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));
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 */
92 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
93 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
94 dcr->VolCatInfo.Slot = vol->Slot;
97 * If the MediaType requested for this volume is not the
98 * same as the current drive, we attempt to find the same
99 * device that was used to write the orginal volume. If
100 * found, we switch to using that device.
102 * N.B. A lot of routines rely on the dcr pointer not changing
103 * read_records.c even has multiple dcrs cached, so we take care
104 * here to release all important parts of the dcr and re-acquire
105 * them such as the block pointer (size may change), but we do
106 * not release the dcr.
108 Dmsg2(50, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
109 if (dcr->media_type[0] && strcmp(dcr->media_type, dev->device->media_type) != 0) {
114 Jmsg3(jcr, M_INFO, 0, _("Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
116 dcr->media_type, dev->device->media_type, dev->print_name());
117 Dmsg3(50, "Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
119 dcr->media_type, dev->device->media_type, dev->print_name());
121 dev->dunblock(DEV_UNLOCKED);
124 memset(&rctx, 0, sizeof(RCTX));
126 jcr->reserve_msgs = New(alist(10, not_owned_by_alist));
127 rctx.any_drive = true;
128 rctx.device_name = vol->device;
129 store = new DIRSTORE;
130 memset(store, 0, sizeof(DIRSTORE));
131 store->name[0] = 0; /* No dir name */
132 bstrncpy(store->media_type, vol->MediaType, sizeof(store->media_type));
133 bstrncpy(store->pool_name, dcr->pool_name, sizeof(store->pool_name));
134 bstrncpy(store->pool_type, dcr->pool_type, sizeof(store->pool_type));
135 store->append = false;
137 clean_device(dcr); /* clean up the dcr */
140 * Search for a new device
142 stat = search_res_for_device(rctx);
143 release_reserve_messages(jcr); /* release queued messages */
144 unlock_reservations();
147 dev = dcr->dev; /* get new device pointer */
148 dev->dblock(BST_DOING_ACQUIRE);
149 dcr->VolumeName[0] = 0;
150 Jmsg(jcr, M_INFO, 0, _("Media Type change. New device %s chosen.\n"),
152 Dmsg1(50, "Media Type change. New device %s chosen.\n", dev->print_name());
154 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
155 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
156 dcr->VolCatInfo.Slot = vol->Slot;
157 bstrncpy(dcr->pool_name, store->pool_name, sizeof(dcr->pool_name));
158 bstrncpy(dcr->pool_type, store->pool_type, sizeof(dcr->pool_type));
161 Jmsg1(jcr, M_FATAL, 0, _("No suitable device found to read Volume \"%s\"\n"),
163 Dmsg1(50, "No suitable device found to read Volume \"%s\"\n", vol->VolumeName);
169 init_device_wait_timers(dcr);
171 tape_previously_mounted = dev->can_read() || dev->can_append() ||
173 tape_initially_mounted = tape_previously_mounted;
176 /* Volume info is always needed because of VolParts */
177 Dmsg0(200, "dir_get_volume_info\n");
178 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_READ)) {
179 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
183 /* If not polling limit retries */
184 if (!dev->poll && retry++ > 10) {
187 dev->clear_labeled(); /* force reread of label */
188 if (job_canceled(jcr)) {
190 Mmsg1(dev->errmsg, _("Job %s canceled.\n"), edit_int64(jcr->JobId, ed1));
191 Jmsg(jcr, M_INFO, 0, dev->errmsg);
192 goto get_out; /* error return */
195 autoload_device(dcr, 0, NULL);
198 * This code ensures that the device is ready for
199 * reading. If it is a file, it opens it.
200 * If it is a tape, it checks the volume name
202 Dmsg1(100, "bstored: open vol=%s\n", dcr->VolumeName);
203 if (dev->open(dcr, OPEN_READ_ONLY) < 0) {
204 Jmsg3(jcr, M_WARNING, 0, _("Read open device %s Volume \"%s\" failed: ERR=%s\n"),
205 dev->print_name(), dcr->VolumeName, dev->bstrerror());
208 Dmsg1(50, "opened dev %s OK\n", dev->print_name());
210 /* Read Volume Label */
211 Dmsg0(50, "calling read-vol-label\n");
212 vol_label_status = read_dev_volume_label(dcr);
213 switch (vol_label_status) {
216 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
220 * Send error message generated by read_dev_volume_label()
221 * only we really had a tape mounted. This supresses superfluous
222 * error messages when nothing is mounted.
