2 Bacula® - The Network Backup Solution
4 Copyright (C) 2002-2010 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 three of the GNU Affero 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 Affero 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 * Routines to acquire and release a device for read/write
31 * Kern Sibbald, August MMII
35 #include "bacula.h" /* pull in global headers */
36 #include "stored.h" /* pull in Storage Deamon headers */
38 /* Forward referenced functions */
39 static void attach_dcr_to_dev(DCR *dcr);
40 static void detach_dcr_from_dev(DCR *dcr);
41 static void set_dcr_from_vol(DCR *dcr, VOL_LIST *vol);
44 /*********************************************************************
45 * Acquire device for reading.
46 * The drive should have previously been reserved by calling
47 * reserve_device_for_read(). We read the Volume label from the block and
48 * leave the block pointers just after the label.
50 * Returns: NULL if failed for any reason
53 bool acquire_device_for_read(DCR *dcr)
55 DEVICE *dev = dcr->dev;
58 bool tape_previously_mounted;
59 bool tape_initially_mounted;
61 bool try_autochanger = true;
66 Dmsg2(950, "dcr=%p dev=%p\n", dcr, dcr->dev);
67 Dmsg2(950, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
68 dev->dblock(BST_DOING_ACQUIRE);
70 if (dev->num_writers > 0) {
71 Jmsg2(jcr, M_FATAL, 0, _("Acquire read: num_writers=%d not zero. Job %d canceled.\n"),
72 dev->num_writers, jcr->JobId);
76 /* Find next Volume, if any */
80 Jmsg(jcr, M_FATAL, 0, _("No volumes specified for reading. Job %s canceled.\n"),
81 edit_int64(jcr->JobId, ed1));
85 for (i=1; i<jcr->CurReadVolume; i++) {
89 Jmsg(jcr, M_FATAL, 0, _("Logic error: no next volume to read. Numvol=%d Curvol=%d\n"),
90 jcr->NumReadVolumes, jcr->CurReadVolume);
91 goto get_out; /* should not happen */
93 set_dcr_from_vol(dcr, vol);
95 Dmsg2(100, "Want Vol=%s Slot=%d\n", vol->VolumeName, vol->Slot);
98 * If the MediaType requested for this volume is not the
99 * same as the current drive, we attempt to find the same
100 * device that was used to write the orginal volume. If
101 * found, we switch to using that device.
103 * N.B. A lot of routines rely on the dcr pointer not changing
104 * read_records.c even has multiple dcrs cached, so we take care
105 * here to release all important parts of the dcr and re-acquire
106 * them such as the block pointer (size may change), but we do
107 * not release the dcr.
109 Dmsg2(50, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
110 if (dcr->media_type[0] && strcmp(dcr->media_type, dev->device->media_type) != 0) {
115 Jmsg3(jcr, M_INFO, 0, _("Changing read device. Want Media Type=\"%s\" have=\"%s\"\n"
117 dcr->media_type, dev->device->media_type, dev->print_name());
118 Dmsg3(50, "Changing read device. Want Media Type=\"%s\" have=\"%s\"\n"
120 dcr->media_type, dev->device->media_type, dev->print_name());
122 dev->dunblock(DEV_UNLOCKED);
125 memset(&rctx, 0, sizeof(RCTX));
128 jcr->reserve_msgs = New(alist(10, not_owned_by_alist));
129 rctx.any_drive = true;
130 rctx.device_name = vol->device;
131 store = new DIRSTORE;
132 memset(store, 0, sizeof(DIRSTORE));
133 store->name[0] = 0; /* No dir name */
134 bstrncpy(store->media_type, vol->MediaType, sizeof(store->media_type));
135 bstrncpy(store->pool_name, dcr->pool_name, sizeof(store->pool_name));
136 bstrncpy(store->pool_type, dcr->pool_type, sizeof(store->pool_type));
137 store->append = false;
139 clean_device(dcr); /* clean up the dcr */
142 * Search for a new device
144 stat = search_res_for_device(rctx);
145 release_reserve_messages(jcr); /* release queued messages */
146 unlock_reservations();
