2 Bacula® - The Network Backup Solution
4 Copyright (C) 2002-2007 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);
43 /*********************************************************************
44 * Acquire device for reading.
45 * The drive should have previously been reserved by calling
46 * reserve_device_for_read(). We read the Volume label from the block and
47 * leave the block pointers just after the label.
49 * Returns: NULL if failed for any reason
52 bool acquire_device_for_read(DCR *dcr)
54 DEVICE *dev = dcr->dev;
57 bool tape_previously_mounted;
58 bool tape_initially_mounted;
60 bool try_autochanger = true;
65 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
66 dev->dblock(BST_DOING_ACQUIRE);
68 if (dev->num_writers > 0) {
69 Jmsg2(jcr, M_FATAL, 0, _("Acquire read: num_writers=%d not zero. Job %d canceled.\n"),
70 dev->num_writers, jcr->JobId);
74 /* Find next Volume, if any */
78 Jmsg(jcr, M_FATAL, 0, _("No volumes specified for reading. Job %s canceled.\n"),
79 edit_int64(jcr->JobId, ed1));
83 for (i=1; i<jcr->CurReadVolume; i++) {
87 Jmsg(jcr, M_FATAL, 0, _("Logic error: no next volume to read. Numvol=%d Curvol=%d\n"),
88 jcr->NumReadVolumes, jcr->CurReadVolume);
89 goto get_out; /* should not happen */
91 bstrncpy(dcr->VolumeName, vol->VolumeName, sizeof(dcr->VolumeName));
92 bstrncpy(dcr->media_type, vol->MediaType, sizeof(dcr->media_type));
93 dcr->VolCatInfo.Slot = vol->Slot;
96 * If the MediaType requested for this volume is not the
97 * same as the current drive, we attempt to find the same
98 * device that was used to write the orginal volume. If
99 * found, we switch to using that device.
101 * N.B. A lot of routines rely on the dcr pointer not changing
102 * read_records.c even has multiple dcrs cached, so we take care
103 * here to release all important parts of the dcr and re-acquire
104 * them such as the block pointer (size may change), but we do
105 * not release the dcr.
107 Dmsg2(50, "MediaType dcr=%s dev=%s\n", dcr->media_type, dev->device->media_type);
108 if (dcr->media_type[0] && strcmp(dcr->media_type, dev->device->media_type) != 0) {
113 Jmsg3(jcr, M_INFO, 0, _("Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
115 dcr->media_type, dev->device->media_type, dev->print_name());
116 Dmsg3(50, "Changing device. Want Media Type=\"%s\" have=\"%s\"\n"
118 dcr->media_type, dev->device->media_type, dev->print_name());
120 dev->dunblock(DEV_UNLOCKED);
123 memset(&rctx, 0, sizeof(RCTX));
125 jcr->reserve_msgs = New(alist(10, not_owned_by_alist));
126 rctx.any_drive = true;
127 rctx.device_name = vol->device;
128 store = new DIRSTORE;
129 memset(store, 0, sizeof(DIRSTORE));
130 store->name[0] = 0; /* No dir name */
131 bstrncpy(store->media_type, vol->MediaType, sizeof(store->media_type));
132 bstrncpy(store->pool_name, dcr->pool_name, sizeof(store->pool_name));
133 bstrncpy(store->pool_type, dcr->pool_type, sizeof(store->pool_type));
134 store->append = false;
136 dcr->keep_dcr = true; /* do not free the dcr */
138 dcr->keep_dcr = false;
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 release = false;
320 bool recycle = false;
321 bool do_mount = 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());
340 if (dev->can_append()) {
341 Dmsg0(190, "device already in append.\n");
343 * Device already in append mode
345 * Check if we have the right Volume mounted
346 * OK if current volume info OK
347 * OK if next volume matches current volume
348 * otherwise mount desired volume obtained from
349 * dir_find_next_appendable_volume
350 * dev->VolHdr.VolumeName is what is in the drive
351 * dcr->VolumeName is what we pass into the routines, or
352 * get back from the subroutines.
