2 Bacula® - The Network Backup Solution
4 Copyright (C) 2002-2008 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 John Walker.
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
36 #include "bacula.h" /* pull in global headers */
37 #include "stored.h" /* pull in Storage Deamon headers */
39 /* Forward referenced functions */
40 static void attach_dcr_to_dev(DCR *dcr);
41 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev);
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 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
67 dev->dblock(BST_DOING_ACQUIRE);
69 if (dev->num_writers > 0) {
70 Jmsg2(jcr, M_FATAL, 0, _("Acquire read: num_writers=%d not zero. Job %d canceled.\n"),
71 dev->num_writers, jcr->JobId);
75 /* Find next Volume, if any */
79 Jmsg(jcr, M_FATAL, 0, _("No volumes specified for reading. Job %s canceled.\n"),
80 edit_int64(jcr->JobId, ed1));
84 for (i=1; i<jcr->CurReadVolume; i++) {
88 Jmsg(jcr, M_FATAL, 0, _("Logic error: no next volume to read. Numvol=%d Curvol=%d\n"),
89 jcr->NumReadVolumes, jcr->CurReadVolume);
90 goto get_out; /* should not happen */
93 * Note, if we want to be able to work from a .bsr file only
94 * for disaster recovery, we must "simulate" reading the catalog
96 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
97 bstrncpy(dcr->VolCatInfo.VolCatName, vol->VolumeName, sizeof(dcr->VolCatInfo.VolCatName));
98 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
99 dcr->VolCatInfo.Slot = vol->Slot;
100 dcr->VolCatInfo.InChanger = vol->Slot > 0;
101 if (reserve_volume(dcr, dcr->VolumeName) == NULL) {
102 Dmsg2(100, "Could not reserve volume %s on %s\n", dcr->VolumeName,
103 dcr->dev->print_name());
104 Jmsg2(jcr, M_FATAL, 0, _("Could not reserve volume %s on %s\n"), dcr->VolumeName,
105 dcr->dev->print_name());
108 if (dev->vol && dev->vol->is_swapping()) {
109 dev->vol->set_slot(vol->Slot);
110 Dmsg3(100, "swapping: slot=%d Vol=%s dev=%s\n", dev->vol->get_slot(),
111 dev->vol->vol_name, dev->print_name());
115 * If the MediaType requested for this volume is not the
116 * same as the current drive, we attempt to find the same
117 * device that was used to write the orginal volume. If
118 * found, we switch to using that device.
120 * N.B. A lot of routines rely on the dcr pointer not changing
121 * read_records.c even has multiple dcrs cached, so we take care
122 * here to release all important parts of the dcr and re-acquire
123 * them such as the block pointer (size may change), but we do
124 * not release the dcr.
126 Dmsg2(50, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
127 if (dcr->media_type[0] && strcmp(dcr->media_type, dev->device->media_type) != 0) {
132 Jmsg3(jcr, M_INFO, 0, _("Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
134 dcr->media_type, dev->device->media_type, dev->print_name());
135 Dmsg3(50, "Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
137 dcr->media_type, dev->device->media_type, dev->print_name());
139 dev->dunblock(DEV_UNLOCKED);
142 memset(&rctx, 0, sizeof(RCTX));
144 jcr->reserve_msgs = New(alist(10, not_owned_by_alist));
145 rctx.any_drive = true;
146 rctx.device_name = vol->device;
147 store = new DIRSTORE;
148 memset(store, 0, sizeof(DIRSTORE));
149 store->name[0] = 0; /* No dir name */
150 bstrncpy(store->media_type, vol->MediaType, sizeof(store->media_type));
151 bstrncpy(store->pool_name, dcr->pool_name, sizeof(store->pool_name));
152 bstrncpy(store->pool_type, dcr->pool_type, sizeof(store->pool_type));
153 store->append = false;
155 clean_device(dcr); /* clean up the dcr */
158 * Search for a new device
160 stat = search_res_for_device(rctx);
161 release_reserve_messages(jcr); /* release queued messages */
162 unlock_reservations();
165 dev = dcr->dev; /* get new device pointer */
166 dev->dblock(BST_DOING_ACQUIRE);
167 dcr->VolumeName[0] = 0;
168 Jmsg(jcr, M_INFO, 0, _("Media Type change. New device %s chosen.\n"),
