xref: /freebsd/sbin/hastd/primary.c (revision 10b59a9b4add0320d52c15ce057dd697261e7dfc)
1 /*-
2  * Copyright (c) 2009 The FreeBSD Foundation
3  * Copyright (c) 2010-2011 Pawel Jakub Dawidek <pawel@dawidek.net>
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/types.h>
35 #include <sys/time.h>
36 #include <sys/bio.h>
37 #include <sys/disk.h>
38 #include <sys/refcount.h>
39 #include <sys/stat.h>
40 
41 #include <geom/gate/g_gate.h>
42 
43 #include <err.h>
44 #include <errno.h>
45 #include <fcntl.h>
46 #include <libgeom.h>
47 #include <pthread.h>
48 #include <signal.h>
49 #include <stdint.h>
50 #include <stdio.h>
51 #include <string.h>
52 #include <sysexits.h>
53 #include <unistd.h>
54 
55 #include <activemap.h>
56 #include <nv.h>
57 #include <rangelock.h>
58 
59 #include "control.h"
60 #include "event.h"
61 #include "hast.h"
62 #include "hast_proto.h"
63 #include "hastd.h"
64 #include "hooks.h"
65 #include "metadata.h"
66 #include "proto.h"
67 #include "pjdlog.h"
68 #include "subr.h"
69 #include "synch.h"
70 
71 /* The is only one remote component for now. */
72 #define	ISREMOTE(no)	((no) == 1)
73 
74 struct hio {
75 	/*
76 	 * Number of components we are still waiting for.
77 	 * When this field goes to 0, we can send the request back to the
78 	 * kernel. Each component has to decrease this counter by one
79 	 * even on failure.
80 	 */
81 	unsigned int		 hio_countdown;
82 	/*
83 	 * Each component has a place to store its own error.
84 	 * Once the request is handled by all components we can decide if the
85 	 * request overall is successful or not.
86 	 */
87 	int			*hio_errors;
88 	/*
89 	 * Structure used to communicate with GEOM Gate class.
90 	 */
91 	struct g_gate_ctl_io	 hio_ggio;
92 	/*
93 	 * Request was already confirmed to GEOM Gate.
94 	 */
95 	bool			 hio_done;
96 	/*
97 	 * Remember replication from the time the request was initiated,
98 	 * so we won't get confused when replication changes on reload.
99 	 */
100 	int			 hio_replication;
101 	TAILQ_ENTRY(hio)	*hio_next;
102 };
103 #define	hio_free_next	hio_next[0]
104 #define	hio_done_next	hio_next[0]
105 
106 /*
107  * Free list holds unused structures. When free list is empty, we have to wait
108  * until some in-progress requests are freed.
109  */
110 static TAILQ_HEAD(, hio) hio_free_list;
111 static pthread_mutex_t hio_free_list_lock;
112 static pthread_cond_t hio_free_list_cond;
113 /*
114  * There is one send list for every component. One requests is placed on all
115  * send lists - each component gets the same request, but each component is
116  * responsible for managing his own send list.
117  */
118 static TAILQ_HEAD(, hio) *hio_send_list;
119 static pthread_mutex_t *hio_send_list_lock;
120 static pthread_cond_t *hio_send_list_cond;
121 /*
122  * There is one recv list for every component, although local components don't
123  * use recv lists as local requests are done synchronously.
124  */
125 static TAILQ_HEAD(, hio) *hio_recv_list;
126 static pthread_mutex_t *hio_recv_list_lock;
127 static pthread_cond_t *hio_recv_list_cond;
128 /*
129  * Request is placed on done list by the slowest component (the one that
130  * decreased hio_countdown from 1 to 0).
131  */
132 static TAILQ_HEAD(, hio) hio_done_list;
133 static pthread_mutex_t hio_done_list_lock;
134 static pthread_cond_t hio_done_list_cond;
135 /*
136  * Structure below are for interaction with sync thread.
137  */
138 static bool sync_inprogress;
139 static pthread_mutex_t sync_lock;
140 static pthread_cond_t sync_cond;
141 /*
142  * The lock below allows to synchornize access to remote connections.
143  */
144 static pthread_rwlock_t *hio_remote_lock;
145 
146 /*
147  * Lock to synchronize metadata updates. Also synchronize access to
148  * hr_primary_localcnt and hr_primary_remotecnt fields.
149  */
150 static pthread_mutex_t metadata_lock;
151 
152 /*
153  * Maximum number of outstanding I/O requests.
154  */
155 #define	HAST_HIO_MAX	256
156 /*
157  * Number of components. At this point there are only two components: local
158  * and remote, but in the future it might be possible to use multiple local
159  * and remote components.
160  */
161 #define	HAST_NCOMPONENTS	2
162 
163 #define	ISCONNECTED(res, no)	\
164 	((res)->hr_remotein != NULL && (res)->hr_remoteout != NULL)
165 
166 #define	QUEUE_INSERT1(hio, name, ncomp)	do {				\
167 	bool _wakeup;							\
168 									\
169 	mtx_lock(&hio_##name##_list_lock[(ncomp)]);			\
170 	_wakeup = TAILQ_EMPTY(&hio_##name##_list[(ncomp)]);		\
171 	TAILQ_INSERT_TAIL(&hio_##name##_list[(ncomp)], (hio),		\
172 	    hio_next[(ncomp)]);						\
173 	mtx_unlock(&hio_##name##_list_lock[ncomp]);			\
174 	if (_wakeup)							\
175 		cv_signal(&hio_##name##_list_cond[(ncomp)]);		\
176 } while (0)
177 #define	QUEUE_INSERT2(hio, name)	do {				\
178 	bool _wakeup;							\
179 									\
180 	mtx_lock(&hio_##name##_list_lock);				\
181 	_wakeup = TAILQ_EMPTY(&hio_##name##_list);			\
182 	TAILQ_INSERT_TAIL(&hio_##name##_list, (hio), hio_##name##_next);\
183 	mtx_unlock(&hio_##name##_list_lock);				\
184 	if (_wakeup)							\
185 		cv_signal(&hio_##name##_list_cond);			\
186 } while (0)
187 #define	QUEUE_TAKE1(hio, name, ncomp, timeout)	do {			\
188 	bool _last;							\
189 									\
190 	mtx_lock(&hio_##name##_list_lock[(ncomp)]);			\
191 	_last = false;							\
192 	while (((hio) = TAILQ_FIRST(&hio_##name##_list[(ncomp)])) == NULL && !_last) { \
193 		cv_timedwait(&hio_##name##_list_cond[(ncomp)],		\
194 		    &hio_##name##_list_lock[(ncomp)], (timeout));	\
195 		if ((timeout) != 0)					\
196 			_last = true;					\
197 	}								\
198 	if (hio != NULL) {						\
199 		TAILQ_REMOVE(&hio_##name##_list[(ncomp)], (hio),	\
200 		    hio_next[(ncomp)]);					\
201 	}								\
202 	mtx_unlock(&hio_##name##_list_lock[(ncomp)]);			\
203 } while (0)
204 #define	QUEUE_TAKE2(hio, name)	do {					\
205 	mtx_lock(&hio_##name##_list_lock);				\
206 	while (((hio) = TAILQ_FIRST(&hio_##name##_list)) == NULL) {	\
207 		cv_wait(&hio_##name##_list_cond,			\
208 		    &hio_##name##_list_lock);				\
209 	}								\
210 	TAILQ_REMOVE(&hio_##name##_list, (hio), hio_##name##_next);	\
211 	mtx_unlock(&hio_##name##_list_lock);				\
212 } while (0)
213 
214 #define	SYNCREQ(hio)		do {					\
215 	(hio)->hio_ggio.gctl_unit = -1;					\
216 	(hio)->hio_ggio.gctl_seq = 1;					\
217 } while (0)
218 #define	ISSYNCREQ(hio)		((hio)->hio_ggio.gctl_unit == -1)
219 #define	SYNCREQDONE(hio)	do { (hio)->hio_ggio.gctl_unit = -2; } while (0)
220 #define	ISSYNCREQDONE(hio)	((hio)->hio_ggio.gctl_unit == -2)
221 
222 static struct hast_resource *gres;
223 
224 static pthread_mutex_t range_lock;
225 static struct rangelocks *range_regular;
226 static bool range_regular_wait;
227 static pthread_cond_t range_regular_cond;
228 static struct rangelocks *range_sync;
229 static bool range_sync_wait;
230 static pthread_cond_t range_sync_cond;
231 static bool fullystarted;
232 
233 static void *ggate_recv_thread(void *arg);
234 static void *local_send_thread(void *arg);
235 static void *remote_send_thread(void *arg);
236 static void *remote_recv_thread(void *arg);
237 static void *ggate_send_thread(void *arg);
238 static void *sync_thread(void *arg);
239 static void *guard_thread(void *arg);
240 
241 static void
242 cleanup(struct hast_resource *res)
243 {
244 	int rerrno;
245 
246 	/* Remember errno. */
247 	rerrno = errno;
248 
249 	/* Destroy ggate provider if we created one. */
250 	if (res->hr_ggateunit >= 0) {
251 		struct g_gate_ctl_destroy ggiod;
252 
253 		bzero(&ggiod, sizeof(ggiod));
254 		ggiod.gctl_version = G_GATE_VERSION;
255 		ggiod.gctl_unit = res->hr_ggateunit;
256 		ggiod.gctl_force = 1;
257 		if (ioctl(res->hr_ggatefd, G_GATE_CMD_DESTROY, &ggiod) < 0) {
258 			pjdlog_errno(LOG_WARNING,
259 			    "Unable to destroy hast/%s device",
260 			    res->hr_provname);
261 		}
262 		res->hr_ggateunit = -1;
263 	}
264 
265 	/* Restore errno. */
266 	errno = rerrno;
267 }
268 
269 static __dead2 void
270 primary_exit(int exitcode, const char *fmt, ...)
271 {
272 	va_list ap;
273 
274 	PJDLOG_ASSERT(exitcode != EX_OK);
275 	va_start(ap, fmt);
276 	pjdlogv_errno(LOG_ERR, fmt, ap);
277 	va_end(ap);
278 	cleanup(gres);
279 	exit(exitcode);
280 }
281 
282 static __dead2 void
283 primary_exitx(int exitcode, const char *fmt, ...)
284 {
285 	va_list ap;
286 
287 	va_start(ap, fmt);
288 	pjdlogv(exitcode == EX_OK ? LOG_INFO : LOG_ERR, fmt, ap);
289 	va_end(ap);
290 	cleanup(gres);
291 	exit(exitcode);
292 }
293 
294 static int
295 hast_activemap_flush(struct hast_resource *res)
296 {
297 	const unsigned char *buf;
298 	size_t size;
299 
300 	buf = activemap_bitmap(res->hr_amp, &size);
301 	PJDLOG_ASSERT(buf != NULL);
302 	PJDLOG_ASSERT((size % res->hr_local_sectorsize) == 0);
303 	if (pwrite(res->hr_localfd, buf, size, METADATA_SIZE) !=
304 	    (ssize_t)size) {
305 		pjdlog_errno(LOG_ERR, "Unable to flush activemap to disk");
306 		return (-1);
307 	}
308 	if (res->hr_metaflush == 1 && g_flush(res->hr_localfd) == -1) {
309 		if (errno == EOPNOTSUPP) {
310 			pjdlog_warning("The %s provider doesn't support flushing write cache. Disabling it.",
311 			    res->hr_localpath);
312 			res->hr_metaflush = 0;
313 		} else {
314 			pjdlog_errno(LOG_ERR,
315 			    "Unable to flush disk cache on activemap update");
316 			return (-1);
317 		}
318 	}
319 	return (0);
320 }
321 
322 static bool
323 real_remote(const struct hast_resource *res)
324 {
325 
326 	return (strcmp(res->hr_remoteaddr, "none") != 0);
327 }
328 
329 static void
330 init_environment(struct hast_resource *res __unused)
331 {
332 	struct hio *hio;
333 	unsigned int ii, ncomps;
334 
335 	/*
336 	 * In the future it might be per-resource value.
