xref: /freebsd/sbin/hastd/secondary.c (revision b7c60aadbbd5c846a250c05791fe7406d6d78bf4)
1 /*-
2  * Copyright (c) 2009-2010 The FreeBSD Foundation
3  * Copyright (c) 2010 Pawel Jakub Dawidek <pjd@FreeBSD.org>
4  * All rights reserved.
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/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/param.h>
35 #include <sys/time.h>
36 #include <sys/bio.h>
37 #include <sys/disk.h>
38 #include <sys/stat.h>
39 
40 #include <err.h>
41 #include <errno.h>
42 #include <fcntl.h>
43 #include <libgeom.h>
44 #include <pthread.h>
45 #include <signal.h>
46 #include <stdint.h>
47 #include <stdio.h>
48 #include <string.h>
49 #include <sysexits.h>
50 #include <unistd.h>
51 
52 #include <activemap.h>
53 #include <nv.h>
54 #include <pjdlog.h>
55 
56 #include "control.h"
57 #include "event.h"
58 #include "hast.h"
59 #include "hast_proto.h"
60 #include "hastd.h"
61 #include "hooks.h"
62 #include "metadata.h"
63 #include "proto.h"
64 #include "subr.h"
65 #include "synch.h"
66 
67 struct hio {
68 	uint64_t	 hio_seq;
69 	int		 hio_error;
70 	void		*hio_data;
71 	uint8_t		 hio_cmd;
72 	uint64_t	 hio_offset;
73 	uint64_t	 hio_length;
74 	TAILQ_ENTRY(hio) hio_next;
75 };
76 
77 static struct hast_resource *gres;
78 
79 /*
80  * Free list holds unused structures. When free list is empty, we have to wait
81  * until some in-progress requests are freed.
82  */
83 static TAILQ_HEAD(, hio) hio_free_list;
84 static pthread_mutex_t hio_free_list_lock;
85 static pthread_cond_t hio_free_list_cond;
86 /*
87  * Disk thread (the one that do I/O requests) takes requests from this list.
88  */
89 static TAILQ_HEAD(, hio) hio_disk_list;
90 static pthread_mutex_t hio_disk_list_lock;
91 static pthread_cond_t hio_disk_list_cond;
92 /*
93  * There is one recv list for every component, although local components don't
94  * use recv lists as local requests are done synchronously.
95  */
96 static TAILQ_HEAD(, hio) hio_send_list;
97 static pthread_mutex_t hio_send_list_lock;
98 static pthread_cond_t hio_send_list_cond;
99 
100 /*
101  * Maximum number of outstanding I/O requests.
102  */
103 #define	HAST_HIO_MAX	256
104 
105 static void *recv_thread(void *arg);
106 static void *disk_thread(void *arg);
107 static void *send_thread(void *arg);
108 
109 #define	QUEUE_INSERT(name, hio)	do {					\
110 	bool _wakeup;							\
111 									\
112 	mtx_lock(&hio_##name##_list_lock);				\
113 	_wakeup = TAILQ_EMPTY(&hio_##name##_list);			\
114 	TAILQ_INSERT_TAIL(&hio_##name##_list, (hio), hio_next);		\
115 	mtx_unlock(&hio_##name##_list_lock);				\
116 	if (_wakeup)							\
117 		cv_signal(&hio_##name##_list_cond);			\
118 } while (0)
119 #define	QUEUE_TAKE(name, hio)	do {					\
120 	mtx_lock(&hio_##name##_list_lock);				\
121 	while (((hio) = TAILQ_FIRST(&hio_##name##_list)) == NULL) {	\
122 		cv_wait(&hio_##name##_list_cond,			\
123 		    &hio_##name##_list_lock);				\
124 	}								\
125 	TAILQ_REMOVE(&hio_##name##_list, (hio), hio_next);		\
126 	mtx_unlock(&hio_##name##_list_lock);				\
127 } while (0)
128 
129 static void
130 hio_clear(struct hio *hio)
131 {
132 
133 	hio->hio_seq = 0;
134 	hio->hio_error = 0;
135 	hio->hio_cmd = HIO_UNDEF;
136 	hio->hio_offset = 0;
137 	hio->hio_length = 0;
138 }
139 
140 static void
141 init_environment(void)
142 {
143 	struct hio *hio;
144 	unsigned int ii;
145 
146 	/*
147 	 * Initialize lists, their locks and theirs condition variables.
