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) {
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) {
236 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
240 * Send error message generated by read_dev_volume_label()
241 * only we really had a tape mounted. This supresses superfluous
242 * error messages when nothing is mounted.
244 if (tape_previously_mounted) {
245 Jmsg(jcr, M_WARNING, 0, "%s", jcr->errmsg);
249 if (tape_initially_mounted) {
250 tape_initially_mounted = false;
253 /* If polling and got a previous bad name, ignore it */
254 if (dev->poll && strcmp(dev->BadVolName, dev->VolHdr.VolumeName) == 0) {
257 bstrncpy(dev->BadVolName, dev->VolHdr.VolumeName, sizeof(dev->BadVolName));
259 if (!unload_autochanger(dcr, -1)) {
260 /* at least free the device so we can re-open with correct volume */
265 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
267 tape_previously_mounted = true;
270 * If the device requires mount, close it, so the device can be ejected.
272 if (dev->requires_mount()) {
276 /* Call autochanger only once unless ask_sysop called */
277 if (try_autochanger) {
279 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
280 dcr->VolumeName, dcr->VolCatInfo.Slot);
281 stat = autoload_device(dcr, 0, NULL);
283 try_autochanger = false;
284 continue; /* try reading volume mounted */
286 /* Try closing and re-opening */
289 if (dev->open(dcr, OPEN_READ_ONLY) >= 0) {
293 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
294 dev->print_name(), dcr->VolumeName, dev->bstrerror());
298 /* Mount a specific volume and no other */
299 Dmsg0(200, "calling dir_ask_sysop\n");
300 if (!dir_ask_sysop_to_mount_volume(dcr, ST_READ)) {
301 goto get_out; /* error return */
303 try_autochanger = true; /* permit using autochanger again */
304 continue; /* try reading again */
309 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
316 set_jcr_job_status(jcr, JS_Running);
317 dir_send_job_status(jcr);
318 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
319 dcr->VolumeName, dev->print_name());
323 dcr->clear_reserved();
325 * Normally we are blocked, but in at least one error case above
326 * we are not blocked because we unsuccessfully tried changing
329 if (dev->is_blocked()) {
330 dev->dunblock(DEV_LOCKED);
332 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
338 * Acquire device for writing. We permit multiple writers.
339 * If this is the first one, we read the label.
341 * Returns: NULL if failed for any reason
343 * Note, normally reserve_device_for_append() is called
344 * before this routine.
346 DCR *acquire_device_for_append(DCR *dcr)
348 DEVICE *dev = dcr->dev;
352 init_device_wait_timers(dcr);
354 dev->dblock(BST_DOING_ACQUIRE);
355 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
356 (dev->is_dvd()?"DVD":"disk"));
359 * With the reservation system, this should not happen
361 if (dev->can_read()) {
362 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
363 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
368 * have_vol defines whether or not mount_next_write_volume should
369 * ask the Director again about what Volume to use.
371 if (dev->can_append() && dcr->is_suitable_volume_mounted() &&
372 strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") != 0) {
373 Dmsg0(190, "device already in append.\n");
375 * At this point, the correct tape is already mounted, so
376 * we do not need to do mount_next_write_volume(), unless
377 * we need to recycle the tape.
379 if (dev->num_writers == 0) {
380 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
382 if (!is_tape_position_ok(jcr, dev)) {
386 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
387 if (!dcr->mount_next_write_volume()) {
388 if (!job_canceled(jcr)) {
389 /* Reduce "noise" -- don't print if job canceled */
390 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
392 Dmsg1(200, "Could not ready device %s for append.\n",
397 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
400 dev->num_writers++; /* we are now a writer */
401 if (jcr->NumWriteVolumes == 0) {
402 jcr->NumWriteVolumes = 1;
404 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
405 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
410 dcr->clear_reserved();
411 dev->dunblock(DEV_LOCKED);
412 return ok ? dcr : NULL;
418 * Check to see if the tape position as defined by the OS is
419 * the same as our concept. If it is not, we bail out, because
420 * it means the user has probably manually rewound the tape.
421 * Note, we check only if num_writers == 0, but this code will
422 * also work fine for any number of writers. If num_writers > 0,
423 * we probably should cancel all jobs using this device, or
424 * perhaps even abort the SD, or at a minimum, mark the tape
425 * in error. Another strategy with num_writers == 0, would be
426 * to rewind the tape and do a new eod() request.
428 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev)
430 if (dev->is_tape() && dev->num_writers == 0) {
431 int32_t file = dev->get_os_tape_file();
432 if (file >= 0 && file != (int32_t)dev->get_file()) {
433 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
434 " on device %s. Expected %d, got %d\n"),
435 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
444 * This job is done, so release the device. From a Unix standpoint,
445 * the device remains open.
447 * Note, if we are spooling, we may enter with the device locked.
448 * However, in all cases, unlock the device when leaving.
451 bool release_device(DCR *dcr)
454 DEVICE *dev = dcr->dev;
458 /* lock only if not already locked by this thread */
459 if (!dcr->is_dev_locked()) {
463 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
465 /* if device is reserved, job never started, so release the reserve here */
466 dcr->clear_reserved();
468 if (dev->can_read()) {
469 dev->clear_read(); /* clear read bit */
470 Dmsg0(100, "dir_update_vol_info. Release0\n");
471 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
474 } else if (dev->num_writers > 0) {
476 * Note if WEOT is set, we are at the end of the tape
477 * and may not be positioned correctly, so the
478 * job_media_record and update_vol_info have already been
479 * done, which means we skip them here.