170 Dmsg1(50, "Media Type change. New device %s chosen.\n", dev->print_name());
172 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
173 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
174 dcr->VolCatInfo.Slot = vol->Slot;
175 bstrncpy(dcr->pool_name, store->pool_name, sizeof(dcr->pool_name));
176 bstrncpy(dcr->pool_type, store->pool_type, sizeof(dcr->pool_type));
179 Jmsg1(jcr, M_FATAL, 0, _("No suitable device found to read Volume \"%s\"\n"),
181 Dmsg1(50, "No suitable device found to read Volume \"%s\"\n", vol->VolumeName);
187 init_device_wait_timers(dcr);
189 tape_previously_mounted = dev->can_read() || dev->can_append() ||
191 tape_initially_mounted = tape_previously_mounted;
194 /* Volume info is always needed because of VolParts */
195 Dmsg0(200, "dir_get_volume_info\n");
196 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_READ)) {
197 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
201 /* If not polling limit retries */
202 if (!dev->poll && retry++ > 10) {
205 dev->clear_labeled(); /* force reread of label */
206 if (job_canceled(jcr)) {
208 Mmsg1(dev->errmsg, _("Job %s canceled.\n"), edit_int64(jcr->JobId, ed1));
209 Jmsg(jcr, M_INFO, 0, dev->errmsg);
210 goto get_out; /* error return */
213 dcr->do_swapping(false/*is_writing*/);
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) {
222 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
223 dev->print_name(), dcr->VolumeName, dev->bstrerror());
226 Dmsg1(50, "opened dev %s OK\n", dev->print_name());
228 /* Read Volume Label */
229 Dmsg0(50, "calling read-vol-label\n");
230 vol_label_status = read_dev_volume_label(dcr);
231 switch (vol_label_status) {
234 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
238 * Send error message generated by read_dev_volume_label()
239 * only we really had a tape mounted. This supresses superfluous
240 * error messages when nothing is mounted.
242 if (tape_previously_mounted) {
243 Jmsg(jcr, M_WARNING, 0, "%s", jcr->errmsg);
247 if (tape_initially_mounted) {
248 tape_initially_mounted = false;
251 /* If polling and got a previous bad name, ignore it */
252 if (dev->poll && strcmp(dev->BadVolName, dev->VolHdr.VolumeName) == 0) {
255 bstrncpy(dev->BadVolName, dev->VolHdr.VolumeName, sizeof(dev->BadVolName));
257 if (!unload_autochanger(dcr, -1)) {
258 /* at least free the device so we can re-open with correct volume */
263 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
265 tape_previously_mounted = true;
268 * If the device requires mount, close it, so the device can be ejected.
270 if (dev->requires_mount()) {
274 /* Call autochanger only once unless ask_sysop called */
275 if (try_autochanger) {
277 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
278 dcr->VolumeName, dcr->VolCatInfo.Slot);
279 stat = autoload_device(dcr, 0, NULL);
281 try_autochanger = false;
282 continue; /* try reading volume mounted */
284 /* Try closing and re-opening */
287 if (dev->open(dcr, OPEN_READ_ONLY) >= 0) {
292 /* Mount a specific volume and no other */
293 Dmsg0(200, "calling dir_ask_sysop\n");
294 if (!dir_ask_sysop_to_mount_volume(dcr, ST_READ)) {
295 goto get_out; /* error return */
297 try_autochanger = true; /* permit using autochanger again */
298 continue; /* try reading again */
303 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
310 set_jcr_job_status(jcr, JS_Running);
311 dir_send_job_status(jcr);
312 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
313 dcr->VolumeName, dev->print_name());
317 dcr->clear_reserved();
319 * Normally we are blocked, but in at least one error case above
320 * we are not blocked because we unsuccessfully tried changing
323 if (dev->is_blocked()) {
324 dev->dunblock(DEV_LOCKED);
326 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
332 * Acquire device for writing. We permit multiple writers.