149 dev = dcr->dev; /* get new device pointer */
150 dev->dblock(BST_DOING_ACQUIRE);
151 dcr->VolumeName[0] = 0;
152 Jmsg(jcr, M_INFO, 0, _("Media Type change. New read device %s chosen.\n"),
154 Dmsg1(50, "Media Type change. New read device %s chosen.\n", dev->print_name());
156 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
157 dcr->setVolCatName(vol->VolumeName);
158 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
159 dcr->VolCatInfo.Slot = vol->Slot;
160 dcr->VolCatInfo.InChanger = vol->Slot > 0;
161 bstrncpy(dcr->pool_name, store->pool_name, sizeof(dcr->pool_name));
162 bstrncpy(dcr->pool_type, store->pool_type, sizeof(dcr->pool_type));
165 Jmsg1(jcr, M_FATAL, 0, _("No suitable device found to read Volume \"%s\"\n"),
167 Dmsg1(50, "No suitable device found to read Volume \"%s\"\n", vol->VolumeName);
171 Dmsg2(400, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
175 if (dev->vol && dev->vol->is_swapping()) {
176 dev->vol->set_slot(vol->Slot);
177 Dmsg3(100, "swapping: slot=%d Vol=%s dev=%s\n", dev->vol->get_slot(),
178 dev->vol->vol_name, dev->print_name());
181 init_device_wait_timers(dcr);
183 tape_previously_mounted = dev->can_read() || dev->can_append() ||
185 tape_initially_mounted = tape_previously_mounted;
188 /* Volume info is always needed because of VolParts */
189 Dmsg1(150, "dir_get_volume_info vol=%s\n", dcr->VolumeName);
190 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_READ)) {
191 Dmsg2(150, "dir_get_vol_info failed for vol=%s: %s\n",
192 dcr->VolumeName, jcr->errmsg);
193 Jmsg1(jcr, M_WARNING, 0, "Read acquire: %s", jcr->errmsg);
195 dev->set_load(); /* set to load volume */
198 /* If not polling limit retries */
199 if (!dev->poll && retry++ > 10) {
202 dev->clear_labeled(); /* force reread of label */
203 if (job_canceled(jcr)) {
205 Mmsg1(dev->errmsg, _("Job %s canceled.\n"), edit_int64(jcr->JobId, ed1));
206 Jmsg(jcr, M_INFO, 0, dev->errmsg);
207 goto get_out; /* error return */
211 dcr->do_swapping(false/*!is_writing*/);
212 dcr->do_load(false /*!is_writing*/);
213 set_dcr_from_vol(dcr, vol); /* refresh dcr with desired volume info */
216 * This code ensures that the device is ready for
217 * reading. If it is a file, it opens it.
218 * If it is a tape, it checks the volume name
220 Dmsg1(100, "bstored: open vol=%s\n", dcr->VolumeName);
221 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
223 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
224 dev->print_name(), dcr->VolumeName, dev->bstrerror());
228 Dmsg1(50, "opened dev %s OK\n", dev->print_name());
230 /* Read Volume Label */
231 Dmsg0(50, "calling read-vol-label\n");
232 vol_label_status = read_dev_volume_label(dcr);
233 switch (vol_label_status) {
235 Dmsg0(50, "Got correct volume.\n");
237 dev->VolCatInfo = dcr->VolCatInfo; /* structure assignment */
240 Dmsg0(50, "IO Error\n");
242 * Send error message generated by read_dev_volume_label()
243 * only we really had a tape mounted. This supresses superfluous
244 * error messages when nothing is mounted.
246 if (tape_previously_mounted) {
247 Jmsg(jcr, M_WARNING, 0, "Read acquire: %s", jcr->errmsg);
251 Dmsg3(50, "Vol name=%s want=%s drv=%s.\n", dev->VolHdr.VolumeName,
252 dcr->VolumeName, dev->print_name());
253 if (dev->is_volume_to_unload()) {
256 dev->set_unload(); /* force unload of unwanted tape */
257 if (!unload_autochanger(dcr, -1)) {
258 /* at least free the device so we can re-open with correct volume */
265 Jmsg1(jcr, M_WARNING, 0, "Read acquire: %s", jcr->errmsg);
267 Dmsg0(50, "default path\n");
268 tape_previously_mounted = true;
271 * If the device requires mount, close it, so the device can be ejected.