354 bstrncpy(dcr->VolumeName, dev->VolHdr.VolumeName, sizeof(dcr->VolumeName));
355 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE) &&
356 !(dir_find_next_appendable_volume(dcr) &&
357 strcmp(dev->VolHdr.VolumeName, dcr->VolumeName) == 0)) { /* wrong tape mounted */
358 Dmsg2(190, "Wrong tape mounted: %s. wants:%s\n", dev->VolHdr.VolumeName,
360 /* Release volume reserved by dir_find_next_appendable_volume() */
361 if (dcr->VolumeName[0]) {
364 if (dev->num_writers != 0) {
365 Jmsg3(jcr, M_FATAL, 0, _("Wanted to append to Volume \"%s\", but device %s is busy writing on \"%s\" .\n"),
366 dcr->VolumeName, dev->print_name(), dev->VolHdr.VolumeName);
367 Dmsg3(200, "Wanted to append to Volume \"%s\", but device %s is busy writing on \"%s\" .\n",
368 dcr->VolumeName, dev->print_name(), dev->VolHdr.VolumeName);
371 /* Wrong tape mounted, release it, then fall through to get correct one */
372 Dmsg0(190, "Wrong tape mounted, release and try mount.\n");
377 * At this point, the correct tape is already mounted, so
378 * we do not need to do mount_next_write_volume(), unless
379 * we need to recycle the tape.
381 recycle = strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") == 0;
382 Dmsg1(190, "Correct tape mounted. recycle=%d\n", recycle);
383 if (recycle && dev->num_writers != 0) {
384 Jmsg(jcr, M_FATAL, 0, _("Cannot recycle volume \"%s\""
385 " on device %s because it is in use by another job.\n"),
386 dev->VolHdr.VolumeName, dev->print_name());
389 if (dev->num_writers == 0) {
390 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
396 * Check to see if the tape position as defined by the OS is
397 * the same as our concept. If it is not, we bail out, because
398 * it means the user has probably manually rewound the tape.
399 * Note, we check only if num_writers == 0, but this code will
400 * also work fine for any number of writers. If num_writers > 0,
401 * we probably should cancel all jobs using this device, or
402 * perhaps even abort the SD, or at a minimum, mark the tape
403 * in error. Another strategy with num_writers == 0, would be
404 * to rewind the tape and do a new eod() request.
406 if (dev->is_tape() && dev->num_writers == 0) {
407 int32_t file = dev->get_os_tape_file();
408 if (file >= 0 && file != (int32_t)dev->get_file()) {
409 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
410 " on device %s. Expected %d, got %d\n"),
411 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
417 /* Not already in append mode, so mount the device */
418 Dmsg0(190, "Not in append mode, try mount.\n");
419 ASSERT(dev->num_writers == 0);
423 if (do_mount || recycle) {
424 Dmsg0(190, "Do mount_next_write_vol\n");
425 bool mounted = mount_next_write_volume(dcr, release);
427 if (!job_canceled(jcr)) {
428 /* Reduce "noise" -- don't print if job canceled */
429 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
431 Dmsg1(200, "Could not ready device %s for append.\n",
436 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
439 dev->num_writers++; /* we are now a writer */
440 if (jcr->NumWriteVolumes == 0) {
441 jcr->NumWriteVolumes = 1;
443 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
444 dir_update_volume_info(dcr, false); /* send Volume info to Director */
446 if (dcr->reserved_device) {
447 dev->reserved_device--;
448 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
449 dev->reserved_device, dev->print_name());
450 dcr->reserved_device = false;
452 dev->dunblock(DEV_LOCKED);
460 if (dcr->reserved_device) {
461 dev->reserved_device--;
462 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
463 dev->reserved_device, dev->print_name());
464 dcr->reserved_device = false;
466 dev->dunblock(DEV_LOCKED);
472 * This job is done, so release the device. From a Unix standpoint,
473 * the device remains open.
475 * Note, if we are spooling, we may enter with the device locked.
476 * However, in all cases, unlock the device when leaving.
479 bool release_device(DCR *dcr)
482 DEVICE *dev = dcr->dev;
485 /* lock only if not already locked by this thread */
486 if (!dcr->is_dev_locked()) {
489 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
491 /* if device is reserved, job never started, so release the reserve here */
492 if (dcr->reserved_device) {
493 dev->reserved_device--;
494 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
495 dcr->reserved_device = false;
498 if (dev->can_read()) {
499 dev->clear_read(); /* clear read bit */
500 Dmsg0(100, "dir_update_vol_info. Release0\n");
501 dir_update_volume_info(dcr, false); /* send Volume info to Director */
503 } else if (dev->num_writers > 0) {
505 * Note if WEOT is set, we are at the end of the tape
506 * and may not be positioned correctly, so the
507 * job_media_record and update_vol_info have already been
508 * done, which means we skip them here.