337 	 */
338 	ncomps = HAST_NCOMPONENTS;
339 
340 	/*
341 	 * Allocate memory needed by lists.
342 	 */
343 	hio_send_list = malloc(sizeof(hio_send_list[0]) * ncomps);
344 	if (hio_send_list == NULL) {
345 		primary_exitx(EX_TEMPFAIL,
346 		    "Unable to allocate %zu bytes of memory for send lists.",
347 		    sizeof(hio_send_list[0]) * ncomps);
348 	}
349 	hio_send_list_lock = malloc(sizeof(hio_send_list_lock[0]) * ncomps);
350 	if (hio_send_list_lock == NULL) {
351 		primary_exitx(EX_TEMPFAIL,
352 		    "Unable to allocate %zu bytes of memory for send list locks.",
353 		    sizeof(hio_send_list_lock[0]) * ncomps);
354 	}
355 	hio_send_list_cond = malloc(sizeof(hio_send_list_cond[0]) * ncomps);
356 	if (hio_send_list_cond == NULL) {
357 		primary_exitx(EX_TEMPFAIL,
358 		    "Unable to allocate %zu bytes of memory for send list condition variables.",
359 		    sizeof(hio_send_list_cond[0]) * ncomps);
360 	}
361 	hio_recv_list = malloc(sizeof(hio_recv_list[0]) * ncomps);
362 	if (hio_recv_list == NULL) {
363 		primary_exitx(EX_TEMPFAIL,
364 		    "Unable to allocate %zu bytes of memory for recv lists.",
365 		    sizeof(hio_recv_list[0]) * ncomps);
366 	}
367 	hio_recv_list_lock = malloc(sizeof(hio_recv_list_lock[0]) * ncomps);
368 	if (hio_recv_list_lock == NULL) {
369 		primary_exitx(EX_TEMPFAIL,
370 		    "Unable to allocate %zu bytes of memory for recv list locks.",
371 		    sizeof(hio_recv_list_lock[0]) * ncomps);
372 	}
373 	hio_recv_list_cond = malloc(sizeof(hio_recv_list_cond[0]) * ncomps);
374 	if (hio_recv_list_cond == NULL) {
375 		primary_exitx(EX_TEMPFAIL,
376 		    "Unable to allocate %zu bytes of memory for recv list condition variables.",
377 		    sizeof(hio_recv_list_cond[0]) * ncomps);
378 	}
379 	hio_remote_lock = malloc(sizeof(hio_remote_lock[0]) * ncomps);
380 	if (hio_remote_lock == NULL) {
381 		primary_exitx(EX_TEMPFAIL,
382 		    "Unable to allocate %zu bytes of memory for remote connections locks.",
383 		    sizeof(hio_remote_lock[0]) * ncomps);
384 	}
385 
386 	/*
387 	 * Initialize lists, their locks and theirs condition variables.
388 	 */
389 	TAILQ_INIT(&hio_free_list);
390 	mtx_init(&hio_free_list_lock);
391 	cv_init(&hio_free_list_cond);
392 	for (ii = 0; ii < HAST_NCOMPONENTS; ii++) {
393 		TAILQ_INIT(&hio_send_list[ii]);
394 		mtx_init(&hio_send_list_lock[ii]);
395 		cv_init(&hio_send_list_cond[ii]);
396 		TAILQ_INIT(&hio_recv_list[ii]);
397 		mtx_init(&hio_recv_list_lock[ii]);
398 		cv_init(&hio_recv_list_cond[ii]);
399 		rw_init(&hio_remote_lock[ii]);
400 	}
401 	TAILQ_INIT(&hio_done_list);
402 	mtx_init(&hio_done_list_lock);
403 	cv_init(&hio_done_list_cond);
404 	mtx_init(&metadata_lock);
405 
406 	/*
407 	 * Allocate requests pool and initialize requests.
408 	 */
409 	for (ii = 0; ii < HAST_HIO_MAX; ii++) {
410 		hio = malloc(sizeof(*hio));
411 		if (hio == NULL) {
412 			primary_exitx(EX_TEMPFAIL,
413 			    "Unable to allocate %zu bytes of memory for hio request.",
414 			    sizeof(*hio));
415 		}
416 		hio->hio_countdown = 0;
417 		hio->hio_errors = malloc(sizeof(hio->hio_errors[0]) * ncomps);
418 		if (hio->hio_errors == NULL) {
419 			primary_exitx(EX_TEMPFAIL,
420 			    "Unable allocate %zu bytes of memory for hio errors.",
421 			    sizeof(hio->hio_errors[0]) * ncomps);
422 		}
423 		hio->hio_next = malloc(sizeof(hio->hio_next[0]) * ncomps);
424 		if (hio->hio_next == NULL) {
425 			primary_exitx(EX_TEMPFAIL,
426 			    "Unable allocate %zu bytes of memory for hio_next field.",
427 			    sizeof(hio->hio_next[0]) * ncomps);
428 		}
429 		hio->hio_ggio.gctl_version = G_GATE_VERSION;
430 		hio->hio_ggio.gctl_data = malloc(MAXPHYS);
431 		if (hio->hio_ggio.gctl_data == NULL) {
432 			primary_exitx(EX_TEMPFAIL,
433 			    "Unable to allocate %zu bytes of memory for gctl_data.",
434 			    MAXPHYS);
435 		}
436 		hio->hio_ggio.gctl_length = MAXPHYS;
437 		hio->hio_ggio.gctl_error = 0;
438 		TAILQ_INSERT_HEAD(&hio_free_list, hio, hio_free_next);
439 	}
440 }
441 
442 static bool
443 init_resuid(struct hast_resource *res)
444 {
445 
446 	mtx_lock(&metadata_lock);
447 	if (res->hr_resuid != 0) {
448 		mtx_unlock(&metadata_lock);
449 		return (false);
450 	} else {
451 		/* Initialize unique resource identifier. */
452 		arc4random_buf(&res->hr_resuid, sizeof(res->hr_resuid));
453 		mtx_unlock(&metadata_lock);
454 		if (metadata_write(res) < 0)
455 			exit(EX_NOINPUT);
456 		return (true);
457 	}
458 }
459 
460 static void
461 init_local(struct hast_resource *res)
462 {
463 	unsigned char *buf;
464 	size_t mapsize;
465 
466 	if (metadata_read(res, true) < 0)
467 		exit(EX_NOINPUT);
468 	mtx_init(&res->hr_amp_lock);
469 	if (activemap_init(&res->hr_amp, res->hr_datasize, res->hr_extentsize,
470 	    res->hr_local_sectorsize, res->hr_keepdirty) < 0) {
471 		primary_exit(EX_TEMPFAIL, "Unable to create activemap");
472 	}
473 	mtx_init(&range_lock);
474 	cv_init(&range_regular_cond);
475 	if (rangelock_init(&range_regular) < 0)
476 		primary_exit(EX_TEMPFAIL, "Unable to create regular range lock");
477 	cv_init(&range_sync_cond);
478 	if (rangelock_init(&range_sync) < 0)
479 		primary_exit(EX_TEMPFAIL, "Unable to create sync range lock");
480 	mapsize = activemap_ondisk_size(res->hr_amp);
481 	buf = calloc(1, mapsize);
482 	if (buf == NULL) {
483 		primary_exitx(EX_TEMPFAIL,
484 		    "Unable to allocate buffer for activemap.");
485 	}
486 	if (pread(res->hr_localfd, buf, mapsize, METADATA_SIZE) !=
487 	    (ssize_t)mapsize) {
488 		primary_exit(EX_NOINPUT, "Unable to read activemap");
489 	}
490 	activemap_copyin(res->hr_amp, buf, mapsize);
491 	free(buf);
492 	if (res->hr_resuid != 0)
493 		return;
494 	/*
495 	 * We're using provider for the first time. Initialize local and remote
496 	 * counters. We don't initialize resuid here, as we want to do it just
497 	 * in time. The reason for this is that we want to inform secondary
498 	 * that there were no writes yet, so there is no need to synchronize
499 	 * anything.
500 	 */
501 	res->hr_primary_localcnt = 0;
502 	res->hr_primary_remotecnt = 0;
503 	if (metadata_write(res) < 0)
504 		exit(EX_NOINPUT);
505 }
506 
507 static int
508 primary_connect(struct hast_resource *res, struct proto_conn **connp)
509 {
510 	struct proto_conn *conn;
511 	int16_t val;
512 
513 	val = 1;
514 	if (proto_send(res->hr_conn, &val, sizeof(val)) < 0) {
515 		primary_exit(EX_TEMPFAIL,
516 		    "Unable to send connection request to parent");
517 	}
518 	if (proto_recv(res->hr_conn, &val, sizeof(val)) < 0) {
519 		primary_exit(EX_TEMPFAIL,
520 		    "Unable to receive reply to connection request from parent");
521 	}
522 	if (val != 0) {
523 		errno = val;
524 		pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
525 		    res->hr_remoteaddr);
526 		return (-1);
527 	}
528 	if (proto_connection_recv(res->hr_conn, true, &conn) < 0) {
529 		primary_exit(EX_TEMPFAIL,
530 		    "Unable to receive connection from parent");
531 	}
532 	if (proto_connect_wait(conn, res->hr_timeout) < 0) {
533 		pjdlog_errno(LOG_WARNING, "Unable to connect to %s",
534 		    res->hr_remoteaddr);
535 		proto_close(conn);
536 		return (-1);
537 	}
538 	/* Error in setting timeout is not critical, but why should it fail? */
539 	if (proto_timeout(conn, res->hr_timeout) < 0)
540 		pjdlog_errno(LOG_WARNING, "Unable to set connection timeout");
541 
542 	*connp = conn;
543 
544 	return (0);
545 }
546 
547 static int
548 init_remote(struct hast_resource *res, struct proto_conn **inp,
549     struct proto_conn **outp)
550 {
551 	struct proto_conn *in, *out;
552 	struct nv *nvout, *nvin;
553 	const unsigned char *token;
554 	unsigned char *map;
555 	const char *errmsg;
556 	int32_t extentsize;
557 	int64_t datasize;
558 	uint32_t mapsize;
559 	size_t size;
560 	int error;
561 
562 	PJDLOG_ASSERT((inp == NULL && outp == NULL) || (inp != NULL && outp != NULL));
563 	PJDLOG_ASSERT(real_remote(res));
564 
565 	in = out = NULL;
566 	errmsg = NULL;
567 
568 	if (primary_connect(res, &out) == -1)
569 		return (ECONNREFUSED);
570 
571 	error = ECONNABORTED;
572 
573 	/*
574 	 * First handshake step.
575 	 * Setup outgoing connection with remote node.
