]>
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> | |
4bac07c9 | 47 | #include <linux/mutex.h> |
ecc635f9 | 48 | #include <asm/xen/hypervisor.h> |
4bac07c9 | 49 | #include <xen/xenbus.h> |
5b25d89e | 50 | #include <xen/xen.h> |
332f791d | 51 | #include "xenbus.h" |
4bac07c9 JF |
52 | |
53 | struct xs_stored_msg { | |
54 | struct list_head list; | |
55 | ||
56 | struct xsd_sockmsg hdr; | |
57 | ||
58 | union { | |
59 | /* Queued replies. */ | |
60 | struct { | |
61 | char *body; | |
62 | } reply; | |
63 | ||
64 | /* Queued watch events. */ | |
65 | struct { | |
66 | struct xenbus_watch *handle; | |
5584ea25 JG |
67 | const char *path; |
68 | const char *token; | |
4bac07c9 JF |
69 | } watch; |
70 | } u; | |
71 | }; | |
72 | ||
73 | struct xs_handle { | |
74 | /* A list of replies. Currently only one will ever be outstanding. */ | |
75 | struct list_head reply_list; | |
76 | spinlock_t reply_lock; | |
77 | wait_queue_head_t reply_waitq; | |
78 | ||
79 | /* | |
80 | * Mutex ordering: transaction_mutex -> watch_mutex -> request_mutex. | |
81 | * response_mutex is never taken simultaneously with the other three. | |
4c31a781 IC |
82 | * |
83 | * transaction_mutex must be held before incrementing | |
84 | * transaction_count. The mutex is held when a suspend is in | |
85 | * progress to prevent new transactions starting. | |
86 | * | |
87 | * When decrementing transaction_count to zero the wait queue | |
88 | * should be woken up, the suspend code waits for count to | |
89 | * reach zero. | |
4bac07c9 JF |
90 | */ |
91 | ||
92 | /* One request at a time. */ | |
93 | struct mutex request_mutex; | |
94 | ||
95 | /* Protect xenbus reader thread against save/restore. */ | |
96 | struct mutex response_mutex; | |
97 | ||
98 | /* Protect transactions against save/restore. */ | |
4c31a781 IC |
99 | struct mutex transaction_mutex; |
100 | atomic_t transaction_count; | |
101 | wait_queue_head_t transaction_wq; | |
4bac07c9 JF |
102 | |
103 | /* Protect watch (de)register against save/restore. */ | |
104 | struct rw_semaphore watch_mutex; | |
105 | }; | |
106 | ||
107 | static struct xs_handle xs_state; | |
108 | ||
109 | /* List of registered watches, and a lock to protect it. */ | |
110 | static LIST_HEAD(watches); | |
111 | static DEFINE_SPINLOCK(watches_lock); | |
112 | ||
113 | /* List of pending watch callback events, and a lock to protect it. */ | |
114 | static LIST_HEAD(watch_events); | |
115 | static DEFINE_SPINLOCK(watch_events_lock); | |
116 | ||
117 | /* | |
118 | * Details of the xenwatch callback kernel thread. The thread waits on the | |
119 | * watch_events_waitq for work to do (queued on watch_events list). When it | |
120 | * wakes up it acquires the xenwatch_mutex before reading the list and | |
121 | * carrying out work. | |
122 | */ | |
123 | static pid_t xenwatch_pid; | |
124 | static DEFINE_MUTEX(xenwatch_mutex); | |
125 | static DECLARE_WAIT_QUEUE_HEAD(watch_events_waitq); | |
126 | ||
127 | static int get_error(const char *errorstring) | |
128 | { | |
129 | unsigned int i; | |
130 | ||
131 | for (i = 0; strcmp(errorstring, xsd_errors[i].errstring) != 0; i++) { | |
132 | if (i == ARRAY_SIZE(xsd_errors) - 1) { | |
283c0972 JP |
133 | pr_warn("xen store gave: unknown error %s\n", |
134 | errorstring); | |
4bac07c9 JF |
135 | return EINVAL; |
136 | } | |
137 | } | |
138 | return xsd_errors[i].errnum; | |
139 | } | |
140 | ||
027bd7e8 KRW |
141 | static bool xenbus_ok(void) |
142 | { | |
143 | switch (xen_store_domain_type) { | |
144 | case XS_LOCAL: | |
145 | switch (system_state) { | |
146 | case SYSTEM_POWER_OFF: | |
147 | case SYSTEM_RESTART: | |
148 | case SYSTEM_HALT: | |
149 | return false; | |
150 | default: | |
151 | break; | |
152 | } | |
153 | return true; | |
