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_suitable_volume_mounted(DCR *dcr);
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));
83 volume_unused(dcr); /* release any current volume */
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 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
94 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
95 dcr->VolCatInfo.Slot = 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 device. Want Media Type=\"%s\" have=\"%s\"\n"
117 dcr->media_type, dev->device->media_type, dev->print_name());
118 Dmsg3(50, "Changing 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));
127 jcr->reserve_msgs = New(alist(10, not_owned_by_alist));
128 rctx.any_drive = true;
129 rctx.device_name = vol->device;
130 store = new DIRSTORE;
131 memset(store, 0, sizeof(DIRSTORE));
132 store->name[0] = 0; /* No dir name */
133 bstrncpy(store->media_type, vol->MediaType, sizeof(store->media_type));
134 bstrncpy(store->pool_name, dcr->pool_name, sizeof(store->pool_name));
135 bstrncpy(store->pool_type, dcr->pool_type, sizeof(store->pool_type));
136 store->append = false;
138 clean_device(dcr); /* clean up the dcr */
141 * Search for a new device
143 stat = search_res_for_device(rctx);
144 release_reserve_messages(jcr); /* release queued messages */
145 unlock_reservations();
148 dev = dcr->dev; /* get new device pointer */
149 dev->dblock(BST_DOING_ACQUIRE);
150 dcr->VolumeName[0] = 0;
151 Jmsg(jcr, M_INFO, 0, _("Media Type change. New device %s chosen.\n"),
153 Dmsg1(50, "Media Type change. New device %s chosen.\n", dev->print_name());
155 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
156 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
157 dcr->VolCatInfo.Slot = vol->Slot;
158 bstrncpy(dcr->pool_name, store->pool_name, sizeof(dcr->pool_name));
159 bstrncpy(dcr->pool_type, store->pool_type, sizeof(dcr->pool_type));
162 Jmsg1(jcr, M_FATAL, 0, _("No suitable device found to read Volume \"%s\"\n"),
164 Dmsg1(50, "No suitable device found to read Volume \"%s\"\n", vol->VolumeName);
170 init_device_wait_timers(dcr);
172 tape_previously_mounted = dev->can_read() || dev->can_append() ||
174 tape_initially_mounted = tape_previously_mounted;
177 /* Volume info is always needed because of VolParts */
178 Dmsg0(200, "dir_get_volume_info\n");
179 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_READ)) {
180 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
184 /* If not polling limit retries */
185 if (!dev->poll && retry++ > 10) {
188 dev->clear_labeled(); /* force reread of label */
189 if (job_canceled(jcr)) {
191 Mmsg1(dev->errmsg, _("Job %s canceled.\n"), edit_int64(jcr->JobId, ed1));
192 Jmsg(jcr, M_INFO, 0, dev->errmsg);
193 goto get_out; /* error return */
196 autoload_device(dcr, 0, NULL);
199 * This code ensures that the device is ready for
200 * reading. If it is a file, it opens it.
201 * If it is a tape, it checks the volume name
203 Dmsg1(100, "bstored: open vol=%s\n", dcr->VolumeName);
204 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
205 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
206 dev->print_name(), dcr->VolumeName, dev->bstrerror());
209 Dmsg1(50, "opened dev %s OK\n", dev->print_name());
211 /* Read Volume Label */
212 Dmsg0(50, "calling read-vol-label\n");
213 vol_label_status = read_dev_volume_label(dcr);
214 switch (vol_label_status) {
217 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
221 * Send error message generated by read_dev_volume_label()
222 * only we really had a tape mounted. This supresses superfluous
223 * error messages when nothing is mounted.
225 if (tape_previously_mounted) {
226 Jmsg(jcr, M_WARNING, 0, "%s", jcr->errmsg);
230 if (tape_initially_mounted) {
231 tape_initially_mounted = false;
234 /* If polling and got a previous bad name, ignore it */
235 if (dev->poll && strcmp(dev->BadVolName, dev->VolHdr.VolumeName) == 0) {
238 bstrncpy(dev->BadVolName, dev->VolHdr.VolumeName, sizeof(dev->BadVolName));
242 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
244 tape_previously_mounted = true;
247 * If the device requires mount, close it, so the device can be ejected.
