2 Bacula® - The Network Backup Solution
4 Copyright (C) 2001-2009 Free Software Foundation Europe e.V.
6 The main author of Bacula is Kern Sibbald, with contributions from
7 many others, a complete list can be found in the file AUTHORS.
8 This program is Free Software; you can redistribute it and/or
9 modify it under the terms of version two of the GNU General Public
10 License as published by the Free Software Foundation and included
13 This program is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; if not, write to the Free Software
20 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
23 Bacula® is a registered trademark of Kern Sibbald.
24 The licensor of Bacula is the Free Software Foundation Europe
25 (FSFE), Fiduciary Program, Sumatrastrasse 25, 8006 Zürich,
26 Switzerland, email:ftf@fsfeurope.org.
29 * This file handles accepting Director Commands
31 * Most Director commands are handled here, with the
32 * exception of the Job command command and subsequent
33 * subcommands that are handled
36 * N.B. in this file, in general we must use P(dev->mutex) rather
37 * than dev->r_lock() so that we can examine the blocked
38 * state rather than blocking ourselves because a Job
39 * thread has the device blocked. In some "safe" cases,
40 * we can do things to a blocked device. CAREFUL!!!!
42 * File daemon commands are handled in fdcmd.c
44 * Kern Sibbald, May MMI
53 /* Exported variables */
55 /* Imported variables */
56 extern BSOCK *filed_chan;
57 extern struct s_last_job last_job;
58 extern bool init_done;
60 /* Static variables */
61 static char derrmsg[] = "3900 Invalid command\n";
62 static char OKsetdebug[] = "3000 OK setdebug=%d\n";
63 static char invalid_cmd[] = "3997 Invalid command for a Director with Monitor directive enabled.\n";
64 static char OK_bootstrap[] = "3000 OK bootstrap\n";
65 static char ERROR_bootstrap[] = "3904 Error bootstrap\n";
67 /* Imported functions */
68 extern void terminate_child();
69 extern bool job_cmd(JCR *jcr);
70 extern bool use_cmd(JCR *jcr);
71 extern bool run_cmd(JCR *jcr);
72 extern bool status_cmd(JCR *sjcr);
73 extern bool qstatus_cmd(JCR *jcr);
74 //extern bool query_cmd(JCR *jcr);
76 /* Forward referenced functions */
77 static bool label_cmd(JCR *jcr);
78 static bool die_cmd(JCR *jcr);
79 static bool relabel_cmd(JCR *jcr);
80 static bool readlabel_cmd(JCR *jcr);
81 static bool release_cmd(JCR *jcr);
82 static bool setdebug_cmd(JCR *jcr);
83 static bool cancel_cmd(JCR *cjcr);
84 static bool mount_cmd(JCR *jcr);
85 static bool unmount_cmd(JCR *jcr);
86 static bool action_on_purge_cmd(JCR *jcr);
87 static bool bootstrap_cmd(JCR *jcr);
88 static bool changer_cmd(JCR *sjcr);
89 static bool do_label(JCR *jcr, int relabel);
90 static DCR *find_device(JCR *jcr, POOL_MEM &dev_name, int drive);
91 static void read_volume_label(JCR *jcr, DCR *dcr, DEVICE *dev, int Slot);
92 static void label_volume_if_ok(DCR *dcr, char *oldname,
93 char *newname, char *poolname,
94 int Slot, int relabel);
95 static bool try_autoload_device(JCR *jcr, DCR *dcr, int slot, const char *VolName);
96 static void send_dir_busy_message(BSOCK *dir, DEVICE *dev);
100 bool (*func)(JCR *jcr);
101 bool monitoraccess; /* set if monitors can access this cmd */
105 * The following are the recognized commands from the Director.
107 static struct s_cmds cmds[] = {
108 {"JobId=", job_cmd, 0}, /* start Job */
109 {"autochanger", changer_cmd, 0},
110 {"bootstrap", bootstrap_cmd, 0},
111 {"cancel", cancel_cmd, 0},
112 {".die", die_cmd, 0},
113 {"label", label_cmd, 0}, /* label a tape */
114 {"mount", mount_cmd, 0},
115 {"readlabel", readlabel_cmd, 0},
116 {"release", release_cmd, 0},
117 {"relabel", relabel_cmd, 0}, /* relabel a tape */
118 {"setdebug=", setdebug_cmd, 0}, /* set debug level */
119 {"status", status_cmd, 1},
120 {".status", qstatus_cmd, 1},
121 {"unmount", unmount_cmd, 0},
122 {"action_on_purge", action_on_purge_cmd, 0},
123 {"use storage=", use_cmd, 0},
125 // {"query", query_cmd, 0},
126 {NULL, NULL} /* list terminator */
131 * Connection request. We accept connections either from the
132 * Director or a Client (File daemon).
134 * Note, we are running as a seperate thread of the Storage daemon.
135 * and it is because a Director has made a connection with
136 * us on the "Message" channel.