148 	 */
149 	TAILQ_INIT(&hio_free_list);
150 	mtx_init(&hio_free_list_lock);
151 	cv_init(&hio_free_list_cond);
152 	TAILQ_INIT(&hio_disk_list);
153 	mtx_init(&hio_disk_list_lock);
154 	cv_init(&hio_disk_list_cond);
155 	TAILQ_INIT(&hio_send_list);
156 	mtx_init(&hio_send_list_lock);
157 	cv_init(&hio_send_list_cond);
158 
159 	/*
160 	 * Allocate requests pool and initialize requests.
161 	 */
162 	for (ii = 0; ii < HAST_HIO_MAX; ii++) {
163 		hio = malloc(sizeof(*hio));
164 		if (hio == NULL) {
165 			pjdlog_exitx(EX_TEMPFAIL,
166 			    "Unable to allocate memory (%zu bytes) for hio request.",
167 			    sizeof(*hio));
168 		}
169 		hio->hio_data = malloc(MAXPHYS);
170 		if (hio->hio_data == NULL) {
171 			pjdlog_exitx(EX_TEMPFAIL,
172 			    "Unable to allocate memory (%zu bytes) for gctl_data.",
173 			    (size_t)MAXPHYS);
174 		}
175 		hio_clear(hio);
176 		TAILQ_INSERT_HEAD(&hio_free_list, hio, hio_next);
177 	}
178 }
179 
180 static void
181 init_local(struct hast_resource *res)
182 {
183 
184 	if (metadata_read(res, true) == -1)
185 		exit(EX_NOINPUT);
186 }
187 
188 static void
189 init_remote(struct hast_resource *res, struct nv *nvin)
190 {
191 	uint64_t resuid;
192 	struct nv *nvout;
193 	unsigned char *map;
194 	size_t mapsize;
195 
196 #ifdef notyet
197 	/* Setup direction. */
198 	if (proto_send(res->hr_remoteout, NULL, 0) == -1)
199 		pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
200 #endif
201 
202 	nvout = nv_alloc();
203 	nv_add_int64(nvout, (int64_t)res->hr_datasize, "datasize");
204 	nv_add_int32(nvout, (int32_t)res->hr_extentsize, "extentsize");
205 	resuid = nv_get_uint64(nvin, "resuid");
206 	res->hr_primary_localcnt = nv_get_uint64(nvin, "localcnt");
207 	res->hr_primary_remotecnt = nv_get_uint64(nvin, "remotecnt");
208 	nv_add_uint64(nvout, res->hr_secondary_localcnt, "localcnt");
209 	nv_add_uint64(nvout, res->hr_secondary_remotecnt, "remotecnt");
210 	mapsize = activemap_calc_ondisk_size(res->hr_local_mediasize -
211 	    METADATA_SIZE, res->hr_extentsize, res->hr_local_sectorsize);
212 	map = malloc(mapsize);
213 	if (map == NULL) {
214 		pjdlog_exitx(EX_TEMPFAIL,
215 		    "Unable to allocate memory (%zu bytes) for activemap.",
216 		    mapsize);
217 	}
218 	/*
219 	 * When we work as primary and secondary is missing we will increase
220 	 * localcnt in our metadata. When secondary is connected and synced
221 	 * we make localcnt be equal to remotecnt, which means nodes are more
222 	 * or less in sync.
223 	 * Split-brain condition is when both nodes are not able to communicate
224 	 * and are both configured as primary nodes. In turn, they can both
225 	 * make incompatible changes to the data and we have to detect that.
226 	 * Under split-brain condition we will increase our localcnt on first
227 	 * write and remote node will increase its localcnt on first write.
228 	 * When we connect we can see that primary's localcnt is greater than
229 	 * our remotecnt (primary was modified while we weren't watching) and
230 	 * our localcnt is greater than primary's remotecnt (we were modified
231 	 * while primary wasn't watching).
232 	 * There are many possible combinations which are all gathered below.
233 	 * Don't pay too much attention to exact numbers, the more important
234 	 * is to compare them. We compare secondary's local with primary's
235 	 * remote and secondary's remote with primary's local.