154 | case XS_PV: | |
155 | case XS_HVM: | |
156 | /* FIXME: Could check that the remote domain is alive, | |
157 | * but it is normally initial domain. */ | |
158 | return true; | |
159 | default: | |
160 | break; | |
161 | } | |
162 | return false; | |
163 | } | |
4bac07c9 JF |
164 | static void *read_reply(enum xsd_sockmsg_type *type, unsigned int *len) |
165 | { | |
166 | struct xs_stored_msg *msg; | |
167 | char *body; | |
168 | ||
169 | spin_lock(&xs_state.reply_lock); | |
170 | ||
171 | while (list_empty(&xs_state.reply_list)) { | |
172 | spin_unlock(&xs_state.reply_lock); | |
027bd7e8 KRW |
173 | if (xenbus_ok()) |
174 | /* XXX FIXME: Avoid synchronous wait for response here. */ | |
175 | wait_event_timeout(xs_state.reply_waitq, | |
176 | !list_empty(&xs_state.reply_list), | |
177 | msecs_to_jiffies(500)); | |
178 | else { | |
179 | /* | |
180 | * If we are in the process of being shut-down there is | |
181 | * no point of trying to contact XenBus - it is either | |
182 | * killed (xenstored application) or the other domain | |
183 | * has been killed or is unreachable. | |
184 | */ | |
185 | return ERR_PTR(-EIO); | |
186 | } | |
4bac07c9 JF |
187 | spin_lock(&xs_state.reply_lock); |
188 | } | |
189 | ||
190 | msg = list_entry(xs_state.reply_list.next, | |
191 | struct xs_stored_msg, list); | |
192 | list_del(&msg->list); | |
193 | ||
194 | spin_unlock(&xs_state.reply_lock); | |
195 | ||
196 | *type = msg->hdr.type; | |
197 | if (len) | |
198 | *len = msg->hdr.len; | |
199 | body = msg->u.reply.body; | |
200 | ||
201 | kfree(msg); | |
202 | ||
203 | return body; | |
204 | } | |
205 | ||
4c31a781 IC |
206 | static void transaction_start(void) |
207 | { | |
208 | mutex_lock(&xs_state.transaction_mutex); | |
209 | atomic_inc(&xs_state.transaction_count); | |
210 | mutex_unlock(&xs_state.transaction_mutex); | |
211 | } | |
212 | ||
213 | static void transaction_end(void) | |
214 | { | |
215 | if (atomic_dec_and_test(&xs_state.transaction_count)) | |
216 | wake_up(&xs_state.transaction_wq); | |
217 | } | |
218 | ||
219 | static void transaction_suspend(void) | |
220 | { | |
221 | mutex_lock(&xs_state.transaction_mutex); | |
222 | wait_event(xs_state.transaction_wq, | |
223 | atomic_read(&xs_state.transaction_count) == 0); | |
224 | } | |
225 | ||
226 | static void transaction_resume(void) | |
227 | { | |
228 | mutex_unlock(&xs_state.transaction_mutex); | |
229 | } | |
230 | ||
4bac07c9 JF |
231 | void *xenbus_dev_request_and_reply(struct xsd_sockmsg *msg) |
232 | { | |
233 | void *ret; | |
e5a79475 | 234 | enum xsd_sockmsg_type type = msg->type; |
4bac07c9 JF |
235 | int err; |
236 | ||
e5a79475 | 237 | if (type == XS_TRANSACTION_START) |
4c31a781 | 238 | transaction_start(); |
4bac07c9 JF |
239 | |
240 | mutex_lock(&xs_state.request_mutex); | |
241 | ||
242 | err = xb_write(msg, sizeof(*msg) + msg->len); | |
243 | if (err) { | |
244 | msg->type = XS_ERROR; | |
245 | ret = ERR_PTR(err); | |
246 | } else | |
247 | ret = read_reply(&msg->type, &msg->len); | |
248 | ||
249 | mutex_unlock(&xs_state.request_mutex); | |
250 | ||
251 | if ((msg->type == XS_TRANSACTION_END) || | |
e5a79475 | 252 | ((type == XS_TRANSACTION_START) && (msg->type == XS_ERROR))) |
4c31a781 | 253 | transaction_end(); |
4bac07c9 JF |
254 | |
255 | return ret; | |
256 | } | |
1107ba88 | 257 | EXPORT_SYMBOL(xenbus_dev_request_and_reply); |
4bac07c9 JF |
258 | |
259 | /* Send message to xs, get kmalloc'ed reply. ERR_PTR() on error. */ | |
260 | static void *xs_talkv(struct xenbus_transaction t, | |
261 | enum xsd_sockmsg_type type, | |
262 | const struct kvec *iovec, | |
263 | unsigned int num_vecs, | |
264 | unsigned int *len) | |
265 | { | |
266 | struct xsd_sockmsg msg; | |
267 | void *ret = NULL; | |
268 | unsigned int i; | |
