3 * Bacula Director -- msgchan.c -- handles the message channel
4 * to the Storage daemon and the File daemon.
6 * Kern Sibbald, August MM
8 * This routine runs as a thread and must be thread reentrant.
10 * Basic tasks done here:
11 * Open a message channel with the Storage daemon
12 * to authenticate ourself and to pass the JobId.
13 * Create a thread to interact with the Storage daemon
14 * who returns a job status and requests Catalog services, etc.
19 Copyright (C) 2000-2005 Kern Sibbald
21 This program is free software; you can redistribute it and/or
22 modify it under the terms of the GNU General Public License as
23 published by the Free Software Foundation; either version 2 of
24 the License, or (at your option) any later version.
26 This program is distributed in the hope that it will be useful,
27 but WITHOUT ANY WARRANTY; without even the implied warranty of
28 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
29 General Public License for more details.
31 You should have received a copy of the GNU General Public
32 License along with this program; if not, write to the Free
33 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
41 /* Commands sent to Storage daemon */
42 static char jobcmd[] = "JobId=%d job=%s job_name=%s client_name=%s "
43 "type=%d level=%d FileSet=%s NoAttr=%d SpoolAttr=%d FileSetMD5=%s "
44 "SpoolData=%d WritePartAfterJob=%d";
45 static char use_device[] = "use device=%s media_type=%s pool_name=%s "
46 "pool_type=%s append=%d\n";
47 static char query_device[] = "query device=%s";
49 /* Response from Storage daemon */
50 static char OKjob[] = "3000 OK Job SDid=%d SDtime=%d Authorization=%100s\n";
51 static char OK_device[] = "3000 OK use device\n";
52 static char OK_query[] = "3001 OK query append=%d read=%d num_writers=%d "
53 "num_waiting=%d open=%d use_count=%d labeled=%d offline=%d autochanger=%d "
54 "media_type=%127s volume_name=%127s";
56 /* Storage Daemon requests */
57 static char Job_start[] = "3010 Job %127s start\n";
58 static char Job_end[] =
59 "3099 Job %127s end JobStatus=%d JobFiles=%d JobBytes=%" lld "\n";
61 /* Forward referenced functions */
62 extern "C" void *msg_thread(void *arg);
65 * Establish a message channel connection with the Storage daemon
66 * and perform authentication.
68 bool connect_to_storage_daemon(JCR *jcr, int retry_interval,
69 int max_retry_time, int verbose)
74 if (jcr->store_bsock) {
75 return true; /* already connected */
77 store = (STORE *)jcr->storage->first();
80 * Open message channel with the Storage daemon
82 Dmsg2(200, "bnet_connect to Storage daemon %s:%d\n", store->address,
84 sd = bnet_connect(jcr, retry_interval, max_retry_time,
85 _("Storage daemon"), store->address,
86 NULL, store->SDport, verbose);
90 sd->res = (RES *)store; /* save pointer to other end */
91 jcr->store_bsock = sd;
93 if (!authenticate_storage_daemon(jcr, store)) {
95 jcr->store_bsock = NULL;
102 * Here we ask the SD to send us the info for a
103 * particular device resource.
105 bool update_device_res(JCR *jcr, DEVICE *dev)
107 POOL_MEM device_name, media_type, volume_name;
108 int dev_open, dev_append, dev_read, dev_labeled;
109 int dev_offline, dev_autochanger;
111 if (!connect_to_storage_daemon(jcr, 5, 30, 0)) {
114 sd = jcr->store_bsock;
115 pm_strcpy(device_name, dev->hdr.name);
116 bash_spaces(device_name);
117 bnet_fsend(sd, query_device, device_name.c_str());
118 if (bget_dirmsg(sd) > 0) {
119 Dmsg1(400, "<stored: %s", sd->msg);
120 if (sscanf(sd->msg, OK_query, &dev_append, &dev_read,
121 &dev->num_writers, &dev->num_waiting, &dev_open,
122 &dev->use_count, &dev_labeled, &dev_offline,
123 &dev_autochanger, media_type.c_str(),
124 volume_name.c_str()) != 11) {
127 unbash_spaces(media_type);
128 unbash_spaces(volume_name);
129 bstrncpy(dev->MediaType, media_type.c_str(), sizeof(dev->MediaType));
130 bstrncpy(dev->VolumeName, volume_name.c_str(), sizeof(dev->VolumeName));
131 dev->open = dev_open;
132 dev->append = dev_append;
133 dev->read = dev_read;
134 dev->labeled = dev_labeled;
135 dev->offline = dev_offline;
136 dev->autochanger = dev_autochanger;
145 * Start a job with the Storage daemon
147 int start_storage_daemon_job(JCR *jcr, alist *store, int append)
153 POOL_MEM device_name, pool_name, pool_type, media_type;
155 sd = jcr->store_bsock;
157 * Now send JobId and permissions, and get back the authorization key.