236 	 * Note that every case where primary's localcnt is smaller than
237 	 * secondary's remotecnt and where secondary's localcnt is smaller than
238 	 * primary's remotecnt should be impossible in practise. We will perform
239 	 * full synchronization then. Those cases are marked with an asterisk.
240 	 * Regular synchronization means that only extents marked as dirty are
241 	 * synchronized (regular synchronization).
242 	 *
243 	 * SECONDARY METADATA PRIMARY METADATA
244 	 * local=3 remote=3   local=2 remote=2*  ?! Full sync from secondary.
245 	 * local=3 remote=3   local=2 remote=3*  ?! Full sync from primary.
246 	 * local=3 remote=3   local=2 remote=4*  ?! Full sync from primary.
247 	 * local=3 remote=3   local=3 remote=2   Primary is out-of-date,
248 	 *                                       regular sync from secondary.
249 	 * local=3 remote=3   local=3 remote=3   Regular sync just in case.
250 	 * local=3 remote=3   local=3 remote=4*  ?! Full sync from primary.
251 	 * local=3 remote=3   local=4 remote=2   Split-brain condition.
252 	 * local=3 remote=3   local=4 remote=3   Secondary out-of-date,
253 	 *                                       regular sync from primary.
254 	 * local=3 remote=3   local=4 remote=4*  ?! Full sync from primary.
255 	 */
256 	if (res->hr_resuid == 0) {
257 		/*
258 		 * Provider is used for the first time. If primary node done no
259 		 * writes yet as well (we will find "virgin" argument) then
260 		 * there is no need to synchronize anything. If primary node
261 		 * done any writes already we have to synchronize everything.
262 		 */
263 		PJDLOG_ASSERT(res->hr_secondary_localcnt == 0);
264 		res->hr_resuid = resuid;
265 		if (metadata_write(res) == -1)
266 			exit(EX_NOINPUT);
267 		if (nv_exists(nvin, "virgin")) {
268 			free(map);
269 			map = NULL;
270 			mapsize = 0;
271 		} else {
272 			memset(map, 0xff, mapsize);
273 		}
274 		nv_add_int8(nvout, 1, "virgin");
275 		nv_add_uint8(nvout, HAST_SYNCSRC_PRIMARY, "syncsrc");
276 	} else if (res->hr_resuid != resuid) {
277 		char errmsg[256];
278 
279 		free(map);
280 		(void)snprintf(errmsg, sizeof(errmsg),
281 		    "Resource unique ID mismatch (primary=%ju, secondary=%ju).",
282 		    (uintmax_t)resuid, (uintmax_t)res->hr_resuid);
283 		pjdlog_error("%s", errmsg);
284 		nv_add_string(nvout, errmsg, "errmsg");
285 		if (hast_proto_send(res, res->hr_remotein, nvout,
286 		    NULL, 0) == -1) {
287 			pjdlog_exit(EX_TEMPFAIL,
288 			    "Unable to send response to %s",
289 			    res->hr_remoteaddr);
290 		}
291 		nv_free(nvout);
292 		exit(EX_CONFIG);
293 	} else if (
294 	    /* Is primary out-of-date? */
295 	    (res->hr_secondary_localcnt > res->hr_primary_remotecnt &&
296 	     res->hr_secondary_remotecnt == res->hr_primary_localcnt) ||
297 	    /* Are the nodes more or less in sync? */
298 	    (res->hr_secondary_localcnt == res->hr_primary_remotecnt &&
299 	     res->hr_secondary_remotecnt == res->hr_primary_localcnt) ||
300 	    /* Is secondary out-of-date? */
301 	    (res->hr_secondary_localcnt == res->hr_primary_remotecnt &&
302 	     res->hr_secondary_remotecnt < res->hr_primary_localcnt)) {
303 		/*
304 		 * Nodes are more or less in sync or one of the nodes is
305 		 * out-of-date.
306 		 * It doesn't matter at this point which one, we just have to
307 		 * send out local bitmap to the remote node.
308 		 */
309 		if (pread(res->hr_localfd, map, mapsize, METADATA_SIZE) !=
310 		    (ssize_t)mapsize) {
311 			pjdlog_exit(LOG_ERR, "Unable to read activemap");
312 		}
313 		if (res->hr_secondary_localcnt > res->hr_primary_remotecnt &&
314 		     res->hr_secondary_remotecnt == res->hr_primary_localcnt) {
315 			/* Primary is out-of-date, sync from secondary. */
316 			nv_add_uint8(nvout, HAST_SYNCSRC_SECONDARY, "syncsrc");
317 		} else {
318 			/*
319 			 * Secondary is out-of-date or counts match.
