1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2009-2010 The FreeBSD Foundation
5 *
6 * This software was developed by Pawel Jakub Dawidek under sponsorship from
7 * the FreeBSD Foundation.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 */
30
31 #include <sys/types.h>
32 #include <sys/wait.h>
33
34 #include <errno.h>
35 #include <pthread.h>
36 #include <signal.h>
37 #include <stdio.h>
38 #include <string.h>
39 #include <unistd.h>
40
41 #include "hast.h"
42 #include "hastd.h"
43 #include "hast_checksum.h"
44 #include "hast_compression.h"
45 #include "hast_proto.h"
46 #include "hooks.h"
47 #include "nv.h"
48 #include "pjdlog.h"
49 #include "proto.h"
50 #include "subr.h"
51
52 #include "control.h"
53
54 void
child_cleanup(struct hast_resource * res)55 child_cleanup(struct hast_resource *res)
56 {
57
58 proto_close(res->hr_ctrl);
59 res->hr_ctrl = NULL;
60 if (res->hr_event != NULL) {
61 proto_close(res->hr_event);
62 res->hr_event = NULL;
63 }
64 if (res->hr_conn != NULL) {
65 proto_close(res->hr_conn);
66 res->hr_conn = NULL;
67 }
68 res->hr_workerpid = 0;
69 }
70
71 static void
control_set_role_common(struct hastd_config * cfg,struct nv * nvout,uint8_t role,struct hast_resource * res,const char * name,unsigned int no)72 control_set_role_common(struct hastd_config *cfg, struct nv *nvout,
73 uint8_t role, struct hast_resource *res, const char *name, unsigned int no)
74 {
75 int oldrole;
76
77 /* Name is always needed. */
78 if (name != NULL)
79 nv_add_string(nvout, name, "resource%u", no);
80
81 if (res == NULL) {
82 PJDLOG_ASSERT(cfg != NULL);
83 PJDLOG_ASSERT(name != NULL);
84
85 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
86 if (strcmp(res->hr_name, name) == 0)
87 break;
88 }
89 if (res == NULL) {
90 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
91 return;
92 }
93 }
94 PJDLOG_ASSERT(res != NULL);
95
96 /* Send previous role back. */
97 nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
98
99 /* Nothing changed, return here. */
100 if (role == res->hr_role)
101 return;
102
103 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
104 pjdlog_info("Role changed to %s.", role2str(role));
105
106 /* Change role to the new one. */
107 oldrole = res->hr_role;
108 res->hr_role = role;
109 pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
110
111 /*
112 * If previous role was primary or secondary we have to kill process
113 * doing that work.
114 */
115 if (res->hr_workerpid != 0) {
116 if (kill(res->hr_workerpid, SIGTERM) == -1) {
117 pjdlog_errno(LOG_WARNING,
118 "Unable to kill worker process %u",
119 (unsigned int)res->hr_workerpid);
120 } else if (waitpid(res->hr_workerpid, NULL, 0) !=
121 res->hr_workerpid) {
122 pjdlog_errno(LOG_WARNING,
123 "Error while waiting for worker process %u",
124 (unsigned int)res->hr_workerpid);
125 } else {
126 pjdlog_debug(1, "Worker process %u stopped.",
127 (unsigned int)res->hr_workerpid);
128 }
129 child_cleanup(res);
130 }
131
132 /* Start worker process if we are changing to primary. */
133 if (role == HAST_ROLE_PRIMARY)
134 hastd_primary(res);
135 pjdlog_prefix_set("%s", "");
136 hook_exec(res->hr_exec, "role", res->hr_name, role2str(oldrole),
137 role2str(res->hr_role), NULL);
138 }
139
140 void
control_set_role(struct hast_resource * res,uint8_t role)141 control_set_role(struct hast_resource *res, uint8_t role)
142 {
143
144 control_set_role_common(NULL, NULL, role, res, NULL, 0);
145 }
146
147 static void
control_status_worker(struct hast_resource * res,struct nv * nvout,unsigned int no)148 control_status_worker(struct hast_resource *res, struct nv *nvout,
149 unsigned int no)
150 {
151 struct nv *cnvin, *cnvout;
152 const char *str;
153 int error;
154
155 cnvin = NULL;
156
157 /*
158 * Prepare and send command to worker process.