482 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
483 if (dev->is_labeled()) {
484 Dmsg2(200, "dir_create_jobmedia. Release vol=%s dev=%s\n",
485 dev->VolCatInfo.VolCatName, dev->print_name());
486 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
487 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
488 dcr->VolCatInfo.VolCatName, jcr->Job);
490 /* If no more writers, and no errors, and wrote something, write an EOF */
491 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
493 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
495 if (!dev->at_weot()) {
496 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
497 /* Note! do volume update before close, which zaps VolCatInfo */
498 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
499 Dmsg2(200, "dir_update_vol_info. Release vol=%s dev=%s\n",
500 dev->VolCatInfo.VolCatName, dev->print_name());
502 if (dev->num_writers == 0) { /* if not being used */
503 // if (!dev->is_busy()) { /* 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.
517 Dmsg3(100, "%d writers, %d reserve, dev=%s\n", dev->num_writers, dev->num_reserved(),
519 debug_list_volumes("acquire:release_device()");
522 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
523 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
524 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
528 /* Fire off Alert command and include any output */
529 if (!job_canceled(jcr) && dcr->device->alert_command) {
533 char line[MAXSTRING];
534 alert = get_pool_memory(PM_FNAME);
535 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
536 bpipe = open_bpipe(alert, 0, "r");
538 while (fgets(line, sizeof(line), bpipe->rfd)) {
539 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
541 status = close_bpipe(bpipe);
547 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
548 alert, be.bstrerror(status));
551 Dmsg1(400, "alert status=%d\n", status);
552 free_pool_memory(alert);
554 pthread_cond_broadcast(&dev->wait_next_vol);
555 Dmsg2(100, "JobId=%u broadcast wait_device_release at %s\n",
556 (uint32_t)jcr->JobId, bstrftimes(tbuf, sizeof(tbuf), (utime_t)time(NULL)));
557 pthread_cond_broadcast(&wait_device_release);
560 detach_dcr_from_dev(dcr);
562 if (jcr->read_dcr == dcr) {
563 jcr->read_dcr = NULL;
565 if (jcr->dcr == dcr) {
570 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
571 (uint32_t)jcr->JobId);
576 * Clean up the device for reuse without freeing the memory
578 bool clean_device(DCR *dcr)
581 dcr->keep_dcr = true; /* do not free the dcr */
582 ok = release_device(dcr);
583 dcr->keep_dcr = false;
588 * Create a new Device Control Record and attach
589 * it to the device (if this is a real job).
590 * Note, this has been updated so that it can be called first
591 * without a DEVICE, then a second or third time with a DEVICE,
592 * and each time, it should cleanup and point to the new device.
593 * This should facilitate switching devices.
594 * Note, each dcr must point to the controlling job (jcr). However,
595 * a job can have multiple dcrs, so we must not store in the jcr's
596 * structure as previously. The higher level routine must store
597 * this dcr in the right place
600 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
603 dcr = (DCR *)malloc(sizeof(DCR));
604 memset(dcr, 0, sizeof(DCR));
605 dcr->tid = pthread_self();
608 dcr->jcr = jcr; /* point back to jcr */
609 /* Set device information, possibly change device */
612 free_block(dcr->block);
614 dcr->block = new_block(dev);
616 free_record(dcr->rec);
618 dcr->rec = new_record();
619 if (dcr->attached_to_dev) {
620 detach_dcr_from_dev(dcr);
622 /* Use job spoolsize prior to device spoolsize */
623 if (jcr->spool_size) {
624 dcr->max_job_spool_size = jcr->spool_size;
626 dcr->max_job_spool_size = dev->device->max_job_spool_size;
628 dcr->device = dev->device;
630 attach_dcr_to_dev(dcr);
636 * Search the dcrs list for the given dcr. If it is found,
637 * as it should be, then remove it. Also zap the jcr pointer
638 * to the dcr if it is the same one.
640 * Note, this code will be turned on when we can write to multiple
641 * dcrs at the same time.
644 static void remove_dcr_from_dcrs(DCR *dcr)
650 int num = jcr->dcrs->size();
651 for (i=0; i < num; i++) {
652 ldcr = (DCR *)jcr->dcrs->get(i);
654 jcr->dcrs->remove(i);
655 if (jcr->dcr == dcr) {
664 static void attach_dcr_to_dev(DCR *dcr)
666 DEVICE *dev = dcr->dev;
669 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
670 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
671 dev->attached_dcrs->append(dcr); /* attach dcr to device */
672 dcr->attached_to_dev = true;
673 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
677 void detach_dcr_from_dev(DCR *dcr)
679 DEVICE *dev = dcr->dev;
680 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
682 /* Detach this dcr only if attached */
683 if (dcr->attached_to_dev && dev) {
685 dcr->unreserve_device();
686 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
687 dcr->attached_to_dev = false;
688 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
694 * Free up all aspects of the given dcr -- i.e. dechain it,
695 * release allocated memory, zap pointers, ...
697 void free_dcr(DCR *dcr)
701 detach_dcr_from_dev(dcr);
704 free_block(dcr->block);
707 free_record(dcr->rec);
709 if (jcr && jcr->dcr == dcr) {