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;
295 dev->dunblock(DEV_LOCKED);
296 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
302 * Acquire device for writing. We permit multiple writers.
303 * If this is the first one, we read the label.
305 * Returns: NULL if failed for any reason
307 * Note, normally reserve_device_for_append() is called
308 * before this routine.
310 DCR *acquire_device_for_append(DCR *dcr)
312 bool release = false;
313 bool recycle = false;
314 bool do_mount = false;
315 DEVICE *dev = dcr->dev;
318 init_device_wait_timers(dcr);
320 dev->dblock(BST_DOING_ACQUIRE);
321 Dmsg1(190, "acquire_append device is %s\n", dev->is_tape()?"tape":
322 (dev->is_dvd()?"DVD":"disk"));
325 * With the reservation system, this should not happen
327 if (dev->can_read()) {
328 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
329 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
333 if (dev->can_append()) {
334 Dmsg0(190, "device already in append.\n");
336 * Device already in append mode
338 * Check if we have the right Volume mounted
339 * OK if current volume info OK
340 * OK if next volume matches current volume
341 * otherwise mount desired volume obtained from
342 * dir_find_next_appendable_volume
343 * dev->VolHdr.VolumeName is what is in the drive
344 * dcr->VolumeName is what we pass into the routines, or
345 * get back from the subroutines.
347 bstrncpy(dcr->VolumeName, dev->VolHdr.VolumeName, sizeof(dcr->VolumeName));
348 if (!dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE) &&
349 !(dir_find_next_appendable_volume(dcr) &&
350 strcmp(dev->VolHdr.VolumeName, dcr->VolumeName) == 0)) { /* wrong tape mounted */
351 Dmsg2(190, "Wrong tape mounted: %s. wants:%s\n", dev->VolHdr.VolumeName,
353 /* Release volume reserved by dir_find_next_appendable_volume() */
354 if (dcr->VolumeName[0]) {
357 if (dev->num_writers != 0) {
358 Jmsg3(jcr, M_FATAL, 0, _("Wanted to append to Volume \"%s\", but device %s is busy writing on \"%s\" .\n"),
359 dcr->VolumeName, dev->print_name(), dev->VolHdr.VolumeName);
360 Dmsg3(200, "Wanted to append to Volume \"%s\", but device %s is busy writing on \"%s\" .\n",
361 dcr->VolumeName, dev->print_name(), dev->VolHdr.VolumeName);
364 /* Wrong tape mounted, release it, then fall through to get correct one */
365 Dmsg0(190, "Wrong tape mounted, release and try mount.\n");
370 * At this point, the correct tape is already mounted, so
371 * we do not need to do mount_next_write_volume(), unless
372 * we need to recycle the tape.
374 recycle = strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") == 0;
375 Dmsg1(190, "Correct tape mounted. recycle=%d\n", recycle);
376 if (recycle && 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());
382 if (dev->num_writers == 0) {
383 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
389 * Check to see if the tape position as defined by the OS is
390 * the same as our concept. If it is not, we bail out, because
391 * it means the user has probably manually rewound the tape.
392 * Note, we check only if num_writers == 0, but this code will
393 * also work fine for any number of writers. If num_writers > 0,
394 * we probably should cancel all jobs using this device, or
395 * perhaps even abort the SD, or at a minimum, mark the tape
396 * in error. Another strategy with num_writers == 0, would be
397 * to rewind the tape and do a new eod() request.
399 if (dev->is_tape() && dev->num_writers == 0) {
400 int32_t file = dev->get_os_tape_file();
401 if (file >= 0 && file != (int32_t)dev->get_file()) {
402 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
403 " on device %s. Expected %d, got %d\n"),
404 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
410 /* Not already in append mode, so mount the device */
411 Dmsg0(190, "Not in append mode, try mount.\n");
412 ASSERT(dev->num_writers == 0);
416 if (do_mount || recycle) {
417 Dmsg0(190, "Do mount_next_write_vol\n");
418 bool mounted = mount_next_write_volume(dcr, release);
420 if (!job_canceled(jcr)) {
421 /* Reduce "noise" -- don't print if job canceled */
422 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
424 Dmsg1(200, "Could not ready device %s for append.\n",
429 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
432 dev->num_writers++; /* we are now a writer */
433 if (jcr->NumWriteVolumes == 0) {
434 jcr->NumWriteVolumes = 1;
436 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
437 dir_update_volume_info(dcr, false); /* send Volume info to Director */
439 if (dcr->reserved_device) {
440 dev->reserved_device--;
441 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
442 dcr->reserved_device = false;
444 dev->dunblock(DEV_LOCKED);
452 if (dcr->reserved_device) {
453 dev->reserved_device--;
454 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
455 dcr->reserved_device = false;
457 dev->dunblock(DEV_LOCKED);
463 * This job is done, so release the device. From a Unix standpoint,
464 * the device remains open.
466 * Note, if we are spooling, we may enter with the device locked.
467 * However, in all cases, unlock the device when leaving.