273 if (dev->requires_mount()) {
278 /* Call autochanger only once unless ask_sysop called */
279 if (try_autochanger) {
281 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
282 dcr->VolumeName, dcr->VolCatInfo.Slot);
283 stat = autoload_device(dcr, 0, NULL);
285 try_autochanger = false;
286 continue; /* try reading volume mounted */
290 /* Mount a specific volume and no other */
291 Dmsg0(200, "calling dir_ask_sysop\n");
292 if (!dir_ask_sysop_to_mount_volume(dcr, ST_READ)) {
293 goto get_out; /* error return */
296 /* Volume info is always needed because of VolParts */
297 Dmsg1(150, "dir_get_volume_info vol=%s\n", dcr->VolumeName);
298 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_READ)) {
299 Dmsg2(150, "dir_get_vol_info failed for vol=%s: %s\n",
300 dcr->VolumeName, jcr->errmsg);
301 Jmsg1(jcr, M_WARNING, 0, "Read acquire: %s", jcr->errmsg);
303 dev->set_load(); /* set to load volume */
305 try_autochanger = true; /* permit trying the autochanger again */
307 continue; /* try reading again */
313 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
320 jcr->setJobStatus(JS_Running);
321 dir_send_job_status(jcr);
322 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
323 dcr->VolumeName, dev->print_name());
327 dcr->clear_reserved();
329 * Normally we are blocked, but in at least one error case above
330 * we are not blocked because we unsuccessfully tried changing
333 if (dev->is_blocked()) {
334 dev->dunblock(DEV_LOCKED);
336 dev->dunlock(); /* dunblock() unlock the device too */
338 Dmsg2(950, "dcr=%p dev=%p\n", dcr, dcr->dev);
339 Dmsg2(950, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
344 * Acquire device for writing. We permit multiple writers.
345 * If this is the first one, we read the label.
347 * Returns: NULL if failed for any reason
349 * Note, normally reserve_device_for_append() is called
350 * before this routine.
352 DCR *acquire_device_for_append(DCR *dcr)
354 DEVICE *dev = dcr->dev;
357 bool have_vol = false;
359 init_device_wait_timers(dcr);
361 P(dev->acquire_mutex); /* only one job at a time */
363 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
364 (dev->is_dvd()?"DVD":"disk"));
367 * With the reservation system, this should not happen
369 if (dev->can_read()) {
370 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
371 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
378 * have_vol defines whether or not mount_next_write_volume should
379 * ask the Director again about what Volume to use.
381 if (dev->can_append() && dcr->is_suitable_volume_mounted() &&
382 strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") != 0) {
383 Dmsg0(190, "device already in append.\n");
385 * At this point, the correct tape is already mounted, so
386 * we do not need to do mount_next_write_volume(), unless
387 * we need to recycle the tape.
389 if (dev->num_writers == 0) {
390 dev->VolCatInfo = dcr->VolCatInfo; /* structure assignment */
392 have_vol = dcr->is_tape_position_ok();
397 block_device(dev, BST_DOING_ACQUIRE);
399 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
400 if (!dcr->mount_next_write_volume()) {
401 if (!job_canceled(jcr)) {
402 /* Reduce "noise" -- don't print if job canceled */
403 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
405 Dmsg1(200, "Could not ready device %s for append.\n",
412 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
417 dev->num_writers++; /* we are now a writer */
418 if (jcr->NumWriteVolumes == 0) {
419 jcr->NumWriteVolumes = 1;
421 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
422 Dmsg4(100, "=== nwriters=%d nres=%d vcatjob=%d dev=%s\n",
423 dev->num_writers, dev->num_reserved(), dev->VolCatInfo.VolCatJobs,
425 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
429 dcr->clear_reserved();
431 V(dev->acquire_mutex);
432 return ok ? dcr : NULL;
436 * This job is done, so release the device. From a Unix standpoint,
437 * the device remains open.