333 * If this is the first one, we read the label.
335 * Returns: NULL if failed for any reason
337 * Note, normally reserve_device_for_append() is called
338 * before this routine.
340 DCR *acquire_device_for_append(DCR *dcr)
342 DEVICE *dev = dcr->dev;
346 init_device_wait_timers(dcr);
348 dev->dblock(BST_DOING_ACQUIRE);
349 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
350 (dev->is_dvd()?"DVD":"disk"));
353 * With the reservation system, this should not happen
355 if (dev->can_read()) {
356 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
357 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
362 * have_vol defines whether or not mount_next_write_volume should
363 * ask the Director again about what Volume to use.
365 if (dev->can_append() && dcr->is_suitable_volume_mounted() &&
366 strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") != 0) {
367 Dmsg0(190, "device already in append.\n");
369 * At this point, the correct tape is already mounted, so
370 * we do not need to do mount_next_write_volume(), unless
371 * we need to recycle the tape.
373 if (dev->num_writers == 0) {
374 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
376 if (!is_tape_position_ok(jcr, dev)) {
380 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
381 if (!dcr->mount_next_write_volume()) {
382 if (!job_canceled(jcr)) {
383 /* Reduce "noise" -- don't print if job canceled */
384 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
386 Dmsg1(200, "Could not ready device %s for append.\n",
391 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
394 dev->num_writers++; /* we are now a writer */
395 if (jcr->NumWriteVolumes == 0) {
396 jcr->NumWriteVolumes = 1;
398 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
399 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
404 dcr->clear_reserved();
405 dev->dunblock(DEV_LOCKED);
406 return ok ? dcr : NULL;
412 * Check to see if the tape position as defined by the OS is
413 * the same as our concept. If it is not, we bail out, because
414 * it means the user has probably manually rewound the tape.
415 * Note, we check only if num_writers == 0, but this code will
416 * also work fine for any number of writers. If num_writers > 0,
417 * we probably should cancel all jobs using this device, or
418 * perhaps even abort the SD, or at a minimum, mark the tape
419 * in error. Another strategy with num_writers == 0, would be
420 * to rewind the tape and do a new eod() request.
422 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev)
424 if (dev->is_tape() && dev->num_writers == 0) {
425 int32_t file = dev->get_os_tape_file();
426 if (file >= 0 && file != (int32_t)dev->get_file()) {
427 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
428 " on device %s. Expected %d, got %d\n"),
429 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
438 * This job is done, so release the device. From a Unix standpoint,
439 * the device remains open.
441 * Note, if we are spooling, we may enter with the device locked.
442 * However, in all cases, unlock the device when leaving.
445 bool release_device(DCR *dcr)
448 DEVICE *dev = dcr->dev;
451 /* lock only if not already locked by this thread */
452 if (!dcr->is_dev_locked()) {
456 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
458 /* if device is reserved, job never started, so release the reserve here */
459 dcr->clear_reserved();
461 if (dev->can_read()) {
462 dev->clear_read(); /* clear read bit */
463 Dmsg0(100, "dir_update_vol_info. Release0\n");
464 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
467 } else if (dev->num_writers > 0) {
469 * Note if WEOT is set, we are at the end of the tape
470 * and may not be positioned correctly, so the
471 * job_media_record and update_vol_info have already been
472 * done, which means we skip them here.
475 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
476 if (dev->is_labeled()) {
477 Dmsg2(200, "dir_create_jobmedia. Release vol=%s dev=%s\n",
478 dev->VolCatInfo.VolCatName, dev->print_name());
479 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
480 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
481 dcr->VolCatInfo.VolCatName, jcr->Job);
483 /* If no more writers, and no errors, and wrote something, write an EOF */
484 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
486 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
488 if (!dev->at_weot()) {
489 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
490 /* Note! do volume update before close, which zaps VolCatInfo */
491 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
492 Dmsg2(200, "dir_update_vol_info. Release vol=%s dev=%s\n",
493 dev->VolCatInfo.VolCatName, dev->print_name());
495 if (!dev->is_busy()) { /* if not being used */
496 volume_unused(dcr); /* we obviously are not using the volume */
502 * If we reach here, it is most likely because the job
503 * has failed, since the device is not in read mode and
504 * there are no writers. It was probably reserved.