470 bool release_device(DCR *dcr)
473 DEVICE *dev = dcr->dev;
476 /* lock only if not already locked by this thread */
477 if (!dcr->is_dev_locked()) {
480 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
482 /* if device is reserved, job never started, so release the reserve here */
483 if (dcr->reserved_device) {
484 dev->reserved_device--;
485 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
486 dcr->reserved_device = false;
489 if (dev->can_read()) {
490 dev->clear_read(); /* clear read bit */
491 Dmsg0(100, "dir_update_vol_info. Release0\n");
492 dir_update_volume_info(dcr, false); /* send Volume info to Director */
494 } else if (dev->num_writers > 0) {
496 * Note if WEOT is set, we are at the end of the tape
497 * and may not be positioned correctly, so the
498 * job_media_record and update_vol_info have already been
499 * done, which means we skip them here.
502 Dmsg1(100, "There are %d writers in release_device\n", dev->num_writers);
503 if (dev->is_labeled()) {
504 Dmsg0(100, "dir_create_jobmedia_record. Release\n");
505 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
506 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
507 dcr->VolCatInfo.VolCatName, jcr->Job);
509 /* If no more writers, write an EOF */
510 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
512 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
514 if (!dev->at_weot()) {
515 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
516 /* Note! do volume update before close, which zaps VolCatInfo */
517 Dmsg0(100, "dir_update_vol_info. Release0\n");
518 dir_update_volume_info(dcr, false); /* send Volume info to Director */
524 * If we reach here, it is most likely because the job
525 * has failed, since the device is not in read mode and
526 * there are no writers. It was probably reserved.
530 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
531 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
532 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
536 /* Fire off Alert command and include any output */
537 if (!job_canceled(jcr) && dcr->device->alert_command) {
541 char line[MAXSTRING];
542 alert = get_pool_memory(PM_FNAME);
543 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
544 bpipe = open_bpipe(alert, 0, "r");
546 while (fgets(line, sizeof(line), bpipe->rfd)) {
547 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
549 status = close_bpipe(bpipe);
555 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
556 alert, be.bstrerror(status));
559 Dmsg1(400, "alert status=%d\n", status);
560 free_pool_memory(alert);
562 pthread_cond_broadcast(&dev->wait_next_vol);
563 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
564 pthread_cond_broadcast(&wait_device_release);
567 detach_dcr_from_dev(dcr);
569 if (jcr->read_dcr == dcr) {
570 jcr->read_dcr = NULL;
572 if (jcr->dcr == dcr) {
577 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
578 (uint32_t)jcr->JobId);
583 * Create a new Device Control Record and attach
584 * it to the device (if this is a real job).
585 * Note, this has been updated so that it can be called first
586 * without a DEVICE, then a second or third time with a DEVICE,
587 * and each time, it should cleanup and point to the new device.
588 * This should facilitate switching devices.
589 * Note, each dcr must point to the controlling job (jcr). However,
590 * a job can have multiple dcrs, so we must not store in the jcr's
591 * structure as previously. The higher level routine must store
592 * this dcr in the right place
595 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
598 dcr = (DCR *)malloc(sizeof(DCR));
599 memset(dcr, 0, sizeof(DCR));
600 dcr->tid = pthread_self();
603 dcr->jcr = jcr; /* point back to jcr */
604 /* Set device information, possibly change device */
607 free_block(dcr->block);
609 dcr->block = new_block(dev);
611 free_record(dcr->rec);
613 dcr->rec = new_record();
614 if (dcr->attached_to_dev) {
615 detach_dcr_from_dev(dcr);
617 dcr->max_job_spool_size = dev->device->max_job_spool_size;
618 dcr->device = dev->device;
620 attach_dcr_to_dev(dcr);
626 * Search the dcrs list for the given dcr. If it is found,
627 * as it should be, then remove it. Also zap the jcr pointer
628 * to the dcr if it is the same one.
630 * Note, this code will be turned on when we can write to multiple
631 * dcrs at the same time.
634 static void remove_dcr_from_dcrs(DCR *dcr)
640 int num = jcr->dcrs->size();
641 for (i=0; i < num; i++) {
642 ldcr = (DCR *)jcr->dcrs->get(i);
644 jcr->dcrs->remove(i);
645 if (jcr->dcr == dcr) {
654 static void attach_dcr_to_dev(DCR *dcr)
656 DEVICE *dev = dcr->dev;
659 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
660 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
661 dev->attached_dcrs->append(dcr); /* attach dcr to device */
662 dcr->attached_to_dev = true;
663 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
667 void detach_dcr_from_dev(DCR *dcr)
669 DEVICE *dev = dcr->dev;
670 Dmsg1(500, "JobId=%u enter detach_dcr_from_dev\n", (uint32_t)dcr->jcr->JobId);
672 /* Detach this dcr only if attached */
673 if (dcr->attached_to_dev && dev) {
675 unreserve_device(dcr);
676 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
677 dcr->attached_to_dev = false;
678 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
684 * Free up all aspects of the given dcr -- i.e. dechain it,
685 * release allocated memory, zap pointers, ...
687 void free_dcr(DCR *dcr)
691 detach_dcr_from_dev(dcr);
694 free_block(dcr->block);
697 free_record(dcr->rec);
699 if (jcr && jcr->dcr == dcr) {