159 */
160 cnvout = nv_alloc();
161 nv_add_uint8(cnvout, CONTROL_STATUS, "cmd");
162 error = nv_error(cnvout);
163 if (error != 0) {
164 pjdlog_common(LOG_ERR, 0, error,
165 "Unable to prepare control header");
166 goto end;
167 }
168 if (hast_proto_send(res, res->hr_ctrl, cnvout, NULL, 0) == -1) {
169 error = errno;
170 pjdlog_errno(LOG_ERR, "Unable to send control header");
171 goto end;
172 }
173
174 /*
175 * Receive response.
176 */
177 if (hast_proto_recv_hdr(res->hr_ctrl, &cnvin) == -1) {
178 error = errno;
179 pjdlog_errno(LOG_ERR, "Unable to receive control header");
180 goto end;
181 }
182
183 error = nv_get_int16(cnvin, "error");
184 if (error != 0)
185 goto end;
186
187 if ((str = nv_get_string(cnvin, "status")) == NULL) {
188 error = ENOENT;
189 pjdlog_errno(LOG_ERR, "Field 'status' is missing.");
190 goto end;
191 }
192 nv_add_string(nvout, str, "status%u", no);
193 nv_add_uint64(nvout, nv_get_uint64(cnvin, "dirty"), "dirty%u", no);
194 nv_add_uint32(nvout, nv_get_uint32(cnvin, "extentsize"),
195 "extentsize%u", no);
196 nv_add_uint32(nvout, nv_get_uint32(cnvin, "keepdirty"),
197 "keepdirty%u", no);
198 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read"),
199 "stat_read%u", no);
200 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write"),
201 "stat_write%u", no);
202 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete"),
203 "stat_delete%u", no);
204 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush"),
205 "stat_flush%u", no);
206 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_activemap_update"),
207 "stat_activemap_update%u", no);
208 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read_error"),
209 "stat_read_error%u", no);
210 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write_error"),
211 "stat_write_error%u", no);
212 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete_error"),
213 "stat_delete_error%u", no);
214 nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush_error"),
215 "stat_flush_error%u", no);
216 nv_add_uint64(nvout, nv_get_uint64(cnvin, "idle_queue_size"),
217 "idle_queue_size%u", no);
218 nv_add_uint64(nvout, nv_get_uint64(cnvin, "local_queue_size"),
219 "local_queue_size%u", no);
220 nv_add_uint64(nvout, nv_get_uint64(cnvin, "send_queue_size"),
221 "send_queue_size%u", no);
222 nv_add_uint64(nvout, nv_get_uint64(cnvin, "recv_queue_size"),
223 "recv_queue_size%u", no);
224 nv_add_uint64(nvout, nv_get_uint64(cnvin, "done_queue_size"),
225 "done_queue_size%u", no);
226 end:
227 if (cnvin != NULL)
228 nv_free(cnvin);
229 if (cnvout != NULL)
230 nv_free(cnvout);
231 if (error != 0)
232 nv_add_int16(nvout, error, "error");
233 }
234
235 static void
control_status(struct hastd_config * cfg,struct nv * nvout,struct hast_resource * res,const char * name,unsigned int no)236 control_status(struct hastd_config *cfg, struct nv *nvout,
237 struct hast_resource *res, const char *name, unsigned int no)
238 {
239
240 PJDLOG_ASSERT(cfg != NULL);
241 PJDLOG_ASSERT(nvout != NULL);
242 PJDLOG_ASSERT(name != NULL);
243
244 /* Name is always needed. */
245 nv_add_string(nvout, name, "resource%u", no);
246
247 if (res == NULL) {
248 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
249 if (strcmp(res->hr_name, name) == 0)
250 break;
251 }
252 if (res == NULL) {
253 nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
254 return;
255 }
256 }
257 PJDLOG_ASSERT(res != NULL);
258 nv_add_string(nvout, res->hr_provname, "provname%u", no);
259 nv_add_string(nvout, res->hr_localpath, "localpath%u", no);
260 nv_add_string(nvout, res->hr_remoteaddr, "remoteaddr%u", no);
261 if (res->hr_sourceaddr[0] != '\0')
262 nv_add_string(nvout, res->hr_sourceaddr, "sourceaddr%u", no);
263 switch (res->hr_replication) {
264 case HAST_REPLICATION_FULLSYNC:
265 nv_add_string(nvout, "fullsync", "replication%u", no);
266 break;
267 case HAST_REPLICATION_MEMSYNC:
268 nv_add_string(nvout, "memsync", "replication%u", no);
269 break;
270 case HAST_REPLICATION_ASYNC:
271 nv_add_string(nvout, "async", "replication%u", no);
272 break;
273 default:
274 nv_add_string(nvout, "unknown", "replication%u", no);
275 break;
276 }
277 nv_add_string(nvout, checksum_name(res->hr_checksum),
278 "checksum%u", no);
279 nv_add_string(nvout, compression_name(res->hr_compression),
280 "compression%u", no);
281 nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
282 nv_add_int32(nvout, res->hr_workerpid, "workerpid%u", no);
283
284 switch (res->hr_role) {
285 case HAST_ROLE_PRIMARY:
286 PJDLOG_ASSERT(res->hr_workerpid != 0);
287 /* FALLTHROUGH */
288 case HAST_ROLE_SECONDARY:
289 if (res->hr_workerpid != 0)
290 break;
291 /* FALLTHROUGH */
292 default:
293 return;
294 }
295
296 /*
297 * If we are here, it means that we have a worker process, which we
298 * want to ask some questions.