138 * Basic tasks done here:
139 * - Create a JCR record
140 * - If it was from the FD, call handle_filed_connection()
141 * - Authenticate the Director
142 * - We wait for a command
143 * - We execute the command
144 * - We continue or exit depending on the return status
146 void *handle_connection_request(void *arg)
148 BSOCK *bs = (BSOCK *)arg;
156 if (bs->recv() <= 0) {
157 Emsg1(M_ERROR, 0, _("Connection request from %s failed.\n"), bs->who());
163 * Do a sanity check on the message received
165 if (bs->msglen < 25 || bs->msglen > (int)sizeof(name)) {
166 Dmsg1(000, "<filed: %s", bs->msg);
167 Emsg2(M_ERROR, 0, _("Invalid connection from %s. Len=%d\n"), bs->who(), bs->msglen);
172 * See if this is a File daemon connection. If so
175 Dmsg1(110, "Conn: %s", bs->msg);
176 if (debug_level == 3) {
177 Dmsg1(000, "<filed: %s", bs->msg);
179 if (sscanf(bs->msg, "Hello Start Job %127s", name) == 1) {
180 Dmsg1(110, "Got a FD connection at %s\n", bstrftimes(tbuf, sizeof(tbuf),
181 (utime_t)time(NULL)));
182 Dmsg1(50, "%s", bs->msg);
183 handle_filed_connection(bs, name);
188 * This is a connection from the Director, so setup a JCR
190 Dmsg1(110, "Got a DIR connection at %s\n", bstrftimes(tbuf, sizeof(tbuf),
191 (utime_t)time(NULL)));
192 jcr = new_jcr(sizeof(JCR), stored_free_jcr); /* create Job Control Record */
193 jcr->dir_bsock = bs; /* save Director bsock */
194 jcr->dir_bsock->set_jcr(jcr);
195 jcr->dcrs = New(alist(10, not_owned_by_alist));
196 /* Initialize FD start condition variable */
197 int errstat = pthread_cond_init(&jcr->job_start_wait, NULL);
200 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.bstrerror(errstat));
204 Dmsg0(1000, "stored in start_job\n");
207 * Authenticate the Director
209 if (!authenticate_director(jcr)) {
210 Jmsg(jcr, M_FATAL, 0, _("Unable to authenticate Director\n"));
213 Dmsg0(90, "Message channel init completed.\n");
215 for (quit=false; !quit;) {
217 if ((bnet_stat = bs->recv()) <= 0) {
218 break; /* connection terminated */
220 Dmsg1(199, "<dird: %s\n", bs->msg);
221 /* Ensure that device initialization is complete */
226 for (i=0; cmds[i].cmd; i++) {
227 if (strncmp(cmds[i].cmd, bs->msg, strlen(cmds[i].cmd)) == 0) {
228 if ((!cmds[i].monitoraccess) && (jcr->director->monitor)) {
229 Dmsg1(100, "Command \"%s\" is invalid.\n", cmds[i].cmd);
230 bs->fsend(invalid_cmd);
231 bs->signal(BNET_EOD);
234 Dmsg1(200, "Do command: %s\n", cmds[i].cmd);
235 if (!cmds[i].func(jcr)) { /* do command */
236 quit = true; /* error, get out */
237 Dmsg1(190, "Command %s reqeusts quit\n", cmds[i].cmd);
239 found = true; /* indicate command found */
243 if (!found) { /* command not found */
249 generate_daemon_event(jcr, "JobEnd");
250 dequeue_messages(jcr); /* send any queued messages */
251 bs->signal(BNET_TERMINATE);
258 * Force SD to die, and hopefully dump itself. Turned on only
259 * in development version.