510 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
511 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
512 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
516 /* Fire off Alert command and include any output */
517 if (!job_canceled(jcr) && dcr->device->alert_command) {
521 char line[MAXSTRING];
522 alert = get_pool_memory(PM_FNAME);
523 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
524 bpipe = open_bpipe(alert, 0, "r");
526 while (fgets(line, sizeof(line), bpipe->rfd)) {
527 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
529 status = close_bpipe(bpipe);
535 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
536 alert, be.bstrerror(status));
539 Dmsg1(400, "alert status=%d\n", status);
540 free_pool_memory(alert);
542 pthread_cond_broadcast(&dev->wait_next_vol);
543 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
544 pthread_cond_broadcast(&wait_device_release);
547 detach_dcr_from_dev(dcr);
549 if (jcr->read_dcr == dcr) {
550 jcr->read_dcr = NULL;
552 if (jcr->dcr == dcr) {
557 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
558 (uint32_t)jcr->JobId);
563 * Clean up the device for reuse without freeing the memory
565 bool clean_device(DCR *dcr)
568 dcr->keep_dcr = true; /* do not free the dcr */
569 ok = release_device(dcr);
570 dcr->keep_dcr = false;
575 * Create a new Device Control Record and attach
576 * it to the device (if this is a real job).
577 * Note, this has been updated so that it can be called first
578 * without a DEVICE, then a second or third time with a DEVICE,
579 * and each time, it should cleanup and point to the new device.
580 * This should facilitate switching devices.
581 * Note, each dcr must point to the controlling job (jcr). However,
582 * a job can have multiple dcrs, so we must not store in the jcr's
583 * structure as previously. The higher level routine must store
584 * this dcr in the right place
587 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
590 dcr = (DCR *)malloc(sizeof(DCR));
591 memset(dcr, 0, sizeof(DCR));
592 dcr->tid = pthread_self();
595 dcr->jcr = jcr; /* point back to jcr */
596 /* Set device information, possibly change device */
599 free_block(dcr->block);
601 dcr->block = new_block(dev);
603 free_record(dcr->rec);
605 dcr->rec = new_record();
606 if (dcr->attached_to_dev) {
607 detach_dcr_from_dev(dcr);
609 /* Use job spoolsize prior to device spoolsize */
610 if (jcr->spool_size) {
611 dcr->max_job_spool_size = jcr->spool_size;
613 dcr->max_job_spool_size = dev->device->max_job_spool_size;
615 dcr->device = dev->device;
617 attach_dcr_to_dev(dcr);
623 * Search the dcrs list for the given dcr. If it is found,
624 * as it should be, then remove it. Also zap the jcr pointer
625 * to the dcr if it is the same one.
627 * Note, this code will be turned on when we can write to multiple
628 * dcrs at the same time.
631 static void remove_dcr_from_dcrs(DCR *dcr)
637 int num = jcr->dcrs->size();
638 for (i=0; i < num; i++) {
639 ldcr = (DCR *)jcr->dcrs->get(i);
641 jcr->dcrs->remove(i);
642 if (jcr->dcr == dcr) {
651 static void attach_dcr_to_dev(DCR *dcr)
653 DEVICE *dev = dcr->dev;
656 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
657 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
658 dev->attached_dcrs->append(dcr); /* attach dcr to device */
659 dcr->attached_to_dev = true;
660 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
664 void detach_dcr_from_dev(DCR *dcr)
666 DEVICE *dev = dcr->dev;
667 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
669 /* Detach this dcr only if attached */
670 if (dcr->attached_to_dev && dev) {
672 dcr->unreserve_device();
673 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
674 dcr->attached_to_dev = false;
675 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
681 * Free up all aspects of the given dcr -- i.e. dechain it,
682 * release allocated memory, zap pointers, ...
684 void free_dcr(DCR *dcr)
688 detach_dcr_from_dev(dcr);
691 free_block(dcr->block);
694 free_record(dcr->rec);
696 if (jcr && jcr->dcr == dcr) {