1 /*-
2 * collectd - src/mcelog.c
3 * MIT License
4 *
5 * Copyright(c) 2016 Intel Corporation. All rights reserved.
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
20 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
22 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
23 * DEALINGS IN THE SOFTWARE.
25 * Authors:
26 * Maryam Tahhan <maryam.tahhan@intel.com>
27 * Volodymyr Mytnyk <volodymyrx.mytnyk@intel.com>
28 * Taras Chornyi <tarasx.chornyi@intel.com>
29 * Krzysztof Matczak <krzysztofx.matczak@intel.com>
30 */
32 #include "common.h"
33 #include "collectd.h"
35 #include <poll.h>
36 #include <sys/socket.h>
37 #include <sys/un.h>
38 #include <unistd.h>
40 #define MCELOG_PLUGIN "mcelog"
41 #define MCELOG_BUFF_SIZE 1024
42 #define MCELOG_POLL_TIMEOUT 1000 /* ms */
43 #define MCELOG_SOCKET_STR "SOCKET"
44 #define MCELOG_DIMM_NAME "DMI_NAME"
45 #define MCELOG_CORRECTED_ERR "corrected memory errors:"
46 #define MCELOG_UNCORRECTED_ERR "uncorrected memory errors:"
48 typedef struct mcelog_config_s {
49 char logfile[PATH_MAX]; /* mcelog logfile */
50 pthread_t tid; /* poll thread id */
51 } mcelog_config_t;
53 typedef struct socket_adapter_s socket_adapter_t;
55 struct socket_adapter_s {
56 int sock_fd; /* mcelog server socket fd */
57 struct sockaddr_un unix_sock; /* mcelog client socket */
58 pthread_rwlock_t lock;
59 /* function pointers for socket operations */
60 int (*write)(socket_adapter_t *self, const char *msg, const size_t len);
61 int (*reinit)(socket_adapter_t *self);
62 int (*receive)(socket_adapter_t *self, FILE **p_file);
63 int (*close)(socket_adapter_t *self);
64 };
66 typedef struct mcelog_memory_rec_s {
67 int corrected_err_total; /* x total*/
68 int corrected_err_timed; /* x in 24h*/
69 char corrected_err_timed_period[DATA_MAX_NAME_LEN];
70 int uncorrected_err_total; /* x total*/
71 int uncorrected_err_timed; /* x in 24h*/
72 char uncorrected_err_timed_period[DATA_MAX_NAME_LEN];
73 char location[DATA_MAX_NAME_LEN]; /* SOCKET x CHANNEL x DIMM x*/
74 char dimm_name[DATA_MAX_NAME_LEN]; /* DMI_NAME "DIMM_F1" */
75 } mcelog_memory_rec_t;
77 static int socket_close(socket_adapter_t *self);
78 static int socket_write(socket_adapter_t *self, const char *msg,
79 const size_t len);
80 static int socket_reinit(socket_adapter_t *self);
81 static int socket_receive(socket_adapter_t *self, FILE **p_file);
83 static mcelog_config_t g_mcelog_config = {
84 .logfile = "/var/log/mcelog", .tid = 0,
85 };
87 static socket_adapter_t socket_adapter = {
88 .sock_fd = -1,
89 .unix_sock =
90 {
91 .sun_family = AF_UNIX, .sun_path = "/var/run/mcelog-client",
92 },
93 .lock = PTHREAD_RWLOCK_INITIALIZER,
94 .close = socket_close,
95 .write = socket_write,
96 .reinit = socket_reinit,
97 .receive = socket_receive,
98 };
100 static _Bool mcelog_thread_running = 0;
102 static int mcelog_config(oconfig_item_t *ci) {
103 for (int i = 0; i < ci->children_num; i++) {
104 oconfig_item_t *child = ci->children + i;
105 if (strcasecmp("McelogClientSocket", child->key) == 0) {
106 if (cf_util_get_string_buffer(child, socket_adapter.unix_sock.sun_path,
107 sizeof(socket_adapter.unix_sock.sun_path)) <
108 0) {
109 ERROR("%s: Invalid configuration option: \"%s\".", MCELOG_PLUGIN,
110 child->key);
111 return -1;
112 }
113 } else if (strcasecmp("McelogLogfile", child->key) == 0) {
114 if (cf_util_get_string_buffer(child, g_mcelog_config.logfile,