261 static bool die_cmd(JCR *jcr)
266 Pmsg0(000, "I have been requested to die ...");
267 a = djcr->JobId; /* ref NULL pointer */
275 * Set debug level as requested by the Director
278 static bool setdebug_cmd(JCR *jcr)
280 BSOCK *dir = jcr->dir_bsock;
281 int level, trace_flag;
283 Dmsg1(10, "setdebug_cmd: %s", dir->msg);
284 if (sscanf(dir->msg, "setdebug=%d trace=%d", &level, &trace_flag) != 2 || level < 0) {
285 dir->fsend(_("3991 Bad setdebug command: %s\n"), dir->msg);
289 set_trace(trace_flag);
290 return dir->fsend(OKsetdebug, level);
297 static bool cancel_cmd(JCR *cjcr)
299 BSOCK *dir = cjcr->dir_bsock;
301 char Job[MAX_NAME_LENGTH];
304 if (sscanf(dir->msg, "cancel Job=%127s", Job) == 1) {
305 if (!(jcr=get_jcr_by_full_name(Job))) {
306 dir->fsend(_("3904 Job %s not found.\n"), Job);
308 oldStatus = jcr->JobStatus;
309 set_jcr_job_status(jcr, JS_Canceled);
310 if (!jcr->authenticated && oldStatus == JS_WaitFD) {
311 pthread_cond_signal(&jcr->job_start_wait); /* wake waiting thread */
313 if (jcr->file_bsock) {
314 bnet_sig(jcr->file_bsock, BNET_TERMINATE);
316 /* Still waiting for FD to connect, release it */
317 pthread_cond_signal(&jcr->job_start_wait); /* wake waiting job */
319 /* If thread waiting on mount, wake him */
320 if (jcr->dcr && jcr->dcr->dev && jcr->dcr->dev->waiting_for_mount()) {
321 pthread_cond_broadcast(&jcr->dcr->dev->wait_next_vol);
322 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
323 pthread_cond_broadcast(&wait_device_release);
325 if (jcr->read_dcr && jcr->read_dcr->dev && jcr->read_dcr->dev->waiting_for_mount()) {
326 pthread_cond_broadcast(&jcr->read_dcr->dev->wait_next_vol);
327 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
328 pthread_cond_broadcast(&wait_device_release);
330 Jmsg(jcr, M_INFO, 0, _("JobId=%d Job=\"%s\" marked to be canceled.\n"),
331 (int)jcr->JobId, jcr->Job);
332 dir->fsend(_("3000 Job %s marked to be canceled.\n"), jcr->Job);
336 dir->fsend(_("3903 Error scanning cancel command.\n"));
338 dir->signal(BNET_EOD);
346 static bool label_cmd(JCR *jcr)
348 return do_label(jcr, 0);
351 static bool relabel_cmd(JCR *jcr)
353 return do_label(jcr, 1);
356 static bool do_label(JCR *jcr, int relabel)
358 POOLMEM *newname, *oldname, *poolname, *mtype;
360 BSOCK *dir = jcr->dir_bsock;
367 newname = get_memory(dir->msglen+1);
368 oldname = get_memory(dir->msglen+1);
369 poolname = get_memory(dir->msglen+1);
370 mtype = get_memory(dir->msglen+1);
372 if (sscanf(dir->msg, "relabel %127s OldName=%127s NewName=%127s PoolName=%127s "
373 "MediaType=%127s Slot=%d drive=%d",
374 dev_name.c_str(), oldname, newname, poolname, mtype,
375 &slot, &drive) == 7) {
380 if (sscanf(dir->msg, "label %127s VolumeName=%127s PoolName=%127s "
381 "MediaType=%127s Slot=%d drive=%d",
382 dev_name.c_str(), newname, poolname, mtype, &slot, &drive) == 6) {
387 unbash_spaces(newname);
388 unbash_spaces(oldname);
389 unbash_spaces(poolname);
390 unbash_spaces(mtype);
391 dcr = find_device(jcr, dev_name, drive);
394 dev->dlock(); /* Use P to avoid indefinite block */
395 if (!dev->is_open() && !dev->is_busy()) {
396 Dmsg1(400, "Can %slabel. Device is not open\n", relabel?"re":"");
397 label_volume_if_ok(dcr, oldname, newname, poolname, slot, relabel);
399 /* Under certain "safe" conditions, we can steal the lock */
400 } else if (dev->can_steal_lock()) {
401 Dmsg0(400, "Can relabel. can_steal_lock\n");
402 label_volume_if_ok(dcr, oldname, newname, poolname, slot, relabel);
403 } else if (dev->is_busy() || dev->is_blocked()) {
404 send_dir_busy_message(dir, dev);
405 } else { /* device not being used */
406 Dmsg0(400, "Can relabel. device not used\n");
407 label_volume_if_ok(dcr, oldname, newname, poolname, slot, relabel);
412 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), dev_name.c_str());
415 /* NB dir->msg gets clobbered in bnet_fsend, so save command */
416 pm_strcpy(jcr->errmsg, dir->msg);
417 dir->fsend(_("3903 Error scanning label command: %s\n"), jcr->errmsg);
419 free_memory(oldname);
420 free_memory(newname);
421 free_memory(poolname);
423 dir->signal(BNET_EOD);
428 * Read the tape label and determine if we can safely
429 * label the tape (not a Bacula volume), then label it.