320 			 * Sync from primary.
321 			 */
322 			nv_add_uint8(nvout, HAST_SYNCSRC_PRIMARY, "syncsrc");
323 		}
324 	} else if (res->hr_secondary_localcnt > res->hr_primary_remotecnt &&
325 	     res->hr_primary_localcnt > res->hr_secondary_remotecnt) {
326 		/*
327 		 * Not good, we have split-brain condition.
328 		 */
329 		free(map);
330 		pjdlog_error("Split-brain detected, exiting.");
331 		nv_add_string(nvout, "Split-brain condition!", "errmsg");
332 		if (hast_proto_send(res, res->hr_remotein, nvout,
333 		    NULL, 0) == -1) {
334 			pjdlog_exit(EX_TEMPFAIL,
335 			    "Unable to send response to %s",
336 			    res->hr_remoteaddr);
337 		}
338 		nv_free(nvout);
339 		/* Exit on split-brain. */
340 		event_send(res, EVENT_SPLITBRAIN);
341 		exit(EX_CONFIG);
342 	} else /* if (res->hr_secondary_localcnt < res->hr_primary_remotecnt ||
343 	    res->hr_primary_localcnt < res->hr_secondary_remotecnt) */ {
344 		/*
345 		 * This should never happen in practise, but we will perform
346 		 * full synchronization.
347 		 */
348 		PJDLOG_ASSERT(res->hr_secondary_localcnt < res->hr_primary_remotecnt ||
349 		    res->hr_primary_localcnt < res->hr_secondary_remotecnt);
350 		mapsize = activemap_calc_ondisk_size(res->hr_local_mediasize -
351 		    METADATA_SIZE, res->hr_extentsize,
352 		    res->hr_local_sectorsize);
353 		memset(map, 0xff, mapsize);
354 		if (res->hr_secondary_localcnt > res->hr_primary_remotecnt) {
355 			/* In this one of five cases sync from secondary. */
356 			nv_add_uint8(nvout, HAST_SYNCSRC_SECONDARY, "syncsrc");
357 		} else {
358 			/* For the rest four cases sync from primary. */
359 			nv_add_uint8(nvout, HAST_SYNCSRC_PRIMARY, "syncsrc");
360 		}
361 		pjdlog_warning("This should never happen, asking for full synchronization (primary(local=%ju, remote=%ju), secondary(local=%ju, remote=%ju)).",
362 		    (uintmax_t)res->hr_primary_localcnt,
363 		    (uintmax_t)res->hr_primary_remotecnt,
364 		    (uintmax_t)res->hr_secondary_localcnt,
365 		    (uintmax_t)res->hr_secondary_remotecnt);
366 	}
367 	nv_add_uint32(nvout, (uint32_t)mapsize, "mapsize");
368 	if (hast_proto_send(res, res->hr_remotein, nvout, map, mapsize) == -1) {
369 		pjdlog_exit(EX_TEMPFAIL, "Unable to send activemap to %s",
370 		    res->hr_remoteaddr);
371 	}
372 	if (map != NULL)
373 		free(map);
374 	nv_free(nvout);
375 #ifdef notyet
376 	/* Setup direction. */
377 	if (proto_recv(res->hr_remotein, NULL, 0) == -1)
378 		pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
379 #endif
380 }
381 
382 void
383 hastd_secondary(struct hast_resource *res, struct nv *nvin)
384 {
385 	sigset_t mask;
386 	pthread_t td;
387 	pid_t pid;
388 	int error, mode, debuglevel;
389 
390 	/*
391 	 * Create communication channel between parent and child.
392 	 */
393 	if (proto_client(NULL, "socketpair://", &res->hr_ctrl) == -1) {
394 		KEEP_ERRNO((void)pidfile_remove(pfh));
395 		pjdlog_exit(EX_OSERR,
396 		    "Unable to create control sockets between parent and child");
397 	}
398 	/*
399 	 * Create communication channel between child and parent.