115 sizeof(g_mcelog_config.logfile)) < 0) {
116 ERROR("%s: Invalid configuration option: \"%s\".", MCELOG_PLUGIN,
117 child->key);
118 return -1;
119 }
120 } else {
121 ERROR("%s: Invalid configuration option: \"%s\".", MCELOG_PLUGIN,
122 child->key);
123 return -1;
124 }
125 }
126 return (0);
127 }
129 static int socket_close(socket_adapter_t *self) {
130 int ret = 0;
131 pthread_rwlock_rdlock(&self->lock);
132 if (fcntl(self->sock_fd, F_GETFL) != -1) {
133 if (shutdown(self->sock_fd, SHUT_RDWR) != 0) {
134 char errbuf[MCELOG_BUFF_SIZE];
135 ERROR("%s: Socket shutdown failed: %s", MCELOG_PLUGIN,
136 sstrerror(errno, errbuf, sizeof(errbuf)));
137 ret = -1;
138 }
139 close(self->sock_fd);
140 }
141 pthread_rwlock_unlock(&self->lock);
142 return ret;
143 }
145 static int socket_write(socket_adapter_t *self, const char *msg,
146 const size_t len) {
147 int ret = 0;
148 pthread_rwlock_rdlock(&self->lock);
149 if (swrite(self->sock_fd, msg, len) < 0)
150 ret = -1;
151 pthread_rwlock_unlock(&self->lock);
152 return ret;
153 }
155 static void mcelog_dispatch_notification(notification_t *n) {
156 if (!n) {
157 ERROR(MCELOG_PLUGIN ": %s: NULL pointer", __FUNCTION__);
158 return;
159 }
161 sstrncpy(n->host, hostname_g, sizeof(n->host));
162 sstrncpy(n->type, "gauge", sizeof(n->type));
163 plugin_dispatch_notification(n);
164 }
166 static int socket_reinit(socket_adapter_t *self) {
167 char errbuff[MCELOG_BUFF_SIZE];
168 int ret = -1;
169 cdtime_t interval = plugin_get_interval();
170 struct timeval socket_timeout = CDTIME_T_TO_TIMEVAL(interval);
172 /* synchronization via write lock since sock_fd may be changed here */
173 pthread_rwlock_wrlock(&self->lock);
174 self->sock_fd = socket(PF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
175 if (self->sock_fd < 0) {
176 ERROR("%s: Could not create a socket. %s", MCELOG_PLUGIN,
177 sstrerror(errno, errbuff, sizeof(errbuff)));
178 pthread_rwlock_unlock(&self->lock);
179 return ret;
180 }
182 /* Set socket timeout option */
183 if (setsockopt(self->sock_fd, SOL_SOCKET, SO_SNDTIMEO,
184 &socket_timeout, sizeof(socket_timeout)) < 0)
185 ERROR("%s: Failed to set the socket timeout option.", MCELOG_PLUGIN);
187 /* downgrading to read lock due to possible recursive read locks
188 * in self->close(self) call */
189 pthread_rwlock_unlock(&self->lock);
190 pthread_rwlock_rdlock(&self->lock);
191 if (connect(self->sock_fd, (struct sockaddr *)&(self->unix_sock),
192 sizeof(self->unix_sock)) < 0) {
193 ERROR("%s: Failed to connect to mcelog server. %s", MCELOG_PLUGIN,
194 sstrerror(errno, errbuff, sizeof(errbuff)));
195 self->close(self);
196 ret = -1;
197 } else {
198 ret = 0;
199 mcelog_dispatch_notification(
200 &(notification_t){.severity = NOTIF_OKAY,
201 .time = cdtime(),
202 .message = "Connected to mcelog server",
203 .plugin = MCELOG_PLUGIN,
204 .type_instance = "mcelog_status"});
205 }
206 pthread_rwlock_unlock(&self->lock);
207 return ret;
208 }
210 static int mcelog_prepare_notification(notification_t *n,
211 mcelog_memory_rec_t mr) {
212 if (n == NULL)
213 return (-1);
215 if (plugin_notification_meta_add_string(n, MCELOG_SOCKET_STR, mr.location) <
216 0) {
217 ERROR("%s: add memory location meta data failed", MCELOG_PLUGIN);
218 return (-1);
219 }
220 if (strlen(mr.dimm_name) > 0)
221 if (plugin_notification_meta_add_string(n, MCELOG_DIMM_NAME, mr.dimm_name) <
222 0) {
