
764 | up_write(&xs_state.watch_mutex); | |
4c31a781 | 765 | mutex_unlock(&xs_state.transaction_mutex); |
4bac07c9 JF |
766 | } |
767 | ||
768 | static int xenwatch_thread(void *unused) | |
769 | { | |
770 | struct list_head *ent; | |
771 | struct xs_stored_msg *msg; | |
772 | ||
773 | for (;;) { | |
774 | wait_event_interruptible(watch_events_waitq, | |
775 | !list_empty(&watch_events)); | |
776 | ||
777 | if (kthread_should_stop()) | |
778 | break; | |
779 | ||
780 | mutex_lock(&xenwatch_mutex); | |
781 | ||
782 | spin_lock(&watch_events_lock); | |
783 | ent = watch_events.next; | |
784 | if (ent != &watch_events) | |
785 | list_del(ent); | |
786 | spin_unlock(&watch_events_lock); | |
787 | ||
788 | if (ent != &watch_events) { | |
789 | msg = list_entry(ent, struct xs_stored_msg, list); | |
790 | msg->u.watch.handle->callback( | |
791 | msg->u.watch.handle, | |
792 | (const char **)msg->u.watch.vec, | |
793 | msg->u.watch.vec_size); | |
794 | kfree(msg->u.watch.vec); | |
795 | kfree(msg); | |
796 | } | |
797 | ||
798 | mutex_unlock(&xenwatch_mutex); | |
799 | } | |
800 | ||
801 | return 0; | |
802 | } | |
803 | ||
804 | static int process_msg(void) | |
805 | { | |
806 | struct xs_stored_msg *msg; | |
807 | char *body; | |
808 | int err; | |
809 | ||
810 | /* | |
811 | * We must disallow save/restore while reading a xenstore message. | |
812 | * A partial read across s/r leaves us out of sync with xenstored. | |
813 | */ | |
814 | for (;;) { | |
815 | err = xb_wait_for_data_to_read(); | |
816 | if (err) | |
817 | return err; | |
818 | mutex_lock(&xs_state.response_mutex); | |
819 | if (xb_data_to_read()) | |
820 | break; | |
821 | /* We raced with save/restore: pending data 'disappeared'. */ | |
822 | mutex_unlock(&xs_state.response_mutex); | |
823 | } | |
824 | ||
825 | ||
a144ff09 | 826 | msg = kmalloc(sizeof(*msg), GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
827 | if (msg == NULL) { |
828 | err = -ENOMEM; | |
829 | goto out; | |
830 | } | |
831 | ||
832 | err = xb_read(&msg->hdr, sizeof(msg->hdr)); | |
833 | if (err) { | |
834 | kfree(msg); | |
835 | goto out; | |
836 | } | |
837 | ||
9e7860ce IC |
838 | if (msg->hdr.len > XENSTORE_PAYLOAD_MAX) { |
839 | kfree(msg); | |
840 | err = -EINVAL; | |
841 | goto out; | |
842 | } | |
843 | ||
a144ff09 | 844 | body = kmalloc(msg->hdr.len + 1, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
845 | if (body == NULL) { |
846 | kfree(msg); | |
847 | err = -ENOMEM; | |
848 | goto out; | |
849 | } | |
850 | ||
851 | err = xb_read(body, msg->hdr.len); | |
852 | if (err) { | |
853 | kfree(body); | |
854 | kfree(msg); | |
855 | goto out; | |
856 | } | |
857 | body[msg->hdr.len] = '\0'; | |
858 | ||
859 | if (msg->hdr.type == XS_WATCH_EVENT) { | |
860 | msg->u.watch.vec = split(body, msg->hdr.len, | |
861 | &msg->u.watch.vec_size); | |
862 | if (IS_ERR(msg->u.watch.vec)) { | |
4bac07c9 | 863 | err = PTR_ERR(msg->u.watch.vec); |
98ac0e53 | 864 | kfree(msg); |
4bac07c9 JF |
865 | goto out; |
866 | } | |
867 | ||
868 | spin_lock(&watches_lock); | |
869 | msg->u.watch.handle = find_watch( | |
870 | msg->u.watch.vec[XS_WATCH_TOKEN]); | |
871 | if (msg->u.watch.handle != NULL) { | |
872 | spin_lock(&watch_events_lock); | |
873 | list_add_tail(&msg->list, &watch_events); | |
874 | wake_up(&watch_events_waitq); | |
875 | spin_unlock(&watch_events_lock); | |
876 | } else { | |
877 | kfree(msg->u.watch.vec); | |
878 | kfree(msg); | |
879 | } | |
880 | spin_unlock(&watches_lock); | |
881 | } else { | |
882 | msg->u.reply.body = body; | |
883 | spin_lock(&xs_state.reply_lock); | |
884 | list_add_tail(&msg->list, &xs_state.reply_list); | |
885 | spin_unlock(&xs_state.reply_lock); | |
886 | wake_up(&xs_state.reply_waitq); | |
887 | } | |
888 | ||
889 | out: | |
890 | mutex_unlock(&xs_state.response_mutex); | |
891 | return err; | |
892 | } | |
893 | ||
894 | static int xenbus_thread(void *unused) | |
895 | { | |
896 | int err; | |
897 | ||
898 | for (;;) { | |
899 | err = process_msg(); | |
900 | if (err) | |
283c0972 | 901 | pr_warn("error %d while reading message\n", err); |
4bac07c9 JF |
902 | if (kthread_should_stop()) |
903 | break; | |
904 | } | |
905 | ||
906 | return 0; | |
907 | } | |
908 | ||
909 | int xs_init(void) | |
910 | { | |
911 | int err; | |
912 | struct task_struct *task; | |
913 | ||
914 | INIT_LIST_HEAD(&xs_state.reply_list); | |
915 | spin_lock_init(&xs_state.reply_lock); | |
916 | init_waitqueue_head(&xs_state.reply_waitq); | |
917 | ||
918 | mutex_init(&xs_state.request_mutex); | |
919 | mutex_init(&xs_state.response_mutex); | |
4c31a781 | 920 | mutex_init(&xs_state.transaction_mutex); |
4bac07c9 | 921 | init_rwsem(&xs_state.watch_mutex); |
4c31a781 IC |
922 | atomic_set(&xs_state.transaction_count, 0); |
923 | init_waitqueue_head(&xs_state.transaction_wq); | |
4bac07c9 JF |
924 | |
925 | /* Initialize the shared memory rings to talk to xenstored */ | |
926 | err = xb_init_comms(); | |
927 | if (err) | |
928 | return err; | |
929 | ||
930 | task = kthread_run(xenwatch_thread, NULL, "xenwatch"); | |
931 | if (IS_ERR(task)) | |
932 | return PTR_ERR(task); | |
933 | xenwatch_pid = task->pid; | |
934 | ||
935 | task = kthread_run(xenbus_thread, NULL, "xenbus"); | |
936 | if (IS_ERR(task)) | |
937 | return PTR_ERR(task); | |
938 | ||
254d1a3f OH |
939 | /* shutdown watches for kexec boot */ |
940 | xs_reset_watches(); | |
941 | ||
4bac07c9 JF |
942 | return 0; |
943 | } |