269 | int err; | |
270 | ||
271 | msg.tx_id = t.id; | |
272 | msg.req_id = 0; | |
273 | msg.type = type; | |
274 | msg.len = 0; | |
275 | for (i = 0; i < num_vecs; i++) | |
276 | msg.len += iovec[i].iov_len; | |
277 | ||
278 | mutex_lock(&xs_state.request_mutex); | |
279 | ||
280 | err = xb_write(&msg, sizeof(msg)); | |
281 | if (err) { | |
282 | mutex_unlock(&xs_state.request_mutex); | |
283 | return ERR_PTR(err); | |
284 | } | |
285 | ||
286 | for (i = 0; i < num_vecs; i++) { | |
287 | err = xb_write(iovec[i].iov_base, iovec[i].iov_len); | |
288 | if (err) { | |
289 | mutex_unlock(&xs_state.request_mutex); | |
290 | return ERR_PTR(err); | |
291 | } | |
292 | } | |
293 | ||
294 | ret = read_reply(&msg.type, len); | |
295 | ||
296 | mutex_unlock(&xs_state.request_mutex); | |
297 | ||
298 | if (IS_ERR(ret)) | |
299 | return ret; | |
300 | ||
301 | if (msg.type == XS_ERROR) { | |
302 | err = get_error(ret); | |
303 | kfree(ret); | |
304 | return ERR_PTR(-err); | |
305 | } | |
306 | ||
307 | if (msg.type != type) { | |
283c0972 JP |
308 | pr_warn_ratelimited("unexpected type [%d], expected [%d]\n", |
309 | msg.type, type); | |
4bac07c9 JF |
310 | kfree(ret); |
311 | return ERR_PTR(-EINVAL); | |
312 | } | |
313 | return ret; | |
314 | } | |
315 | ||
316 | /* Simplified version of xs_talkv: single message. */ | |
317 | static void *xs_single(struct xenbus_transaction t, | |
318 | enum xsd_sockmsg_type type, | |
319 | const char *string, | |
320 | unsigned int *len) | |
321 | { | |
322 | struct kvec iovec; | |
323 | ||
324 | iovec.iov_base = (void *)string; | |
325 | iovec.iov_len = strlen(string) + 1; | |
326 | return xs_talkv(t, type, &iovec, 1, len); | |
327 | } | |
328 | ||
329 | /* Many commands only need an ack, don't care what it says. */ | |
330 | static int xs_error(char *reply) | |
331 | { | |
332 | if (IS_ERR(reply)) | |
333 | return PTR_ERR(reply); | |
334 | kfree(reply); | |
335 | return 0; | |
336 | } | |
337 | ||
338 | static unsigned int count_strings(const char *strings, unsigned int len) | |
339 | { | |
340 | unsigned int num; | |
341 | const char *p; | |
342 | ||
343 | for (p = strings, num = 0; p < strings + len; p += strlen(p) + 1) | |
344 | num++; | |
345 | ||
346 | return num; | |
347 | } | |
348 | ||
349 | /* Return the path to dir with /name appended. Buffer must be kfree()'ed. */ | |
350 | static char *join(const char *dir, const char *name) | |
351 | { | |
352 | char *buffer; | |
353 | ||
354 | if (strlen(name) == 0) | |
a144ff09 | 355 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s", dir); |
4bac07c9 | 356 | else |
a144ff09 | 357 | buffer = kasprintf(GFP_NOIO | __GFP_HIGH, "%s/%s", dir, name); |
4bac07c9 JF |
358 | return (!buffer) ? ERR_PTR(-ENOMEM) : buffer; |
359 | } | |
360 | ||
361 | static char **split(char *strings, unsigned int len, unsigned int *num) | |
362 | { | |
363 | char *p, **ret; | |
364 | ||
365 | /* Count the strings. */ | |
366 | *num = count_strings(strings, len); | |
367 | ||
368 | /* Transfer to one big alloc for easy freeing. */ | |
a144ff09 | 369 | ret = kmalloc(*num * sizeof(char *) + len, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
370 | if (!ret) { |
371 | kfree(strings); | |
372 | return ERR_PTR(-ENOMEM); | |
373 | } | |
374 | memcpy(&ret[*num], strings, len); | |
375 | kfree(strings); | |
376 | ||
377 | strings = (char *)&ret[*num]; | |
378 | for (p = strings, *num = 0; p < strings + len; p += strlen(p) + 1) | |
379 | ret[(*num)++] = p; | |
380 | ||
381 | return ret; | |
382 | } | |
383 | ||
384 | char **xenbus_directory(struct xenbus_transaction t, | |
385 | const char *dir, const char *node, unsigned int *num) | |
386 | { | |
387 | char *strings, *path; | |
388 | unsigned int len; | |
389 | ||
