control.c revision 225736 
1/*-
2 * Copyright (c) 2009-2010 The FreeBSD Foundation
3 * All rights reserved.
4 *
5 * This software was developed by Pawel Jakub Dawidek under sponsorship from
6 * the FreeBSD Foundation.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 */
29
30#include <sys/cdefs.h>
31__FBSDID("$FreeBSD: stable/9/sbin/hastd/control.c 223780 2011-07-05 06:12:28Z trociny $");
32
33#include <sys/types.h>
34#include <sys/wait.h>
35
36#include <assert.h>
37#include <errno.h>
38#include <pthread.h>
39#include <signal.h>
40#include <stdio.h>
41#include <string.h>
42#include <unistd.h>
43
44#include "hast.h"
45#include "hastd.h"
46#include "hast_checksum.h"
47#include "hast_compression.h"
48#include "hast_proto.h"
49#include "hooks.h"
50#include "nv.h"
51#include "pjdlog.h"
52#include "proto.h"
53#include "subr.h"
54
55#include "control.h"
56
57void
58child_cleanup(struct hast_resource *res)
59{
60
61	proto_close(res->hr_ctrl);
62	res->hr_ctrl = NULL;
63	if (res->hr_event != NULL) {
64		proto_close(res->hr_event);
65		res->hr_event = NULL;
66	}
67	if (res->hr_conn != NULL) {
68		proto_close(res->hr_conn);
69		res->hr_conn = NULL;
70	}
71	res->hr_workerpid = 0;
72}
73
74static void
75control_set_role_common(struct hastd_config *cfg, struct nv *nvout,
76    uint8_t role, struct hast_resource *res, const char *name, unsigned int no)
77{
78	int oldrole;
79
80	/* Name is always needed. */
81	if (name != NULL)
82		nv_add_string(nvout, name, "resource%u", no);
83
84	if (res == NULL) {
85		assert(cfg != NULL);
86		assert(name != NULL);
87
88		TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
89			if (strcmp(res->hr_name, name) == 0)
90				break;
91		}
92		if (res == NULL) {
93			nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
94			return;
95		}
96	}
97	assert(res != NULL);
98
99	/* Send previous role back. */
100	nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
101
102	/* Nothing changed, return here. */
103	if (role == res->hr_role)
104		return;
105
106	pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
107	pjdlog_info("Role changed to %s.", role2str(role));
108
109	/* Change role to the new one. */
110	oldrole = res->hr_role;
111	res->hr_role = role;
112	pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
113
114	/*
115	 * If previous role was primary or secondary we have to kill process
116	 * doing that work.
117	 */
118	if (res->hr_workerpid != 0) {
119		if (kill(res->hr_workerpid, SIGTERM) < 0) {
120			pjdlog_errno(LOG_WARNING,
121			    "Unable to kill worker process %u",
122			    (unsigned int)res->hr_workerpid);
123		} else if (waitpid(res->hr_workerpid, NULL, 0) !=
124		    res->hr_workerpid) {
125			pjdlog_errno(LOG_WARNING,
126			    "Error while waiting for worker process %u",
127			    (unsigned int)res->hr_workerpid);
128		} else {
129			pjdlog_debug(1, "Worker process %u stopped.",
130			    (unsigned int)res->hr_workerpid);
131		}
132		child_cleanup(res);
133	}
134
135	/* Start worker process if we are changing to primary. */
136	if (role == HAST_ROLE_PRIMARY)
137		hastd_primary(res);
138	pjdlog_prefix_set("%s", "");
139	hook_exec(res->hr_exec, "role", res->hr_name, role2str(oldrole),
140	    role2str(res->hr_role), NULL);
141}
142
143void
144control_set_role(struct hast_resource *res, uint8_t role)
145{
146
147	control_set_role_common(NULL, NULL, role, res, NULL, 0);
148}
149
150static void
151control_status_worker(struct hast_resource *res, struct nv *nvout,
152    unsigned int no)
153{
154	struct nv *cnvin, *cnvout;
155	const char *str;
156	int error;
157
158	cnvin = NULL;
159
160	/*
161	 * Prepare and send command to worker process.