400 	 */
401 	if (proto_client(NULL, "socketpair://", &res->hr_event) == -1) {
402 		KEEP_ERRNO((void)pidfile_remove(pfh));
403 		pjdlog_exit(EX_OSERR,
404 		    "Unable to create event sockets between child and parent");
405 	}
406 
407 	pid = fork();
408 	if (pid == -1) {
409 		KEEP_ERRNO((void)pidfile_remove(pfh));
410 		pjdlog_exit(EX_OSERR, "Unable to fork");
411 	}
412 
413 	if (pid > 0) {
414 		/* This is parent. */
415 		proto_close(res->hr_remotein);
416 		res->hr_remotein = NULL;
417 		proto_close(res->hr_remoteout);
418 		res->hr_remoteout = NULL;
419 		/* Declare that we are receiver. */
420 		proto_recv(res->hr_event, NULL, 0);
421 		/* Declare that we are sender. */
422 		proto_send(res->hr_ctrl, NULL, 0);
423 		res->hr_workerpid = pid;
424 		return;
425 	}
426 
427 	gres = res;
428 	mode = pjdlog_mode_get();
429 	debuglevel = pjdlog_debug_get();
430 
431 	/* Declare that we are sender. */
432 	proto_send(res->hr_event, NULL, 0);
433 	/* Declare that we are receiver. */
434 	proto_recv(res->hr_ctrl, NULL, 0);
435 	descriptors_cleanup(res);
436 
437 	descriptors_assert(res, mode);
438 
439 	pjdlog_init(mode);
440 	pjdlog_debug_set(debuglevel);
441 	pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
442 	setproctitle("%s (%s)", res->hr_name, role2str(res->hr_role));
443 
444 	PJDLOG_VERIFY(sigemptyset(&mask) == 0);
445 	PJDLOG_VERIFY(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
446 
447 	/* Error in setting timeout is not critical, but why should it fail? */
448 	if (proto_timeout(res->hr_remotein, 2 * HAST_KEEPALIVE) == -1)
449 		pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
450 	if (proto_timeout(res->hr_remoteout, res->hr_timeout) == -1)
451 		pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
452 
453 	init_local(res);
454 	init_environment();
455 
456 	if (drop_privs(res) != 0)
457 		exit(EX_CONFIG);
458 	pjdlog_info("Privileges successfully dropped.");
459 
460 	/*
461 	 * Create the control thread before sending any event to the parent,
462 	 * as we can deadlock when parent sends control request to worker,
463 	 * but worker has no control thread started yet, so parent waits.
464 	 * In the meantime worker sends an event to the parent, but parent
465 	 * is unable to handle the event, because it waits for control
466 	 * request response.
467 	 */
468 	error = pthread_create(&td, NULL, ctrl_thread, res);
469 	PJDLOG_ASSERT(error == 0);
470 
471 	init_remote(res, nvin);
472 	event_send(res, EVENT_CONNECT);
473 
474 	error = pthread_create(&td, NULL, recv_thread, res);
475 	PJDLOG_ASSERT(error == 0);
476 	error = pthread_create(&td, NULL, disk_thread, res);
477 	PJDLOG_ASSERT(error == 0);
478 	(void)send_thread(res);
479 }
480 
481 static void
482 reqlog(int loglevel, int debuglevel, int error, struct hio *hio,
483     const char *fmt, ...)