576 	 */
577 	nvout = nv_alloc();
578 	nv_add_string(nvout, res->hr_name, "resource");
579 	if (nv_error(nvout) != 0) {
580 		pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
581 		    "Unable to allocate header for connection with %s",
582 		    res->hr_remoteaddr);
583 		nv_free(nvout);
584 		goto close;
585 	}
586 	if (hast_proto_send(res, out, nvout, NULL, 0) < 0) {
587 		pjdlog_errno(LOG_WARNING,
588 		    "Unable to send handshake header to %s",
589 		    res->hr_remoteaddr);
590 		nv_free(nvout);
591 		goto close;
592 	}
593 	nv_free(nvout);
594 	if (hast_proto_recv_hdr(out, &nvin) < 0) {
595 		pjdlog_errno(LOG_WARNING,
596 		    "Unable to receive handshake header from %s",
597 		    res->hr_remoteaddr);
598 		goto close;
599 	}
600 	errmsg = nv_get_string(nvin, "errmsg");
601 	if (errmsg != NULL) {
602 		pjdlog_warning("%s", errmsg);
603 		if (nv_exists(nvin, "wait"))
604 			error = EBUSY;
605 		nv_free(nvin);
606 		goto close;
607 	}
608 	token = nv_get_uint8_array(nvin, &size, "token");
609 	if (token == NULL) {
610 		pjdlog_warning("Handshake header from %s has no 'token' field.",
611 		    res->hr_remoteaddr);
612 		nv_free(nvin);
613 		goto close;
614 	}
615 	if (size != sizeof(res->hr_token)) {
616 		pjdlog_warning("Handshake header from %s contains 'token' of wrong size (got %zu, expected %zu).",
617 		    res->hr_remoteaddr, size, sizeof(res->hr_token));
618 		nv_free(nvin);
619 		goto close;
620 	}
621 	bcopy(token, res->hr_token, sizeof(res->hr_token));
622 	nv_free(nvin);
623 
624 	/*
625 	 * Second handshake step.
626 	 * Setup incoming connection with remote node.
627 	 */
628 	if (primary_connect(res, &in) == -1)
629 		goto close;
630 
631 	nvout = nv_alloc();
632 	nv_add_string(nvout, res->hr_name, "resource");
633 	nv_add_uint8_array(nvout, res->hr_token, sizeof(res->hr_token),
634 	    "token");
635 	if (res->hr_resuid == 0) {
636 		/*
637 		 * The resuid field was not yet initialized.
638 		 * Because we do synchronization inside init_resuid(), it is
639 		 * possible that someone already initialized it, the function
640 		 * will return false then, but if we successfully initialized
641 		 * it, we will get true. True means that there were no writes
642 		 * to this resource yet and we want to inform secondary that
643 		 * synchronization is not needed by sending "virgin" argument.
644 		 */
645 		if (init_resuid(res))
646 			nv_add_int8(nvout, 1, "virgin");
647 	}
648 	nv_add_uint64(nvout, res->hr_resuid, "resuid");
649 	nv_add_uint64(nvout, res->hr_primary_localcnt, "localcnt");
650 	nv_add_uint64(nvout, res->hr_primary_remotecnt, "remotecnt");
651 	if (nv_error(nvout) != 0) {
652 		pjdlog_common(LOG_WARNING, 0, nv_error(nvout),
653 		    "Unable to allocate header for connection with %s",
654 		    res->hr_remoteaddr);
655 		nv_free(nvout);
656 		goto close;
657 	}
658 	if (hast_proto_send(res, in, nvout, NULL, 0) < 0) {
659 		pjdlog_errno(LOG_WARNING,
660 		    "Unable to send handshake header to %s",
661 		    res->hr_remoteaddr);
662 		nv_free(nvout);
663 		goto close;
664 	}
665 	nv_free(nvout);
666 	if (hast_proto_recv_hdr(out, &nvin) < 0) {
667 		pjdlog_errno(LOG_WARNING,
668 		    "Unable to receive handshake header from %s",
669 		    res->hr_remoteaddr);
670 		goto close;
671 	}
672 	errmsg = nv_get_string(nvin, "errmsg");
673 	if (errmsg != NULL) {
674 		pjdlog_warning("%s", errmsg);
675 		nv_free(nvin);
676 		goto close;
677 	}
678 	datasize = nv_get_int64(nvin, "datasize");
679 	if (datasize != res->hr_datasize) {
680 		pjdlog_warning("Data size differs between nodes (local=%jd, remote=%jd).",
681 		    (intmax_t)res->hr_datasize, (intmax_t)datasize);
682 		nv_free(nvin);
683 		goto close;
684 	}
685 	extentsize = nv_get_int32(nvin, "extentsize");
686 	if (extentsize != res->hr_extentsize) {
687 		pjdlog_warning("Extent size differs between nodes (local=%zd, remote=%zd).",
688 		    (ssize_t)res->hr_extentsize, (ssize_t)extentsize);
689 		nv_free(nvin);
690 		goto close;
691 	}
692 	res->hr_secondary_localcnt = nv_get_uint64(nvin, "localcnt");
693 	res->hr_secondary_remotecnt = nv_get_uint64(nvin, "remotecnt");
694 	res->hr_syncsrc = nv_get_uint8(nvin, "syncsrc");
695 	if (nv_exists(nvin, "virgin")) {
696 		/*
697 		 * Secondary was reinitialized, bump localcnt if it is 0 as
698 		 * only we have the data.
699 		 */
700 		PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_PRIMARY);
701 		PJDLOG_ASSERT(res->hr_secondary_localcnt == 0);
702 
703 		if (res->hr_primary_localcnt == 0) {
704 			PJDLOG_ASSERT(res->hr_secondary_remotecnt == 0);
705 
706 			mtx_lock(&metadata_lock);
707 			res->hr_primary_localcnt++;
708 			pjdlog_debug(1, "Increasing localcnt to %ju.",
709 			    (uintmax_t)res->hr_primary_localcnt);
710 			(void)metadata_write(res);
711 			mtx_unlock(&metadata_lock);
712 		}
713 	}
714 	map = NULL;
715 	mapsize = nv_get_uint32(nvin, "mapsize");
716 	if (mapsize > 0) {
717 		map = malloc(mapsize);
718 		if (map == NULL) {
719 			pjdlog_error("Unable to allocate memory for remote activemap (mapsize=%ju).",
720 			    (uintmax_t)mapsize);
721 			nv_free(nvin);
722 			goto close;
723 		}
724 		/*
725 		 * Remote node have some dirty extents on its own, lets
726 		 * download its activemap.
727 		 */
728 		if (hast_proto_recv_data(res, out, nvin, map,
729 		    mapsize) < 0) {
730 			pjdlog_errno(LOG_ERR,
731 			    "Unable to receive remote activemap");
732 			nv_free(nvin);
733 			free(map);
734 			goto close;
735 		}
736 		/*
737 		 * Merge local and remote bitmaps.
738 		 */
739 		activemap_merge(res->hr_amp, map, mapsize);
740 		free(map);
741 		/*
742 		 * Now that we merged bitmaps from both nodes, flush it to the
743 		 * disk before we start to synchronize.
744 		 */
745 		(void)hast_activemap_flush(res);
746 	}
747 	nv_free(nvin);
748 #ifdef notyet
749 	/* Setup directions. */
750 	if (proto_send(out, NULL, 0) == -1)
751 		pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
752 	if (proto_recv(in, NULL, 0) == -1)
753 		pjdlog_errno(LOG_WARNING, "Unable to set connection direction");
754 #endif
755 	pjdlog_info("Connected to %s.", res->hr_remoteaddr);
756 	if (inp != NULL && outp != NULL) {
757 		*inp = in;
758 		*outp = out;
759 	} else {
760 		res->hr_remotein = in;
761 		res->hr_remoteout = out;
762 	}
763 	event_send(res, EVENT_CONNECT);
764 	return (0);
765 close:
766 	if (errmsg != NULL && strcmp(errmsg, "Split-brain condition!") == 0)
767 		event_send(res, EVENT_SPLITBRAIN);
768 	proto_close(out);
769 	if (in != NULL)
770 		proto_close(in);
771 	return (error);
772 }
773 
774 static void
775 sync_start(void)
776 {
777 
778 	mtx_lock(&sync_lock);
779 	sync_inprogress = true;
780 	mtx_unlock(&sync_lock);
781 	cv_signal(&sync_cond);
782 }
783 
784 static void
785 sync_stop(void)
786 {
787 
788 	mtx_lock(&sync_lock);
789 	if (sync_inprogress)
790 		sync_inprogress = false;
791 	mtx_unlock(&sync_lock);
792 }
793 
794 static void
795 init_ggate(struct hast_resource *res)
796 {
797 	struct g_gate_ctl_create ggiocreate;
798 	struct g_gate_ctl_cancel ggiocancel;
799 
800 	/*
801 	 * We communicate with ggate via /dev/ggctl. Open it.
802 	 */
803 	res->hr_ggatefd = open("/dev/" G_GATE_CTL_NAME, O_RDWR);
804 	if (res->hr_ggatefd < 0)
805 		primary_exit(EX_OSFILE, "Unable to open /dev/" G_GATE_CTL_NAME);
806 	/*
807 	 * Create provider before trying to connect, as connection failure
808 	 * is not critical, but may take some time.
809 	 */
810 	bzero(&ggiocreate, sizeof(ggiocreate));
811 	ggiocreate.gctl_version = G_GATE_VERSION;
812 	ggiocreate.gctl_mediasize = res->hr_datasize;
813 	ggiocreate.gctl_sectorsize = res->hr_local_sectorsize;
814 	ggiocreate.gctl_flags = 0;
815 	ggiocreate.gctl_maxcount = 0;
816 	ggiocreate.gctl_timeout = 0;
817 	ggiocreate.gctl_unit = G_GATE_NAME_GIVEN;
818 	snprintf(ggiocreate.gctl_name, sizeof(ggiocreate.gctl_name), "hast/%s",
819 	    res->hr_provname);
820 	if (ioctl(res->hr_ggatefd, G_GATE_CMD_CREATE, &ggiocreate) == 0) {
821 		pjdlog_info("Device hast/%s created.", res->hr_provname);
822 		res->hr_ggateunit = ggiocreate.gctl_unit;
823 		return;
824 	}
825 	if (errno != EEXIST) {
826 		primary_exit(EX_OSERR, "Unable to create hast/%s device",
827 		    res->hr_provname);
828 	}
829 	pjdlog_debug(1,
830 	    "Device hast/%s already exists, we will try to take it over.",
831 	    res->hr_provname);
832 	/*
833 	 * If we received EEXIST, we assume that the process who created the
834 	 * provider died and didn't clean up. In that case we will start from
835 	 * where he left of.
836 	 */
837 	bzero(&ggiocancel, sizeof(ggiocancel));
838 	ggiocancel.gctl_version = G_GATE_VERSION;
839 	ggiocancel.gctl_unit = G_GATE_NAME_GIVEN;
840 	snprintf(ggiocancel.gctl_name, sizeof(ggiocancel.gctl_name), "hast/%s",
841 	    res->hr_provname);
842 	if (ioctl(res->hr_ggatefd, G_GATE_CMD_CANCEL, &ggiocancel) == 0) {
843 		pjdlog_info("Device hast/%s recovered.", res->hr_provname);
844 		res->hr_ggateunit = ggiocancel.gctl_unit;
845 		return;
846 	}
847 	primary_exit(EX_OSERR, "Unable to take over hast/%s device",
848 	    res->hr_provname);
849 }
850 
851 void
852 hastd_primary(struct hast_resource *res)
853 {
854 	pthread_t td;
855 	pid_t pid;
856 	int error, mode, debuglevel;
857 
858 	/*
859 	 * Create communication channel for sending control commands from
860 	 * parent to child.
861 	 */
862 	if (proto_client(NULL, "socketpair://", &res->hr_ctrl) < 0) {
863 		/* TODO: There's no need for this to be fatal error. */
864 		KEEP_ERRNO((void)pidfile_remove(pfh));
865 		pjdlog_exit(EX_OSERR,
866 		    "Unable to create control sockets between parent and child");
867 	}
868 	/*
869 	 * Create communication channel for sending events from child to parent.