162	 */
163	cnvout = nv_alloc();
164	nv_add_uint8(cnvout, CONTROL_STATUS, "cmd");
165	error = nv_error(cnvout);
166	if (error != 0) {
167		pjdlog_common(LOG_ERR, 0, error,
168		    "Unable to prepare control header");
169		goto end;
170	}
171	if (hast_proto_send(res, res->hr_ctrl, cnvout, NULL, 0) < 0) {
172		error = errno;
173		pjdlog_errno(LOG_ERR, "Unable to send control header");
174		goto end;
175	}
176
177	/*
178	 * Receive response.
179	 */
180	if (hast_proto_recv_hdr(res->hr_ctrl, &cnvin) < 0) {
181		error = errno;
182		pjdlog_errno(LOG_ERR, "Unable to receive control header");
183		goto end;
184	}
185
186	error = nv_get_int16(cnvin, "error");
187	if (error != 0)
188		goto end;
189
190	if ((str = nv_get_string(cnvin, "status")) == NULL) {
191		error = ENOENT;
192		pjdlog_errno(LOG_ERR, "Field 'status' is missing.");
193		goto end;
194	}
195	nv_add_string(nvout, str, "status%u", no);
196	nv_add_uint64(nvout, nv_get_uint64(cnvin, "dirty"), "dirty%u", no);
197	nv_add_uint32(nvout, nv_get_uint32(cnvin, "extentsize"),
198	    "extentsize%u", no);
199	nv_add_uint32(nvout, nv_get_uint32(cnvin, "keepdirty"),
200	    "keepdirty%u", no);
201	nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read"),
202	    "stat_read%u", no);
203	nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write"),
204	    "stat_write%u", no);
205	nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete"),
206	    "stat_delete%u", no);
207	nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush"),
208	    "stat_flush%u", no);
209	nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_activemap_update"),
210	    "stat_activemap_update%u", no);
211end:
212	if (cnvin != NULL)
213		nv_free(cnvin);
214	if (cnvout != NULL)
215		nv_free(cnvout);
216	if (error != 0)
217		nv_add_int16(nvout, error, "error");
218}
219
220static void
221control_status(struct hastd_config *cfg, struct nv *nvout,
222    struct hast_resource *res, const char *name, unsigned int no)
223{
224
225	assert(cfg != NULL);
226	assert(nvout != NULL);
227	assert(name != NULL);
228
229	/* Name is always needed. */
230	nv_add_string(nvout, name, "resource%u", no);
231
232	if (res == NULL) {
233		TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
234			if (strcmp(res->hr_name, name) == 0)
235				break;
236		}
237		if (res == NULL) {
238			nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
239			return;
240		}
241	}
242	assert(res != NULL);
243	nv_add_string(nvout, res->hr_provname, "provname%u", no);
244	nv_add_string(nvout, res->hr_localpath, "localpath%u", no);
245	nv_add_string(nvout, res->hr_remoteaddr, "remoteaddr%u", no);
246	if (res->hr_sourceaddr[0] != '\0')
247		nv_add_string(nvout, res->hr_sourceaddr, "sourceaddr%u", no);
248	switch (res->hr_replication) {
249	case HAST_REPLICATION_FULLSYNC:
250		nv_add_string(nvout, "fullsync", "replication%u", no);
251		break;
252	case HAST_REPLICATION_MEMSYNC:
253		nv_add_string(nvout, "memsync", "replication%u", no);
254		break;
255	case HAST_REPLICATION_ASYNC:
256		nv_add_string(nvout, "async", "replication%u", no);
257		break;
258	default:
259		nv_add_string(nvout, "unknown", "replication%u", no);
260		break;
261	}
262	nv_add_string(nvout, checksum_name(res->hr_checksum),
263	    "checksum%u", no);
264	nv_add_string(nvout, compression_name(res->hr_compression),
265	    "compression%u", no);
266	nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
267
268	switch (res->hr_role) {
269	case HAST_ROLE_PRIMARY:
270		assert(res->hr_workerpid != 0);
271		/* FALLTHROUGH */
272	case HAST_ROLE_SECONDARY:
273		if (res->hr_workerpid != 0)
274			break;
275		/* FALLTHROUGH */
276	default:
277		return;
278	}
279
280	/*
281	 * If we are here, it means that we have a worker process, which we
282	 * want to ask some questions.