299 */
300 control_status_worker(res, nvout, no);
301 }
302
303 void
control_handle(struct hastd_config * cfg)304 control_handle(struct hastd_config *cfg)
305 {
306 struct proto_conn *conn;
307 struct nv *nvin, *nvout;
308 unsigned int ii;
309 const char *str;
310 uint8_t cmd, role;
311 int error;
312
313 if (proto_accept(cfg->hc_controlconn, &conn) == -1) {
314 pjdlog_errno(LOG_ERR, "Unable to accept control connection");
315 return;
316 }
317
318 cfg->hc_controlin = conn;
319 nvin = nvout = NULL;
320 role = HAST_ROLE_UNDEF;
321
322 if (hast_proto_recv_hdr(conn, &nvin) == -1) {
323 pjdlog_errno(LOG_ERR, "Unable to receive control header");
324 nvin = NULL;
325 goto close;
326 }
327
328 /* Obtain command code. 0 means that nv_get_uint8() failed. */
329 cmd = nv_get_uint8(nvin, "cmd");
330 if (cmd == 0) {
331 pjdlog_error("Control header is missing 'cmd' field.");
332 goto close;
333 }
334
335 /* Allocate outgoing nv structure. */
336 nvout = nv_alloc();
337 if (nvout == NULL) {
338 pjdlog_error("Unable to allocate header for control response.");
339 goto close;
340 }
341
342 error = 0;
343
344 str = nv_get_string(nvin, "resource0");
345 if (str == NULL) {
346 pjdlog_error("Control header is missing 'resource0' field.");
347 error = EHAST_INVALID;
348 goto fail;
349 }
350 if (cmd == HASTCTL_CMD_SETROLE) {
351 role = nv_get_uint8(nvin, "role");
352 switch (role) {
353 case HAST_ROLE_INIT:
354 case HAST_ROLE_PRIMARY:
355 case HAST_ROLE_SECONDARY:
356 break;
357 default:
358 pjdlog_error("Invalid role received (%hhu).", role);
359 error = EHAST_INVALID;
360 goto fail;
361 }
362 }
363 if (strcmp(str, "all") == 0) {
364 struct hast_resource *res;
365
366 /* All configured resources. */
367
368 ii = 0;
369 TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
370 switch (cmd) {
371 case HASTCTL_CMD_SETROLE:
372 control_set_role_common(cfg, nvout, role, res,
373 res->hr_name, ii++);
374 break;
375 case HASTCTL_CMD_STATUS:
376 control_status(cfg, nvout, res, res->hr_name,
377 ii++);
378 break;
379 default:
380 pjdlog_error("Invalid command received (%hhu).",
381 cmd);
382 error = EHAST_UNIMPLEMENTED;
383 goto fail;
384 }
385 }
386 } else {
387 /* Only selected resources. */
388
389 for (ii = 0; ; ii++) {
390 str = nv_get_string(nvin, "resource%u", ii);
391 if (str == NULL)
392 break;
393 switch (cmd) {
394 case HASTCTL_CMD_SETROLE:
395 control_set_role_common(cfg, nvout, role, NULL,
396 str, ii);
397 break;
398 case HASTCTL_CMD_STATUS:
399 control_status(cfg, nvout, NULL, str, ii);
400 break;
401 default:
402 pjdlog_error("Invalid command received (%hhu).",
403 cmd);
404 error = EHAST_UNIMPLEMENTED;
405 goto fail;
406 }
407 }
408 }
409 if (nv_error(nvout) != 0)
410 goto close;
411 fail:
412 if (error != 0)
413 nv_add_int16(nvout, error, "error");
414
415 if (hast_proto_send(NULL, conn, nvout, NULL, 0) == -1)
416 pjdlog_errno(LOG_ERR, "Unable to send control response");
417 close:
418 if (nvin != NULL)
419 nv_free(nvin);
420 if (nvout != NULL)
421 nv_free(nvout);
422 proto_close(conn);
423 cfg->hc_controlin = NULL;
424 }
425
426 /*
427 * Thread handles control requests from the parent.