431 * Enter with the mutex set
433 static void label_volume_if_ok(DCR *dcr, char *oldname,
434 char *newname, char *poolname,
435 int slot, int relabel)
437 BSOCK *dir = dcr->jcr->dir_bsock;
439 DEVICE *dev = dcr->dev;
442 const char *volname = (relabel == 1) ? oldname : newname;
445 steal_device_lock(dev, &hold, BST_WRITING_LABEL);
446 Dmsg1(100, "Stole device %s lock, writing label.\n", dev->print_name());
449 Dmsg0(90, "try_autoload_device - looking for volume_info\n");
450 if (!try_autoload_device(dcr->jcr, dcr, slot, volname)) {
451 goto bail_out; /* error */
454 /* Ensure that the device is open -- autoload_device() closes it */
455 if (dev->is_tape()) {
456 mode = OPEN_READ_WRITE;
458 mode = CREATE_READ_WRITE;
462 dev->truncating = true; /* let open() know we will truncate it */
464 /* Set old volume name for open if relabeling */
465 bstrncpy(dcr->VolCatInfo.VolCatName, volname, sizeof(dcr->VolCatInfo.VolCatName));
466 if (dev->open(dcr, mode) < 0) {
467 dir->fsend(_("3910 Unable to open device %s: ERR=%s\n"),
468 dev->print_name(), dev->bstrerror());
472 /* See what we have for a Volume */
473 label_status = read_dev_volume_label(dcr);
475 /* Set new volume name */
476 bstrncpy(dcr->VolCatInfo.VolCatName, newname, sizeof(dcr->VolCatInfo.VolCatName));
477 switch(label_status) {
479 case VOL_VERSION_ERROR:
480 case VOL_LABEL_ERROR:
484 "3920 Cannot label Volume because it is already labeled: \"%s\"\n"),
485 dev->VolHdr.VolumeName);
489 /* Relabel request. If oldname matches, continue */
490 if (strcmp(oldname, dev->VolHdr.VolumeName) != 0) {
491 dir->fsend(_("3921 Wrong volume mounted.\n"));
494 if (dev->label_type != B_BACULA_LABEL) {
495 dir->fsend(_("3922 Cannot relabel an ANSI/IBM labeled Volume.\n"));
498 /* Fall through wanted! */
501 if (!write_new_volume_label_to_dev(dcr, newname, poolname,
502 relabel, true /* write dvd now */)) {
503 dir->fsend(_("3912 Failed to label Volume: ERR=%s\n"), dev->bstrerror());
506 bstrncpy(dcr->VolumeName, newname, sizeof(dcr->VolumeName));
507 /* The following 3000 OK label. string is scanned in ua_label.c */
508 dir->fsend("3000 OK label. VolBytes=%s DVD=%d Volume=\"%s\" Device=%s\n",
509 edit_uint64(dev->VolCatInfo.VolCatBytes, ed1),
510 dev->is_dvd()?1:0, newname, dev->print_name());
513 dir->fsend(_("3914 Failed to label Volume (no media): ERR=%s\n"), dev->bstrerror());
516 dir->fsend(_("3913 Cannot label Volume. "
517 "Unknown status %d from read_volume_label()\n"), label_status);
522 if (!dev->is_open()) {
525 volume_unused(dcr); /* no longer using volume */
526 give_back_device_lock(dev, &hold);
532 * Read the tape label
534 * Enter with the mutex set
536 static bool read_label(DCR *dcr)
540 BSOCK *dir = jcr->dir_bsock;
542 DEVICE *dev = dcr->dev;
544 steal_device_lock(dev, &hold, BST_DOING_ACQUIRE);
546 dcr->VolumeName[0] = 0;
547 dev->clear_labeled(); /* force read of label */
548 switch (read_dev_volume_label(dcr)) {
550 dir->fsend(_("3001 Mounted Volume: %s\n"), dev->VolHdr.VolumeName);
554 dir->fsend(_("3902 Cannot mount Volume on Storage Device %s because:\n%s"),
555 dev->print_name(), jcr->errmsg);
560 give_back_device_lock(dev, &hold);
565 * Searches for device by name, and if found, creates a dcr and
568 static DCR *find_device(JCR *jcr, POOL_MEM &devname, int drive)
571 AUTOCHANGER *changer;
575 unbash_spaces(devname);
576 foreach_res(device, R_DEVICE) {
577 /* Find resource, and make sure we were able to open it */
578 if (strcmp(device->hdr.name, devname.c_str()) == 0) {
580 device->dev = init_dev(jcr, device);
583 Jmsg(jcr, M_WARNING, 0, _("\n"
584 " Device \"%s\" requested by DIR could not be opened or does not exist.\n"),
588 Dmsg1(20, "Found device %s\n", device->hdr.name);
594 foreach_res(changer, R_AUTOCHANGER) {
595 /* Find resource, and make sure we were able to open it */
596 if (strcmp(devname.c_str(), changer->hdr.name) == 0) {
597 /* Try each device in this AutoChanger */
598 foreach_alist(device, changer->device) {
599 Dmsg1(100, "Try changer device %s\n", device->hdr.name);