390 | path = join(dir, node); | |
391 | if (IS_ERR(path)) | |
392 | return (char **)path; | |
393 | ||
394 | strings = xs_single(t, XS_DIRECTORY, path, &len); | |
395 | kfree(path); | |
396 | if (IS_ERR(strings)) | |
397 | return (char **)strings; | |
398 | ||
399 | return split(strings, len, num); | |
400 | } | |
401 | EXPORT_SYMBOL_GPL(xenbus_directory); | |
402 | ||
403 | /* Check if a path exists. Return 1 if it does. */ | |
404 | int xenbus_exists(struct xenbus_transaction t, | |
405 | const char *dir, const char *node) | |
406 | { | |
407 | char **d; | |
408 | int dir_n; | |
409 | ||
410 | d = xenbus_directory(t, dir, node, &dir_n); | |
411 | if (IS_ERR(d)) | |
412 | return 0; | |
413 | kfree(d); | |
414 | return 1; | |
415 | } | |
416 | EXPORT_SYMBOL_GPL(xenbus_exists); | |
417 | ||
418 | /* Get the value of a single file. | |
419 | * Returns a kmalloced value: call free() on it after use. | |
420 | * len indicates length in bytes. | |
421 | */ | |
422 | void *xenbus_read(struct xenbus_transaction t, | |
423 | const char *dir, const char *node, unsigned int *len) | |
424 | { | |
425 | char *path; | |
426 | void *ret; | |
427 | ||
428 | path = join(dir, node); | |
429 | if (IS_ERR(path)) | |
430 | return (void *)path; | |
431 | ||
432 | ret = xs_single(t, XS_READ, path, len); | |
433 | kfree(path); | |
434 | return ret; | |
435 | } | |
436 | EXPORT_SYMBOL_GPL(xenbus_read); | |
437 | ||
438 | /* Write the value of a single file. | |
439 | * Returns -err on failure. | |
440 | */ | |
441 | int xenbus_write(struct xenbus_transaction t, | |
442 | const char *dir, const char *node, const char *string) | |
443 | { | |
444 | const char *path; | |
445 | struct kvec iovec[2]; | |
446 | int ret; | |
447 | ||
448 | path = join(dir, node); | |
449 | if (IS_ERR(path)) | |
450 | return PTR_ERR(path); | |
451 | ||
452 | iovec[0].iov_base = (void *)path; | |
453 | iovec[0].iov_len = strlen(path) + 1; | |
454 | iovec[1].iov_base = (void *)string; | |
455 | iovec[1].iov_len = strlen(string); | |
456 | ||
457 | ret = xs_error(xs_talkv(t, XS_WRITE, iovec, ARRAY_SIZE(iovec), NULL)); | |
458 | kfree(path); | |
459 | return ret; | |
460 | } | |
461 | EXPORT_SYMBOL_GPL(xenbus_write); | |
462 | ||
463 | /* Create a new directory. */ | |
464 | int xenbus_mkdir(struct xenbus_transaction t, | |
465 | const char *dir, const char *node) | |
466 | { | |
467 | char *path; | |
468 | int ret; | |
469 | ||
470 | path = join(dir, node); | |
471 | if (IS_ERR(path)) | |
472 | return PTR_ERR(path); | |
473 | ||
474 | ret = xs_error(xs_single(t, XS_MKDIR, path, NULL)); | |
475 | kfree(path); | |
476 | return ret; | |
477 | } | |
478 | EXPORT_SYMBOL_GPL(xenbus_mkdir); | |
479 | ||
480 | /* Destroy a file or directory (directories must be empty). */ | |
481 | int xenbus_rm(struct xenbus_transaction t, const char *dir, const char *node) | |
482 | { | |
483 | char *path; | |
484 | int ret; | |
485 | ||
486 | path = join(dir, node); | |
487 | if (IS_ERR(path)) | |
488 | return PTR_ERR(path); | |
489 | ||
490 | ret = xs_error(xs_single(t, XS_RM, path, NULL)); | |
491 | kfree(path); | |
492 | return ret; | |
493 | } | |
494 | EXPORT_SYMBOL_GPL(xenbus_rm); | |
495 | ||
496 | /* Start a transaction: changes by others will not be seen during this | |
497 | * transaction, and changes will not be visible to others until end. | |
498 | */ | |
499 | int xenbus_transaction_start(struct xenbus_transaction *t) | |
500 | { | |
501 | char *id_str; | |
502 | ||
4c31a781 | 503 | transaction_start(); |
4bac07c9 JF |
504 | |
505 | id_str = xs_single(XBT_NIL, XS_TRANSACTION_START, "", NULL); | |
506 | if (IS_ERR(id_str)) { | |
4c31a781 | 507 | transaction_end(); |
4bac07c9 JF |
508 | return PTR_ERR(id_str); |
509 | } | |
510 | ||