484 {
485 	char msg[1024];
486 	va_list ap;
487 	int len;
488 
489 	va_start(ap, fmt);
490 	len = vsnprintf(msg, sizeof(msg), fmt, ap);
491 	va_end(ap);
492 	if ((size_t)len < sizeof(msg)) {
493 		switch (hio->hio_cmd) {
494 		case HIO_READ:
495 			(void)snprintf(msg + len, sizeof(msg) - len,
496 			    "READ(%ju, %ju).", (uintmax_t)hio->hio_offset,
497 			    (uintmax_t)hio->hio_length);
498 			break;
499 		case HIO_DELETE:
500 			(void)snprintf(msg + len, sizeof(msg) - len,
501 			    "DELETE(%ju, %ju).", (uintmax_t)hio->hio_offset,
502 			    (uintmax_t)hio->hio_length);
503 			break;
504 		case HIO_FLUSH:
505 			(void)snprintf(msg + len, sizeof(msg) - len, "FLUSH.");
506 			break;
507 		case HIO_WRITE:
508 			(void)snprintf(msg + len, sizeof(msg) - len,
509 			    "WRITE(%ju, %ju).", (uintmax_t)hio->hio_offset,
510 			    (uintmax_t)hio->hio_length);
511 			break;
512 		case HIO_KEEPALIVE:
513 			(void)snprintf(msg + len, sizeof(msg) - len, "KEEPALIVE.");
514 			break;
515 		default:
516 			(void)snprintf(msg + len, sizeof(msg) - len,
517 			    "UNKNOWN(%u).", (unsigned int)hio->hio_cmd);
518 			break;
519 		}
520 	}
521 	pjdlog_common(loglevel, debuglevel, error, "%s", msg);
522 }
523 
524 static int
525 requnpack(struct hast_resource *res, struct hio *hio, struct nv *nv)
526 {
527 
528 	hio->hio_cmd = nv_get_uint8(nv, "cmd");
529 	if (hio->hio_cmd == 0) {
530 		pjdlog_error("Header contains no 'cmd' field.");
531 		hio->hio_error = EINVAL;
532 		goto end;
533 	}
534 	if (hio->hio_cmd != HIO_KEEPALIVE) {
535 		hio->hio_seq = nv_get_uint64(nv, "seq");
536 		if (hio->hio_seq == 0) {
537 			pjdlog_error("Header contains no 'seq' field.");
538 			hio->hio_error = EINVAL;
539 			goto end;
540 		}
541 	}
542 	switch (hio->hio_cmd) {
543 	case HIO_FLUSH:
544 	case HIO_KEEPALIVE:
545 		break;
546 	case HIO_READ:
547 	case HIO_WRITE:
548 	case HIO_DELETE:
549 		hio->hio_offset = nv_get_uint64(nv, "offset");
550 		if (nv_error(nv) != 0) {
551 			pjdlog_error("Header is missing 'offset' field.");
552 			hio->hio_error = EINVAL;
553 			goto end;
554 		}
555 		hio->hio_length = nv_get_uint64(nv, "length");
556 		if (nv_error(nv) != 0) {
557 			pjdlog_error("Header is missing 'length' field.");
558 			hio->hio_error = EINVAL;
559 			goto end;
560 		}
561 		if (hio->hio_length == 0) {
562 			pjdlog_error("Data length is zero.");
563 			hio->hio_error = EINVAL;
564 			goto end;
565 		}
566 		if (hio->hio_length > MAXPHYS) {
567 			pjdlog_error("Data length is too large (%ju > %ju).",
568 			    (uintmax_t)hio->hio_length, (uintmax_t)MAXPHYS);
569 			hio->hio_error = EINVAL;
570 			goto end;
571 		}
572 		if ((hio->hio_offset % res->hr_local_sectorsize) != 0) {
573 			pjdlog_error("Offset %ju is not multiple of sector size.",
574 			    (uintmax_t)hio->hio_offset);
575 			hio->hio_error = EINVAL;
576 			goto end;
577 		}
578 		if ((hio->hio_length % res->hr_local_sectorsize) != 0) {
579 			pjdlog_error("Length %ju is not multiple of sector size.",
580 			    (uintmax_t)hio->hio_length);
581 			hio->hio_error = EINVAL;
582 			goto end;
583 		}
584 		if (hio->hio_offset + hio->hio_length >
585 		    (uint64_t)res->hr_datasize) {
586 			pjdlog_error("Data offset is too large (%ju > %ju).",
587 			    (uintmax_t)(hio->hio_offset + hio->hio_length),
588 			    (uintmax_t)res->hr_datasize);
589 			hio->hio_error = EINVAL;
590 			goto end;
591 		}
592 		break;
593 	default:
594 		pjdlog_error("Header contains invalid 'cmd' (%hhu).",
595 		    hio->hio_cmd);
596 		hio->hio_error = EINVAL;
597 		goto end;
598 	}
599 	hio->hio_error = 0;
600 end:
601 	return (hio->hio_error);
602 }
603 
604 static __dead2 void
605 secondary_exit(int exitcode, const char *fmt, ...)
606 {
607 	va_list ap;
608 
609 	PJDLOG_ASSERT(exitcode != EX_OK);
610 	va_start(ap, fmt);
611 	pjdlogv_errno(LOG_ERR, fmt, ap);
612 	va_end(ap);
613 	event_send(gres, EVENT_DISCONNECT);
614 	exit(exitcode);
615 }
616 
617 /*
618  * Thread receives requests from the primary node.