601 device->dev = init_dev(jcr, device);
604 Dmsg1(100, "Device %s could not be opened. Skipped\n", devname.c_str());
605 Jmsg(jcr, M_WARNING, 0, _("\n"
606 " Device \"%s\" in changer \"%s\" requested by DIR could not be opened or does not exist.\n"),
607 device->hdr.name, devname.c_str());
610 if (!device->dev->autoselect) {
611 Dmsg1(100, "Device %s not autoselect skipped.\n", devname.c_str());
612 continue; /* device is not available */
614 if (drive < 0 || drive == (int)device->dev->drive_index) {
615 Dmsg1(20, "Found changer device %s\n", device->hdr.name);
619 Dmsg3(100, "Device %s drive wrong: want=%d got=%d skipping\n",
620 devname.c_str(), drive, (int)device->dev->drive_index);
622 break; /* we found it but could not open a device */
628 Dmsg1(100, "Found device %s\n", device->hdr.name);
629 dcr = new_dcr(jcr, NULL, device->dev);
630 dcr->device = device;
637 * Mount command from Director
639 static bool mount_cmd(JCR *jcr)
642 BSOCK *dir = jcr->dir_bsock;
649 ok = sscanf(dir->msg, "mount %127s drive=%d slot=%d", devname.c_str(),
652 ok = sscanf(dir->msg, "mount %127s drive=%d", devname.c_str(), &drive) == 2;
655 dcr = find_device(jcr, devname, drive);
658 dev->dlock(); /* Use P to avoid indefinite block */
659 Dmsg2(100, "mount cmd blocked=%d must_unload=%d\n", dev->blocked(),
661 switch (dev->blocked()) { /* device blocked? */
662 case BST_WAITING_FOR_SYSOP:
663 /* Someone is waiting, wake him */
664 Dmsg0(100, "Waiting for mount. Attempting to wake thread\n");
665 dev->set_blocked(BST_MOUNT);
666 dir->fsend("3001 OK mount requested. %sDevice=%s\n",
667 slot>0?_("Specified slot ignored. "):"",
669 pthread_cond_broadcast(&dev->wait_next_vol);
670 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)dcr->jcr->JobId);
671 pthread_cond_broadcast(&wait_device_release);
674 /* In both of these two cases, we (the user) unmounted the Volume */
675 case BST_UNMOUNTED_WAITING_FOR_SYSOP:
677 if (dev->is_autochanger() && slot > 0) {
678 try_autoload_device(jcr, dcr, slot, "");
680 /* We freed the device, so reopen it and wake any waiting threads */
681 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
682 dir->fsend(_("3901 Unable to open device %s: ERR=%s\n"),
683 dev->print_name(), dev->bstrerror());
684 if (dev->blocked() == BST_UNMOUNTED) {
685 /* We blocked the device, so unblock it */
686 Dmsg0(100, "Unmounted. Unblocking device\n");
691 read_dev_volume_label(dcr);
692 if (dev->blocked() == BST_UNMOUNTED) {
693 /* We blocked the device, so unblock it */
694 Dmsg0(100, "Unmounted. Unblocking device\n");
695 read_label(dcr); /* this should not be necessary */
698 Dmsg0(100, "Unmounted waiting for mount. Attempting to wake thread\n");
699 dev->set_blocked(BST_MOUNT);
701 if (dev->is_labeled()) {
702 dir->fsend(_("3001 Device %s is mounted with Volume \"%s\"\n"),
703 dev->print_name(), dev->VolHdr.VolumeName);
705 dir->fsend(_("3905 Device %s open but no Bacula volume is mounted.\n"
706 "If this is not a blank tape, try unmounting and remounting the Volume.\n"),
709 pthread_cond_broadcast(&dev->wait_next_vol);
710 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)dcr->jcr->JobId);
711 pthread_cond_broadcast(&wait_device_release);
714 case BST_DOING_ACQUIRE:
715 dir->fsend(_("3001 Device %s is doing acquire.\n"),
719 case BST_WRITING_LABEL:
720 dir->fsend(_("3903 Device %s is being labeled.\n"),
724 case BST_NOT_BLOCKED:
725 if (dev->is_autochanger() && slot > 0) {
726 try_autoload_device(jcr, dcr, slot, "");
728 if (dev->is_open()) {
729 if (dev->is_labeled()) {
730 dir->fsend(_("3001 Device %s is mounted with Volume \"%s\"\n"),
731 dev->print_name(), dev->VolHdr.VolumeName);
733 dir->fsend(_("3905 Device %s open but no Bacula volume is mounted.\n"
734 "If this is not a blank tape, try unmounting and remounting the Volume.\n"),
737 } else if (dev->is_tape()) {
738 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
739 dir->fsend(_("3901 Unable to open device %s: ERR=%s\n"),
740 dev->print_name(), dev->bstrerror());
744 if (dev->is_labeled()) {