428 */
429 void *
ctrl_thread(void * arg)430 ctrl_thread(void *arg)
431 {
432 struct hast_resource *res = arg;
433 struct nv *nvin, *nvout;
434 uint8_t cmd;
435
436 for (;;) {
437 if (hast_proto_recv_hdr(res->hr_ctrl, &nvin) == -1) {
438 if (sigexit_received)
439 pthread_exit(NULL);
440 pjdlog_errno(LOG_ERR,
441 "Unable to receive control message");
442 kill(getpid(), SIGTERM);
443 pthread_exit(NULL);
444 }
445 cmd = nv_get_uint8(nvin, "cmd");
446 if (cmd == 0) {
447 pjdlog_error("Control message is missing 'cmd' field.");
448 nv_free(nvin);
449 continue;
450 }
451 nvout = nv_alloc();
452 switch (cmd) {
453 case CONTROL_STATUS:
454 if (res->hr_remotein != NULL &&
455 res->hr_remoteout != NULL) {
456 nv_add_string(nvout, "complete", "status");
457 } else {
458 nv_add_string(nvout, "degraded", "status");
459 }
460 nv_add_uint32(nvout, (uint32_t)res->hr_extentsize,
461 "extentsize");
462 if (res->hr_role == HAST_ROLE_PRIMARY) {
463 nv_add_uint32(nvout,
464 (uint32_t)res->hr_keepdirty, "keepdirty");
465 nv_add_uint64(nvout,
466 (uint64_t)(activemap_ndirty(res->hr_amp) *
467 res->hr_extentsize), "dirty");
468 } else {
469 nv_add_uint32(nvout, (uint32_t)0, "keepdirty");
470 nv_add_uint64(nvout, (uint64_t)0, "dirty");
471 }
472 nv_add_uint64(nvout, res->hr_stat_read, "stat_read");
473 nv_add_uint64(nvout, res->hr_stat_write, "stat_write");
474 nv_add_uint64(nvout, res->hr_stat_delete,
475 "stat_delete");
476 nv_add_uint64(nvout, res->hr_stat_flush, "stat_flush");
477 nv_add_uint64(nvout, res->hr_stat_activemap_update,
478 "stat_activemap_update");
479 nv_add_uint64(nvout, res->hr_stat_read_error,
480 "stat_read_error");
481 nv_add_uint64(nvout, res->hr_stat_write_error +
482 res->hr_stat_activemap_write_error,
483 "stat_write_error");
484 nv_add_uint64(nvout, res->hr_stat_delete_error,
485 "stat_delete_error");
486 nv_add_uint64(nvout, res->hr_stat_flush_error +
487 res->hr_stat_activemap_flush_error,
488 "stat_flush_error");
489 res->output_status_aux(nvout);
490 nv_add_int16(nvout, 0, "error");
491 break;
492 case CONTROL_RELOAD:
493 /*
494 * When parent receives SIGHUP and discovers that
495 * something related to us has changes, it sends reload
496 * message to us.
497 */
498 PJDLOG_ASSERT(res->hr_role == HAST_ROLE_PRIMARY);
499 primary_config_reload(res, nvin);
500 nv_add_int16(nvout, 0, "error");
501 break;
502 default:
503 nv_add_int16(nvout, EINVAL, "error");
504 break;
505 }
506 nv_free(nvin);
507 if (nv_error(nvout) != 0) {
508 pjdlog_error("Unable to create answer on control message.");
509 nv_free(nvout);
510 continue;
511 }
512 if (hast_proto_send(NULL, res->hr_ctrl, nvout, NULL, 0) == -1) {
513 pjdlog_errno(LOG_ERR,
514 "Unable to send reply to control message");
515 }
516 nv_free(nvout);
517 }
518 /* NOTREACHED */
519 return (NULL);
520 }
521