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);
42 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev);
45 /*********************************************************************
46 * Acquire device for reading.
47 * The drive should have previously been reserved by calling
48 * reserve_device_for_read(). We read the Volume label from the block and
49 * leave the block pointers just after the label.
51 * Returns: NULL if failed for any reason
54 bool acquire_device_for_read(DCR *dcr)
56 DEVICE *dev = dcr->dev;
59 bool tape_previously_mounted;
60 bool tape_initially_mounted;
62 bool try_autochanger = true;
67 Dmsg1(950, "jcr->dcr=%p\n", jcr->dcr);
68 dev->dblock(BST_DOING_ACQUIRE);
70 if (dev->num_writers > 0) {
71 Jmsg2(jcr, M_FATAL, 0, _("Acquire read: num_writers=%d not zero. Job %d canceled.\n"),
72 dev->num_writers, jcr->JobId);
76 /* Find next Volume, if any */
80 Jmsg(jcr, M_FATAL, 0, _("No volumes specified for reading. Job %s canceled.\n"),
81 edit_int64(jcr->JobId, ed1));
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, ST_READ)) {
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 DEVICE *dev = dcr->dev;
322 init_device_wait_timers(dcr);
324 dev->dblock(BST_DOING_ACQUIRE);
325 Dmsg1(100, "acquire_append device is %s\n", dev->is_tape()?"tape":
326 (dev->is_dvd()?"DVD":"disk"));
329 * With the reservation system, this should not happen
331 if (dev->can_read()) {
332 Jmsg1(jcr, M_FATAL, 0, _("Want to append, but device %s is busy reading.\n"), dev->print_name());
333 Dmsg1(200, "Want to append but device %s is busy reading.\n", dev->print_name());
338 * have_vol defines whether or not mount_next_write_volume should
339 * ask the Director again about what Volume to use.
341 if (dev->can_append() && is_suitable_volume_mounted(dcr) &&
342 strcmp(dcr->VolCatInfo.VolCatStatus, "Recycle") != 0) {
343 Dmsg0(190, "device already in append.\n");
345 * At this point, the correct tape is already mounted, so
346 * we do not need to do mount_next_write_volume(), unless
347 * we need to recycle the tape.
349 if (dev->num_writers == 0) {
350 memcpy(&dev->VolCatInfo, &dcr->VolCatInfo, sizeof(dev->VolCatInfo));
352 if (!is_tape_position_ok(jcr, dev)) {
356 Dmsg1(190, "jid=%u Do mount_next_write_vol\n", (uint32_t)jcr->JobId);
357 if (!dcr->mount_next_write_volume()) {
358 if (!job_canceled(jcr)) {
359 /* Reduce "noise" -- don't print if job canceled */
360 Jmsg(jcr, M_FATAL, 0, _("Could not ready device %s for append.\n"),
362 Dmsg2(200, "jid=%u Could not ready device %s for append.\n",
363 (uint32_t)jcr->JobId, dev->print_name());
367 Dmsg2(190, "Output pos=%u:%u\n", dcr->dev->file, dcr->dev->block_num);
370 dev->num_writers++; /* we are now a writer */
371 if (jcr->NumWriteVolumes == 0) {
372 jcr->NumWriteVolumes = 1;
374 dev->VolCatInfo.VolCatJobs++; /* increment number of jobs on vol */
375 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
377 if (dcr->reserved_device) {
378 dev->reserved_device--;
379 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
380 dev->reserved_device, dev->print_name());
381 dcr->reserved_device = false;
383 dev->dunblock(DEV_LOCKED);
391 if (dcr->reserved_device) {
392 dev->reserved_device--;
393 Dmsg3(100, "jid=%u Dec reserve=%d dev=%s\n", (uint32_t)jcr->JobId,
394 dev->reserved_device, dev->print_name());
395 dcr->reserved_device = false;
397 dev->dunblock(DEV_LOCKED);
404 * Check to see if the tape position as defined by the OS is
405 * the same as our concept. If it is not, we bail out, because
406 * it means the user has probably manually rewound the tape.
407 * Note, we check only if num_writers == 0, but this code will
408 * also work fine for any number of writers. If num_writers > 0,
409 * we probably should cancel all jobs using this device, or
410 * perhaps even abort the SD, or at a minimum, mark the tape
411 * in error. Another strategy with num_writers == 0, would be
412 * to rewind the tape and do a new eod() request.