249 if (dev->requires_mount()) {
253 /* Call autochanger only once unless ask_sysop called */
254 if (try_autochanger) {
256 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
257 dcr->VolumeName, dcr->VolCatInfo.Slot);
258 stat = autoload_device(dcr, 0, NULL);
260 try_autochanger = false;
261 continue; /* try reading volume mounted */
265 /* Mount a specific volume and no other */
266 Dmsg0(200, "calling dir_ask_sysop\n");
267 if (!dir_ask_sysop_to_mount_volume(dcr)) {
268 goto get_out; /* error return */
270 try_autochanger = true; /* permit using autochanger again */
271 continue; /* try reading again */
276 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
283 set_jcr_job_status(jcr, JS_Running);
284 dir_send_job_status(jcr);
285 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
286 dcr->VolumeName, dev->print_name());
290 if (dcr && dcr->reserved_device) {
291 dev->reserved_device--;
292 Dmsg2(50, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
293 dcr->reserved_device = false;
296 * Normally we are blocked, but in at least one error case above
297 * we are not blocked because we unsuccessfully tried changing
300 if (dev->is_blocked()) {
301 dev->dunblock(DEV_LOCKED);
303 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
309 * Acquire device for writing. We permit multiple writers.
310 * If this is the first one, we read the label.
312 * Returns: NULL if failed for any reason
314 * Note, normally reserve_device_for_append() is called
315 * before this routine.
317 DCR *acquire_device_for_append(DCR *dcr)
319 bool do_mount = false;
320 bool release = false;
322 DEVICE *dev = dcr->dev;
325 init_device_wait_timers(dcr);
327 dev->dblock(BST_DOING_ACQUIRE);
328 Dmsg1(190, "acquire_append device is %s\n", dev->is_tape()?"tape":
329 (dev->is_dvd()?"DVD":"disk"));
332 * With the reservation system, this should not happen
334 if (dev->can_read()) {
335 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
336 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
341 * have_vol defines whether or not mount_next_write_volume should
342 * ask the Director again about what Volume to use.
344 have_vol = is_suitable_volume_mounted(dcr);
345 if (dev->can_append()) {
346 Dmsg0(190, "device already in append.\n");
348 * Device already in append mode
350 * Check if we have the right Volume mounted
351 * OK if current volume info OK
352 * OK if next volume matches current volume
353 * otherwise mount desired volume obtained from
354 * dir_find_next_appendable_volume
355 * dev->VolHdr.VolumeName is what is in the drive
356 * dcr->VolumeName is what we pass into the routines, or
357 * get back from the subroutines.
360 !(dir_find_next_appendable_volume(dcr) &&
361 strcmp(dev->VolHdr.VolumeName, dcr->VolumeName) == 0)) { /* wrong tape mounted */
362 /* Wrong tape mounted, release it, then fall through to get correct one */
363 Dmsg0(50, "Wrong tape mounted, release and try mount.\n");
368 * At this point, the correct tape is already mounted, so
369 * we do not need to do mount_next_write_volume(), unless
370 * we need to recycle the tape.
372 do_mount = strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") == 0;
373 Dmsg2(190, "jid=%u Correct tape mounted. recycle=%d\n",
374 (uint32_t)jcr->JobId, do_mount);
376 if (do_mount && dev->num_writers != 0) {
377 Jmsg(jcr, M_FATAL, 0, _("Cannot recycle volume \"%s\""
378 " on device %s because it is in use by another job.\n"),
379 dev->VolHdr.VolumeName, dev->print_name());
383 if (dev->num_writers == 0) {
384 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
390 * Check to see if the tape position as defined by the OS is
391 * the same as our concept. If it is not, we bail out, because
392 * it means the user has probably manually rewound the tape.