159 bash_spaces(jcr->job->hdr.name);
160 bash_spaces(jcr->client->hdr.name);
161 bash_spaces(jcr->fileset->hdr.name);
162 if (jcr->fileset->MD5[0] == 0) {
163 bstrncpy(jcr->fileset->MD5, "**Dummy**", sizeof(jcr->fileset->MD5));
165 bnet_fsend(sd, jobcmd, jcr->JobId, jcr->Job, jcr->job->hdr.name,
166 jcr->client->hdr.name, jcr->JobType, jcr->JobLevel,
167 jcr->fileset->hdr.name, !jcr->pool->catalog_files,
168 jcr->job->SpoolAttributes, jcr->fileset->MD5, jcr->spool_data, jcr->write_part_after_job);
169 Dmsg1(200, "Jobcmd=%s\n", sd->msg);
170 unbash_spaces(jcr->job->hdr.name);
171 unbash_spaces(jcr->client->hdr.name);
172 unbash_spaces(jcr->fileset->hdr.name);
173 if (bget_dirmsg(sd) > 0) {
174 Dmsg1(110, "<stored: %s", sd->msg);
175 if (sscanf(sd->msg, OKjob, &jcr->VolSessionId,
176 &jcr->VolSessionTime, &auth_key) != 3) {
177 Dmsg1(100, "BadJob=%s\n", sd->msg);
178 Jmsg(jcr, M_FATAL, 0, _("Storage daemon rejected Job command: %s\n"), sd->msg);
181 jcr->sd_auth_key = bstrdup(auth_key);
182 Dmsg1(150, "sd_auth_key=%s\n", jcr->sd_auth_key);
185 Jmsg(jcr, M_FATAL, 0, _("<stored: bad response to Job command: %s\n"),
190 // foreach_alist(storage, store) {
191 storage = (STORE *)store->first();
192 pm_strcpy(device_name, storage->dev_name());
193 pm_strcpy(media_type, storage->media_type);
194 pm_strcpy(pool_type, jcr->pool->pool_type);
195 pm_strcpy(pool_name, jcr->pool->hdr.name);
196 bash_spaces(device_name);
197 bash_spaces(media_type);
198 bash_spaces(pool_type);
199 bash_spaces(pool_name);
200 bnet_fsend(sd, use_device, device_name.c_str(),
201 media_type.c_str(), pool_name.c_str(), pool_type.c_str(),
203 Dmsg1(200, ">stored: %s", sd->msg);
204 ok = response(jcr, sd, OK_device, "Use Device", NO_DISPLAY);
206 pm_strcpy(pool_type, sd->msg); /* save message */
207 Jmsg(jcr, M_FATAL, 0, _("\n"
208 " Storage daemon didn't accept Device \"%s\" because:\n %s"),
209 device_name.c_str(), pool_type.c_str()/* sd->msg */);
213 bnet_fsend(sd, "run");
219 * Start a thread to handle Storage daemon messages and
222 int start_storage_daemon_message_thread(JCR *jcr)
228 jcr->use_count++; /* mark in use by msg thread */
229 jcr->sd_msg_thread_done = false;
230 jcr->SD_msg_chan = 0;
232 Dmsg0(100, "Start SD msg_thread.\n");
233 if ((status=pthread_create(&thid, NULL, msg_thread, (void *)jcr)) != 0) {
235 Jmsg1(jcr, M_ABORT, 0, _("Cannot create message thread: %s\n"), be.strerror(status));
237 Dmsg0(100, "SD msg_thread started.\n");
238 /* Wait for thread to start */
239 while (jcr->SD_msg_chan == 0) {
245 extern "C" void msg_thread_cleanup(void *arg)
247 JCR *jcr = (JCR *)arg;
248 Dmsg0(200, "End msg_thread\n");
249 db_end_transaction(jcr, jcr->db); /* terminate any open transaction */
251 jcr->sd_msg_thread_done = true;
252 pthread_cond_broadcast(&jcr->term_wait); /* wakeup any waiting threads */
253 jcr->SD_msg_chan = 0;
255 free_jcr(jcr); /* release jcr */
259 * Handle the message channel (i.e. requests from the
261 * Note, we are running in a separate thread.