414 static bool is_tape_position_ok(JCR *jcr, DEVICE *dev)
416 if (dev->is_tape() && dev->num_writers == 0) {
417 int32_t file = dev->get_os_tape_file();
418 if (file >= 0 && file != (int32_t)dev->get_file()) {
419 Jmsg(jcr, M_FATAL, 0, _("Invalid tape position on volume \"%s\""
420 " on device %s. Expected %d, got %d\n"),
421 dev->VolHdr.VolumeName, dev->print_name(), dev->get_file(), file);
428 static bool is_suitable_volume_mounted(DCR *dcr)
430 DEVICE *dev = dcr->dev;
432 /* Volume mounted? */
433 if (dev->VolHdr.VolumeName[0] == 0 || dcr->swap_dev || dcr->unload_device) {
434 return false; /* no */
436 bstrncpy(dcr->VolumeName, dev->VolHdr.VolumeName, sizeof(dcr->VolumeName));
437 return dir_get_volume_info(dcr, GET_VOL_INFO_FOR_WRITE);
441 * This job is done, so release the device. From a Unix standpoint,
442 * the device remains open.
444 * Note, if we are spooling, we may enter with the device locked.
445 * However, in all cases, unlock the device when leaving.
448 bool release_device(DCR *dcr)
451 DEVICE *dev = dcr->dev;
454 /* lock only if not already locked by this thread */
455 if (!dcr->is_dev_locked()) {
459 Dmsg2(100, "release_device device %s is %s\n", dev->print_name(), dev->is_tape()?"tape":"disk");
461 /* if device is reserved, job never started, so release the reserve here */
462 if (dcr->reserved_device) {
463 dev->reserved_device--;
464 Dmsg2(100, "Dec reserve=%d dev=%s\n", dev->reserved_device, dev->print_name());
465 dcr->reserved_device = false;
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 Dmsg0(100, "dir_create_jobmedia_record. Release\n");
485 if (!dev->at_weot() && !dir_create_jobmedia_record(dcr)) {
486 Jmsg(jcr, M_FATAL, 0, _("Could not create JobMedia record for Volume=\"%s\" Job=%s\n"),
487 dcr->VolCatInfo.VolCatName, jcr->Job);
489 /* If no more writers, and no errors, and wrote something, write an EOF */
490 if (!dev->num_writers && dev->can_write() && dev->block_num > 0) {
492 write_ansi_ibm_labels(dcr, ANSI_EOF_LABEL, dev->VolHdr.VolumeName);
494 if (!dev->num_writers) { /* if no more writers */
495 volume_unused(dcr); /* we obviously are not using the volume */
497 if (!dev->at_weot()) {
498 dev->VolCatInfo.VolCatFiles = dev->file; /* set number of files */
499 /* Note! do volume update before close, which zaps VolCatInfo */
500 Dmsg0(100, "dir_update_vol_info. Release0\n");
501 dir_update_volume_info(dcr, false, false); /* send Volume info to Director */
507 * If we reach here, it is most likely because the job
508 * has failed, since the device is not in read mode and
509 * there are no writers. It was probably reserved.
515 /* If no writers, close if file or !CAP_ALWAYS_OPEN */
516 if (dev->num_writers == 0 && (!dev->is_tape() || !dev->has_cap(CAP_ALWAYSOPEN))) {
517 dvd_remove_empty_part(dcr); /* get rid of any empty spool part */
521 /* Fire off Alert command and include any output */
522 if (!job_canceled(jcr) && dcr->device->alert_command) {
526 char line[MAXSTRING];
527 alert = get_pool_memory(PM_FNAME);
528 alert = edit_device_codes(dcr, alert, dcr->device->alert_command, "");
529 bpipe = open_bpipe(alert, 0, "r");
531 while (fgets(line, sizeof(line), bpipe->rfd)) {
532 Jmsg(jcr, M_ALERT, 0, _("Alert: %s"), line);
534 status = close_bpipe(bpipe);
540 Jmsg(jcr, M_ALERT, 0, _("3997 Bad alert command: %s: ERR=%s.\n"),
541 alert, be.bstrerror(status));
544 Dmsg1(400, "alert status=%d\n", status);
545 free_pool_memory(alert);
547 pthread_cond_broadcast(&dev->wait_next_vol);
548 Dmsg1(100, "JobId=%u broadcast wait_device_release\n", (uint32_t)jcr->JobId);
549 pthread_cond_broadcast(&wait_device_release);
552 detach_dcr_from_dev(dcr);
554 if (jcr->read_dcr == dcr) {
555 jcr->read_dcr = NULL;
557 if (jcr->dcr == dcr) {
562 Dmsg2(100, "===== Device %s released by JobId=%u\n", dev->print_name(),
563 (uint32_t)jcr->JobId);
568 * Clean up the device for reuse without freeing the memory
570 bool clean_device(DCR *dcr)
573 dcr->keep_dcr = true; /* do not free the dcr */
574 ok = release_device(dcr);
575 dcr->keep_dcr = false;
580 * Create a new Device Control Record and attach
581 * it to the device (if this is a real job).