745 dir->fsend(_("3001 Device %s is already mounted with Volume \"%s\"\n"),
746 dev->print_name(), dev->VolHdr.VolumeName);
748 dir->fsend(_("3905 Device %s open but no Bacula volume is mounted.\n"
749 "If this is not a blank tape, try unmounting and remounting the Volume.\n"),
752 } else if (dev->is_unmountable()) {
754 dir->fsend(_("3002 Device %s is mounted.\n"),
757 dir->fsend(_("3907 %s"), dev->bstrerror());
759 } else { /* must be file */
760 dir->fsend(_("3906 File device %s is always mounted.\n"),
762 pthread_cond_broadcast(&dev->wait_next_vol);
763 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)dcr->jcr->JobId);
764 pthread_cond_broadcast(&wait_device_release);
769 dir->fsend(_("3905 Bizarre wait state %d\n"), dev->blocked());
775 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
778 pm_strcpy(jcr->errmsg, dir->msg);
779 dir->fsend(_("3909 Error scanning mount command: %s\n"), jcr->errmsg);
781 dir->signal(BNET_EOD);
786 * unmount command from Director
788 static bool unmount_cmd(JCR *jcr)
791 BSOCK *dir = jcr->dir_bsock;
796 if (sscanf(dir->msg, "unmount %127s drive=%d", devname.c_str(), &drive) == 2) {
797 dcr = find_device(jcr, devname, drive);
800 dev->dlock(); /* Use P to avoid indefinite block */
801 if (!dev->is_open()) {
802 if (!dev->is_busy()) {
803 unload_autochanger(dcr, -1);
805 if (dev->is_unmountable()) {
806 if (dev->unmount(0)) {
807 dir->fsend(_("3002 Device %s unmounted.\n"),
810 dir->fsend(_("3907 %s"), dev->bstrerror());
813 Dmsg0(90, "Device already unmounted\n");
814 dir->fsend(_("3901 Device %s is already unmounted.\n"),
817 } else if (dev->blocked() == BST_WAITING_FOR_SYSOP) {
818 Dmsg2(90, "%d waiter dev_block=%d. doing unmount\n", dev->num_waiting,
820 if (!unload_autochanger(dcr, -1)) {
821 /* ***FIXME**** what is this ???? */
825 if (dev->is_unmountable() && !dev->unmount(0)) {
826 dir->fsend(_("3907 %s"), dev->bstrerror());
828 dev->set_blocked(BST_UNMOUNTED_WAITING_FOR_SYSOP);
829 dir->fsend(_("3001 Device %s unmounted.\n"),
833 } else if (dev->blocked() == BST_DOING_ACQUIRE) {
834 dir->fsend(_("3902 Device %s is busy in acquire.\n"),
837 } else if (dev->blocked() == BST_WRITING_LABEL) {
838 dir->fsend(_("3903 Device %s is being labeled.\n"),
841 } else if (dev->is_busy()) {
842 send_dir_busy_message(dir, dev);
843 } else { /* device not being used */
844 Dmsg0(90, "Device not in use, unmounting\n");
845 /* On FreeBSD, I am having ASSERT() failures in block_device()
846 * and I can only imagine that the thread id that we are
847 * leaving in no_wait_id is being re-used. So here,
848 * we simply do it by hand. Gross, but a solution.
850 /* block_device(dev, BST_UNMOUNTED); replace with 2 lines below */
851 dev->set_blocked(BST_UNMOUNTED);
852 clear_thread_id(dev->no_wait_id);
853 if (!unload_autochanger(dcr, -1)) {
857 if (dev->is_unmountable() && !dev->unmount(0)) {
858 dir->fsend(_("3907 %s"), dev->bstrerror());
860 dir->fsend(_("3002 Device %s unmounted.\n"),
867 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
870 /* NB dir->msg gets clobbered in bnet_fsend, so save command */
871 pm_strcpy(jcr->errmsg, dir->msg);
872 dir->fsend(_("3907 Error scanning unmount command: %s\n"), jcr->errmsg);
874 dir->signal(BNET_EOD);
879 * The truncate command will recycle a volume. The director can call this
880 * after purging a volume so that disk space will not be wasted. Only useful
881 * for File Storage, of course.
884 static bool action_on_purge_cmd(JCR *jcr)
888 BSOCK *dir = jcr->dir_bsock;
893 devname.check_size(dir->msglen+1);
894 volumename.check_size(dir->msglen+1);
896 if (sscanf(dir->msg, "action_on_purge %127s vol=%s action=%d",
897 devname.c_str(), volumename.c_str(), &action) != 3) {
898 dir->fsend(_("3916 Error scanning action_on_purge command\n"));
901 unbash_spaces(volumename.c_str());
902 unbash_spaces(devname.c_str());
904 /* FIXME: autochanger, drive = 0? how can we avoid that? we only work on
907 if ((dcr = find_device(jcr, devname, 0)) == NULL) {
908 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