263 extern "C" void *msg_thread(void *arg)
265 JCR *jcr = (JCR *)arg;
268 char Job[MAX_NAME_LENGTH];
273 pthread_detach(pthread_self());
274 jcr->SD_msg_chan = pthread_self();
275 pthread_cleanup_push(msg_thread_cleanup, arg);
276 sd = jcr->store_bsock;
278 /* Read the Storage daemon's output.
280 Dmsg0(100, "Start msg_thread loop\n");
281 while ((stat=bget_dirmsg(sd)) >= 0) {
282 Dmsg1(200, "<stored: %s", sd->msg);
283 if (sscanf(sd->msg, Job_start, &Job) == 1) {
286 if (sscanf(sd->msg, Job_end, &Job, &JobStatus, &JobFiles,
288 jcr->SDJobStatus = JobStatus; /* termination status */
289 jcr->SDJobFiles = JobFiles;
290 jcr->SDJobBytes = JobBytes;
294 if (is_bnet_error(sd)) {
295 jcr->SDJobStatus = JS_ErrorTerminated;
297 pthread_cleanup_pop(1);
301 void wait_for_storage_daemon_termination(JCR *jcr)
303 int cancel_count = 0;
304 /* Now wait for Storage daemon to terminate our message thread */
305 set_jcr_job_status(jcr, JS_WaitSD);
307 while (!jcr->sd_msg_thread_done) {
310 struct timespec timeout;
312 gettimeofday(&tv, &tz);
314 timeout.tv_sec = tv.tv_sec + 10; /* wait 10 seconds */
315 Dmsg0(300, "I'm waiting for message thread termination.\n");
316 pthread_cond_timedwait(&jcr->term_wait, &jcr->mutex, &timeout);
317 if (job_canceled(jcr)) {
320 /* Give SD 30 seconds to clean up after cancel */
321 if (cancel_count == 3) {
326 set_jcr_job_status(jcr, JS_Terminated);
332 extern "C" void *device_thread(void *arg)
339 pthread_detach(pthread_self());
340 jcr = new_control_jcr("*DeviceInit*", JT_SYSTEM);
341 for (i=0; i < MAX_TRIES; i++) {
342 if (!connect_to_storage_daemon(jcr, 10, 30, 1)) {
343 Dmsg0(000, "Failed connecting to SD.\n");
347 foreach_res(dev, R_DEVICE) {
348 if (!update_device_res(jcr, dev)) {
349 Dmsg1(900, "Error updating device=%s\n", dev->hdr.name);
351 Dmsg1(900, "Updated Device=%s\n", dev->hdr.name);
355 bnet_close(jcr->store_bsock);
356 jcr->store_bsock = NULL;
365 * Start a thread to handle getting Device resource information
366 * from SD. This is called once at startup of the Director.
368 void init_device_resources()
373 Dmsg0(100, "Start Device thread.\n");
374 if ((status=pthread_create(&thid, NULL, device_thread, NULL)) != 0) {
376 Jmsg1(NULL, M_ABORT, 0, _("Cannot create message thread: %s\n"), be.strerror(status));