582 * Note, this has been updated so that it can be called first
583 * without a DEVICE, then a second or third time with a DEVICE,
584 * and each time, it should cleanup and point to the new device.
585 * This should facilitate switching devices.
586 * Note, each dcr must point to the controlling job (jcr). However,
587 * a job can have multiple dcrs, so we must not store in the jcr's
588 * structure as previously. The higher level routine must store
589 * this dcr in the right place
592 DCR *new_dcr(JCR *jcr, DCR *dcr, DEVICE *dev)
595 dcr = (DCR *)malloc(sizeof(DCR));
596 memset(dcr, 0, sizeof(DCR));
597 dcr->tid = pthread_self();
600 dcr->jcr = jcr; /* point back to jcr */
601 /* Set device information, possibly change device */
604 free_block(dcr->block);
606 dcr->block = new_block(dev);
608 free_record(dcr->rec);
610 dcr->rec = new_record();
611 if (dcr->attached_to_dev) {
612 detach_dcr_from_dev(dcr);
614 /* Use job spoolsize prior to device spoolsize */
615 if (jcr->spool_size) {
616 dcr->max_job_spool_size = jcr->spool_size;
618 dcr->max_job_spool_size = dev->device->max_job_spool_size;
620 dcr->device = dev->device;
622 attach_dcr_to_dev(dcr);
628 * Search the dcrs list for the given dcr. If it is found,
629 * as it should be, then remove it. Also zap the jcr pointer
630 * to the dcr if it is the same one.
632 * Note, this code will be turned on when we can write to multiple
633 * dcrs at the same time.
636 static void remove_dcr_from_dcrs(DCR *dcr)
642 int num = jcr->dcrs->size();
643 for (i=0; i < num; i++) {
644 ldcr = (DCR *)jcr->dcrs->get(i);
646 jcr->dcrs->remove(i);
647 if (jcr->dcr == dcr) {
656 static void attach_dcr_to_dev(DCR *dcr)
658 DEVICE *dev = dcr->dev;
661 if (jcr) Dmsg1(500, "JobId=%u enter attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
662 if (!dcr->attached_to_dev && dev->initiated && jcr && jcr->JobType != JT_SYSTEM) {
663 dev->attached_dcrs->append(dcr); /* attach dcr to device */
664 dcr->attached_to_dev = true;
665 Dmsg1(500, "JobId=%u attach_dcr_to_dev\n", (uint32_t)jcr->JobId);
669 void detach_dcr_from_dev(DCR *dcr)
671 DEVICE *dev = dcr->dev;
672 Dmsg0(500, "Enter detach_dcr_from_dev\n"); /* jcr is NULL in some cases */
674 /* Detach this dcr only if attached */
675 if (dcr->attached_to_dev && dev) {
677 unreserve_device(dcr);
678 dcr->dev->attached_dcrs->remove(dcr); /* detach dcr from device */
679 dcr->attached_to_dev = false;
680 // remove_dcr_from_dcrs(dcr); /* remove dcr from jcr list */
686 * Free up all aspects of the given dcr -- i.e. dechain it,
687 * release allocated memory, zap pointers, ...
689 void free_dcr(DCR *dcr)
693 detach_dcr_from_dev(dcr);
696 free_block(dcr->block);
699 free_record(dcr->rec);
701 if (jcr && jcr->dcr == dcr) {