511 | t->id = simple_strtoul(id_str, NULL, 0); | |
512 | kfree(id_str); | |
513 | return 0; | |
514 | } | |
515 | EXPORT_SYMBOL_GPL(xenbus_transaction_start); | |
516 | ||
517 | /* End a transaction. | |
518 | * If abandon is true, transaction is discarded instead of committed. | |
519 | */ | |
520 | int xenbus_transaction_end(struct xenbus_transaction t, int abort) | |
521 | { | |
522 | char abortstr[2]; | |
523 | int err; | |
524 | ||
525 | if (abort) | |
526 | strcpy(abortstr, "F"); | |
527 | else | |
528 | strcpy(abortstr, "T"); | |
529 | ||
530 | err = xs_error(xs_single(t, XS_TRANSACTION_END, abortstr, NULL)); | |
531 | ||
4c31a781 | 532 | transaction_end(); |
4bac07c9 JF |
533 | |
534 | return err; | |
535 | } | |
536 | EXPORT_SYMBOL_GPL(xenbus_transaction_end); | |
537 | ||
538 | /* Single read and scanf: returns -errno or num scanned. */ | |
539 | int xenbus_scanf(struct xenbus_transaction t, | |
540 | const char *dir, const char *node, const char *fmt, ...) | |
541 | { | |
542 | va_list ap; | |
543 | int ret; | |
544 | char *val; | |
545 | ||
546 | val = xenbus_read(t, dir, node, NULL); | |
547 | if (IS_ERR(val)) | |
548 | return PTR_ERR(val); | |
549 | ||
550 | va_start(ap, fmt); | |
551 | ret = vsscanf(val, fmt, ap); | |
552 | va_end(ap); | |
553 | kfree(val); | |
554 | /* Distinctive errno. */ | |
555 | if (ret == 0) | |
556 | return -ERANGE; | |
557 | return ret; | |
558 | } | |
559 | EXPORT_SYMBOL_GPL(xenbus_scanf); | |
560 | ||
9c53a179 JG |
561 | /* Read an (optional) unsigned value. */ |
562 | unsigned int xenbus_read_unsigned(const char *dir, const char *node, | |
563 | unsigned int default_val) | |
564 | { | |
565 | unsigned int val; | |
566 | int ret; | |
567 | ||
568 | ret = xenbus_scanf(XBT_NIL, dir, node, "%u", &val); | |
569 | if (ret <= 0) | |
570 | val = default_val; | |
571 | ||
572 | return val; | |
573 | } | |
574 | EXPORT_SYMBOL_GPL(xenbus_read_unsigned); | |
575 | ||
4bac07c9 JF |
576 | /* Single printf and write: returns -errno or 0. */ |
577 | int xenbus_printf(struct xenbus_transaction t, | |
578 | const char *dir, const char *node, const char *fmt, ...) | |
579 | { | |
580 | va_list ap; | |
581 | int ret; | |
a800651e | 582 | char *buf; |
4bac07c9 JF |
583 | |
584 | va_start(ap, fmt); | |
a800651e | 585 | buf = kvasprintf(GFP_NOIO | __GFP_HIGH, fmt, ap); |
4bac07c9 JF |
586 | va_end(ap); |
587 | ||
a800651e IC |
588 | if (!buf) |
589 | return -ENOMEM; | |
590 | ||
591 | ret = xenbus_write(t, dir, node, buf); | |
4bac07c9 | 592 | |
a800651e | 593 | kfree(buf); |
4bac07c9 JF |
594 | |
595 | return ret; | |
596 | } | |
597 | EXPORT_SYMBOL_GPL(xenbus_printf); | |
598 | ||
599 | /* Takes tuples of names, scanf-style args, and void **, NULL terminated. */ | |
600 | int xenbus_gather(struct xenbus_transaction t, const char *dir, ...) | |
601 | { | |
602 | va_list ap; | |
603 | const char *name; | |
604 | int ret = 0; | |
605 | ||
606 | va_start(ap, dir); | |
607 | while (ret == 0 && (name = va_arg(ap, char *)) != NULL) { | |
608 | const char *fmt = va_arg(ap, char *); | |
609 | void *result = va_arg(ap, void *); | |
610 | char *p; | |
611 | ||
612 | p = xenbus_read(t, dir, name, NULL); | |
613 | if (IS_ERR(p)) { | |
614 | ret = PTR_ERR(p); | |
615 | break; | |
616 | } | |
617 | if (fmt) { | |
618 | if (sscanf(p, fmt, result) == 0) | |
619 | ret = -EINVAL; | |
620 | kfree(p); | |
621 | } else | |
622 | *(char **)result = p; | |
623 | } | |
624 | va_end(ap); | |
625 | return ret; | |
626 | } | |
627 | EXPORT_SYMBOL_GPL(xenbus_gather); | |
628 | ||
629 | static int xs_watch(const char *path, const char *token) | |
630 | { | |
631 | struct kvec iov[2]; | |
632 | ||
633 | iov[0].iov_base = (void *)path; | |
634 | iov[0].iov_len = strlen(path) + 1; | |
635 | iov[1].iov_base = (void *)token; | |
