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 NewVol=%d\n";
45 static char use_storage[] = "use storage=%s media_type=%s pool_name=%s "
46 "pool_type=%s append=%d copy=%d stripe=%d\n";
47 static char use_device[] = "use device=%s\n";
48 //static char query_device[] = "query device=%s";
50 /* Response from Storage daemon */
51 static char OKjob[] = "3000 OK Job SDid=%d SDtime=%d Authorization=%100s\n";
52 static char OK_device[] = "3000 OK use device device=%s\n";
54 /* Storage Daemon requests */
55 static char Job_start[] = "3010 Job %127s start\n";
56 static char Job_end[] =
57 "3099 Job %127s end JobStatus=%d JobFiles=%d JobBytes=%" lld "\n";
59 /* Forward referenced functions */
60 extern "C" void *msg_thread(void *arg);
63 * Establish a message channel connection with the Storage daemon
64 * and perform authentication.
66 bool connect_to_storage_daemon(JCR *jcr, int retry_interval,
67 int max_retry_time, int verbose)
72 if (jcr->store_bsock) {
73 return true; /* already connected */
75 store = (STORE *)jcr->storage->first();
78 * Open message channel with the Storage daemon
80 Dmsg2(100, "bnet_connect to Storage daemon %s:%d\n", store->address,
82 sd = bnet_connect(jcr, retry_interval, max_retry_time,
83 _("Storage daemon"), store->address,
84 NULL, store->SDport, verbose);
88 sd->res = (RES *)store; /* save pointer to other end */
89 jcr->store_bsock = sd;
91 if (!authenticate_storage_daemon(jcr, store)) {
93 jcr->store_bsock = NULL;
100 * Here we ask the SD to send us the info for a
101 * particular device resource.
104 bool update_device_res(JCR *jcr, DEVICE *dev)
106 POOL_MEM device_name;
108 if (!connect_to_storage_daemon(jcr, 5, 30, 0)) {
111 sd = jcr->store_bsock;
112 pm_strcpy(device_name, dev->hdr.name);
113 bash_spaces(device_name);
114 bnet_fsend(sd, query_device, device_name.c_str());
115 Dmsg1(100, ">stored: %s\n", sd->msg);
116 /* The data is returned through Device_update */
117 if (bget_dirmsg(sd) <= 0) {
125 * Start a job with the Storage daemon
127 int start_storage_daemon_job(JCR *jcr, alist *store, int append)
133 POOL_MEM store_name, device_name, pool_name, pool_type, media_type;
138 sd = jcr->store_bsock;
140 * Now send JobId and permissions, and get back the authorization key.
142 bash_spaces(jcr->job->hdr.name);
143 bash_spaces(jcr->client->hdr.name);
144 bash_spaces(jcr->fileset->hdr.name);
145 if (jcr->fileset->MD5[0] == 0) {
146 bstrncpy(jcr->fileset->MD5, "**Dummy**", sizeof(jcr->fileset->MD5));
148 bnet_fsend(sd, jobcmd, jcr->JobId, jcr->Job, jcr->job->hdr.name,
149 jcr->client->hdr.name, jcr->JobType, jcr->JobLevel,
150 jcr->fileset->hdr.name, !jcr->pool->catalog_files,
151 jcr->job->SpoolAttributes, jcr->fileset->MD5, jcr->spool_data,
152 jcr->write_part_after_job, jcr->job->NewVolEachJob);
153 Dmsg1(100, ">stored: %s\n", sd->msg);
154 unbash_spaces(jcr->job->hdr.name);
155 unbash_spaces(jcr->client->hdr.name);
156 unbash_spaces(jcr->fileset->hdr.name);
157 if (bget_dirmsg(sd) > 0) {
158 Dmsg1(100, "<stored: %s", sd->msg);
159 if (sscanf(sd->msg, OKjob, &jcr->VolSessionId,
160 &jcr->VolSessionTime, &auth_key) != 3) {
161 Dmsg1(100, "BadJob=%s\n", sd->msg);
162 Jmsg(jcr, M_FATAL, 0, _("Storage daemon rejected Job command: %s\n"), sd->msg);
165 jcr->sd_auth_key = bstrdup(auth_key);
166 Dmsg1(150, "sd_auth_key=%s\n", jcr->sd_auth_key);
169 Jmsg(jcr, M_FATAL, 0, _("<stored: bad response to Job command: %s\n"),
174 pm_strcpy(pool_type, jcr->pool->pool_type);
175 pm_strcpy(pool_name, jcr->pool->hdr.name);
176 bash_spaces(pool_type);
177 bash_spaces(pool_name);
178 edit_int64(jcr->PoolId, PoolId);
181 * We have two loops here. The first comes from the
182 * Storage = associated with the Job, and we need
183 * to attach to each one.
184 * The inner loop loops over all the alternative devices
185 * associated with each Storage. It selects the first
188 * Note, the outer loop is not yet implemented.