870 	 */
871 	if (proto_client(NULL, "socketpair://", &res->hr_event) < 0) {
872 		/* TODO: There's no need for this to be fatal error. */
873 		KEEP_ERRNO((void)pidfile_remove(pfh));
874 		pjdlog_exit(EX_OSERR,
875 		    "Unable to create event sockets between child and parent");
876 	}
877 	/*
878 	 * Create communication channel for sending connection requests from
879 	 * child to parent.
880 	 */
881 	if (proto_client(NULL, "socketpair://", &res->hr_conn) < 0) {
882 		/* TODO: There's no need for this to be fatal error. */
883 		KEEP_ERRNO((void)pidfile_remove(pfh));
884 		pjdlog_exit(EX_OSERR,
885 		    "Unable to create connection sockets between child and parent");
886 	}
887 
888 	pid = fork();
889 	if (pid < 0) {
890 		/* TODO: There's no need for this to be fatal error. */
891 		KEEP_ERRNO((void)pidfile_remove(pfh));
892 		pjdlog_exit(EX_TEMPFAIL, "Unable to fork");
893 	}
894 
895 	if (pid > 0) {
896 		/* This is parent. */
897 		/* Declare that we are receiver. */
898 		proto_recv(res->hr_event, NULL, 0);
899 		proto_recv(res->hr_conn, NULL, 0);
900 		/* Declare that we are sender. */
901 		proto_send(res->hr_ctrl, NULL, 0);
902 		res->hr_workerpid = pid;
903 		return;
904 	}
905 
906 	gres = res;
907 	mode = pjdlog_mode_get();
908 	debuglevel = pjdlog_debug_get();
909 
910 	/* Declare that we are sender. */
911 	proto_send(res->hr_event, NULL, 0);
912 	proto_send(res->hr_conn, NULL, 0);
913 	/* Declare that we are receiver. */
914 	proto_recv(res->hr_ctrl, NULL, 0);
915 	descriptors_cleanup(res);
916 
917 	descriptors_assert(res, mode);
918 
919 	pjdlog_init(mode);
920 	pjdlog_debug_set(debuglevel);
921 	pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
922 	setproctitle("%s (%s)", res->hr_name, role2str(res->hr_role));
923 
924 	init_local(res);
925 	init_ggate(res);
926 	init_environment(res);
927 
928 	if (drop_privs(res) != 0) {
929 		cleanup(res);
930 		exit(EX_CONFIG);
931 	}
932 	pjdlog_info("Privileges successfully dropped.");
933 
934 	/*
935 	 * Create the guard thread first, so we can handle signals from the
936 	 * very begining.
937 	 */
938 	error = pthread_create(&td, NULL, guard_thread, res);
939 	PJDLOG_ASSERT(error == 0);
940 	/*
941 	 * Create the control thread before sending any event to the parent,
942 	 * as we can deadlock when parent sends control request to worker,
943 	 * but worker has no control thread started yet, so parent waits.
944 	 * In the meantime worker sends an event to the parent, but parent
945 	 * is unable to handle the event, because it waits for control
946 	 * request response.
947 	 */
948 	error = pthread_create(&td, NULL, ctrl_thread, res);
949 	PJDLOG_ASSERT(error == 0);
950 	if (real_remote(res)) {
951 		error = init_remote(res, NULL, NULL);
952 		if (error == 0) {
953 			sync_start();
954 		} else if (error == EBUSY) {
955 			time_t start = time(NULL);
956 
957 			pjdlog_warning("Waiting for remote node to become %s for %ds.",
958 			    role2str(HAST_ROLE_SECONDARY),
959 			    res->hr_timeout);
960 			for (;;) {
961 				sleep(1);
962 				error = init_remote(res, NULL, NULL);
963 				if (error != EBUSY)
964 					break;
965 				if (time(NULL) > start + res->hr_timeout)
966 					break;
967 			}
968 			if (error == EBUSY) {
969 				pjdlog_warning("Remote node is still %s, starting anyway.",
970 				    role2str(HAST_ROLE_PRIMARY));
971 			}
972 		}
973 	}
974 	error = pthread_create(&td, NULL, ggate_recv_thread, res);
975 	PJDLOG_ASSERT(error == 0);
976 	error = pthread_create(&td, NULL, local_send_thread, res);
977 	PJDLOG_ASSERT(error == 0);
978 	error = pthread_create(&td, NULL, remote_send_thread, res);
979 	PJDLOG_ASSERT(error == 0);
980 	error = pthread_create(&td, NULL, remote_recv_thread, res);
981 	PJDLOG_ASSERT(error == 0);
982 	error = pthread_create(&td, NULL, ggate_send_thread, res);
983 	PJDLOG_ASSERT(error == 0);
984 	fullystarted = true;
985 	(void)sync_thread(res);
986 }
987 
988 static void
989 reqlog(int loglevel, int debuglevel, struct g_gate_ctl_io *ggio, const char *fmt, ...)
990 {
991 	char msg[1024];
992 	va_list ap;
993 	int len;
994 
995 	va_start(ap, fmt);
996 	len = vsnprintf(msg, sizeof(msg), fmt, ap);
997 	va_end(ap);
998 	if ((size_t)len < sizeof(msg)) {
999 		switch (ggio->gctl_cmd) {
1000 		case BIO_READ:
1001 			(void)snprintf(msg + len, sizeof(msg) - len,
1002 			    "READ(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
1003 			    (uintmax_t)ggio->gctl_length);
1004 			break;
1005 		case BIO_DELETE:
1006 			(void)snprintf(msg + len, sizeof(msg) - len,
1007 			    "DELETE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
1008 			    (uintmax_t)ggio->gctl_length);
1009 			break;
1010 		case BIO_FLUSH:
1011 			(void)snprintf(msg + len, sizeof(msg) - len, "FLUSH.");
1012 			break;
1013 		case BIO_WRITE:
1014 			(void)snprintf(msg + len, sizeof(msg) - len,
1015 			    "WRITE(%ju, %ju).", (uintmax_t)ggio->gctl_offset,
1016 			    (uintmax_t)ggio->gctl_length);
1017 			break;
1018 		default:
1019 			(void)snprintf(msg + len, sizeof(msg) - len,
1020 			    "UNKNOWN(%u).", (unsigned int)ggio->gctl_cmd);
1021 			break;
1022 		}
1023 	}
1024 	pjdlog_common(loglevel, debuglevel, -1, "%s", msg);
1025 }
1026 
1027 static void
1028 remote_close(struct hast_resource *res, int ncomp)
1029 {
1030 
1031 	rw_wlock(&hio_remote_lock[ncomp]);
1032 	/*
1033 	 * Check for a race between dropping rlock and acquiring wlock -
1034 	 * another thread can close connection in-between.
1035 	 */
1036 	if (!ISCONNECTED(res, ncomp)) {
1037 		PJDLOG_ASSERT(res->hr_remotein == NULL);
1038 		PJDLOG_ASSERT(res->hr_remoteout == NULL);
1039 		rw_unlock(&hio_remote_lock[ncomp]);
1040 		return;
1041 	}
1042 
1043 	PJDLOG_ASSERT(res->hr_remotein != NULL);
1044 	PJDLOG_ASSERT(res->hr_remoteout != NULL);
1045 
1046 	pjdlog_debug(2, "Closing incoming connection to %s.",
1047 	    res->hr_remoteaddr);
1048 	proto_close(res->hr_remotein);
1049 	res->hr_remotein = NULL;
1050 	pjdlog_debug(2, "Closing outgoing connection to %s.",
1051 	    res->hr_remoteaddr);
1052 	proto_close(res->hr_remoteout);
1053 	res->hr_remoteout = NULL;
1054 
1055 	rw_unlock(&hio_remote_lock[ncomp]);
1056 
1057 	pjdlog_warning("Disconnected from %s.", res->hr_remoteaddr);
1058 
1059 	/*
1060 	 * Stop synchronization if in-progress.
1061 	 */
1062 	sync_stop();
1063 
1064 	event_send(res, EVENT_DISCONNECT);
1065 }
1066 
1067 /*
1068  * Acknowledge write completion to the kernel, but don't update activemap yet.
1069  */
1070 static void
1071 write_complete(struct hast_resource *res, struct hio *hio)
1072 {
1073 	struct g_gate_ctl_io *ggio;
1074 	unsigned int ncomp;
1075 
1076 	PJDLOG_ASSERT(!hio->hio_done);
1077 
1078 	ggio = &hio->hio_ggio;
1079 	PJDLOG_ASSERT(ggio->gctl_cmd == BIO_WRITE);
1080 
1081 	/*
1082 	 * Bump local count if this is first write after
1083 	 * connection failure with remote node.
1084 	 */
1085 	ncomp = 1;
1086 	rw_rlock(&hio_remote_lock[ncomp]);
1087 	if (!ISCONNECTED(res, ncomp)) {
1088 		mtx_lock(&metadata_lock);
1089 		if (res->hr_primary_localcnt == res->hr_secondary_remotecnt) {
1090 			res->hr_primary_localcnt++;
1091 			pjdlog_debug(1, "Increasing localcnt to %ju.",
1092 			    (uintmax_t)res->hr_primary_localcnt);
1093 			(void)metadata_write(res);
1094 		}
1095 		mtx_unlock(&metadata_lock);
1096 	}
1097 	rw_unlock(&hio_remote_lock[ncomp]);
1098 	if (ioctl(res->hr_ggatefd, G_GATE_CMD_DONE, ggio) < 0)
1099 		primary_exit(EX_OSERR, "G_GATE_CMD_DONE failed");
1100 	hio->hio_done = true;
1101 }
1102 
1103 /*
1104  * Thread receives ggate I/O requests from the kernel and passes them to
1105  * appropriate threads:
1106  * WRITE - always goes to both local_send and remote_send threads
1107  * READ (when the block is up-to-date on local component) -
1108  *	only local_send thread
1109  * READ (when the block isn't up-to-date on local component) -
1110  *	only remote_send thread
1111  * DELETE - always goes to both local_send and remote_send threads
1112  * FLUSH - always goes to both local_send and remote_send threads
1113  */
1114 static void *
1115 ggate_recv_thread(void *arg)
1116 {
1117 	struct hast_resource *res = arg;
1118 	struct g_gate_ctl_io *ggio;
1119 	struct hio *hio;
1120 	unsigned int ii, ncomp, ncomps;
1121 	int error;
1122 
1123 	for (;;) {
1124 		pjdlog_debug(2, "ggate_recv: Taking free request.");
1125 		QUEUE_TAKE2(hio, free);
1126 		pjdlog_debug(2, "ggate_recv: (%p) Got free request.", hio);
1127 		ggio = &hio->hio_ggio;
1128 		ggio->gctl_unit = res->hr_ggateunit;
1129 		ggio->gctl_length = MAXPHYS;
1130 		ggio->gctl_error = 0;
1131 		hio->hio_done = false;
1132 		hio->hio_replication = res->hr_replication;
1133 		pjdlog_debug(2,
1134 		    "ggate_recv: (%p) Waiting for request from the kernel.",
1135 		    hio);
1136 		if (ioctl(res->hr_ggatefd, G_GATE_CMD_START, ggio) < 0) {
1137 			if (sigexit_received)
1138 				pthread_exit(NULL);
1139 			primary_exit(EX_OSERR, "G_GATE_CMD_START failed");
1140 		}
1141 		error = ggio->gctl_error;
1142 		switch (error) {
1143 		case 0:
1144 			break;
1145 		case ECANCELED:
1146 			/* Exit gracefully. */
1147 			if (!sigexit_received) {
1148 				pjdlog_debug(2,
1149 				    "ggate_recv: (%p) Received cancel from the kernel.",
1150 				    hio);
1151 				pjdlog_info("Received cancel from the kernel, exiting.");
1152 			}
1153 			pthread_exit(NULL);
1154 		case ENOMEM:
1155 			/*
1156 			 * Buffer too small? Impossible, we allocate MAXPHYS
1157 			 * bytes - request can't be bigger than that.