223 ERROR("%s: add DIMM name meta data failed", MCELOG_PLUGIN);
224 return (-1);
225 }
226 if (plugin_notification_meta_add_signed_int(n, MCELOG_CORRECTED_ERR,
227 mr.corrected_err_total) < 0) {
228 ERROR("%s: add corrected errors meta data failed", MCELOG_PLUGIN);
229 return (-1);
230 }
231 if (plugin_notification_meta_add_signed_int(
232 n, "corrected memory timed errors", mr.corrected_err_timed) < 0) {
233 ERROR("%s: add corrected timed errors meta data failed", MCELOG_PLUGIN);
234 return (-1);
235 }
236 if (plugin_notification_meta_add_string(n, "corrected errors time period",
237 mr.corrected_err_timed_period) < 0) {
238 ERROR("%s: add corrected errors period meta data failed", MCELOG_PLUGIN);
239 return (-1);
240 }
241 if (plugin_notification_meta_add_signed_int(n, MCELOG_UNCORRECTED_ERR,
242 mr.uncorrected_err_total) < 0) {
243 ERROR("%s: add corrected errors meta data failed", MCELOG_PLUGIN);
244 return (-1);
245 }
246 if (plugin_notification_meta_add_signed_int(
247 n, "uncorrected memory timed errors", mr.uncorrected_err_timed) < 0) {
248 ERROR("%s: add corrected timed errors meta data failed", MCELOG_PLUGIN);
249 return (-1);
250 }
251 if (plugin_notification_meta_add_string(n, "uncorrected errors time period",
252 mr.uncorrected_err_timed_period) <
253 0) {
254 ERROR("%s: add corrected errors period meta data failed", MCELOG_PLUGIN);
255 return (-1);
256 }
258 return (0);
259 }
261 static int mcelog_submit(mcelog_memory_rec_t mr) {
263 value_list_t vl = VALUE_LIST_INIT;
264 vl.values_len = 1;
265 vl.time = cdtime();
267 sstrncpy(vl.plugin, MCELOG_PLUGIN, sizeof(vl.plugin));
268 sstrncpy(vl.type, "errors", sizeof(vl.type));
269 if (strlen(mr.dimm_name) > 0) {
270 ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%s_%s",
271 mr.location, mr.dimm_name);
272 } else
273 sstrncpy(vl.plugin_instance, mr.location, sizeof(vl.plugin_instance));
275 sstrncpy(vl.type_instance, "corrected_memory_errors",
276 sizeof(vl.type_instance));
277 vl.values = &(value_t){.derive = (derive_t)mr.corrected_err_total};
278 plugin_dispatch_values(&vl);
280 ssnprintf(vl.type_instance, sizeof(vl.type_instance),
281 "corrected_memory_errors_in_%s", mr.corrected_err_timed_period);
282 vl.values = &(value_t){.derive = (derive_t)mr.corrected_err_timed};
283 plugin_dispatch_values(&vl);
285 sstrncpy(vl.type_instance, "uncorrected_memory_errors",
286 sizeof(vl.type_instance));
287 vl.values = &(value_t){.derive = (derive_t)mr.uncorrected_err_total};
288 plugin_dispatch_values(&vl);
290 ssnprintf(vl.type_instance, sizeof(vl.type_instance),
291 "uncorrected_memory_errors_in_%s", mr.uncorrected_err_timed_period);
292 vl.values = &(value_t){.derive = (derive_t)mr.uncorrected_err_timed};
293 plugin_dispatch_values(&vl);
295 return 0;
296 }
298 static int parse_memory_info(FILE *p_file, mcelog_memory_rec_t *memory_record) {
299 char buf[DATA_MAX_NAME_LEN] = {0};
300 while (fgets(buf, sizeof(buf), p_file)) {
301 /* Got empty line or "done" */
302 if ((!strncmp("\n", buf, strlen(buf))) ||
303 (!strncmp(buf, "done\n", strlen(buf))))
304 return 1;
305 if (strlen(buf) < 5)
306 continue;
307 if (!strncmp(buf, MCELOG_SOCKET_STR, strlen(MCELOG_SOCKET_STR))) {
308 sstrncpy(memory_record->location, buf, strlen(buf));
309 /* replace spaces with '_' */
310 for (size_t i = 0; i < strlen(memory_record->location); i++)
311 if (memory_record->location[i] == ' ')
312 memory_record->location[i] = '_';