224 if (tape_previously_mounted) {
225 Jmsg(jcr, M_WARNING, 0, "%s", jcr->errmsg);
229 if (tape_initially_mounted) {
230 tape_initially_mounted = false;
233 /* If polling and got a previous bad name, ignore it */
234 if (dev->poll && strcmp(dev->BadVolName, dev->VolHdr.VolumeName) == 0) {
237 bstrncpy(dev->BadVolName, dev->VolHdr.VolumeName, sizeof(dev->BadVolName));
241 Jmsg1(jcr, M_WARNING, 0, "%s", jcr->errmsg);
243 tape_previously_mounted = true;
246 * If the device requires mount, close it, so the device can be ejected.
248 if (dev->requires_mount()) {
252 /* Call autochanger only once unless ask_sysop called */
253 if (try_autochanger) {
255 Dmsg2(200, "calling autoload Vol=%s Slot=%d\n",
256 dcr->VolumeName, dcr->VolCatInfo.Slot);
257 stat = autoload_device(dcr, 0, NULL);
259 try_autochanger = false;
260 continue; /* try reading volume mounted */
264 /* Mount a specific volume and no other */
265 Dmsg0(200, "calling dir_ask_sysop\n");
266 if (!dir_ask_sysop_to_mount_volume(dcr, ST_READ)) {
267 goto get_out; /* error return */
269 try_autochanger = true; /* permit using autochanger again */
270 continue; /* try reading again */
275 Jmsg1(jcr, M_FATAL, 0, _("Too many errors trying to mount device %s for reading.\n"),
282 set_jcr_job_status(jcr, JS_Running);
283 dir_send_job_status(jcr);
284 Jmsg(jcr, M_INFO, 0, _("Ready to read from volume \"%s\" on device %s.\n"),
285 dcr->VolumeName, dev->print_name());
289 if (dcr && dcr->reserved_device) {
290 dev->reserved_device--;
291 Dmsg2(50, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
292 dcr->reserved_device = false;
295 * Normally we are blocked, but in at least one error case above
296 * we are not blocked because we unsuccessfully tried changing
299 if (dev->is_blocked()) {
300 dev->dunblock(DEV_LOCKED);
302 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
308 * Acquire device for writing. We permit multiple writers.
309 * If this is the first one, we read the label.
311 * Returns: NULL if failed for any reason
313 * Note, normally reserve_device_for_append() is called
314 * before this routine.
316 DCR *acquire_device_for_append(DCR *dcr)
318 bool do_mount = false;
319 bool release = false;
321 DEVICE *dev = dcr->dev;
324 init_device_wait_timers(dcr);
326 dev->dblock(BST_DOING_ACQUIRE);
327 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
328 (dev->is_dvd()?"DVD":"disk"));
331 * With the reservation system, this should not happen
333 if (dev->can_read()) {
334 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
335 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
340 * have_vol defines whether or not mount_next_write_volume should
341 * ask the Director again about what Volume to use.
343 have_vol = is_suitable_volume_mounted(dcr);
344 if (dev->can_append()) {
345 Dmsg0(190, "device already in append.\n");
347 * Device already in append mode
349 * Check if we have the right Volume mounted
350 * OK if current volume info OK
351 * OK if next volume matches current volume
352 * otherwise mount desired volume obtained from
353 * dir_find_next_appendable_volume
354 * dev->VolHdr.VolumeName is what is in the drive
355 * dcr->VolumeName is what we pass into the routines, or
356 * get back from the subroutines.
359 !(dir_find_next_appendable_volume(dcr) &&
360 strcmp(dev->VolHdr.VolumeName, dcr->VolumeName) == 0)) { /* wrong tape mounted */
361 /* Wrong tape mounted, release it, then fall through to get correct one */
362 Dmsg3(50, "Wrong tape mounted. Wanted:%s, got:%s, dev=%s release and try mount.\n",
363 dcr->VolumeName, dev->VolHdr.VolumeName, dev->print_name());
364 /* Do not release if no Volume in drive */
365 if (dev->VolHdr.VolumeName[0]) {
371 * At this point, the correct tape is already mounted, so
372 * we do not need to do mount_next_write_volume(), unless
373 * we need to recycle the tape.