1158 			 */
1159 			/* FALLTHROUGH */
1160 		case ENXIO:
1161 		default:
1162 			primary_exitx(EX_OSERR, "G_GATE_CMD_START failed: %s.",
1163 			    strerror(error));
1164 		}
1165 
1166 		ncomp = 0;
1167 		ncomps = HAST_NCOMPONENTS;
1168 
1169 		for (ii = 0; ii < ncomps; ii++)
1170 			hio->hio_errors[ii] = EINVAL;
1171 		reqlog(LOG_DEBUG, 2, ggio,
1172 		    "ggate_recv: (%p) Request received from the kernel: ",
1173 		    hio);
1174 
1175 		/*
1176 		 * Inform all components about new write request.
1177 		 * For read request prefer local component unless the given
1178 		 * range is out-of-date, then use remote component.
1179 		 */
1180 		switch (ggio->gctl_cmd) {
1181 		case BIO_READ:
1182 			res->hr_stat_read++;
1183 			ncomps = 1;
1184 			mtx_lock(&metadata_lock);
1185 			if (res->hr_syncsrc == HAST_SYNCSRC_UNDEF ||
1186 			    res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1187 				/*
1188 				 * This range is up-to-date on local component,
1189 				 * so handle request locally.
1190 				 */
1191 				 /* Local component is 0 for now. */
1192 				ncomp = 0;
1193 			} else /* if (res->hr_syncsrc ==
1194 			    HAST_SYNCSRC_SECONDARY) */ {
1195 				PJDLOG_ASSERT(res->hr_syncsrc ==
1196 				    HAST_SYNCSRC_SECONDARY);
1197 				/*
1198 				 * This range is out-of-date on local component,
1199 				 * so send request to the remote node.
1200 				 */
1201 				 /* Remote component is 1 for now. */
1202 				ncomp = 1;
1203 			}
1204 			mtx_unlock(&metadata_lock);
1205 			break;
1206 		case BIO_WRITE:
1207 			res->hr_stat_write++;
1208 			if (res->hr_resuid == 0 &&
1209 			    res->hr_primary_localcnt == 0) {
1210 				/* This is first write. */
1211 				res->hr_primary_localcnt = 1;
1212 			}
1213 			for (;;) {
1214 				mtx_lock(&range_lock);
1215 				if (rangelock_islocked(range_sync,
1216 				    ggio->gctl_offset, ggio->gctl_length)) {
1217 					pjdlog_debug(2,
1218 					    "regular: Range offset=%jd length=%zu locked.",
1219 					    (intmax_t)ggio->gctl_offset,
1220 					    (size_t)ggio->gctl_length);
1221 					range_regular_wait = true;
1222 					cv_wait(&range_regular_cond, &range_lock);
1223 					range_regular_wait = false;
1224 					mtx_unlock(&range_lock);
1225 					continue;
1226 				}
1227 				if (rangelock_add(range_regular,
1228 				    ggio->gctl_offset, ggio->gctl_length) < 0) {
1229 					mtx_unlock(&range_lock);
1230 					pjdlog_debug(2,
1231 					    "regular: Range offset=%jd length=%zu is already locked, waiting.",
1232 					    (intmax_t)ggio->gctl_offset,
1233 					    (size_t)ggio->gctl_length);
1234 					sleep(1);
1235 					continue;
1236 				}
1237 				mtx_unlock(&range_lock);
1238 				break;
1239 			}
1240 			mtx_lock(&res->hr_amp_lock);
1241 			if (activemap_write_start(res->hr_amp,
1242 			    ggio->gctl_offset, ggio->gctl_length)) {
1243 				res->hr_stat_activemap_update++;
1244 				(void)hast_activemap_flush(res);
1245 			}
1246 			mtx_unlock(&res->hr_amp_lock);
1247 			break;
1248 		case BIO_DELETE:
1249 			res->hr_stat_delete++;
1250 			break;
1251 		case BIO_FLUSH:
1252 			res->hr_stat_flush++;
1253 			break;
1254 		}
1255 		pjdlog_debug(2,
1256 		    "ggate_recv: (%p) Moving request to the send queues.", hio);
1257 		refcount_init(&hio->hio_countdown, ncomps);
1258 		for (ii = ncomp; ii < ncomps; ii++)
1259 			QUEUE_INSERT1(hio, send, ii);
1260 	}
1261 	/* NOTREACHED */
1262 	return (NULL);
1263 }
1264 
1265 /*
1266  * Thread reads from or writes to local component.
1267  * If local read fails, it redirects it to remote_send thread.
1268  */
1269 static void *
1270 local_send_thread(void *arg)
1271 {
1272 	struct hast_resource *res = arg;
1273 	struct g_gate_ctl_io *ggio;
1274 	struct hio *hio;
1275 	unsigned int ncomp, rncomp;
1276 	ssize_t ret;
1277 
1278 	/* Local component is 0 for now. */
1279 	ncomp = 0;
1280 	/* Remote component is 1 for now. */
1281 	rncomp = 1;
1282 
1283 	for (;;) {
1284 		pjdlog_debug(2, "local_send: Taking request.");
1285 		QUEUE_TAKE1(hio, send, ncomp, 0);
1286 		pjdlog_debug(2, "local_send: (%p) Got request.", hio);
1287 		ggio = &hio->hio_ggio;
1288 		switch (ggio->gctl_cmd) {
1289 		case BIO_READ:
1290 			ret = pread(res->hr_localfd, ggio->gctl_data,
1291 			    ggio->gctl_length,
1292 			    ggio->gctl_offset + res->hr_localoff);
1293 			if (ret == ggio->gctl_length)
1294 				hio->hio_errors[ncomp] = 0;
1295 			else if (!ISSYNCREQ(hio)) {
1296 				/*
1297 				 * If READ failed, try to read from remote node.
1298 				 */
1299 				if (ret < 0) {
1300 					reqlog(LOG_WARNING, 0, ggio,
1301 					    "Local request failed (%s), trying remote node. ",
1302 					    strerror(errno));
1303 				} else if (ret != ggio->gctl_length) {
1304 					reqlog(LOG_WARNING, 0, ggio,
1305 					    "Local request failed (%zd != %jd), trying remote node. ",
1306 					    ret, (intmax_t)ggio->gctl_length);
1307 				}
1308 				QUEUE_INSERT1(hio, send, rncomp);
1309 				continue;
1310 			}
1311 			break;
1312 		case BIO_WRITE:
1313 			ret = pwrite(res->hr_localfd, ggio->gctl_data,
1314 			    ggio->gctl_length,
1315 			    ggio->gctl_offset + res->hr_localoff);
1316 			if (ret < 0) {
1317 				hio->hio_errors[ncomp] = errno;
1318 				reqlog(LOG_WARNING, 0, ggio,
1319 				    "Local request failed (%s): ",
1320 				    strerror(errno));
1321 			} else if (ret != ggio->gctl_length) {
1322 				hio->hio_errors[ncomp] = EIO;
1323 				reqlog(LOG_WARNING, 0, ggio,
1324 				    "Local request failed (%zd != %jd): ",
1325 				    ret, (intmax_t)ggio->gctl_length);
1326 			} else {
1327 				hio->hio_errors[ncomp] = 0;
1328 				if (hio->hio_replication ==
1329 				    HAST_REPLICATION_ASYNC) {
1330 					ggio->gctl_error = 0;
1331 					write_complete(res, hio);
1332 				}
1333 			}
1334 			break;
1335 		case BIO_DELETE:
1336 			ret = g_delete(res->hr_localfd,
1337 			    ggio->gctl_offset + res->hr_localoff,
1338 			    ggio->gctl_length);
1339 			if (ret < 0) {
1340 				hio->hio_errors[ncomp] = errno;
1341 				reqlog(LOG_WARNING, 0, ggio,
1342 				    "Local request failed (%s): ",
1343 				    strerror(errno));
1344 			} else {
1345 				hio->hio_errors[ncomp] = 0;
1346 			}
1347 			break;
1348 		case BIO_FLUSH:
1349 			if (!res->hr_localflush) {
1350 				ret = -1;
1351 				errno = EOPNOTSUPP;
1352 				break;
1353 			}
1354 			ret = g_flush(res->hr_localfd);
1355 			if (ret < 0) {
1356 				if (errno == EOPNOTSUPP)
1357 					res->hr_localflush = false;
1358 				hio->hio_errors[ncomp] = errno;
1359 				reqlog(LOG_WARNING, 0, ggio,
1360 				    "Local request failed (%s): ",
1361 				    strerror(errno));
1362 			} else {
1363 				hio->hio_errors[ncomp] = 0;
1364 			}
1365 			break;
1366 		}
1367 		if (!refcount_release(&hio->hio_countdown))
1368 			continue;
1369 		if (ISSYNCREQ(hio)) {
1370 			mtx_lock(&sync_lock);
1371 			SYNCREQDONE(hio);
1372 			mtx_unlock(&sync_lock);
1373 			cv_signal(&sync_cond);
1374 		} else {
1375 			pjdlog_debug(2,
1376 			    "local_send: (%p) Moving request to the done queue.",
1377 			    hio);
1378 			QUEUE_INSERT2(hio, done);
1379 		}
1380 	}
1381 	/* NOTREACHED */
1382 	return (NULL);
1383 }
1384 
1385 static void
1386 keepalive_send(struct hast_resource *res, unsigned int ncomp)
1387 {
1388 	struct nv *nv;
1389 
1390 	rw_rlock(&hio_remote_lock[ncomp]);
1391 
1392 	if (!ISCONNECTED(res, ncomp)) {
1393 		rw_unlock(&hio_remote_lock[ncomp]);
1394 		return;
1395 	}
1396 
1397 	PJDLOG_ASSERT(res->hr_remotein != NULL);
1398 	PJDLOG_ASSERT(res->hr_remoteout != NULL);
1399 
1400 	nv = nv_alloc();
1401 	nv_add_uint8(nv, HIO_KEEPALIVE, "cmd");
1402 	if (nv_error(nv) != 0) {
1403 		rw_unlock(&hio_remote_lock[ncomp]);
1404 		nv_free(nv);
1405 		pjdlog_debug(1,
1406 		    "keepalive_send: Unable to prepare header to send.");
1407 		return;
1408 	}
1409 	if (hast_proto_send(res, res->hr_remoteout, nv, NULL, 0) < 0) {
1410 		rw_unlock(&hio_remote_lock[ncomp]);
1411 		pjdlog_common(LOG_DEBUG, 1, errno,
1412 		    "keepalive_send: Unable to send request");
1413 		nv_free(nv);
1414 		remote_close(res, ncomp);
1415 		return;
1416 	}
1417 
1418 	rw_unlock(&hio_remote_lock[ncomp]);
1419 	nv_free(nv);
1420 	pjdlog_debug(2, "keepalive_send: Request sent.");
1421 }
1422 
1423 /*
1424  * Thread sends request to secondary node.