636 | iov[1].iov_len = strlen(token) + 1; | |
637 | ||
638 | return xs_error(xs_talkv(XBT_NIL, XS_WATCH, iov, | |
639 | ARRAY_SIZE(iov), NULL)); | |
640 | } | |
641 | ||
642 | static int xs_unwatch(const char *path, const char *token) | |
643 | { | |
644 | struct kvec iov[2]; | |
645 | ||
646 | iov[0].iov_base = (char *)path; | |
647 | iov[0].iov_len = strlen(path) + 1; | |
648 | iov[1].iov_base = (char *)token; | |
649 | iov[1].iov_len = strlen(token) + 1; | |
650 | ||
651 | return xs_error(xs_talkv(XBT_NIL, XS_UNWATCH, iov, | |
652 | ARRAY_SIZE(iov), NULL)); | |
653 | } | |
654 | ||
655 | static struct xenbus_watch *find_watch(const char *token) | |
656 | { | |
657 | struct xenbus_watch *i, *cmp; | |
658 | ||
659 | cmp = (void *)simple_strtoul(token, NULL, 16); | |
660 | ||
661 | list_for_each_entry(i, &watches, list) | |
662 | if (i == cmp) | |
663 | return i; | |
664 | ||
665 | return NULL; | |
666 | } | |
cb6b6df1 KRW |
667 | /* |
668 | * Certain older XenBus toolstack cannot handle reading values that are | |
669 | * not populated. Some Xen 3.4 installation are incapable of doing this | |
670 | * so if we are running on anything older than 4 do not attempt to read | |
671 | * control/platform-feature-xs_reset_watches. | |
672 | */ | |
e9d1aa05 | 673 | static bool xen_strict_xenbus_quirk(void) |
cb6b6df1 | 674 | { |
7644bdac | 675 | #ifdef CONFIG_X86 |
cb6b6df1 KRW |
676 | uint32_t eax, ebx, ecx, edx, base; |
677 | ||
678 | base = xen_cpuid_base(); | |
679 | cpuid(base + 1, &eax, &ebx, &ecx, &edx); | |
4bac07c9 | 680 | |
cb6b6df1 KRW |
681 | if ((eax >> 16) < 4) |
682 | return true; | |
7644bdac | 683 | #endif |
cb6b6df1 KRW |
684 | return false; |
685 | ||
686 | } | |
254d1a3f OH |
687 | static void xs_reset_watches(void) |
688 | { | |
999c9af9 | 689 | int err; |
254d1a3f | 690 | |
ecc635f9 | 691 | if (!xen_hvm_domain() || xen_initial_domain()) |
254d1a3f OH |
692 | return; |
693 | ||
cb6b6df1 KRW |
694 | if (xen_strict_xenbus_quirk()) |
695 | return; | |
696 | ||
999c9af9 JG |
697 | if (!xenbus_read_unsigned("control", |
698 | "platform-feature-xs_reset_watches", 0)) | |
254d1a3f OH |
699 | return; |
700 | ||
701 | err = xs_error(xs_single(XBT_NIL, XS_RESET_WATCHES, "", NULL)); | |
702 | if (err && err != -EEXIST) | |
283c0972 | 703 | pr_warn("xs_reset_watches failed: %d\n", err); |
254d1a3f OH |
704 | } |
705 | ||
4bac07c9 JF |
706 | /* Register callback to watch this node. */ |
707 | int register_xenbus_watch(struct xenbus_watch *watch) | |
708 | { | |
709 | /* Pointer in ascii is the token. */ | |
710 | char token[sizeof(watch) * 2 + 1]; | |
711 | int err; | |
712 | ||
713 | sprintf(token, "%lX", (long)watch); | |
714 | ||
715 | down_read(&xs_state.watch_mutex); | |
716 | ||
717 | spin_lock(&watches_lock); | |
718 | BUG_ON(find_watch(token)); | |
719 | list_add(&watch->list, &watches); | |
720 | spin_unlock(&watches_lock); | |
721 | ||
722 | err = xs_watch(watch->node, token); | |
723 | ||
c4c303c7 | 724 | if (err) { |
4bac07c9 JF |
725 | spin_lock(&watches_lock); |
726 | list_del(&watch->list); | |
727 | spin_unlock(&watches_lock); | |
728 | } | |
729 | ||
730 | up_read(&xs_state.watch_mutex); | |
731 | ||
732 | return err; | |
733 | } | |
734 | EXPORT_SYMBOL_GPL(register_xenbus_watch); | |
735 | ||
736 | void unregister_xenbus_watch(struct xenbus_watch *watch) | |
737 | { | |
738 | struct xs_stored_msg *msg, *tmp; | |
739 | char token[sizeof(watch) * 2 + 1]; | |
740 | int err; | |
741 | ||
742 | sprintf(token, "%lX", (long)watch); | |
743 | ||
744 | down_read(&xs_state.watch_mutex); | |
745 | ||
746 | spin_lock(&watches_lock); | |
747 | BUG_ON(!find_watch(token)); | |
748 | list_del(&watch->list); | |
749 | spin_unlock(&watches_lock); | |
750 | ||