511 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
512 if (dev->is_labeled()) {
513 Dmsg0(100, "dir_create_jobmedia_record. Release\n");
514 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
515 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
516 dcr->VolCatInfo.VolCatName, jcr->Job);
518 /* If no more writers, write an EOF */
519 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
521 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
523 if (!dev->at_weot()) {
524 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
525 /* Note! do volume update before close, which zaps VolCatInfo */
526 Dmsg0(100, "dir_update_vol_info. Release0\n");
527 dir_update_volume_info(dcr, false); /* send Volume info to Director */
533 * If we reach here, it is most likely because the job
534 * has failed, since the device is not in read mode and
535 * there are no writers. It was probably reserved.
539 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
540 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
541 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
545 /* Fire off Alert command and include any output */
546 if (!job_canceled(jcr) && dcr->device->alert_command) {
550 char line[MAXSTRING];
551 alert = get_pool_memory(PM_FNAME);
552 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
553 bpipe = open_bpipe(alert, 0, "r");
555 while (fgets(line, sizeof(line), bpipe->rfd)) {
556 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
558 status = close_bpipe(bpipe);
564 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
565 alert, be.bstrerror(status));
568 Dmsg1(400, "alert status=%d\n", status);
569 free_pool_memory(alert);
571 pthread_cond_broadcast(&dev->wait_next_vol);
572 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
573 pthread_cond_broadcast(&wait_device_release);
576 detach_dcr_from_dev(dcr);
578 if (jcr->read_dcr == dcr) {
579 jcr->read_dcr = NULL;
581 if (jcr->dcr == dcr) {
586 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
587 (uint32_t)jcr->JobId);
592 * Create a new Device Control Record and attach
593 * it to the device (if this is a real job).
594 * Note, this has been updated so that it can be called first
595 * without a DEVICE, then a second or third time with a DEVICE,
596 * and each time, it should cleanup and point to the new device.
597 * This should facilitate switching devices.
598 * Note, each dcr must point to the controlling job (jcr). However,
599 * a job can have multiple dcrs, so we must not store in the jcr's
600 * structure as previously. The higher level routine must store
601 * this dcr in the right place
604 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
607 dcr = (DCR *)malloc(sizeof(DCR));
608 memset(dcr, 0, sizeof(DCR));
609 dcr->tid = pthread_self();
612 dcr->jcr = jcr; /* point back to jcr */
613 /* Set device information, possibly change device */
616 free_block(dcr->block);
618 dcr->block = new_block(dev);
620 free_record(dcr->rec);
622 dcr->rec = new_record();
623 if (dcr->attached_to_dev) {
624 detach_dcr_from_dev(dcr);
626 dcr->max_job_spool_size = dev->device->max_job_spool_size;
627 dcr->device = dev->device;
629 attach_dcr_to_dev(dcr);
635 * Search the dcrs list for the given dcr. If it is found,
636 * as it should be, then remove it. Also zap the jcr pointer
637 * to the dcr if it is the same one.
639 * Note, this code will be turned on when we can write to multiple
640 * dcrs at the same time.
643 static void remove_dcr_from_dcrs(DCR *dcr)
649 int num = jcr->dcrs->size();
650 for (i=0; i < num; i++) {
651 ldcr = (DCR *)jcr->dcrs->get(i);
653 jcr->dcrs->remove(i);
654 if (jcr->dcr == dcr) {
663 static void attach_dcr_to_dev(DCR *dcr)
665 DEVICE *dev = dcr->dev;
668 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
669 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
670 dev->attached_dcrs->append(dcr); /* attach dcr to device */
671 dcr->attached_to_dev = true;
672 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
676 void detach_dcr_from_dev(DCR *dcr)
678 DEVICE *dev = dcr->dev;
679 Dmsg1(500, "JobId=%u enter detach_dcr_from_dev\n", (uint32_t)dcr->jcr->JobId);
681 /* Detach this dcr only if attached */
682 if (dcr->attached_to_dev && dev) {
684 unreserve_device(dcr);
685 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
686 dcr->attached_to_dev = false;
687 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
693 * Free up all aspects of the given dcr -- i.e. dechain it,
694 * release allocated memory, zap pointers, ...
696 void free_dcr(DCR *dcr)
700 detach_dcr_from_dev(dcr);
703 free_block(dcr->block);
706 free_record(dcr->rec);
708 if (jcr && jcr->dcr == dcr) {