313 DEBUG("%s: Got SOCKET INFO %s", MCELOG_PLUGIN, memory_record->location);
314 }
315 if (!strncmp(buf, MCELOG_DIMM_NAME, strlen(MCELOG_DIMM_NAME))) {
316 char *name = NULL;
317 char *saveptr = NULL;
318 name = strtok_r(buf, "\"", &saveptr);
319 if (name != NULL && saveptr != NULL) {
320 name = strtok_r(NULL, "\"", &saveptr);
321 if (name != NULL) {
322 sstrncpy(memory_record->dimm_name, name,
323 sizeof(memory_record->dimm_name));
324 DEBUG("%s: Got DIMM NAME %s", MCELOG_PLUGIN,
325 memory_record->dimm_name);
326 }
327 }
328 }
329 if (!strncmp(buf, MCELOG_CORRECTED_ERR, strlen(MCELOG_CORRECTED_ERR))) {
330 /* Get next line*/
331 if (fgets(buf, sizeof(buf), p_file) != NULL) {
332 sscanf(buf, "\t%d total", &(memory_record->corrected_err_total));
333 DEBUG("%s: Got corrected error total %d", MCELOG_PLUGIN,
334 memory_record->corrected_err_total);
335 }
336 if (fgets(buf, sizeof(buf), p_file) != NULL) {
337 sscanf(buf, "\t%d in %s", &(memory_record->corrected_err_timed),
338 memory_record->corrected_err_timed_period);
339 DEBUG("%s: Got timed corrected errors %d in %s", MCELOG_PLUGIN,
340 memory_record->corrected_err_total,
341 memory_record->corrected_err_timed_period);
342 }
343 }
344 if (!strncmp(buf, MCELOG_UNCORRECTED_ERR, strlen(MCELOG_UNCORRECTED_ERR))) {
345 if (fgets(buf, sizeof(buf), p_file) != NULL) {
346 sscanf(buf, "\t%d total", &(memory_record->uncorrected_err_total));
347 DEBUG("%s: Got uncorrected error total %d", MCELOG_PLUGIN,
348 memory_record->uncorrected_err_total);
349 }
350 if (fgets(buf, sizeof(buf), p_file) != NULL) {
351 sscanf(buf, "\t%d in %s", &(memory_record->uncorrected_err_timed),
352 memory_record->uncorrected_err_timed_period);
353 DEBUG("%s: Got timed uncorrected errors %d in %s", MCELOG_PLUGIN,
354 memory_record->uncorrected_err_total,
355 memory_record->uncorrected_err_timed_period);
356 }
357 }
358 memset(buf, 0, sizeof(buf));
359 }
360 /* parsing definitely finished */
361 return 0;
362 }
364 static void poll_worker_cleanup(void *arg) {
365 mcelog_thread_running = 0;
366 FILE *p_file = *((FILE **)arg);
367 if (p_file != NULL)
368 fclose(p_file);
369 free(arg);
370 }
372 static int socket_receive(socket_adapter_t *self, FILE **pp_file) {
373 int res = -1;
374 pthread_rwlock_rdlock(&self->lock);
375 struct pollfd poll_fd = {
376 .fd = self->sock_fd, .events = POLLIN | POLLPRI,
377 };
379 if ((res = poll(&poll_fd, 1, MCELOG_POLL_TIMEOUT)) <= 0) {
380 if (res != 0 && errno != EINTR) {
381 char errbuf[MCELOG_BUFF_SIZE];
382 ERROR("mcelog: poll failed: %s",
383 sstrerror(errno, errbuf, sizeof(errbuf)));
384 }
385 pthread_rwlock_unlock(&self->lock);
386 return res;
387 }
389 if (poll_fd.revents & (POLLERR | POLLHUP | POLLNVAL)) {
390 /* connection is broken */
391 ERROR("%s: Connection to socket is broken", MCELOG_PLUGIN);
392 if (poll_fd.revents & (POLLERR | POLLHUP)) {
393 mcelog_dispatch_notification(
394 &(notification_t){.severity = NOTIF_FAILURE,
395 .time = cdtime(),
396 .message = "Connection to mcelog socket is broken.",
397 .plugin = MCELOG_PLUGIN,
398 .type_instance = "mcelog_status"});
399 }
400 pthread_rwlock_unlock(&self->lock);
401 return -1;
402 }
404 if (!(poll_fd.revents & (POLLIN | POLLPRI))) {
405 INFO("%s: No data to read", MCELOG_PLUGIN);
406 pthread_rwlock_unlock(&self->lock);
407 return 0;
408 }
410 if ((*pp_file = fdopen(dup(self->sock_fd), "r")) == NULL)
411 res = -1;