751 | err = xs_unwatch(watch->node, token); | |
752 | if (err) | |
283c0972 | 753 | pr_warn("Failed to release watch %s: %i\n", watch->node, err); |
4bac07c9 JF |
754 | |
755 | up_read(&xs_state.watch_mutex); | |
756 | ||
757 | /* Make sure there are no callbacks running currently (unless | |
758 | its us) */ | |
759 | if (current->pid != xenwatch_pid) | |
760 | mutex_lock(&xenwatch_mutex); | |
761 | ||
762 | /* Cancel pending watch events. */ | |
763 | spin_lock(&watch_events_lock); | |
764 | list_for_each_entry_safe(msg, tmp, &watch_events, list) { | |
765 | if (msg->u.watch.handle != watch) | |
766 | continue; | |
767 | list_del(&msg->list); | |
5584ea25 | 768 | kfree(msg->u.watch.path); |
4bac07c9 JF |
769 | kfree(msg); |
770 | } | |
771 | spin_unlock(&watch_events_lock); | |
772 | ||
773 | if (current->pid != xenwatch_pid) | |
774 | mutex_unlock(&xenwatch_mutex); | |
775 | } | |
776 | EXPORT_SYMBOL_GPL(unregister_xenbus_watch); | |
777 | ||
778 | void xs_suspend(void) | |
779 | { | |
4c31a781 | 780 | transaction_suspend(); |
4bac07c9 JF |
781 | down_write(&xs_state.watch_mutex); |
782 | mutex_lock(&xs_state.request_mutex); | |
783 | mutex_lock(&xs_state.response_mutex); | |
784 | } | |
785 | ||
786 | void xs_resume(void) | |
787 | { | |
788 | struct xenbus_watch *watch; | |
789 | char token[sizeof(watch) * 2 + 1]; | |
790 | ||
de5b31bd IC |
791 | xb_init_comms(); |
792 | ||
4bac07c9 JF |
793 | mutex_unlock(&xs_state.response_mutex); |
794 | mutex_unlock(&xs_state.request_mutex); | |
4c31a781 | 795 | transaction_resume(); |
4bac07c9 JF |
796 | |
797 | /* No need for watches_lock: the watch_mutex is sufficient. */ | |
798 | list_for_each_entry(watch, &watches, list) { | |
799 | sprintf(token, "%lX", (long)watch); | |
800 | xs_watch(watch->node, token); | |
801 | } | |
802 | ||
803 | up_write(&xs_state.watch_mutex); | |
804 | } | |
805 | ||
806 | void xs_suspend_cancel(void) | |
807 | { | |
808 | mutex_unlock(&xs_state.response_mutex); | |
809 | mutex_unlock(&xs_state.request_mutex); | |
810 | up_write(&xs_state.watch_mutex); | |
4c31a781 | 811 | mutex_unlock(&xs_state.transaction_mutex); |
4bac07c9 JF |
812 | } |
813 | ||
814 | static int xenwatch_thread(void *unused) | |
815 | { | |
816 | struct list_head *ent; | |
817 | struct xs_stored_msg *msg; | |
818 | ||
819 | for (;;) { | |
820 | wait_event_interruptible(watch_events_waitq, | |
821 | !list_empty(&watch_events)); | |
822 | ||
823 | if (kthread_should_stop()) | |
824 | break; | |
825 | ||
826 | mutex_lock(&xenwatch_mutex); | |
827 | ||
828 | spin_lock(&watch_events_lock); | |
829 | ent = watch_events.next; | |
830 | if (ent != &watch_events) | |
831 | list_del(ent); | |
832 | spin_unlock(&watch_events_lock); | |
833 | ||
834 | if (ent != &watch_events) { | |
835 | msg = list_entry(ent, struct xs_stored_msg, list); | |
5584ea25 JG |
836 | msg->u.watch.handle->callback(msg->u.watch.handle, |
837 | msg->u.watch.path, | |
838 | msg->u.watch.token); | |
839 | kfree(msg->u.watch.path); | |
4bac07c9 JF |
840 | kfree(msg); |
841 | } | |
842 | ||
843 | mutex_unlock(&xenwatch_mutex); | |
844 | } | |
845 | ||
846 | return 0; | |
847 | } | |
848 | ||
849 | static int process_msg(void) | |
850 | { | |
851 | struct xs_stored_msg *msg; | |
852 | char *body; | |
853 | int err; | |
854 | ||
855 | /* | |
856 | * We must disallow save/restore while reading a xenstore message. | |
857 | * A partial read across s/r leaves us out of sync with xenstored. | |
858 | */ | |
859 | for (;;) { | |
860 | err = xb_wait_for_data_to_read(); | |
861 | if (err) | |
862 | return err; | |
863 | mutex_lock(&xs_state.response_mutex); | |
864 | if (xb_data_to_read()) | |
865 | break; | |
866 | /* We raced with save/restore: pending data 'disappeared'. */ | |
867 | mutex_unlock(&xs_state.response_mutex); | |