914 /* Store the VolumeName for opening and re-labeling the volume */
915 bstrncpy(dcr->VolumeName, volumename.c_str(), sizeof(dcr->VolumeName));
916 bstrncpy(dev->VolHdr.VolumeName, volumename.c_str(), sizeof(dev->VolHdr.VolumeName));
918 /* Re-write the label with the recycle flag */
919 if (rewrite_volume_label(dcr, true)) {
920 dir->fsend(_("3917 Volume recycled\n"));
922 dir->fsend(_("3918 Recycle failed\n"));
926 dir->signal(BNET_EOD);
931 * Release command from Director. This rewinds the device and if
932 * configured does a offline and ensures that Bacula will
933 * re-read the label of the tape before continuing. This gives
934 * the operator the chance to change the tape anytime before the
937 static bool release_cmd(JCR *jcr)
940 BSOCK *dir = jcr->dir_bsock;
945 if (sscanf(dir->msg, "release %127s drive=%d", devname.c_str(), &drive) == 2) {
946 dcr = find_device(jcr, devname, drive);
949 dev->dlock(); /* Use P to avoid indefinite block */
950 if (!dev->is_open()) {
951 if (!dev->is_busy()) {
952 unload_autochanger(dcr, -1);
954 Dmsg0(90, "Device already released\n");
955 dir->fsend(_("3921 Device %s already released.\n"),
958 } else if (dev->blocked() == BST_WAITING_FOR_SYSOP) {
959 Dmsg2(90, "%d waiter dev_block=%d.\n", dev->num_waiting,
961 unload_autochanger(dcr, -1);
962 dir->fsend(_("3922 Device %s waiting for sysop.\n"),
965 } else if (dev->blocked() == BST_UNMOUNTED_WAITING_FOR_SYSOP) {
966 Dmsg2(90, "%d waiter dev_block=%d. doing unmount\n", dev->num_waiting,
968 dir->fsend(_("3922 Device %s waiting for mount.\n"),
971 } else if (dev->blocked() == BST_DOING_ACQUIRE) {
972 dir->fsend(_("3923 Device %s is busy in acquire.\n"),
975 } else if (dev->blocked() == BST_WRITING_LABEL) {
976 dir->fsend(_("3914 Device %s is being labeled.\n"),
979 } else if (dev->is_busy()) {
980 send_dir_busy_message(dir, dev);
981 } else { /* device not being used */
982 Dmsg0(90, "Device not in use, releasing\n");
983 dcr->release_volume();
984 dir->fsend(_("3022 Device %s released.\n"),
990 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
993 /* NB dir->msg gets clobbered in bnet_fsend, so save command */
994 pm_strcpy(jcr->errmsg, dir->msg);
995 dir->fsend(_("3927 Error scanning release command: %s\n"), jcr->errmsg);
997 dir->signal(BNET_EOD);
1001 static pthread_mutex_t bsr_mutex = PTHREAD_MUTEX_INITIALIZER;
1002 static uint32_t bsr_uniq = 0;
1004 static bool get_bootstrap_file(JCR *jcr, BSOCK *sock)
1006 POOLMEM *fname = get_pool_memory(PM_FNAME);
1010 if (jcr->RestoreBootstrap) {
1011 unlink(jcr->RestoreBootstrap);
1012 free_pool_memory(jcr->RestoreBootstrap);
1016 Mmsg(fname, "%s/%s.%s.%d.bootstrap", me->working_directory, me->hdr.name,
1017 jcr->Job, bsr_uniq);
1019 Dmsg1(400, "bootstrap=%s\n", fname);
1020 jcr->RestoreBootstrap = fname;
1021 bs = fopen(fname, "a+b"); /* create file */
1024 Jmsg(jcr, M_FATAL, 0, _("Could not create bootstrap file %s: ERR=%s\n"),
1025 jcr->RestoreBootstrap, be.bstrerror());
1028 Dmsg0(10, "=== Bootstrap file ===\n");
1029 while (sock->recv() >= 0) {
1030 Dmsg1(10, "%s", sock->msg);
1031 fputs(sock->msg, bs);
1034 Dmsg0(10, "=== end bootstrap file ===\n");
1035 jcr->bsr = parse_bsr(jcr, jcr->RestoreBootstrap);
1037 Jmsg(jcr, M_FATAL, 0, _("Error parsing bootstrap file.\n"));
1040 if (debug_level >= 10) {
1041 dump_bsr(jcr->bsr, true);
1043 /* If we got a bootstrap, we are reading, so create read volume list */
1044 create_restore_volume_list(jcr);
1048 unlink(jcr->RestoreBootstrap);
1049 free_pool_memory(jcr->RestoreBootstrap);
1050 jcr->RestoreBootstrap = NULL;
1052 sock->fsend(ERROR_bootstrap);
1055 return sock->fsend(OK_bootstrap);
1058 static bool bootstrap_cmd(JCR *jcr)
1060 return get_bootstrap_file(jcr, jcr->dir_bsock);
1064 * Autochanger command from Director
1066 static bool changer_cmd(JCR *jcr)
1069 BSOCK *dir = jcr->dir_bsock;
1072 const char *cmd = NULL;
1075 * A safe_cmd may call autochanger script but does not load/unload
1076 * slots so it can be done at the same time that the drive is open.