619  */
620 static void *
621 recv_thread(void *arg)
622 {
623 	struct hast_resource *res = arg;
624 	struct hio *hio;
625 	struct nv *nv;
626 
627 	for (;;) {
628 		pjdlog_debug(2, "recv: Taking free request.");
629 		QUEUE_TAKE(free, hio);
630 		pjdlog_debug(2, "recv: (%p) Got request.", hio);
631 		if (hast_proto_recv_hdr(res->hr_remotein, &nv) == -1) {
632 			secondary_exit(EX_TEMPFAIL,
633 			    "Unable to receive request header");
634 		}
635 		if (requnpack(res, hio, nv) != 0) {
636 			nv_free(nv);
637 			pjdlog_debug(2,
638 			    "recv: (%p) Moving request to the send queue.",
639 			    hio);
640 			QUEUE_INSERT(send, hio);
641 			continue;
642 		}
643 		switch (hio->hio_cmd) {
644 		case HIO_READ:
645 			res->hr_stat_read++;
646 			break;
647 		case HIO_WRITE:
648 			res->hr_stat_write++;
649 			break;
650 		case HIO_DELETE:
651 			res->hr_stat_delete++;
652 			break;
653 		case HIO_FLUSH:
654 			res->hr_stat_flush++;
655 			break;
656 		case HIO_KEEPALIVE:
657 			break;
658 		default:
659 			PJDLOG_ABORT("Unexpected command (cmd=%hhu).",
660 			    hio->hio_cmd);
661 		}
662 		reqlog(LOG_DEBUG, 2, -1, hio,
663 		    "recv: (%p) Got request header: ", hio);
664 		if (hio->hio_cmd == HIO_KEEPALIVE) {
665 			nv_free(nv);
666 			pjdlog_debug(2,
667 			    "recv: (%p) Moving request to the free queue.",
668 			    hio);
669 			hio_clear(hio);
670 			QUEUE_INSERT(free, hio);
671 			continue;
672 		} else if (hio->hio_cmd == HIO_WRITE) {
673 			if (hast_proto_recv_data(res, res->hr_remotein, nv,
674 			    hio->hio_data, MAXPHYS) == -1) {
675 				secondary_exit(EX_TEMPFAIL,
676 				    "Unable to receive request data");
677 			}
678 		}
679 		nv_free(nv);
680 		pjdlog_debug(2, "recv: (%p) Moving request to the disk queue.",
681 		    hio);
682 		QUEUE_INSERT(disk, hio);
683 	}
684 	/* NOTREACHED */
685 	return (NULL);
686 }
687 
688 /*
689  * Thread reads from or writes to local component and also handles DELETE and
690  * FLUSH requests.
691  */
692 static void *
693 disk_thread(void *arg)
694 {
695 	struct hast_resource *res = arg;
696 	struct hio *hio;
697 	ssize_t ret;
698 	bool clear_activemap, logerror;
699 
700 	clear_activemap = true;
701 
702 	for (;;) {
703 		pjdlog_debug(2, "disk: Taking request.");
704 		QUEUE_TAKE(disk, hio);
705 		while (clear_activemap) {
706 			unsigned char *map;
707 			size_t mapsize;
708 
709 			/*
710 			 * When first request is received, it means that primary
711 			 * already received our activemap, merged it and stored
712 			 * locally. We can now safely clear our activemap.