283	 */
284	control_status_worker(res, nvout, no);
285}
286
287void
288control_handle(struct hastd_config *cfg)
289{
290	struct proto_conn *conn;
291	struct nv *nvin, *nvout;
292	unsigned int ii;
293	const char *str;
294	uint8_t cmd, role;
295	int error;
296
297	if (proto_accept(cfg->hc_controlconn, &conn) < 0) {
298		pjdlog_errno(LOG_ERR, "Unable to accept control connection");
299		return;
300	}
301
302	cfg->hc_controlin = conn;
303	nvin = nvout = NULL;
304	role = HAST_ROLE_UNDEF;
305
306	if (hast_proto_recv_hdr(conn, &nvin) < 0) {
307		pjdlog_errno(LOG_ERR, "Unable to receive control header");
308		nvin = NULL;
309		goto close;
310	}
311
312	/* Obtain command code. 0 means that nv_get_uint8() failed. */
313	cmd = nv_get_uint8(nvin, "cmd");
314	if (cmd == 0) {
315		pjdlog_error("Control header is missing 'cmd' field.");
316		error = EHAST_INVALID;
317		goto close;
318	}
319
320	/* Allocate outgoing nv structure. */
321	nvout = nv_alloc();
322	if (nvout == NULL) {
323		pjdlog_error("Unable to allocate header for control response.");
324		error = EHAST_NOMEMORY;
325		goto close;
326	}
327
328	error = 0;
329
330	str = nv_get_string(nvin, "resource0");
331	if (str == NULL) {
332		pjdlog_error("Control header is missing 'resource0' field.");
333		error = EHAST_INVALID;
334		goto fail;
335	}
336	if (cmd == HASTCTL_CMD_SETROLE) {
337		role = nv_get_uint8(nvin, "role");
338		switch (role) {
339		case HAST_ROLE_INIT:
340		case HAST_ROLE_PRIMARY:
341		case HAST_ROLE_SECONDARY:
342			break;
343		default:
344			pjdlog_error("Invalid role received (%hhu).", role);
345			error = EHAST_INVALID;
346			goto fail;
347		}
348	}
349	if (strcmp(str, "all") == 0) {
350		struct hast_resource *res;
351
352		/* All configured resources. */
353
354		ii = 0;
355		TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
356			switch (cmd) {
357			case HASTCTL_CMD_SETROLE:
358				control_set_role_common(cfg, nvout, role, res,
359				    res->hr_name, ii++);
360				break;
361			case HASTCTL_CMD_STATUS:
362				control_status(cfg, nvout, res, res->hr_name,
363				    ii++);
364				break;
365			default:
366				pjdlog_error("Invalid command received (%hhu).",
367				    cmd);
368				error = EHAST_UNIMPLEMENTED;
369				goto fail;
370			}
371		}
372	} else {
373		/* Only selected resources. */
374
375		for (ii = 0; ; ii++) {
376			str = nv_get_string(nvin, "resource%u", ii);
377			if (str == NULL)
378				break;
379			switch (cmd) {
380			case HASTCTL_CMD_SETROLE:
381				control_set_role_common(cfg, nvout, role, NULL,
382				    str, ii);
383				break;
384			case HASTCTL_CMD_STATUS:
385				control_status(cfg, nvout, NULL, str, ii);
386				break;
387			default:
388				pjdlog_error("Invalid command received (%hhu).",
389				    cmd);
390				error = EHAST_UNIMPLEMENTED;
391				goto fail;
392			}
393		}
394	}
395	if (nv_error(nvout) != 0)
396		goto close;
397fail:
398	if (error != 0)
399		nv_add_int16(nvout, error, "error");
400
401	if (hast_proto_send(NULL, conn, nvout, NULL, 0) < 0)
402		pjdlog_errno(LOG_ERR, "Unable to send control response");
403close:
404	if (nvin != NULL)
405		nv_free(nvin);
406	if (nvout != NULL)
407		nv_free(nvout);
408	proto_close(conn);
409	cfg->hc_controlin = NULL;
410}
411
412/*
413 * Thread handles control requests from the parent.