868 | } | |
869 | ||
870 | ||
a144ff09 | 871 | msg = kmalloc(sizeof(*msg), GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
872 | if (msg == NULL) { |
873 | err = -ENOMEM; | |
874 | goto out; | |
875 | } | |
876 | ||
877 | err = xb_read(&msg->hdr, sizeof(msg->hdr)); | |
878 | if (err) { | |
879 | kfree(msg); | |
880 | goto out; | |
881 | } | |
882 | ||
9e7860ce IC |
883 | if (msg->hdr.len > XENSTORE_PAYLOAD_MAX) { |
884 | kfree(msg); | |
885 | err = -EINVAL; | |
886 | goto out; | |
887 | } | |
888 | ||
a144ff09 | 889 | body = kmalloc(msg->hdr.len + 1, GFP_NOIO | __GFP_HIGH); |
4bac07c9 JF |
890 | if (body == NULL) { |
891 | kfree(msg); | |
892 | err = -ENOMEM; | |
893 | goto out; | |
894 | } | |
895 | ||
896 | err = xb_read(body, msg->hdr.len); | |
897 | if (err) { | |
898 | kfree(body); | |
899 | kfree(msg); | |
900 | goto out; | |
901 | } | |
902 | body[msg->hdr.len] = '\0'; | |
903 | ||
904 | if (msg->hdr.type == XS_WATCH_EVENT) { | |
5584ea25 JG |
905 | if (count_strings(body, msg->hdr.len) != 2) { |
906 | err = -EINVAL; | |
98ac0e53 | 907 | kfree(msg); |
5584ea25 | 908 | kfree(body); |
4bac07c9 JF |
909 | goto out; |
910 | } | |
5584ea25 JG |
911 | msg->u.watch.path = (const char *)body; |
912 | msg->u.watch.token = (const char *)strchr(body, '\0') + 1; | |
4bac07c9 JF |
913 | |
914 | spin_lock(&watches_lock); | |
5584ea25 | 915 | msg->u.watch.handle = find_watch(msg->u.watch.token); |
4bac07c9 JF |
916 | if (msg->u.watch.handle != NULL) { |
917 | spin_lock(&watch_events_lock); | |
918 | list_add_tail(&msg->list, &watch_events); | |
919 | wake_up(&watch_events_waitq); | |
920 | spin_unlock(&watch_events_lock); | |
921 | } else { | |
5584ea25 | 922 | kfree(body); |
4bac07c9 JF |
923 | kfree(msg); |
924 | } | |
925 | spin_unlock(&watches_lock); | |
926 | } else { | |
927 | msg->u.reply.body = body; | |
928 | spin_lock(&xs_state.reply_lock); | |
929 | list_add_tail(&msg->list, &xs_state.reply_list); | |
930 | spin_unlock(&xs_state.reply_lock); | |
931 | wake_up(&xs_state.reply_waitq); | |
932 | } | |
933 | ||
934 | out: | |
935 | mutex_unlock(&xs_state.response_mutex); | |
936 | return err; | |
937 | } | |
938 | ||
939 | static int xenbus_thread(void *unused) | |
940 | { | |
941 | int err; | |
942 | ||
943 | for (;;) { | |
944 | err = process_msg(); | |
945 | if (err) | |
283c0972 | 946 | pr_warn("error %d while reading message\n", err); |
4bac07c9 JF |
947 | if (kthread_should_stop()) |
948 | break; | |
949 | } | |
950 | ||
951 | return 0; | |
952 | } | |
953 | ||
954 | int xs_init(void) | |
955 | { | |
956 | int err; | |
957 | struct task_struct *task; | |
958 | ||
959 | INIT_LIST_HEAD(&xs_state.reply_list); | |
960 | spin_lock_init(&xs_state.reply_lock); | |
961 | init_waitqueue_head(&xs_state.reply_waitq); | |
962 | ||
963 | mutex_init(&xs_state.request_mutex); | |
964 | mutex_init(&xs_state.response_mutex); | |
4c31a781 | 965 | mutex_init(&xs_state.transaction_mutex); |
4bac07c9 | 966 | init_rwsem(&xs_state.watch_mutex); |
4c31a781 IC |
967 | atomic_set(&xs_state.transaction_count, 0); |
968 | init_waitqueue_head(&xs_state.transaction_wq); | |
4bac07c9 JF |
969 | |
970 | /* Initialize the shared memory rings to talk to xenstored */ | |
971 | err = xb_init_comms(); | |
972 | if (err) | |
973 | return err; | |
974 | ||
975 | task = kthread_run(xenwatch_thread, NULL, "xenwatch"); | |
976 | if (IS_ERR(task)) | |
977 | return PTR_ERR(task); | |
978 | xenwatch_pid = task->pid; | |
979 | ||
980 | task = kthread_run(xenbus_thread, NULL, "xenbus"); | |
981 | if (IS_ERR(task)) | |
982 | return PTR_ERR(task); | |
983 | ||
254d1a3f OH |
984 | /* shutdown watches for kexec boot */ |
985 | xs_reset_watches(); | |
986 | ||
4bac07c9 JF |
987 | return 0; |
988 | } |