1078 bool safe_cmd = false;
1080 if (sscanf(dir->msg, "autochanger listall %127s", devname.c_str()) == 1) {
1082 safe_cmd = ok = true;
1083 } else if (sscanf(dir->msg, "autochanger list %127s", devname.c_str()) == 1) {
1085 safe_cmd = ok = true;
1086 } else if (sscanf(dir->msg, "autochanger slots %127s", devname.c_str()) == 1) {
1088 safe_cmd = ok = true;
1089 } else if (sscanf(dir->msg, "autochanger drives %127s", devname.c_str()) == 1) {
1091 safe_cmd = ok = true;
1094 dcr = find_device(jcr, devname, -1);
1097 dev->dlock(); /* Use P to avoid indefinite block */
1098 if (!dev->device->changer_res) {
1099 dir->fsend(_("3995 Device %s is not an autochanger.\n"),
1101 /* Under certain "safe" conditions, we can steal the lock */
1102 } else if (safe_cmd || !dev->is_open() || dev->can_steal_lock()) {
1103 autochanger_cmd(dcr, dir, cmd);
1104 } else if (dev->is_busy() || dev->is_blocked()) {
1105 send_dir_busy_message(dir, dev);
1106 } else { /* device not being used */
1107 autochanger_cmd(dcr, dir, cmd);
1112 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
1114 } else { /* error on scanf */
1115 pm_strcpy(jcr->errmsg, dir->msg);
1116 dir->fsend(_("3908 Error scanning autocharger drives/list/slots command: %s\n"),
1119 dir->signal(BNET_EOD);
1124 * Read and return the Volume label
1126 static bool readlabel_cmd(JCR *jcr)
1129 BSOCK *dir = jcr->dir_bsock;
1135 if (sscanf(dir->msg, "readlabel %127s Slot=%d drive=%d", devname.c_str(),
1136 &Slot, &drive) == 3) {
1137 dcr = find_device(jcr, devname, drive);
1140 dev->dlock(); /* Use P to avoid indefinite block */
1141 if (!dev->is_open()) {
1142 read_volume_label(jcr, dcr, dev, Slot);
1144 /* Under certain "safe" conditions, we can steal the lock */
1145 } else if (dev->can_steal_lock()) {
1146 read_volume_label(jcr, dcr, dev, Slot);
1147 } else if (dev->is_busy() || dev->is_blocked()) {
1148 send_dir_busy_message(dir, dev);
1149 } else { /* device not being used */
1150 read_volume_label(jcr, dcr, dev, Slot);
1155 dir->fsend(_("3999 Device \"%s\" not found or could not be opened.\n"), devname.c_str());
1158 pm_strcpy(jcr->errmsg, dir->msg);
1159 dir->fsend(_("3909 Error scanning readlabel command: %s\n"), jcr->errmsg);
1161 dir->signal(BNET_EOD);
1167 * Read the tape label
1169 * Enter with the mutex set
1171 static void read_volume_label(JCR *jcr, DCR *dcr, DEVICE *dev, int Slot)
1173 BSOCK *dir = jcr->dir_bsock;
1177 steal_device_lock(dev, &hold, BST_WRITING_LABEL);
1179 if (!try_autoload_device(jcr, dcr, Slot, "")) {
1180 goto bail_out; /* error */
1183 dev->clear_labeled(); /* force read of label */
1184 switch (read_dev_volume_label(dcr)) {
1186 /* DO NOT add quotes around the Volume name. It is scanned in the DIR */
1187 dir->fsend(_("3001 Volume=%s Slot=%d\n"), dev->VolHdr.VolumeName, Slot);
1188 Dmsg1(100, "Volume: %s\n", dev->VolHdr.VolumeName);
1191 dir->fsend(_("3902 Cannot mount Volume on Storage Device %s because:\n%s"),
1192 dev->print_name(), jcr->errmsg);
1197 give_back_device_lock(dev, &hold);
1201 static bool try_autoload_device(JCR *jcr, DCR *dcr, int slot, const char *VolName)
1203 BSOCK *dir = jcr->dir_bsock;
1205 bstrncpy(dcr->VolumeName, VolName, sizeof(dcr->VolumeName));
1206 dcr->VolCatInfo.Slot = slot;
1207 dcr->VolCatInfo.InChanger = slot > 0;
1208 if (autoload_device(dcr, 0, dir) < 0) { /* autoload if possible */
1214 static void send_dir_busy_message(BSOCK *dir, DEVICE *dev)
1216 if (dev->is_blocked()) {
1217 switch (dev->blocked()) {
1219 dir->fsend(_("3931 Device %s is BLOCKED. user unmounted.\n"),
1222 case BST_UNMOUNTED_WAITING_FOR_SYSOP:
1223 dir->fsend(_("3932 Device %s is BLOCKED. user unmounted during wait for media/mount.\n"),
1226 case BST_WAITING_FOR_SYSOP:
1227 dir->fsend(_("3933 Device %s is BLOCKED waiting for media.\n"),
1230 case BST_DOING_ACQUIRE:
1231 dir->fsend(_("3934 Device %s is being initialized.\n"),
1234 case BST_WRITING_LABEL:
1235 dir->fsend(_("3935 Device %s is blocked labeling a Volume.\n"),
1239 dir->fsend(_("3935 Device %s is blocked for unknown reason.\n"),
1243 } else if (dev->can_read()) {
1244 dir->fsend(_("3936 Device %s is busy reading.\n"),
1245 dev->print_name());;
1247 dir->fsend(_("3937 Device %s is busy with %d writer(s).\n"),
1248 dev->print_name(), dev->num_writers);