414 */
415void *
416ctrl_thread(void *arg)
417{
418	struct hast_resource *res = arg;
419	struct nv *nvin, *nvout;
420	uint8_t cmd;
421
422	for (;;) {
423		if (hast_proto_recv_hdr(res->hr_ctrl, &nvin) < 0) {
424			if (sigexit_received)
425				pthread_exit(NULL);
426			pjdlog_errno(LOG_ERR,
427			    "Unable to receive control message");
428			kill(getpid(), SIGTERM);
429			pthread_exit(NULL);
430		}
431		cmd = nv_get_uint8(nvin, "cmd");
432		if (cmd == 0) {
433			pjdlog_error("Control message is missing 'cmd' field.");
434			nv_free(nvin);
435			continue;
436		}
437		nvout = nv_alloc();
438		switch (cmd) {
439		case CONTROL_STATUS:
440			if (res->hr_remotein != NULL &&
441			    res->hr_remoteout != NULL) {
442				nv_add_string(nvout, "complete", "status");
443			} else {
444				nv_add_string(nvout, "degraded", "status");
445			}
446			nv_add_uint32(nvout, (uint32_t)res->hr_extentsize,
447			    "extentsize");
448			if (res->hr_role == HAST_ROLE_PRIMARY) {
449				nv_add_uint32(nvout,
450				    (uint32_t)res->hr_keepdirty, "keepdirty");
451				nv_add_uint64(nvout,
452				    (uint64_t)(activemap_ndirty(res->hr_amp) *
453				    res->hr_extentsize), "dirty");
454			} else {
455				nv_add_uint32(nvout, (uint32_t)0, "keepdirty");
456				nv_add_uint64(nvout, (uint64_t)0, "dirty");
457			}
458			nv_add_uint64(nvout, res->hr_stat_read, "stat_read");
459			nv_add_uint64(nvout, res->hr_stat_write, "stat_write");
460			nv_add_uint64(nvout, res->hr_stat_delete,
461			    "stat_delete");
462			nv_add_uint64(nvout, res->hr_stat_flush, "stat_flush");
463			nv_add_uint64(nvout, res->hr_stat_activemap_update,
464			    "stat_activemap_update");
465			nv_add_int16(nvout, 0, "error");
466			break;
467		case CONTROL_RELOAD:
468			/*
469			 * When parent receives SIGHUP and discovers that
470			 * something related to us has changes, it sends reload
471			 * message to us.
472			 */
473			assert(res->hr_role == HAST_ROLE_PRIMARY);
474			primary_config_reload(res, nvin);
475			nv_add_int16(nvout, 0, "error");
476			break;
477		default:
478			nv_add_int16(nvout, EINVAL, "error");
479			break;
480		}
481		nv_free(nvin);
482		if (nv_error(nvout) != 0) {
483			pjdlog_error("Unable to create answer on control message.");
484			nv_free(nvout);
485			continue;
486		}
487		if (hast_proto_send(NULL, res->hr_ctrl, nvout, NULL, 0) < 0) {
488			pjdlog_errno(LOG_ERR,
489			    "Unable to send reply to control message");
490		}
491		nv_free(nvout);
492	}
493	/* NOTREACHED */
494	return (NULL);
495}
496