393 * Note, we check only if num_writers == 0, but this code will
394 * also work fine for any number of writers. If num_writers > 0,
395 * we probably should cancel all jobs using this device, or
396 * perhaps even abort the SD, or at a minimum, mark the tape
397 * in error. Another strategy with num_writers == 0, would be
398 * to rewind the tape and do a new eod() request.
400 if (dev->is_tape() && dev->num_writers == 0) {
401 int32_t file = dev->get_os_tape_file();
402 if (file >= 0 && file != (int32_t)dev->get_file()) {
403 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
404 " on device %s. Expected %d, got %d\n"),
405 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
411 /* Not already in append mode, so mount the device */
412 Dmsg2(190, "jid=%u Not in append mode, try mount have_vol=%d\n",
413 (uint32_t)jcr->JobId, have_vol);
415 ASSERT(dev->num_writers == 0);
419 if (do_mount || !have_vol) {
420 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
421 bool mounted = mount_next_write_volume(dcr, have_vol, release);
423 if (!job_canceled(jcr)) {
424 /* Reduce "noise" -- don't print if job canceled */
425 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
427 Dmsg2(200, "jid=%u Could not ready device %s for append.\n",
428 (uint32_t)jcr->JobId, dev->print_name());
432 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
435 dev->num_writers++; /* we are now a writer */
436 if (jcr->NumWriteVolumes == 0) {
437 jcr->NumWriteVolumes = 1;
439 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
440 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
442 if (dcr->reserved_device) {
443 dev->reserved_device--;
444 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
445 dev->reserved_device, dev->print_name());
446 dcr->reserved_device = false;
448 dev->dunblock(DEV_LOCKED);
456 if (dcr->reserved_device) {
457 dev->reserved_device--;
458 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
459 dev->reserved_device, dev->print_name());
460 dcr->reserved_device = false;
462 dev->dunblock(DEV_LOCKED);
467 static bool is_suitable_volume_mounted(DCR *dcr)
469 DEVICE *dev = dcr->dev;
471 /* Volume mounted? */
472 if (dev->VolHdr.VolumeName[0] == 0) {
473 return false; /* no */
475 bstrncpy(dcr->VolumeName, dev->VolHdr.VolumeName, sizeof(dcr->VolumeName));
476 return dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE);
480 * This job is done, so release the device. From a Unix standpoint,
481 * the device remains open.
483 * Note, if we are spooling, we may enter with the device locked.
484 * However, in all cases, unlock the device when leaving.
487 bool release_device(DCR *dcr)
490 DEVICE *dev = dcr->dev;
493 /* lock only if not already locked by this thread */
494 if (!dcr->is_dev_locked()) {
497 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
499 /* if device is reserved, job never started, so release the reserve here */
500 if (dcr->reserved_device) {
501 dev->reserved_device--;
502 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
503 dcr->reserved_device = false;
506 if (dev->can_read()) {
507 dev->clear_read(); /* clear read bit */
508 Dmsg0(100, "dir_update_vol_info. Release0\n");
509 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
511 } else if (dev->num_writers > 0) {
513 * Note if WEOT is set, we are at the end of the tape
514 * and may not be positioned correctly, so the
515 * job_media_record and update_vol_info have already been
516 * done, which means we skip them here.
519 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
520 if (dev->is_labeled()) {
521 Dmsg0(100, "dir_create_jobmedia_record. Release\n");
522 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
523 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
524 dcr->VolCatInfo.VolCatName, jcr->Job);
526 /* If no more writers, write an EOF */
527 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
529 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
531 if (!dev->at_weot()) {
532 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
533 /* Note! do volume update before close, which zaps VolCatInfo */
534 Dmsg0(100, "dir_update_vol_info. Release0\n");
535 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
541 * If we reach here, it is most likely because the job
542 * has failed, since the device is not in read mode and
543 * there are no writers. It was probably reserved.