439 * Note, if we were spooling, we may enter with the device blocked.
440 * We unblock at the end, only if it was us who blocked the
444 bool release_device(DCR *dcr)
447 DEVICE *dev = dcr->dev;
450 int was_blocked = BST_NOT_BLOCKED;
453 if (!dev->is_blocked()) {
454 block_device(dev, BST_RELEASING);
456 was_blocked = dev->blocked();
457 dev->set_blocked(BST_RELEASING);
460 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
462 /* if device is reserved, job never started, so release the reserve here */
463 dcr->clear_reserved();
465 if (dev->can_read()) {
466 VOLUME_CAT_INFO *vol = &dev->VolCatInfo;
467 dev->clear_read(); /* clear read bit */
468 Dmsg2(150, "dir_update_vol_info. label=%d Vol=%s\n",
469 dev->is_labeled(), vol->VolCatName);
470 if (dev->is_labeled() && vol->VolCatName[0] != 0) {
471 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
472 remove_read_volume(jcr, dcr->VolumeName);
475 } else if (dev->num_writers > 0) {
477 * Note if WEOT is set, we are at the end of the tape
478 * and may not be positioned correctly, so the
479 * job_media_record and update_vol_info have already been
480 * done, which means we skip them here.
483 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
484 if (dev->is_labeled()) {
485 Dmsg2(200, "dir_create_jobmedia. Release vol=%s dev=%s\n",
486 dev->getVolCatName(), dev->print_name());
487 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
488 Jmsg2(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
489 dcr->getVolCatName(), jcr->Job);
491 /* If no more writers, and no errors, and wrote something, write an EOF */
492 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
494 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
496 if (!dev->at_weot()) {
497 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
498 /* Note! do volume update before close, which zaps VolCatInfo */
499 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
500 Dmsg2(200, "dir_update_vol_info. Release vol=%s dev=%s\n",
501 dev->getVolCatName(), dev->print_name());
503 if (dev->num_writers == 0) { /* if not being used */
504 volume_unused(dcr); /* we obviously are not using the volume */
510 * If we reach here, it is most likely because the job
511 * has failed, since the device is not in read mode and
512 * there are no writers. It was probably reserved.
516 Dmsg3(100, "%d writers, %d reserve, dev=%s\n", dev->num_writers, dev->num_reserved(),
519 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
520 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
521 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
526 /* Fire off Alert command and include any output */
527 if (!job_canceled(jcr) && dcr->device->alert_command) {
531 char line[MAXSTRING];
532 alert = get_pool_memory(PM_FNAME);
533 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
534 /* Wait maximum 5 minutes */
535 bpipe = open_bpipe(alert, 60 * 5, "r");
537 while (fgets(line, sizeof(line), bpipe->rfd)) {
538 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
540 status = close_bpipe(bpipe);
546 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
547 alert, be.bstrerror(status));
550 Dmsg1(400, "alert status=%d\n", status);
551 free_pool_memory(alert);
553 pthread_cond_broadcast(&dev->wait_next_vol);
554 Dmsg2(100, "JobId=%u broadcast wait_device_release at %s\n",
555 (uint32_t)jcr->JobId, bstrftimes(tbuf, sizeof(tbuf), (utime_t)time(NULL)));
556 pthread_cond_broadcast(&wait_device_release);
560 * If we are the thread that blocked the device, then unblock it
562 if (pthread_equal(dev->no_wait_id, pthread_self())) {
565 /* Otherwise, reset the prior block status and unlock */
566 dev->set_blocked(was_blocked);
571 detach_dcr_from_dev(dcr);
575 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
576 (uint32_t)jcr->JobId);
581 * Clean up the device for reuse without freeing the memory
583 bool clean_device(DCR *dcr)
586 dcr->keep_dcr = true; /* do not free the dcr */
587 ok = release_device(dcr);
588 dcr->keep_dcr = false;
593 * Create a new Device Control Record and attach
594 * it to the device (if this is a real job).