1425  */
1426 static void *
1427 remote_send_thread(void *arg)
1428 {
1429 	struct hast_resource *res = arg;
1430 	struct g_gate_ctl_io *ggio;
1431 	time_t lastcheck, now;
1432 	struct hio *hio;
1433 	struct nv *nv;
1434 	unsigned int ncomp;
1435 	bool wakeup;
1436 	uint64_t offset, length;
1437 	uint8_t cmd;
1438 	void *data;
1439 
1440 	/* Remote component is 1 for now. */
1441 	ncomp = 1;
1442 	lastcheck = time(NULL);
1443 
1444 	for (;;) {
1445 		pjdlog_debug(2, "remote_send: Taking request.");
1446 		QUEUE_TAKE1(hio, send, ncomp, HAST_KEEPALIVE);
1447 		if (hio == NULL) {
1448 			now = time(NULL);
1449 			if (lastcheck + HAST_KEEPALIVE <= now) {
1450 				keepalive_send(res, ncomp);
1451 				lastcheck = now;
1452 			}
1453 			continue;
1454 		}
1455 		pjdlog_debug(2, "remote_send: (%p) Got request.", hio);
1456 		ggio = &hio->hio_ggio;
1457 		switch (ggio->gctl_cmd) {
1458 		case BIO_READ:
1459 			cmd = HIO_READ;
1460 			data = NULL;
1461 			offset = ggio->gctl_offset;
1462 			length = ggio->gctl_length;
1463 			break;
1464 		case BIO_WRITE:
1465 			cmd = HIO_WRITE;
1466 			data = ggio->gctl_data;
1467 			offset = ggio->gctl_offset;
1468 			length = ggio->gctl_length;
1469 			break;
1470 		case BIO_DELETE:
1471 			cmd = HIO_DELETE;
1472 			data = NULL;
1473 			offset = ggio->gctl_offset;
1474 			length = ggio->gctl_length;
1475 			break;
1476 		case BIO_FLUSH:
1477 			cmd = HIO_FLUSH;
1478 			data = NULL;
1479 			offset = 0;
1480 			length = 0;
1481 			break;
1482 		default:
1483 			PJDLOG_ABORT("invalid condition");
1484 		}
1485 		nv = nv_alloc();
1486 		nv_add_uint8(nv, cmd, "cmd");
1487 		nv_add_uint64(nv, (uint64_t)ggio->gctl_seq, "seq");
1488 		nv_add_uint64(nv, offset, "offset");
1489 		nv_add_uint64(nv, length, "length");
1490 		if (nv_error(nv) != 0) {
1491 			hio->hio_errors[ncomp] = nv_error(nv);
1492 			pjdlog_debug(2,
1493 			    "remote_send: (%p) Unable to prepare header to send.",
1494 			    hio);
1495 			reqlog(LOG_ERR, 0, ggio,
1496 			    "Unable to prepare header to send (%s): ",
1497 			    strerror(nv_error(nv)));
1498 			/* Move failed request immediately to the done queue. */
1499 			goto done_queue;
1500 		}
1501 		/*
1502 		 * Protect connection from disappearing.
1503 		 */
1504 		rw_rlock(&hio_remote_lock[ncomp]);
1505 		if (!ISCONNECTED(res, ncomp)) {
1506 			rw_unlock(&hio_remote_lock[ncomp]);
1507 			hio->hio_errors[ncomp] = ENOTCONN;
1508 			goto done_queue;
1509 		}
1510 		/*
1511 		 * Move the request to recv queue before sending it, because
1512 		 * in different order we can get reply before we move request
1513 		 * to recv queue.
1514 		 */
1515 		pjdlog_debug(2,
1516 		    "remote_send: (%p) Moving request to the recv queue.",
1517 		    hio);
1518 		mtx_lock(&hio_recv_list_lock[ncomp]);
1519 		wakeup = TAILQ_EMPTY(&hio_recv_list[ncomp]);
1520 		TAILQ_INSERT_TAIL(&hio_recv_list[ncomp], hio, hio_next[ncomp]);
1521 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1522 		if (hast_proto_send(res, res->hr_remoteout, nv, data,
1523 		    data != NULL ? length : 0) < 0) {
1524 			hio->hio_errors[ncomp] = errno;
1525 			rw_unlock(&hio_remote_lock[ncomp]);
1526 			pjdlog_debug(2,
1527 			    "remote_send: (%p) Unable to send request.", hio);
1528 			reqlog(LOG_ERR, 0, ggio,
1529 			    "Unable to send request (%s): ",
1530 			    strerror(hio->hio_errors[ncomp]));
1531 			remote_close(res, ncomp);
1532 			/*
1533 			 * Take request back from the receive queue and move
1534 			 * it immediately to the done queue.
1535 			 */
1536 			mtx_lock(&hio_recv_list_lock[ncomp]);
1537 			TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1538 			    hio_next[ncomp]);
1539 			mtx_unlock(&hio_recv_list_lock[ncomp]);
1540 			goto done_queue;
1541 		}
1542 		rw_unlock(&hio_remote_lock[ncomp]);
1543 		nv_free(nv);
1544 		if (wakeup)
1545 			cv_signal(&hio_recv_list_cond[ncomp]);
1546 		continue;
1547 done_queue:
1548 		nv_free(nv);
1549 		if (ISSYNCREQ(hio)) {
1550 			if (!refcount_release(&hio->hio_countdown))
1551 				continue;
1552 			mtx_lock(&sync_lock);
1553 			SYNCREQDONE(hio);
1554 			mtx_unlock(&sync_lock);
1555 			cv_signal(&sync_cond);
1556 			continue;
1557 		}
1558 		if (ggio->gctl_cmd == BIO_WRITE) {
1559 			mtx_lock(&res->hr_amp_lock);
1560 			if (activemap_need_sync(res->hr_amp, ggio->gctl_offset,
1561 			    ggio->gctl_length)) {
1562 				(void)hast_activemap_flush(res);
1563 			}
1564 			mtx_unlock(&res->hr_amp_lock);
1565 		}
1566 		if (!refcount_release(&hio->hio_countdown))
1567 			continue;
1568 		pjdlog_debug(2,
1569 		    "remote_send: (%p) Moving request to the done queue.",
1570 		    hio);
1571 		QUEUE_INSERT2(hio, done);
1572 	}
1573 	/* NOTREACHED */
1574 	return (NULL);
1575 }
1576 
1577 /*
1578  * Thread receives answer from secondary node and passes it to ggate_send
1579  * thread.
1580  */
1581 static void *
1582 remote_recv_thread(void *arg)
1583 {
1584 	struct hast_resource *res = arg;
1585 	struct g_gate_ctl_io *ggio;
1586 	struct hio *hio;
1587 	struct nv *nv;
1588 	unsigned int ncomp;
1589 	uint64_t seq;
1590 	int error;
1591 
1592 	/* Remote component is 1 for now. */
1593 	ncomp = 1;
1594 
1595 	for (;;) {
1596 		/* Wait until there is anything to receive. */
1597 		mtx_lock(&hio_recv_list_lock[ncomp]);
1598 		while (TAILQ_EMPTY(&hio_recv_list[ncomp])) {
1599 			pjdlog_debug(2, "remote_recv: No requests, waiting.");
1600 			cv_wait(&hio_recv_list_cond[ncomp],
1601 			    &hio_recv_list_lock[ncomp]);
1602 		}
1603 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1604 
1605 		rw_rlock(&hio_remote_lock[ncomp]);
1606 		if (!ISCONNECTED(res, ncomp)) {
1607 			rw_unlock(&hio_remote_lock[ncomp]);
1608 			/*
1609 			 * Connection is dead, so move all pending requests to
1610 			 * the done queue (one-by-one).
1611 			 */
1612 			mtx_lock(&hio_recv_list_lock[ncomp]);
1613 			hio = TAILQ_FIRST(&hio_recv_list[ncomp]);
1614 			PJDLOG_ASSERT(hio != NULL);
1615 			TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1616 			    hio_next[ncomp]);
1617 			mtx_unlock(&hio_recv_list_lock[ncomp]);
1618 			goto done_queue;
1619 		}
1620 		if (hast_proto_recv_hdr(res->hr_remotein, &nv) < 0) {
1621 			pjdlog_errno(LOG_ERR,
1622 			    "Unable to receive reply header");
1623 			rw_unlock(&hio_remote_lock[ncomp]);
1624 			remote_close(res, ncomp);
1625 			continue;
1626 		}
1627 		rw_unlock(&hio_remote_lock[ncomp]);
1628 		seq = nv_get_uint64(nv, "seq");
1629 		if (seq == 0) {
1630 			pjdlog_error("Header contains no 'seq' field.");
1631 			nv_free(nv);
1632 			continue;
1633 		}
1634 		mtx_lock(&hio_recv_list_lock[ncomp]);
1635 		TAILQ_FOREACH(hio, &hio_recv_list[ncomp], hio_next[ncomp]) {
1636 			if (hio->hio_ggio.gctl_seq == seq) {
1637 				TAILQ_REMOVE(&hio_recv_list[ncomp], hio,
1638 				    hio_next[ncomp]);
1639 				break;
1640 			}
1641 		}
1642 		mtx_unlock(&hio_recv_list_lock[ncomp]);
1643 		if (hio == NULL) {
1644 			pjdlog_error("Found no request matching received 'seq' field (%ju).",
1645 			    (uintmax_t)seq);
1646 			nv_free(nv);
1647 			continue;
1648 		}
1649 		ggio = &hio->hio_ggio;
1650 		error = nv_get_int16(nv, "error");
1651 		if (error != 0) {
1652 			/* Request failed on remote side. */
1653 			hio->hio_errors[ncomp] = error;
1654 			reqlog(LOG_WARNING, 0, ggio,
1655 			    "Remote request failed (%s): ", strerror(error));
1656 			nv_free(nv);
1657 			goto done_queue;
1658 		}
1659 		switch (ggio->gctl_cmd) {
1660 		case BIO_READ:
1661 			rw_rlock(&hio_remote_lock[ncomp]);
1662 			if (!ISCONNECTED(res, ncomp)) {
1663 				rw_unlock(&hio_remote_lock[ncomp]);
1664 				nv_free(nv);
1665 				goto done_queue;
1666 			}
1667 			if (hast_proto_recv_data(res, res->hr_remotein, nv,
1668 			    ggio->gctl_data, ggio->gctl_length) < 0) {
1669 				hio->hio_errors[ncomp] = errno;
1670 				pjdlog_errno(LOG_ERR,
1671 				    "Unable to receive reply data");
1672 				rw_unlock(&hio_remote_lock[ncomp]);
1673 				nv_free(nv);
1674 				remote_close(res, ncomp);
1675 				goto done_queue;
1676 			}
1677 			rw_unlock(&hio_remote_lock[ncomp]);
1678 			break;
1679 		case BIO_WRITE:
1680 		case BIO_DELETE:
1681 		case BIO_FLUSH:
1682 			break;
1683 		default:
1684 			PJDLOG_ABORT("invalid condition");
1685 		}
1686 		hio->hio_errors[ncomp] = 0;
1687 		nv_free(nv);
1688 done_queue:
1689 		if (!refcount_release(&hio->hio_countdown))
1690 			continue;
1691 		if (ISSYNCREQ(hio)) {
1692 			mtx_lock(&sync_lock);
1693 			SYNCREQDONE(hio);
1694 			mtx_unlock(&sync_lock);
1695 			cv_signal(&sync_cond);
1696 		} else {
1697 			pjdlog_debug(2,
1698 			    "remote_recv: (%p) Moving request to the done queue.",
1699 			    hio);
1700 			QUEUE_INSERT2(hio, done);
1701 		}
1702 	}
1703 	/* NOTREACHED */
1704 	return (NULL);
1705 }
1706 
1707 /*
1708  * Thread sends answer to the kernel.
1709  */
1710 static void *
1711 ggate_send_thread(void *arg)
1712 {
1713 	struct hast_resource *res = arg;
1714 	struct g_gate_ctl_io *ggio;
1715 	struct hio *hio;
1716 	unsigned int ii, ncomps;
1717 
1718 	ncomps = HAST_NCOMPONENTS;
1719 
1720 	for (;;) {
1721 		pjdlog_debug(2, "ggate_send: Taking request.");
1722 		QUEUE_TAKE2(hio, done);
1723 		pjdlog_debug(2, "ggate_send: (%p) Got request.", hio);
1724 		ggio = &hio->hio_ggio;
1725 		for (ii = 0; ii < ncomps; ii++) {
1726 			if (hio->hio_errors[ii] == 0) {
1727 				/*
1728 				 * One successful request is enough to declare
1729 				 * success.