413 pthread_rwlock_unlock(&self->lock);
414 return res;
415 }
417 static void *poll_worker(__attribute__((unused)) void *arg) {
418 char errbuf[MCELOG_BUFF_SIZE];
419 mcelog_thread_running = 1;
420 FILE **pp_file = calloc(1, sizeof(*pp_file));
421 if (pp_file == NULL) {
422 ERROR("mcelog: memory allocation failed: %s",
423 sstrerror(errno, errbuf, sizeof(errbuf)));
424 pthread_exit((void *)1);
425 }
427 pthread_cleanup_push(poll_worker_cleanup, pp_file);
429 while (1) {
430 /* blocking call */
431 int res = socket_adapter.receive(&socket_adapter, pp_file);
432 if (res < 0) {
433 socket_adapter.close(&socket_adapter);
434 while (socket_adapter.reinit(&socket_adapter) != 0) {
435 nanosleep(&CDTIME_T_TO_TIMESPEC(MS_TO_CDTIME_T(MCELOG_POLL_TIMEOUT)),
436 NULL);
437 }
438 continue;
439 }
440 /* timeout or no data to read */
441 else if (res == 0)
442 continue;
444 if (*pp_file == NULL)
445 continue;
447 mcelog_memory_rec_t memory_record = {0};
448 while (parse_memory_info(*pp_file, &memory_record)) {
449 /* Check if location was successfully parsed */
450 if (memory_record.location[0] == '\0') {
451 memset(&memory_record, 0, sizeof(memory_record));
452 continue;
453 }
455 notification_t n = {.severity = NOTIF_OKAY,
456 .time = cdtime(),
457 .message = "Got memory errors info.",
458 .plugin = MCELOG_PLUGIN,
459 .type_instance = "memory_erros"};
461 if (mcelog_prepare_notification(&n, memory_record) == 0)
462 mcelog_dispatch_notification(&n);
463 if (mcelog_submit(memory_record) != 0)
464 ERROR("%s: Failed to submit memory errors", MCELOG_PLUGIN);
465 memset(&memory_record, 0, sizeof(memory_record));
466 }
468 fclose(*pp_file);
469 *pp_file = NULL;
470 }
472 mcelog_thread_running = 0;
473 pthread_cleanup_pop(1);
474 return NULL;
475 }
477 static int mcelog_init(void) {
478 if (socket_adapter.reinit(&socket_adapter) != 0) {
479 ERROR("%s: Cannot connect to client socket", MCELOG_PLUGIN);
480 return -1;
481 }
483 if (plugin_thread_create(&g_mcelog_config.tid, NULL, poll_worker, NULL,
484 NULL) != 0) {
485 ERROR("%s: Error creating poll thread.", MCELOG_PLUGIN);
486 return -1;
487 }
488 return 0;
489 }
491 static int get_memory_machine_checks(void) {
492 static const char dump[] = "dump all bios\n";
493 int ret = socket_adapter.write(&socket_adapter, dump, sizeof(dump));
494 if (ret != 0)
495 ERROR("%s: SENT DUMP REQUEST FAILED", MCELOG_PLUGIN);
496 else
497 DEBUG("%s: SENT DUMP REQUEST OK", MCELOG_PLUGIN);
498 return ret;
499 }
501 static int mcelog_read(__attribute__((unused)) user_data_t *ud) {
502 DEBUG("%s: %s", MCELOG_PLUGIN, __FUNCTION__);
504 if (get_memory_machine_checks() != 0)
505 ERROR("%s: MACHINE CHECK INFO NOT AVAILABLE", MCELOG_PLUGIN);
507 return 0;
508 }
510 static int mcelog_shutdown(void) {
511 int ret = 0;
512 if (mcelog_thread_running) {
513 pthread_cancel(g_mcelog_config.tid);
514 if (pthread_join(g_mcelog_config.tid, NULL) != 0) {
515 ERROR("%s: Stopping thread failed.", MCELOG_PLUGIN);
516 ret = -1;
517 }
518 }
520 ret = socket_adapter.close(&socket_adapter) || ret;
521 pthread_rwlock_destroy(&(socket_adapter.lock));
522 return -ret;
523 }
525 void module_register(void) {
526 plugin_register_complex_config(MCELOG_PLUGIN, mcelog_config);
527 plugin_register_init(MCELOG_PLUGIN, mcelog_init);
528 plugin_register_complex_read(NULL, MCELOG_PLUGIN, mcelog_read, 0, NULL);
529 plugin_register_shutdown(MCELOG_PLUGIN, mcelog_shutdown);
530 }