190 // foreach_alist(storage, store) {
191 storage = (STORE *)store->first();
192 pm_strcpy(store_name, storage->hdr.name);
193 bash_spaces(store_name);
194 pm_strcpy(media_type, storage->media_type);
195 bash_spaces(media_type);
196 bnet_fsend(sd, use_storage, store_name.c_str(), media_type.c_str(),
197 pool_name.c_str(), pool_type.c_str(), append, copy, stripe);
200 /* Loop over alternative storage Devices until one is OK */
201 foreach_alist(dev, storage->device) {
202 pm_strcpy(device_name, dev->hdr.name);
203 bash_spaces(device_name);
204 bnet_fsend(sd, use_device, device_name.c_str());
205 Dmsg1(100, ">stored: %s", sd->msg);
207 bnet_sig(sd, BNET_EOD); /* end of Devices */
208 bnet_sig(sd, BNET_EOD); /* end of Storages */
209 if (bget_dirmsg(sd) > 0) {
210 Dmsg1(100, "<stored: %s", sd->msg);
211 /* ****FIXME**** save actual device name */
212 ok = sscanf(sd->msg, OK_device, device_name.c_str()) == 1;
215 pm_strcpy(err_msg, sd->msg); /* save message */
216 Jmsg(jcr, M_WARNING, 0, _("\n"
217 " Storage daemon didn't accept Device \"%s\" because:\n %s"),
218 device_name.c_str(), err_msg.c_str()/* sd->msg */);
225 ok = bnet_fsend(sd, "run");
226 Dmsg1(100, ">stored: %s\n", sd->msg);
232 * Start a thread to handle Storage daemon messages and
235 int start_storage_daemon_message_thread(JCR *jcr)
241 jcr->use_count++; /* mark in use by msg thread */
242 jcr->sd_msg_thread_done = false;
243 jcr->SD_msg_chan = 0;
245 Dmsg0(100, "Start SD msg_thread.\n");
246 if ((status=pthread_create(&thid, NULL, msg_thread, (void *)jcr)) != 0) {
248 Jmsg1(jcr, M_ABORT, 0, _("Cannot create message thread: %s\n"), be.strerror(status));
250 Dmsg0(100, "SD msg_thread started.\n");
251 /* Wait for thread to start */
252 while (jcr->SD_msg_chan == 0) {
258 extern "C" void msg_thread_cleanup(void *arg)
260 JCR *jcr = (JCR *)arg;
261 Dmsg0(200, "End msg_thread\n");
262 db_end_transaction(jcr, jcr->db); /* terminate any open transaction */
264 jcr->sd_msg_thread_done = true;
265 pthread_cond_broadcast(&jcr->term_wait); /* wakeup any waiting threads */
266 jcr->SD_msg_chan = 0;
268 free_jcr(jcr); /* release jcr */
272 * Handle the message channel (i.e. requests from the
274 * Note, we are running in a separate thread.
276 extern "C" void *msg_thread(void *arg)
278 JCR *jcr = (JCR *)arg;
281 char Job[MAX_NAME_LENGTH];
286 pthread_detach(pthread_self());
287 jcr->SD_msg_chan = pthread_self();
288 pthread_cleanup_push(msg_thread_cleanup, arg);
289 sd = jcr->store_bsock;
291 /* Read the Storage daemon's output.
293 Dmsg0(100, "Start msg_thread loop\n");
294 while ((stat=bget_dirmsg(sd)) >= 0) {
295 Dmsg1(200, "<stored: %s", sd->msg);
296 if (sscanf(sd->msg, Job_start, &Job) == 1) {
299 if (sscanf(sd->msg, Job_end, &Job, &JobStatus, &JobFiles,
301 jcr->SDJobStatus = JobStatus; /* termination status */
302 jcr->SDJobFiles = JobFiles;
303 jcr->SDJobBytes = JobBytes;
307 if (is_bnet_error(sd)) {
308 jcr->SDJobStatus = JS_ErrorTerminated;
310 pthread_cleanup_pop(1);
314 void wait_for_storage_daemon_termination(JCR *jcr)
316 int cancel_count = 0;
317 /* Now wait for Storage daemon to terminate our message thread */
318 set_jcr_job_status(jcr, JS_WaitSD);
320 while (!jcr->sd_msg_thread_done) {
323 struct timespec timeout;
325 gettimeofday(&tv, &tz);
327 timeout.tv_sec = tv.tv_sec + 10; /* wait 10 seconds */
328 Dmsg0(300, "I'm waiting for message thread termination.\n");
329 pthread_cond_timedwait(&jcr->term_wait, &jcr->mutex, &timeout);
330 if (job_canceled(jcr)) {
333 /* Give SD 30 seconds to clean up after cancel */
334 if (cancel_count == 3) {
339 set_jcr_job_status(jcr, JS_Terminated);
345 extern "C" void *device_thread(void *arg)
352 pthread_detach(pthread_self());
353 jcr = new_control_jcr("*DeviceInit*", JT_SYSTEM);
354 for (i=0; i < MAX_TRIES; i++) {
355 if (!connect_to_storage_daemon(jcr, 10, 30, 1)) {
356 Dmsg0(000, "Failed connecting to SD.\n");
360 foreach_res(dev, R_DEVICE) {
361 if (!update_device_res(jcr, dev)) {
362 Dmsg1(900, "Error updating device=%s\n", dev->hdr.name);
364 Dmsg1(900, "Updated Device=%s\n", dev->hdr.name);
368 bnet_close(jcr->store_bsock);
369 jcr->store_bsock = NULL;
378 * Start a thread to handle getting Device resource information
379 * from SD. This is called once at startup of the Director.
381 void init_device_resources()
386 Dmsg0(100, "Start Device thread.\n");
387 if ((status=pthread_create(&thid, NULL, device_thread, NULL)) != 0) {
389 Jmsg1(NULL, M_ABORT, 0, _("Cannot create message thread: %s\n"), be.strerror(status));