1730 				 */
1731 				ggio->gctl_error = 0;
1732 				break;
1733 			}
1734 		}
1735 		if (ii == ncomps) {
1736 			/*
1737 			 * None of the requests were successful.
1738 			 * Use the error from local component except the
1739 			 * case when we did only remote request.
1740 			 */
1741 			if (ggio->gctl_cmd == BIO_READ &&
1742 			    res->hr_syncsrc == HAST_SYNCSRC_SECONDARY)
1743 				ggio->gctl_error = hio->hio_errors[1];
1744 			else
1745 				ggio->gctl_error = hio->hio_errors[0];
1746 		}
1747 		if (ggio->gctl_error == 0 && ggio->gctl_cmd == BIO_WRITE) {
1748 			mtx_lock(&res->hr_amp_lock);
1749 			if (activemap_write_complete(res->hr_amp,
1750 			    ggio->gctl_offset, ggio->gctl_length)) {
1751 				res->hr_stat_activemap_update++;
1752 				(void)hast_activemap_flush(res);
1753 			}
1754 			mtx_unlock(&res->hr_amp_lock);
1755 		}
1756 		if (ggio->gctl_cmd == BIO_WRITE) {
1757 			/*
1758 			 * Unlock range we locked.
1759 			 */
1760 			mtx_lock(&range_lock);
1761 			rangelock_del(range_regular, ggio->gctl_offset,
1762 			    ggio->gctl_length);
1763 			if (range_sync_wait)
1764 				cv_signal(&range_sync_cond);
1765 			mtx_unlock(&range_lock);
1766 			if (!hio->hio_done)
1767 				write_complete(res, hio);
1768 		} else {
1769 			if (ioctl(res->hr_ggatefd, G_GATE_CMD_DONE, ggio) < 0) {
1770 				primary_exit(EX_OSERR,
1771 				    "G_GATE_CMD_DONE failed");
1772 			}
1773 		}
1774 		pjdlog_debug(2,
1775 		    "ggate_send: (%p) Moving request to the free queue.", hio);
1776 		QUEUE_INSERT2(hio, free);
1777 	}
1778 	/* NOTREACHED */
1779 	return (NULL);
1780 }
1781 
1782 /*
1783  * Thread synchronize local and remote components.
1784  */
1785 static void *
1786 sync_thread(void *arg __unused)
1787 {
1788 	struct hast_resource *res = arg;
1789 	struct hio *hio;
1790 	struct g_gate_ctl_io *ggio;
1791 	struct timeval tstart, tend, tdiff;
1792 	unsigned int ii, ncomp, ncomps;
1793 	off_t offset, length, synced;
1794 	bool dorewind;
1795 	int syncext;
1796 
1797 	ncomps = HAST_NCOMPONENTS;
1798 	dorewind = true;
1799 	synced = 0;
1800 	offset = -1;
1801 
1802 	for (;;) {
1803 		mtx_lock(&sync_lock);
1804 		if (offset >= 0 && !sync_inprogress) {
1805 			gettimeofday(&tend, NULL);
1806 			timersub(&tend, &tstart, &tdiff);
1807 			pjdlog_info("Synchronization interrupted after %#.0T. "
1808 			    "%NB synchronized so far.", &tdiff,
1809 			    (intmax_t)synced);
1810 			event_send(res, EVENT_SYNCINTR);
1811 		}
1812 		while (!sync_inprogress) {
1813 			dorewind = true;
1814 			synced = 0;
1815 			cv_wait(&sync_cond, &sync_lock);
1816 		}
1817 		mtx_unlock(&sync_lock);
1818 		/*
1819 		 * Obtain offset at which we should synchronize.
1820 		 * Rewind synchronization if needed.
1821 		 */
1822 		mtx_lock(&res->hr_amp_lock);
1823 		if (dorewind)
1824 			activemap_sync_rewind(res->hr_amp);
1825 		offset = activemap_sync_offset(res->hr_amp, &length, &syncext);
1826 		if (syncext != -1) {
1827 			/*
1828 			 * We synchronized entire syncext extent, we can mark
1829 			 * it as clean now.
1830 			 */
1831 			if (activemap_extent_complete(res->hr_amp, syncext))
1832 				(void)hast_activemap_flush(res);
1833 		}
1834 		mtx_unlock(&res->hr_amp_lock);
1835 		if (dorewind) {
1836 			dorewind = false;
1837 			if (offset < 0)
1838 				pjdlog_info("Nodes are in sync.");
1839 			else {
1840 				pjdlog_info("Synchronization started. %NB to go.",
1841 				    (intmax_t)(res->hr_extentsize *
1842 				    activemap_ndirty(res->hr_amp)));
1843 				event_send(res, EVENT_SYNCSTART);
1844 				gettimeofday(&tstart, NULL);
1845 			}
1846 		}
1847 		if (offset < 0) {
1848 			sync_stop();
1849 			pjdlog_debug(1, "Nothing to synchronize.");
1850 			/*
1851 			 * Synchronization complete, make both localcnt and
1852 			 * remotecnt equal.
1853 			 */
1854 			ncomp = 1;
1855 			rw_rlock(&hio_remote_lock[ncomp]);
1856 			if (ISCONNECTED(res, ncomp)) {
1857 				if (synced > 0) {
1858 					int64_t bps;
1859 
1860 					gettimeofday(&tend, NULL);
1861 					timersub(&tend, &tstart, &tdiff);
1862 					bps = (int64_t)((double)synced /
1863 					    ((double)tdiff.tv_sec +
1864 					    (double)tdiff.tv_usec / 1000000));
1865 					pjdlog_info("Synchronization complete. "
1866 					    "%NB synchronized in %#.0lT (%NB/sec).",
1867 					    (intmax_t)synced, &tdiff,
1868 					    (intmax_t)bps);
1869 					event_send(res, EVENT_SYNCDONE);
1870 				}
1871 				mtx_lock(&metadata_lock);
1872 				res->hr_syncsrc = HAST_SYNCSRC_UNDEF;
1873 				res->hr_primary_localcnt =
1874 				    res->hr_secondary_remotecnt;
1875 				res->hr_primary_remotecnt =
1876 				    res->hr_secondary_localcnt;
1877 				pjdlog_debug(1,
1878 				    "Setting localcnt to %ju and remotecnt to %ju.",
1879 				    (uintmax_t)res->hr_primary_localcnt,
1880 				    (uintmax_t)res->hr_primary_remotecnt);
1881 				(void)metadata_write(res);
1882 				mtx_unlock(&metadata_lock);
1883 			}
1884 			rw_unlock(&hio_remote_lock[ncomp]);
1885 			continue;
1886 		}
1887 		pjdlog_debug(2, "sync: Taking free request.");
1888 		QUEUE_TAKE2(hio, free);
1889 		pjdlog_debug(2, "sync: (%p) Got free request.", hio);
1890 		/*
1891 		 * Lock the range we are going to synchronize. We don't want
1892 		 * race where someone writes between our read and write.
1893 		 */
1894 		for (;;) {
1895 			mtx_lock(&range_lock);
1896 			if (rangelock_islocked(range_regular, offset, length)) {
1897 				pjdlog_debug(2,
1898 				    "sync: Range offset=%jd length=%jd locked.",
1899 				    (intmax_t)offset, (intmax_t)length);
1900 				range_sync_wait = true;
1901 				cv_wait(&range_sync_cond, &range_lock);
1902 				range_sync_wait = false;
1903 				mtx_unlock(&range_lock);
1904 				continue;
1905 			}
1906 			if (rangelock_add(range_sync, offset, length) < 0) {
1907 				mtx_unlock(&range_lock);
1908 				pjdlog_debug(2,
1909 				    "sync: Range offset=%jd length=%jd is already locked, waiting.",
1910 				    (intmax_t)offset, (intmax_t)length);
1911 				sleep(1);
1912 				continue;
1913 			}
1914 			mtx_unlock(&range_lock);
1915 			break;
1916 		}
1917 		/*
1918 		 * First read the data from synchronization source.
1919 		 */
1920 		SYNCREQ(hio);
1921 		ggio = &hio->hio_ggio;
1922 		ggio->gctl_cmd = BIO_READ;
1923 		ggio->gctl_offset = offset;
1924 		ggio->gctl_length = length;
1925 		ggio->gctl_error = 0;
1926 		hio->hio_done = false;
1927 		hio->hio_replication = res->hr_replication;
1928 		for (ii = 0; ii < ncomps; ii++)
1929 			hio->hio_errors[ii] = EINVAL;
1930 		reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1931 		    hio);
1932 		pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1933 		    hio);
1934 		mtx_lock(&metadata_lock);
1935 		if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1936 			/*
1937 			 * This range is up-to-date on local component,
1938 			 * so handle request locally.
1939 			 */
1940 			 /* Local component is 0 for now. */
1941 			ncomp = 0;
1942 		} else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1943 			PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1944 			/*
1945 			 * This range is out-of-date on local component,
1946 			 * so send request to the remote node.
1947 			 */
1948 			 /* Remote component is 1 for now. */
1949 			ncomp = 1;
1950 		}
1951 		mtx_unlock(&metadata_lock);
1952 		refcount_init(&hio->hio_countdown, 1);
1953 		QUEUE_INSERT1(hio, send, ncomp);
1954 
1955 		/*
1956 		 * Let's wait for READ to finish.
1957 		 */
1958 		mtx_lock(&sync_lock);
1959 		while (!ISSYNCREQDONE(hio))
1960 			cv_wait(&sync_cond, &sync_lock);
1961 		mtx_unlock(&sync_lock);
1962 
1963 		if (hio->hio_errors[ncomp] != 0) {
1964 			pjdlog_error("Unable to read synchronization data: %s.",
1965 			    strerror(hio->hio_errors[ncomp]));
1966 			goto free_queue;
1967 		}
1968 
1969 		/*
1970 		 * We read the data from synchronization source, now write it
1971 		 * to synchronization target.
1972 		 */
1973 		SYNCREQ(hio);
1974 		ggio->gctl_cmd = BIO_WRITE;
1975 		for (ii = 0; ii < ncomps; ii++)
1976 			hio->hio_errors[ii] = EINVAL;
1977 		reqlog(LOG_DEBUG, 2, ggio, "sync: (%p) Sending sync request: ",
1978 		    hio);
1979 		pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
1980 		    hio);
1981 		mtx_lock(&metadata_lock);
1982 		if (res->hr_syncsrc == HAST_SYNCSRC_PRIMARY) {
1983 			/*
1984 			 * This range is up-to-date on local component,
1985 			 * so we update remote component.
1986 			 */
1987 			 /* Remote component is 1 for now. */
1988 			ncomp = 1;
1989 		} else /* if (res->hr_syncsrc == HAST_SYNCSRC_SECONDARY) */ {
1990 			PJDLOG_ASSERT(res->hr_syncsrc == HAST_SYNCSRC_SECONDARY);
1991 			/*
1992 			 * This range is out-of-date on local component,
1993 			 * so we update it.
1994 			 */
1995 			 /* Local component is 0 for now. */
1996 			ncomp = 0;
1997 		}
1998 		mtx_unlock(&metadata_lock);
1999 
2000 		pjdlog_debug(2, "sync: (%p) Moving request to the send queue.",
2001 		    hio);
2002 		refcount_init(&hio->hio_countdown, 1);
2003 		QUEUE_INSERT1(hio, send, ncomp);
2004 
2005 		/*
2006 		 * Let's wait for WRITE to finish.