713 			 */
714 			mapsize =
715 			    activemap_calc_ondisk_size(res->hr_local_mediasize -
716 			    METADATA_SIZE, res->hr_extentsize,
717 			    res->hr_local_sectorsize);
718 			map = calloc(1, mapsize);
719 			if (map == NULL) {
720 				pjdlog_warning("Unable to allocate memory to clear local activemap.");
721 				break;
722 			}
723 			if (pwrite(res->hr_localfd, map, mapsize,
724 			    METADATA_SIZE) != (ssize_t)mapsize) {
725 				pjdlog_errno(LOG_WARNING,
726 				    "Unable to store cleared activemap");
727 				free(map);
728 				break;
729 			}
730 			free(map);
731 			clear_activemap = false;
732 			pjdlog_debug(1, "Local activemap cleared.");
733 			break;
734 		}
735 		reqlog(LOG_DEBUG, 2, -1, hio, "disk: (%p) Got request: ", hio);
736 		logerror = true;
737 		/* Handle the actual request. */
738 		switch (hio->hio_cmd) {
739 		case HIO_READ:
740 			ret = pread(res->hr_localfd, hio->hio_data,
741 			    hio->hio_length,
742 			    hio->hio_offset + res->hr_localoff);
743 			if (ret == -1)
744 				hio->hio_error = errno;
745 			else if (ret != (int64_t)hio->hio_length)
746 				hio->hio_error = EIO;
747 			else
748 				hio->hio_error = 0;
749 			break;
750 		case HIO_WRITE:
751 			ret = pwrite(res->hr_localfd, hio->hio_data,
752 			    hio->hio_length,
753 			    hio->hio_offset + res->hr_localoff);
754 			if (ret == -1)
755 				hio->hio_error = errno;
756 			else if (ret != (int64_t)hio->hio_length)
757 				hio->hio_error = EIO;
758 			else
759 				hio->hio_error = 0;
760 			break;
761 		case HIO_DELETE:
762 			ret = g_delete(res->hr_localfd,
763 			    hio->hio_offset + res->hr_localoff,
764 			    hio->hio_length);
765 			if (ret == -1)
766 				hio->hio_error = errno;
767 			else
768 				hio->hio_error = 0;
769 			break;
770 		case HIO_FLUSH:
771 			if (!res->hr_localflush) {
772 				ret = -1;
773 				hio->hio_error = EOPNOTSUPP;
774 				logerror = false;
775 				break;
776 			}
777 			ret = g_flush(res->hr_localfd);
778 			if (ret == -1) {
779 				if (errno == EOPNOTSUPP)
780 					res->hr_localflush = false;
781 				hio->hio_error = errno;
782 			} else {
783 				hio->hio_error = 0;
784 			}
785 			break;
786 		default:
787 			PJDLOG_ABORT("Unexpected command (cmd=%hhu).",
788 			    hio->hio_cmd);
789 		}
790 		if (logerror && hio->hio_error != 0) {
791 			reqlog(LOG_ERR, 0, hio->hio_error, hio,
792 			    "Request failed: ");
793 		}
794 		pjdlog_debug(2, "disk: (%p) Moving request to the send queue.",
795 		    hio);
796 		QUEUE_INSERT(send, hio);
797 	}
798 	/* NOTREACHED */
799 	return (NULL);
800 }
801 
802 /*
803  * Thread sends requests back to primary node.
804  */
805 static void *
806 send_thread(void *arg)
807 {
808 	struct hast_resource *res = arg;
809 	struct nv *nvout;
810 	struct hio *hio;
811 	void *data;
812 	size_t length;
813 
814 	for (;;) {
815 		pjdlog_debug(2, "send: Taking request.");
816 		QUEUE_TAKE(send, hio);
817 		reqlog(LOG_DEBUG, 2, -1, hio, "send: (%p) Got request: ", hio);
818 		nvout = nv_alloc();
819 		/* Copy sequence number. */
820 		nv_add_uint64(nvout, hio->hio_seq, "seq");
821 		switch (hio->hio_cmd) {
822 		case HIO_READ:
823 			if (hio->hio_error == 0) {
824 				data = hio->hio_data;
825 				length = hio->hio_length;
826 				break;
827 			}
828 			/*
829 			 * We send no data in case of an error.
830 			 */
831 			/* FALLTHROUGH */
832 		case HIO_DELETE:
833 		case HIO_FLUSH:
834 		case HIO_WRITE:
835 			data = NULL;
836 			length = 0;
837 			break;
838 		default:
839 			PJDLOG_ABORT("Unexpected command (cmd=%hhu).",
840 			    hio->hio_cmd);
841 		}
842 		if (hio->hio_error != 0)
843 			nv_add_int16(nvout, hio->hio_error, "error");
844 		if (hast_proto_send(res, res->hr_remoteout, nvout, data,
845 		    length) == -1) {
846 			secondary_exit(EX_TEMPFAIL, "Unable to send reply");
847 		}
848 		nv_free(nvout);
849 		pjdlog_debug(2, "send: (%p) Moving request to the free queue.",
850 		    hio);
851 		hio_clear(hio);
852 		QUEUE_INSERT(free, hio);
853 	}
854 	/* NOTREACHED */
855 	return (NULL);
856 }
857