547 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
548 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
549 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
553 /* Fire off Alert command and include any output */
554 if (!job_canceled(jcr) && dcr->device->alert_command) {
558 char line[MAXSTRING];
559 alert = get_pool_memory(PM_FNAME);
560 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
561 bpipe = open_bpipe(alert, 0, "r");
563 while (fgets(line, sizeof(line), bpipe->rfd)) {
564 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
566 status = close_bpipe(bpipe);
572 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
573 alert, be.bstrerror(status));
576 Dmsg1(400, "alert status=%d\n", status);
577 free_pool_memory(alert);
579 pthread_cond_broadcast(&dev->wait_next_vol);
580 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
581 pthread_cond_broadcast(&wait_device_release);
584 detach_dcr_from_dev(dcr);
586 if (jcr->read_dcr == dcr) {
587 jcr->read_dcr = NULL;
589 if (jcr->dcr == dcr) {
594 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
595 (uint32_t)jcr->JobId);
600 * Clean up the device for reuse without freeing the memory
602 bool clean_device(DCR *dcr)
605 dcr->keep_dcr = true; /* do not free the dcr */
606 ok = release_device(dcr);
607 dcr->keep_dcr = false;
612 * Create a new Device Control Record and attach
613 * it to the device (if this is a real job).
614 * Note, this has been updated so that it can be called first
615 * without a DEVICE, then a second or third time with a DEVICE,
616 * and each time, it should cleanup and point to the new device.
617 * This should facilitate switching devices.
618 * Note, each dcr must point to the controlling job (jcr). However,
619 * a job can have multiple dcrs, so we must not store in the jcr's
620 * structure as previously. The higher level routine must store
621 * this dcr in the right place
624 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
627 dcr = (DCR *)malloc(sizeof(DCR));
628 memset(dcr, 0, sizeof(DCR));
629 dcr->tid = pthread_self();
632 dcr->jcr = jcr; /* point back to jcr */
633 /* Set device information, possibly change device */
636 free_block(dcr->block);
638 dcr->block = new_block(dev);
640 free_record(dcr->rec);
642 dcr->rec = new_record();
643 if (dcr->attached_to_dev) {
644 detach_dcr_from_dev(dcr);
646 /* Use job spoolsize prior to device spoolsize */
647 if (jcr->spool_size) {
648 dcr->max_job_spool_size = jcr->spool_size;
650 dcr->max_job_spool_size = dev->device->max_job_spool_size;
652 dcr->device = dev->device;
654 attach_dcr_to_dev(dcr);
660 * Search the dcrs list for the given dcr. If it is found,
661 * as it should be, then remove it. Also zap the jcr pointer
662 * to the dcr if it is the same one.
664 * Note, this code will be turned on when we can write to multiple
665 * dcrs at the same time.
668 static void remove_dcr_from_dcrs(DCR *dcr)
674 int num = jcr->dcrs->size();
675 for (i=0; i < num; i++) {
676 ldcr = (DCR *)jcr->dcrs->get(i);
678 jcr->dcrs->remove(i);
679 if (jcr->dcr == dcr) {
688 static void attach_dcr_to_dev(DCR *dcr)
690 DEVICE *dev = dcr->dev;
693 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
694 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
695 dev->attached_dcrs->append(dcr); /* attach dcr to device */
696 dcr->attached_to_dev = true;
697 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
701 void detach_dcr_from_dev(DCR *dcr)
703 DEVICE *dev = dcr->dev;
704 Dmsg1(500, "JobId=%u enter detach_dcr_from_dev\n", (uint32_t)dcr->jcr->JobId);
706 /* Detach this dcr only if attached */
707 if (dcr->attached_to_dev && dev) {
709 unreserve_device(dcr);
710 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
711 dcr->attached_to_dev = false;
712 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
718 * Free up all aspects of the given dcr -- i.e. dechain it,
719 * release allocated memory, zap pointers, ...
721 void free_dcr(DCR *dcr)
725 detach_dcr_from_dev(dcr);
728 free_block(dcr->block);
731 free_record(dcr->rec);
733 if (jcr && jcr->dcr == dcr) {