595 * Note, this has been updated so that it can be called first
596 * without a DEVICE, then a second or third time with a DEVICE,
597 * and each time, it should cleanup and point to the new device.
598 * This should facilitate switching devices.
599 * Note, each dcr must point to the controlling job (jcr). However,
600 * a job can have multiple dcrs, so we must not store in the jcr's
601 * structure as previously. The higher level routine must store
602 * this dcr in the right place
605 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
609 dcr = (DCR *)malloc(sizeof(DCR));
610 memset(dcr, 0, sizeof(DCR));
611 dcr->tid = pthread_self();
613 if ((errstat = pthread_mutex_init(&dcr->m_mutex, NULL)) != 0) {
615 dev->dev_errno = errstat;
616 Mmsg1(dev->errmsg, _("Unable to init mutex: ERR=%s\n"), be.bstrerror(errstat));
617 Jmsg0(jcr, M_ERROR_TERM, 0, dev->errmsg);
620 dcr->jcr = jcr; /* point back to jcr */
621 /* Set device information, possibly change device */
624 free_block(dcr->block);
626 dcr->block = new_block(dev);
628 free_record(dcr->rec);
630 dcr->rec = new_record();
631 if (dcr->attached_to_dev) {
632 detach_dcr_from_dev(dcr);
634 /* Use job spoolsize prior to device spoolsize */
635 if (jcr->spool_size) {
636 dcr->max_job_spool_size = jcr->spool_size;
638 dcr->max_job_spool_size = dev->device->max_job_spool_size;
640 dcr->device = dev->device;
642 attach_dcr_to_dev(dcr);
648 * Search the dcrs list for the given dcr. If it is found,
649 * as it should be, then remove it. Also zap the jcr pointer
650 * to the dcr if it is the same one.
652 * Note, this code will be turned on when we can write to multiple
653 * dcrs at the same time.
656 static void remove_dcr_from_dcrs(DCR *dcr)
662 int num = jcr->dcrs->size();
663 for (i=0; i < num; i++) {
664 ldcr = (DCR *)jcr->dcrs->get(i);
666 jcr->dcrs->remove(i);
667 if (jcr->dcr == dcr) {
676 static void attach_dcr_to_dev(DCR *dcr)
684 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
685 /* ***FIXME*** return error if dev not initiated */
686 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->getJobType() != JT_SYSTEM) {
688 dev->attached_dcrs->append(dcr); /* attach dcr to device */
690 dcr->attached_to_dev = true;
691 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
697 * DCR is locked before calling this routine
699 static void locked_detach_dcr_from_dev(DCR *dcr)
701 DEVICE *dev = dcr->dev;
702 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
704 /* Detach this dcr only if attached */
705 if (dcr->attached_to_dev && dev) {
706 dcr->unreserve_device();
708 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
709 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
712 dcr->attached_to_dev = false;
716 static void detach_dcr_from_dev(DCR *dcr)
719 locked_detach_dcr_from_dev(dcr);
724 * Free up all aspects of the given dcr -- i.e. dechain it,
725 * release allocated memory, zap pointers, ...
727 void free_dcr(DCR *dcr)
733 locked_detach_dcr_from_dev(dcr);
736 free_block(dcr->block);
739 free_record(dcr->rec);
741 if (jcr && jcr->dcr == dcr) {
744 if (jcr && jcr->read_dcr == dcr) {
745 jcr->read_dcr = NULL;
748 pthread_mutex_destroy(&dcr->m_mutex);
752 static void set_dcr_from_vol(DCR *dcr, VOL_LIST *vol)
755 * Note, if we want to be able to work from a .bsr file only
756 * for disaster recovery, we must "simulate" reading the catalog
758 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
759 dcr->setVolCatName(vol->VolumeName);
760 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
761 dcr->VolCatInfo.Slot = vol->Slot;
762 dcr->VolCatInfo.InChanger = vol->Slot > 0;