375 do_mount = strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") == 0;
376 Dmsg2(190, "jid=%u Correct tape mounted. recycle=%d\n",
377 (uint32_t)jcr->JobId, do_mount);
378 if (dev->num_writers == 0) {
379 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
385 * Check to see if the tape position as defined by the OS is
386 * the same as our concept. If it is not, we bail out, because
387 * it means the user has probably manually rewound the tape.
388 * Note, we check only if num_writers == 0, but this code will
389 * also work fine for any number of writers. If num_writers > 0,
390 * we probably should cancel all jobs using this device, or
391 * perhaps even abort the SD, or at a minimum, mark the tape
392 * in error. Another strategy with num_writers == 0, would be
393 * to rewind the tape and do a new eod() request.
395 if (dev->is_tape() && dev->num_writers == 0) {
396 int32_t file = dev->get_os_tape_file();
397 if (file >= 0 && file != (int32_t)dev->get_file()) {
398 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
399 " on device %s. Expected %d, got %d\n"),
400 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
406 /* Not already in append mode, so mount the device */
407 Dmsg2(190, "jid=%u Not in append mode, try mount have_vol=%d\n",
408 (uint32_t)jcr->JobId, have_vol);
410 ASSERT(dev->num_writers == 0);
414 if (do_mount || !have_vol) {
415 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
416 bool mounted = mount_next_write_volume(dcr, have_vol, release);
418 if (!job_canceled(jcr)) {
419 /* Reduce "noise" -- don't print if job canceled */
420 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
422 Dmsg2(200, "jid=%u Could not ready device %s for append.\n",
423 (uint32_t)jcr->JobId, dev->print_name());
427 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
430 dev->num_writers++; /* we are now a writer */
431 if (jcr->NumWriteVolumes == 0) {
432 jcr->NumWriteVolumes = 1;
434 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
435 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
437 if (dcr->reserved_device) {
438 dev->reserved_device--;
439 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
440 dev->reserved_device, dev->print_name());
441 dcr->reserved_device = false;
443 dev->dunblock(DEV_LOCKED);
451 if (dcr->reserved_device) {
452 dev->reserved_device--;
453 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
454 dev->reserved_device, dev->print_name());
455 dcr->reserved_device = false;
457 dev->dunblock(DEV_LOCKED);
462 static bool is_suitable_volume_mounted(DCR *dcr)
464 DEVICE *dev = dcr->dev;
466 /* Volume mounted? */
467 if (dev->VolHdr.VolumeName[0] == 0) {
468 return false; /* no */
470 bstrncpy(dcr->VolumeName, dev->VolHdr.VolumeName, sizeof(dcr->VolumeName));
471 return dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE);
475 * This job is done, so release the device. From a Unix standpoint,
476 * the device remains open.
478 * Note, if we are spooling, we may enter with the device locked.
479 * However, in all cases, unlock the device when leaving.
482 bool release_device(DCR *dcr)
485 DEVICE *dev = dcr->dev;
488 /* lock only if not already locked by this thread */
489 if (!dcr->is_dev_locked()) {
492 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
494 /* if device is reserved, job never started, so release the reserve here */
495 if (dcr->reserved_device) {
496 dev->reserved_device--;
497 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
498 dcr->reserved_device = false;
501 if (dev->can_read()) {
502 dev->clear_read(); /* clear read bit */
503 Dmsg0(100, "dir_update_vol_info. Release0\n");
504 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
507 } else if (dev->num_writers > 0) {
509 * Note if WEOT is set, we are at the end of the tape
510 * and may not be positioned correctly, so the
511 * job_media_record and update_vol_info have already been
512 * done, which means we skip them here.
515 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
516 if (dev->is_labeled()) {
517 Dmsg0(100, "dir_create_jobmedia_record. Release\n");
518 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
519 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
520 dcr->VolCatInfo.VolCatName, jcr->Job);
522 /* If no more writers, and no errors, and wrote something, write an EOF */
523 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
525 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
527 if (!dev->num_writers) { /* if no more writers */
528 volume_unused(dcr); /* we obviously are not using the volume */
530 if (!dev->at_weot()) {
531 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
532 /* Note! do volume update before close, which zaps VolCatInfo */
533 Dmsg0(100, "dir_update_vol_info. Release0\n");
534 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
540 * If we reach here, it is most likely because the job
541 * has failed, since the device is not in read mode and
542 * 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 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
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) {