2007 		 */
2008 		mtx_lock(&sync_lock);
2009 		while (!ISSYNCREQDONE(hio))
2010 			cv_wait(&sync_cond, &sync_lock);
2011 		mtx_unlock(&sync_lock);
2012 
2013 		if (hio->hio_errors[ncomp] != 0) {
2014 			pjdlog_error("Unable to write synchronization data: %s.",
2015 			    strerror(hio->hio_errors[ncomp]));
2016 			goto free_queue;
2017 		}
2018 
2019 		synced += length;
2020 free_queue:
2021 		mtx_lock(&range_lock);
2022 		rangelock_del(range_sync, offset, length);
2023 		if (range_regular_wait)
2024 			cv_signal(&range_regular_cond);
2025 		mtx_unlock(&range_lock);
2026 		pjdlog_debug(2, "sync: (%p) Moving request to the free queue.",
2027 		    hio);
2028 		QUEUE_INSERT2(hio, free);
2029 	}
2030 	/* NOTREACHED */
2031 	return (NULL);
2032 }
2033 
2034 void
2035 primary_config_reload(struct hast_resource *res, struct nv *nv)
2036 {
2037 	unsigned int ii, ncomps;
2038 	int modified, vint;
2039 	const char *vstr;
2040 
2041 	pjdlog_info("Reloading configuration...");
2042 
2043 	PJDLOG_ASSERT(res->hr_role == HAST_ROLE_PRIMARY);
2044 	PJDLOG_ASSERT(gres == res);
2045 	nv_assert(nv, "remoteaddr");
2046 	nv_assert(nv, "sourceaddr");
2047 	nv_assert(nv, "replication");
2048 	nv_assert(nv, "checksum");
2049 	nv_assert(nv, "compression");
2050 	nv_assert(nv, "timeout");
2051 	nv_assert(nv, "exec");
2052 	nv_assert(nv, "metaflush");
2053 
2054 	ncomps = HAST_NCOMPONENTS;
2055 
2056 #define MODIFIED_REMOTEADDR	0x01
2057 #define MODIFIED_SOURCEADDR	0x02
2058 #define MODIFIED_REPLICATION	0x04
2059 #define MODIFIED_CHECKSUM	0x08
2060 #define MODIFIED_COMPRESSION	0x10
2061 #define MODIFIED_TIMEOUT	0x20
2062 #define MODIFIED_EXEC		0x40
2063 #define MODIFIED_METAFLUSH	0x80
2064 	modified = 0;
2065 
2066 	vstr = nv_get_string(nv, "remoteaddr");
2067 	if (strcmp(gres->hr_remoteaddr, vstr) != 0) {
2068 		/*
2069 		 * Don't copy res->hr_remoteaddr to gres just yet.
2070 		 * We want remote_close() to log disconnect from the old
2071 		 * addresses, not from the new ones.
2072 		 */
2073 		modified |= MODIFIED_REMOTEADDR;
2074 	}
2075 	vstr = nv_get_string(nv, "sourceaddr");
2076 	if (strcmp(gres->hr_sourceaddr, vstr) != 0) {
2077 		strlcpy(gres->hr_sourceaddr, vstr, sizeof(gres->hr_sourceaddr));
2078 		modified |= MODIFIED_SOURCEADDR;
2079 	}
2080 	vint = nv_get_int32(nv, "replication");
2081 	if (gres->hr_replication != vint) {
2082 		gres->hr_replication = vint;
2083 		modified |= MODIFIED_REPLICATION;
2084 	}
2085 	vint = nv_get_int32(nv, "checksum");
2086 	if (gres->hr_checksum != vint) {
2087 		gres->hr_checksum = vint;
2088 		modified |= MODIFIED_CHECKSUM;
2089 	}
2090 	vint = nv_get_int32(nv, "compression");
2091 	if (gres->hr_compression != vint) {
2092 		gres->hr_compression = vint;
2093 		modified |= MODIFIED_COMPRESSION;
2094 	}
2095 	vint = nv_get_int32(nv, "timeout");
2096 	if (gres->hr_timeout != vint) {
2097 		gres->hr_timeout = vint;
2098 		modified |= MODIFIED_TIMEOUT;
2099 	}
2100 	vstr = nv_get_string(nv, "exec");
2101 	if (strcmp(gres->hr_exec, vstr) != 0) {
2102 		strlcpy(gres->hr_exec, vstr, sizeof(gres->hr_exec));
2103 		modified |= MODIFIED_EXEC;
2104 	}
2105 	vint = nv_get_int32(nv, "metaflush");
2106 	if (gres->hr_metaflush != vint) {
2107 		gres->hr_metaflush = vint;
2108 		modified |= MODIFIED_METAFLUSH;
2109 	}
2110 
2111 	/*
2112 	 * Change timeout for connected sockets.
2113 	 * Don't bother if we need to reconnect.
2114 	 */
2115 	if ((modified & MODIFIED_TIMEOUT) != 0 &&
2116 	    (modified & (MODIFIED_REMOTEADDR | MODIFIED_SOURCEADDR)) == 0) {
2117 		for (ii = 0; ii < ncomps; ii++) {
2118 			if (!ISREMOTE(ii))
2119 				continue;
2120 			rw_rlock(&hio_remote_lock[ii]);
2121 			if (!ISCONNECTED(gres, ii)) {
2122 				rw_unlock(&hio_remote_lock[ii]);
2123 				continue;
2124 			}
2125 			rw_unlock(&hio_remote_lock[ii]);
2126 			if (proto_timeout(gres->hr_remotein,
2127 			    gres->hr_timeout) < 0) {
2128 				pjdlog_errno(LOG_WARNING,
2129 				    "Unable to set connection timeout");
2130 			}
2131 			if (proto_timeout(gres->hr_remoteout,
2132 			    gres->hr_timeout) < 0) {
2133 				pjdlog_errno(LOG_WARNING,
2134 				    "Unable to set connection timeout");
2135 			}
2136 		}
2137 	}
2138 	if ((modified & (MODIFIED_REMOTEADDR | MODIFIED_SOURCEADDR)) != 0) {
2139 		for (ii = 0; ii < ncomps; ii++) {
2140 			if (!ISREMOTE(ii))
2141 				continue;
2142 			remote_close(gres, ii);
2143 		}
2144 		if (modified & MODIFIED_REMOTEADDR) {
2145 			vstr = nv_get_string(nv, "remoteaddr");
2146 			strlcpy(gres->hr_remoteaddr, vstr,
2147 			    sizeof(gres->hr_remoteaddr));
2148 		}
2149 	}
2150 #undef	MODIFIED_REMOTEADDR
2151 #undef	MODIFIED_SOURCEADDR
2152 #undef	MODIFIED_REPLICATION
2153 #undef	MODIFIED_CHECKSUM
2154 #undef	MODIFIED_COMPRESSION
2155 #undef	MODIFIED_TIMEOUT
2156 #undef	MODIFIED_EXEC
2157 #undef	MODIFIED_METAFLUSH
2158 
2159 	pjdlog_info("Configuration reloaded successfully.");
2160 }
2161 
2162 static void
2163 guard_one(struct hast_resource *res, unsigned int ncomp)
2164 {
2165 	struct proto_conn *in, *out;
2166 
2167 	if (!ISREMOTE(ncomp))
2168 		return;
2169 
2170 	rw_rlock(&hio_remote_lock[ncomp]);
2171 
2172 	if (!real_remote(res)) {
2173 		rw_unlock(&hio_remote_lock[ncomp]);
2174 		return;
2175 	}
2176 
2177 	if (ISCONNECTED(res, ncomp)) {
2178 		PJDLOG_ASSERT(res->hr_remotein != NULL);
2179 		PJDLOG_ASSERT(res->hr_remoteout != NULL);
2180 		rw_unlock(&hio_remote_lock[ncomp]);
2181 		pjdlog_debug(2, "remote_guard: Connection to %s is ok.",
2182 		    res->hr_remoteaddr);
2183 		return;
2184 	}
2185 
2186 	PJDLOG_ASSERT(res->hr_remotein == NULL);
2187 	PJDLOG_ASSERT(res->hr_remoteout == NULL);
2188 	/*
2189 	 * Upgrade the lock. It doesn't have to be atomic as no other thread
2190 	 * can change connection status from disconnected to connected.
2191 	 */
2192 	rw_unlock(&hio_remote_lock[ncomp]);
2193 	pjdlog_debug(2, "remote_guard: Reconnecting to %s.",
2194 	    res->hr_remoteaddr);
2195 	in = out = NULL;
2196 	if (init_remote(res, &in, &out) == 0) {
2197 		rw_wlock(&hio_remote_lock[ncomp]);
2198 		PJDLOG_ASSERT(res->hr_remotein == NULL);
2199 		PJDLOG_ASSERT(res->hr_remoteout == NULL);
2200 		PJDLOG_ASSERT(in != NULL && out != NULL);
2201 		res->hr_remotein = in;
2202 		res->hr_remoteout = out;
2203 		rw_unlock(&hio_remote_lock[ncomp]);
2204 		pjdlog_info("Successfully reconnected to %s.",
2205 		    res->hr_remoteaddr);
2206 		sync_start();
2207 	} else {
2208 		/* Both connections should be NULL. */
2209 		PJDLOG_ASSERT(res->hr_remotein == NULL);
2210 		PJDLOG_ASSERT(res->hr_remoteout == NULL);
2211 		PJDLOG_ASSERT(in == NULL && out == NULL);
2212 		pjdlog_debug(2, "remote_guard: Reconnect to %s failed.",
2213 		    res->hr_remoteaddr);
2214 	}
2215 }
2216 
2217 /*
2218  * Thread guards remote connections and reconnects when needed, handles
2219  * signals, etc.
2220  */
2221 static void *
2222 guard_thread(void *arg)
2223 {
2224 	struct hast_resource *res = arg;
2225 	unsigned int ii, ncomps;
2226 	struct timespec timeout;
2227 	time_t lastcheck, now;
2228 	sigset_t mask;
2229 	int signo;
2230 
2231 	ncomps = HAST_NCOMPONENTS;
2232 	lastcheck = time(NULL);
2233 
2234 	PJDLOG_VERIFY(sigemptyset(&mask) == 0);
2235 	PJDLOG_VERIFY(sigaddset(&mask, SIGINT) == 0);
2236 	PJDLOG_VERIFY(sigaddset(&mask, SIGTERM) == 0);
2237 
2238 	timeout.tv_sec = HAST_KEEPALIVE;
2239 	timeout.tv_nsec = 0;
2240 	signo = -1;
2241 
2242 	for (;;) {
2243 		switch (signo) {
2244 		case SIGINT:
2245 		case SIGTERM:
2246 			sigexit_received = true;
2247 			primary_exitx(EX_OK,
2248 			    "Termination signal received, exiting.");
2249 			break;
2250 		default:
2251 			break;
2252 		}
2253 
2254 		/*
2255 		 * Don't check connections until we fully started,
2256 		 * as we may still be looping, waiting for remote node
2257 		 * to switch from primary to secondary.
2258 		 */
2259 		if (fullystarted) {
2260 			pjdlog_debug(2, "remote_guard: Checking connections.");
2261 			now = time(NULL);
2262 			if (lastcheck + HAST_KEEPALIVE <= now) {
2263 				for (ii = 0; ii < ncomps; ii++)
2264 					guard_one(res, ii);
2265 				lastcheck = now;
2266 			}
2267 		}
2268 		signo = sigtimedwait(&mask, NULL, &timeout);
2269 	}
2270 	/* NOTREACHED */
2271 	return (NULL);
2272 }
2273